poly_int: subreg_get_info
[official-gcc.git] / gcc / ChangeLog
blobe7901e52116e6a0f95cf9eca0f6cdde6cc02ac0f
1 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
2             Alan Hayward  <alan.hayward@arm.com>
3             David Sherwood  <david.sherwood@arm.com>
5         * rtlanal.c (subreg_get_info): Handle polynomial mode sizes.
7 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
8             Alan Hayward  <alan.hayward@arm.com>
9             David Sherwood  <david.sherwood@arm.com>
11         * combine.c (can_change_dest_mode): Handle polynomial
12         REGMODE_NATURAL_SIZE.
13         * expmed.c (store_bit_field_1): Likewise.
14         * expr.c (store_constructor): Likewise.
15         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
16         and polynomial REGMODE_NATURAL_SIZE.
17         (gen_lowpart_common): Likewise.
18         * reginfo.c (record_subregs_of_mode): Likewise.
19         * rtlanal.c (read_modify_subreg_p): Likewise.
21 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22             Alan Hayward  <alan.hayward@arm.com>
23             David Sherwood  <david.sherwood@arm.com>
25         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
26         numbers of elements.
28 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
29             Alan Hayward  <alan.hayward@arm.com>
30             David Sherwood  <david.sherwood@arm.com>
32         * match.pd: Cope with polynomial numbers of vector elements.
34 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
35             Alan Hayward  <alan.hayward@arm.com>
36             David Sherwood  <david.sherwood@arm.com>
38         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
39         in a POINTER_PLUS_EXPR.
41 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
42             Alan Hayward  <alan.hayward@arm.com>
43             David Sherwood  <david.sherwood@arm.com>
45         * omp-simd-clone.c (simd_clone_subparts): New function.
46         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
47         (ipa_simd_modify_function_body): Likewise.
49 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
50             Alan Hayward  <alan.hayward@arm.com>
51             David Sherwood  <david.sherwood@arm.com>
53         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
54         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
55         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
56         (expand_vector_condition, vector_element): Likewise.
57         (subparts_gt): New function.
58         (get_compute_type): Use subparts_gt.
59         (count_type_subparts): Delete.
60         (expand_vector_operations_1): Use subparts_gt instead of
61         count_type_subparts.
63 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
64             Alan Hayward  <alan.hayward@arm.com>
65             David Sherwood  <david.sherwood@arm.com>
67         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
68         (vect_compile_time_alias): ...this new function.  Do the calculation
69         on poly_ints rather than trees.
70         (vect_prune_runtime_alias_test_list): Update call accordingly.
72 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
73             Alan Hayward  <alan.hayward@arm.com>
74             David Sherwood  <david.sherwood@arm.com>
76         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
77         numbers of units.
78         (vect_schedule_slp_instance): Likewise.
80 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
81             Alan Hayward  <alan.hayward@arm.com>
82             David Sherwood  <david.sherwood@arm.com>
84         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
85         constant and extern definitions for variable-length vectors.
86         (vect_get_constant_vectors): Note that the number of units
87         is known to be constant.
89 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
90             Alan Hayward  <alan.hayward@arm.com>
91             David Sherwood  <david.sherwood@arm.com>
93         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
94         of units as polynomial.  Choose between WIDE and NARROW based
95         on multiple_p.
97 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
98             Alan Hayward  <alan.hayward@arm.com>
99             David Sherwood  <david.sherwood@arm.com>
101         * tree-vect-stmts.c (simd_clone_subparts): New function.
102         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
104 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
105             Alan Hayward  <alan.hayward@arm.com>
106             David Sherwood  <david.sherwood@arm.com>
108         * tree-vect-stmts.c (vectorizable_call): Treat the number of
109         vectors as polynomial.  Use build_index_vector for
110         IFN_GOMP_SIMD_LANE.
112 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
113             Alan Hayward  <alan.hayward@arm.com>
114             David Sherwood  <david.sherwood@arm.com>
116         * tree-vect-stmts.c (get_load_store_type): Treat the number of
117         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
118         for variable-length vectors.
119         (vectorizable_mask_load_store): Treat the number of units as
120         polynomial, asserting that it is constant if the condition has
121         already been enforced.
122         (vectorizable_store, vectorizable_load): Likewise.
124 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
125             Alan Hayward  <alan.hayward@arm.com>
126             David Sherwood  <david.sherwood@arm.com>
128         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
129         of units as polynomial.  Punt if we can't tell at compile time
130         which vector contains the final result.
132 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
133             Alan Hayward  <alan.hayward@arm.com>
134             David Sherwood  <david.sherwood@arm.com>
136         * tree-vect-loop.c (vectorizable_induction): Treat the number
137         of units as polynomial.  Punt on SLP inductions.  Use an integer
138         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
139         cast of such a series for variable-length floating-point
140         reductions.
142 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
143             Alan Hayward  <alan.hayward@arm.com>
144             David Sherwood  <david.sherwood@arm.com>
146         * tree.h (build_index_vector): Declare.
147         * tree.c (build_index_vector): New function.
148         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
149         of units as polynomial, forcibly converting it to a constant if
150         vectorizable_reduction has already enforced the condition.
151         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
152         to create a {1,2,3,...} vector.
153         (vectorizable_reduction): Treat the number of units as polynomial.
154         Choose vectype_in based on the largest scalar element size rather
155         than the smallest number of units.  Enforce the restrictions
156         relied on above.
158 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
159             Alan Hayward  <alan.hayward@arm.com>
160             David Sherwood  <david.sherwood@arm.com>
162         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
163         number of units as polynomial.
165 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
166             Alan Hayward  <alan.hayward@arm.com>
167             David Sherwood  <david.sherwood@arm.com>
169         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
170         * target.def (autovectorize_vector_sizes): Return the vector sizes
171         by pointer, using vector_sizes rather than a bitmask.
172         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
173         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
174         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
175         Likewise.
176         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
177         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
178         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
179         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
180         * omp-general.c (omp_max_vf): Likewise.
181         * omp-low.c (omp_clause_aligned_alignment): Likewise.
182         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
183         * tree-vect-loop.c (vect_analyze_loop): Likewise.
184         * tree-vect-slp.c (vect_slp_bb): Likewise.
185         * doc/tm.texi: Regenerate.
186         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
187         to a poly_uint64.
188         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
189         the vector size as a poly_uint64 rather than an unsigned int.
190         (current_vector_size): Change from an unsigned int to a poly_uint64.
191         (get_vectype_for_scalar_type): Update accordingly.
192         * tree.h (build_truth_vector_type): Take the size and number of
193         units as a poly_uint64 rather than an unsigned int.
194         (build_vector_type): Add a temporary overload that takes
195         the number of units as a poly_uint64 rather than an unsigned int.
196         * tree.c (make_vector_type): Likewise.
197         (build_truth_vector_type): Take the number of units as a poly_uint64
198         rather than an unsigned int.
200 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
201             Alan Hayward  <alan.hayward@arm.com>
202             David Sherwood  <david.sherwood@arm.com>
204         * target.def (get_mask_mode): Take the number of units and length
205         as poly_uint64s rather than unsigned ints.
206         * targhooks.h (default_get_mask_mode): Update accordingly.
207         * targhooks.c (default_get_mask_mode): Likewise.
208         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
209         * doc/tm.texi: Regenerate.
211 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
212             Alan Hayward  <alan.hayward@arm.com>
213             David Sherwood  <david.sherwood@arm.com>
215         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
216         * omp-general.c (omp_max_vf): Likewise.
217         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
218         (expand_omp_simd): Handle polynomial safelen.
219         * omp-low.c (omplow_simd_context): Add a default constructor.
220         (omplow_simd_context::max_vf): Change from int to poly_uint64.
221         (lower_rec_simd_input_clauses): Update accordingly.
222         (lower_rec_input_clauses): Likewise.
224 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
225             Alan Hayward  <alan.hayward@arm.com>
226             David Sherwood  <david.sherwood@arm.com>
228         * tree-vectorizer.h (vect_nunits_for_cost): New function.
229         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
230         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
231         (vect_analyze_slp_cost): Likewise.
232         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
233         (vect_model_load_cost): Likewise.
235 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
236             Alan Hayward  <alan.hayward@arm.com>
237             David Sherwood  <david.sherwood@arm.com>
239         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
240         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
241         from an unsigned int * to a poly_uint64_pod *.
242         (calculate_unrolling_factor): New function.
243         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
245 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
246             Alan Hayward  <alan.hayward@arm.com>
247             David Sherwood  <david.sherwood@arm.com>
249         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
250         from an unsigned int to a poly_uint64.
251         (_loop_vec_info::slp_unrolling_factor): Likewise.
252         (_loop_vec_info::vectorization_factor): Change from an int
253         to a poly_uint64.
254         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
255         (vect_get_num_vectors): New function.
256         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
257         (vect_get_num_copies): Use vect_get_num_vectors.
258         (vect_analyze_data_ref_dependences): Change max_vf from an int *
259         to an unsigned int *.
260         (vect_analyze_data_refs): Change min_vf from an int * to a
261         poly_uint64 *.
262         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
263         than an unsigned HOST_WIDE_INT.
264         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
265         (vect_analyze_data_ref_dependence): Change max_vf from an int *
266         to an unsigned int *.
267         (vect_analyze_data_ref_dependences): Likewise.
268         (vect_compute_data_ref_alignment): Handle polynomial vf.
269         (vect_enhance_data_refs_alignment): Likewise.
270         (vect_prune_runtime_alias_test_list): Likewise.
271         (vect_shift_permute_load_chain): Likewise.
272         (vect_supportable_dr_alignment): Likewise.
273         (dependence_distance_ge_vf): Take the vectorization factor as a
274         poly_uint64 rather than an unsigned HOST_WIDE_INT.
275         (vect_analyze_data_refs): Change min_vf from an int * to a
276         poly_uint64 *.
277         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
278         vfm1 as a poly_uint64 rather than an int.  Make the same change
279         for the returned bound_scalar.
280         (vect_gen_vector_loop_niters): Handle polynomial vf.
281         (vect_do_peeling): Likewise.  Update call to
282         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
283         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
284         be constant.
285         * tree-vect-loop.c (vect_determine_vectorization_factor)
286         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
287         (vect_get_known_peeling_cost): Likewise.
288         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
289         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
290         (vect_transform_loop): Likewise.  Use the lowest possible VF when
291         updating the upper bounds of the loop.
292         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
293         rather than an int.
294         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
295         polynomial unroll factors.
296         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
297         (vect_make_slp_decision): Likewise.
298         (vect_supported_load_permutation_p): Likewise, and polynomial
299         vf too.
300         (vect_analyze_slp_cost): Handle polynomial vf.
301         (vect_slp_analyze_node_operations): Likewise.
302         (vect_slp_analyze_bb_1): Likewise.
303         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
304         than an unsigned HOST_WIDE_INT.
305         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
306         (vectorizable_load): Handle polynomial vf.
307         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
308         a poly_uint64.
309         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
311 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
312             Alan Hayward  <alan.hayward@arm.com>
313             David Sherwood  <david.sherwood@arm.com>
315         * match.pd: Handle bit operations involving three constants
316         and try to fold one pair.
318 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
320         * tree-vect-loop-manip.c: Include gimple-fold.h.
321         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
322         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
323         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
324         Add a path that uses a step of VF instead of 1, but disable it
325         for now.
326         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
327         and niters_no_overflow parameters.  Update calls to
328         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
329         Create a new SSA name if the latter choses to use a ste other
330         than zero, and return it via niters_vector_mult_vf_var.
331         * tree-vect-loop.c (vect_transform_loop): Update calls to
332         vect_do_peeling, vect_gen_vector_loop_niters and
333         slpeel_make_loop_iterate_ntimes.
334         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
335         (vect_gen_vector_loop_niters): Update declarations after above changes.
337 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
339         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
340         128-bit round to integer instructions.
341         (ceil<mode>2): Likewise.
342         (btrunc<mode>2): Likewise.
343         (round<mode>2): Likewise.
345 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
347         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
348         unaligned VSX load/store on P8/P9.
349         (expand_block_clear): Allow the use of unaligned VSX
350         load/store on P8/P9.
352 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
354         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
355         New function.
356         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
357         swap associated with both a load and a store.
359 2018-01-02  Andrew Waterman  <andrew@sifive.com>
361         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
362         * config/riscv/riscv.md (clear_cache): Use it.
364 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
366         * web.c: Remove out-of-date comment.
368 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
370         * expr.c (fixup_args_size_notes): Check that any existing
371         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
372         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
373         (emit_single_push_insn): ...here.
375 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
377         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
378         (const_vector_encoded_nelts): New function.
379         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
380         (const_vector_int_elt, const_vector_elt): Declare.
381         * emit-rtl.c (const_vector_int_elt_1): New function.
382         (const_vector_elt): Likewise.
383         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
384         of CONST_VECTOR_ELT.
386 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
388         * expr.c: Include rtx-vector-builder.h.
389         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
390         directly on the tree encoding.
391         (const_vector_from_tree): Likewise.
392         * optabs.c: Include rtx-vector-builder.h.
393         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
394         sequence of "u" values.
395         * vec-perm-indices.c: Include rtx-vector-builder.h.
396         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
397         directly on the vec_perm_indices encoding.
399 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
401         * doc/rtl.texi (const_vector): Describe new encoding scheme.
402         * Makefile.in (OBJS): Add rtx-vector-builder.o.
403         * rtx-vector-builder.h: New file.
404         * rtx-vector-builder.c: Likewise.
405         * rtl.h (rtx_def::u2): Add a const_vector field.
406         (CONST_VECTOR_NPATTERNS): New macro.
407         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
408         (CONST_VECTOR_DUPLICATE_P): Likewise.
409         (CONST_VECTOR_STEPPED_P): Likewise.
410         (CONST_VECTOR_ENCODED_ELT): Likewise.
411         (const_vec_duplicate_p): Check for a duplicated vector encoding.
412         (unwrap_const_vec_duplicate): Likewise.
413         (const_vec_series_p): Check for a non-duplicated vector encoding.
414         Say that the function only returns true for integer vectors.
415         * emit-rtl.c: Include rtx-vector-builder.h.
416         (gen_const_vec_duplicate_1): Delete.
417         (gen_const_vector): Call gen_const_vec_duplicate instead of
418         gen_const_vec_duplicate_1.
419         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
420         (gen_const_vec_duplicate): Use rtx_vector_builder.
421         (gen_const_vec_series): Likewise.
422         (gen_rtx_CONST_VECTOR): Likewise.
423         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
424         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
425         Build a new vector rather than modifying a CONST_VECTOR in-place.
426         (handle_special_swappables): Update call accordingly.
427         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
428         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
429         Build a new vector rather than modifying a CONST_VECTOR in-place.
430         (handle_special_swappables): Update call accordingly.
432 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
434         * simplify-rtx.c (simplify_const_binary_operation): Use
435         CONST_VECTOR_ELT instead of XVECEXP.
437 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
439         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
440         the selector elements to be different from the data elements
441         if the selector is a VECTOR_CST.
442         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
443         ssizetype for the selector.
445 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
447         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
448         before testing each element individually.
449         * tree-vect-generic.c (lower_vec_perm): Likewise.
451 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
453         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
454         * selftest-run-tests.c (selftest::run_tests): Call it.
455         * vector-builder.h (vector_builder::operator ==): New function.
456         (vector_builder::operator !=): Likewise.
457         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
458         (vec_perm_indices::all_from_input_p): New function.
459         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
460         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
461         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
462         instead of reading the VECTOR_CST directly.  Detect whether both
463         vector inputs are the same before constructing the vec_perm_indices,
464         and update the number of inputs argument accordingly.  Use the
465         utility functions added above.  Only construct sel2 if we need to.
467 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
469         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
470         the broadcast of the low byte.
471         (expand_mult_highpart): Use an explicit encoding for the permutes.
472         * optabs-query.c (can_mult_highpart_p): Likewise.
473         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
474         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
475         (vectorizable_bswap): Likewise.
476         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
477         explicit encoding for the power-of-2 permutes.
478         (vect_permute_store_chain): Likewise.
479         (vect_grouped_load_supported): Likewise.
480         (vect_permute_load_chain): Likewise.
482 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
484         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
485         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
486         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
487         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
488         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
489         (vect_gen_perm_mask_any): Likewise.
491 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
493         * int-vector-builder.h: New file.
494         * vec-perm-indices.h: Include int-vector-builder.h.
495         (vec_perm_indices): Redefine as an int_vector_builder.
496         (auto_vec_perm_indices): Delete.
497         (vec_perm_builder): Redefine as a stand-alone class.
498         (vec_perm_indices::vec_perm_indices): New function.
499         (vec_perm_indices::clamp): Likewise.
500         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
501         (vec_perm_indices::new_vector): New function.
502         (vec_perm_indices::new_expanded_vector): Update for new
503         vec_perm_indices class.
504         (vec_perm_indices::rotate_inputs): New function.
505         (vec_perm_indices::all_in_range_p): Operate directly on the
506         encoded form, without computing elided elements.
507         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
508         encoding.  Update for new vec_perm_indices class.
509         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
510         the given vec_perm_builder.
511         (expand_vec_perm_var): Update vec_perm_builder constructor.
512         (expand_mult_highpart): Use vec_perm_builder instead of
513         auto_vec_perm_indices.
514         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
515         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
516         or double series encoding as appropriate.
517         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
518         vec_perm_indices instead of auto_vec_perm_indices.
519         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
520         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
521         (vect_permute_store_chain): Likewise.
522         (vect_grouped_load_supported): Likewise.
523         (vect_permute_load_chain): Likewise.
524         (vect_shift_permute_load_chain): Likewise.
525         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
526         (vect_transform_slp_perm_load): Likewise.
527         (vect_schedule_slp_instance): Likewise.
528         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
529         (vectorizable_mask_load_store): Likewise.
530         (vectorizable_bswap): Likewise.
531         (vectorizable_store): Likewise.
532         (vectorizable_load): Likewise.
533         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
534         vec_perm_indices instead of auto_vec_perm_indices.  Use
535         tree_to_vec_perm_builder to read the vector from a tree.
536         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
537         vec_perm_builder instead of a vec_perm_indices.
538         (have_whole_vector_shift): Use vec_perm_builder and
539         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
540         truncation to calc_vec_perm_mask_for_shift.
541         (vect_create_epilog_for_reduction): Likewise.
542         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
543         from auto_vec_perm_indices to vec_perm_indices.
544         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
545         instead of changing individual elements.
546         (aarch64_vectorize_vec_perm_const): Use new_vector to install
547         the vector in d.perm.
548         * config/arm/arm.c (expand_vec_perm_d::perm): Change
549         from auto_vec_perm_indices to vec_perm_indices.
550         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
551         instead of changing individual elements.
552         (arm_vectorize_vec_perm_const): Use new_vector to install
553         the vector in d.perm.
554         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
555         Update vec_perm_builder constructor.
556         (rs6000_expand_interleave): Likewise.
557         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
558         (rs6000_expand_interleave): Likewise.
560 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
562         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
563         to qimode could truncate the indices.
564         * optabs.c (expand_vec_perm_var): Likewise.
566 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
568         * Makefile.in (OBJS): Add vec-perm-indices.o.
569         * vec-perm-indices.h: New file.
570         * vec-perm-indices.c: Likewise.
571         * target.h (vec_perm_indices): Replace with a forward class
572         declaration.
573         (auto_vec_perm_indices): Move to vec-perm-indices.h.
574         * optabs.h: Include vec-perm-indices.h.
575         (expand_vec_perm): Delete.
576         (selector_fits_mode_p, expand_vec_perm_var): Declare.
577         (expand_vec_perm_const): Declare.
578         * target.def (vec_perm_const_ok): Replace with...
579         (vec_perm_const): ...this new hook.
580         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
581         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
582         * doc/tm.texi: Regenerate.
583         * optabs.def (vec_perm_const): Delete.
584         * doc/md.texi (vec_perm_const): Likewise.
585         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
586         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
587         expand_vec_perm for constant permutation vectors.  Assert that
588         the mode of variable permutation vectors is the integer equivalent
589         of the mode that is being permuted.
590         * optabs-query.h (selector_fits_mode_p): Declare.
591         * optabs-query.c: Include vec-perm-indices.h.
592         (selector_fits_mode_p): New function.
593         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
594         is defined, instead of checking whether the vec_perm_const_optab
595         exists.  Use targetm.vectorize.vec_perm_const instead of
596         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
597         fit in the vector mode before using a variable permute.
598         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
599         vec_perm_indices instead of an rtx.
600         (expand_vec_perm): Replace with...
601         (expand_vec_perm_const): ...this new function.  Take the selector
602         as a vec_perm_indices rather than an rtx.  Also take the mode of
603         the selector.  Update call to shift_amt_for_vec_perm_mask.
604         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
605         Use vec_perm_indices::new_expanded_vector to expand the original
606         selector into bytes.  Check whether the indices fit in the vector
607         mode before using a variable permute.
608         (expand_vec_perm_var): Make global.
609         (expand_mult_highpart): Use expand_vec_perm_const.
610         * fold-const.c: Includes vec-perm-indices.h.
611         * tree-ssa-forwprop.c: Likewise.
612         * tree-vect-data-refs.c: Likewise.
613         * tree-vect-generic.c: Likewise.
614         * tree-vect-loop.c: Likewise.
615         * tree-vect-slp.c: Likewise.
616         * tree-vect-stmts.c: Likewise.
617         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
618         Delete.
619         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
620         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
621         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
622         (aarch64_vectorize_vec_perm_const): ...this new function.
623         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
624         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
625         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
626         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
627         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
628         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
629         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
630         into...
631         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
632         check for NEON modes.
633         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
634         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
635         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
636         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
637         into...
638         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
639         the old VEC_PERM_CONST conditions.
640         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
641         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
642         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
643         (ia64_vectorize_vec_perm_const_ok): Merge into...
644         (ia64_vectorize_vec_perm_const): ...this new function.
645         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
646         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
647         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
648         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
649         * config/mips/mips.c (mips_expand_vec_perm_const)
650         (mips_vectorize_vec_perm_const_ok): Merge into...
651         (mips_vectorize_vec_perm_const): ...this new function.
652         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
653         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
654         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
655         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
656         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
657         (rs6000_expand_vec_perm_const): Delete.
658         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
659         Delete.
660         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
661         (altivec_expand_vec_perm_const_le): Take each operand individually.
662         Operate on constant selectors rather than rtxes.
663         (altivec_expand_vec_perm_const): Likewise.  Update call to
664         altivec_expand_vec_perm_const_le.
665         (rs6000_expand_vec_perm_const): Delete.
666         (rs6000_vectorize_vec_perm_const_ok): Delete.
667         (rs6000_vectorize_vec_perm_const): New function.
668         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
669         an element count and rtx array.
670         (rs6000_expand_extract_even): Update call accordingly.
671         (rs6000_expand_interleave): Likewise.
672         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
673         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
674         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
675         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
676         (rs6000_expand_vec_perm_const): Delete.
677         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
678         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
679         (altivec_expand_vec_perm_const_le): Take each operand individually.
680         Operate on constant selectors rather than rtxes.
681         (altivec_expand_vec_perm_const): Likewise.  Update call to
682         altivec_expand_vec_perm_const_le.
683         (rs6000_expand_vec_perm_const): Delete.
684         (rs6000_vectorize_vec_perm_const_ok): Delete.
685         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
686         reference to the SPE evmerge intructions.
687         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
688         an element count and rtx array.
689         (rs6000_expand_extract_even): Update call accordingly.
690         (rs6000_expand_interleave): Likewise.
691         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
692         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
693         new function.
694         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
696 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
698         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
699         vector mode and that that mode matches the mode of the data
700         being permuted.
701         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
702         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
703         directly using expand_vec_perm_1 when forcing selectors into
704         registers.
705         (expand_vec_perm_var): New function, split out from expand_vec_perm.
707 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
709         * optabs-query.h (can_vec_perm_p): Delete.
710         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
711         * optabs-query.c (can_vec_perm_p): Split into...
712         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
713         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
714         particular selector is valid.
715         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
716         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
717         (vect_grouped_load_supported): Likewise.
718         (vect_shift_permute_load_chain): Likewise.
719         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
720         (vect_transform_slp_perm_load): Likewise.
721         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
722         (vectorizable_bswap): Likewise.
723         (vect_gen_perm_mask_checked): Likewise.
724         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
725         implementations of variable permutation vectors into account
726         when deciding which selector to use.
727         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
728         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
729         with a false third argument.
730         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
731         to test whether the constant selector is valid and can_vec_perm_var_p
732         to test whether a variable selector is valid.
734 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
736         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
737         * optabs-query.c (can_vec_perm_p): Likewise.
738         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
739         instead of vec_perm_indices.
740         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
741         (vect_gen_perm_mask_checked): Likewise,
742         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
743         (vect_gen_perm_mask_checked): Likewise,
745 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
747         * optabs-query.h (qimode_for_vec_perm): Declare.
748         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
749         (qimode_for_vec_perm): ...this new function.
750         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
752 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
754         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
755         does not have a conditional at the top.
757 2018-01-02  Richard Biener  <rguenther@suse.de>
759         * ipa-inline.c (big_speedup_p): Fix expression.
761 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
763         PR target/81616
764         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
765         for generic 4->6.
767 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
769         PR target/81616
770         Generic tuning.
771         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
772         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
773         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
774         cond_taken_branch_cost 3->4.
776 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
778         PR tree-optimization/83581
779         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
780         TODO_cleanup_cfg if any changes have been made.
782         PR middle-end/83608
783         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
784         convert_modes if target mode has the right side, but different mode
785         class.
787         PR middle-end/83609
788         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
789         last argument when extracting from CONCAT.  If either from_real or
790         from_imag is NULL, use expansion through memory.  If result is not
791         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
792         the parts directly to inner mode, if even that fails, use expansion
793         through memory.
795         PR middle-end/83623
796         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
797         check for bswap in mode rather than HImode and use that in expand_unop
798         too.
800 2017-12-31  Jakub Jelinek  <jakub@redhat.com>
802         PR target/83536
803         * config/i386/i386.c (ix86_attribute_table): Remove excess
804         initializer for "shared" attribute.
806 2017-12-30  Tom de Vries  <tom@codesourcery.com>
808         PR libgomp/83046
809         * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
810         DECL_PRESERVE_P.
811         * lto-streamer-out.c (prune_offload_funcs): New function.  Remove
812         offload_funcs entries that no longer have a corresponding cgraph_node.
813         Mark the remaining ones as DECL_PRESERVE_P.
814         (output_lto): Call prune_offload_funcs.
816 2017-12-30  Jakub Jelinek  <jakub@redhat.com>
818         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
819         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
820         vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
821         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
822         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
823         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
824         vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
825         vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
826         vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
827         vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
828         vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
829         vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
830         vaesenc_<mode>, vpclmulqdq_<mode>,
831         avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
833 2017-12-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
835         * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
836         _Float<N> and _Float<N>X variants for rounding built-in
837         functions.
838         (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
839         (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
840         (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
841         (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
842         (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
843         * builtins.c (mathfn_built_in_2): Likewise.
844         * internal-fn.def (CEIL): Likewise.
845         (FLOOR): Likewise.
846         (NEARBYINT): Likewise.
847         (RINT): Likewise.
848         (ROUND): Likewise.
849         (TRUNC): Likewise.
850         * convert.c (convert_to_integer_1): Likewise.
851         * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
852         (integer_valued_real_call_p): Likewise.
853         * fold-const-call.c (fold_const_call_ss): Likewise.
854         * gencfn-macros.c (print_case_cfn): Change CFN and operator
855         printers to take a const char * suffix instead of a bool.
856         (print_define_operator_list): Likewise.
857         (fltall_suffixes): New list of suffixes, that include the
858         traditional suffixes as well as all of the _Float<N> and
859         _Float<N>X suffixes.
860         (main): For _Float<N> and _Float<N>X functions, emit both
861         <name>_FN and <name>_ALL variants.  The <macro>_FN variant only
862         has the _Float<N> and _Float<N>X case names or operators.  The
863         <name>_ALL variant has both the traditional and the
864         _Float<N>/_Float<N>X case names or operators.
865         * match.pd (COPYSIGN optimizations): Provide optimizations for
866         _Float<N> and _Float<N>X types where possible.
867         (MIN/MAX optimizations): Likewise.
868         (sqrt optimizations): Likewise.
869         (rounding optimizations): Likewise.
871 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
873         * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
874         gen_int_mode rather than GEN_INT.
875         * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
877 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
879         * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
880         instead of CONSTANT_P.
881         (gen_vec_series): Likewise.
882         * simplify-rtx.c (simplify_binary_operation_1): Likewise.
884 2017-12-28  Andreas Schwab  <schwab@linux-m68k.org>
886         * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
887         CC_STATUS_INIT.
889 2017-12-27  Kugan Vivekanandarajah  <kuganv@linaro.org>
891         * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
892         (aarch64_ld1x2<VDC:mode>): Likewise.
893         (aarch64_simd_ld1<mode>_x2): Likewise.
894         (aarch64_simd_ld1<mode>_x2): Likewise.
895         * config/aarch64/arm_neon.h (vld1_u8_x2): New.
896         (vld1_s8_x2): Likewise.
897         (vld1_u16_x2): Likewise.
898         (vld1_s16_x2): Likewise.
899         (vld1_u32_x2): Likewise.
900         (vld1_s32_x2): Likewise.
901         (vld1_u64_x2): Likewise.
902         (vld1_s64_x2): Likewise.
903         (vld1_f16_x2): Likewise.
904         (vld1_f32_x2): Likewise.
905         (vld1_f64_x2): Likewise.
906         (vld1_p8_x2): Likewise.
907         (vld1_p16_x2): Likewise.
908         (vld1_p64_x2): Likewise.
909         (vld1q_u8_x2): Likewise.
910         (vld1q_s8_x2): Likewise.
911         (vld1q_u16_x2): Likewise.
912         (vld1q_s16_x2): Likewise.
913         (vld1q_u32_x2): Likewise.
914         (vld1q_s32_x2): Likewise.
915         (vld1q_u64_x2): Likewise.
916         (vld1q_s64_x2): Likewise.
917         (vld1q_f16_x2): Likewise.
918         (vld1q_f32_x2): Likewise.
919         (vld1q_f64_x2): Likewise.
920         (vld1q_p8_x2): Likewise.
921         (vld1q_p16_x2): Likewise.
922         (vld1q_p64_x2): Likewise.
924 2017-12-27  Martin Liska  <mliska@suse.cz>
926         PR tree-optimization/83552
927         * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
928         of get_string_lenth to a SSA_NAME if not a GIMPLE value.
930 2017-12-27  Tom de Vries  <tom@codesourcery.com>
932         * config/nvptx/nvptx.c (nvptx_option_override): Disable
933         -gstatement-frontiers.
935 2017-12-26  Alexander Monakov  <amonakov@ispras.ru>
937         PR rtl-optimization/83513
938         * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
939         before priority comparison.
941 2017-12-25  Jakub Jelinek  <jakub@redhat.com>
943         PR target/83488
944         * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
945         ix86_isa_flags2 to ix86_isa_flags.
946         * config/i386/i386-c.c (ix86_target_macros_internal): Test
947         OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
948         isa_flags rather than isa_flags2.
949         * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
950         and -mavx512bitalg from isa2_opts to isa_opts.
951         (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
952         in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
953         and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
954         Formatting fixes.
955         (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
956         OPTION_MASK_ISA_AVX512F ored with another option similarly to
957         OPTION_MASK_ISA_AVX512VL.  Even for OPTION_MASK_ISA_AVX512VL don't
958         clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
959         (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
960         OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
961         OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
962         * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
963         builtins from bdesc_args2 to bdesc_args section.
964         (__builtin_ia32_compressstoreuqi512_mask,
965         __builtin_ia32_compressstoreuhi512_mask,
966         __builtin_ia32_compressstoreuqi256_mask,
967         __builtin_ia32_expandloadqi512_mask,
968         __builtin_ia32_expandloadqi512_maskz,
969         __builtin_ia32_expandloadhi512_mask,
970         __builtin_ia32_expandloadhi512_maskz,
971         __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
972         __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
973         __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
974         __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
975         __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
976         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
977         __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
978         __builtin_ia32_vpshldv_v32hi_maskz,
979         __builtin_ia32_vpopcountb_v64qi_mask,
980         __builtin_ia32_vpopcountw_v32hi_mask,
981         __builtin_ia32_vpshufbitqmb512_mask,
982         __builtin_ia32_vpshufbitqmb256_mask): Add
983         " | OPTION_MASK_ISA_AVX512BW".
984         (__builtin_ia32_expandloadqi256_mask,
985         __builtin_ia32_expandloadqi256_maskz,
986         __builtin_ia32_vpopcountb_v32qi_mask): Add
987         " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
988         (__builtin_ia32_expandloadhi256_mask,
989         __builtin_ia32_expandloadhi256_maskz,
990         __builtin_ia32_expandloadqi128_mask,
991         __builtin_ia32_expandloadqi128_maskz,
992         __builtin_ia32_expandloadhi128_mask,
993         __builtin_ia32_expandloadhi128_maskz,
994         __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
995         __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
996         __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
997         __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
998         __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
999         __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
1000         __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
1001         __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
1002         __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
1003         __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
1004         __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
1005         __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
1006         __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
1007         __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
1008         __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
1009         __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
1010         __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
1011         __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
1012         __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
1013         __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
1014         __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
1015         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
1016         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
1017         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
1018         __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
1019         __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
1020         __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
1021         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
1022         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
1023         __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
1024         __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
1025         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
1026         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1027         __builtin_ia32_vpopcountw_v8hi_mask): Add
1028         " | OPTION_MASK_ISA_AVX512VL".
1029         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
1030         _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1031         _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
1032         _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1033         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
1034         _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
1035         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
1036         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
1037         _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
1038         _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
1039         _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
1040         * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
1041         _mm_mask_bitshuffle_epi64_mask): Likewise.
1042         * common/config/i386/i386-common.c
1043         (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
1044         OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
1045         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1046         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
1047         OPTION_MASK_ISA_AVX512BITALG_UNSET.
1048         (OPTION_MASK_ISA2_AVX512F_UNSET,
1049         OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
1050         (ix86_handle_option): For -mno-general-regs-only, clear from
1051         ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
1052         just OPTION_MASK_ISA_MPX.  For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
1053         -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
1054         ix86_isa_flags2.  For -mno-avx512f likewise, instead of masking
1055         individually listed ISAs.  For -m{,no-}avx512{vpopcntdq,bitalg} adjust
1056         for moving from ix86_isa_flags2 to ix86_isa_flags.
1058 2017-12-23  Jakub Jelinek  <jakub@redhat.com>
1060         PR c++/83553
1061         * fold-const.c (struct contains_label_data): New type.
1062         (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
1063         inside of a SWITCH_BODY seen during the walk.
1064         (contains_label_p): Use walk_tree instead of
1065         walk_tree_without_duplicates, prepare data for contains_label_1 and
1066         provide own pset.
1068 2017-12-22  Jakub Jelinek  <jakub@redhat.com>
1070         PR debug/83547
1071         * tree-iterator.c (alloc_stmt_list): Start with cleared
1072         TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
1073         or old one reused.
1075         PR target/83488
1076         * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
1077         enable avx512f explicitly in #pragma GCC target.
1078         * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
1079         __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
1080         __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
1081         __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
1082         __builtin_ia32_vpdpbusds_v8si_mask,
1083         __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
1084         __builtin_ia32_vpdpbusds_v4si_mask,
1085         __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
1086         __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
1087         __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
1088         __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
1089         __builtin_ia32_vpdpwssds_v8si_mask,
1090         __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
1091         __builtin_ia32_vpdpwssds_v4si_mask,
1092         __builtin_ia32_vpdpwssds_v4si_maskz): Use
1093         OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
1094         just OPTION_MASK_ISA_AVX512VNNI.
1096 2017-12-22  Martin Jambor  <mjambor@suse.cz>
1098         PR lto/82027
1099         * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
1100         clones.
1102 2017-12-22  Julia Koval  <julia.koval@intel.com>
1103             Sebastian Peryt  <sebastian.peryt@intel.com>
1105         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
1106         OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
1107         (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
1108         * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
1109         * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
1110         _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
1111         _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
1112         _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
1113         _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
1114         _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
1115         _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
1116         _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
1117         _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
1118         * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
1119         _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
1120         _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
1121         _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
1122         _mm256_maskz_popcnt_epi64): New intrinsics.
1123         * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
1124         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
1125         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
1126         V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
1127         V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
1128         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
1129         __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
1130         __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
1131         __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
1132         __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
1133         __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
1134         __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
1135         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
1136         __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
1137         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1138         __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
1139         __builtin_ia32_vpshufbitqmb256_mask,
1140         __builtin_ia32_vpshufbitqmb512_mask): New builtins.
1141         * config/i386/i386-c.c (__AVX512BITALG__): New.
1142         * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
1143         (ix86_valid_target_attribute_inner_p): Ditto.
1144         (ix86_expand_args_builtin): Handle new types.
1145         * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
1146         * config/i386/i386.opt: Add -mavx512bitalg.
1147         * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
1148         avx512bitalgintrin.h.
1149         * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
1150         (vpopcount<mode><mask_name>): Add more types.
1151         (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
1152         * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
1154 2017-12-22  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
1156         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
1157         Or in OPTION_MASK_ISA_AVX512F_SET.
1158         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1159         OPTION_MASK_ISA_AVX512VNNI_UNSET.
1160         (ix86_handle_option): Adjust for
1161         OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
1162         * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
1163         section to ARGS.
1164         * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
1165         isa_flag instead of isa_flag2.
1166         * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
1167         isa_opts2 to isa_opts.
1168         * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
1169         to ix86_isa_flags.
1171 2017-12-22  Mike Stump  <mikestump@comcast.net>
1172             Eric Botcazou  <ebotcazou@adacore.com>
1174         * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
1176 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1178         PR middle-end/83487
1179         * config/i386/i386.c (ix86_function_arg_boundary): Return
1180         PARM_BOUNDARY for TYPE_EMPTY_P types.
1182         PR c/83448
1183         * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
1184         if navail is >= dir.len.
1186 2017-12-21  Steve Ellcey  <sellcey@cavium.com>
1188         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
1189         triplet for ilp32.
1191 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1193         PR rtl-optimization/80747
1194         PR rtl-optimization/83512
1195         * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
1196         succ edge from ENTRY, copy partition from e->dest to the newly
1197         created bb.
1198         * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
1199         ENTRY, use BB_PARTITION of its successor block as current_partition.
1200         Don't copy partition when splitting succ edge from ENTRY.
1202         PR tree-optimization/83523
1203         * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
1204         for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
1205         (convert_mult_to_fma): Likewise.
1207         PR tree-optimization/83521
1208         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
1209         gimple_build_assign without code on result of
1210         fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
1211         a VIEW_CONVERT_EXPR.
1213 2017-12-21  Andrew Pinski  <apinski@cavium.com>
1214             Steve Ellcey  <sellcey@cavium.com>
1216         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
1217         multi-arch for ilp32.
1219 2017-12-21  Uros Bizjak  <ubizjak@gmail.com>
1221         PR target/83467
1222         * config/i386/i386.md (*ashl<mode>3_mask): Add operand
1223         constraints to operand 2.
1224         (*ashl<mode>3_mask_1): Ditto.
1225         (*<shift_insn><mode>3_mask): Ditto.
1226         (*<shift_insn><mode>3_mask_1): Ditto.
1227         (*<rotate_insn><mode>3_mask): Ditto.
1228         (*<rotate_insn><mode>3_mask_1): Ditto.
1230 2017-12-21  Alexandre Oliva  <aoliva@redhat.com>
1232         * reorg.c (make_return_insns): Reemit each insn with its own location.
1234 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1236         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
1237         generation for cases where splatting a value is not useful.
1238         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
1239         across a vec_duplicate and a paradoxical subreg forming a vector
1240         mode to a vec_concat.
1242 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1244         * combine.c (simplify_set): Do not transform subregs to zero_extends
1245         if the destination is not a scalar int mode.
1247 2017-12-21  Eric Botcazou  <ebotcazou@adacore.com>
1249         PR c++/82872
1250         * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1251         shared zero if the input has overflowed.
1253 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1255         * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1256         support for Cortex-A55 and Cortex-A75.
1258 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1260         * common/config/arm/arm-common.c (compare_opt_names): Add function
1261         comment.  Use strcmp instead of manual loop.
1263 2017-12-21  Martin Liska  <mliska@suse.cz>
1265         PR gcov-profile/83509
1266         * gcov-dump.c (dump_gcov_file): Do not read info about
1267         support_unexecuted_blocks for gcda files.
1269 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1271         PR rtl-optimization/82973
1272         * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1273         (valid_for_const_vector_p): ... this.
1274         * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1275         (valid_for_const_vector_p): ... this.  Adjust function comment.
1276         (gen_vec_duplicate): Adjust caller.
1277         * optabs.c (expand_vector_broadcast): Likewise.
1278         * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1279         CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1280         constant.
1281         (simplify_const_binary_operation): Likewise.  Use CONST_FIXED_P macro
1282         instead of GET_CODE == CONST_FIXED.
1283         (simplify_subreg): Use CONST_FIXED_P macro instead of
1284         GET_CODE == CONST_FIXED.
1286         PR target/83488
1287         * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1288         -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1289         -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1290         (ix86_option_override_internal): Adjust for
1291         OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1292         and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1293         (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1294         (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1295         Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1296         and OPTION_MASK_ISA_CLZERO builtins.  Use def_builtin instead of
1297         def_builtin2 for CET builtins.
1298         (ix86_expand_builtin): Remove bdesc_special_args2 handling.  Fix
1299         up formatting in IX86_BUILTIN_RDPID code.
1300         * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1301         section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1302         * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1303         ix86_isa_flags2 to ix86_isa_flags.
1304         (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1305         ix86_isa_flags2.
1306         * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1307         OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1308         Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1309         of isa_flag2.
1310         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1311         Or in OPTION_MASK_ISA_AVX512F_SET.
1312         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1313         OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1314         (ix86_handle_option): Adjust for
1315         OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1316         and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1318 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1319             Alan Hayward  <alan.hayward@arm.com>
1320             David Sherwood  <david.sherwood@arm.com>
1322         * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1323         factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1324         * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1325         Track polynomial offsets.
1327 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1328             Alan Hayward  <alan.hayward@arm.com>
1329             David Sherwood  <david.sherwood@arm.com>
1331         * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
1332         Treat drb->init as a poly_int.  Fail if its misalignment wrt
1333         vector_alignment isn't known.
1335 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1336             Alan Hayward  <alan.hayward@arm.com>
1337             David Sherwood  <david.sherwood@arm.com>
1339         * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1340         field.
1341         (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1342         (vect_loop_versioning): Take the loop versioning threshold as a
1343         separate parameter.
1344         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1345         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1346         versioning_threshold.
1347         (vect_analyze_loop_2): Compute the loop versioning threshold
1348         whenever loop versioning is needed, and store it in the new
1349         field rather than combining it with the cost model threshold.
1350         (vect_transform_loop): Update call to vect_loop_versioning.
1351         Try to combine the loop versioning and cost thresholds here.
1353 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1354             Alan Hayward  <alan.hayward@arm.com>
1355             David Sherwood  <david.sherwood@arm.com>
1357         * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1358         poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1359         * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1360         * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1361         an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1362         (group_compare_offset): Update accordingly.
1363         (split_small_address_groups_p): Likewise.
1364         (record_use): Take addr_offset as a poly_uint64 rather than
1365         an unsigned HOST_WIDE_INT.
1366         (strip_offset): Return the offset as a poly_uint64 rather than
1367         an unsigned HOST_WIDE_INT.
1368         (record_group_use, split_address_groups): Track polynomial offsets.
1369         (add_iv_candidate_for_use): Likewise.
1370         (addr_offset_valid_p): Take the offset as a poly_int64 rather
1371         than a HOST_WIDE_INT.
1372         (strip_offset_1): Return the offset as a poly_int64 rather than
1373         a HOST_WIDE_INT.
1375 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1376             Alan Hayward  <alan.hayward@arm.com>
1377             David Sherwood  <david.sherwood@arm.com>
1379         * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1380         rather than a HOST_WIDE_INT.
1381         * tree.c (get_binfo_at_offset): Likewise.
1383 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1384             Alan Hayward  <alan.hayward@arm.com>
1385             David Sherwood  <david.sherwood@arm.com>
1387         * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1388         rather than a HOST_WIDE_INT.
1389         * tree-sra.c (build_ref_for_offset): Likewise.
1391 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1392             Alan Hayward  <alan.hayward@arm.com>
1393             David Sherwood  <david.sherwood@arm.com>
1395         * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1396         than an offset_int.
1397         * tree.c (mem_ref_offset): Likewise.
1398         (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1399         * builtins.c (get_object_alignment_2): Likewise.
1400         * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1401         * gimple-fold.c (get_base_constructor): Likewise.
1402         * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1403         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1404         Likewise.
1405         * ipa-polymorphic-call.c
1406         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1407         * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1408         (get_ancestor_addr_info): Likewise.
1409         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1410         * match.pd: Likewise.
1411         * tree-data-ref.c (dr_analyze_innermost): Likewise.
1412         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1413         * tree-eh.c (tree_could_trap_p): Likewise.
1414         * tree-object-size.c (addr_object_size): Likewise.
1415         * tree-ssa-address.c (copy_ref_info): Likewise.
1416         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1417         (indirect_refs_may_alias_p): Likewise.
1418         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1419         * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1420         (non_rewritable_mem_ref_base): Likewise.
1421         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1422         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1423         * varasm.c (decode_addr_const): Likewise.
1425 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1426             Alan Hayward  <alan.hayward@arm.com>
1427             David Sherwood  <david.sherwood@arm.com>
1429         * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1430         offsets for MEM_REFs.
1432 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1433             Alan Hayward  <alan.hayward@arm.com>
1434             David Sherwood  <david.sherwood@arm.com>
1436         * tree-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1437         as a poly_uint64 rather than an unsigned int.
1438         * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1440 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1441             Alan Hayward  <alan.hayward@arm.com>
1442             David Sherwood  <david.sherwood@arm.com>
1444         * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1445         to poly_int64.
1446         (decode_addr_const): Update accordingly.
1448 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1449             Alan Hayward  <alan.hayward@arm.com>
1450             David Sherwood  <david.sherwood@arm.com>
1452         * tree.h (bit_field_size, bit_field_offset): New functions.
1453         * hsa-gen.c (gen_hsa_addr): Use them.
1454         * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1455         (simplify_vector_constructor): Likewise.
1456         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1457         * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1458         BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1459         * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1460         tree_fits_uhwi_p.
1462 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1463             Alan Hayward  <alan.hayward@arm.com>
1464             David Sherwood  <david.sherwood@arm.com>
1466         * expr.h (emit_group_load, emit_group_load_into_temps)
1467         (emit_group_store): Take the size as a poly_int64 rather than an int.
1468         * expr.c (emit_group_load_1, emit_group_load): Likewise.
1469         (emit_group_load_into_temp, emit_group_store): Likewise.
1471 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1472             Alan Hayward  <alan.hayward@arm.com>
1473             David Sherwood  <david.sherwood@arm.com>
1475         * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1476         an unsigned int to a poly_uint64.
1477         * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1478         sizes as poly_uint64s rather than unsigned ints.
1479         * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1480         Likewise.
1482 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1483             Alan Hayward  <alan.hayward@arm.com>
1484             David Sherwood  <david.sherwood@arm.com>
1486         * lra-constraints.c (emit_inc): Change inc_amount from an int
1487         to a poly_int64.
1489 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1490             Alan Hayward  <alan.hayward@arm.com>
1491             David Sherwood  <david.sherwood@arm.com>
1493         * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1494         to a poly_uint64.
1495         (add_stack_var, stack_var_cmp, partition_stack_vars)
1496         (dump_stack_var_partition): Update accordingly.
1497         (alloc_stack_frame_space): Take the size as a poly_int64 rather
1498         than a HOST_WIDE_INT.
1499         (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1500         (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1501         (account_stack_vars, expand_one_var): Likewise.  Return a poly_uint64
1502         rather than a HOST_WIDE_INT.
1504 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1505             Alan Hayward  <alan.hayward@arm.com>
1506             David Sherwood  <david.sherwood@arm.com>
1508         * target.def (return_pops_args): Treat both the input and output
1509         sizes as poly_int64s rather than HOST_WIDE_INTS.
1510         * targhooks.h (default_return_pops_args): Update accordingly.
1511         * targhooks.c (default_return_pops_args): Likewise.
1512         * doc/tm.texi: Regenerate.
1513         * emit-rtl.h (incoming_args): Change pops_args, size and
1514         outgoing_args_size from int to poly_int64_pod.
1515         * function.h (expr_status): Change x_pending_stack_adjust and
1516         x_stack_pointer_delta from int to poly_int64.
1517         (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1518         (ARGS_SIZE_RTX): Update accordingly.
1519         * calls.c (highest_outgoing_arg_in_use): Change from int to
1520         unsigned int.
1521         (stack_usage_watermark, stored_args_watermark): New variables.
1522         (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1523         (emit_call_1): Change the stack_size and rounded_stack_size
1524         parameters from HOST_WIDE_INT to poly_int64.  Track n_popped
1525         as a poly_int64.
1526         (save_fixed_argument_area): Check stack_usage_watermark.
1527         (initialize_argument_information): Change old_pending_adj from
1528         a HOST_WIDE_INT * to a poly_int64_pod *.
1529         (compute_argument_block_size): Return the size as a poly_int64
1530         rather than an int.
1531         (finalize_must_preallocate): Track polynomial argument sizes.
1532         (compute_argument_addresses): Likewise.
1533         (internal_arg_pointer_based_exp): Track polynomial offsets.
1534         (mem_overlaps_already_clobbered_arg_p): Rename to...
1535         (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1536         size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1537         Check stored_args_used_watermark.
1538         (load_register_parameters): Update accordingly.
1539         (check_sibcall_argument_overlap_1): Likewise.
1540         (combine_pending_stack_adjustment_and_call): Take the unadjusted
1541         args size as a poly_int64 rather than an int.  Return a bool
1542         indicating whether the optimization was possible and return
1543         the new adjustment by reference.
1544         (check_sibcall_argument_overlap): Track polynomail argument sizes.
1545         Update stored_args_watermark.
1546         (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1547         (expand_call): Likewise.  Maintain stack_usage_watermark and
1548         stored_args_watermark.  Update calls to
1549         combine_pending_stack_adjustment_and_call.
1550         (emit_library_call_value_1): Handle polynomial argument sizes.
1551         Call stack_region_maybe_used_p and mark_stack_region_used.
1552         Maintain stack_usage_watermark.
1553         (store_one_arg): Likewise.  Update call to
1554         mem_overlaps_already_clobbered_arg_p.
1555         * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1556         HOST_WIDE_INT.
1557         * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1558         * config/microblaze/microblaze.c (microblaze_function_prologue):
1559         Likewise.
1560         * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1561         TARGET_RETURN_POPS_ARGS interface.
1562         (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1563         to HOST_WIDE_INT.
1564         * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1565         * config/i386/i386.c (ix86_return_pops_args): Update for new
1566         TARGET_RETURN_POPS_ARGS interface.
1567         (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1568         * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1569         * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1570         TARGET_RETURN_POPS_ARGS interface.
1571         * config/vax/vax.c (vax_return_pops_args): Likewise.
1572         * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1573         (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1574         * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1575         * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1576         * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1577         * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1578         * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1579         * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1580         * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1581         * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1582         and x_stack_pointer_delta from int to poly_int64.
1583         * dojump.c (do_pending_stack_adjust): Update accordingly.
1584         * explow.c (allocate_dynamic_stack_space): Handle polynomial
1585         stack_pointer_deltas.
1586         * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1587         (pad_to_arg_alignment): Track polynomial offsets.
1588         (assign_parm_find_stack_rtl): Likewise.
1589         (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1590         * toplev.c (output_stack_usage): Update reference to
1591         current_function_pushed_stack_size.
1593 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1594             Alan Hayward  <alan.hayward@arm.com>
1595             David Sherwood  <david.sherwood@arm.com>
1597         * function.c (in_arg_offset, var_offset, dynamic_offset)
1598         (out_arg_offset, cfa_offset): Change from int to poly_int64.
1599         (instantiate_new_reg): Return the new offset as a poly_int64_pod
1600         rather than a HOST_WIDE_INT.
1601         (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1602         (instantiate_virtual_regs_in_insn): Likewise.
1604 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1605             Alan Hayward  <alan.hayward@arm.com>
1606             David Sherwood  <david.sherwood@arm.com>
1608         * rtl.h (get_args_size, add_args_size_note): New functions.
1609         (find_args_size_adjust): Return a poly_int64 rather than a
1610         HOST_WIDE_INT.
1611         (fixup_args_size_notes): Likewise.  Make the same change to the
1612         end_args_size parameter.
1613         * rtlanal.c (get_args_size, add_args_size_note): New functions.
1614         * builtins.c (expand_builtin_trap): Use add_args_size_note.
1615         * calls.c (emit_call_1): Likewise.
1616         * explow.c (adjust_stack_1): Likewise.
1617         * cfgcleanup.c (old_insns_match_p): Update use of
1618         find_args_size_adjust.
1619         * combine.c (distribute_notes): Track polynomial arg sizes.
1620         * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1621         end_true_args_size, beg_delay_args_size and end_delay_args_size
1622         from HOST_WIDE_INT to poly_int64.
1623         (add_cfi_args_size): Take the args_size as a poly_int64 rather
1624         than a HOST_WIDE_INT.
1625         (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1626         (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1627         polynomial arg sizes.
1628         * emit-rtl.c (try_split): Use get_args_size.
1629         * recog.c (peep2_attempt): Likewise.
1630         * reload1.c (reload_as_needed): Likewise.
1631         * expr.c (find_args_size_adjust): Return the adjustment as a
1632         poly_int64 rather than a HOST_WIDE_INT.
1633         (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1634         to a poly_int64 and change the return type in the same way.
1635         (emit_single_push_insn): Track polynomial arg sizes.
1637 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1638             Alan Hayward  <alan.hayward@arm.com>
1639             David Sherwood  <david.sherwood@arm.com>
1641         * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1642         from HOST_WIDE_INT to poly_int64.
1643         * expr.c (push_block, emit_push_insn): Likewise.
1645 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1646             Alan Hayward  <alan.hayward@arm.com>
1647             David Sherwood  <david.sherwood@arm.com>
1649         * function.h (frame_space): Change start and length from HOST_WIDE_INT
1650         to poly_int64.
1651         (get_frame_size): Return the size as a poly_int64 rather than a
1652         HOST_WIDE_INT.
1653         (frame_offset_overflow): Take the offset as a poly_int64 rather
1654         than a HOST_WIDE_INT.
1655         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1656         (assign_stack_temp): Likewise for the size.
1657         * function.c (get_frame_size): Return a poly_int64 rather than
1658         a HOST_WIDE_INT.
1659         (frame_offset_overflow): Take the offset as a poly_int64 rather
1660         than a HOST_WIDE_INT.
1661         (try_fit_stack_local): Take the start, length and size as poly_int64s
1662         rather than HOST_WIDE_INTs.  Return the offset as a poly_int64_pod
1663         rather than a HOST_WIDE_INT.
1664         (add_frame_space): Take the start and end as poly_int64s rather than
1665         HOST_WIDE_INTs.
1666         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1667         (assign_stack_temp): Likewise for the size.
1668         (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1669         to poly_int64.
1670         (find_temp_slot_from_address): Handle polynomial offsets.
1671         (combine_temp_slots): Likewise.
1672         * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1673         to poly_int64.
1674         * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1675         poly_int64 rather than a HOST_WIDE_INT.
1676         (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1677         than a HOST_WIDE_INT.
1678         (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1679         polynomial frame offsets.
1680         * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1681         as a poly_int64 rather than an int.
1682         * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1683         * config/v850/v850-protos.h (compute_frame_size): Likewise.
1684         * config/v850/v850.c (compute_frame_size): Likewise.
1685         * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1686         * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1687         * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1688         * config/pa/pa.c (pa_compute_frame_size): Likewise.
1689         * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1690         rather than a HOST_WIDE_INT.
1691         * explow.c (get_dynamic_stack_base): Likewise.
1692         * final.c (final_start_function): Use the constant lower bound
1693         of the frame size for -Wframe-larger-than.
1694         * ira.c (do_reload): Adjust for new get_frame_size return type.
1695         * lra.c (lra): Likewise.
1696         * reload1.c (reload): Likewise.
1697         * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1698         * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1699         * rtlanal.c (get_initial_register_offset): Return the offset as
1700         a poly_int64 rather than a HOST_WIDE_INT.
1702 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1703             Alan Hayward  <alan.hayward@arm.com>
1704             David Sherwood  <david.sherwood@arm.com>
1706         * reload1.c (elim_table): Change initial_offset, offset and
1707         previous_offset from HOST_WIDE_INT to poly_int64_pod.
1708         (offsets_at): Change the target array's element type from
1709         HOST_WIDE_INT to poly_int64_pod.
1710         (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1711         (elimination_costs_in_insn, update_eliminable_offsets)
1712         (verify_initial_elim_offsets, set_offsets_for_label)
1713         (init_eliminable_invariants): Update after above changes.
1715 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1716             Alan Hayward  <alan.hayward@arm.com>
1717             David Sherwood  <david.sherwood@arm.com>
1719         * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1720         * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1721         (decomposition): Change start and end from HOST_WIDE_INT
1722         to poly_int64_pod.
1723         (decompose, immune_p): Update accordingly.
1724         (find_inc_amount): Return a poly_int64 rather than an int.
1725         * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1726         rather than an int.
1728 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1729             Alan Hayward  <alan.hayward@arm.com>
1730             David Sherwood  <david.sherwood@arm.com>
1732         * tree.h (get_inner_reference): Return the bitsize and bitpos
1733         as poly_int64_pods rather than HOST_WIDE_INT.
1734         * fold-const.h (ptr_difference_const): Return the pointer difference
1735         as a poly_int64_pod rather than a HOST_WIDE_INT.
1736         * expr.c (get_inner_reference): Return the bitsize and bitpos
1737         as poly_int64_pods rather than HOST_WIDE_INT.
1738         (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1739         offsets and sizes.
1740         * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1741         rather than a HOST_WIDE_INT.  Update call to get_inner_reference.
1742         (optimize_bit_field_compare): Update call to get_inner_reference.
1743         (decode_field_reference): Likewise.
1744         (fold_unary_loc): Track polynomial offsets and sizes.
1745         (split_address_to_core_and_offset): Return the bitpos as a
1746         poly_int64_pod rather than a HOST_WIDE_INT.
1747         (ptr_difference_const): Likewise for the pointer difference.
1748         * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1749         * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1750         * dbxout.c (dbxout_expand_expr): Likewise.
1751         * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1752         (loc_list_from_tree_1, fortran_common): Likewise.
1753         * gimple-laddress.c (pass_laddress::execute): Likewise.
1754         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1755         * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1756         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1757         * tree-affine.c (tree_to_aff_combination): Likewise.
1758         (get_inner_reference_aff): Likewise.
1759         * tree-data-ref.c (split_constant_offset_1): Likewise.
1760         (dr_analyze_innermost): Likewise.
1761         * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1762         * tree-sra.c (ipa_sra_check_caller): Likewise.
1763         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1764         * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1765         (instrument_bool_enum_load, instrument_object_size): Likewise.
1766         * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1767         to get_inner_reference.
1768         * hsa-gen.c (gen_hsa_addr): Likewise.
1769         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1770         * tsan.c (instrument_expr): Likewise.
1771         * match.pd: Update call to ptr_difference_const.
1773 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1774             Alan Hayward  <alan.hayward@arm.com>
1775             David Sherwood  <david.sherwood@arm.com>
1777         * fold-const.c (fold_comparison): Track sizes and offsets as
1778         poly_int64s rather than HOST_WIDE_INTs when folding address
1779         comparisons.
1781 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1782             Alan Hayward  <alan.hayward@arm.com>
1783             David Sherwood  <david.sherwood@arm.com>
1785         * expr.h (get_bit_range): Return the bitstart and bitend as
1786         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Return the bitpos
1787         as a poly_int64 rather than a HOST_WIDE_INT.
1788         * expr.c (get_bit_range): Likewise.
1789         (expand_assignment): Update call accordingly.
1790         * fold-const.c (optimize_bit_field_compare): Likewise.
1792 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1793             Alan Hayward  <alan.hayward@arm.com>
1794             David Sherwood  <david.sherwood@arm.com>
1796         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1797         (aarch64_legitimate_address_p): Use it instead of an rtx code,
1798         as an optional final parameter.
1799         * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1800         (aarch64_legitimate_address_p): Likewise.
1801         (aarch64_print_address_internal): Take an aarch64_addr_query_type
1802         instead of an rtx code.
1803         (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1804         (aarch64_legitimate_address_hook_p): Likewise.
1805         (aarch64_print_ldpstp_address): Likewise.
1806         (aarch64_print_operand_address): Likewise.
1807         (aarch64_address_cost): Likewise.
1808         * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1809         * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1810         (aarch64_mem_pair_lanes_operand): Likewise.
1812 2017-12-20  Richard Biener  <rguenther@suse.de>
1814         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1815         update_stmt_if_modified.
1817 2017-12-20  Wilco Dijkstra  <wdijkstr@arm.com>
1819         PR tree-optimization/83491
1820         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1821         before walking uses.  Improve coding style and comments.
1823 2017-12-20  Tom de Vries  <tom@codesourcery.com>
1825         * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1827 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
1829         PR ipa/83506
1830         * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1831         todo_flags_finish.
1832         (pass_ipa_free_fn_summary): Add small_p private data member,
1833         initialize to false in the ctor.
1834         (pass_ipa_free_fn_summary::clone,
1835         pass_ipa_free_fn_summary::set_pass_param,
1836         pass_ipa_free_fn_summary::gate): New methods.
1837         (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1838         | TODO_dump_symtab if small_p.
1839         * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1840         entry and add another instance of the pass with false parm after
1841         ipa-pure-const.
1842         * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1843         ipa_free_fn_summary here.
1845 2017-12-20  Paolo Carlini  <paolo.carlini@oracle.com>
1847         * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1849 2017-12-20  Martin Sebor  <msebor@redhat.com>
1851         PR testsuite/83131
1852         * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1853         (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1854         (expand_builtin_strncmp): Same.
1856 2017-12-20  Alexandre Oliva  <aoliva@redhat.com>
1858         PR bootstrap/83396
1859         * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1860         allowed debug stmts before labels.
1861         (expand_gimple_basic_block): Likewise.
1862         * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1863         * gimple-iterator.h (gsi_after_labels): Likewise.
1864         * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1865         rename reused variable, and simplify using gsi_move_before.
1866         * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1867         * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1868         (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1869         (gimple_verify_flow_info, gimple_block_label): Likewise.
1870         (make_blocks): Move debug markers after adjacent labels.
1871         * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1872         allowed debug insns outside blocks.
1873         * df-scan.c (df_insn_delete): Likewise.
1874         * lra-constraints.c (update_ebb_live_info): Likewise.
1875         * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1876         (vt_initialize, delete_vta_debug_insns): Likewise.
1877         (reemit_marker_as_note): Drop BB parm.  Adjust callers.
1879 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1880             Alan Hayward  <alan.hayward@arm.com>
1881             David Sherwood  <david.sherwood@arm.com>
1883         * poly-int-types.h (round_down_to_byte_boundary): New macro.
1884         (round_up_to_byte_boundary): Likewise.
1885         * expr.h (get_bit_range): Add temporary shim.
1886         * gimple-ssa-store-merging.c (store_operand_info): Change the
1887         bitsize, bitpos, bitregion_start and bitregion_end fields from
1888         unsigned HOST_WIDE_INT to poly_uint64.
1889         (merged_store_group): Likewise load_align_base.
1890         (compatible_load_p, compatible_load_p): Update accordingly.
1891         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1892         (split_group, imm_store_chain_info::output_merged_store): Likewise.
1893         (mem_valid_for_store_merging): Return the bitsize, bitpos,
1894         bitregion_start and bitregion_end as poly_uint64s rather than
1895         unsigned HOST_WIDE_INTs.  Track polynomial offsets internally.
1896         (handled_load): Take the bitsize, bitpos,
1897         bitregion_start and bitregion_end as poly_uint64s rather than
1898         unsigned HOST_WIDE_INTs.
1899         (pass_store_merging::process_store): Update call to
1900         mem_valid_for_store_merging.
1902 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1903             Alan Hayward  <alan.hayward@arm.com>
1904             David Sherwood  <david.sherwood@arm.com>
1906         * builtins.c (get_object_alignment_2): Track polynomial offsets
1907         and sizes.  Update the alignment handling.
1909 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1910             Alan Hayward  <alan.hayward@arm.com>
1911             David Sherwood  <david.sherwood@arm.com>
1913         * tree.h (get_inner_reference): Add a version that returns the
1914         offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1915         * cfgexpand.c (expand_debug_expr): Track polynomial offsets.  Simply
1916         the case in which bitpos is not associated with the first byte.
1918 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1919             Alan Hayward  <alan.hayward@arm.com>
1920             David Sherwood  <david.sherwood@arm.com>
1922         * tree-affine.h (get_inner_reference_aff): Return the size as a
1923         poly_widest_int.
1924         * tree-affine.c (get_inner_reference_aff): Likewise.
1925         * tree-data-ref.c (dr_may_alias_p): Update accordingly.
1926         * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
1928 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1929             Alan Hayward  <alan.hayward@arm.com>
1930             David Sherwood  <david.sherwood@arm.com>
1932         * fold-const.c (pointer_may_wrap_p): Take the offset as a
1933         HOST_WIDE_INT rather than a poly_int64.
1935 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1936             Alan Hayward  <alan.hayward@arm.com>
1937             David Sherwood  <david.sherwood@arm.com>
1939         * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
1940         HOST_WIDE_INT to poly_int64_pod.
1941         (perform_symbolic_merge): Update accordingly.
1942         (bswap_replace): Likewise.
1944 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1945             Alan Hayward  <alan.hayward@arm.com>
1946             David Sherwood  <david.sherwood@arm.com>
1948         * tree-affine.h (aff_tree::offset): Change from widest_int
1949         to poly_widest_int.
1950         (wide_int_ext_for_comb): Delete.
1951         (aff_combination_const, aff_comb_cannot_overlap_p): Take the
1952         constants as poly_widest_int rather than widest_int.
1953         (aff_combination_constant_multiple_p): Return the multiplier
1954         as a poly_widest_int.
1955         (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
1956         polynomial offsets.
1957         * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
1958         version static and add an overload for poly_widest_int.
1959         (aff_combination_const, aff_combination_add_cst)
1960         (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
1961         the constants as poly_widest_int rather than widest_int.
1962         (tree_to_aff_combination): Generalize INTEGER_CST case to
1963         poly_int_tree_p.
1964         (aff_combination_to_tree): Track offsets as poly_widest_ints.
1965         (aff_combination_add_product, aff_combination_mult): Handle
1966         polynomial offsets.
1967         (aff_combination_constant_multiple_p): Return the multiplier
1968         as a poly_widest_int.
1969         * tree-predcom.c (determine_offset): Return the offset as a
1970         poly_widest_int.
1971         (split_data_refs_to_components, suitable_component_p): Update
1972         accordingly.
1973         (valid_initializer_p): Update call to
1974         aff_combination_constant_multiple_p.
1975         * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
1976         * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
1977         as a poly_int64 rather than a HOST_WIDE_INT.
1978         (get_address_cost): Handle polynomial offsets.
1979         (iv_elimination_compare_lt): Likewise.
1980         (rewrite_use_nonlinear_expr): Likewise.
1982 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1983             Alan Hayward  <alan.hayward@arm.com>
1984             David Sherwood  <david.sherwood@arm.com>
1986         * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
1987         as a poly_int64_pod rather than a HOST_WIDE_INT.
1988         (get_addr_base_and_unit_offset): Likewise.
1989         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1990         (get_addr_base_and_unit_offset): Likewise.
1991         * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
1992         to poly_int64 in example.
1993         * fold-const.c (fold_binary_loc): Update call to
1994         get_addr_base_and_unit_offset.
1995         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1996         (maybe_canonicalize_mem_ref_addr): Likewise.
1997         (gimple_fold_stmt_to_constant_1): Likewise.
1998         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1999         Likewise.
2000         * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
2001         * match.pd: Likewise.
2002         * omp-low.c (lower_omp_target): Likewise.
2003         * tree-sra.c (build_ref_for_offset): Likewise.
2004         (build_debug_ref_for_model): Likewise.
2005         * tree-ssa-address.c (maybe_fold_tmr): Likewise.
2006         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
2007         * tree-ssa-ccp.c (optimize_memcpy): Likewise.
2008         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
2009         (constant_pointer_difference): Likewise.
2010         * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
2011         * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
2012         * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
2013         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
2014         (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
2015         (set_ssa_val_to): Likewise.
2016         * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
2017         (maybe_diag_stxncpy_trunc): Likewise.
2018         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
2019         * tree.c (build_simple_mem_ref_loc): Likewise.
2020         (array_at_struct_end_p): Likewise.
2022 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2023             Alan Hayward  <alan.hayward@arm.com>
2024             David Sherwood  <david.sherwood@arm.com>
2026         * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
2027         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
2028         (get_ref_base_and_extent_hwi): Declare.
2029         * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
2030         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
2031         (get_ref_base_and_extent_hwi): New function.
2032         * cfgexpand.c (expand_debug_expr): Update call to
2033         get_ref_base_and_extent.
2034         * dwarf2out.c (add_var_loc_to_decl): Likewise.
2035         * gimple-fold.c (get_base_constructor): Return the offset as a
2036         poly_int64_pod rather than a HOST_WIDE_INT.
2037         (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
2038         * ipa-polymorphic-call.c
2039         (ipa_polymorphic_call_context::set_by_invariant)
2040         (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
2041         (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
2042         (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
2043         rather than get_ref_base_and_extent.
2044         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
2045         * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
2046         (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
2047         Likewise.
2048         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
2049         call to get_ref_base_and_extent.
2050         * tree-sra.c (create_access, get_access_for_expr): Likewise.
2051         * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
2052         (stmt_kills_ref_p): Likewise.
2053         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2054         * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
2055         Likewise.
2056         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
2057         Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
2058         when calling native_encode_expr.
2059         * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
2060         call to get_ref_base_and_extent.
2061         (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
2062         get_ref_base_and_extent.
2063         * var-tracking.c (track_expr_p): Likewise.
2065 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2066             Alan Hayward  <alan.hayward@arm.com>
2067             David Sherwood  <david.sherwood@arm.com>
2069         * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
2070         HOST_WIDE_INT to poly_int64_pod.
2071         * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
2072         polynomail parameter offsets.
2074 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2075             Alan Hayward  <alan.hayward@arm.com>
2076             David Sherwood  <david.sherwood@arm.com>
2078         * gengtype.c (main): Handle poly_int64_pod.
2079         * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
2080         (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
2081         (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
2082         from HOST_WIDE_INT to poly_int64_pod.
2083         * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
2084         (copy_cfa): New function.
2085         (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
2086         (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
2087         (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
2088         (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
2089         (initial_return_save): Treat offsets as poly_ints.
2090         (def_cfa_0): Likewise.  Cache the CFA in dw_cfi_cfa_loc if either
2091         offset is nonconstant.
2092         (reg_save): Take the offset as a poly_int64.  Fall back to
2093         DW_CFA_expression for nonconstant offsets.
2094         (queue_reg_save): Take the offset as a poly_int64.
2095         * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
2097 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2098             Alan Hayward  <alan.hayward@arm.com>
2099             David Sherwood  <david.sherwood@arm.com>
2101         * rtl.h (operand_subword, operand_subword_force): Take the offset
2102         as a poly_uint64 an unsigned int.
2103         * emit-rtl.c (operand_subword, operand_subword_force): Likewise.
2105 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2106             Alan Hayward  <alan.hayward@arm.com>
2107             David Sherwood  <david.sherwood@arm.com>
2109         * doc/rtl.texi: Update documentation of SUBREG_BYTE.  Document the
2110         'p' format code.  Use INT_LIST rather than SUBREG as the example of
2111         a code with an XINT and an XEXP.  Remove the implication that
2112         accessing an rtx field using XINT is expected to work.
2113         * rtl.def (SUBREG): Change format from "ei" to "ep".
2114         * rtl.h (rtunion::rt_subreg): New field.
2115         (XCSUBREG): New macro.
2116         (SUBREG_BYTE): Use it.
2117         (subreg_shape): Change offset from an unsigned int to a poly_uint16.
2118         Update constructor accordingly.
2119         (subreg_shape::operator ==): Update accordingly.
2120         (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
2121         than an unsigned int.
2122         (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
2123         a poly_uint64 rather than an unsigned int.
2124         (subreg_lsb_1): Likewise.  Take the offset as a poly_uint64 rather
2125         than an unsigned int.
2126         (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
2127         (subreg_size_highpart_offset): Return a poly_uint64 rather than
2128         an unsigned int.  Take the sizes as poly_uint64s.
2129         (subreg_offset_from_lsb): Return a poly_uint64 rather than
2130         an unsigned int.  Take the shift as a poly_uint64 rather than
2131         an unsigned int.
2132         (subreg_regno_offset, subreg_offset_representable_p): Take the offset
2133         as a poly_uint64 rather than an unsigned int.
2134         (simplify_subreg_regno): Likewise.
2135         (byte_lowpart_offset): Return the memory offset as a poly_int64
2136         rather than an int.
2137         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2138         poly_uint64 rather than an unsigned int.
2139         (simplify_subreg, simplify_gen_subreg, subreg_get_info)
2140         (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
2141         poly_uint64 rather than an unsigned int.
2142         * rtl.c (rtx_format): Describe 'p' in comment.
2143         (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
2144         * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
2145         offset as a poly_uint64 rather than an unsigned int.
2146         (byte_lowpart_offset): Return the memory offset as a poly_int64
2147         rather than an int.
2148         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2149         poly_uint64 rather than an unsigned int.
2150         (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
2151         mode sizes as poly_uint64s rather than unsigned ints.  Return a
2152         poly_uint64 rather than an unsigned int.
2153         (subreg_lowpart_p): Treat subreg offsets as poly_ints.
2154         (copy_insn_1): Handle 'p'.
2155         * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
2156         (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
2157         an unsigned int.  Return the shift in the same way.
2158         (subreg_lsb): Return the shift as a poly_uint64 rather than an
2159         unsigned int.
2160         (subreg_size_offset_from_lsb): Take the sizes and shift as
2161         poly_uint64s rather than unsigned ints.  Return the offset as
2162         a poly_uint64.
2163         (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
2164         (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
2165         an unsigned int.
2166         * rtlhash.c (add_rtx): Handle 'p'.
2167         * genemit.c (gen_exp): Likewise.
2168         * gengenrtl.c (type_from_format, gendef): Likewise.
2169         * gensupport.c (subst_pattern_match, get_alternatives_number)
2170         (collect_insn_data, alter_predicate_for_insn, alter_constraints)
2171         (subst_dup): Likewise.
2172         * gengtype.c (adjust_field_rtx_def): Likewise.
2173         * genrecog.c (find_operand, find_matching_operand, validate_pattern)
2174         (match_pattern_2): Likewise.
2175         (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
2176         (rtx_test::subreg_field): New function.
2177         (operator ==, safe_to_hoist_p, transition_parameter_type)
2178         (print_nonbool_test, print_test): Handle SUBREG_FIELD.
2179         * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
2180         * genpeep.c (match_rtx): Likewise.
2181         * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
2182         (rtx_writer::print_rtx_operand): Handle 'p'.
2183         (print_value): Handle SUBREG.
2184         * read-rtl.c (apply_int_iterator): Likewise.
2185         (rtx_reader::read_rtx_operand): Handle 'p'.
2186         * alias.c (rtx_equal_for_memref_p): Likewise.
2187         * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
2188         * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
2189         as poly_ints.
2190         * calls.c (expand_call): Likewise.
2191         * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
2192         (make_extraction, gen_lowpart_for_combine): Likewise.
2193         * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
2194         Likewise.
2195         * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
2196         rather than an unsigned int.  Treat subreg offsets as poly_ints.
2197         (exp_equiv_p): Handle 'p'.
2198         (hash_rtx_cb): Likewise.  Treat subreg offsets as poly_ints.
2199         (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
2200         * dse.c (find_shift_sequence): Likewise.
2201         * dwarf2out.c (rtl_for_decl_location): Likewise.
2202         * expmed.c (extract_low_bits): Likewise.
2203         * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
2204         (expand_expr_real_2): Likewise.
2205         * final.c (alter_subreg): Likewise.
2206         (leaf_renumber_regs_insn): Handle 'p'.
2207         * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
2208         Treat subreg offsets as poly_ints.
2209         * fwprop.c (forward_propagate_and_simplify): Likewise.
2210         * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
2211         * ira.c (get_subreg_tracking_sizes): Likewise.
2212         * ira-conflicts.c (go_through_subreg): Likewise.
2213         * ira-lives.c (process_single_reg_class_operands): Likewise.
2214         * jump.c (rtx_renumbered_equal_p): Likewise.  Handle 'p'.
2215         * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
2216         as a poly_uint64 rather than an unsigned int.
2217         (simplify_gen_subreg_concatn, resolve_simple_move): Treat
2218         subreg offsets as poly_ints.
2219         * lra-constraints.c (operands_match_p): Handle 'p'.
2220         (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
2221         * lra-spills.c (assign_mem_slot): Likewise.
2222         * postreload.c (move2add_valid_value_p): Likewise.
2223         * recog.c (general_operand, indirect_operand): Likewise.
2224         * regcprop.c (copy_value, maybe_mode_change): Likewise.
2225         (copyprop_hardreg_forward_1): Likewise.
2226         * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
2227         (record_subregs_of_mode): Likewise.
2228         * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
2229         * reload.c (operands_match_p): Handle 'p'.
2230         (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
2231         * reload1.c (alter_reg, choose_reload_regs): Likewise.
2232         (compute_reload_subreg_offset): Likewise, and return an poly_int64.
2233         * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
2234         (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
2235         (simplify_const_poly_int_tests<N>::run): Likewise.
2236         (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
2237         a poly_uint64 rather than an unsigned int.
2238         * valtrack.c (debug_lowpart_subreg): Likewise.
2239         * var-tracking.c (var_lowpart): Likewise.
2240         (loc_cmp): Handle 'p'.
2242 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2243             Alan Hayward  <alan.hayward@arm.com>
2244             David Sherwood  <david.sherwood@arm.com>
2246         * ira.c (get_subreg_tracking_sizes): New function.
2247         (init_live_subregs): Take an integer size rather than a register.
2248         (build_insn_chain): Use get_subreg_tracking_sizes.  Update calls
2249         to init_live_subregs.
2251 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2252             Alan Hayward  <alan.hayward@arm.com>
2253             David Sherwood  <david.sherwood@arm.com>
2255         * expr.c (store_constructor_field): Change bitsize from a
2256         unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2257         HOST_WIDE_INT to a poly_int64.
2258         (store_constructor): Change size from a HOST_WIDE_INT to
2259         a poly_int64.
2260         (store_field): Likewise bitsize and bitpos.
2262 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2263             Alan Hayward  <alan.hayward@arm.com>
2264             David Sherwood  <david.sherwood@arm.com>
2266         * expmed.h (store_bit_field): Change bitregion_start and
2267         bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2268         * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2269         (store_bit_field_1, store_integral_bit_field, store_bit_field)
2270         (store_fixed_bit_field, store_split_bit_field): Likewise.
2271         * expr.c (store_constructor_field, store_field): Likewise.
2272         (optimize_bitfield_assignment_op): Likewise.  Make the same change
2273         to bitsize and bitpos.
2274         * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2275         and m_bitregion_end from HOST_WIDE_INT to poly_int64.  Make the
2276         same change in the constructor arguments.
2277         (get_best_mode): Change bitregion_start and bitregion_end from
2278         unsigned HOST_WIDE_INT to poly_uint64.
2279         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2280         Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2281         poly_int64.
2282         (bit_field_mode_iterator::next_mode): Update for new types
2283         of m_bitregion_start and m_bitregion_end.
2284         (get_best_mode): Change bitregion_start and bitregion_end from
2285         unsigned HOST_WIDE_INT to poly_uint64.
2287 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2288             Alan Hayward  <alan.hayward@arm.com>
2289             David Sherwood  <david.sherwood@arm.com>
2291         * rtl.h (simplify_gen_subreg): Add a temporary overload that
2292         accepts poly_uint64 offsets.
2293         * expmed.h (extract_bit_field): Take bitsize and bitnum as
2294         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2295         * expmed.c (lowpart_bit_field_p): Likewise.
2296         (extract_bit_field_as_subreg): New function, split out from...
2297         (extract_bit_field_1): ...here.  Take bitsize and bitnum as
2298         poly_uint64s rather than unsigned HOST_WIDE_INTs.  For vector
2299         extractions, check that BITSIZE matches the size of the extracted
2300         value and that BITNUM is an exact multiple of that size.
2301         If all else fails, try forcing the value into memory if
2302         BITNUM is variable, and adjusting the address so that the
2303         offset is constant.  Split the part that can only handle constant
2304         bitsize and bitnum out into...
2305         (extract_integral_bit_field): ...this new function.
2306         (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2307         rather than unsigned HOST_WIDE_INTs.
2309 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2310             Alan Hayward  <alan.hayward@arm.com>
2311             David Sherwood  <david.sherwood@arm.com>
2313         * expmed.h (store_bit_field): Take bitsize and bitnum as
2314         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2315         * expmed.c (simple_mem_bitfield_p): Likewise.  Add a parameter
2316         that returns the byte size.
2317         (store_bit_field_1): Take bitsize and bitnum as
2318         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Update call
2319         to simple_mem_bitfield_p.  Split the part that can only handle
2320         constant bitsize and bitnum out into...
2321         (store_integral_bit_field): ...this new function.
2322         (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2323         than unsigned HOST_WIDE_INTs.
2324         (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2326 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2327             Alan Hayward  <alan.hayward@arm.com>
2328             David Sherwood  <david.sherwood@arm.com>
2330         * lra-int.h (lra_reg): Change offset from int to poly_int64.
2331         (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2332         to poly_int64.
2333         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2334         update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2335         (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2336         offset as a poly_int64 rather than an int.
2337         * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2338         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2339         * lra-constraints.c (equiv_address_substitution): Track offsets
2340         as poly_int64s.
2341         (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2342         (curr_insn_transform): Handle the new form of sp_offset.
2343         * lra-eliminations.c (lra_elim_table): Change previous_offset
2344         and offset from HOST_WIDE_INT to poly_int64.
2345         (print_elim_table, update_reg_eliminate): Update accordingly.
2346         (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2347         (get_elimination): Update accordingly.
2348         (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2349         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2350         update_sp_offset from a HOST_WIDE_INT to a poly_int64.  Handle
2351         poly_int64 offsets generally.
2352         (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2353         (mark_not_eliminable, init_elimination): Update accordingly.
2354         (remove_reg_equal_offset_note): Return a bool and pass the new
2355         offset back by pointer as a poly_int64.
2356         * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2357         rather than a HOST_WIDE_INT.
2358         (do_remat): Track offsets poly_int64s.
2359         * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2361 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2362             Alan Hayward  <alan.hayward@arm.com>
2363             David Sherwood  <david.sherwood@arm.com>
2365         * rtl.h (mem_attrs): Add a default constructor.  Change size and
2366         offset from HOST_WIDE_INT to poly_int64.
2367         * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2368         (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2369         (widen_memory_access): Take the sizes and offsets as poly_int64s
2370         rather than HOST_WIDE_INTs.
2371         * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2372         (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2373         and ints.
2374         (adjust_offset_for_component_ref): Change the offset from a
2375         HOST_WIDE_INT to a poly_int64.
2376         (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2377         * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2378         * dce.c (find_call_stack_args): Likewise.
2379         * dse.c (record_store): Likewise.
2380         * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2381         * print-rtl.c (rtx_writer::print_rtx): Likewise.
2382         * read-rtl-function.c (test_loading_mem): Likewise.
2383         * rtlanal.c (may_trap_p_1): Likewise.
2384         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2385         * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2386         * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2387         (mem_attrs::mem_attrs): New function.
2388         (set_mem_attributes_minus_bitpos): Change bitpos from a
2389         HOST_WIDE_INT to poly_int64.
2390         (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2391         (clear_mem_offset, clear_mem_size, change_address)
2392         (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2393         initialize mem_attrs.
2394         (set_mem_offset, set_mem_size, adjust_address_1)
2395         (adjust_automodify_address_1, offset_address, widen_memory_access):
2396         Likewise.  Take poly_int64s rather than HOST_WIDE_INT.
2398 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2399             Alan Hayward  <alan.hayward@arm.com>
2400             David Sherwood  <david.sherwood@arm.com>
2402         * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2403         as poly_int64s rather than HOST_WIDE_INTs.  Use a size of -1
2404         rather than 0 to represent an unknown size.  Assert that the size
2405         is known when the mode isn't BLKmode.
2406         (may_trap_p_1): Use -1 for unknown sizes.
2407         (rtx_addr_can_trap_p): Likewise.  Pass BLKmode rather than VOIDmode.
2409 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2410             Alan Hayward  <alan.hayward@arm.com>
2411             David Sherwood  <david.sherwood@arm.com>
2413         * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2414         to poly_int64.  Update commentary for positions_needed.large.
2415         (read_info_type): Change offset and width from HOST_WIDE_INT
2416         to poly_int64.
2417         (set_usage_bits): Likewise.
2418         (canon_address): Return the offset as a poly_int64 rather than
2419         a HOST_WIDE_INT.  Use strip_offset_and_add.
2420         (set_all_positions_unneeded, any_positions_needed_p): Use
2421         positions_needed.large to track stores with non-constant widths.
2422         (all_positions_needed_p): Likewise.  Take the offset and width
2423         as poly_int64s rather than ints.  Assert that rhs is nonnull.
2424         (record_store): Cope with non-constant offsets and widths.
2425         Nullify the rhs of an earlier store if we can't tell which bytes
2426         of it are needed.
2427         (find_shift_sequence): Take the access_size and shift as poly_int64s
2428         rather than ints.
2429         (get_stored_val): Take the read_offset and read_width as poly_int64s
2430         rather than HOST_WIDE_INTs.
2431         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2432         non-constant offsets and widths.
2434 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2435             Alan Hayward  <alan.hayward@arm.com>
2436             David Sherwood  <david.sherwood@arm.com>
2438         * inchash.h (inchash::hash::add_poly_int): New function.
2439         * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2440         Use poly_int64 rather than HOST_WIDE_INT.
2441         (ao_ref::max_size_known_p): New function.
2442         * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2443         rather than HOST_WIDE_INT.
2444         * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2445         to temporaries until its interface is adjusted to match.
2446         (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2447         (aliasing_component_refs_p, decl_refs_may_alias_p)
2448         (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2449         the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2450         (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2451         ao_ref fields.
2452         * alias.c (ao_ref_from_mem): Likewise.
2453         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2454         * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2455         (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2456         (maybe_trim_complex_store, maybe_trim_constructor_store)
2457         (live_bytes_read, dse_classify_store): Likewise.
2458         * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2459         (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2460         (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2461         (vn_reference_lookup_3): Likewise.
2462         * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
2464 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2465             Alan Hayward  <alan.hayward@arm.com>
2466             David Sherwood  <david.sherwood@arm.com>
2468         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2469         (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2470         instead of ranges_overlap_p.
2472 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2473             Alan Hayward  <alan.hayward@arm.com>
2474             David Sherwood  <david.sherwood@arm.com>
2476         * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2477         sizes as poly_int64s rather than HOST_WIDE_INTs.
2479 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2480             Alan Hayward  <alan.hayward@arm.com>
2481             David Sherwood  <david.sherwood@arm.com>
2483         * gimple-fold.h (fold_ctor_reference): Take the offset and size
2484         as poly_uint64 rather than unsigned HOST_WIDE_INT.
2485         * gimple-fold.c (fold_ctor_reference): Likewise.
2487 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2488             Alan Hayward  <alan.hayward@arm.com>
2489             David Sherwood  <david.sherwood@arm.com>
2491         * target.def (dwarf_poly_indeterminate_value): New hook.
2492         * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2493         * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2494         * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2495         * doc/tm.texi: Regenerate.
2496         * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2497         offset as a poly_int64.
2498         * dwarf2out.c (new_reg_loc_descr): Move later in file.  Take the
2499         offset as a poly_int64.
2500         (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2501         Take the offset as a poly_int64.
2502         (build_cfa_loc): Likewise.  Use loc_descr_plus_const.
2503         (frame_pointer_fb_offset): Change to a poly_int64.
2504         (int_loc_descriptor): Take the offset as a poly_int64.  Use
2505         targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2506         (based_loc_descr): Take the offset as a poly_int64.
2507         Use strip_offset_and_add to handle (plus X (const)).
2508         Use new_reg_loc_descr instead of an open-coded version of the
2509         previous implementation.
2510         (mem_loc_descriptor): Handle CONST_POLY_INT.
2511         (compute_frame_pointer_to_fb_displacement): Take the offset as a
2512         poly_int64.  Use strip_offset_and_add to handle (plus X (const)).
2514 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2515             Alan Hayward  <alan.hayward@arm.com>
2516             David Sherwood  <david.sherwood@arm.com>
2518         * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2519         (gen_rtx_REG_offset): Take the offset as a poly_int64.
2520         * inchash.h (inchash::hash::add_poly_hwi): New function.
2521         * gengtype.c (main): Register poly_int64.
2522         * emit-rtl.c (reg_attr_hasher::hash): Use inchash.  Treat the
2523         offset as a poly_int.
2524         (reg_attr_hasher::equal): Use must_eq to compare offsets.
2525         (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2526         offset as a poly_int64.
2527         (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2528         * print-rtl.c (print_poly_int): New function.
2529         (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2530         a poly_int.
2531         * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2532         functions.
2533         (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2534         (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2535         (vt_get_decl_and_offset): Return the offset as a poly_int64.
2536         Enforce track_offset_p for parts of a PARALLEL.
2537         (vt_add_function_parameter): Use const_offset for the final
2538         offset to track.  Use get_tracked_reg_offset for the parts
2539         of a PARALLEL.
2541 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2542             Alan Hayward  <alan.hayward@arm.com>
2543             David Sherwood  <david.sherwood@arm.com>
2545         * target.def (truly_noop_truncation): Take poly_uint64s instead of
2546         unsigned ints.  Change default to hook_bool_puint64_puint64_true.
2547         * doc/tm.texi: Regenerate.
2548         * hooks.h (hook_bool_uint_uint_true): Delete.
2549         (hook_bool_puint64_puint64_true): Declare.
2550         * hooks.c (hook_bool_uint_uint_true): Delete.
2551         (hook_bool_puint64_puint64_true): New function.
2552         * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2553         instead of unsigned ints.
2554         * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2555         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
2557 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2558             Alan Hayward  <alan.hayward@arm.com>
2559             David Sherwood  <david.sherwood@arm.com>
2561         * optabs.h (expand_operand): Add an int_value field.
2562         (create_expand_operand): Add an int_value parameter and use it
2563         to initialize the new expand_operand field.
2564         (create_integer_operand): Replace with a declaration of a function
2565         that accepts poly_int64s.  Move the implementation to...
2566         * optabs.c (create_integer_operand): ...here.
2567         (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2568         the mode preserves the value of int_value, instead of calling
2569         const_int_operand on the rtx.  Use gen_int_mode to generate
2570         the new rtx.
2572 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2573             Alan Hayward  <alan.hayward@arm.com>
2574             David Sherwood  <david.sherwood@arm.com>
2576         * dumpfile.h (dump_dec): Declare.
2577         * dumpfile.c (dump_dec): New function.
2578         * pretty-print.h (pp_wide_integer): Turn into a function and
2579         declare a poly_int version.
2580         * pretty-print.c (pp_wide_integer): New function for poly_ints.
2582 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2583             Alan Hayward  <alan.hayward@arm.com>
2584             David Sherwood  <david.sherwood@arm.com>
2586         * doc/generic.texi (POLY_INT_CST): Document.
2587         * tree.def (POLY_INT_CST): New tree code.
2588         * treestruct.def (TS_POLY_INT_CST): New tree layout.
2589         * tree-core.h (tree_poly_int_cst): New struct.
2590         (tree_node): Add a poly_int_cst field.
2591         * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2592         (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2593         instead of a wide_int_ref.
2594         (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2595         of a HOST_WIDE_INT.
2596         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2597         instead of an unsigned HOST_WIDE_INT.
2598         (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2599         (ptrdiff_tree_p): Declare.
2600         (tree_to_poly_int64, tree_to_poly_uint64): Likewise.  Provide
2601         extern inline implementations if the target doesn't use POLY_INT_CST.
2602         (poly_int_tree_p): New function.
2603         (wi::unextended_tree): New class.
2604         (wi::int_traits <unextended_tree>): New override.
2605         (wi::extended_tree): Add a default constructor.
2606         (wi::extended_tree::get_tree): New function.
2607         (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2608         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2609         (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2610         (wi::tree_to_poly_wide_ref): New typedefs.
2611         (wi::ints_for): Provide overloads for extended_tree and
2612         unextended_tree.
2613         (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2614         (wi::to_wide): New functions.
2615         (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2616         * tree.c (poly_int_cst_hasher): New struct.
2617         (poly_int_cst_hash_table): New variable.
2618         (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2619         (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2620         POLY_INT_CST.
2621         (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2622         (init_ttree): Initialize poly_int_cst_hash_table.
2623         (build_int_cst, build_int_cst_type, build_invariant_address): Take
2624         a poly_int64 instead of a HOST_WIDE_INT.
2625         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2626         instead of an unsigned HOST_WIDE_INT.
2627         (wide_int_to_tree): Rename to...
2628         (wide_int_to_tree_1): ...this.
2629         (build_new_poly_int_cst, build_poly_int_cst): New functions.
2630         (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2631         (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2632         (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2633         (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2634         * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2635         TS_POLY_INT_CST.
2636         * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2637         (streamer_read_tree_body): Likewise.
2638         * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2639         (streamer_write_tree_body): Likewise.
2640         * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2641         * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2642         * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2643         * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2644         * gimple-expr.h (is_gimple_constant): Likewise.
2645         * gimplify.c (maybe_with_size_expr): Likewise.
2646         * print-tree.c (print_node): Likewise.
2647         * tree-data-ref.c (data_ref_compare_tree): Likewise.
2648         * tree-pretty-print.c (dump_generic_node): Likewise.
2649         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2650         * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2651         * tree-vrp.c (compare_values_warnv): Likewise.
2652         * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2653         (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2654         (force_expr_to_var_cost): Likewise.
2655         * tree-ssa-loop.c (for_each_index): Likewise.
2656         * fold-const.h (build_invariant_address, size_int_kind): Take a
2657         poly_int64 instead of a HOST_WIDE_INT.
2658         * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2659         (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2660         POLY_INT_CST.
2661         (size_binop_loc): Likewise.  Allow int_const_binop_1 to fail.
2662         (int_const_binop_2): New function, split out from...
2663         (int_const_binop_1): ...here.  Handle POLY_INT_CST.
2664         (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2665         * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2666         * gimple-ssa-strength-reduction.c (slsr_process_add)
2667         (slsr_process_mul): Check for INTEGER_CSTs before using them
2668         as candidates.
2669         * stor-layout.c (bits_from_bytes): New function.
2670         (bit_from_pos): Use it.
2671         (layout_type): Likewise.  For vectors, multiply the TYPE_SIZE_UNIT
2672         by BITS_PER_UNIT to get the TYPE_SIZE.
2673         * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2674         MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2676 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2677             Alan Hayward  <alan.hayward@arm.com>
2678             David Sherwood  <david.sherwood@arm.com>
2680         * doc/rtl.texi (const_poly_int): Document.  Also document the
2681         rtl sharing behavior.
2682         * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2683         * rtl.h (const_poly_int_def): New struct.
2684         (rtx_def::u): Add a cpi field.
2685         (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2686         (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2687         (wi::rtx_to_poly_wide_ref): New typedef
2688         (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2689         (poly_int_rtx_p): New functions.
2690         (trunc_int_for_mode): Declare a poly_int64 version.
2691         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2692         (immed_wide_int_const): Take a poly_wide_int_ref rather than
2693         a wide_int_ref.
2694         (strip_offset): Declare.
2695         (strip_offset_and_add): New function.
2696         * rtl.def (CONST_POLY_INT): New rtx code.
2697         * rtl.c (rtx_size): Handle CONST_POLY_INT.
2698         (shared_const_p): Use poly_int_rtx_p.
2699         * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2700         HOST_WIDE_INT.
2701         (gen_int_shift_amount): Likewise.
2702         * emit-rtl.c (const_poly_int_hasher): New class.
2703         (const_poly_int_htab): New variable.
2704         (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2705         (const_poly_int_hasher::hash): New function.
2706         (const_poly_int_hasher::equal): Likewise.
2707         (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2708         (immed_wide_int_const): Rename to...
2709         (immed_wide_int_const_1): ...this and make static.
2710         (immed_wide_int_const): New function, taking a poly_wide_int_ref
2711         instead of a wide_int_ref.
2712         (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2713         (gen_lowpart_common): Handle CONST_POLY_INT.
2714         * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2715         * cselib.c (cselib_hash_rtx): Likewise.
2716         * dwarf2out.c (const_ok_for_output_1): Likewise.
2717         * expr.c (convert_modes): Likewise.
2718         * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2719         * rtlhash.c (add_rtx): Likewise.
2720         * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2721         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2722         Handle existing CONST_POLY_INT rtxes.
2723         * expmed.h (expand_shift): Take a poly_int64 instead of a
2724         HOST_WIDE_INT.
2725         * expmed.c (expand_shift): Likewise.
2726         * rtlanal.c (strip_offset): New function.
2727         (commutative_operand_precedence): Give CONST_POLY_INT the same
2728         precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2729         and CONST_INT.
2730         * rtl-tests.c (const_poly_int_tests): New struct.
2731         (rtl_tests_c_tests): Use it.
2732         * simplify-rtx.c (simplify_const_unary_operation): Handle
2733         CONST_POLY_INT.
2734         (simplify_const_binary_operation): Likewise.
2735         (simplify_binary_operation_1): Fold additions of symbolic constants
2736         and CONST_POLY_INTs.
2737         (simplify_subreg): Handle extensions and truncations of
2738         CONST_POLY_INTs.
2739         (simplify_const_poly_int_tests): New struct.
2740         (simplify_rtx_c_tests): Use it.
2741         * wide-int.h (storage_ref): Add default constructor.
2742         (wide_int_ref_storage): Likewise.
2743         (trailing_wide_ints): Use GTY((user)).
2744         (trailing_wide_ints::operator[]): Add a const version.
2745         (trailing_wide_ints::get_precision): New function.
2746         (trailing_wide_ints::extra_size): Likewise.
2748 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2749             Alan Hayward  <alan.hayward@arm.com>
2750             David Sherwood  <david.sherwood@arm.com>
2752         * emit-rtl.h (gen_int_shift_amount): Declare.
2753         * emit-rtl.c (gen_int_shift_amount): New function.
2754         * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2755         instead of GEN_INT.
2756         * calls.c (shift_return_value): Likewise.
2757         * cse.c (fold_rtx): Likewise.
2758         * dse.c (find_shift_sequence): Likewise.
2759         * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2760         (expand_shift, expand_smod_pow2): Likewise.
2761         * lower-subreg.c (shift_cost): Likewise.
2762         * optabs.c (expand_superword_shift, expand_doubleword_mult)
2763         (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2764         (expand_vec_perm_var): Likewise.
2765         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2766         (simplify_binary_operation_1): Likewise.
2767         * combine.c (try_combine, find_split_point, force_int_to_mode)
2768         (simplify_shift_const_1, simplify_shift_const): Likewise.
2769         (change_zero_ext): Likewise.  Use simplify_gen_binary.
2771 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2773         * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2775 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2777         * doc/invoke.texi (ARM Options): Document accepted extension options
2778         for -march=armv8.3-a.
2780 2017-12-20  Richard Earnshaw  <rearnsha@arm.com>
2782         PR target/83105
2783         * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2784         or --with-float=softfp, set the default CPU to arm10e.
2786 2017-12-20  Eric Botcazou  <ebotcazou@adacore.com>
2788         * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2789         * config/visium/predicates.md (const_shift_operand): Likewise.
2790         * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2791         (visium_legitimize_reload_address): Likewise.
2793 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2795         PR target/82975
2796         * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2798 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
2800         PR c++/83490
2801         * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2803 2017-12-20  Julia Koval  <julia.koval@intel.com>
2805         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2806         OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2807         (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2808         * config.gcc: Include vpclmulqdqintrin.h.
2809         * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2810         * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2811         -mvpclmulqdq.
2812         * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2813         __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2814         * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2815         * config/i386/i386.c (isa2_opts): Add -mcx16.
2816         (isa_opts): Add -mpclmulqdq, remove -mcx16.
2817         (ix86_option_override_internal): Move mcx16 to flags2.
2818         (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2819         (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2820         * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2821         * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2822         * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2823         * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2824         * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2825         _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2826         * doc/invoke.texi: Add -mvpclmulqdq.
2828 2017-12-20  Tom de Vries  <tom@codesourcery.com>
2830         PR middle-end/83423
2831         * config/i386/i386.c (ix86_static_chain): Move
2832         DECL_STATIC_CHAIN test ...
2833         * calls.c (rtx_for_static_chain): ... here.  New function.
2834         * calls.h (rtx_for_static_chain): Declare.
2835         * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2836         instead of targetm.calls.static_chain.
2837         * df-scan.c (df_get_entry_block_def_set): Same.
2839 2017-12-19  Tom de Vries  <tom@codesourcery.com>
2841         PR tree-optimization/83493
2842         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2843         and lb.
2845 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2847         * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2848         inform with hardcoded english plural handling.
2850 2017-12-18  Jeff Law  <law@redhat.com>
2852         PR tree-optimization/83477
2853         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2854         a non-virtual PHI, always push a new range.
2856 2017-12-19  Martin Sebor  <msebor@redhat.com>
2858         PR middle-end/77608
2859         * builtins.c (compute_objsize): Handle non-constant offsets.
2861 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2863         PR tree-optimization/83444
2864         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2865         character load case, if get_stridx on MEM_REF's operand doesn't
2866         look usable, retry with get_addr_stridx.
2868 2017-12-19  Alexandre Oliva  <aoliva@redhat.com>
2870         PR debug/83422
2871         * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2872         (variable_tracking_main_1): Keep markers even when VTA fails.
2874         PR bootstrap/83396
2875         * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2876         even if there are markers before it.
2877         * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2879 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2881         * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2882         typical order conditions.
2883         * sel-sched.c (extract_new_fences_from): Likewise.
2884         * config/visium/constraints.md (J, K, L): Likewise.
2885         * config/visium/predicates.md (const_shift_operand): Likewise.
2886         * config/visium/visium.c (visium_legitimize_address,
2887         visium_legitimize_reload_address): Likewise.
2888         * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2889         * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2890         * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2891         * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2892         SET_DUMP_DETAIL): Likewise.
2893         * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2894         * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2895         avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2896         avr_set_core_architecture, avr_set_current_function,
2897         avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2898         output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2899         avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2900         avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2901         avr_map_decompose, avr_fold_builtin): Likewise.
2902         * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2903         * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2904         * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2905         * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2906         * config/m32c/m32c.c (m32c_conditional_register_usage,
2907         m32c_address_cost): Likewise.
2908         * config/m32c/predicates.md (shiftcount_operand,
2909         longshiftcount_operand): Likewise.
2910         * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2911         * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2912         can_use_cdx_ldstw): Likewise.
2913         * config/nios2/nios2.h (CDX_REG_P): Likewise.
2914         * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2915         Likewise.
2916         * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2917         * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
2918         * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
2919         * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
2920         * config/vax/vax.c (adjacent_operands_p): Likewise.
2921         * config/ft32/constraints.md (L, b, KA): Likewise.
2922         * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
2923         Likewise.
2924         * cfgexpand.c (expand_stack_alignment): Likewise.
2925         * gcse.c (insert_expr_in_table): Likewise.
2926         * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
2927         * cgraphunit.c (cgraph_node::expand): Likewise.
2928         * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
2929         * emit-rtl.c (add_insn): Likewise.
2930         * input.c (dump_location_info): Likewise.
2931         * passes.c (NEXT_PASS): Likewise.
2932         * read-rtl-function.c (parse_note_insn_name,
2933         function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
2934         Likewise.
2935         * sched-rgn.c (sched_rgn_init): Likewise.
2936         * diagnostic-show-locus.c (layout::show_ruler): Likewise.
2937         * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
2938         if_then_else_cond, simplify_shift_const_1, simplify_comparison):
2939         Likewise.
2940         * explow.c (eliminate_constant_term): Likewise.
2941         * final.c (leaf_renumber_regs_insn): Likewise.
2942         * cfgrtl.c (print_rtl_with_bb): Likewise.
2943         * genhooks.c (emit_init_macros): Likewise.
2944         * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
2945         * tree-data-ref.c (conflict_fn): Likewise.
2946         * selftest.c (assert_streq): Likewise.
2947         * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
2948         * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
2949         fold_binary_loc, multiple_of_p): Likewise.
2950         * reload.c (push_reload, find_equiv_reg): Likewise.
2951         * et-forest.c (et_nca, et_below): Likewise.
2952         * dbxout.c (dbxout_symbol_location): Likewise.
2953         * reorg.c (relax_delay_slots): Likewise.
2954         * dojump.c (do_compare_rtx_and_jump): Likewise.
2955         * gengtype-parse.c (type): Likewise.
2956         * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
2957         simplify_const_relational_operation): Likewise.
2958         * reload1.c (do_output_reload): Likewise.
2959         * dumpfile.c (get_dump_file_info_by_switch): Likewise.
2960         * gengtype.c (type_for_name): Likewise.
2961         * gimple-ssa-sprintf.c (format_directive): Likewise.
2963 2017-12-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2965         PR target/82975
2966         * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
2967         accessing it.  Adjust comment.
2969 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2971         PR middle-end/81914
2972         * predict.c (zero_one_minusone): New function.
2973         (apply_return_prediction): Avoid return prediction for functions
2974         returning only -1, 0 and 1 values, unless they only return -1 and 0
2975         or 0 and 1.
2977 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2979         * config/arc/arc.c (legitimate_scaled_address_p): Clean
2980         fall-through warning.
2981         (arc_compute_frame_size): Remove unused variables.
2982         (arc_print_operand): Fix fprintif format.
2983         (arc_can_follow_jump): Clean fall-through warning.
2985 2017-12-19  Marek Polacek  <polacek@redhat.com>
2987         PR c++/83489
2988         * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
2989         on an error node.
2991 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2993         * config/arc/arc.c (overriderregs): New variable.
2994         (arc_override_options): Track fixed/call saved/call options.
2995         (arc_conditional_register_usage): Check against overrideregs
2996         variable whenever we change register properties.
2998 2017-12-19  Nathan Sidwell  <nathan@acm.org>
3000         * opts.c (finish_options): Don't prefix dump_base_name if it
3001         already contains directories.
3003 2017-12-19  Martin Liska  <mliska@suse.cz>
3005         PR rtl-optimization/82675
3006         * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
3007         more element in sbitmap.
3009 2017-12-19  Paolo Carlini  <paolo.carlini@oracle.com>
3011         * gimplify.c (gimplify_expr): Use error_operand_p.
3013 2017-12-19  Sebastian Huber  <sebastian.huber@embedded-brains.de>
3015         PR target/83387
3016         * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
3017         multilib.
3019 2017-12-19  Daniel Cederman  <cederman@gaisler.com>
3021         * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
3022         the jump is to a label.
3024 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
3026         PR tree-optimization/83444
3027         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
3028         character loads.
3030         PR ipa/82801
3031         PR ipa/83346
3032         * ipa-inline.c (flatten_remove_node_hook): New function.
3033         (ipa_inline): Keep only nodes with flatten attribute at the end of
3034         the array in the order from ipa_reverse_postorder, only walk that
3035         portion of array for flattening, if there is more than one such
3036         node, temporarily register a removal hook and ignore removed nodes.
3038         PR tree-optimization/80631
3039         * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
3040         induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
3041         IFN_REDUC_MAX or IFN_REDUC_MIN.
3043 2017-12-18  Jeff Law  <law@redhat.com>
3045         * tree-ssa-dom.c (record_equivalences_from_phis): Do not
3046         record symbolic equivalences from backedges in the CFG.
3048         Revert
3049         2017-11-19  Jeff Law  <law@redhat.com>
3051         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
3052         of degenerates resulting from ignoring an edge.
3054 2017-12-18  Martin Sebor  <msebor@redhat.com>
3056         PR middle-end/83373
3057         PR tree-optimization/78450
3058         * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
3059         (handle_builtin_strlen): Call it.
3061 2017-12-18  Segher Boessenkool  <segher@kernel.crashing.org>
3063         PR rtl-optimization/83424
3064         * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
3066 2017-12-18  Marek Polacek  <polacek@redhat.com>
3068         PR middle-end/83463
3069         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
3070         Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
3071         values.
3073 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3075         * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
3076         (umaddsidi4, umaddsidi_split): Likewise.
3078 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3080         * config/arc/arc.c (arc_legitimate_constant_p): Always check all
3081         constants.
3083 2017-12-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
3085         PR target/83420
3086         * config/s390/s390.c (s390_option_override): Avoid strncpy.
3088 2017-12-18  Richard Biener  <rguenther@suse.de>
3090         PR tree-optimization/81877
3091         * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
3092         (outermost_indep_loop): Adjust.
3093         (ref_indep_loop_p_1): Likewise.  Remove safelen handling again.
3094         (can_sm_ref_p): Adjust.
3096 2017-12-18  Richard Biener  <rguenther@suse.de>
3098         PR middle-end/77291
3099         * tree.c (array_at_struct_end_p): Return true if the underlying
3100         object has space for at least one element in excess of what
3101         the array domain specifies.
3103 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3105         * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
3106         example to avoid overfull hbox.
3107         * doc/invoke.texi (Option Summary): Add missing @gol.
3108         (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
3109         overfull hbox.
3111 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3112             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3114         * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
3115         to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
3116         (AArch64 Options): Document -mverbose-cost-dump.
3117         (ARM Options): Likewise, plus -mflip-thumb.
3119 2017-12-17  Martin Sebor  <msebor@redhat.com>
3121         PR bootstrap/83446
3122         * gimple-ssa-warn-restrict.c
3123         (builtin_memref::offset_out_of_bounds): Correct the handling of
3124         anti-ranges.
3126 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3128         * doc/invoke.texi (ARC Options): Add missing -mlra entry.
3130 2017-12-17  John David Anglin  <danglin@gcc.gnu.org>
3132         * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
3134 2017-12-17  Markus Trippelsdorf  <markus@trippelsdorf.de>
3136         * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
3137         latencies.
3139 2017-12-16  Sandra Loosemore  <sandra@codesourcery.com>
3141         * doc/invoke.texi: Fix some typos.
3143 2017-12-16  Martin Sebor  <msebor@redhat.com>
3145         PR tree-optimization/78918
3146         * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
3147         * builtins.c (check_sizes): Rename...
3148         (check_access): ...to this.  Rename function arguments for clarity.
3149         (check_memop_sizes): Adjust names.
3150         (expand_builtin_memchr, expand_builtin_memcpy): Same.
3151         (expand_builtin_memmove, expand_builtin_mempcpy): Same.
3152         (expand_builtin_strcat, expand_builtin_stpncpy): Same.
3153         (check_strncat_sizes, expand_builtin_strncat): Same.
3154         (expand_builtin_strncpy, expand_builtin_memset): Same.
3155         (expand_builtin_bzero, expand_builtin_memcmp): Same.
3156         (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
3157         (maybe_emit_sprintf_chk_warning): Same.
3158         (expand_builtin_strcpy): Adjust.
3159         (expand_builtin_stpcpy): Same.
3160         (expand_builtin_with_bounds): Detect out-of-bounds accesses
3161         in pointer-checking forms of memcpy, memmove, and mempcpy.
3162         (gcall_to_tree_minimal, max_object_size): Define new functions.
3163         * builtins.h (max_object_size): Declare.
3164         * calls.c (alloc_max_size): Call max_object_size instead of
3165         hardcoding ssizetype limit.
3166         (get_size_range): Handle new argument.
3167         * calls.h (get_size_range): Add a new argument.
3168         * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
3169         * doc/invoke.texi (-Wrestrict): Adjust, add example.
3170         * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
3171         operations.
3172         (gimple_fold_builtin_memory_chk): Same.
3173         (gimple_fold_builtin_stxcpy_chk): New function.
3174         * gimple-ssa-warn-restrict.c: New source.
3175         * gimple-ssa-warn-restrict.h: New header.
3176         * gimple.c (gimple_build_call_from_tree): Propagate location.
3177         * passes.def (pass_warn_restrict): Add new pass.
3178         * tree-pass.h (make_pass_warn_restrict): Declare.
3179         * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
3180         operations.
3181         (handle_builtin_strcat): Same.
3182         (strlen_optimize_stmt): Rename...
3183         (strlen_check_and_optimize_stmt): ...to this.  Handle strncat,
3184         stpncpy, strncpy, and their checking forms.
3186 2017-12-16  Jan Hubicka  <hubicka@ucw.cz>
3188         PR rtl-optimization/82849
3189         * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
3190         and get_max_loop_iterations_int.
3192 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3193             Alan Hayward  <alan.hayward@arm.com>
3194             David Sherwood  <david.sherwood@arm.com>
3196         * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
3197         (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
3198         size as a poly_uint64.
3199         (mode_for_vector, mode_for_int_vector): Take the number of vector
3200         elements as a poly_uint64.
3201         * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
3202         size as a poly_uint64.
3203         (mode_for_vector, mode_for_int_vector): Take the number of vector
3204         elements as a poly_uint64.
3206 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3207             Alan Hayward  <alan.hayward@arm.com>
3208             David Sherwood  <david.sherwood@arm.com>
3210         * machmode.h (MACRO_MODE): New macro.
3211         * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
3212         * targhooks.c (default_libcall_value, default_secondary_reload)
3213         (default_memory_move_cost, default_register_move_cost)
3214         (default_class_max_nregs): Likewise.
3216 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3217             Alan Hayward  <alan.hayward@arm.com>
3218             David Sherwood  <david.sherwood@arm.com>
3220         * doc/sourcebuild.texi: Document IN_TARGET_CODE.
3221         * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
3222         target C file.
3223         * genautomata.c (main): Likewise.
3224         * genconditions.c (write_header): Likewise.
3225         * genemit.c (main): Likewise.
3226         * genextract.c (print_header): Likewise.
3227         * genopinit.c (main): Likewise.
3228         * genoutput.c (output_prologue): Likewise.
3229         * genpeep.c (main): Likewise.
3230         * genpreds.c (write_insn_preds_c): Likewise.
3231         * genrecog.c (writer_header): Likewise.
3232         * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
3233         * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
3234         * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
3235         * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
3236         * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
3237         * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
3238         * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
3239         * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
3240         * config/arc/arc.c (IN_TARGET_CODE): Likewise.
3241         * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
3242         * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
3243         * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3244         * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3245         * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3246         * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3247         * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3248         * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3249         * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3250         * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3251         * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3252         * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3253         * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3254         * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3255         * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3256         * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3257         * config/darwin.c (IN_TARGET_CODE): Likewise.
3258         * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3259         * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3260         * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3261         * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3262         * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3263         * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3264         * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3265         * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3266         * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3267         * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3268         * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3269         * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3270         * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3271         * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3272         * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3273         * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3274         * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3275         * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3276         * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3277         * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3278         * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3279         * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3280         * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3281         * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3282         * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3283         * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3284         * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3285         * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3286         * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3287         * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3288         * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3289         * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3290         * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3291         * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3292         * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3293         * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3294         * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3295         * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3296         * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3297         * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3298         * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3299         * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3300         * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3301         * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3302         * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3303         * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3304         * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3305         * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3306         * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3307         * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3308         * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3309         * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3310         * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3311         * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3312         * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3313         * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3314         * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3315         * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3316         * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3317         * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3318         * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3319         * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3320         * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3321         * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3322         * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3323         * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3324         * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3325         * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3326         * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3327         * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3328         * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3329         * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3330         * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3331         * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3332         * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3333         * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3334         * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3335         * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3336         * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3337         * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3338         * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3339         * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3340         * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3341         * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3342         * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3343         * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3344         * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3345         * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3346         * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3347         * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3348         * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3349         * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3350         * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3351         * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3352         * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3353         * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3354         * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3355         * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3356         * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3357         * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3358         * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3359         * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3360         * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3361         * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3362         * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3363         * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3365 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3366             Alan Hayward  <alan.hayward@arm.com>
3367             David Sherwood  <david.sherwood@arm.com>
3369         * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3370         checks for MEM_REF.
3372 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3373             Alan Hayward  <alan.hayward@arm.com>
3374             David Sherwood  <david.sherwood@arm.com>
3376         * doc/generic.texi (VEC_SERIES_EXPR): Document.
3377         * doc/md.texi (vec_series@var{m}): Document.
3378         * tree.def (VEC_SERIES_EXPR): New tree code.
3379         * tree.h (build_vec_series): Declare.
3380         * tree.c (build_vec_series): New function.
3381         * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3382         * tree-pretty-print.c (dump_generic_node): Likewise.
3383         * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3384         * tree-inline.c (estimate_operator_cost): Likewise.
3385         * expr.c (expand_expr_real_2): Likewise.
3386         * optabs-tree.c (optab_for_tree_code): Likewise.
3387         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3388         * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3389         * expmed.c (make_tree): Handle VEC_SERIES.
3390         * optabs.def (vec_series_optab): New optab.
3391         * optabs.h (expand_vec_series_expr): Declare.
3392         * optabs.c (expand_vec_series_expr): New function.
3393         * tree-vect-generic.c (expand_vector_operations_1): Check that
3394         the operands also have vector type.
3396 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3397             Alan Hayward  <alan.hawyard@arm.com>
3398             David Sherwood  <david.sherwood@arm.com>
3400         * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3401         (VEC_COND_EXPR): Add missing @tindex.
3402         * doc/md.texi (vec_duplicate@var{m}): Document.
3403         * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3404         * tree.c (build_vector_from_val): Add stubbed-out handling of
3405         variable-length vectors, using VEC_DUPLICATE_EXPR.
3406         (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3407         * cfgexpand.c (expand_debug_expr): Likewise.
3408         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3409         * tree-inline.c (estimate_operator_cost): Likewise.
3410         * tree-pretty-print.c (dump_generic_node): Likewise.
3411         * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3412         * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3413         (test_vec_duplicate_folding): New function.
3414         (fold_const_c_tests): Call it.
3415         * optabs.def (vec_duplicate_optab): New optab.
3416         * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3417         * optabs.h (expand_vector_broadcast): Declare.
3418         * optabs.c (expand_vector_broadcast): Make non-static.  Try using
3419         vec_duplicate_optab.
3420         * expr.c (store_constructor): Try using vec_duplicate_optab for
3421         uniform vectors.
3422         (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3424 2017-12-15  Markus Trippelsdorf  <markus@trippelsdorf.de>
3426         PR target/83358
3427         * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3428         div/mod latencies a bit.
3430 2017-12-15  Jeff Law  <law@redhat.com>
3432         PR tree-optimization/36550
3433         * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3434         (mark_threaded_blocks): Rewrite code to avoid block copying when
3435         optimizing for size.  Don't pessimize blocks which will be
3436         copied, but all the statements will be dead.
3438 2017-12-15  Alexandre Oliva  <aoliva@redhat.com>
3440         PR tree-optimization/81165
3441         * tree-ssa-threadupdate.c (uses_in_bb): New.
3442         (estimate_threading_killed_stmts): New.
3443         * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3444         * tree-ssa-threadedge.c
3445         (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3446         when its hit.
3448 2017-12-15  Jeff Law  <law@redhat.com>
3450         PR tree-optimization/83410
3451         * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3452         threads when parallelizing loops.
3454 2017-12-15  Jakub Jelinek  <jakub@redhat.com>
3456         * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3457         handler fields.
3458         * config/alpha/alpha.c (vms_attribute_table): Swap
3459         affects_type_identity and handler fields, adjust comments.
3460         * config/mips/mips.c (mips_attribute_table): Likewise.
3461         * config/visium/visium.c (visium_attribute_table): Likewise.
3462         * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3463         * config/microblaze/microblaze.c (microblaze_attribute_table):
3464         Likewise.
3465         * config/spu/spu.c (spu_attribute_table): Likewise.
3466         * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3467         * config/arc/arc.c (arc_attribute_table): Likewise.
3468         * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3469         * config/v850/v850.c (v850_handle_interrupt_attribute,
3470         v850_handle_data_area_attribute): Formatting fixes.
3471         (v850_attribute_table): Swap affects_type_identity and handler
3472         fields, adjust comments.
3473         * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3474         * config/arm/arm.c (arm_attribute_table): Likewise.
3475         * config/avr/avr.c (avr_attribute_table): Likewise.
3476         * config/s390/s390.c (s390_attribute_table): Likewise.
3477         * config/sh/sh.c (sh_attribute_table): Likewise.
3478         * config/i386/i386.c (ix86_handle_cconv_attribute,
3479         ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3480         (ix86_attribute_table): Swap affects_type_identity and handler
3481         fields, adjust comments.
3482         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3483         * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3484         * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3485         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3486         * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3487         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3488         * config/rx/rx.c (rx_attribute_table): Likewise.
3489         * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3490         * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3491         * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3492         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3493         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3494         * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3495         * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3496         * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3497         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3498         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3499         * doc/plugins.texi (user_attr): Likewise.  Add NULL for
3500         exclude.
3501         * attribs.c (empty_attribute_table): Swap affects_type_identity and
3502         handler fields.
3503         (register_scoped_attributes, decl_attributes): Formatting fixes.
3505         PR tree-optimization/83269
3506         * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3507         subtraction in arg0's type if type is signed and arg0 is unsigned.
3508         Formatting fix.
3510         PR sanitizer/81281
3511         * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3512         convert? on @0 instead of convert.  Check type of @1, not @0.
3513         ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3514         convert? on @0 instead of convert.  Check type of @1, not @0.
3515         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3516         only optimize if either both @1 and @2 types are narrower
3517         precision, or both are wider or equal precision, and in the former
3518         case only if both have undefined overflow.
3520 2017-12-15  Richard Biener  <rguenther@suse.de>
3522         PR lto/83388
3523         * internal-fn.def (IFN_NOP): Add.
3524         * internal-fn.c (expand_NOP): Do nothing.
3525         * lto-streamer-in.c (input_function): Instead of removing
3526         sanitizer calls replace them with IFN_NOP calls.
3528 2017-12-15  Richard Sandiford  <richard.sandiford@linaro.org>
3529             Alan Hayward  <alan.hayward@arm.com>
3530             David Sherwood  <david.sherwood@arm.com>
3532         * dse.c (store_info, read_info_type): Replace begin and end with
3533         offset and width.
3534         (print_range): New function.
3535         (set_all_positions_unneeded, any_positions_needed_p)
3536         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3537         accordingly.
3538         (record_store): Likewise.  Optimize the case in which all positions
3539         are unneeded.
3540         (get_stored_val): Replace read_begin and read_end with read_offset
3541         and read_width.
3542         (replace_read): Update call accordingly.
3544 2017-12-15  Bin Cheng  <bin.cheng@arm.com>
3546         * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3547         (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3548         (loop_cand::loop_cand): Initialize above members.
3549         (loop_cand::supported_operations): Delete.
3550         (loop_cand::can_interchange_p): Inline above function.
3551         (loop_cand::classify_simple_reduction): Record number of constant
3552         initialized simple reductions.
3553         (should_interchange_loops): New parameters.  Check stmt cost of loops
3554         to be interchange.
3555         (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3556         Update call to should_interchange_loops.
3557         (should_interchange_loop_nest): Update call to
3558         should_interchange_loops.
3560 2017-12-15  Eric Botcazou  <ebotcazou@adacore.com>
3562         PR target/66488
3563         * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3564         * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3565         * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3567 2017-12-15  Julia Koval  <julia.koval@intel.com>
3569         * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3570         __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3571         New.
3572         * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3573         * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3574         _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3576 2017-12-15  Julia Koval  <julia.koval@intel.com>
3578         * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3579         __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3580         * config/i386/sse.md (vaesenc_<mode>): New pattern.
3581         * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3582         _mm_aesenc_epi128): New intrinsics.
3584 2017-12-15  Julia Koval  <julia.koval@intel.com>
3586         * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3587         __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3588         New.
3589         * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3590         * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3591         _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3593 2017-12-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3595         * gimple-ssa-strength-reduction.c (analyze_increments):
3596         Distinguish replacement costs for constant strides from those for
3597         unknown strides.
3599 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3601         * var-tracking.c (variable_tracking_main_1): Formatting fix.
3603 2017-12-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3605         * doc/invoke.texi: Document -Wcast-function-type.
3606         * recog.h (stored_funcptr): Change signature.
3607         * tree-dump.c (dump_node): Avoid warning.
3608         * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3610 2017-12-14  Qing Zhao  <qing.zhao@oracle.com>
3612         PR middle_end/79538
3613         * gimple-fold.c (get_range_strlen): Add the handling of non-member 
3614         array.
3616 2017-12-14  David Malcolm  <dmalcolm@redhat.com>
3618         PR tree-optimization/83312
3619         * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3620         * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3621         "val" param, and to cope with arbitrary basic blocks.
3622         (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3623         handle NULL_TREE for "val", dropping "bb" param.
3624         (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3625         drop "bb" param.  Handle NULL_TREE for "val".
3626         (find_case_label_for_value): Make "switch_stmt" param const.
3627         * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3628         of dom_walker.
3629         (vrp_prop::check_all_array_refs): Reimplement as...
3630         (check_array_bounds_dom_walker::before_dom_children): ...this new
3631         vfunc.  Replace linear search through BB block list, excluding
3632         those with non-executable in-edges via dominator walk.
3634 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3636         * config/arm/arm.opt (mverbose-cost-dump): New option.
3637         * config/arm/arm.c (arm_rtx_costs): Use it.
3639 2017-12-14  Andreas Schwab  <schwab@linux-m68k.org>
3641         PR bootstrap/83396
3642         * reload1.c (emit_input_reload_insns): Skip debug markers.
3644 2017-12-14  Alexandre Oliva  <aoliva@redhat.com>
3646         * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3647         for its nop_endbr.
3649         PR bootstrap/83396
3650         * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3651         * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3652         * config/sh/sh.c: Skip debug insns besides notes.
3653         * config/sh/sh.md: Likewise.
3654         * config/sh/sh_treg_combine.cc: Likewise.
3655         * config/sh/sync.md: Likewise.
3657 2017-12-14  Tom de Vries  <tom@codesourcery.com>
3659         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3660         Add item for weak_undefined.
3662 2017-12-14  Richard Biener  <rguenther@suse.de>
3664         PR tree-optimization/67842
3665         * sese.h (bb_in_region): Remove #if 0'ed code.
3667 2017-12-14  Richard Biener  <rguenther@suse.de>
3669         PR tree-optimization/83326
3670         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3671         may_be_zero parameter and handle it by not marking the first
3672         peeled copy as not exiting the loop.
3673         (try_peel_loop): Likewise.
3674         (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3675         to handle the case of constant or zero iterations and perform
3676         loop header copying on-the-fly.
3678 2017-12-14  Richard Biener  <rguenther@suse.de>
3680         PR tree-optimization/83418
3681         * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3682         Instead of asserting we don't get unfolded comparisons deal with them.
3684 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3686         PR bootstrap/83396
3687         * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3688         basic blocks.  Assert debug bind insns don't appear outside of bbs,
3689         don't reset them.  Assert insns without BLOCK_FOR_INSN are outside of
3690         bb.  Simplify.
3692         PR tree-optimization/83198
3693         * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3694         dir.modifier, regardless of TREE_TYPE (arg).  Assume non-REAL_CST
3695         value if arg is a REAL_CST with incompatible type.
3697 2017-12-14  Sudakshina Das  <sudi.das@arm.com>
3698             Bin Cheng  <bin.cheng@arm.com>
3700         PR target/81228
3701         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3702         to CCFPEmode.
3703         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3704         LTGT.
3706 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3708         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3709         cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3711 2017-12-14  Richard Sandiford  <richard.sandiford@linaro.org>
3712             Alan Hayward  <alan.hayward@arm.com>
3713             David Sherwood  <david.sherwood@arm.com>
3715         * poly-int.h: New file.
3716         * poly-int-types.h: Likewise.
3717         * coretypes.h: Include them.
3718         (POLY_INT_CONVERSION): Define.
3719         * target.def (estimated_poly_value): New hook.
3720         * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3721         * doc/tm.texi: Regenerate.
3722         * doc/poly-int.texi: New file.
3723         * doc/gccint.texi: Include it.
3724         * doc/rtl.texi: Describe restrictions on subreg modes.
3725         * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3726         * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3727         (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3728         * targhooks.h (default_estimated_poly_value): Declare.
3729         * targhooks.c (default_estimated_poly_value): New function.
3730         * target.h (estimated_poly_value): Likewise.
3731         * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3732         (wi::unary_traits): Delete.
3733         (wi::binary_traits::signed_shift_result_type): Define for
3734         offset_int << HOST_WIDE_INT, etc.
3735         (generic_wide_int::operator <<=): Define for all types and use
3736         wi::lshift instead of <<.
3737         (wi::hwi_with_prec): Add a default constructor.
3738         (wi::ints_for): New class.
3739         (operator <<): Define for all wide-int types.
3740         (operator /): New function.
3741         (operator %): Likewise.
3742         * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3743         (ASSERT_MAYBE_NE_AT): New macros.
3745 2017-12-13  Eric Botcazou  <ebotcazou@adacore.com>
3746             Dominik Vogt  <vogt@linux.vnet.ibm.com>
3748         PR middle-end/78468
3749         * emit-rtl.c (init_emit): Remove ??? comment.
3750         * explow.c (get_dynamic_stack_size): Take known alignment of stack
3751         pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3752         * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3753         alignment of 3 virtual registers to BITS_PER_WORD.
3755         * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3757 2017-12-13  Peter Bergner  <bergner@vnet.ibm.com>
3759         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3760         * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3762 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3764         PR bootstrap/83396
3765         * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3767 2017-12-13  Jakub Jelinek  <jakub@redhat.com>
3769         * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3770         with the exception of debug begin stmt markers appear before
3771         labels.
3773         PR bootstrap/83396
3774         * final.c (rest_of_handle_final): Call variable_tracking_main only
3775         if !flag_var_tracking.
3777 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3778             Jakub Jelinek  <jakub@redhat.com>
3780         PR bootstrap/83396
3781         PR debug/83391
3782         * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3783         labels debug stmts that can only appear after labels.
3785 2017-12-13  Alexander Monakov  <amonakov@ispras.ru>
3787         PR rtl-optimization/82398
3788         * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3789         EXPR_USEFULNESS in priority comparison.
3791 2017-12-13  Segher Boessenkool  <segher@kernel.crashing.org>
3793         PR rtl-optimization/83393
3794         * combine.c (move_deaths): If reg_stat points to a too new insn in
3795         last_death, do not use it: find the proper insn instead.
3797 2017-12-12  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3799         PR target/82190
3800         * config/rs6000/rs6000-string.c (expand_block_compare,
3801         expand_strn_compare): Fix set_mem_size() calls.
3803 2017-12-12  Jeff Law  <law@redhat.com>
3805         PR tree-optimization/83298
3806         PR tree-optimization/83362
3807         PR tree-optimization/83383
3808         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3809         push_value_range a public interface.  Add new argument to
3810         record_ranges_from_stmt.
3811         * gimple-ssa-evrp-analyze.c
3812         (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3813         Update comments.  Handle recording temporary equivalences.
3814         * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3815         new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3816         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3817         * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3818         gimple-ssa-evrp-analyze.h.
3819         (record_temporary_equivalences_from_phis): Add new argument.  When
3820         the PHI arg is an SSA_NAME, set the result's range to the range
3821         of the PHI arg.
3822         (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3823         from statements too.
3824         (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3825         Pass it down to children as needed.
3826         (thread_outgoing_edges): Likewise.
3827         (thread_across_edge): Likewise.   Push/pop range state as needed.
3828         * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3830 2017-12-12  Julia Koval  <julia.koval@intel.com>
3832         * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3833         (PTA_CANNONLAKE): Remove PTA_CLWB.
3835 2017-12-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3837         PR target/83332
3838         * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3839         (vcondv2div2df): Likewise.
3840         (vconduv2dfv2di): Likewise.
3842 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3844         * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3845         BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3846         BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3847         ATTR_MATHFN_FPROUNDING_ERRNO.
3849 2017-12-12  Richard Biener  <rguenther@suse.de>
3851         PR tree-optimization/83385
3852         * graphite-scop-detection.c (get_order, order): Remove.
3853         (bb_to_rpo): New global.
3854         (cmp_pbbs): Adjust.
3855         (build_scops): Sort pbbs in RPO order.
3857 2017-12-12  James Greenhalgh  <james.greenhalgh@arm.com>
3859         * combine.c (simplify_set): Do not transform subregs to zero_extends
3860         if the destination mode is a vector mode.
3862 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3864         PR tree-optimization/83359
3865         * tree-cfg.h (fold_loop_internal_call): Declare.
3866         * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3867         * tree-cfg.c (fold_loop_internal_call): ... here.  No longer static.
3868         (find_loop_dist_alias): New function.
3869         (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3870         updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3871         calls.
3873         PR tree-optimization/80631
3874         * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3875         (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3876         arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3877         hardcoding zero as the value if COND_EXPR is never true.  For
3878         INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3879         INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3880         hardcoding MAX_EXPR as the reduction operation.
3881         (is_nonwrapping_integer_induction): Allow negative step.
3882         (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3883         vect_create_epilog_for_reduction, if no value is suitable, don't
3884         use INTEGER_INDUC_COND_REDUCTION for now.  Formatting fixes.
3886 2017-12-12  Richard Biener  <rguenther@suse.de>
3888         PR tree-optimization/81889
3889         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3890         range info from the non-wrapping IV instead of just the range
3891         of the type.
3893 2017-12-12  Julia Koval  <julia.koval@intel.com>
3895         * config.gcc: Add vaesintrin.h.
3896         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3897         New type.
3898         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3899         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3900         New builtins.
3901         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3902         * config/i386/immintrin.h: Include vaesintrin.h.
3903         * config/i386/sse.md (vaesdec_<mode>): New pattern.
3904         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3905         _mm_aesdec_epi128): New intrinsics.
3907 2017-12-12  Julia Koval  <julia.koval@intel.com>
3909         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3910         OPTION_MASK_ISA_VAES_UNSET): New.
3911         (ix86_handle_option): Handle -mvaes.
3912         * config/i386/cpuid.h: Define bit_VAES.
3913         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3914         * config/i386/i386-c.c (__VAES__): New.
3915         * config/i386/i386.c (ix86_target_string): Add -mvaes.
3916         (ix86_valid_target_attribute_inner_p): Ditto.
3917         * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
3918         * config/i386/i386.opt: Add -mvaes.
3919         * doc/invoke.texi: Ditto.
3921 2017-12-12  Alexandre Oliva  <aoliva@redhat.com>
3923         * debug.h (gcc_debug_hooks): Add inline_entry.
3924         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
3925         * debug.c (do_nothing_debug_hooks): Likewise.
3926         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
3927         * dwarf2out.c (dwarf2_debug_hooks): Likewise.
3928         (dwarf2_lineno_debug_hooks): Likewise.
3930         * common.opt (gstatement-frontiers): New, setting
3931         debug_nonbind_markers_p.
3932         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
3933         * toplev.c (process_options): Autodetect value for debug statement
3934         frontiers option.
3935         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
3936         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
3938         * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
3939         markers.  Integrate source bind into debug stmt expand loop.
3940         (pass_expand::execute): Check debug marker limit.  Avoid deep
3941         TER and expand debug locations for debug bind insns only.
3942         * cse.c (insn_live_p): Keep nonbind markers and debug bindings
3943         followed by them.
3944         * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
3945         * final.c (reemit_insn_block_notes): Take current block from
3946         nonbind markers.  Declare note where it's first set.
3947         (final_scan_insn): Handle begin stmt notes.  Emit is_stmt according to
3948         begin stmt markers if enabled.
3949         (notice_source_line): Handle nonbind markers.  Fail if their
3950         location is unknown or that of builtins.
3951         (rest_of_handle_final): Convert begin stmt markers to notes if
3952         var-tracking didn't run.
3953         (rest_of_clean_state): Skip begin stmt markers.
3954         * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
3955         markers.
3956         * function.c (allocate_struct_function): Set begin_stmt_markers.
3957         * function.h (struct function): Add debug_marker_count counter
3958         and debug_nonbind_markers flag.
3959         * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
3960         * gimple-low.c (lower_function_body): Adjust
3961         debug_nonbind_markers.
3962         (lower_stmt): Drop or skip gimple debug stmts.
3963         (lower_try_catch): Skip debug stmts.
3964         * gimple.c (gimple_build_debug_begin_stmt): New.
3965         (gimple_copy): Increment debug_marker_count if copying one.
3966         * gimple.h (gimple_build_debug_begin_stmt): Declare.
3967         * gimplify.c (rexpr_location): New.
3968         (rexpr_has_location): New.
3969         (warn_switch_unreachable_r): Handle gimple debug stmts.
3970         (shortcut_cond_r): Call expr_location.
3971         (find_goto): New.
3972         (find_goto_label): New.
3973         (shortcut_cond_expr): Call expr_has_location, expr_location, and
3974         find_goto_label.
3975         (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
3976         expr_location.
3977         (gimplify_expr): Handle begin stmt markers.  Reject debug expr decls.
3978         * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New.  Add to...
3979         (LANG_HOOKS_INITIALIZER): ... this.
3980         * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
3981         * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
3982         debug insns.
3983         (update_ebb_live_info): Skip debug insn markers.
3984         * lra.c (debug_insn_static_data): Rename to...
3985         (debug_bind_static_data): ... this.
3986         (debug_marker_static_data): New.
3987         (lra_set_insn_recog_data): Select one of the above depending
3988         on debug insn kind.
3989         (lra_update_isn_regno_info): Don't assume debug insns have
3990         freqs.
3991         (push_insns): Skip debug insns.
3992         * lto-streamer-in.c (input_function): Drop debug stmts
3993         depending on active options.  Adjust debug_nonbind_markers.
3994         * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
3995         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3996         begin stmt marker notes.
3997         (print_insn): Likewise.
3998         * recog.c (extract_insn): Recognize rtl for debug markers.
3999         * rtl.def (DEBUG_MARKER): New.
4000         * tree-inline.c: Include params.h.
4001         (remap_gimple_stmt): Handle nonbind markers.
4002         (maybe_move_debug_stmts_to_successors): Likewise.
4003         (copy_debug_stmt): Likewise.
4004         * tree-iterator.c (append_to_statement_list_1): Append begin stmt
4005         markers regardless of no side effects.
4006         (tsi_link_before): Don't update container's side effects when adding
4007         a begin stmt marker.
4008         (tsi_link_after): Likewise.
4009         (expr_first): Skip begin stmt markers.
4010         (expr_last): Likewise.
4011         * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
4012         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4013         Disregard nonbind markers.
4014         * tree.c (make_node_stat): Don't set side effects for begin stmt
4015         markers.
4016         (build1_stat): Likewise.
4017         * tree.def (DEBUG_BEGIN_STMT): New.
4018         * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
4019         * var-tracking.c (delete_debug_insns): Renamed to...
4020         (delete_vta_debug_insns): ... this.
4021         (reemit_marker_as_note): New.
4022         (vt_initialize): Reemit markers.
4023         (delete_vta_debug_insns): Likewise.
4024         (vt_debug_insns_local): Reemit or delete markers.
4025         (variable_tracking_main_1): Likewise.
4026         * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
4027         * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
4028         (gimple_debug_nonbind_marker_p): New.
4029         (gimple_build_debug_bind): Adjust.
4030         (gimple_build_debug_begin_stmt): New.
4031         * doc/invoke.texi (max-debug-marker-count): New param.
4032         * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
4033         (debug_parameter_ref, debug_marker): New.
4034         (NOTE_INSN_BEGIN_STMT): New.
4035         (DEBUG_INSN): Describe begin stmt markers.
4037         * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
4038         without debug insns, we wouldn't, but clean up debug insns
4039         after a control flow insn nevertheless.
4041         * cfgcleanup.c (delete_unreachable_blocks): Use alternate
4042         block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
4043         * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
4044         * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
4045         (rtl_tidy_fallthru_edge): Likewise.
4046         (rtl_verify_fallthru): Likewise.
4047         (rtl_verify_bb_layout): Likewise.
4048         (skip_insns_after_block): Likewise.
4049         (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
4050         * dwarf2out.c: Include print-rtl.h.
4051         (dwarf2out_next_real_insn): New.
4052         (dwarf2out_var_location): Call it.  Disregard begin stmt markers.
4053         Dump debug binds in asm comments.
4054         * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
4055         * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
4056         callers to use gsi_start_nondebug_bb instead.
4057         (gsi_after_labels): Skip gimple debug stmts.
4058         (gsi_start_nondebug): New.
4059         * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
4060         (proper_loop_form_for_interchange): Adjust.
4061         * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
4062         * gimple.h (gimple_seq_last_nondebug_stmt): New.
4063         * gimplify.c (last_stmt_in_scope): Skip debug stmts.
4064         (collect_fallthrough_labels): Likewise.
4065         (should_warn_for_implicit_fallthrough): Likewise.
4066         (warn_implicit_fallthrough_r): Likewise.
4067         (expand_FALLTHROUGH_r): Likewise.
4068         * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
4069         (graphite_copy_stmts_from_block): Skip nonbind markers.
4070         * haifa-sched.c (sched_extend_bb): Skip debug insns.
4071         * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
4072         * jump.c (clean_barriers): Skip debug insns.
4073         * omp-expand.c (expand_parallel_call): Skip debug insns.
4074         (expand_task_call): Likewise.
4075         (remove_exit_barrier): Likewise.
4076         (expand_omp_taskreg): Likewise.
4077         (expand_omp_for_init_counts): Likewise.
4078         (expand_omp_for_generic): Likewise.
4079         (expand_omp_for_static_nochunk): Likewise.
4080         (expand_omp_for_static_chunk): Likewise.
4081         (expand_omp_simd): Likewise.
4082         (expand_omp_taskloop_for_outer): Likewise.
4083         (expand_omp_taskloop_for_inner): Likewise.
4084         (expand_oacc_for): Likewise.
4085         (expand_omp_sections): Likewise.
4086         (expand_omp_single): Likewise.
4087         (expand_omp_synch): Likewise.
4088         (expand_omp_atomic_load): Likewise.
4089         (expand_omp_atomic_store): Likewise.
4090         (expand_omp_atomic_fetch_op): Likewise.
4091         (expand_omp_atomic_pipeline): Likewise.
4092         (expand_omp_atomic_mutex): Likewise.
4093         (expand_omp_target): Likewise.
4094         (grid_expand_omp_for_loop): Likewise.
4095         (grid_expand_target_grid_body): Likewise.
4096         (build_omp_regions_1): Likewise.
4097         * omp-low.c (check_combined_parallel): Skip debug stmts.
4098         * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
4099         * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
4100         * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
4101         test.
4102         * tree-cfg.c (make_blobs_1): Skip debug stmts.
4103         (make_edges): Likewise.
4104         (cleanup_dead_labels): Likewise.
4105         (gimple_can_merge_blocks_p): Likewise.
4106         (stmt_starts_bb_p): Likewise.
4107         (gimple_block_label): Likewise.
4108         (gimple_redirect_edge_and_branch): Likewise.
4109         * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
4110         of debug stmts.
4111         (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
4112         TDF_SLIM.
4113         * tree-pretty-print (print_declaration): Omit initializer in slim
4114         dumps.
4115         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
4116         markers.
4117         (eliminate_unnecessary_stmts): Stabilize block removal order.
4118         * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
4119         * var-tracking.c (get_first_insn): New.
4120         (vt_emit_notes): Call it.
4121         (vt_initialize): Walk any insns before the first BB.
4122         (delete_debug_insns): Likewise.
4124         * gimple.h (enum gimple_debug_subcode): Add
4125         GIMPLE_DEBUG_BEGIN_STMT.
4126         (gimple_debug_begin_stmt_p): New.
4127         (gimple_debug_nonbind_marker_p): New.
4128         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
4129         (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
4130         (MAY_HAVE_DEBUG_STMTS): ... this.  Check both.
4131         * insn-notes.def (BEGIN_STMT): New.
4132         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
4133         (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
4134         (MAY_HAVE_DEBUG_INSNS): ... this.  Check both.
4135         (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
4136         (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
4137         (INSN_DEBUG_MARKER_KIND): New.
4138         (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
4139         (INSN_VAR_LOCATION): Check for VAR_LOCATION.
4140         (INSN_VAR_LOCATION_PTR): New.
4141         * cfgexpand.c (expand_debug_locations): Handle debug bind insns
4142         only.
4143         (expand_gimple_basic_block): Likewise.  Emit debug temps for TER
4144         deps only if debug bind insns are enabled.
4145         (pass_expand::execute): Avoid deep TER and expand
4146         debug locations for debug bind insns only.
4147         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
4148         debug stmts special handling down to debug bind stmts.
4149         * combine.c (try_combine): Narrow debug insns special handling
4150         down to debug bind insns.
4151         * cse.c (delete_trivially_dead_insns): Handle debug bindings.
4152         Narrow debug insns preexisting special handling down to debug
4153         bind insns.
4154         * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
4155         handling down to debug bind insns.
4156         * function.c (instantiate_virtual_regs): Skip debug markers,
4157         adjust handling of debug binds.
4158         * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
4159         temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
4160         * haifa-sched.c (schedule_insn): Narrow special handling of debug
4161         insns to debug bind insns.
4162         * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
4163         special handling of debug stmts to debug bind stmts.
4164         * ipa-split.c (split_function): Likewise.
4165         * ira.c (combine_and_move_insns): Adjust debug bind insns only.
4166         * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
4167         debug insns.
4168         * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
4169         * regrename.c (build_def_use): Likewise.
4170         * regcprop.c (copyprop_hardreg_forward_1): Likewise.
4171         (pass_cprop_hardreg): Narrow special casing of debug insns to
4172         debug bind insns.
4173         * regstat.c (regstat_init_n_sets_and_refs): Likewise.
4174         * reload1.c (reload): Likewise.
4175         * sese.c (sese_insert_phis_for_liveouts): Narrow special
4176         casing of debug stmts to debug bind stmts.
4177         * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
4178         * ssa-iterators.h (num_imm_uses): Likewise.
4179         * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
4180         debug stmts to debug bind stmts.
4181         * tree-inline.c (tree_function_versioning): Narrow special casing
4182         of debug stmts to debug bind stmts.
4183         * tree-loop-distribution.c (generate_loops_for_partition):
4184         Narrow special casing of debug stmts to debug bind stmts.
4185         * tree-sra.c (analyze_access_subtree): Narrow special casing
4186         of debug stmts to debug bind stmts.
4187         * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
4188         stmts to debug bind stmts.
4189         * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
4190         casing of debug stmts to debug bind stmts.
4191         * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
4192         * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
4193         casing of debug stmts to debug bind stmts.
4194         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4195         Likewise.
4196         * tree-ssa.c (flush_pending_stmts): Narrow special casing of
4197         debug stmts to debug bind stmts.
4198         (gimple_replace_ssa_lhs): Likewise.
4199         (insert_debug_temp_for_var_def): Likewise.
4200         (insert_debug_temps_for_defs): Likewise.
4201         (reset_debug_uses): Likewise.
4202         * tree-ssanames.c (release_ssa_name_fn): Likewise.
4203         * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
4204         (adjust_debug_stmts): Likewise.
4205         (adjust_phi_and_debug_stmts): Likewise.
4206         (vect_do_peeling): Likewise.
4207         * tree-vect-loop.c (vect_transform_loop): Likewise.
4208         * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
4209         * var-tracking.c (adjust_mems): Narrow special casing of debug
4210         insns to debug bind insns.
4211         (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
4212         (compute_bb_dataflow, vt_find_locations): Likewise.
4213         (vt_expand_loc, emit_notes_for_changes): Likewise.
4214         (vt_init_cfa_base): Likewise.
4215         (vt_emit_notes): Likewise.
4216         (vt_initialize): Likewise.
4217         (vt_finalize): Likewise.
4219         * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
4220         (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
4221         (next_nonnote_nondebug_insn_bb): New.
4222         (prev_nonnote_nondebug_insn_bb): New.
4223         (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
4224         * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
4225         (prev_nonnote_nondebug_insn_bb): Declare.
4226         (next_nonnote_nondebug_insn_bb): Declare.
4227         * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
4228         * cfgrtl.c (get_last_bb_insn): Likewise.
4229         * lra.c (push_insns): Likewise.
4231 2017-12-11  David Malcolm  <dmalcolm@redhat.com>
4233         PR c/82050
4234         * selftest-run-tests.c (selftest::run_tests): Move start/finish code
4235         to...
4236         * selftest.c (selftest::test_runner::test_runner): New ctor.
4237         (selftest::test_runner::~test_runner): New dtor.
4238         * selftest.h (class selftest::test_runner): New class.
4240 2017-12-11  Carl Love  <cel@us.ibm.com>
4242         * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
4243         vec_extract_fp32_from_shortl]): Add #defines.
4244         * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4245         * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4246         ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4247         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4248         ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4249         ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4250         * doc/extend.texi: Add documentation for the added builtins.
4252 2017-12-11  Alexandre Oliva  <aoliva@redhat.com>
4254         PR rtl-optimization/80693
4255         PR rtl-optimization/81019
4256         PR rtl-optimization/81020
4257         * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4258         are not mentioned in i3.  Place the REG_UNUSED note on i2,
4259         possibly modified to REG_DEAD, if it did not originate in i3.
4261 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4263         * recog.c (store_data_bypass_p_1): New function.
4264         (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER.  Use
4265         store_data_bypass_p_1 to avoid code duplication.  Formatting fixes.
4267 2017-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
4269         PR rtl-optimization/83361
4270         * ifcvt.c (if_convert): Call fixup_partitions.
4272 2017-12-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
4274         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4275         early folding of splat_u{8,16,32}.
4277 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4279         * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4280         output_operand_lossage first argument with capital letter.
4281         (aarch64_override_options): Don't start error and sorry first argument
4282         with capital letter.
4284 2017-12-11  Andi Kleen  <ak@linux.intel.com>
4286         PR gcov-profile/83355
4287         * auto-profile.c (string_table::get_index_by_decl): Don't
4288         recurse when abstract origin points to itself.
4290 2017-12-11  Bin Cheng  <bin.cheng@arm.com>
4292         PR tree-optimization/83320
4293         * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4294         (prune_datarefs_not_in_loop): Ditto.
4296 2017-12-10  Gerald Pfeifer  <gerald@pfeifer.com>
4298         * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4300 2017-12-10  Jakub Jelinek  <jakub@redhat.com>
4302         PR tree-optimization/83337
4303         * gimple-loop-interchange.cc (compute_access_stride): Handle
4304         bitfield DRs properly.
4306 2017-12-09  Jakub Jelinek  <jakub@redhat.com>
4308         PR tree-optimization/83338
4309         * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4310         vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4311         vector of unsigned integers to vector of signed integers.
4313 2017-12-08  Vladimir Makarov  <vmakarov@redhat.com>
4315         PR rtl-optimization/83317
4316         * lra-constraints.c (process_address_1): Add insn code check.
4318 2017-12-08  Michael Matz  <matz@suse.de>
4320         Fix PR tree-optimization/83323
4321         * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4322         head-controlled loops and loop BBs.
4323         * common.opt (funroll-and-jam): Remove, instead ...
4324         (floop-unroll-and-jam): ... reuse this option.
4325         * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4326         * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4327         (-floop-unroll-and-jam): ... this option.
4329 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4331         * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4332         a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4333         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4335 2017-12-08  Richard Biener  <rguenther@suse.de>
4337         PR middle-end/81782
4338         * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4339         handle accesses outside of zero-sized vars.
4341 2017-12-08  Martin Jambor  <mjambor@suse.cz>
4343         PR tree-optimization/83141
4344         * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4345         test for MEM_REFs implicitely changing types with padding.  Remove
4346         inline keyword.
4347         (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4349 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4351         * config/arc/arc.c (arc_attribute_table): Add exclusions to
4352         the comment.
4353         * config/avr/avr.c (avr_attribute_table): Likewise.
4354         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4355         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4356         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4357         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4358         of struct attribute_spec.
4359         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4361 2017-12-08  Ulrich Weigand  <uweigand@de.ibm.com>
4363         PR target/82960
4364         * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4366 2017-12-08  Jan Hubicka  <hubicka@ucw.cz>
4368         PR middle-end/83609
4369         * profile-count.c (profile_count::from_gcov_type): Move from
4370         profile-count.h; handle overflow.
4371         * profile-count.h (profile_count::from_gcov_type): Move offline.
4373 2017-12-08  Segher Boessenkool  <segher@kernel.crashing.org>
4375         PR rtl-optimization/83304
4376         * combine.c (move_deaths): If we do not know where a register died,
4377         search for it.
4379 2017-12-08  Richard Biener  <rguenther@suse.de>
4381         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4382         Provide -fopt-info-loop feedback when we interchange in a nest.
4384 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4386         * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4387         for armv6 ARM CPU IDs.
4389 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4391         * common/config/arm/arm-common.c: Include <algorithm>.
4392         (INCLUDE_VECTOR): Define.
4393         (compare_opt_names): New function.
4394         (arm_rewrite_selected_arch): Only strip out extensions that can be
4395         expressed through -mfpu.  Sort the remaining extensions
4396         alphabetically.
4398 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4400         * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4401         (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4402         (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4403         * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4405 2017-06-08  Tristan Gingold  <gindold@adacore.com>
4407         PR ada/81470
4408         * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4409         .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4411 2017-12-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4413         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4414         of struct attribute_spec.
4416 2017-12-08  Julia Koval  <julia.koval@intel.com>
4418         * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4419         _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4420         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4421         _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4422         _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4423         _mm_maskz_dpwssds_epi32): Ditto.
4425 2017-12-08  Richard Biener  <rguenther@suse.de>
4427         PR tree-optimization/81303
4428         * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4429         conditions try to create a comparison vector type matching
4430         the data vector type.
4431         (vectorizable_condition): Adjust.
4432         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4433         Leave invariant conditions alone in case we can vectorize those.
4435 2017-12-08  Julia Koval  <julia.koval@intel.com>
4437         * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4438         _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4439         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4440         _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4441         _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4443 2017-12-08  Julia Koval  <julia.koval@intel.com>
4445         * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4446         _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4447         * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4448         _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4449         _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4450         _mm_maskz_dpbusds_epi32): New intrinsics.
4452 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4454         * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4455         operand.
4456         (ld<bh>io_signed): Likewise.
4457         (st<bhw>io): Likewise.
4458         * config/nios2/predicates.md (ldstio_memory_operand): Allow
4459         SMALL_INT12 constant integer operand.
4461 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4463         * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4464         Rename to...
4465         (nios2_large_constant_allowed): ...this.  Adjust uses.
4466         (nios2_plus_symbolic_constant_p): Rename to...
4467         (nios2_plus_large_constant_p): ...this.  Adjust uses.
4468         (nios2_legitimate_address_p): Correct CONST_INT handling.
4469         (nios2_symbolic_memory_operand_p): Rename to...
4470         (nios2_large_constant_memory_operand_p): ...this.  Adjust uses.
4471         (nios2_large_constant_p): Check for large constant integers too.
4472         (nios2_split_large_constant): Handle constant integers.
4473         (nios2_split_symbolic_memory_operand): Rename to...
4474         (nios2_split_large_constant_memory_operand): ...this.  Adjust uses.
4475         (nios2_legitimize_constant_address): Handle constant integers.
4476         (r0rel_constant_p): Handle small constant integers.
4477         (nios2_print_operand_address): Handle r0-relative integer addresses.
4478         * config/nios2/nios2-protos.h: Adjust for renamed functions.
4479         * config/nios2/nios2.md: Adjust for renamed functions.
4481 2017-12-07  Andrew Waterman  <andrew@sifive.com>
4483         * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4484         (TARGET_HAVE_SRODATA_SECTION): New define.
4485         (riscv_select_section): New function.
4487 2017-12-08  Joseph Myers  <joseph@codesourcery.com>
4488             Alexander Monakov  <amonakov@ispras.ru>
4489             Jakub Jelinek  <jakub@redhat.com>
4491         PR target/81906
4492         * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4494 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4496         * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4497         a bool success value.  Don't call output_operand_lossage here.
4498         (aarch64_print_ldpstp_address): Return a bool success value.
4499         (aarch64_print_operand_address): Call output_addr_const if
4500         aarch64_print_address_internal fails.
4501         (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4502         'y'; call output_operand_lossage instead.  Call output_operand_lossage
4503         if aarch64_print_ldpstp_address fails.
4505 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4507         * tree-vector-builder.h
4508         (tree_vector_builder::binary_encoded_nelts): Declare.
4509         * tree-vector-builder.c
4510         (tree_vector_builder::binary_encoded_nelts): New function.
4511         * fold-const.c (negate_expr_p): Likewise.
4512         (operand_equal_p, fold_checksum_tree): Likewise.
4513         * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4514         * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4515         (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4516         (uniform_vector_p): Likewise.
4517         * varasm.c (const_hash_1, compare_constant): Likewise.
4518         * tree-ssa-ccp.c: Include tree-vector-builder.h.
4519         (valid_lattice_transition): Operate directly on the VECTOR_CST
4520         encoding.
4521         * ipa-icf.c: Include tree-vector-builder.h.
4522         (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4523         * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4525 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4527         * tree.c (build_vector): Delete.
4528         * tree.h (build_vector): Make static and move into the self-testing
4529         block.
4531 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4533         * vector-builder.h (vector_builder::derived): New const overload.
4534         (vector_builder::elt): New function.
4535         * tree-vector-builder.h (tree_vector_builder::type): New function.
4536         (tree_vector_builder::apply_step): Declare.
4537         * tree-vector-builder.c (tree_vector_builder::apply_step): New
4538         function.
4539         * gimple-fold.h (tree_vector_builder): Declare.
4540         (gimple_build_vector): Take a tree_vector_builder instead of a
4541         type and vector of elements.
4542         * gimple-fold.c (gimple_build_vector): Likewise.
4543         * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4544         accordingly.
4545         (get_initial_defs_for_reduction): Likewise.
4546         (vectorizable_induction): Likewise.
4548 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4550         * tree-vector-builder.h
4551         (tree_vector_builder::new_binary_operation): Declare.
4552         * tree-vector-builder.c
4553         (tree_vector_builder::new_binary_operation): New function.
4554         * fold-const.c (fold_relational_const): Use it.
4555         (const_binop): Likewise.  Check that both input vectors have
4556         the same number of elements, thus excluding things like WIDEN_SUM.
4557         Check whether it is possible to operate directly on the encodings
4558         of stepped inputs.
4560 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4562         * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4563         new_unary_operation, operating only on the encoded elements.
4564         (const_unop): Likewise.
4565         (exact_inverse): Likewise.
4566         (distributes_over_addition_p): New function.
4567         (const_binop): Use tree_vector_builder and new_unary_operation
4568         for combinations of VECTOR_CST and INTEGER_CST.  Operate only
4569         on the encoded elements unless the encoding is strided and the
4570         operation does not distribute over addition.
4571         (fold_convert_const):  Use tree_vector_builder and
4572         new_unary_operation.  Operate only on the encoded elements
4573         for truncating integer conversions, or for non-stepped encodings.
4575 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4577         * config/sparc/sparc.c: Include tree-vector-builder.h.
4578         (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4579         * expmed.c: Include tree-vector-builder.h.
4580         (make_tree): Use tree_vector_builder instead of build_vector.
4581         * fold-const.c: Include tree-vector-builder.h.
4582         (const_binop): Use tree_vector_builder instead of build_vector.
4583         (const_unop): Likewise.
4584         (native_interpret_vector): Likewise.
4585         (fold_vec_perm): Likewise.
4586         (fold_ternary_loc): Likewise.
4587         * gimple-fold.c: Include tree-vector-builder.h.
4588         (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4589         of build_vector.
4590         * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4591         (simplify_vector_constructor): Use tree_vector_builder instead
4592         of build_vector.
4593         * tree-vect-generic.c: Include tree-vector-builder.h.
4594         (add_rshift): Use tree_vector_builder instead of build_vector.
4595         (expand_vector_divmod): Likewise.
4596         (optimize_vector_constructor): Likewise.
4597         * tree-vect-loop.c: Include tree-vector-builder.h.
4598         (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4599         of build_vector.  Explicitly use a stepped encoding for
4600         { 1, 2, 3, ... }.
4601         * tree-vect-slp.c: Include tree-vector-builder.h.
4602         (vect_get_constant_vectors): Use tree_vector_builder instead
4603         of build_vector.
4604         (vect_transform_slp_perm_load): Likewise.
4605         (vect_schedule_slp_instance): Likewise.
4606         * tree-vect-stmts.c: Include tree-vector-builder.h.
4607         (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4608         (vect_gen_perm_mask_any): Likewise.
4609         (vectorizable_call): Likewise.  Explicitly use a stepped encoding.
4610         * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4611         of build_vector.
4612         (build_vector_from_val): Likewise.  Explicitly use a duplicate
4613         encoding.
4615 2017-12-07  Richard Sandiford  <richard.sandiford@arm.com>
4617         * doc/generic.texi (VECTOR_CST): Describe new representation of
4618         vector constants.
4619         * vector-builder.h: New file.
4620         * tree-vector-builder.h: Likewise.
4621         * tree-vector-builder.c: Likewise.
4622         * Makefile.in (OBJS): Add tree-vector-builder.o.
4623         * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4624         * tree-core.h (tree_base): Add a vector_cst field to the u union.
4625         (tree_vector): Change the number of elements to
4626         vector_cst_encoded_nelts.
4627         * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4628         (VECTOR_CST_ELTS): Delete.
4629         (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4630         (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4631         (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4632         (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4633         (VECTOR_CST_ENCODED_ELT): Likewise.
4634         (vector_cst_encoded_nelts): New function.
4635         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4636         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4637         (vector_cst_int_elt, vector_cst_elt): Declare.
4638         * tree.c: Include tree-vector-builder.h.
4639         (tree_code_size): Abort if passed VECTOR_CST.
4640         (tree_size): Update for new VECTOR_CST layout.
4641         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4642         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4643         (build_vector): Use tree_vector_builder.
4644         (vector_cst_int_elt, vector_cst_elt): New functions.
4645         (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4646         encoded elements and then create the vector in the canonical form.
4647         (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4648         (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4649         (tree_c_tests): Call test_vector_cst_patterns.
4650         * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4651         VECTOR_CST fields.
4652         (hash_tree): Likewise.
4653         * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4654         (streamer_write_tree_header): Likewise.
4655         * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4656         (streamer_alloc_tree): Likewise.  Update call to make_vector.
4657         * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4659 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4661         * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4662         (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4663         * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4665 2017-12-07  Bin Cheng  <bin.cheng@arm.com>
4666             Richard Biener  <rguenther@suse.de>
4668         PR tree-optimization/81303
4669         * Makefile.in (gimple-loop-interchange.o): New object file.
4670         * common.opt (floop-interchange): Reuse the option from graphite.
4671         * doc/invoke.texi (-floop-interchange): Ditto.  New document for
4672         -floop-interchange and mention it for -O3.
4673         * opts.c (default_options_table): Enable -floop-interchange at -O3.
4674         * gimple-loop-interchange.cc: New file.
4675         * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4676         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4677         * passes.def (pass_linterchange): New pass.
4678         * timevar.def (TV_LINTERCHANGE): New time var.
4679         * tree-pass.h (make_pass_linterchange): New declaration.
4680         * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4681         interchange.  Record IV before/after increment in new parameters.
4682         * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4683         * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4684         path check into...
4685         (check_reduction_path): ...New function here.
4686         * tree-vectorizer.h (check_reduction_path): New declaration.
4688 2017-12-07  Vladimir Makarov  <vmakarov@redhat.com>
4690         PR target/83252
4691         PR rtl-optimization/80818
4692         * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4693         always early clobbered.
4694         * lra-lives.c (process_bb_lives): Check input hard regs for early
4695         clobbered non-operand hard reg.
4697 2017-12-07  Jakub Jelinek  <jakub@redhat.com>
4699         PR middle-end/83164
4700         * tree-cfg.c (verify_gimple_assign_binary): Don't require
4701         types_compatible_p, just that TYPE_MODE is the same.
4703 2017-12-07  Martin Sebor  <msebor@redhat.com>
4705         PR c/81544
4706         * attribs.c (empty_attribute_table): Initialize new member of
4707         struct attribute_spec.
4708         (decl_attributes): Add argument.  Handle mutually exclusive
4709         combinations of attributes.
4710         (selftests::test_attribute_exclusions): New function.
4711         (selftests::attribute_c_tests): Ditto.
4712         * attribs.h (decl_attributes): Add default argument.
4713         * selftest.h (attribute_c_tests): Declare.
4714         * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4715         * tree-core.h (attribute_spec::exclusions, exclude): New type and
4716         member.
4717         * doc/extend.texi (Common Function Attributes): Update const and pure.
4718         * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4719         of struct attribute_spec.
4720         * config/arc/arc.c (arc_attribute_table): Same.
4721         * config/arm/arm.c (arm_attribute_table): Same.
4722         * config/avr/avr.c ( avr_attribute_table): Same.
4723         * config/bfin/bfin.c (bfin_attribute_table): Same.
4724         * config/cr16/cr16.c (cr16_attribute_table): Same.
4725         * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4726         * config/h8300/h8300.c (h8300_attribute_table): Same.
4727         * config/i386/i386.c (ix86_attribute_table): Same.
4728         * config/ia64/ia64.c (ia64_attribute_table): Same.
4729         * config/m32c/m32c.c (m32c_attribute_table): Same.
4730         * config/m32r/m32r.c (m32r_attribute_table): Same.
4731         * config/m68k/m68k.c (m68k_attribute_table): Same.
4732         * config/mcore/mcore.c (mcore_attribute_table): Same.
4733         * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4734         * config/mips/mips.c (mips_attribute_table): Same.
4735         * config/msp430/msp430.c (msp430_attribute_table): Same.
4736         * config/nds32/nds32.c (nds32_attribute_table): Same.
4737         * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4738         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4739         * config/rl78/rl78.c (rl78__attribute_table): Same.
4740         * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4741         * onfig/rx/rx.c (rx_attribute_table): Same.
4742         * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4743         * config/sh/sh.c (sh_attribute_table): Same.
4744         * config/sparc/sparc.c (sparc_attribute_table): Same.
4745         * config/spu/spu.c (spu_attribute_table): Same.
4746         * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4747         * config/v850/v850.c (v850_attribute_table): Same.
4748         * config/visium/visium.c (visium_attribute_table): Same.
4750 2017-12-07  Tamar Christina  <tamar.christina@arm.com>
4752         PR target/82641
4753         * config/arm/arm.c (INCLUDE_STRING): Define.
4754         (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4755         (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4756         and .fpu.
4758 2017-12-07  Michael Matz  <matz@suse.de>
4760         Add unroll and jam pass
4762         * gimple-loop-jam.c: New file.
4763         * Makefile.in (OBJS): Add gimple-loop-jam.o.
4764         * common.opt (funroll-and-jam): New option.
4765         * opts.c (default_options_table): Add unroll-and-jam at -O3.
4766         * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4767         (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4768         * passes.def: Add pass_loop_jam.
4769         * timevar.def (TV_LOOP_JAM): Add.
4770         * tree-pass.h (make_pass_loop_jam): Declare.
4771         * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4772         * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4773         * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4774         to flow_loop_tree_node_add.
4775         (duplicate_subloops, copy_loops_to): Append to sibling list.
4776         * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4777         * doc/invoke.texi (-funroll-and-jam): Document new option.
4778         (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4780 2017-12-07  Richard Biener  <rguenther@suse.de>
4782         PR tree-optimization/83296
4783         PR tree-optimization/67769
4784         * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4785         flow sensitive info in an unrelated BB.
4786         (value_replacement): Use reset_flow_sensitive_info.
4787         (minmax_replacement): Reset flow sensitive info on the def
4788         we move.  Do not reset flow sensitive info in the whole BB
4789         we move the stmt to.
4790         (abs_replacement): Likewise.
4792 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4794         PR target/43871
4795         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4796         rs6000_cpu to the given -mcpu=, or to the default processor.
4798 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4800         * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4801         * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4802         instead of rs6000_cpu_attr.
4803         (rs6000_adjust_cost): Ditto.
4804         (is_microcoded_insn): Ditto.
4805         (rs6000_adjust_priority): Ditto.
4806         (rs6000_issue_rate): Ditto.
4807         (rs6000_use_sched_lookahead): Ditto.
4808         (rs6000_use_sched_lookahead_guard): Ditto.
4809         (rs6000_sched_reorder): Ditto.
4810         (force_new_group): Ditto.
4811         * config/rs6000/rs6000.md (cpu attribute): Ditto.
4812         (group_ending_nop): Ditto.
4814 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4816         * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4817         * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4818         rs6000_tune.  Use rs6000_tune instead of rs6000_cpu where appropriate.
4819         (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4820         appropriate.
4821         (rs6000_reassociation_width): Ditto.
4822         (rs6000_emit_epilogue): Ditto.
4823         (rs6000_adjust_cost): Ditto.
4824         (is_microcoded_insn): Ditto.
4825         (is_cracked_insn): Ditto.
4826         (rs6000_adjust_priority): Ditto.
4827         (rs6000_sched_reorder): Ditto.
4828         (rs6000_sched_reorder2): Ditto.
4829         (insn_must_be_first_in_group): Ditto.
4830         (insn_must_be_last_in_group): Ditto.
4831         (rs6000_register_move_cost): Ditto.
4832         * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4833         rs6000_cpu.
4835 2017-12-07  Julia Koval  <julia.koval@intel.com>
4837         * config.gcc: Add vaesintrin.h.
4838         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4839         New type.
4840         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4841         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4842         New builtins.
4843         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4844         * config/i386/immintrin.h: Include vaesintrin.h.
4845         * config/i386/sse.md (vaesdec_<mode>): New pattern.
4846         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4847         _mm_aesdec_epi128): New intrinsics.
4849 2017-12-06  David Malcolm  <dmalcolm@redhat.com>
4851         * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4852         * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4853         c-family/c-spellcheck.cc.
4854         (best_macro_match::best_macro_match): Likewise.
4855         * spellcheck-tree.h
4856         (struct edit_distance_traits<cpp_hashnode *>): Move to
4857         c-family/c-spellcheck.h.
4858         (class best_macro_match): Likewise.
4860 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4862         PR tree-optimization/83293
4863         * gimple-ssa-strength-reduction.c (insert_initializers): Use
4864         GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4865         might insert into empty bb.
4867         PR sanitizer/81281
4868         * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4869         simplify for plus with :c added, and pointer_plus without that.
4870         ((T)P - (T)(P + A) -> -(T) A): Likewise.  If type is integral
4871         with undefined overflow and the conversion is not widening,
4872         perform negation in utype and only convert to type afterwards.
4873         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4874         simplify for plus with :c added, and pointer_plus without that.
4875         If type is integral with undefined overflow and the conversion is
4876         not widening, perform minus in utype and only convert to type
4877         afterwards.  Move the last pointer_diff_expr simplify into the
4878         two outermost ifs.
4880 2017-12-06  Martin Sebor  <msebor@redhat.com>
4882         PR tree-optimization/82646
4883         * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4884         strncpy, not maxlen.
4886 2017-12-06  Martin Sebor  <msebor@redhat.com>
4888         * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4889         nonstring.
4891         PR tree-optimization/83075
4892         * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4893         strncat/strncpy don't change length of source string.
4895 2017-12-06  Eric Botcazou  <ebotcazou@adacore.com>
4897         Revert
4898         2017-11-29  Martin Aberg  <maberg@gaisler.com>
4900         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4901         to prevent b2bst errata sequence.
4902         (sqrtdf2_fix): Likewise.
4904 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4906         PR tree-optimization/81945
4907         * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4908         * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4909         to dest_cfun has orig_loop_num set, either remap it to the new
4910         loop number if the loop got moved too, or clear it.
4912 2017-12-05  Steve Ellcey  <sellcey@cavium.com>
4914         * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4915         to reservation.
4916         (thunderx2t99_nothing): New insn reservation.
4917         (thunderx2t99_mrs): New insn reservation.
4918         (thunderx2t99_multiple): New insn reservation.
4919         (thunderx2t99_alu_basi): Add bfx to reservation.
4920         (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
4922 2017-12-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4924         PR target/82248
4925         * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
4927 2017-12-05  Bin Cheng  <bin.cheng@arm.com>
4929         * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
4930         tree-ssa-pre.c::remove_dead_inserted_code.
4931         * tree-ssa-dce.h: New file.
4932         * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
4933         (remove_dead_inserted_code): Move and rename to function
4934         tree-ssa-dce.c::simple_dce_from_worklist.
4935         (pass_pre::execute): Update use.
4937 2017-12-05  Richard Biener  <rguenther@suse.de>
4939         PR tree-optimization/83277
4940         * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
4941         to code-gen liveout vars.
4943 2017-12-05  Richard Sandiford  <richard.sandiford@linaro.org>
4945         * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
4946         (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
4947         (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
4948         split condition.
4950 2017-12-05  Max Filippov  <jcmvbkbc@gmail.com>
4952         * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
4953         function.
4954         (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
4955         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
4956         ASAN is enabled.
4958 2017-12-05  Richard Biener  <rguenther@suse.de>
4960         * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
4961         * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
4962         (pass_data_cse_sincos): Use TV_TREE_SINCOS.
4963         (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
4965 2017-12-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4967         * dbxout.c (dbxout_block): Grow buf to 30 bytes.
4969 2017-12-05  Martin Liska  <mliska@suse.cz>
4970             Jakub Jelinek  <jakub@redhat.com>
4972         * doc/invoke.texi: Document the options.
4973         * flag-types.h (enum sanitize_code): Add
4974         SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4975         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
4976         of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4977         * opts.c: Define new sanitizer options.
4978         * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
4979         (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
4981 2017-12-05  Julia Koval  <julia.koval@intel.com>
4983         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
4984         OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
4985         (ix86_handle_option): Handle -mavx512vnni.
4986         * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
4987         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
4988         * config/i386/i386-c (__AVX512VNNI__): New.
4989         * config/i386/i386.c (ix86_target_string): Handle new option.
4990         (ix86_valid_target_attribute_inner_p): Handle new option.
4991         * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
4992         * config/i386/i386.opt (mavx512vnni): New option.
4994 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4996         PR target/81616
4997         * athlon.md: Disable for generic.
4998         * haswell.md: Enable for generic.
4999         * i386.c (ix86_sched_init_global): Add core hooks for generic.
5000         * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
5001         to 4.
5002         (ix86_adjust_cost): Move generic to haswell path.
5004 2017-12-04  Eric Botcazou  <ebotcazou@adacore.com>
5006         * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
5007         instead of MEM_P in a couple more places.  Fix formatting issues.
5009 2017-12-04  Jim Wilson  <jimw@sifive.com>
5011         * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
5012         instead of GP_REG_LAST-1.
5013         (riscv_adjust_libcall_cfi_prologue): Likewise.
5014         (riscv_adjust_libcall_cri_epilogue): Likewise.
5015         * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
5016         comment.
5018 2017-12-04  Luis Machado  <luis.machado@linaro.org>
5020         * ipa-pure-const.c (check_decl): Add missing newline.
5021         (state_from_flags): Likewise.
5023 2017-12-04  Jeff Law  <law@redhat.com>
5025         PR tree-optimizatin/78496
5026         * gimple-ssa-evrp-analyze.h
5027         (evrp_range_analyzer::get_vr_values): Simplify.
5028         * gimple-ssa-evrp-analyze.c: Corresponding changes.
5029         * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
5030         and gimple-ssa-evrp-analyze.h.
5031         (dom_opt_dom_walker class): Add evrp_range_analyzer member.
5032         (simplify_stmt_for_jump_threading): Copy a blob of code from
5033         tree-vrp.c to use ranges to simplify statements.
5034         (dom_opt_dom_walker::before_dom_children): Call
5035         evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
5036         (dom_opt_dom_walker::after_dom_children): Similarly for
5037         evrp_range_analyzer::leave.
5038         (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
5039         conditionals.
5041         * gimple-ssa-evrp-analyze.c
5042         (evrp_range_analyzer::extract_range_from_stmt):  Always use
5043         vr_values::update_value_range so preexisting range info is
5044         medged with new range info, even if the new range is VR_VARYING.
5046 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5048         * combine.c: Adjust comment.
5049         (use_crosses_set_p): Delete.
5050         (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
5051         (try_combine): Ditto.
5053 2017-12-04  Richard Biener  <rguenther@suse.de>
5055         PR tree-optimization/83255
5056         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
5057         Re-add zero-iteration check.
5059 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5061         PR rtl-optimization/83245
5062         * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
5063         hard registers as earlyclobber, also if not in an asm.
5065 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5067         PR bootstrap/83265
5068         Revert
5069         2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5071         PR target/43871
5072         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5073         rs6000_cpu based on cpu_index, not tune_index.
5075 2017-12-04  Richard Biener  <rguenther@suse.de>
5077         PR tree-optimization/83238
5078         * graphite-scop-detection.c (scop_detection::merge_sese): Make
5079         code match comment, rejecting invalid SESE regions.
5081 2017-12-03  John David Anglin  <danglin@gcc.gnu.org>
5083         * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
5084         require base operand is a REG_POINTER prior to reload on targets
5085         with non-equivalent space registers.
5087 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5089         * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
5090         (update_specialized_profile): Fix updating of counts.
5091         (perhaps_add_new_callers): Likewise.
5093 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5095         PR target/81616
5096         * x86-tune.def: Remove obsolette FIXMEs.
5097         (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
5098         (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
5099         X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
5100         Enable for generic.
5101         (X86_TUNE_PAD_RETURNS): Disable for generic.
5103 2017-12-02  Jakub Jelinek  <jakub@redhat.com>
5105         PR tree-optimization/83170
5106         PR tree-optimization/83241
5107         * gimple-ssa-store-merging.c
5108         (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
5109         gimple_vuse (ins_stmt) in case it has changed.
5110         (imm_store_chain_info::output_merged_store): Likewise.
5112         * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
5113         POINTER_DIFF_EXPR.
5115         PR c++/81212
5116         * tree-cfg.c (pass_warn_function_return::execute): Handle
5117         __builtin_ubsan_handle_missing_return like __builtin_unreachable
5118         with BUILTINS_LOCATION.
5120         PR target/78643
5121         PR target/80583
5122         * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
5123         is BLKmode for vector field with vector raw mode, use TYPE_MODE
5124         instead of DECL_MODE.
5126         * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
5127         last argument to rtx pointer.
5128         * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
5129         with OPERANDS.  For AVX+ 128-bit VEX encoded instructions over 256-bit
5130         or 512-bit.  If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
5131         depending on the chosen ISAs.
5132         * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
5133         *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
5134         *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
5135         callers.
5136         * config/i386/sse.md (mov<mode>_internal): Likewise.
5137         * config/i386/mmx.md (*mov<mode>_internal): Likewise.
5139 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5141         * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
5143 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5145         * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
5146         parameters from prototype.
5147         * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
5148         parameters.  Don't print a comment.
5149         (emit_fusion_gpr_load): Adjust.
5150         (emit_fusion_load_store): Adjust.
5151         * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
5152         * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
5153         comment on the second line.
5155 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5157         PR target/43871
5158         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5159         rs6000_cpu based on cpu_index, not tune_index.
5161 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5163         * final.c (output_asm_name): Print insn_cost.  Shorten output.  Print
5164         which_alternative instead of which_alternative + 1.
5165         (output_asm_insn): Print an extra tab if the template is short.
5167 2017-12-01  Jim Wilson  <jimw@sifive.com>
5169         * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
5170         comment.
5171         * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
5172         * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
5173         reference.
5174         * doc/tm.texi: Regenerate.
5176 2017-12-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
5178         PR target/81959
5179         * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
5180         whether we can allocate pseudos before trying to fix an address.
5181         * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
5182         memory address is indexed or indirect.
5183         (floatuns_<mode>si2_hw2): Likewise.
5185 2017-12-01  Jason Merrill  <jason@redhat.com>
5187         * Makefile.in (TAGS): Add c-family/*.cc.
5189 2017-12-01  Wilco Dijkstra  <wdijkstr@arm.com>
5191         * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
5192         (call_value_insn): Likewise.
5193         (sibcall_insn): Likewise.
5194         (sibcall_value_insn): Likewise.
5195         (movsi_aarch64): Likewise.
5196         (movdi_aarch64): Likewise.
5197         (add_losym_): Likewise.
5198         (ldr_got_small_): Likewise.
5199         (ldr_got_small_sidi): Likewise.
5200         (ldr_got_small_28k_): Likewise.
5201         (ldr_got_small_28k_sidi): Likewise.
5202         * config/aarch64/aarch64.c (aarch64_print_address_internal):
5203         Move output_addr_const to symbolic case. Add error check.
5205 2017-12-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5207         * config/s390/predicates.md (plus16_Q_operand): New predicate.
5208         * config/s390/s390.md: Disable MVC merging peephole if it would
5209         disable operand forwarding.
5210         (new peephole2): Split MVCs if it would turn them into up to 2
5211         forwardable MVCs.
5213 2017-12-01  Richard Biener  <rguenther@suse.de>
5215         PR tree-optimization/83232
5216         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
5217         detection of same access. Instead of breaking the group here
5218         do not consider the duplicate.  Add comment explaining real fix.
5220 2017-12-01  Jonathan Wakely  <jwakely@redhat.com>
5222         * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
5224 2017-12-01  Sudakshina Das  <sudi.das@arm.com>
5226         * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
5228 2017-12-01  Jakub Jelinek  <jakub@redhat.com>
5230         * function.h (struct function): Remove cilk_frame_decl,
5231         is_cilk_function and calls_cilk_spawn fields.
5232         * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
5233         field.
5234         * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
5235         cilk_elemental field.
5236         * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
5237         * target.def: Adjust comment.
5238         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5239         Don't test cilk_elemental.
5241         PR tree-optimization/83233
5242         * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
5243         bswap_stat name for the struct.
5245         PR c/79153
5246         * tree.h (SWITCH_BREAK_LABEL_P): Define.
5247         * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5248         starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5249         SWITCH_BREAK_LABEL_P set on the label.
5250         (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5251         added for default case if it was missing and not all cases covered.
5252         Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5253         switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5254         set on the label.
5255         * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5256         to avoid -Wimplicit-fallthrough warning.
5257         * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5258         FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5260         PR tree-optimization/83221
5261         * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5262         down by 16.
5263         (init_reassoc): Formatting fix.
5265         PR sanitizer/81275
5266         * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5267         C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5268         is diagnosed.
5270         PR sanitizer/83219
5271         * tree-cfg.c: Include asan.h.
5272         (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5274 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5276         * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5278 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5280         * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5281         vpcmpeqd instruction.
5283 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5285         * config/i386/i386.c (standard_sse_constant_opcode): Fix
5286         registers type for 128bit mode.
5288 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5290         * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5291         of ;;.
5292         * gengtype-state.c (read_state_pair): Likewise.
5293         * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5294         * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5295         * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5296         * ifcvt.c (noce_try_store_flag_constants): Likewise.
5297         * tree-ssa-ccp.c (ccp_finalize): Likewise.
5298         * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5299         * builtins.c (fold_builtin_3): Likewise.
5300         * graphite-scop-detection.c
5301         (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5302         * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5303         Likewise.
5305 2017-12-01  Maxim Ostapenko  <m.ostapenko@samsung.com>
5307         PR sanitizer/81697
5308         * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5309         parameter. Return true if ignore_decl_rtl_set_p is true and other
5310         conditions are satisfied.
5311         * asan.h (asan_protect_global): Add new parameter.
5312         * varasm.c (categorize_decl_for_section): Pass true as second parameter
5313         to asan_protect_global calls.
5315 2017-11-30  Jim Wilson  <jimw@sifive.com>
5317         * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5318         -mno-memcpy options.  For -mplt, -mfdiv, -mdiv, -msave-restore, and
5319         -mstrict-align, add info on default value.  Delete redundant lines for
5320         -mabi.  Add missing -mexplicit-relocs docs.
5322 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5324         * config/arc/arc.md (trap): New pattern.
5326 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5328         * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5329         instruction to end into a delay slot.
5330         * config/arc/arc.md (cond_delay_insn): Check if the instruction
5331         can be placed into a delay slot against reg_note.
5333 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5335         * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5336         labels number of usages.
5338 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5340         * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5341         function.
5342         (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5344 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5346         PR target/83210
5347         * internal-fn.c (expand_mul_overflow): Optimize unsigned
5348         multiplication by power of 2 constant into two shifts + comparison.
5350 2017-11-30  Jan Hubicka  <hubicka@ucw.cz>
5352         PR target/81616
5353         * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5355 2017-11-30  Richard Biener  <rguenther@suse.de>
5357         PR tree-optimization/83202
5358         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5359         allow_peel argument and guard peeling.
5360         (canonicalize_loop_induction_variables): Likewise.
5361         (canonicalize_induction_variables): Pass false.
5362         (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5363         peeling from cunrolli.
5365 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5367         * combine.c (try_combine): Print a message to dump file whenever
5368         I0, I1, or I2 cannot be combined into I3.
5370 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5372         PR rtl-optimization/83156
5373         PR rtl-optimization/82621
5374         * combine.c (try_combine): Don't split an I2 if one of the dests is
5375         set again before I3.  Allow unused dests.
5377 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5379         * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5381 2017-11-29  Vladimir Makarov  <vmakarov@redhat.com>
5383         PR rtl-optimization/80818
5384         * lra.c (collect_non_operand_hard_regs): New arg insn.  Pass it
5385         recursively.  Use insn code for clobber.
5386         (lra_set_insn_recog_data): Pass the new arg to
5387         collect_non_operand_hard_regs.
5388         (add_regs_to_insn_regno_info): Pass insn instead of uid.  Use insn
5389         code for clobber.
5390         (lra_update_insn_regno_info): Pass insn to
5391         add_regs_to_insn_regno_info.
5393 2017-11-29  Jim Wilson  <jimw@sifive.com>
5394             Andrew Waterman  <andrew@sifive.com>
5396         * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5397         (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5398         SINGLE_SHIFT_COST.  Case LT and ZERO_EXTEND, likewise.  Case ASHIFT,
5399         use SINGLE_SHIFT_COST.
5400         * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5401         (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5403 2017-11-29  Julia Koval  <julia.koval@intel.com>
5405         * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5406         _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5407         _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5408         _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5409         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5410         _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5411         _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5412         _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5413         _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5414         _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5415         _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5416         * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5417         __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5418         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5419         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5420         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5421         __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5422         __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5423         __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5424         __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5425         __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5426         __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5427         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5428         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5429         __builtin_ia32_vpshldv_v2di_mask,
5430         __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5431         * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5432         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5434 2017-11-29  Julia Koval  <julia.koval@intel.com>
5436         * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5437         _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5438         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5439         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5440         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5441         _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5442         _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5443         _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5444         _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5445         _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5446         _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5447         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5448         V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5449         V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5450         V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5451         V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5452         V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5453         V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5454         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5455         * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5456         vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5458 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5460         * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5461         movsi_pic_gotdata_op instruction as a load for the UT699 errata
5462         workaround.
5464 2017-11-29  Martin Aberg  <maberg@gaisler.com>
5466         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5467         to prevent b2bst errata sequence.
5468         (sqrtdf2_fix): Likewise.
5470 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5472         * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5473         (div_sqrt_insn_p): New function.
5474         (sparc_do_work_around_errata): Insert NOP instructions to
5475         prevent sequences that could trigger the TN-0013 errata for
5476         certain LEON3 processors.
5477         (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5478         (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5479         * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5480         (in_branch_delay): Prevent div and sqrt in delay slot if
5481         fix_lost_divsqrt.
5482         * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5484 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5486         * config/sparc/sparc.c (atomic_insn_p): New function.
5487         (sparc_do_work_around_errata): Insert NOP instructions to
5488         prevent sequences that could trigger the TN-0010 errata for
5489         UT700.
5490         * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5491         instruction referable in atomic_insns_p.
5493 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5495         * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5496         (atomic_compare_and_swap_leon3_1): Likewise.
5497         (ldstub): Likewise.
5499 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5501         * config/sparc/sparc.c (fpop_insn_p): New function.
5502         (sparc_do_work_around_errata): Insert NOP instructions to
5503         prevent sequences that could trigger the TN-0012 errata for
5504         GR712RC.
5505         (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5506         * config/sparc/sparc.md (fix_gr712rc): New attribute.
5507         (in_branch_annul_delay): Prevent floating-point instructions
5508         in delay slot of annulled integer branch.
5510 2017-11-29  Richard Biener  <rguenther@suse.de>
5512         PR tree-optimization/83202
5513         * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5514         (bst_fail): Use it.
5515         (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5516         nodes vectorized to the same stmts multiple times.
5517         (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5518         (vect_analyze_slp_instance): Adjust.
5519         (scalar_stmts_to_slp_tree_map_t): New typedef.
5520         (vect_schedule_slp_instance): Add a map recording the SLP node
5521         representing the vectorized stmts for a set of scalar stmts.
5522         Avoid code-generating redundancies.
5523         (vect_schedule_slp): Allocate map and pass it down.
5525 2017-11-29  Nathan Sidwell  <nathan@acm.org>
5527         PR c++/83187
5528         * tree.c (build_complex_type): Fix canonicalization.  Only fill in
5529         type if it is new.
5531 2017-11-29  Wilco Dijkstra  <wdijkstr@arm.com>
5533         * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5534         cases for printing LDP/STP memory addresses.
5535         (aarch64_print_address_internal): Renamed from
5536         aarch64_print_operand_address, added parameter, add Pmode check.
5537         (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5538         (aarch64_print_operand_address): Indirect to
5539         aarch64_print_address_internal.
5540         * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5541         'y' operand output specifier.
5543 2017-11-29  Jakub Jelinek  <jakub@redhat.com>
5545         PR middle-end/83185
5546         * tree.c (build_simple_mem_ref_loc): Handle
5547         get_addr_base_and_unit_offset returning a MEM_REF.
5549         PR middle-end/80929
5550         * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5552         PR target/80819
5553         * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5554         alternative.
5556 2017-11-28  David Malcolm  <dmalcolm@redhat.com>
5558         * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5559         m_x_offset.
5560         (layout::move_to_column): Likewise.
5562 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5564         PR sanitizer/81275
5565         * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5566         is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5568 2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5569             Martin Jambor  <mjambor@suse.cz>
5571         PR ipa/82808
5572         * tree.h (expr_type_first_operand_type_p): Declare
5573         * tree.c (expr_type_first_operand_type_p): New function.
5574         * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5575         (ipa_value_from_jfunc): Adjust declaration.
5576         * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5577         Use it as result type for arithmetics, unless it is NULL in which case
5578         be more conservative.
5579         (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5580         ipa_get_jf_pass_through_result.
5581         (propagate_vals_across_pass_through): Likewise.
5582         (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5583         is to propagate_vals_across_pass_through.
5584         (propagate_constants_across_call): Pass PARM_TYPE to
5585         propagate_scalar_across_jump_function.
5586         (find_more_scalar_values_for_callers_subset): Pass parameter type to
5587         ipa_value_from_jfunc.
5588         (cgraph_edge_brings_all_scalars_for_node): Likewise.
5589         * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5590         to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5591         * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5592         target_type, pass it to ipa_value_from_jfunc.
5593         (update_indirect_edges_after_inlining): Pass parameter type to
5594         try_make_edge_direct_simple_call.
5596 2017-11-28  Jeff Law  <law@redhat.com>
5598         * gimple-ssa-evrp-analyze.c
5599         (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5600         refine ranges if scev_initialized_p returns true.
5601         * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5603 2017-11-28  Julia Koval  <julia.koval@intel.com>
5605         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5606         _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5607         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5608         _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5609         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5610         _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5611         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5612         _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5613         _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5614         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5615         _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5616         * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5617         __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5618         __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5619         __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5620         __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5621         __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5622         __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5623         __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5624         __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5625         __builtin_ia32_vpshrd_v2di_mask): New builtins.
5626         * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5628 2017-11-28  Julia Koval  <julia.koval@intel.com>
5630         * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5631         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5632         _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5633         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5634         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5635         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5636         _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5637         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5638         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5639         _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5640         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5641         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5642         V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5643         V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5644         V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5645         V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5646         V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5647         V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5648         V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5649         V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5650         * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5651         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5652         __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5653         __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5654         __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5655         __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5656         __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5657         __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5658         __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5659         __builtin_ia32_vpshld_v2di_mask): New builtins.
5660         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5661         * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5663 2017-11-28  Richard Biener  <rguenther@suse.de>
5665         PR tree-optimization/80776
5666         * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5667         Declare.
5668         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5669         New function.
5670         (evrp_range_analyzer::record_ranges_from_incoming_edges):
5671         If the incoming edge is an effective fallthru because the other
5672         edge only reaches a __builtin_unreachable () then record ranges
5673         derived from the controlling condition in SSA info.
5674         (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5675         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5677 2017-11-28  Olivier Hainque  <hainque@adacore.com>
5679         * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5680         on mingw build hosts.
5682 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5684         PR debug/81307
5685         * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5687 2017-11-28  Richard Biener  <rguenther@suse.de>
5689         PR middle-end/83141
5690         * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5691         copies generated from memcpy use a character array as reference
5692         type.
5694 2017-11-28  Julia Koval  <julia.koval@intel.com>
5695             Sebastian Peryt  <sebastian.peryt@intel.com>
5697         * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5698         c-family/c-cilkplus.o, c-family/array-notation-common.o,
5699         cilk-common.o, cilk.h, cilk-common.c): Remove.
5700         * builtin-types.def
5701         (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5702         * builtins.c (is_builtin_name): Remove cilkplus condition.
5703         (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5704         * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5705         cilk-builtins.def, cilkplus.def): Remove.
5706         * cif-code.def (CILK_SPAWN): Remove.
5707         * cilk-builtins.def: Delete.
5708         * cilk-common.c: Ditto.
5709         * cilk.h: Ditto.
5710         * cilkplus.def: Ditto.
5711         * config/darwin.h (fcilkplus): Delete.
5712         * cppbuiltin.c: Ditto.
5713         * doc/extend.texi: Remove cilkplus doc.
5714         * doc/generic.texi: Ditto.
5715         * doc/invoke.texi: Ditto.
5716         * doc/passes.texi: Ditto.
5717         * gcc.c (fcilkplus): Remove.
5718         * gengtype.c (cilk.h): Remove.
5719         * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5720         support.
5721         * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5722         Remove.
5723         * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5724         gimplify_call_expr,
5725         is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5726         gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5727         cilkplus conditions.
5728         * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5729         inline_read_section): Ditto.
5730         * ipa-inline-analysis.c (cilk.h): Remove.
5731         * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5732         * lto-wrapper.c (merge_and_complain, append_compiler_options,
5733         append_linker_options): Remove condition for fcilkplus.
5734         * lto/lto-lang.c (cilk.h): Remove.
5735         (lto_init): Remove condition for fcilkplus.
5736         * omp-expand.c (expand_cilk_for_call): Delete.
5737         (expand_omp_taskreg, expand_omp_for_static_chunk,
5738         expand_omp_for): Remove cilkplus
5739         conditions.
5740         (expand_cilk_for): Delete.
5741         * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5742         * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5743         execute_lower_omp, diagnose_sb_0): Ditto.
5744         * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5745         * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5746         * tree-nested.c: Ditto.
5747         * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5748         (dump_generic_node): Ditto.
5749         * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5750         * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5751         Delete.
5752         * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5754 2017-11-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5756         * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5757         * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5758         for vector mode and !TARGET_SIMD.
5760 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5762         * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5763         Adjust comment.
5764         * tree.h (SWITCH_LABELS): Remove.
5765         * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5766         assert SWITCH_BODY is non-NULL.
5767         * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5768         handling.
5769         * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5771         PR tree-optimization/80788
5772         * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5773         has TREE_OVERFLOW set, call drop_tree_overflow.
5775 2017-11-28  Richard Biener  <rguenther@suse.de>
5777         PR tree-optimization/83158
5778         * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5780 2017-11-28  Segher Boessenkool  <segher@kernel.crashing.org>
5782         PR 81288/target
5783         * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5784         TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.
5786 2017-11-27  Segher Boessenkool  <segher@kernel.crashing.org>
5788         * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5789         LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5791 2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5793         PR middle_end/82333
5794         * varasm.c (compare_constant): Take the mode of the constants into
5795         account when comparing floating point constants.
5797 2017-11-27  Gerald Pfeifer  <gerald@pfeifer.com>
5799         * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5800         from explicit instantiation of debug_helper.
5801         * vec.h (DEFINE_DEBUG_VEC): Ditto.
5803 2017-11-27  Richard Biener  <rguenther@suse.de>
5805         * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5806         refactor a bit.
5808 2017-11-27  Richard Biener  <rguenther@suse.de>
5810         * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5811         (type_hash_canon): Also clear int_cst_hash_table entry for
5812         TYPE_MIN/MAX_VALUE.
5813         (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5815 2017-11-27  Tamar Christina  <tamar.christina@arm.com>
5817         * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5818         * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5819         * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5821 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5823         * hash-map.h (gt_cleare_cache): Avoid UB.
5825 2017-11-27  Eric Botcazou  <ebotcazou@adacore.com>
5827         * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5828         * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5829         (decide_unroll_runtime_iterations): Likewise.
5830         (decide_unroll_stupid): Likewise.
5832 2017-11-27  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5834         PR target/83109
5835         * config/i386/i386.md: Add a loop with incssp.
5837 2017-11-27  Martin Jambor  <mjambor@suse.cz>
5839         PR tree-optimization/81248
5840         * tree-sra.c (splice_param_accesses): Remove size check.
5841         (decide_one_param_reduction): Fix size check.
5842         * gimple-pretty-print.c (dump_profile): Silence warning.
5843         * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5845 2017-11-27  Jakub Jelinek  <jakub@redhat.com>
5847         PR debug/81307
5848         * dbxout.c (lastlineno): New variable.
5849         (dbx_debug_hooks): Use dbxout_switch_text_section as
5850         switch_text_section debug hook.
5851         (dbxout_function_end): Switch to current_function_section
5852         rather than function_section.  If crtl->has_bb_partition,
5853         output just one N_FUN, depending on in_cold_section_p.
5854         (dbxout_source_line): Remember last lineno in lastlineno.
5855         (dbxout_switch_text_section): New function.
5856         (dbxout_function_decl): Adjust dbxout_block caller.
5857         (dbx_block_with_cold_children): New function.
5858         (dbxout_block): Return true if any LBRAC/RBRAC have been
5859         emitted.  Use dbx_block_with_cold_children at depth == 0
5860         in second partition.  Add PARENT_BLOCKNUM argument, pass
5861         it optionally adjusted to children.  Output LBRAC/RBRAC
5862         around recursive call only if the block is in the current
5863         partition, if not and anything was output, emit empty
5864         range LBRAC/RBRAC.
5865         * final.c (final_scan_insn): Compute cold_function_name
5866         before calling switch_text_section debug hook.  Call
5867         that hook even if dwarf2out_do_frame if not emitting
5868         dwarf debug info.
5870         PR target/83100
5871         * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5872         TREE_READONLY decls.
5874 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5876         PR rtl-optimization/82488
5877         * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5879 2017-11-26  Julia Koval  <julia.koval@intel.com>
5881         * config/i386/i386.c (processor_target_table): Add skylake_cost for
5882         skylake-avx512.
5883         * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5884         skylake_cost): New.
5886 2017-11-26  Julia Koval  <julia.koval@intel.com>
5888         * config/i386/driver-i386.c (host_detect_local_cpu):
5889         Detect skylake-avx512.
5891 2017-11-26  Julia Koval  <julia.koval@intel.com>
5893         * config.gcc: Add -march=cannonlake.
5894         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5895         * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5896         * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5897         (PTA_CANNONLAKE): New.
5898         (processor_target_table): Add cannonlake.
5899         (ix86_option_override_internal): Ditto.
5900         (fold_builtin_cpu): Ditto.
5901         (get_builtin_code_for_version): Handle cannonlake.
5902         (M_INTEL_COREI7_CANNONLAKE): New.
5903         * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5904         * doc/invoke.texi: Add -march=cannonlake.
5906 2017-11-14  Boris Kolpackov  <boris@codesynthesis.com>
5908         * plugin.c (add_new_plugin): Use platform-specific library extensions.
5909         (try_init_one_plugin): Alternative implementation for MinGW.
5910         * Makefile.in (plugin_implib): New.
5911         (gengtype-lex.c): Fix broken AIX workaround.
5912         * configure: Regenerate.
5913         * doc/plugins.texi: Document support for MinGW.
5915 2017-11-25  Jakub Jelinek  <jakub@redhat.com>
5917         PR rtl-optimization/81553
5918         * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
5919         to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
5920         is a shift where C1 has different mode than the whole shift, use C1's
5921         mode for MULT rather than the shift's mode.
5923         PR target/82848
5924         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
5925         builtins not enabled in the currently selected ISA.
5927 2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>
5929         PR tree-optimization/71026
5930         * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
5931         (insert_reciprocals): Change to insert reciprocals before a division
5932         by a square and to insert the square of a reciprocal.
5933         (execute_cse_reciprocals_1): Change to consider division by a square.
5934         (register_division_in): Add importance parameter.
5936 2017-11-24  Richard Biener  <rguenther@suse.de>
5938         PR tree-optimization/82402
5939         * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
5940         set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
5942 2017-11-24  Marc Glisse  <marc.glisse@inria.fr>
5944         * match.pd (0-ptr): New transformation.
5946 2017-11-24  Jan Hubicka  <hubicka@ucw.cz>
5948         PR bootstrap/83015
5949         * ipa-inline.c (inline_small_functions): Set current badnes correctly
5950         when skipping checking.
5952 2017-11-24  Richard Biener  <rguenther@suse.de>
5954         PR tree-optimization/83128
5955         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
5956         (vn_reference_lookup_3): Likewise.
5958 2017-11-24  Jakub Jelinek  <jakub@redhat.com>
5960         PR sanitizer/83014
5961         * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
5962         instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
5963         tree_to_uhwi twice.
5965         * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
5966         ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
5967         check for the latter with gimple_call_builtin_p.  Do not handle
5968         BUILT_IN_STPNCPY_CHK which is not a pass through call.
5970 2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>
5972         * config/arm/arm_neon.h: Fix pragma GCC push_options before
5973         vdot_u32.
5975 2017-11-23  Julia Koval  <julia.koval@intel.com>
5977         * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
5978         _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
5979         _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
5980         _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
5981         _mm512_maskz_expandloadu_epi16): New intrinsics.
5982         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
5983         _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
5984         _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
5985         _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
5986         _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
5987         _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
5988         _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
5989         _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
5990         _mm256_maskz_expandloadu_epi8): New intrinsics.
5991         * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
5992         V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
5993         V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
5994         V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
5995         * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
5996         * config/i386/sse.md (VI248_VLBW): New iterator.
5997         (expand<mode>_mask, expand<mode>_maskz): New patterns.
5999 2017-11-23  Julia Koval  <julia.koval@intel.com>
6001         * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
6002         * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
6003         _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
6004         _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
6005         _mm512_mask_compressstoreu_epi16): New.
6006         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
6007         _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
6008         _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
6009         _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
6010         _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
6011         _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
6012         _mm256_mask_compressstoreu_epi8): New.
6013         * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
6014         VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
6015         VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
6016         VOID_FTYPE_PV8HI_V8HI_UQI): New types.
6017         * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
6018         __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
6019         __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
6020         __builtin_ia32_compresshi128_mask,
6021         __builtin_ia32_compressstoreuqi512_mask,
6022         __builtin_ia32_compressstoreuhi512_mask,
6023         __builtin_ia32_compressstoreuqi256_mask,
6024         __builtin_ia32_compressstoreuqi128_mask,
6025         __builtin_ia32_compressstoreuhi256_mask,
6026         __builtin_ia32_compressstoreuhi128_mask): New builtins.
6027         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
6028         array for flags2.
6029         (ix86_expand_special_args_builtin): Handle new types.
6030         (s4fma_expand): Handle new builtin array.
6031         * config/i386/immintrin.h: Include new headers.
6032         * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
6033         (compress<mode>_mask, compressstore<mode>_mask): New patterns.
6035 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6037         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
6038         fixes.  Declare temp and g variables at the top in order to avoid
6039         {} in most of the cases.
6041 2017-11-23  Marc Glisse  <marc.glisse@inria.fr>
6043         * match.pd (ptr-0): New transformation.
6045 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
6047         * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
6048         (struct addr_mode_cost_table): New.
6049         (struct tune_params): Add field addr_mode_costs.
6050         * config/arm/arm.c (generic_addr_mode_costs): New.
6051         (arm_slowmul_tune): Initialise addr_mode_costs field.
6052         (arm_fastmul_tune): Likewise.
6053         (arm_strongarm_tune): Likewise.
6054         (arm_xscale_tune): Likewise.
6055         (arm_9e_tune): Likewise.
6056         (arm_marvell_pj4_tune): Likewise.
6057         (arm_v6t2_tune): Likewise.
6058         (arm_cortex_tune): Likewise.
6059         (arm_cortex_a8_tune): Likewise.
6060         (arm_cortex_a7_tune): Likewise.
6061         (arm_cortex_a15_tune): Likewise.
6062         (arm_cortex_a35_tune): Likewise.
6063         (arm_cortex_a53_tune): Likewise.
6064         (arm_cortex_a57_tune): Likewise.
6065         (arm_exynosm1_tune): Likewise.
6066         (arm_xgene1_tune): Likewise.
6067         (arm_cortex_a5_tune): Likewise.
6068         (arm_cortex_a9_tune): Likewise.
6069         (arm_cortex_a12_tune): Likewise.
6070         (arm_cortex_a73_tune): Likewise.
6071         (arm_v7m_tune): Likewise.
6072         (arm_cortex_m7_tune): Likewise.
6073         (arm_v6m_tune): Likewise.
6074         (arm_fa726te_tune): Likewise.
6075         (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
6077 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
6079         * config/arm/arm.c (arm_mem_costs): New function.
6080         (arm_rtx_costs_internal): Use arm_mem_costs.
6082 2017-11-23  Mark Wielaard  <mark@klomp.org>
6084         * dwarf2out.c (init_sections_and_labels): Use generation to create
6085         unique ranges_section_label and ranges_base_label. Return generation.
6086         (output_rnglists): Add generation argument. Use generation to create
6087         unique ranges labels.
6088         (dwarf2out_finish): Get generation from init_sections_and_labels
6089         and pass generation to output_rnglists.
6091 2017-11-23  Mike Stump  <mikestump@comcast.net>
6092             Eric Botcazou  <ebotcazou@adacore.com>
6094         * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
6095         * cfgloop.h (struct loop): Add unroll field.
6096         * function.h (struct function): Add has_unroll bitfield.
6097         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
6098         (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
6099         * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
6100         (pass_rtl_unroll_loops::gate): Likewise.
6101         * loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
6102         for which loop->unroll==1.
6103         (decide_unroll_constant_iterations): Use note for consistency and
6104         take loop->unroll into account.  Return early if loop->unroll is set.
6105         Fix thinko in existing test.
6106         (decide_unroll_runtime_iterations): Use note for consistency and
6107         take loop->unroll into account.
6108         (decide_unroll_stupid): Likewise.
6109         * lto-streamer-in.c (input_cfg): Read loop->unroll.
6110         * lto-streamer-out.c (output_cfg): Write loop->unroll.
6111         * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
6112         New case.
6113         (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
6114         (print_loop): Print loop->unroll if set.
6115         * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
6116         * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
6117         * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
6118         New case.
6119         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
6120         loop->unroll is set and smaller than the trip count.  Otherwise bypass
6121         entirely the heuristics if loop->unroll is set.  Remove dead note.
6122         Fix off-by-one bug in other note.
6123         (try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
6124         (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
6125         is greater than 1.
6126         (tree_unroll_loops_completely): Make static.
6127         (pass_complete_unroll::execute): Use correct type for variable.
6128         (pass_complete_unrolli::execute): Fix formatting.
6129         * tree.def (ANNOTATE_EXPR): Add 3rd operand.
6131 2017-11-23  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6133         * config/i386/i386.h (TARGET_PREFER_AVX256): Also
6134         enable when TARGET_PREFER_AVX128 is set.
6136 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6138         * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
6139         frequency.
6140         * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
6141         (cgraph_edge::maybe_hot_p): Use sreal frequencies.
6143 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6145         * ipa-fnsummary.c (record_modified_bb_info): Use sreal
6146         frequencies. Fix estimation of aggregate parameters.
6148 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6150         * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
6151         when inlining.
6153 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6155         * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
6157 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6159         * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
6160         macro.
6162 2017-11-23  Oleg Endo  <olegendo@gcc.gnu.org>
6164         PR target/83111
6165         * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
6166         sibcall_value_pcrel_fdpic): Use local variable instead of
6167         operands[3].
6168         (calli_tbr_rel): Add missing operand 2.
6169         (call_valuei_tbr_rel): Add missing operand 3.
6171 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6173         PR middle-end/82253
6174         * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
6175         bitpos/bitsize covering the whole destination, use store_expr only if
6176         the complex mode is the same.  Otherwise, use expand_normal and if
6177         it returns CONCAT, subreg each part separately instead of trying to
6178         subreg the whole result.
6180 2017-11-23  Richard Biener  <rguenther@suse.de>
6182         PR tree-optimization/23094
6183         * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
6184         come here from walking over backedges in the first iteration.
6185         (vn_reference_lookup_3): Skip clobbers that store the same value.
6187 2017-11-23  Richard Biener  <rguenther@suse.de>
6189         PR tree-optimization/81403
6190         * tree-ssa-pre.c (get_representative_for): Add parameter specifying
6191         a block we need a leader relative to.
6192         (phi_translate_1): For nary processing require a leader from
6193         get_representative_for given we run expression simplification
6194         using match-and-simplify.  Remove previous fix.
6196 2017-11-22  Jeff Law  <law@redhat.com>
6198         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
6199         Use new method allocate_value_range rather than accessing the
6200         vrp_value_range_pool data member directly.
6201         * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
6202         to use extract_range_from_stmt method to avoid need for
6203         extract_range_from_assignment method.
6204         (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
6205         method rather than setting values_propgated data member directly.
6206         * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
6207         and values propagated data members and extract_range_from_assignment
6208         method.  Reorder private data members to conform to standards.
6209         Add new methods set_lattice_propagation_complete and
6210         allocate_value_range.
6212 2017-11-22  Eric Botcazou  <ebotcazou@adacore.com>
6214         PR rtl-optimization/83030
6215         * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
6216         for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
6217         (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
6218         * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
6219         CROSSING_JUMP_P flag is already set before setting it.
6220         * cfgrtl.c (fixup_partition_crossing): Likewise.
6221         * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
6222         insn as useless.
6224 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6226         * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
6227         Handle the case where both arguments are using gen_const_vec_series.
6229 2017-11-22  David Malcolm  <dmalcolm@redhat.com>
6231         PR c++/62170
6232         * pretty-print.c (pp_format): Move quoting implementation to
6233         pp_begin_quote and pp_end_quote.  Update pp_format_decoder call
6234         to pass address of "quote" local.
6235         (pp_begin_quote): New function.
6236         (pp_end_quote): New function.
6237         * pretty-print.h (printer_fn): Convert penultimate param from bool
6238         to bool *.
6239         (pp_begin_quote): New decl.
6240         (pp_end_quote): New decl.
6241         * tree-diagnostic.c (default_tree_printer): Convert penultimate
6242         param from bool to bool *.
6243         * tree-diagnostic.h (default_tree_printer): Likewise.
6245 2017-11-22  Jeff Law  <law@redhat.com>
6247         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6248         Initialize vr_values.
6249         (evrp_range_analyzer::try_find_new_range): Call methods attached to
6250         vr_values via vr_values class instance rather than delegators.
6251         (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6252         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6253         (evrp_range_analyzer::push_value_range): Likewise.
6254         (evrp_range_analyzer::pop_value_range): Likewise.
6255         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6256         most delegators.  Those remaining are exposed as public interfaces.
6257         Make vr_values a pointer and private.
6258         (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6259         vr_values.
6260         (evrp_range_analyzer::get_vr_value): New method.
6261         * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6262         (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6263         (class evrp_dom_walker): Attach evrp_folder class, initialize
6264         it in the ctor.  Remove temporary delegators.
6265         (evrp_dom_walker::before_dom_children): Call methods in attached
6266         evrp_range_analyzer class via class instance pointer.  Use
6267         free value_range_constant_singleton to remove need for
6268         op_with_constant_singleton_value delegator method.  Do not
6269         create a vrp_prop class instance for every call!  Narrow
6270         scope of a couple variables.
6271         (evrp_dom_walker::cleanup): Call methods in attached
6272         evrp_range_analyzer class via class instance pointer.
6273         * vr-values.h (class vr_values): Privatize many methods and
6274         data members.
6276 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6278         * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6279         * cfgexpand.c (expand_debug_expr): Remove handling for them.
6280         * expr.c (expand_expr_real_2): Likewise.
6281         * fold-const.c (const_unop): Likewise.
6282         * optabs-tree.c (optab_for_tree_code): Likewise.
6283         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6284         * tree-inline.c (estimate_operator_cost): Likewise.
6285         * tree-pretty-print.c (dump_generic_node): Likewise.
6286         (op_code_prio): Likewise.
6287         (op_symbol_code): Likewise.
6288         * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6289         (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6290         * internal-fn.c (direct_internal_fn_optab): New function.
6291         (direct_internal_fn_array, direct_internal_fn_supported_p
6292         (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6293         * fold-const-call.c (fold_const_reduction): New function.
6294         (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6295         CFN_REDUC_MIN.
6296         * tree-vect-loop.c: Include internal-fn.h.
6297         (reduction_code_for_scalar_code): Rename to...
6298         (reduction_fn_for_scalar_code): ...this and return an internal
6299         function.
6300         (vect_model_reduction_cost): Take an internal_fn rather than
6301         a tree_code.
6302         (vect_create_epilog_for_reduction): Likewise.  Build calls rather
6303         than assignments.
6304         (vectorizable_reduction): Use internal functions rather than tree
6305         codes for the reduction operation.  Update calls to the functions
6306         above.
6307         * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6308         Use calls to internal functions rather than REDUC tree codes.
6309         * config/aarch64/aarch64-simd.md: Update comment accordingly.
6311 2017-11-22  Olivier Hainque  <hainque@adacore.com>
6313         * config/vxworks.c (vxworks_override_options): Pick default
6314         dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6315         * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6316         DWARF_GNAT_ENCODINGS_DEFAULT.
6317         * config/vxworksae.h: Likewise.
6319 2017-11-22  Marc Glisse  <marc.glisse@inria.fr>
6321         PR tree-optimization/83104
6322         * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6323         not result.
6325 2017-11-22  Marek Polacek  <polacek@redhat.com>
6326             H.J. Lu  <hongjiu.lu@intel.com>
6327             Jason Merrill  <jason@redhat.com>
6329         PR c++/60336
6330         PR middle-end/67239
6331         PR target/68355
6332         * calls.c (initialize_argument_information): Call
6333         warn_parameter_passing_abi target hook.
6334         (store_one_arg): Use 0 for empty record size.  Don't push 0 size
6335         argument onto stack.
6336         (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6337         * common.opt: Update -fabi-version description.
6338         * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6339         (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6340         int_size_in_bytes.
6341         (ix86_is_empty_record): New function.
6342         (ix86_warn_parameter_passing_abi): New function.
6343         (TARGET_EMPTY_RECORD_P): Redefine.
6344         (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6345         * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6346         * doc/tm.texi: Regenerated.
6347         * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6348         TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6349         * dwarf2out.c (get_ultimate_context): Move to tree.c.
6350         * explow.c (hard_function_value): Call arg_int_size_in_bytes
6351         instead of int_size_in_bytes.
6352         * expr.c (copy_blkmode_to_reg): Likewise.
6353         * function.c (aggregate_value_p): Return 0 for empty types.
6354         (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6355         target hook.
6356         (locate_and_pad_parm): Call arg size_in_bytes instead
6357         size_in_bytes.
6358         * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6359         * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6360         * target.def (empty_record_p, warn_parameter_passing_abi): New target
6361         hooks.
6362         * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6363         (std_gimplify_va_arg_expr): Skip empty records.  Call
6364         arg_size_in_bytes instead size_in_bytes.
6365         * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6366         * tree-core.h (tree_type_common): Add empty_flag.
6367         (tree_decl_common): Update comments.
6368         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6369         DECL_PADDING_P.
6370         (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6371         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6372         DECL_PADDING_P.
6373         (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6374         * tree.c (default_is_empty_type): New function.
6375         (default_is_empty_record): New function.
6376         (arg_int_size_in_bytes): New function.
6377         (arg_size_in_bytes): New function.
6378         (get_ultimate_context): New function.
6379         * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6380         TRANSLATION_UNIT_WARN_EMPTY_P.
6381         (default_is_empty_record, arg_int_size_in_bytes,
6382         arg_size_in_bytes, get_ultimate_context): Declare.
6384 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6386         * config/arm/arm.c (cmse_clear_registers): New function.
6387         (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6388         code by call to cmse_clear_registers.
6389         (cmse_nonsecure_entry_clear_before_return): Likewise.
6391 2017-11-22  Tamar Christina  <tamar.christina@arm.com>
6393         * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6394         (vdot_s32, vdotq_s32): New.
6395         (vdot_lane_u32, vdotq_lane_u32): New.
6396         (vdot_lane_s32, vdotq_lane_s32): New.
6399 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6401         PR middle-end/82547
6402         * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6403         for unsigned values with fewer HWIs than the precision.
6404         (test_overflow): New function.
6405         (wide_int_cc_tests): Call it.
6407 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6409         * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6410         has the same mode class as Pmode.
6412 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6414         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6415         padding_bits_to_clear_ptr.
6416         (cmse_nonsecure_entry_clear_before_return): Likewise.
6418 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6420         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6421         auto_sbitap instead of integer bitfield to control register needing
6422         clearing.
6424 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6426         PR tree-optimization/83044
6427         * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6428         INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6429         that need the upper bound.  Subtract offset from
6430         get_addr_base_and_unit_offset only if positive and subtract it
6431         before division by eltsize rather than after it.
6433         PR debug/83084
6434         * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6435         debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6436         (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6437         asm.
6439         PR middle-end/82875
6440         * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6441         expand_binop with *mul_widen_optab, make sure at least one of the
6442         operands doesn't have VOIDmode.
6444         PR debug/83034
6445         * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6447         PR rtl-optimization/82044
6448         PR tree-optimization/82042
6449         * dse.c (record_store): Check for overflow.
6450         (check_mem_read_rtx): Properly check for overflow if width == -1, call
6451         add_wild_read instead of clear_rhs_from_active_local_stores on
6452         overflow and log it into dump_file.
6454 2017-11-22  Richard Biener  <rguenther@suse.de>
6456         * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6457         fake edges to exit when looking for a place to insert.
6458         * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6459         and remove.
6460         (insert_into_preds_of_block): Commit edge insertion immediately,
6461         assert that doesn't require new BBs.
6462         (fini_pre): Release expressions.
6463         (pass_pre::execute): Shuffle things around a bit, if the fn
6464         is too large do not compute AVAIL either as this is really the
6465         quadratic bit.
6467 2017-11-22  Richard Biener  <rguenther@suse.de>
6469         PR tree-optimization/83089
6470         * tree-if-conv.c (pass_if_conversion::execute): If anything
6471         changed reset SCEV and free the number of iteration estimates.
6473 2017-11-21  Martin Sebor  <msebor@redhat.com>
6475         PR tree-optimization/82945
6476         * calls.h (warn_nonstring_bound): Remove unused function.
6478 2017-11-21  Martin Sebor  <msebor@redhat.com>
6480         PR tree-optimization/82945
6481         * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6482         * calls.h (maybe_warn_nonstring_arg): Declare new function.
6483         * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6484         functions.
6485         (initialize_argument_information): Call maybe_warn_nonstring_arg.
6486         * calls.h (get_attr_nonstring_decl): Declare new function.
6487         * doc/extend.texi (attribute nonstring): Update.
6488         * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6489         get_attr_nonstring_decl and handle it.
6490         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same.  Improve
6491         detection of nul-termination.
6492         (strlen_to_stridx): Change to a pointer.
6493         (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6494         (pass_strlen::execute): Same.
6496 2017-11-21  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6498         * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6499         for the new option -mprefer-vector-width=[none|128|256|512].
6500         * config/i386/i386.c (ix86_target_string): remove old style options
6501         -mprefer-avx256 and make -mprefer-avx128 as alias.
6502         (ix86_option_override_internal):  Apply defaults for the
6503         -mprefer-vector-width=[128|256] option.
6504         * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6505         Implement macros to work with -mprefer-vector-width=.
6506         * config/i386/i386.opt: Implemented option
6507         -mprefer-vector-width=[none|128|256|512].
6508         * doc/invoke.texi: Documentation for
6509         -mprefer-vector-width=[none|128|256|512].
6511 2017-11-21  Pat Haugen  <pthaugen@us.ibm.com>
6513         * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6515 2017-11-21  Jonathan Wakely  <jwakely@redhat.com>
6517         * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6519 2017-11-21  Marc Glisse  <marc.glisse@inria.fr>
6521         * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6522         POINTER_PLUS_EXPR.
6523         * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6524         * expr.c (expand_expr_real_2): Likewise.
6525         * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6526         fold_binary_loc): Likewise.
6527         * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6528         P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6529         (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6530         MINUS_EXPR transformations.
6531         * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6532         * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6533         * tree-inline.c (estimate_operator_cost): Likewise.
6534         * tree-pretty-print.c (dump_generic_node, op_code_prio,
6535         op_symbol_code): Likewise.
6536         * tree-vect-stmts.c (vectorizable_operation): Likewise.
6537         * vr-values.c (extract_range_from_binary_expr): Likewise.
6538         * varasm.c (initializer_constant_valid_p_1): Likewise.
6539         * tree.def: New tree code POINTER_DIFF_EXPR.
6541 2017-11-21  Uros Bizjak  <ubizjak@gmail.com>
6543         * config/i386/i386.md (*bswap<mode>2_movbe): Add
6544         integer suffix to movbe mnemonic.
6545         (*bswaphi2_movbe): Ditto.
6546         (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6548 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6550         PR c++/83045
6551         * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6552         Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6553         Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6555 2017-11-21  Martin Liska  <mliska@suse.cz>
6557         * tree-inline.c (expand_call_inline): Remove not needed
6558         xstrdup_for_dump.
6560 2017-11-21  James Cowgill  <James.Cowgill@imgtec.com>
6561             Jakub Jelinek  <jakub@redhat.com>
6563         PR target/82880
6564         * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6565         Remove static keyword from f variable.
6567 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6569         PR tree-optimization/83086
6570         * gimple-ssa-store-merging.c
6571         (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6572         rather than n.base_addr.
6574 2017-11-21  Martin Liska  <mliska@suse.cz>
6576         PR rtl-optimization/82044
6577         PR tree-optimization/82042
6578         * dse.c (check_mem_read_rtx): Check for overflow.
6580 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6582         * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6583         typo in comment.
6585 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6587         * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6588         and its corresponding call arg location note.
6590 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6591             Andrew Burgess  <andrew.burgess@embecosm.com>
6593         * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6594         declaration.
6595         (arc_return_slot_offset): Likewise.
6596         (arc_eh_return_address_location): New declaration.
6597         * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6598         (MUST_SAVE_REGISTER): Add exception handler case.
6599         (MUST_SAVE_RETURN_ADDR): Likewise.
6600         (arc_frame_pointer_required): Likewise.
6601         (arc_frame_pointer_needed): New function.
6602         (arc_compute_frame_size): Changed.
6603         (arc_expand_prologue): Likewise.
6604         (arc_expand_epilogue): Likewise.
6605         (arc_initial_elimination_offset): Likewise.
6606         (arc_return_slot_offset): Delete.
6607         (arc_eh_return_address_location): New function.
6608         (arc_builtin_setjmp_frame_value): Likewise.
6609         * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6610         (EH_RETURN_STACKADJ_RTX): Define.
6611         (EH_RETURN_HANDLER_RTX): Likewise.
6612         * config/arc/arc.md (eh_return): Delete.
6614 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6616         * print-tree.h (debug_vec_tree): Remove prototype.
6617         * gdbinit.in (pvt): Remove macro.
6619 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6621         PR tree-optimization/83047
6622         * gimple-ssa-store-merging.c
6623         (imm_store_chain_info::output_merged_store): If the loads with the
6624         same vuse are in different basic blocks, for load_gsi pick a load
6625         location that is dominated by the other loads.
6627         PR c++/83059
6628         * config/i386/i386.c (ix86_memmodel_check): Start
6629         -Winvalid-memory-model diagnostics with lowercase letter.
6631         PR debug/82718
6632         * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6633         set in_cold_section_p to the partition containing loc_list->first.
6634         When seeing loc_list->last_before_switch node, update secname and
6635         perform range_across_switch second partition handling only after that.
6637         PR debug/82933
6638         * run-rtl-passes.c: Include debug.h.
6639         (run_rtl_passes): Call debug_hooks->assembly_start.
6640         * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6641         multiple times.
6643         PR target/82981
6644         * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6645         OPTAB_DIRECT in calls to expand_simple_binop.
6647 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6649         PR c/81404
6650         * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6652 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6654         PR c++/72786
6655         * spellcheck.h (best_match::blithely_get_best_candidate): New
6656         accessor.
6658 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6660         * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6661         with lower case letter.
6663 2017-11-20  Uros Bizjak  <ubizjak@gmail.com>
6665         * config/i386/i386.md (bswaphi2): New expander.
6666         (*bswaphi2_movbe): New insn pattern.
6667         (bswaphi -> rorhi pepehole2): New peephole pattern.
6669 2017-11-20  Jeff Law  <law@redhat.com>
6671         * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6672         * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6673         * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6674         * gimple-ssa-evrp.c: Remove bits moved into new files.  Include
6675         gimple-ssa-evrp-analyze.h.
6677         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6678         set BB_VISITED here.
6679         (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6681 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6683         * tree-predcom.c: Add general comment on Store-Store chains.
6684         (split_data_refs_to_components): Postpone clearing eliminate_store_p
6685         flag in component.
6686         (get_chain_last_ref_at): Rename into...
6687         (get_chain_last_write_at): ...this.
6688         (get_chain_last_write_before_load): New function.
6689         (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6690         CT_STORE_STORE when write reference is added.
6691         (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6692         (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6693         (initialize_root_vars_store_elim_1): Ditto.
6694         (initialize_root_vars_store_elim_2): Ditto.  Replace rhs once default
6695         definition is created.
6696         (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6697         chain by replacing it with dominant stored value.
6699 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6701         * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6703 2017-11-20  Marc Glisse  <marc.glisse@inria.fr>
6705         * vr-values.c (extract_range_from_binary_expr): Use a full range
6706         for VR_VARYING.
6708 2017-11-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6710         * config/arm/arm.md (R4_REGNUM): Define constant.
6711         (nonsecure_call_internal): Remove r4 clobber.
6712         (nonsecure_call_value_internal): Likewise.
6713         * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6714         clobber and resequence match_operands.
6715         (nonsecure_call_value_reg_thumb1_v5): Likewise.
6716         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6717         (nonsecure_call_value_reg_thumb2): Likewise.
6719 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6721         PR tree-optimization/78821
6722         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6723         if base is TARGET_MEM_REF.  If base is not MEM_REF, set base_addr
6724         to the address of the base rather than the base itself.
6725         (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6726         (find_bswap_or_nop_finalize): New function.
6727         (find_bswap_or_nop): Use it.
6728         (bswap_replace): Return a tree rather than bool, change first
6729         argument from gimple * to gimple_stmt_iterator, allow inserting
6730         into an empty sequence, allow ins_stmt to be NULL - then emit
6731         all stmts into gsi.  Fix up MEM_REF address gimplification.
6732         (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6733         (struct store_immediate_info): Add N and INS_STMT non-static
6734         data members.
6735         (store_immediate_info::store_immediate_info): Initialize them
6736         from newly added ctor args.
6737         (merged_store_group::apply_stores): Formatting fixes.  Sort by
6738         bitpos at the end.
6739         (stmts_may_clobber_ref_p): For stores call also
6740         refs_anti_dependent_p.
6741         (gather_bswap_load_refs): New function.
6742         (imm_store_chain_info::try_coalesce_bswap): New method.
6743         (imm_store_chain_info::coalesce_immediate_stores): Use it.
6744         (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6745         (imm_store_chain_info::output_merged_store): Fail if number of
6746         new estimated stmts is bigger or equal than old.  Handle LROTATE_EXPR
6747         and NOP_EXPR rhs_code.
6748         (pass_store_merging::process_store): Compute n and ins_stmt, if
6749         ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6750         LROTATE_EXPR rhs_code.  Pass n and ins_stmt to store_immediate_info
6751         ctor.
6752         (pass_store_merging::execute): Calculate dominators.
6754         * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6755         BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6756         CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6757         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6758         find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6759         class pass_optimize_bswap, bswap_replace,
6760         pass_optimize_bswap::execute): Moved to ...
6761         * gimple-ssa-store-merging.c: ... this file.
6762         Include optabs-tree.h.
6763         (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6764         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6765         find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6766         anonymous namespace, remove static keywords.
6767         (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6768         (pass_optimize_bswap::execute): ... rather than here.  Formatting fix.
6770 2017-11-20  Jan Hubicka  <hubicka@ucw.cz>
6772         PR bootstrap/83062
6773         * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6775 2017-11-20  Aldy Hernandez  <aldyh@redhat.com>
6777         * vec.h (debug_helper): New function.
6778         (DEFINE_DEBUG_VEC): New macro.
6779         * hash-set.h (debug_helper): New function.
6780         (DEFINE_DEBUG_HASH_SET): New macro.
6781         * cfg.c (debug_slim (edge)): New function.
6782         Call DEFINE_DEBUG_VEC for edges.
6783         Call DEFINE_DEBUG_HASH_SET for edges.
6784         * cfghooks.c (debug_slim (basic_block)): New function.
6785         Call DEFINE_DEBUG_VEC for basic blocks.
6786         Call DEFINE_DEBUG_HASH_SET for basic blocks.
6787         * print-tree.c (debug_slim): New function to handle trees.
6788         Call DEFINE_DEBUG_VEC for trees.
6789         Call DEFINE_DEBUG_HASH_SET for trees.
6790         (debug (vec<tree, va_gc>) &): Remove.
6791         (debug (<vec<tree, va_gc>) *): Remove.
6792         * print-rtl.c (debug_slim): New function to handle const_rtx.
6793         Call DEFINE_DEBUG_VEC for rtx_def.
6794         Call DEFINE_DEBUG_VEC for rtx_insn.
6795         Call DEFINE_DEBUG_HASH_SET for rtx_def.
6796         Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6797         * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6798         (debug (vec<rtx_insn *> *ptr): Remove.
6799         (debug_insn_vector): Remove.
6800         * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6802 2017-11-20  Tom de Vries  <tom@codesourcery.com>
6804         PR rtl-optimization/82020
6805         * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6806         IF_THEN_ELSE condition.
6808 2017-11-19  Jeff Law  <law@redhat.com>
6810         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6811         of degenerates resulting from ignoring an edge.
6813 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6815         PR ipa/81360
6816         * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6818 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6820         PR ipa/83001
6821         * profile-count.c (profile_count::to_sreal_scale): Fix return value
6822         for uninitialied counts.
6824 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6826         PR ipa/60243
6827         * tree-inline.c (estimate_num_insns): Set to 1 at least.
6829 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6831         PR target/82713
6832         * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6834 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6836         * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6837         "do while (0)".
6839 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6841         * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6842         "do {} while (0)".
6844 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6846         * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6847         macro body.
6849 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6851         * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6852         "do {} while (0)".
6853         * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6855 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6857         * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6858         "do {} while (0)".
6859         * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6860         semicolon after MCORE_EXPORT_NAME call.
6862 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6864         PR target/82961
6865         * vmsdbgout.c (vmsdbgout_early_finish): New function.
6866         (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6868 2017-11-18  Jan Hubicka  <hubicka@ucw.cz>
6870         * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6871         (cgraph_edge::clone): Cleanup updating of profile.
6872         * ipa-cp.c (update_profiling_info): Likewise.
6873         * ipa-inline-transform.c (inline_transform): Likewise.
6874         * ipa-inline.c (inline_small_functions): Add missing space to dump.
6875         * ipa-split.c (execute_split_functions): Do not split when function
6876         is cold.
6877         * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6878         * profile-count.c (profile_count::dump): Add global0.
6879         (profile_count::to_cgraph_frequency): Do not ICE when entry is
6880         undefined.
6881         (profile_count::to_sreal_scale): Likewise.
6882         (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6883         (profile_count::combine_with_ipa_count): New function.
6884         * profile-count.h (profile_guessed_global0adjusted): New.
6885         (profile_count::adjusted_zero): New.
6886         (profile_count::global0adjusted): New.
6887         (profile_count::combine_with_ipa_count): New.
6888         * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6889         correct profile of return block of split functions.
6890         (copy_cfg_body): Remove unused profile_count.
6891         (copy_body): Likewise.
6892         (expand_call_inline): Update.
6893         (tree_function_versioning): Update.
6895 2017-11-18  Aldy Hernandez  <aldyh@redhat.com>
6897         * hash-set.h (hash_set::empty): New.
6898         * tree-ssa-threadbackward.h: Delete.
6899         * tree-ssa-threadbackward.c (class thread_jumps): New.
6900         Move max_threaded_paths into class.
6901         (fsm_find_thread_path): Remove arguments that are now in class.
6902         (profitable_jump_thread_path): Rename to...
6903         (thread_jumps::profitable_jump_thread_path): ...this.
6904         (convert_and_register_jump_thread_path): Rename to...
6905         (thread_jumps::convert_and_register_current_path): ...this.
6906         (check_subpath_and_update_thread_path): Rename to...
6907         (thread_jumps::check_subpath_and_update_thread_path): ...this.
6908         (register_jump_thread_path_if_profitable): Rename to...
6909         (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6910         (handle_phi): Rename to...
6911         (thread_jumps::handle_phi): ...this.
6912         (handle_assignment): Rename to...
6913         (thread_jumps::handle_assignment): ...this.
6914         (fsm_find_control_statement_thread_paths): Rename to...
6915         (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6916         (find_jump_threads_backwards): Rename to...
6917         (thread_jumps::find_jump_threads_backwards): ...this.
6918         Initialize path local data.
6919         (pass_thread_jumps::execute): Call find_jump_threads_backwards
6920         from within thread_jumps class.
6921         (pass_early_thread_jumps::execute): Same.
6923 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6925         * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
6927 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6929         * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
6930         to canonical location.
6932 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6934         PR target/81356
6935         * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
6936         Remove.
6937         (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
6939 2017-11-17  H.J. Lu  <hongjiu.lu@intel.com>
6941         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
6942         rcrt1.o%s/grcrt1.o%s for -static-pie.
6944 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6946         * i386.c (ix86_multiplication_cost, ix86_division_cost,
6947         ix86_shift_rotate_cost): Break out from ...
6948         (ix86_rtx_costs): ... here.
6949         (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
6950         vector operations.
6952 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6954         * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
6955         when entry block was promoted unlikely.
6956         (estimate_bb_frequencies): Increase frequency scale.
6957         * profile-count.h (profile_count): Export precision info.
6959 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6961         * tree-tailcall.c (eliminate_tail_call): Be more careful about not
6962         disturbin profile of entry block.
6964 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6966         * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
6967         roundoff errors.
6969 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6971         * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
6972         * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
6974 2017-11-17  Jeff Law  <law@redhat.com>
6976         * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
6977         from evrp_dom_walker class.  Various methods moved into new class.
6978         (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
6979         (evrp_range_analyzer::enter): New method.
6980         (evrp_range_analyzer::leave): New method.
6981         (evrp_dom_walker): Remove delegators no longer needed by this class.
6982         Replace vr_values data member with evrp_range_analyzer
6984         * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
6985         method extracted from evrp_dom_walker::before_dom_children.
6986         (evrp_dom_walker::record_ranges_from_stmt): Likewise.
6987         (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
6989         * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
6990         Add private copy constructor and move assignment operators.
6991         Privatize methods and class data where trivially possible.
6992         (evrp_dom_walker::cleanup): New function, extracted from
6993         execute_early_vrp.  Simplify access to class data.
6995         * vr-values.h (get_output_for_vrp): Prototype.
6996         * vr-values.c (get_output_for_vrp): New function extracted from
6997         vrp_visit_assignment_or_call and extract_range_from_stmt.
6998         (vrp_visit_assignment_or_call): Use get_output_for_vrp.  Simplify.
7000 2017-11-17  Luis Machado  <luis.machado@linaro.org>
7002         * config/aarch64/aarch64.c
7003         (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
7004         (qdf24xx_tunings) <autoprefetcher_model>: Set to
7005         tune_params::AUTOPREFETCHER_WEAK.
7007 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
7009         PR target/82641
7010         * config/arm/arm.c (arm_valid_target_attribute_rec):
7011         Parse "arch=" and "+<ext>".
7012         (arm_valid_target_attribute_tree): Re-init global options.
7013         (arm_option_override): Make non-static.
7014         (arm_options_perform_arch_sanity_checks): Make errors fatal.
7015         * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
7016         (__ARM_FEATURE_CRC32): Support undef.
7017         * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
7018         * doc/extend.texi (ARM Function Attributes): Add pragma and target.
7020 2017-11-17  David Malcolm  <dmalcolm@redhat.com>
7022         * gdbinit.in (break-on-diagnostic): New command.
7024 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7026         * config/i386/i386.c (ix86_expand_epilogue): Change simple
7027         return to indirect jump for EH return if control-flow
7028         protection is enabled. Change explicit 'false' argument in
7029         pro_epilogue_adjust_stack with a value of
7030         flag_cf_protection.
7031         * config/i386/i386.md (simple_return_indirect_internal):
7032         Remove SImode restriction to support 64-bit.
7034 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
7036         * combine.c (added_notes_insn): New.
7037         (try_combine): Handle added_notes_insn like added_links_insn.
7038         Rewrite return value code.
7039         (distribute_notes): Set added_notes_insn to the earliest insn we added
7040         a note to.
7042 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
7044         PR rtl-optimization/82621
7045         * combine.c (try_combine): Do not split PARALLELs of two SETs if the
7046         dest of one of those SETs is unused.
7048 2017-11-17  Richard Biener  <rguenther@suse.de>
7050         PR fortran/83017
7051         * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
7052         * tree-pretty-print.c (dump_generic_node): Handle
7053         annot_expr_parallel_kind.
7054         * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
7055         * gimplify.c (gimple_boolify): Likewise.
7057 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7059         * config.gcc (extra_headers): Add cet.h for x86 targets.
7060         * config/i386/cet.h: New file.
7061         * doc/install.texi: Add --enable-cet/--disable-cet.
7063 2017-11-17  Richard Biener  <rguenther@suse.de>
7065         PR tree-optimization/83017
7066         * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
7067         (gen_parallel_loop): Properly count iterations.
7068         (parallelize_loops): Handle loop->can_be_parallel independent
7069         of flag_loop_parallelize_all.  Make static profitability test match
7070         the runtime one.
7071         * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
7072         * invoke.texi (parloops-min-per-thread): Document.
7074 2017-11-17  Vineet Gupta  <vgupta@synopsys.com>
7076         * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
7077         upstreaming review comments.
7079 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
7081         * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
7082         with fast unaligned access.
7083         * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
7085 2017-11-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
7087         * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
7088         to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
7090 2017-11-17  Richard Biener  <rguenther@suse.de>
7092         * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
7093         folding of references.
7095 2017-11-17  Qing Zhao  <qing.zhao@oracle.com>
7097         PR middle-end/78809
7098         * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
7099         of replacing call to strncmp with corresponding call to strcmp when
7100         meeting conditions.
7102 2017-11-17  Sergey Shalnov  <Sergey.Shalnov@intel.com>
7104         * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
7105         option prefer-avx256 for skylake-avx512 configuration.
7106         * config/i386/i386.c (ix86_option_override_internal): Ditto.
7107         (get_builtin_code_for_version): Ditto.
7109 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7110             Monk Chiang  <sh.chiang04@gmail.com>
7112         * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
7113         (FIXED_REGISTERS): Reserve more register numbers.
7114         (CALL_USED_REGISTERS): Likewise.
7115         (REG_ALLOC_ORDER): Likewise.
7116         (REG_CLASS_CONTENTS): Likewise.
7117         (REGISTER_NAMES): Likewise.
7119 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7120             Kito Cheng  <kito.cheng@gmail.com>
7122         * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
7123         V2SI.
7124         * config/nds32/iterators.md: Add vector mode iterators and attributes.
7126 2017-11-16  Steven Munroe  <munroesj@gcc.gnu.org>
7128         * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
7129         parameter list for vec_splats.
7131 2017-11-16  Joseph Myers  <joseph@codesourcery.com>
7133         * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
7134         date of C17.
7135         (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
7137 2017-11-16  Thomas Koenig  <tkoenig@gcc.gnu.org>
7139         PR bootstrap/82856
7140         * doc/install.texi: Document incompatibility of Perl >=5.6.26
7141         with the required version of automake 1.11.6.
7143 2017-11-16  Pat Haugen  <pthaugen@us.ibm.com>
7145         * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
7146         for it.
7147         (DU_C2_3_power9): Correct reservation combinations.
7148         (FP_DIV_power9, VEC_DIV_power9): New.
7149         (power9-alu): Split out rotate/shift...
7150         (power9-rot): ...to here, correct dispatch resource.
7151         (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
7152         resource.
7153         (power9-fp): Correct latency.
7154         (power9-sdiv): Add div/sqrt resource.
7155         (power9-ddiv): Correct latency, add div/sqrt resource.
7156         (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
7157         (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
7158         resource.
7159         (power9-qpdiv, power9-qpmul): Adjust resource usage.
7161 2017-11-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
7163         * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
7164         switch statement mapping KF built-ins to TF built-ins if we don't
7165         have the proper ISA 3.0 assembler support.
7167 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7169         * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
7170         (lower_emutls_function_body): Do not compute it.
7172 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7174         * ipa-split.c (split_bb_info): Turn time to sreal.
7175         (split_point): Likewise.
7176         (dump_split_point): Likewise.
7177         (fine_split_points): Likewise.
7178         (execute_split_functions): Only zero split_bbs; turn time to sreals.
7180 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7182         * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
7183         in sreal.
7185 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7187         * predict.c (combine_predictions_for_bb): Preserve zero predicted
7188         edges.
7189         (expensive_function_p): Remove useless assert.
7190         (determine_unlikely_bbs): Propagate also forward; determine cold blocks
7192 2017-11-16  Martin Sebor  <msebor@redhat.com>
7194         PR tree-optimization/82588
7195         PR tree-optimization/82583
7196         * tree-vrp.c (check_array_ref): Handle flexible array members,
7197         string literals, and inner indices.
7198         (search_for_addr_array): Add detail to diagnostics.
7200 2017-11-16  Nathan Sidwell  <nathan@acm.org>
7202         PR c++/82836
7203         PR c++/82737
7204         * tree.h (COPY_DECL_RTL): Rename parms for clarity.
7205         (SET_DECL_ASSEMBLER_NAME): Forward to
7206         overwrite_decl_assembler_name.
7207         (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
7208         (overwrite_decl_assembler_name): Declare.
7209         * tree.c (overwrite_decl_assembler_name): New.
7210         * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
7211         (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
7212         (LANG_HOOKS_INITIALIZER): Add it.
7213         * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
7214         * langhooks.c (lhd_set_decl_assembler_name): Use
7215         SET_DECL_ASSEMBLER_NAME.
7216         (lhd_overwrite_decl_assembler_name): Default implementation.
7218 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7219             Jackson Woodruff  <jackson.woodruff@arm.com>
7221         PR tree-optimization/71026
7222         * match.pd: Canonicalize constant multiplies in division.
7224 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7226         * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
7227         * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
7228         Remove OPT_fomit_frame_pointer entry.
7229         * common/config/arc/arc-common.c: Likewise.
7230         * common/config/arm/arm-common.c: Likewise.
7231         * common/config/avr/avr-common.c: Likewise.
7232         * common/config/c6x/c6x-common.c: Likewise.
7233         * common/config/cr16/cr16-common.c: Likewise.
7234         * common/config/cris/cris-common.c: Likewise.
7235         * common/config/epiphany/epiphany-common.c: Likewise.
7236         * common/config/fr30/fr30-common.c: Likewise.
7237         * common/config/frv/frv-common.c: Likewise.
7238         * common/config/ia64/ia64-common.c: Likewise.
7239         * common/config/iq2000/iq2000-common.c: Likewise.
7240         * common/config/lm32/lm32-common.c: Likewise.
7241         * common/config/m32r/m32r-common.c: Likewise.
7242         * common/config/mcore/mcore-common.c: Likewise.
7243         * common/config/microblaze/microblaze-common.c: Likewise.
7244         * common/config/mips/mips-common.c: Likewise.
7245         * common/config/mmix/mmix-common.c: Likewise.
7246         * common/config/mn10300/mn10300-common.c: Likewise.
7247         * common/config/nios2/nios2-common.c: Likewise.
7248         * common/config/pa/pa-common.c: Likewise.
7249         * common/config/pdp11/pdp11-common.c: Likewise.
7250         * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7251         * common/config/riscv/riscv-common.c: Likewise.
7252         * common/config/rs6000/rs6000-common.c: Likewise.
7253         * common/config/rx/rx-common.c: Likewise.
7254         * common/config/s390/s390-common.c: Likewise.
7255         * common/config/sh/sh-common.c: Likewise.
7256         * common/config/sparc/sparc-common.c: Likewise.
7257         * common/config/tilegx/tilegx-common.c: Likewise.
7258         * common/config/tilepro/tilepro-common.c: Likewise.
7259         * common/config/v850/v850-common.c: Likewise.
7260         * common/config/visium/visium-common.c: Likewise.
7261         * common/config/xstormy16/xstormy16-common.c: Likewise.
7262         * common/config/xtensa/xtensa-common.c: Likewise.
7263         * invoke.texi (-fomit-frame-pointer): Update documentation.
7265 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7267         * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7269 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7271         * cfg.c (scale_bbs_frequencies_int,
7272         cale_bbs_frequencies_gcov_type): Remove.
7273         * cfg.h (scale_bbs_frequencies_int,
7274         cale_bbs_frequencies_gcov_type): Remove.
7276 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7278         * tree-ssa-loop-manip.c
7279         (scale_dominated_blocks_in_loop): Update to profile counts.
7280         (tree_transform_and_unroll_loop): Likewise.
7282 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7284         * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7285         scale_bbs_frequencies_int.
7287 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7289         * final.c (compute_alignments): Use counts rather than frequencies.
7291 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7293         * cfgloopanal.c: Include sreal.h
7294         (average_num_loop_insns): Use counts and sreal for accounting.
7296 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7298         * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7299         manipulation.
7301 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7302             Kito Cheng  <kito.cheng@gmail.com>
7304         * config/nds32/constraints.md: Provide more constraints.
7305         * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7306         REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7307         support constraints usage.
7309 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7311         * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7313 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7314             Kito Cheng  <kito.cheng@gmail.com>
7316         * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7317         * config/nds32/nds32.opt: Refine the layout.
7318         * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7319         TARGET_EXT_STRING): Support new options.
7320         * config/nds32/nds32.h: Likewise.
7321         * config/nds32/nds32.md: Likewise.
7322         * config/nds32/nds32-predicates.c: Likewise.
7323         * config/nds32/constraints.md: Likewise.
7324         * common/config/nds32/nds32-common.c: Likewise.
7326 2017-11-16  Julia Koval  <julia.koval@intel.com>
7328         PR target/82983
7329         * config/i386/gfniintrin.h: Add sse check.
7330         * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7332 2017-11-16  Julia Koval  <julia.koval@intel.com>
7334         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7335         OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7336         (ix86_handle_option): Handle -mavx512vbmi2.
7337         * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7338         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7339         * config/i386/i386-c.c (__AVX512VBMI2__): New.
7340         * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7341         (ix86_valid_target_attribute_inner_p): Ditto.
7342         * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7343         * config/i386/i386.opt (mavx512vbmi2): New option.
7344         * doc/invoke.texi: Add new option.
7346 2017-11-16  Julia Koval  <julia.koval@intel.com>
7348         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7349         _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7350         _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7351         _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7352         _mm512_gf2p8mul_epi8): New intrinsics.
7353         * config/i386/i386-builtin-types.def
7354         (V64QI_FTYPE_V64QI_V64QI): New type.
7355         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7356         __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7357         __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7358         __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7359         * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7360         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7362 2017-11-15  Uros Bizjak  <ubizjak@gmail.com>
7364         * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7365         explicitly as a stream of bytes.
7367 2017-11-15  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7369         * config/rs6000/altivec.h (vec_xst_be): New #define.
7370         * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7371         and externalize from *altivec_vperm_<mode>_internal.
7372         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7373         instantiation.
7374         (XL_BE_V8HI): Likewise.
7375         (XL_BE_V4SI): Likewise.
7376         (XL_BE_V4SI): Likewise.
7377         (XL_BE_V2DI): Likewise.
7378         (XL_BE_V4SF): Likewise.
7379         (XL_BE_V2DF): Likewise.
7380         (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7381         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7382         all array entries with these keys: VSX_BUILTIN_VEC_XL,
7383         VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST.  Add entries for key
7384         VSX_BUILTIN_VEC_XST_BE.
7385         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7386         (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7387         built-ins.
7388         (altivec_init_builtins): Replace conditional calls to def_builtin
7389         for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7390         __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7391         with unconditional calls.  Remove calls to def_builtin for
7392         __builtin_vsx_le_be_<mode>.  Add a call to def_builtin for
7393         __builtin_vec_xst_be.
7394         * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7395         to define_expand, and add alternate RTL generation for P8.
7396         (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7397         vsx_ld_elemrev_v8hi.
7398         (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7399         add alternate RTL generation for P8.
7400         (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7401         vsx_ld_elemrev_v16qi.
7402         (vsx_st_elemrev_v8hi): Convert define_insn
7403         to define_expand, and add alternate RTL generation for P8.
7404         (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7405         vsx_st_elemrev_v8hi.
7406         (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7407         add alternate RTL generation for P8.
7408         (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7409         vsx_st_elemrev_v16qi.
7411 2017-11-15  H.J. Lu  <hongjiu.lu@intel.com>
7413         PR target/82990
7414         * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7415         TARGET_AVX512ER check.
7416         (ix86_option_override_internal): Set MASK_VZEROUPPER if
7417         neither -mzeroupper nor -mno-zeroupper is used and
7418         TARGET_EMIT_VZEROUPPER is set.
7419         * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7420         * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7422 2017-11-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7424         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7425         folding of vector compares.
7426         (fold_build_vec_cmp): New helper function.
7427         (fold_compare_helper): New helper function.
7428         (builtin_function_type): Add compare builtins to the list of functions
7429         having unsigned arguments.  Cosmetic updates to comment indentation.
7430         * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7431         the not+eq combination.
7433 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7435         PR tree-optimization/82726
7436         PR tree-optimization/70754
7437         * tree-predcom.c (order_drefs_by_pos): New function.
7438         (combine_chains): Move code setting has_max_use_after to...
7439         (try_combine_chains): ...here.  New parameter.  Sort combined chains
7440         according to position information.
7441         (tree_predictive_commoning_loop): Update call to above function.
7442         (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7444 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7446         PR tree-optimization/82726
7447         Revert
7448         2017-01-23  Bin Cheng  <bin.cheng@arm.com>
7450         PR tree-optimization/70754
7451         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7452         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
7453         combined stmt before it if not NULL.
7454         (combine_chains): Process refs reversely and compute dominance point
7455         for root ref.
7457         Revert
7458         2017-02-23  Bin Cheng  <bin.cheng@arm.com>
7460         PR tree-optimization/79663
7461         * tree-predcom.c (combine_chains): Process refs in reverse order
7462         only for ZERO length chains, and add explaining comment.
7464 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7466         * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7467         * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7468         * doc/invoke.texi (ARM Options): Add armv8.3-a.
7470 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7472         * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7474 2017-11-15  Martin Liska  <mliska@suse.cz>
7476         * tree-cfg.c (pass_warn_function_return::execute):
7477         Compare warn_return_type for greater than zero.
7479 2017-11-15  Sebastian Peryt  <sebastian.peryt@intel.com>
7481         PR target/82941
7482         PR target/82942
7483         * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7484         to return true on Xeon and not on Xeon Phi.
7485         (ix86_check_avx256_register): Changed to ...
7486         (ix86_check_avx_upper_register): ... this. Add extra check for
7487         VALID_AVX512F_REG_OR_XI_MODE.
7488         (ix86_avx_u128_mode_needed): Changed
7489         ix86_check_avx256_register to ix86_check_avx_upper_register.
7490         (ix86_check_avx256_stores): Changed to ...
7491         (ix86_check_avx_upper_stores): ... this. Changed
7492         ix86_check_avx256_register to ix86_check_avx_upper_register.
7493         (ix86_avx_u128_mode_after): Changed
7494         avx_reg256_found to avx_upper_reg_found. Changed
7495         ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7496         (ix86_avx_u128_mode_entry): Changed
7497         ix86_check_avx256_register to ix86_check_avx_upper_register.
7498         (ix86_avx_u128_mode_exit): Ditto.
7499         * config/i386/i386.h: (host_detect_local_cpu): New define.
7501 2017-11-15  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
7503         * config/arm/xgene1.md (xgene1): Split into automatons
7504         xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7505         (xgene1_f_load): Adjust reservations and/or types.
7506         (xgene1_f_store): Likewise.
7507         (xgene1_load_pair): Likewise.
7508         (xgene1_store_pair): Likewise.
7509         (xgene1_fp_load1): Likewise.
7510         (xgene1_load1): Likewise.
7511         (xgene1_store1): Likewise.
7512         (xgene1_move): Likewise.
7513         (xgene1_alu): Likewise.
7514         (xgene1_simd): Likewise.
7515         (xgene1_bfm): Likewise.
7516         (xgene1_neon_load1): Likewise.
7517         (xgene1_neon_store1): Likewise.
7518         (xgene1_neon_logic): Likewise.
7519         (xgene1_neon_st1): Likewise.
7520         (xgene1_neon_ld1r): Likewise.
7521         (xgene1_alu_cond): Added.
7522         (xgene1_shift_reg): Likwise.
7523         (xgene1_bfx): Likewise.
7524         (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7526 2017-11-15  Jakub Jelinek  <jakub@redhat.com>
7528         PR target/82981
7529         * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7530         ssa-iterators.h.
7531         (can_widen_mult_without_libcall): New function.
7532         (expand_mul_overflow): If only checking unsigned mul overflow,
7533         not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7534         Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7535         no other way works.  Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7536         (expand_DIVMOD): Formatting fix.
7537         * expmed.h (expand_mult): Add NO_LIBCALL argument.
7538         * expmed.c (expand_mult): Likewise.  Use OPTAB_WIDEN rather
7539         than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7541         PR tree-optimization/82977
7542         * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7543         constructed temporary to strlen_to_stridx.put.
7545 2017-11-15  Martin Liska  <mliska@suse.cz>
7547         * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7548         * configure: Regenerate.
7550 2017-11-15  Martin Liska  <mliska@suse.cz>
7552         PR target/82927
7553         * config/sh/sh-mem.cc: Use proper probability for
7554         REG_BR_PROB_NOTE.
7556 2017-11-14  Jeff Law  <law@redhat.com>
7558         * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7559         the red zone for stack_clash_protection_final_dynamic_probe targets
7560         when the total dynamic stack size is zero bytes.
7562         * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7563         blocks is post order.
7565 2017-11-15  Alexandre Oliva  <aoliva@redhat.com>
7567         * dumpfile.h (TDF_COMPARE_DEBUG): New.
7568         * final.c (rest_of_clean_state): Set it for the
7569         -fcompare-debug dump.
7570         * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7571         class when TDF_COMPARE_DEBUG is set.
7573         * dwarf2out.c (gen_producer_string): Discard
7574         OPT_fcompare_debug.
7576 2017-11-15  Joseph Myers  <joseph@codesourcery.com>
7578         PR c/81156
7579         * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7580         * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7581         (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7582         (__TGMATH_REAL_2_3): Remove macros.
7583         (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7584         (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7585         __builtin_tgmath.
7586         (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7587         __TGMATH_REAL_2.
7588         (remquo): Define using __TGMATH_REAL_3.
7590 2017-11-14  Jeff Law  <law@redhat.com>
7592         * vr-values.c: New file with contents extracted from tree-vrp.c.
7593         * Makefile.in (OBJS): Add vr-values.o
7594         * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7595         (set_value_range, set_and_canonicalize_value_range): Likewise.
7596         (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7597         (value_range_constant_singleton, symbolic_range_p): Likewise.
7598         (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7599         (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7600         (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7601         (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7602         (find_case_label_range, find_case_label_index): Likewise.
7603         (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7604         (range_int_cst_singleton_p, value_inside_range): Likewise.
7605         (get_single_symbol): Likewise.
7606         (switch_update): Move structure definition here.
7607         (to_remove_edges, to_update_switch_stmts): Provide externs.
7608         * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7609         (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7610         (vrp_val_is_min, set_value_range): Likewise.
7611         (set_and_canonicalize_value_range, copy_value_range): Likewise.
7612         (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7613         (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7614         (range_is_nonnull, range_int_cst_p): Likewwise.
7615         (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7616         (get_single_symbol, operand_less_p): Likewise
7617         (compare_values_warnv, compare_values): Likewise.
7618         (value_inside_range, value_range_constant_singleton): Likewise.
7619         (zero_nonzero_bitgs_from_vr): Likewise.
7620         (extract_range_from_binary_expr_1): Likewise.
7621         (overflow_comparison_p): Likewise.
7622         (to_remove_edges, to_update_switch_stmts): Likewise.
7623         (find_case_label-index, find_case_label_range): Likewise.
7624         (switch_update, set_value_range_to_nonnegative): Remove.
7625         (set_value_range_to_truthvalue): Likewise.
7626         (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7627         (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7628         (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7629         (find_case_label_ranges, test_for_singularity): Likewise.
7630         (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7631         (x_vr_values): Move to its remaining use site.
7633 2017-11-10  Jeff Law  <law@redhat.com>
7635         * vr-values.h (VR_INITIALIZER): Move #define here.
7636         * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7637         * Makefile.in (OBJS): Add tree-evrp.o
7638         * tree-vrp.h (assert_info): Move structure definition here.
7639         (set_value_range_to_varying): Prototype.
7640         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7641         (infer_value_range, register_edge_assert_for): Likewise.
7642         (stmt_interesting_for_vrp): Likewise.
7643         * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7644         (set_value_range_to_varying): No longer static.
7645         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7646         (infer_value_range, register_edge_assert_for): Likewise.
7648 2017-11-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
7650         * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7651         generate the XXBRD instruction.
7653         * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7654         (rs6000_builtin_type_compatible): Treat _Float128 and long double
7655         as being compatible if -mabi=ieeelongdouble.
7656         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7657         to setup float128 built-ins with hardware support.
7658         (BU_FLOAT128_HW_2): Likewise.
7659         (BU_FLOAT128_HW_3): Likewise.
7660         (BU_FLOAT128_HW_VSX_1): Likewise.
7661         (BU_FLOAT128_HW_VSX_2): Likewise.
7662         (scalar_extract_expq): Change float128 built-in functions to
7663         accommodate having both KFmode and TFmode functions.  Use the
7664         KFmode variant as the default.
7665         (scalar_extract_sigq): Likewise.
7666         (scalar_test_neg_qp): Likewise.
7667         (scalar_insert_exp_q): Likewise.
7668         (scalar_insert_exp_qp): Likewise.
7669         (scalar_test_data_class_qp): Likewise.
7670         (sqrtf128_round_to_odd): Delete processing the round to odd
7671         built-in functions as special built-in functions, and define them
7672         as float128 built-ins.  Use the KFmode variant as the default.
7673         (truncf128_round_to_odd): Likewise.
7674         (addf128_round_to_odd): Likewise.
7675         (subf128_round_to_odd): Likewise.
7676         (mulf128_round_to_odd): Likewise.
7677         (divf128_round_to_odd): Likewise.
7678         (fmaf128_round_to_odd): Likewise.
7679         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7680         support for KFmode and TFmode xststdcqp calls.
7681         (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7682         point, switch the built-in handlers for the get/set float128
7683         exponent, get float128 mantissa, float128 test built-ins, and the
7684         float128 round to odd built-in functions.  Eliminate creating the
7685         float128 round to odd built-in functions as special built-ins.
7686         (rs6000_init_builtins): Eliminate special creation of the float128
7687         round to odd built-in functions.
7688         * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7689         function insns to support both TFmode and KFmode variants.
7690         (xsxsigqp_<mode>): Likewise.
7691         (xsiexpqpf_<mode>): Likewise.
7692         (xsiexpqp_<mode>): Likewise.
7693         (xststdcqp_<mode>): Likewise.
7694         (xststdcnegqp_<mode>): Likewise.
7695         (xststdcqp_<mode>): Likewise.
7697 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7699         * tree-ssa-threadupdate.c (compute_path_counts): Remove
7700         unused path_in_freq_ptr parameter.
7701         (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7703 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7705         * ipa-inline.c (edge_badness): Dump sreal frequency.
7706         (compute_inlined_call_time): Match natural implementaiton ...
7707         * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7708         forgotten division by CGRAPH_FREQ_BASE.
7710 2017-11-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7712         * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7713         Solaris 11.  Update comment.
7714         * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7715         renaming.
7716         * config/sol2.h (STARTFILE_SPEC): Likewise.
7717         * configure: Regenerate.
7719 2017-11-14  Carl Love  <cel@us.ibm.com>
7721         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7722         le_ and be_ prefixes to swap* variables.  Remove
7723         if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7724         statements.
7726 2017-11-14  Jason Merrill  <jason@redhat.com>
7728         Support GTY((cache)) on hash_map.
7729         * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7730         (ggc_cache_remove): Override it instead of ggc_mx.
7731         * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7732         (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7733         * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7734         (simple_cache_map_traits): Override maybe_mx.
7735         * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7736         (hash_map): Friend gt_cleare_cache.
7737         (gt_cleare_cache): New.
7738         * tree.h (tree_cache_traits): New hash_map traits class.
7739         (tree_cache_map): New typedef.
7741 2017-11-14  Richard Biener  <rguenther@suse.de>
7743         * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7744         paramter and handling.
7745         (cleanup_control_flow_bb): Likewise.
7746         (cleanup_control_flow_pre): New helper performing a DFS walk
7747         to call cleanup_control_flow_bb in PRE order.
7748         (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7749         via cleanup_control_flow_pre.
7751 2017-11-14  James Greenhalgh  <james.greenhalgh@arm.com>
7753         * config/aarch64/aarch64-simd.md
7754         (aarch64_simd_bsl<mode>_internal): Remove DImode.
7755         (*aarch64_simd_bsl<mode>_alt): Likewise.
7756         (aarch64_simd_bsldi_internal): New.
7757         (aarch64_simd_bsldi_alt): Likewise.
7759 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7761         * tracer.c (better_p): Do not compare frequencies.
7762         * reg-stack.c (better_edge): Likewise.
7763         * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7764         and back.
7766 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7768         * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7769         * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7770         * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7771         * lto-streamer-in.c (input_function): Use update_max_bb_count.
7772         * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7773         * predict.c (maybe_hot_frequency_p): Inline to ...
7774         (maybe_hot_count_p): ... here; rewrite to counts.
7775         (counts_to_freqs): Rename to ...
7776         (update_max_bb_count): ... this one.
7777         (expensive_function_p): Use counts.
7778         (estimate_bb_frequencies): Update.
7779         (rebuild_frequencies): Update.
7780         * predict.h (counts_to_freqs): Rename to ...
7781         (update_max_bb_count): ... this one.
7782         * profile.c (compute_branch_probabilities): Add debug info
7783         * tree-inline.c (expand_call_inline): Update debug info.
7784         (optimize_inline_calls): Use update_max_bb_count..
7785         (tree_function_versioning): Use update_max_bb_count..
7786         * value-prof.c (gimple_value_profile_transformations):
7787         Do not use update_max_bb_count.
7789 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7791         * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7792         always use frequencies.
7794 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7796         * bb-reorder.c: Remove frequencies from comments.
7797         (better_edge_p): Use profile counts.
7798         (find_traces): Dump profile counts.
7799         (rotate_loop): Use profile counts.
7800         (find_traces_1_round): Likewise.
7801         (connect_better_edge_p): Use counts instead of probabilities for
7802         reverse walk.
7803         (copy_bb_p): Drop early check for non-0 frequency.
7804         (sanitize_hot_paths): Update comments.
7806 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7808         * ipa-split.c (struct split_point): Add count.
7809         (consider_split): Do not compute incoming frequency; compute incoming
7810         count and store it to split_point.
7811         (split_function): Set count of the call to split part correctly.
7813 2017-11-13  Carl Love  <cel@us.ibm.com>
7815         * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7817 2017-11-13  Tom Tromey  <tom@tromey.com>
7819         * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7821 2017-11-13  Carl Love  <cel@us.ibm.com>
7823         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7824         Add support for builtins:
7825         unsigned int vec_first_{,miss}_match_{,or_eos}index,
7826         vector {un,}signed {char,int,short},
7827         vector {un,}signed {char,int,short}) arguments.
7828         * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7829         VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7830         Add BU_P9V_AV_2 expansions for the builtins.
7831         * config/rs6000/altivec.h (vec_first_match_index,
7832         vec_first_mismatch_index, vec_first_match_or_eos_index,
7833         vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7834         * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7835         new extern declaration.
7836         * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7837         * config/rs6000/vsx.md (first_match_index_<mode>,
7838         first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7839         first_mismatch_or_eos_index_<mode>): Add define expand.
7840         (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7841         * doc/extend.texi: Update the built-in documenation file for the new
7842         built-in functions.
7844 2017-11-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
7846         * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7847         into the min/max operations for _Float<N> and _Float<N>X types.
7849 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7851         PR lto/81351
7852         * dwarf2out.c (do_eh_frame): New static variable.
7853         (dwarf2out_begin_prologue): Set it.
7854         (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7856 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7858         * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7860         * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7862 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7864         PR lto/81351
7865         * debug.h (dwarf2out_do_eh_frame): Declare.
7866         * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7867         (dwarf2out_do_frame): Use it.
7868         (dwarf2out_do_cfi_asm): Likewise.
7869         * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7870         (dwarf2out_assembly_start): Likewise.
7871         (dwarf2out_begin_prologue): Fix comment.
7872         * toplev.c (compile_file): Always call dwarf2out_frame_finish
7873         if the target needs either debug or unwind DWARF2 info.
7874         * lto-opts.c (lto_write_options): Do not save -fexceptions,
7875         -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7876         -ftrapping-math, -ftrapv and -fwrapv.
7878 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7880         * cgraph.c (cgraph_edge::sreal_frequency): New function.
7881         * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7882         * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7883         (estimate_edge_size_and_time): Likewise.
7884         (ipa_merge_fn_summary_after_inlining): Likewise.
7885         * ipa-inline.c (cgraph_freq_base_rec): Remove.
7886         (compute_uninlined_call_time): Use sreal_frequency.
7887         (compute_inlined_call_time): Likewise.
7888         (ipa_inline): Do not initialize cgraph_freq_base_rec.
7889         * profile-count.c: Include sreal.h.
7890         (profile_count::to_sreal_scale): New.
7891         * profile-count.h: Forward declare sreal.
7892         (profile_count::to_sreal_scale): Declare.
7894 2017-11-13  Nathan Sidwell  <nathan@acm.org>
7896         * diagnostic.c (maybe_line_and_column): New.
7897         (diagnostic_get_location_text): Use it.
7898         (diagnostic_report_current_module): Likewise.
7899         (test_diagnostic_get_location_text): Add tests.
7901 2017-11-13  Luis Machado  <luis.machado@linaro.org>
7903         * doc/md.texi (Specifying processor pipeline description): Fix
7904         incorrect latency for the div instruction example.
7906 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7908         PR tree-optimization/78821
7909         * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7910         that bit_not_p is the same.
7911         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7912         (split_group): Count precisely bit_not_p bits in each statement.
7913         (invert_op): New function.
7914         (imm_store_chain_info::output_merged_store): Use invert_op to
7915         emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7916         but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
7918 2017-11-13  Martin Liska  <mliska@suse.cz>
7920         * gcov.c (struct coverage_info): Remove typedef of coverage_t.
7921         (struct source_info): Likewise.
7922         (add_branch_counts): Likewise.
7923         (add_line_counts): Likewise.
7924         (function_summary): Likewise.
7925         (output_intermediate_line): Likewise.
7926         (generate_results): Likewise.
7928 2017-11-13  Martin Liska  <mliska@suse.cz>
7930         * gcov.c (struct block_info): Remove typedef for block_t.
7931         (struct line_info): Likewise.
7932         (line_info::has_block): Likewise.
7933         (EXIT_BLOCK): Likewise.
7934         (unblock): Likewise.
7935         (circuit): Likewise.
7936         (get_cycles_count): Likewise.
7937         (process_file): Likewise.
7938         (read_graph_file): Likewise.
7939         (solve_flow_graph): Likewise.
7940         (find_exception_blocks): Likewise.
7941         (add_line_counts): Likewise.
7942         (accumulate_line_info): Likewise.
7943         (output_line_details): Likewise.
7945 2017-11-13  Martin Liska  <mliska@suse.cz>
7947         * gcov.c (struct arc_info): Remove typedef for arc_t.
7948         (struct line_info): Likewise.
7949         (add_branch_counts): Likewise.
7950         (output_branch_count): Likewise.
7951         (function_info::~function_info): Likewise.
7952         (circuit): Likewise.
7953         (output_intermediate_line): Likewise.
7954         (read_graph_file): Likewise.
7955         (solve_flow_graph): Likewise.
7956         (find_exception_blocks): Likewise.
7957         (add_line_counts): Likewise.
7958         (accumulate_line_info): Likewise.
7959         (output_line_details): Likewise.
7960         (output_function_details): Likewise.
7962 2017-11-13  Martin Liska  <mliska@suse.cz>
7964         * gcov.c (struct function_info): Remove typedef for function_t.
7965         (struct source_info): Likewise.
7966         (source_info::get_functions_at_location): Likewise.
7967         (solve_flow_graph): Likewise.
7968         (find_exception_blocks): Likewise.
7969         (add_line_counts): Likewise.
7970         (output_intermediate_file): Likewise.
7971         (process_file): Likewise.
7972         (generate_results): Likewise.
7973         (release_structures): Likewise.
7974         (read_graph_file): Likewise.
7975         (read_count_file): Likewise.
7976         (accumulate_line_counts): Likewise.
7977         (output_lines): Likewise.
7979 2017-11-13  Martin Liska  <mliska@suse.cz>
7981         * gcov.c (function_info::function_info): Remove num_counts
7982         and add vector<gcov_type>.
7983         (function_info::~function_info): Use the vector.
7984         (process_file): Likewise.
7985         (read_graph_file): Likewise.
7986         (read_count_file): Likewise.
7987         (solve_flow_graph): Likewise.
7989 2017-11-13  Martin Liska  <mliska@suse.cz>
7991         * gcov.c (function_info::is_artificial): New function.
7992         (process_file): Erase all artificial early.
7993         (generate_results): Skip as all artificial are already
7994         removed.
7996 2017-11-13  Martin Liska  <mliska@suse.cz>
7998         * gcov.c (read_graph_file): Store to global vector of functions.
7999         (read_count_file): Iterate the vector.
8000         (process_file): Likewise.
8001         (generate_results): Likewise.
8002         (release_structures): Likewise.
8004 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
8006         PR tree-optimization/82954
8007         * gimple-ssa-store-merging.c
8008         (imm_store_chain_info::coalesce_immediate_stores): If
8009         !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
8011 2017-11-13  Richard Sandiford  <richard.sandiford@linaro.org>
8013         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
8014         Upddate call to ENDIAN_LANE_N.
8015         (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
8016         (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
8017         (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
8018         and use aarch64_endian_lane_rtx.
8019         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
8021 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8023         * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
8025 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8027         * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
8029 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8031         * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
8032         body.
8033         * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
8034         * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
8035         * defaults.h (ASM_OUTPUT_LABELREF): Same.
8037 2017-11-11  Martin Sebor  <msebor@redhat.com>
8039         PR c/81117
8040         * doc/extend.texi (attribute nonstring): Remove spurious argument.
8042         PR bootstrap/82948
8043         * prefic.c (translate_name): Replace strncpy with memcpy to
8044         avoid -Wstringop-truncation.
8046 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8048         * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
8050 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8052         * predict.c (maybe_hot_frequency_p): Do not use cfun.
8054 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8056         * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
8057         merging.
8059 2017-11-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
8061         * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
8062         enable generating XXBRH if the value is in a vector register.
8063         (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
8064         value is in a vector register.
8065         (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
8066         register to register bswap64's instead of doing the GPR sequence
8067         used on previous machines.
8068         (bswapdi2_xxbrd): New insn.
8069         (bswapdi2_reg): Disallow on ISA 3.0.
8070         (register to register bswap64 splitter): Do not split the insn on
8071         ISA 3.0 systems that use XXBRD.
8073 2017-11-10  Martin Sebor  <msebor@redhat.com>
8075         PR c/81117
8076         * config/darwin-c.c (framework_construct_pathname): Replace strncpy
8077         with memcpy.
8078         (find_subframework_file): Same.
8080 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8082         * auto-profile.c (afdo_indirect_call): Drop frequency.
8083         * cgraph.c (symbol_table::create_edge): Drop frequency argument.
8084         (cgraph_node::create_edge): Drop frequency argument.
8085         (cgraph_node::create_indirect_edge): Drop frequency argument.
8086         (cgraph_edge::make_speculative): Drop frequency arguments.
8087         (cgraph_edge::resolve_speculation): Do not update frequencies
8088         (cgraph_edge::dump_edge_flags): Do not dump frequency.
8089         (cgraph_node::dump): Check consistency in IPA mode.
8090         (cgraph_edge::maybe_hot_p): Use IPA counter.
8091         (cgraph_edge::verify_count_and_frequency): Rename to ...
8092         (cgraph_edge::verify_count): ... this one; drop frequency checking.
8093         (cgraph_node::verify_node): Update.
8094         * cgraph.h (struct cgraph_edge): Drop frequency.
8095         (cgraph_edge::frequency): New function.
8096         * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
8097         frequencies.
8098         (cgraph_edge::rebuild_edges): Likewise.
8099         * cgraphclones.c (cgraph_edge::clone): Scale only counts.
8100         (duplicate_thunk_for_node): Do not pass frequency.
8101         (cgraph_node::create_clone): Scale only counts.
8102         (cgraph_node::create_virtual_clone): Do not pass frequency.
8103         (cgraph_node::create_edge_including_clones): Do not pass frequency.
8104         (cgraph_node::create_version_clone): Do not pass frequency.
8105         * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
8106         (cgraph_node::expand_thunk): Do not pass frequency.
8107         (cgraph_node::create_wrapper): Do not pass frequency.
8108         * gimple-iterator.c (update_call_edge_frequencies): Do not pass
8109         frequency.
8110         * gimple-streamer-in.c (input_bb): Scale only IPA counts.
8111         * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
8112         * ipa-cp.c (ipcp_lattice::print): Use frequency function.
8113         (gather_caller_stats): Use frequency function.
8114         (ipcp_cloning_candidate_p): Use frequency function.
8115         (ipcp_propagate_stage): Use frequency function.
8116         (get_info_about_necessary_edges): Use frequency function.
8117         (update_profiling_info): Update only IPA profile.
8118         (update_specialized_profile): Use frequency functoin.
8119         (perhaps_add_new_callers): Update only IPA profile.
8120         * ipa-devirt.c (ipa_devirt): Use IPA profile.
8121         * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
8122         (dump_ipa_call_summary): Use frequency function.
8123         (estimate_edge_size_and_time): Use frequency function.
8124         (ipa_merge_fn_summary_after_inlining): Use frequency function.
8125         * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
8126         * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
8127         (update_noncloned_counts): ... ths one; scale counts only.
8128         (clone_inlined_nodes): Do not scale frequency.
8129         (inline_call): Do not pass frequency.
8130         * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
8131         (compute_inlined_call_time): Use IPA profile.
8132         (want_inline_small_function_p): Use IPA profile.
8133         (want_inline_self_recursive_call_p): Use IPA profile.
8134         (edge_badness): Use IPA profile.
8135         (lookup_recursive_calls): Use IPA profile.
8136         (recursive_inlining): Do not pass frequency.
8137         (resolve_noninline_speculation): Do not update frequency.
8138         (inline_small_functions): Collect max of IPA profile.
8139         (dump_overall_stats): Dump IPA porfile.
8140         (dump_inline_stats): Dump IPA porfile.
8141         (ipa_inline): Collect IPA stats.
8142         * ipa-inline.h (clone_inlined_nodes): Update prototype.
8143         * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
8144         (ipa_propagate_frequency): Use frequency function.
8145         (ipa_profile): Cleanup.
8146         * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
8147         * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
8148         * lto-cgraph.c (lto_output_edge): Do not stream frequency.
8149         (input_node): Do not stream frequency.
8150         (input_edge): Do not stream frequency.
8151         (merge_profile_summaries): Scale only IPA profiles.
8152         * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
8153         * predict.c (drop_profile): Do not recompute frequency.
8154         * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
8155         (ipa_tm_insert_gettmclone_call): Do not pass frequency.
8156         * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
8157         * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
8158         * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
8159         * tree-inline.c (copy_bb): Do not scale frequency.
8160         (expand_call_inline): Do not scale frequency.
8161         (tree_function_versioning): Do not scale frequency.
8162         * ubsan.c (ubsan_create_edge): Do not pass frequency.
8164 2017-11-10  Julia Koval  <julia.koval@intel.com>
8166         * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
8167         (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
8168         (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
8169         (_mm256_maskz_gf2p8affine_epi64_epi8)
8170         (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
8171         (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
8172         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
8173         (__builtin_ia32_vgf2p8affineqb_v32qi)
8174         (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
8175         * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
8177 2017-11-10  Tamar Christina  <tamar.christina@arm.com>
8179         PR target/82641
8180         * config/arm/arm.c
8181         (arm_option_override): Refactor.
8182         (arm_option_reconfigure_globals): New.
8183         (arm_options_perform_arch_sanity_checks): New.
8184         * config/arm/arm-protos.h (arm_option_reconfigure_globals):
8185         New prototype.
8186         (arm_options_perform_arch_sanity_checks): Likewise
8188 2017-11-10  Pat Haugen  <pthaugen@us.ibm.com>
8190         * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
8191         (power9-qpmul): New.
8192         * rs6000/rs6000.md ("type" attr): Add qmul.
8193         (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
8194         *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
8195         *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
8197 2017-11-10  Martin Sebor  <msebor@redhat.com>
8199         PR c/81117
8200         * builtins.c (compute_objsize): Handle arrays that
8201         compute_builtin_object_size likes to fail for.  Make extern.
8202         * builtins.h (compute_objsize): Declare.
8203         (check_strncpy_sizes): New function.
8204         (expand_builtin_strncpy): Call check_strncpy_sizes.
8205         * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
8206         -Wstringop-truncation.
8207         (gimple_fold_builtin_strncat): Same.
8208         * gimple.c (gimple_build_call_from_tree): Set call location.
8209         * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
8210         (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
8211         (handle_builtin_stxncpy, handle_builtin_strncat): Same.
8212         (handle_builtin_strlen): Use strlen_to_stridx.
8213         (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
8214         stpncpy.
8215         Use strlen_to_stridx.
8216         (pass_strlen::execute): Release strlen_to_stridx.
8217         * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
8218         (-Wstringop-truncation): Document new option.
8220 2017-11-10  Martin Liska  <mliska@suse.cz>
8222         PR gcov-profile/82702
8223         * gcov.c (main): Handle intermediate files in a different
8224         way.
8225         (get_gcov_intermediate_filename): New function.
8226         (output_gcov_file): Remove support of intermediate files.
8227         (generate_results): Allocate intermediate file.
8228         (release_structures): Clean-up properly fn_end.
8229         (output_intermediate_file): Start iterating with line 1.
8231 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8233         PR tree-optimization/82929
8234         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8235         ops_swapped_p non-static data member.
8236         (store_immediate_info::store_immediate_info): Clear it.
8237         (imm_store_chain_info::coalesce_immediate_stores): If swapping
8238         ops set ops_swapped_p.
8239         (count_multiple_uses): Handle ops_swapped_p.
8241 2017-11-10  Martin Liska  <mliska@suse.cz>
8243         * coverage.c (coverage_init): Stream information about
8244         support of has_unexecuted_blocks.
8245         * doc/gcov.texi: Document that.
8246         * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8247         * gcov.c (read_graph_file): Likewise.
8248         (output_line_beginning): Fix a small issue with
8249         color output.
8251 2017-11-10  Bin Cheng  <bin.cheng@arm.com>
8253         * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8254         reference of trivial component.
8256 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8258         PR bootstrap/82916
8259         * gimple-ssa-store-merging.c
8260         (pass_store_merging::terminate_all_aliasing_chains): For
8261         gimple_store_p stmts also call refs_output_dependent_p.
8263         PR rtl-optimization/82913
8264         * compare-elim.c (try_merge_compare): Punt if def_insn is not
8265         single set.
8267 2017-11-09  Jeff Law  <law@redhat.com>
8269         * vr-values.h: New file with vr_values class.
8270         * tree-vrp.c: Include vr-values.h
8271         (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8272         data objects into the vr_values class.
8273         (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8274         (get_value_range): Make it a member function within vr_values class.
8275         (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8276         (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8277         (op_with_constant_singleton_value_range): Likewise.
8278         (extract_range_for_var_from_comparison_expr): Likewise.
8279         (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8280         (extract_range_from_binary_expr): Likewise.
8281         (extract_range_from_unary_expr): Likewise.
8282         (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8283         (check_for_binary_op_overflow, extract_range_basic): Likewise.
8284         (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8285         (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8286         (compare_name_with_value, compare_names): Likewise.
8287         (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8288         (vrp_evaluate_conditional_warnv_with_ops): Likewise.  Remove prototype.
8289         (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8290         (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8291         (extract_range_from_phi_node): Likewise.
8292         (simplify_truth_ops_using_ranges): Likewise.
8293         (simplify_div_or_mod_using_ranges): Likewise.
8294         (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8295         (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8296         (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8297         (simplify_float_conversion_using_ranges): Likewise.
8298         (simplify_internal_call_using_ranges): Likewise.
8299         (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8300         (vrp_visit_assignment_or_call): Likewise.  Smuggle class instance
8301         poitner via x_vr_values for calls into gimple folder.
8302         (vrp_initialize_lattice): Make this the vr_values ctor.
8303         (vrp_free_lattice): Make this the vr_values dtor.
8304         (set_vr_value): New function.
8305         (class vrp_prop): Add vr_values data member.  Add various member
8306         functions as well as member functions that delegate to vr_values.
8307         (check_array_ref): Make a member function within vrp_prop class.
8308         (search_for_addr_array, vrp_initialize): Likewise.
8309         (vrp_finalize): Likewise.  Revamp to avoid direct access to
8310         vr_value, values_propagated, etc.
8311         (check_array_bounds): Extract vrp_prop class instance pointer from
8312         walk info structure.  Use it to call member functions.
8313         (check_all_array_refs): Make a member function within vrp_prop class.
8314         Smuggle class instance pointer via walk info structure.
8315         (x_vr_values): New local static.
8316         (vrp_valueize): Use x_vr_values to get class instance.
8317         (vr_valueize_1): Likewise.
8318         (class vrp_folder): Add vr_values data member.  Add various member
8319         functions as well as member functions that delegate to vr_values.
8320         (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8321         (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8322         class instance from vr_values.  Use it to call member functions.
8323         (vrp_dom_walker): Add vr_values data member.
8324         (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8325         instance via x_vr_values.
8326         (identify_jump_threads): Accept vr_values as argument.  Store
8327         it into the walker structure.
8328         (evrp_dom_walker): Add vr_values class data member.  Add various
8329         delegators.
8330         (evrp_dom_walker::try_find_new_range): Use vr_values data
8331         member to access the memory allocator.
8332         (evrp_dom_walker::before_dom_children): Store vr_values class
8333         instance into the vrp_folder class.
8334         (evrp_dom_walker::push_value_range): Rework to avoid direct
8335         access to num_vr_values and vr_value.
8336         (evrp_dom_walker::pop_value_range): Likewise.
8337         (execute_early_vrp): Remove call to vrp_initialize_lattice.
8338         Use vr_values to get to dump_all_value_ranges member function.
8339         Remove call to vrp_free_lattice.  Call vrp_initialize, vrp_finalize,
8340         and simplify_cond_using_ranges_2 via vrp_prop class instance.
8341         Pass vr_values class instance down to identify_jump_threads.
8342         Remove call to vrp_free_lattice.
8343         (debug_all_value_ranges): Remove.
8345         * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8346         (vrp_folder): Likewise.
8348         * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8349         Get it from the existing bitmap instead.
8350         (vrp_intersect_ranges_1): Likewise.
8352 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8354         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8355         bit_not_p field.
8356         (store_immediate_info::store_immediate_info): Add bitnotp argument,
8357         set bit_not_p to it.
8358         (imm_store_chain_info::coalesce_immediate_stores): Break group
8359         if bit_not_p is different.
8360         (count_multiple_uses, split_group,
8361         imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8362         (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8363         (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8364         result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8365         to store_immediate_info ctor.
8367 2017-11-09  Jim Wilson  <jimw@sifive.com>
8369         * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8370         (scan_prog_file): Likewise.
8372 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8374         * bb-reorder.c (max_entry_frequency): Remove.
8375         (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8376         connect_traces, push_to_next_round_p): Remove prototypes.
8377         (find_traces_1_round): Use counts only.
8378         (push_to_next_round_p): Likewise.
8379         (find_traces): Likewise.
8380         (rotate_loop): Likewise.
8381         (find_traces_1_round): Likewise.
8382         (connect_traces): Likewise.
8383         (edge_order): Likewise.
8385 2017-11-09  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8387         * config/arm/arm.c (output_return_instruction): Add comments to
8388         indicate requirement for cmse_nonsecure_entry return to account
8389         for the size of clearing instruction output here.
8390         (thumb_exit): Likewise.
8391         * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8392         return in hardfloat mode.
8394 2017-11-09  Segher Boessenkool  <segher@kernel.crashing.org>
8396         * config/rs6000/rs6000.c (machine_function): Add a bool,
8397         "toc_is_wrapped_separately".
8398         (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8399         if it wasn't explicitly set or unset, we are optimizing for speed, and
8400         doing separate shrink-wrapping.
8401         (rs6000_get_separate_components): Enable the TOC component if
8402         saving the TOC register in the prologue.
8403         (rs6000_components_for_bb): Handle the TOC component.
8404         (rs6000_emit_prologue_components): Store the TOC register where needed.
8405         (rs6000_set_handled_components): Mark TOC as handled, if handled.
8406         (rs6000_emit_prologue): Don't save the TOC if that is already done.
8408 2017-11-09  Martin Jambor  <mjambor@suse.cz>
8410         * ipa-param-manipulation.c: New file.
8411         * ipa-param-manipulation.h: Likewise.
8412         * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8413         (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8414         * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8415         (ipa_parm_adjustment): Likewise.
8416         (ipa_parm_adjustment_vec): Likewise.
8417         (ipa_get_vector_of_formal_parms): Moved declaration to
8418         ipa-param-manipulation.h.
8419         (ipa_get_vector_of_formal_parm_types): Likewise.
8420         (ipa_modify_formal_parameters): Likewise.
8421         (ipa_modify_call_arguments): Likewise.
8422         (ipa_combine_adjustments): Likewise.
8423         (ipa_dump_param_adjustments): Likewise.
8424         (ipa_modify_expr): Likewise.
8425         (ipa_get_adjustment_candidate): Likewise.
8426         * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8427         ipa-param-manipulation.c.
8428         (ipa_get_vector_of_formal_parm_types): Likewise.
8429         (ipa_modify_formal_parameters): Likewise.
8430         (ipa_modify_call_arguments): Likewise.
8431         (ipa_modify_expr): Likewise.
8432         (get_ssa_base_param): Likewise.
8433         (ipa_get_adjustment_candidate): Likewise.
8434         (index_in_adjustments_multiple_times_p): Likewise.
8435         (ipa_combine_adjustments): Likewise.
8436         (ipa_dump_param_adjustments): Likewise.
8437         * tree-sra.c: Also include ipa-param-manipulation.h
8438         * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8439         ipa-param.h.
8441 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8442             Alan Hayward  <alan.hayward@arm.com>
8443             David Sherwood  <david.sherwood@arm.com>
8445         * doc/sourcebuild.texi (vect_masked_store): Document.
8447 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8448             Alan Hayward  <alan.hayward@arm.com>
8449             David Sherwood  <david.sherwood@arm.com>
8451         * doc/sourcebuild.texi (vect_align_stack_vars): Document.
8453 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8454             Alan Hayward  <alan.hayward@arm.com>
8455             David Sherwood  <david.sherwood@arm.com>
8457         * doc/sourcebuild.texi (vect_variable_length): Document.
8459 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8460             Alan Hayward  <alan.hayward@arm.com>
8461             David Sherwood  <david.sherwood@arm.com>
8463         * doc/sourcebuild.texi (vect_unaligned_possible): Document.
8465 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8466             Alan Hayward  <alan.hayward@arm.com>
8467             David Sherwood  <david.sherwood@arm.com>
8469         * doc/sourcebuild.texi (vect_element_align_preferred): Document.
8471 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8472             Alan Hayward  <alan.hayward@arm.com>
8473             David Sherwood  <david.sherwood@arm.com>
8475         * doc/sourcebuild.texi (vect_perm_short, vect_perm_byte): Document
8476         previously undocumented selectors.
8477         (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8479 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8481         * doc/rtl.texi (const_vector): Say that elements can be
8482         const_wide_ints too.
8483         * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8484         * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8485         (gen_vec_duplicate): Use it instead of CONSTANT_P.
8486         * optabs.c (expand_vector_broadcast): Likewise.
8488 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8489             Alan Hayward  <alan.hayward@arm.com>
8490             David Sherwood  <david.sherwood@arm.com>
8492         * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8493         scaled index even if the unscaled address was invalid.
8494         Don't increase the complexity of using a scale in that case.
8496 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8497             Alan Hayward  <alan.hayward@arm.com>
8498             David Sherwood  <david.sherwood@arm.com>
8500         * doc/rtl.texi: Rewrite the subreg rules so that they partition
8501         the inner register into REGMODE_NATURAL_SIZE bytes rather than
8502         UNITS_PER_WORD bytes.
8503         * emit-rtl.c (validate_subreg): Divide subregs into blocks
8504         based on REGMODE_NATURAL_SIZE of the inner mode.
8505         (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8506         !SCALAR_FLOAT_MODE_P cases.  Use REGMODE_NATURAL_SIZE for the latter.
8507         * expmed.c (lowpart_bit_field_p): Divide the value up into
8508         chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8509         * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8510         whether something is likely to occupy more than one register.
8512 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8514         PR ipa/82879
8515         * ipa-inline-transform.c (update_noncloned_frequencies): Use
8516         profile_count::adjust_for_ipa_scaling.
8517         * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8518         * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8519         function.
8520         * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8522 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8524         * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8525         (split_group): Add total_orig and total_new arguments, estimate the
8526         number of statements related to the store group without store merging
8527         and with store merging.
8528         (imm_store_chain_info::output_merged_store): Adjust split_group
8529         callers, punt if estimated number of statements with store merging
8530         is not smaller than estimated number of statements without it.
8531         Formatting fix.
8532         (handled_load): Remove has_single_use checks.
8533         (pass_store_merging::process_store): Likewise.
8535 2017-11-09  Richard Biener  <rguenther@suse.de>
8537         PR tree-optimization/82902
8538         * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8540 2017-11-09  Martin Liska  <mliska@suse.cz>
8542         PR target/82863
8543         * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8544         uninitialized.
8546 2017-11-09  Martin Liska  <mliska@suse.cz>
8548         PR tree-optimization/82669
8549         * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8551 2017-11-09  Martin Liska  <mliska@suse.cz>
8553         PR gcov-profile/48463
8554         * coverage.c (coverage_begin_function): Output also end locus
8555         of a function and information whether the function is
8556         artificial.
8557         * gcov-dump.c (tag_function): Parse and print the information.
8558         * gcov.c (INCLUDE_MAP): Add include.
8559         (INCLUDE_SET): Likewise.
8560         (struct line_info): Move earlier in the source file because
8561         of vector<line_info> in function_info structure.
8562         (line_info::line_info): Likewise.
8563         (line_info::has_block): Likewise.
8564         (struct source_info): Add new member index.
8565         (source_info::get_functions_at_location): New function.
8566         (function_info::group_line_p): New function.
8567         (output_intermediate_line): New function.
8568         (output_intermediate_file): Use the mentioned function.
8569         (struct function_start): New.
8570         (struct function_start_pair_hash): Likewise.
8571         (process_file): Add code that identifies group functions.
8572         Assign lines either to global or function scope.
8573         (generate_results): Skip artificial functions.
8574         (find_source): Assign index for each source file.
8575         (read_graph_file): Read new flag artificial and end_line.
8576         (add_line_counts): Assign it either to global of function scope.
8577         (accumulate_line_counts): Isolate core of the function to
8578         accumulate_line_info and call it for both function and global
8579         scope lines.
8580         (accumulate_line_info): New function.
8581         (output_line_beginning): Fix GNU coding style.
8582         (print_source_line): New function.
8583         (output_line_details): Likewise.
8584         (output_function_details): Likewise.
8585         (output_lines): Iterate both source (global) scope and function
8586         scope.
8587         (struct function_line_start_cmp): New class.
8588         * doc/gcov.texi: Reflect changes in documentation.
8590 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8592         PR debug/82837
8593         * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8594         (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8595         and similarly for not instead of neg.
8597 2017-11-08  Andi Kleen  <ak@linux.intel.com>
8599         * config/i386/i386.opt: Add -mforce-indirect-call.
8600         * config/i386/predicates.md: Check for flag_force_indirect_call.
8601         * doc/invoke.texi: Document -mforce-indirect-call
8603 2017-11-08  Kito Cheng  <kito.cheng@gmail.com>
8605         * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8606         New extern.
8607         (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8608         config/riscv/riscv.c (predict.h): New include.
8609         (riscv_slow_unaligned_access_p): No longer static.
8610         (riscv_block_move_straight): Add require.
8611         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8613 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8615         PR target/82855
8616         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8617         <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8618         nonimmediate_operand predicate for operand 1 instead of
8619         register_operand.
8621 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8623         * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8624         New pattern.
8625         * config/aarch64/constraints.md (Uml): New constraint.
8626         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8627         predicate.
8629 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8631         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8632         of two vec_duplicates into a vec_concat.
8634 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8636         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8637         Simplify vec_merge of vec_duplicate and vec_concat.
8638         * config/aarch64/constraints.md (Utq): New constraint.
8639         * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8640         define_insn.
8642 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8644         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8645         Simplify vec_merge of vec_duplicate and const_vector.
8646         * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8647         New predicate.
8648         * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8649         mode iterator.  Update predicate on operand 1 to
8650         handle non-const_vec constants.  Delete constraints.
8651         (*aarch64_combinez_be<mode>): Likewise for operand 2.
8653 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8655         PR tree-optimization/78821
8656         * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8657         data member.
8658         (store_operand_info::store_operand_info): Initialize it to false.
8659         (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8660         ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8661         store in the group, and if chain_info is non-NULL, to ignore altogether
8662         that chain.
8663         (compatible_load_p): Fail if bit_not_p does not match.
8664         (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8665         (handled_load): Fill in bit_not_p.  Handle BIT_NOT_EXPR.
8666         (pass_store_merging::process_store): Adjust
8667         terminate_all_aliasing_chains calls to pass NULL in all current spots,
8668         call terminate_all_aliasing_chains newly when adding a store into
8669         a chain with non-NULL chain_info.
8671 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8673         * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8675 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8677         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8678         Remove.
8679         (aarch64_layout_frame): Initialise emit_frame_chain.
8680         (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8681         (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8683 2017-11-08  Martin Liska  <mliska@suse.cz>
8685         * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8686         of gimple_call_internal_p.
8688 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8690         * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8692 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8694         * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8695         "do {} while (0)".
8697 2017-11-08  Martin Liska  <mliska@suse.cz>
8699         PR sanitizer/82792
8700         * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8702 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8704         * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8706 2017-11-07  Jakub Jelinek  <jakub@redhat.com>
8708         PR target/82855
8709         * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8710         (*cmp<mode>_ccz_1): New insn with $k alternative.
8712         PR target/82855
8713         * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8714         RTX_COMM_COMPARE as commutative as well.
8715         (ix86_binary_operator_ok): Formatting fix.
8716         * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8717         *<code><mode>3<mask_name><round_saeonly_name>,
8718         *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8719         *<s>mul<mode>3_highpart<mask_name>,
8720         *vec_widen_umult_even_v16si<mask_name>,
8721         *vec_widen_umult_even_v8si<mask_name>,
8722         *vec_widen_umult_even_v4si<mask_name>,
8723         *vec_widen_smult_even_v16si<mask_name>,
8724         *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8725         *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8726         *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8727         *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8728         *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8729         <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8730         *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8731         *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8732         *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8733         !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8734         ix86_binary_operator_ok.  Formatting fixes.
8735         (*<plusminus_insn><mode>3<mask_name><round_name>,
8736         *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8737         fixes.
8739 2017-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
8741         * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8742         ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8743         eq and ne if TARGET_ISEL.
8744         (cmp): New code_iterator.
8745         (UNS, UNSU_, UNSIK): New code_attrs.
8746         (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8747         ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8748         to...
8749         ("eq<mode>3"): ... this.
8750         ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8751         to...
8752         ("ne<mode>3"): ... this.
8754 2017-11-07  Julia Koval  <julia.koval@intel.com>
8756         PR target/82812
8757         * common/config/i386/i386-common.c
8758         (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8759         (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8760         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8761         * config/i386/i386.opt: Ditto.
8762         * config/i386/i386.c (ix86_target_string): Ditto.
8763         (ix86_option_override_internal): Ditto.
8764         (ix86_init_mpx_builtins): Move MPX to args2.
8765         (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8766         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8767         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8768         __builtin_ia32_vgf2p8affineinvqb_v32qi,
8769         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8770         __builtin_ia32_vgf2p8affineinvqb_v16qi,
8771         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8773 2017-11-07  Uros Bizjak  <ubizjak@gmail.com>
8775         PR target/80425
8776         * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8777         and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8778         (zero-extendsidi peephole2): Remove peephole.
8780 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8782         PR c/53037
8783         * stor-layout.c: Include attribs.h.
8784         (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8785         explicit lookup of "aligned" attribute.
8787 2017-11-07  Andrew Waterman  <andrew@sifive.com>
8789         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8790         (riscv_expand_block_move): Likewise.
8791         * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8792         implementation.
8793         (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8794         (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8795         * config/riscv/riscv.c (riscv_block_move_straight): New function.
8796         (riscv_adjust_block_mem): Likewise.
8797         (riscv_block_move_loop): Likewise.
8798         (riscv_expand_block_move): Likewise.
8799         * config/riscv/riscv.md (movmemsi): New pattern.
8801 2017-11-07  Michael Clark  <michaeljclark@mac.com>
8803         * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8804         (MUSL_DYNAMIC_LINKER): Likewise.
8806 2017-11-07  Richard Sandiford  <richard.sandiford@linaro.org>
8808         * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8809         (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8810         nonmmory_operand.
8812 2017-11-07  Richard Biener  <rguenther@suse.de>
8814         * match.pd: Fix build.
8816 2017-11-07  Wilco Dijkstra  <wdijkstr@arm.com>
8817             Jackson Woodruff  <jackson.woodruff@arm.com>
8819         PR tree-optimization/71026
8820         * match.pd: Canonicalize negate in division.
8822 2017-11-07  Sudakshina Das  <sudi.das@arm.com>
8824         PR middle-end/80131
8825         * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8827 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8829         * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8830         (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8832 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8834         * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8835         non-scalar integral types.
8836         * match.pd (negate_expr_p): Handle MINUS_EXPR.
8837         (-(A-B), -(~A)): New transformations.
8839 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8841         * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8842         semicolon after "do {} while (0)".
8843         * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8844         * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8845         * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8846         * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8847         * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8849 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8851         * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8852         after "do {} while (0)".
8853         * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8854         * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8855         * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8856         * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8857         * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8859 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8861         PR other/82784
8862         * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8863         "while {} do (0)".
8864         (arm_rtx_costs_internal): Add missing semicolon after
8865         HANDLE_NARROW_SHIFT_ARITH call.
8867 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8869         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8870         disable isel if it was not set explicitly.
8872 2017-11-06  James Bowman  <james.bowman@ftdichip.com>
8874         * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8875         (add_type_attribute) likewise.
8877 2017-11-06  H.J. Lu  <hongjiu.lu@intel.com>
8879         * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8880         of struct ix86_frame.
8881         (ix86_initial_elimination_offset): Likewise.
8882         (ix86_expand_split_stack_prologue): Likewise.
8884 2017-11-06  Marc Glisse  <marc.glisse@inria.fr>
8886         * tree-vrp.h (enum value_range_type): Update stale comment.
8888 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8889             Alan Hayward  <alan.hayward@arm.com>
8890             David Sherwood  <david.sherwood@arm.com>
8892         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8893         (aarch64_expand_vec_perm_const): Take the number of units too.
8894         * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8895         (aarch64_expand_vec_perm_const): Likewise.
8896         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8897         (vec_perm<mode>): Update accordingly.
8899 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8900             Alan Hayward  <alan.hayward@arm.com>
8901             David Sherwood  <david.sherwood@arm.com>
8903         * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8904         Take the number of units too.
8905         * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8906         (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8907         but check for a vector mode before rather than after the call.
8908         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8909         (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8910         (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8911         (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8912         (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8913         (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8914         (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8915         (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8916         (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8917         (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
8918         (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
8919         (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
8920         (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
8921         (aarch64_sqdmull2_n<mode>): Update accordingly.
8923 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8924             Alan Hayward  <alan.hayward@arm.com>
8925             David Sherwood  <david.sherwood@arm.com>
8927         * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
8928         the number of units too.
8929         * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
8930         * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
8931         (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
8932         (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
8933         (vec_store_lanesxi<mode>): Update accordingly.
8935 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8936             Alan Hayward  <alan.hayward@arm.com>
8937             David Sherwood  <david.sherwood@arm.com>
8939         * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
8940         * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
8941         * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
8942         of units rather than the mode.
8943         * config/aarch64/iterators.md (nunits): New mode attribute.
8944         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
8945         Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
8946         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
8947         (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
8948         (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
8949         (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
8950         (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
8951         (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
8952         (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
8953         (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
8954         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
8955         (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
8956         (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
8957         (reduc_<maxmin_uns>_scal_<mode>): Likewise.
8958         (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
8959         (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
8960         (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
8961         (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
8962         (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
8963         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
8964         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
8965         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
8966         (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
8967         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
8968         (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
8969         (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
8970         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
8971         (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
8972         (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
8973         (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
8974         (aarch64_vec_load_lanesci_lane<mode>): Likewise.
8975         (aarch64_vec_store_lanesci_lane<mode>): Likewise.
8976         (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
8977         (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
8978         (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
8979         (aarch64_simd_vec_setv2di): Likewise.
8981 2017-11-06  Carl Love  <cel@us.ibm.com>
8983         * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
8984         definitions.
8985         (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
8986         * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
8987         to power 8.
8988         * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
8989         extern declaration.
8990         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
8991         * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
8992         Add power 8 macro expansions.
8993         (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
8994         * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
8995         power 8 instructions.  (VSX_XXBR): Add iterator.
8997 2017-11-06  Wilco Dijkstra  <wdijkstr@arm.com>
8999         * config/arm/arm.md (predicable_short_it): Change default to "no",
9000         improve documentation, remove uses that are identical to the default.
9001         (enabled_for_depr_it): Rename to enabled_for_short_it.
9002         * gcc/config/arm/arm-fixed.md (predicable_short_it):
9003         Remove default uses.
9004         * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
9005         * gcc/config/arm/sync.md (predicable_short_it): Likewise.
9006         * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
9007         * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
9009 2017-11-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
9011         PR target/82748
9012         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
9013         float128 helper macros, which are no longer used after deleting
9014         the old 'q' built-in functions, and moving the round to odd
9015         built-in functions to being special built-in functions.
9016         (BU_FLOAT128_2): Likewise.
9017         (BU_FLOAT128_1_HW): Likewise.
9018         (BU_FLOAT128_2_HW): Likewise.
9019         (BU_FLOAT128_3_HW): Likewise.
9020         (FABSQ): Delete old 'q' built-in functions.
9021         (COPYSIGNQ): Likewise.
9022         (SQRTF128_ODD): Move round to odd built-in functions to be
9023         special built-in functions, so that we can handle
9024         -mabi=ieeelongdouble.
9025         (TRUNCF128_ODD): Likewise.
9026         (ADDF128_ODD): Likewise.
9027         (SUBF128_ODD): Likewise.
9028         (MULF128_ODD): Likewise.
9029         (DIVF128_ODD): Likewise.
9030         (FMAF128_ODD): Likewise.
9031         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
9032         built-in names to 'f128'.
9033         * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
9034         old 'q' built-in functions, as the machine independent code for
9035         'f128' built-in functions handles this.
9036         (rs6000_expand_builtin): Add expansion for float128 round to odd
9037         functions, keying off on -mabi=ieeelongdouble of whether to use
9038         the KFmode or TFmode variant.
9039         (rs6000_init_builtins): Initialize the _Float128 round to odd
9040         built-in functions.
9041         * doc/extend.texi (PowerPC Built-in Functions): Document the old
9042         _Float128 'q' built-in functions are now mapped into the new
9043         'f128' built-in functions.
9045 2017-11-06  David Edelsohn  <dje.gcc@gmail.com>
9047         * collect2.c (add_lto_object): Compile for OBJECT_COFF.
9048         (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
9050 2017-11-06  David Malcolm  <dmalcolm@redhat.com>
9052         PR jit/82826
9053         * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
9054         * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
9055         * toplev.c: Include "ipa-fnsummary.h".
9056         (toplev::finalize): Call ipa_fnsummary_c_finalize.
9058 2017-11-06  Jakub Jelinek  <jakub@redhat.com>
9060         PR tree-optimization/82838
9061         * gimple-ssa-store-merging.c
9062         (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
9063         on a separate gimple_seq which is then appended to seq.
9065 2017-11-06  Jeff Law  <law@redhat.com>
9067         PR target/82788
9068         * config/i386/i386.c (PROBE_INTERVAL): Remove.
9069         (get_probe_interval): New functions.
9070         (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
9071         (ix86_adjust_stack_and_probe): Likewise.
9072         (output_adjust_stack_and_probe): Likewise.
9073         (ix86_emit_probe_stack_range): Likewise.
9074         (ix86_expand_prologue): Likewise.
9076 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9078         PR tree-optimization/82816
9079         * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
9080         if the modes of the two types are the same.
9081         (convert_plusminus_to_widen): Likewise.
9083 2017-11-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9085         * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
9086         p9_vadu<mode>3.
9087         (usadv16qi): New define_expand.
9088         (usadv8hi): New define_expand.
9090 2017-11-06  Jan Hubicka  <hubicka@ucw.cz>
9092         PR bootstrap/82832
9093         * ipa-inline-transform.c (update_noncloned_frequencies): Always
9094         scale.
9095         (inline_transform): Likewise.
9096         * predict.c (counts_to_freqs): Remove useless conditional.
9097         * profile-count.h (profile_count::apply_scale): Move sanity check.
9098         * tree-inline.c (copy_bb): Always scale.
9099         (copy_cfg_body): Likewise.
9101 2017-11-06  Christophe Lyon  <christophe.lyon@linaro.org>
9103         PR target/67591
9104         * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
9105         attribute.
9106         (*cmp_ite0): Add enabled_for_depr_it attribute.
9107         (*cmp_ite1): Likewise.
9109 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
9111         * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
9112         TYPE_MFCRF.
9114 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9116         * tree-vrp.c (vrp_int_const_binop): Return true on success and
9117         return the value by pointer.
9118         (extract_range_from_multiplicative_op_1): Update accordingly.
9119         Return as soon as an operation fails.
9121 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9123         PR other/82784
9124         * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
9125         (DEF_SANITIZER_BUILTIN): ... here.
9126         (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
9127         DEF_SANITIZER_BUILTIN in if stmt.  Add missing semicolon.
9129 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9131         PR other/82784
9132         * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
9133         macro body.
9134         (ASM_OUTPUT_CASE_LABEL): Add semicolon after
9135         ASM_OUTPUT_BEFORE_CASE_LABEL call.
9136         * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
9137         after macro body.
9138         * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9139         * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9140         * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9142 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9144         PR other/82784
9145         * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
9146         "do {} while (0)".
9148 2017-11-04  Michael Clark  <michaeljclark@mac.com>
9150         * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
9151         config/riscv/riscv.md (addsi3): Use 'i' for immediates.
9152         (adddi3): Likewise.
9153         (*addsi3_extended): Likewise.
9154         (*addsi3_extended2): Likewise.
9155         (<optab>si3): Likewise.
9156         (<optab>di3): Likewise.
9157         (<optab><mode>3): Likewise.
9158         (<*optabe>si3_internal): Likewise.
9159         (zero_extendqi<SUPERQI:mode>2): Likewise.
9160         (*add<mode>hi3): Likewise.
9161         (*xor<mode>hi3): Likewise.
9162         (<optab>di3): Likewise.
9163         (*<optab>si3_extend): Likewise.
9164         (*sge<u>_<X:mode><GPR:mode>): Likewise.
9165         (*slt<u>_<X:mode><GPR:mode>): Likewise.
9166         (*sle<u>_<X:mode><GPR:mode>): Likewise.
9168 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9170         * config/riscv/riscv.c (riscv_option_override): Conditionally set
9171         TARGET_STRICT_ALIGN based upon -mtune argument.
9173 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9175         * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
9177 2017-11-04  Daniel Santos  <daniel.santos@pobox.com>
9179         * config/i386/i386.c (choose_basereg): Use optional scratch
9180         register and add assertion.
9181         (x86_emit_outlined_ms2sysv_save): Use scratch register when
9182         needed, and don't allocate stack.
9183         (ix86_expand_prologue): Rearrange where SSE saves/stub call is
9184         emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
9185         (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
9187 2017-11-03  Jeff Law  <law@redhat.com>
9189         * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
9190         (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
9191         to probe at the start of a noreturn function.
9193 2017-11-03  Jakub Jelinek  <jakub@redhat.com>
9195         PR tree-optimization/78821
9196         * gimple-ssa-store-merging.c: Update the file comment.
9197         (MAX_STORE_ALIAS_CHECKS): Define.
9198         (struct store_operand_info): New type.
9199         (store_operand_info::store_operand_info): New constructor.
9200         (struct store_immediate_info): Add rhs_code and ops data members.
9201         (store_immediate_info::store_immediate_info): Add rhscode, op0r
9202         and op1r arguments to the ctor, initialize corresponding data members.
9203         (struct merged_store_group): Add load_align_base and load_align
9204         data members.
9205         (merged_store_group::merged_store_group): Initialize them.
9206         (merged_store_group::do_merge): Update them.
9207         (merged_store_group::apply_stores): Pick the constant for
9208         encode_tree_to_bitpos from one of the two operands, or skip
9209         encode_tree_to_bitpos if neither operand is a constant.
9210         (class pass_store_merging): Add process_store method decl.  Remove
9211         bool argument from terminate_all_aliasing_chains method decl.
9212         (pass_store_merging::terminate_all_aliasing_chains): Remove
9213         var_offset_p argument and corresponding handling.
9214         (stmts_may_clobber_ref_p): New function.
9215         (compatible_load_p): New function.
9216         (imm_store_chain_info::coalesce_immediate_stores): Terminate group
9217         if there is overlap and rhs_code is not INTEGER_CST.  For
9218         non-overlapping stores terminate group if rhs is not mergeable.
9219         (get_alias_type_for_stmts): Change first argument from
9220         auto_vec<gimple *> & to vec<gimple *> &.  Add IS_LOAD, CLIQUEP and
9221         BASEP arguments.  If IS_LOAD is true, look at rhs1 of the stmts
9222         instead of lhs.  Compute *CLIQUEP and *BASEP in addition to the
9223         alias type.
9224         (get_location_for_stmts): Change first argument from
9225         auto_vec<gimple *> & to vec<gimple *> &.
9226         (struct split_store): Remove orig_stmts data member, add orig_stores.
9227         (split_store::split_store): Create orig_stores rather than orig_stmts.
9228         (find_constituent_stmts): Renamed to ...
9229         (find_constituent_stores): ... this.  Change second argument from
9230         vec<gimple *> * to vec<store_immediate_info *> *, push pointers
9231         to info structures rather than the statements.
9232         (split_group): Rename ALLOW_UNALIGNED argument to
9233         ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
9234         it.  Adjust find_constituent_stores caller.
9235         (imm_store_chain_info::output_merged_store): Handle rhs_code other
9236         than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
9237         get_location_for_stmts callers.  Set MR_DEPENDENCE_CLIQUE and
9238         MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
9239         (mem_valid_for_store_merging): New function.
9240         (handled_load): New function.
9241         (pass_store_merging::process_store): New method.
9242         (pass_store_merging::execute): Use process_store method.  Adjust
9243         terminate_all_aliasing_chains caller.
9245 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9247         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9248         Return true for more constants, symbols and label references.
9249         (aarch64_valid_floating_const): Remove unused function.
9251 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9253         PR target/82786
9254         * config/aarch64/aarch64.c (aarch64_layout_frame):
9255         Undo forcing of LR at bottom of frame.
9257 2017-11-03  Jeff Law  <law@redhat.com>
9259         PR target/82823
9260         * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9261         for int_registers_saved.
9263         * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9264         extracted from tree-ssa-dom.c.
9265         * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9266         * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9267         (record_equivalences_from_incoming_edge): Add additional argument
9268         to single_pred_edge_ignoring_loop_edges call.
9269         * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9270         (uncprop_dom_walker::before_dom_children): Add additional argument
9271         to single_pred_edge_ignoring_loop_edges call.
9272         * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9273         single_pred_edge_ignoring_loop_edges rather than open coding.
9274         * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9276 2017-11-03  Marc Glisse  <marc.glisse@inria.fr>
9278         * match.pd (-(-A)): Rewrite.
9280 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9282         * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9283         (rs6000_emit_int_cmove): New declaration.
9284         * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9285         (rs6000_emit_sISEL): Delete.
9286         (rs6000_emit_int_cmove): Make non-static.
9287         * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9288         instead of rs6000_emit_sISEL.
9290 2017-11-03  Jan Hubicka  <hubicka@ucw.cz>
9292         * asan.c (create_cond_insert_point): Maintain profile.
9293         * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9294         merged.
9295         * basic-block.h (struct basic_block_def): Remove frequency.
9296         (EDGE_FREQUENCY): Use to_frequency
9297         * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9298         heuristics.
9299         (find_traces): Update to use to_frequency.
9300         (find_traces_1_round): Likewise; use only IPA counts.
9301         (bb_to_key): Likewise.
9302         (connect_traces): Use IPA counts only.
9303         (copy_bb_p): Update to use to_frequency.
9304         (fix_up_crossing_landing_pad): Likewise.
9305         (sanitize_hot_paths): Likewise.
9306         * bt-load.c (basic_block_freq): Likewise.
9307         * cfg.c (init_flow): Set count_max to uninitialized.
9308         (check_bb_profile): Remove frequencies; check counts.
9309         (dump_bb_info): Do not dump frequencies.
9310         (update_bb_profile_for_threading): Update counts only.
9311         (scale_bbs_frequencies_int): Likewise.
9312         (MAX_SAFE_MULTIPLIER): Remove.
9313         (scale_bbs_frequencies_gcov_type): Update counts only.
9314         (scale_bbs_frequencies_profile_count): Update counts only.
9315         (scale_bbs_frequencies): Update counts only.
9316         * cfg.h (struct control_flow_graph): Add count-max.
9317         (update_bb_profile_for_threading): Update prototype.
9318         * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9319         (find_many_sub_basic_blocks): Likewise.
9320         * cfgcleanup.c (try_forward_edges): Likewise.
9321         (try_crossjump_to_edge): Likewise.
9322         * cfgexpand.c (expand_gimple_cond): Likewise.
9323         (expand_gimple_tailcall): Likewise.
9324         (construct_init_block): Likewise.
9325         (construct_exit_block): Likewise.
9326         * cfghooks.c (verify_flow_info): Check consistency of counts.
9327         (dump_bb_for_graph): Do not dump frequencies.
9328         (split_block_1): Do not update frequencies.
9329         (split_edge): Do not update frequencies.
9330         (make_forwarder_block): Do not update frequencies.
9331         (duplicate_block): Do not update frequencies.
9332         (account_profile_record): Do not update frequencies.
9333         * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9334         for global heuristics.
9335         * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9336         (expected_loop_iterations_unbounded): Use counts only.
9337         * cfgloopmanip.c (scale_loop_profile): Simplify.
9338         (create_empty_loop_on_edge): Simplify
9339         (loopify): Simplify
9340         (duplicate_loop_to_header_edge): Simplify
9341         * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9342         (update_br_prob_note): Take care of removing note when profile
9343         becomes undefined.
9344         (relink_block_chain): Do not dump frequency.
9345         (rtl_account_profile_record): Use to_frequency.
9346         * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9347         (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9348         (cgraph_update_edges_for_call_stmt_node): Likewise.
9349         (cgraph_edge::verify_count_and_frequency): Update.
9350         (cgraph_node::verify_node): Temporarily disable frequency verification.
9351         * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9352         to_cgraph_frequency.
9353         (cgraph_edge::rebuild_edges): Convert to ipa counts.
9354         * cgraphunit.c (init_lowered_empty_function): Do not initialize
9355         frequencies.
9356         (cgraph_node::expand_thunk): Update profile.
9357         * except.c (dw2_build_landing_pads): Do not update frequency.
9358         * final.c (compute_alignments): Use to_frequency.
9359         (dump_basic_block_info): Do not dump frequency.
9360         * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9361         (dump_gimple_bb_header): Do not dump frequency.
9362         * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9363         do update count.
9364         * gimple-streamer-in.c (input_bb): Do not stream frequency.
9365         * gimple-streamer-out.c (output_bb): Do not stream frequency.
9366         * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9367         (init_before_recovery): Do not update frequency.
9368         (sched_create_recovery_edges): Do not update frequency.
9369         * hsa-gen.c (convert_switch_statements): Do not update frequency.
9370         * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9371         (ipa_cp_c_finalize): Set max_count to uninitialized.
9372         * ipa-fnsummary.c (get_minimal_bb): Use counts.
9373         (param_change_prob): Use counts.
9374         * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9375         local profiles.
9376         * ipa-split.c (consider_split): Use to_frequency.
9377         (split_function): Use to_frequency.
9378         * ira-build.c (loop_compare_func): Likewise.
9379         (mark_loops_for_removal): Likewise.
9380         (mark_all_loops_for_removal): Likewise.
9381         * loop-doloop.c (doloop_modify): Do not update frequency.
9382         * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9383         frequency.
9384         * lto-streamer-in.c (input_function): Update count_max.
9385         * omp-expand.c (expand_omp_taskreg): Update count_max.
9386         * omp-simd-clone.c (simd_clone_adjust): Update profile.
9387         * predict.c (maybe_hot_frequency_p): Use to_frequency.
9388         (maybe_hot_count_p): Use ipa counts only.
9389         (maybe_hot_bb_p): Simplify.
9390         (maybe_hot_edge_p): Simplify.
9391         (probably_never_executed): Do not take frequency argument.
9392         (probably_never_executed_bb_p): Do not pass frequency.
9393         (probably_never_executed_edge_p): Likewise.
9394         (combine_predictions_for_bb): Check that profile is nonzero.
9395         (propagate_freq): Do not set frequency.
9396         (drop_profile): Simplify.
9397         (counts_to_freqs): Simplify.
9398         (expensive_function_p): Use to_frequency.
9399         (propagate_unlikely_bbs_forward): Simplify.
9400         (determine_unlikely_bbs): Simplify.
9401         (estimate_bb_frequencies): Add hack to silence graphite issues.
9402         (compute_function_frequency): Use ipa counts.
9403         (pass_profile::execute): Update.
9404         (rebuild_frequencies): Use counts only.
9405         (force_edge_cold): Use counts only.
9406         * profile-count.c (profile_count::dump): Dump new count types.
9407         (profile_count::differs_from_p): Check compatiblity.
9408         (profile_count::to_frequency): New function.
9409         (profile_count::to_cgraph_frequency): New function.
9410         * profile-count.h (struct function): Declare.
9411         (enum profile_quality): Add profile_guessed_local and
9412         profile_guessed_global0.
9413         (class profile_proability): Decrease number of bits to 29;
9414         update from_reg_br_prob_note and to_reg_br_prob_note.
9415         (class profile_count: Update comment; decrease number of bits
9416         to 61. Check compatibility.
9417         (profile_count::compatible_p): New private member function.
9418         (profile_count::ipa_p): New member function.
9419         (profile_count::operator<): Handle global zero correctly.
9420         (profile_count::operator>): Handle global zero correctly.
9421         (profile_count::operator<=): Handle global zero correctly.
9422         (profile_count::operator>=): Handle global zero correctly.
9423         (profile_count::nonzero_p): New member function.
9424         (profile_count::force_nonzero): New member function.
9425         (profile_count::max): New member function.
9426         (profile_count::apply_scale): Handle IPA scalling.
9427         (profile_count::guessed_local): New member function.
9428         (profile_count::global0): New member function.
9429         (profile_count::ipa): New member function.
9430         (profile_count::to_frequency): Declare.
9431         (profile_count::to_cgraph_frequency): Declare.
9432         * profile.c (OVERLAP_BASE): Delete.
9433         (compute_frequency_overlap): Delete.
9434         (compute_branch_probabilities): Do not use compute_frequency_overlap.
9435         * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9436         * sched-ebb.c (rank): Use counts only.
9437         * shrink-wrap.c (handle_simple_exit): Use counts only.
9438         (try_shrink_wrapping): Use counts only.
9439         (place_prologue_for_one_component): Use counts only.
9440         * tracer.c (find_best_predecessor): Use to_frequency.
9441         (find_trace): Use to_frequency.
9442         (tail_duplicate): Use to_frequency.
9443         * trans-mem.c (expand_transaction): Do not update frequency.
9444         * tree-call-cdce.c: Do not update frequency.
9445         * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9446         (gimple_merge_blocks): Likewise.
9447         (gimple_split_edge): Likewise.
9448         (gimple_duplicate_sese_region): Likewise.
9449         (gimple_duplicate_sese_tail): Likewise.
9450         (move_sese_region_to_fn): Likewise.
9451         (gimple_account_profile_record): Likewise.
9452         (insert_cond_bb): Likewise.
9453         * tree-complex.c (expand_complex_div_wide): Likewise.
9454         * tree-eh.c (lower_resx): Update profile.
9455         * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9456         frequencies.
9457         (initialize_cfun): Do not initialize frequencies
9458         (freqs_to_counts): Delete.
9459         (copy_cfg_body): Ignore count parameter.
9460         (copy_body): Update.
9461         (expand_call_inline): Update count_max.
9462         (optimize_inline_calls): Update count_max.
9463         (tree_function_versioning): Update count_max.
9464         * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9465         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9466         frequency.
9467         * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9468         * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9469         (try_peel_loop): Likewise.
9470         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9471         to_frequency.
9472         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9473         (tree_transform_and_unroll_loop): Do not use frequencies
9474         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9475         Use reliable prediction only.
9476         * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9477         * tree-ssa-sink.c (select_best_block): Use to_frequency.
9478         * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9479         probability scaling.
9480         * tree-ssa-threadupdate.c (create_block_for_threading): Do
9481         not update frequency
9482         (any_remaining_duplicated_blocks): Likewise.
9483         (update_profile): Likewise.
9484         (estimated_freqs_path): Delete.
9485         (freqs_to_counts_path): Delete.
9486         (clear_counts_path): Delete.
9487         (ssa_fix_duplicate_block_edges): Likewise.
9488         (duplicate_thread_path): Likewise.
9489         * tree-switch-conversion.c (gen_inbound_check): Use counts.
9490         * tree-tailcall.c (decrease_profile): Do not update frequency.
9491         (eliminate_tail_call): Likewise.
9492         * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9493         * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9494         (optimize_mask_stores): Likewise.
9495         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9496         * ubsan.c (ubsan_expand_null_ifn): Update profile.
9497         (ubsan_expand_ptr_ifn): Update profile.
9498         * value-prof.c (gimple_ic): Simplify.
9499         * value-prof.h (gimple_ic): Update prototype.
9500         * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9501         * ipa-inline.c (compute_uninlined_call_time): Be sure that
9502         counts are nonzero.
9503         (want_inline_self_recursive_call_p): Likewise.
9504         (resolve_noninline_speculation): Only cummulate defined counts.
9505         (inline_small_functions): Use nonzero_p.
9506         (ipa_inline): Do not access freed node.
9508 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9510         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9511         Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9513 2017-11-03  Kito Cheng  <kito.cheng@gmail.com>
9515         * config/riscv/riscv.c (riscv_legitimize_move): Handle
9516         non-legitimate address.
9518 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9520         * config/rs6000/rs6000.md (*lt0_disi): Delete.
9521         (*lt0_<mode>di, *lt0_<mode>si): New.
9523 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9525         * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9526         TARGET_PAIRED_FLOAT.
9528 2017-11-03  Siddhesh Poyarekar  <siddhesh.poyarekar@linaro.org>
9529             Jim Wilson  <jim.wilson@linaro.org>
9531         * config/aarch64/aarch64-cores.def (saphira): New CPU.
9532         * config/aarch64/aarch64-tune.md: Regenerated.
9533         * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9534         * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9536 2017-11-03  Cupertino Miranda  <cmiranda@synopsys.com>
9538         * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9539         (arc_expand_prologue): Restore blink for millicode.
9540         * config/arc/linux.h (LINK_EH_SPEC): Defined.
9542 2017-11-03  Richard Sandiford  <richard.sandiford@linaro.org>
9544         PR target/82809
9545         * config/i386/i386.c (ix86_vector_duplicate_value): Use
9546         gen_vec_duplicate after forcing the scalar into a register.
9548 2017-11-02  Segher Boessenkool  <segher@kernel.crashing.org>
9550         * combine (try_combine): Print the insns input to try_combine to the
9551         dump file.
9553 2017-11-02  Steve Ellcey  <sellcey@cavium.com>
9555         PR target/79868
9556         * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9557         Remove second argument from aarch64_process_target_attr call.
9558         * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9559         Ditto.
9560         * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9561         field type.
9562         (aarch64_handle_attr_arch): Remove second argument.
9563         (aarch64_handle_attr_cpu): Ditto.
9564         (aarch64_handle_attr_tune): Ditto.
9565         (aarch64_handle_attr_isa_flags): Ditto.
9566         (aarch64_process_one_target_attr): Ditto.
9567         (aarch64_process_target_attr): Ditto.
9568         (aarch64_option_valid_attribute_p): Remove second argument.
9569         on aarch64_process_target_attr call.
9571 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9573         * diagnostic.c: Include "selftest-diagnostic.h".
9574         (selftest::assert_location_text): New function.
9575         (selftest::test_diagnostic_get_location_text): New function.
9576         (selftest::diagnostic_c_tests): Call it.
9578 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9580         * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9581         * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9582         (class selftest::test_diagnostic_context): Move to...
9583         * selftest-diagnostic.c: New file.
9584         * selftest-diagnostic.h: New file.
9586 2017-11-02  James Bowman  <james.bowman@ftdichip.com>
9588         * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9589         offset range for FT32B.
9590         * config/ft32/ft32.h: option "mcompress" enables relaxation.
9591         * config/ft32/ft32.md: Add TARGET_NOPM.
9592         * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9593         * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9595 2017-11-02  Wilco Dijkstra  <wdijkstr@arm.com>
9597         * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9599 2017-11-02  Jeff Law  <law@redhat.com>
9601         * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9602         virtual keyword on FINAL OVERRIDE members.
9604         * tree-ssa-propagate.h (ssa_propagation_engine): Group
9605         virtuals together.  Add virtual destructor.
9606         (substitute_and_fold_engine): Similarly.
9608 2017-11-02  Jan Hubicka  <hubicka@ucw.cz>
9610         * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9612 2017-11-02  Richard Biener  <rguenther@suse.de>
9614         PR tree-optimization/82795
9615         * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9617 2017-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9619         * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9620         gcc_SUN_LD_VERSION.
9621         (gcc_GAS_CHECK_FEATURE): Remove.
9622         * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9623         gcc_AC_INITFINI_ARRAY here.  Update for Solaris 11.4 changes.
9624         * configure: Regenerate.
9626 2017-11-02  Claudiu Zissulescu <claziss@synopsys.com>
9628         * config/arc/arc.c (hwloop_optimize): Account for empty
9629         body loops.
9631 2017-11-02  Richard Biener  <rguenther@suse.de>
9633         PR middle-end/82765
9634         * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9635         Truncate ARRAY_REF index and element size.
9637 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9639         * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9641 2017-11-01  Jeff Law  <law@redhat.com>
9643         * tree-ssa-ccp.c (ccp_folder): New class derived from
9644         substitute_and_fold_engine.
9645         (ccp_folder::get_value): New member function.
9646         (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9647         (ccp_fold_stmt): Remove prototype.
9648         (ccp_finalize): Call substitute_and_fold from the ccp_class.
9649         * tree-ssa-copy.c (copy_folder): New class derived from
9650         substitute_and_fold_engine.
9651         (copy_folder::get_value): Renamed from get_value.
9652         (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9653         * tree-vrp.c (vrp_folder): New class derived from
9654         substitute_and_fold_engine.
9655         (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9656         (vrp_folder::get_value): New member function.
9657         (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9658         (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9659         * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9660         provide a class interface to folder/substitute routines.
9661         (ssa_prop_fold_stmt_fn): Remove typedef.
9662         (ssa_prop_get_value_fn): Likewise.
9663         (subsitute_and_fold): Remove prototype.
9664         (replace_uses_in): Likewise.
9665         * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9666         Renamed from replace_uses_in.  Call the virtual member function
9667         (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9668         (substitute_and_fold_dom_walker): Remove initialization of
9669         data member entries for calbacks.  Add substitute_and_fold_engine
9670         member and initialize it.
9671         (substitute_and_fold_dom_walker::before_dom_children0: Use the
9672         member functions for get_value, replace_phi_args_in c
9673         replace_uses_in, and fold_stmt calls.
9674         (substitute_and_fold_engine::substitute_and_fold): Renamed from
9675         substitute_and_fold.  Remove assert.   Update ctor call.
9677         * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9678         (ssa_prop_visit_phi_fn): Likewise.
9679         (class ssa_propagation_engine): New class to provide an interface
9680         into ssa_propagate.
9681         * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9682         variable.
9683         (ssa_prop_visit_phi): Likewise.
9684         (ssa_propagation_engine::simulate_stmt): Moved into class.
9685         Call visit_phi/visit_stmt from the class rather than via
9686         file scoped static variables.
9687         (ssa_propagation_engine::simulate_block): Moved into class.
9688         (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9689         (ssa_propagation_engine::ssa_propagate): Similarly.  No longer
9690         set file scoped statics for the visit_stmt/visit_phi callbacks.
9691         * tree-complex.c (complex_propagate): New class derived from
9692         ssa_propagation_engine.
9693         (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9694         (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9695         (tree_lower_complex): Call ssa_propagate via the complex_propagate
9696         class.
9697         * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9698         ssa_propagation_engine.
9699         (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9700         (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9701         (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9702         * tree-ssa-copy.c (copy_prop): New class derived from
9703         ssa_propagation_engine.
9704         (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9705         (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9706         (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9707         * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9708         (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9709         (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9710         (execute_vrp): Call ssa_propagate from the vrp_prop class.
9712 2017-11-01  Jakub Jelinek  <jakub@redhat.com>
9714         PR rtl-optimization/82778
9715         PR rtl-optimization/82597
9716         * compare-elim.c (struct comparison): Add in_a_setter field.
9717         (find_comparison_dom_walker::before_dom_children): Remove killed
9718         bitmap and df_simulate_find_defs call, instead walk the defs.
9719         Compute last_setter and initialize in_a_setter.  Merge definitions
9720         with first initialization for a few variables.
9721         (try_validate_parallel): Use insn_invalid_p instead of
9722         recog_memoized.  Return insn rather than just the pattern.
9723         (try_merge_compare): Fix up comment.  Don't uselessly test if
9724         in_a is a REG_P.  Use cmp->in_a_setter instead of walking UD
9725         chains.
9726         (execute_compare_elim_after_reload): Remove df_chain_add_problem
9727         call.
9729 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9730             Alan Hayward  <alan.hayward@arm.com>
9731             David Sherwood  <david.sherwood@arm.com>
9733         * config/aarch64/aarch64.c (aarch64_rtx_costs): Use
9734         aarch64_hard_regno_nregs to get the number of registers
9735         in a mode.
9737 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9738             Alan Hayward  <alan.hayward@arm.com>
9739             David Sherwood  <david.sherwood@arm.com>
9741         * config/aarch64/constraints.md (Upl): Rename to...
9742         (Uaa): ...this.
9743         * config/aarch64/aarch64.md
9744         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9745         Update accordingly.
9747 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9748             Alan Hayward  <alan.hayward@arm.com>
9749             David Sherwood  <david.sherwood@arm.com>
9751         * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9752         (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9753         earlier in file.
9755 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9756             Alan Hayward  <alan.hayward@arm.com>
9757             David Sherwood  <david.sherwood@arm.com>
9759         * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9760         (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9761         (aarch64_evpc_dup): Generate rtl direcly, rather than using
9762         named expanders.
9763         (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9764         of a single element.
9765         * config/aarch64/iterators.md: Add a comment above the permute
9766         unspecs to say that they are generated directly by
9767         aarch64_expand_vec_perm_const.
9768         * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9770 2017-11-01  Nathan Sidwell  <nathan@acm.org>
9772         * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9774 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9776         * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9777         and medany code models, and describe what they do.
9779 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9781         Revert accidental duplicate:
9783         * combine.c (can_change_dest_mode): Reject changes in
9784         REGMODE_NATURAL_SIZE.
9786 2017-11-01  Segher Boessenkool  <segher@kernel.crashing.org>
9788         PR rtl-optimization/64682
9789         PR rtl-optimization/69567
9790         PR rtl-optimization/69737
9791         PR rtl-optimization/82683
9792         * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9793         register mentioned in the note, drop the note, unless it came from I3,
9794         in which case it should go to I3 again.
9796 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9798         * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9799         and return false if not.
9800         (clear_bytes_written_by, live_bytes_read): Update accordingly.
9802 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9804         * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9805         range is known to be empty.
9807 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9808             Alan Hayward  <alan.hayward@arm.com>
9809             David Sherwood  <david.sherwood@arm.com>
9811         * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9812         and CONST_VECTOR_NUNITS instead of computing the number of units from
9813         the byte sizes of the vector and element.
9814         (simplify_binary_operation_1): Likewise.
9815         (simplify_const_binary_operation): Likewise.
9816         (simplify_ternary_operation): Likewise.
9818 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9819             Alan Hayward  <alan.hayward@arm.com>
9820             David Sherwood  <david.sherwood@arm.com>
9822         * var-tracking.c (INT_MEM_OFFSET): Replace with...
9823         (int_mem_offset): ...this new function.
9824         (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9825         (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9826         (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9827         Update accordingly.
9829 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9830             Alan Hayward  <alan.hayward@arm.com>
9831             David Sherwood  <david.sherwood@arm.com>
9833         * lower-subreg.c (interesting_mode_p): New function.
9834         (compute_costs, find_decomposable_subregs, decompose_register)
9835         (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9836         (resolve_clobber, dump_choices): Use it.
9838 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9839             Alan Hayward  <alan.hayward@arm.com>
9840             David Sherwood  <david.sherwood@arm.com>
9842         * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
9844 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9845             Alan Hayward  <alan.hayward@arm.com>
9846             David Sherwood  <david.sherwood@arm.com>
9848         * alias.c (find_base_value, find_base_term): Only process integer
9849         truncations.  Check the precision rather than the size.
9851 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9852             Alan Hayward  <alan.hayward@arm.com>
9853             David Sherwood  <david.sherwood@arm.com>
9855         * machmode.h (is_narrower_int_mode): New function
9856         * optabs.c (expand_float, expand_fix): Use it.
9857         * dwarf2out.c (rotate_loc_descriptor): Likewise.
9859 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9860             Alan Hayward  <alan.hayward@arm.com>
9861             David Sherwood  <david.sherwood@arm.com>
9863         * rtl.h (narrower_subreg_mode): New function.
9864         * ira-color.c (update_costs_from_allocno): Use it.
9866 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9867             Alan Hayward  <alan.hayward@arm.com>
9868             David Sherwood  <david.sherwood@arm.com>
9870         * optabs-query.h (convert_optab_p): New function, split out from...
9871         (convert_optab_handler): ...here.
9872         (widening_optab_handler): Delete.
9873         (find_widening_optab_handler): Remove permit_non_widening parameter.
9874         (find_widening_optab_handler_and_mode): Likewise.  Provide an
9875         override that operates on mode class wrappers.
9876         * optabs-query.c (widening_optab_handler): Delete.
9877         (find_widening_optab_handler_and_mode): Remove permit_non_widening
9878         parameter.  Assert that the two modes are the same class and that
9879         the "from" mode is narrower than the "to" mode.  Use
9880         convert_optab_handler instead of widening_optab_handler.
9881         * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9882         instead of widening_optab_handler.
9883         * expr.c (expand_expr_real_2): Update calls to
9884         find_widening_optab_handler.
9885         * optabs.c (expand_widen_pattern_expr): Likewise.
9886         (expand_binop_directly): Take the insn_code as a parameter.
9887         (expand_binop): Only call find_widening_optab_handler for
9888         conversion optabs; use optab_handler otherwise.  Update calls
9889         to find_widening_optab_handler and expand_binop_directly.
9890         Use convert_optab_handler instead of widening_optab_handler.
9891         * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9892         find_widening_optab_handler and use scalar_mode rather than
9893         machine_mode.
9894         (convert_plusminus_to_widen): Likewise.
9896 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9897             Alan Hayward  <alan.hayward@arm.com>
9898             David Sherwood  <david.sherwood@arm.com>
9900         * machmode.h (fixed_size_mode): New class.
9901         * rtl.h (get_pool_mode): Return fixed_size_mode.
9902         * gengtype.c (main): Add fixed_size_mode.
9903         * target.def (get_raw_result_mode): Return a fixed_size_mode.
9904         (get_raw_arg_mode): Likewise.
9905         * doc/tm.texi: Regenerate.
9906         * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9907         * targhooks.c (default_get_reg_raw_mode): Likewise.
9908         * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9909         * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9910         * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9911         (msp430_get_raw_result_mode): Likewise.
9912         * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9913         * dbxout.c (dbxout_parms): Require fixed-size modes.
9914         * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9915         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9916         * omp-low.c (lower_oacc_reductions): Likewise.
9917         * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
9918         (simplify_subreg): Update accordingly.
9919         * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
9920         (force_const_mem): Update accordingly.  Return NULL_RTX for modes
9921         that aren't fixed-size.
9922         (get_pool_mode): Return a fixed_size_mode.
9923         (output_constant_pool_2): Take a fixed_size_mode.
9925 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9926             Alan Hayward  <alan.hayward@arm.com>
9927             David Sherwood  <david.sherwood@arm.com>
9929         * doc/rtl.texi (vec_series): Document.
9930         (const): Say that the operand can be a vec_series.
9931         * rtl.def (VEC_SERIES): New rtx code.
9932         * rtl.h (const_vec_series_p_1): Declare.
9933         (const_vec_series_p): New function.
9934         * emit-rtl.h (gen_const_vec_series): Declare.
9935         (gen_vec_series): Likewise.
9936         * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
9937         (gen_vec_series): Likewise.
9938         * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
9939         * simplify-rtx.c (simplify_unary_operation): Handle negations
9940         of vector series.
9941         (simplify_binary_operation_series): New function.
9942         (simplify_binary_operation_1): Use it.  Handle VEC_SERIES.
9943         (test_vector_ops_series): New function.
9944         (test_vector_ops): Call it.
9945         * config/powerpcspe/altivec.md (altivec_lvsl): Use
9946         gen_const_vec_series.
9947         (altivec_lvsr): Likewise.
9948         * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
9950 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9951             Alan Hayward  <alan.hayward@arm.com>
9952             David Sherwood  <david.sherwood@arm.com>
9954         * doc/rtl.texi (const): Update description of address constants.
9955         Say that vector constants are allowed too.
9956         * common.md (E, F): Use CONSTANT_P instead of checking for
9957         CONST_VECTOR.
9958         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9959         checking for CONST_VECTOR.
9960         * expmed.c (make_tree): Use build_vector_from_val for a CONST
9961         VEC_DUPLICATE.
9962         * expr.c (expand_expr_real_2): Check for vector modes instead
9963         of checking for CONST_VECTOR.
9964         * rtl.h (const_vec_p): New function.
9965         (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
9966         (unwrap_const_vec_duplicate): Handle them here too.
9968 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9969             David Malcolm  <dmalcolm@redhat.com>
9970             Alan Hayward  <alan.hayward@arm.com>
9971             David Sherwood  <david.sherwood@arm.com>
9973         * rtl.h (vec_duplicate_p): New function.
9974         * selftest-rtl.c (assert_rtx_eq_at): New function.
9975         * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
9976         (assert_rtx_eq_at): Declare.
9977         * selftest.h (selftest::simplify_rtx_c_tests): Declare.
9978         * selftest-run-tests.c (selftest::run_tests): Call it.
9979         * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
9980         (simplify_unary_operation_1): Recursively handle vector duplicates.
9981         (simplify_binary_operation_1): Likewise.  Handle VEC_SELECTs of
9982         vector duplicates.
9983         (simplify_subreg): Handle subregs of vector duplicates.
9984         (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
9985         (selftest::simplify_rtx_c_tests): New functions.
9987 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9988             Alan Hayward  <alan.hayward@arm.com>
9989             David Sherwood  <david.sherwood@arm.com>
9991         * emit-rtl.h (gen_const_vec_duplicate): Declare.
9992         (gen_vec_duplicate): Likewise.
9993         * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
9994         out from...
9995         (gen_const_vector): ...here.
9996         (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
9997         (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
9998         whose elements are all equal.
9999         * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
10000         * simplify-rtx.c (simplify_const_unary_operation): Likewise.
10001         (simplify_relational_operation): Likewise.
10002         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
10003         Likewise.
10004         (aarch64_simd_dup_constant): Use gen_vec_duplicate.
10005         (aarch64_expand_vector_init): Likewise.
10006         * config/arm/arm.c (neon_vdup_constant): Likewise.
10007         (neon_expand_vector_init): Likewise.
10008         (arm_expand_vec_perm): Use gen_const_vec_duplicate.
10009         (arm_block_set_unaligned_vect): Likewise.
10010         (arm_block_set_aligned_vect): Likewise.
10011         * config/arm/neon.md (neon_copysignf<mode>): Likewise.
10012         * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
10013         (expand_vec_perm_even_odd_pack): Likewise.
10014         (ix86_vector_duplicate_value): Use gen_vec_duplicate.
10015         * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
10016         * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
10017         gen_const_vec_duplicate.
10018         * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
10019         * config/mips/mips.c (mips_gen_const_int_vector): Use
10020         gen_const_vec_duplicate.
10021         (mips_expand_vector_init): Use CONST0_RTX.
10022         * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
10023         (define_split): Use gen_const_vec_duplicate.
10024         * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
10025         (define_split): Use gen_const_vec_duplicate.
10026         * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
10027         (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
10028         * config/spu/spu.c (spu_const): Likewise.
10030 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
10031             Alan Hayward  <alan.hayward@arm.com>
10032             David Sherwood  <david.sherwood@arm.com>
10034         * combine.c (can_change_dest_mode): Reject changes in
10035         REGMODE_NATURAL_SIZE.
10037 2017-10-31  Sandra Loosemore  <sandra@codesourcery.com>
10039         * configure.ac (--enable-libssp): New.
10040         (gcc_cv_libc_provides_ssp): Check for explicit setting before
10041         trying to determine target-specific default.  Adjust indentation.
10042         * configure: Regenerated.
10043         * doc/install.texi (Configuration): Expand --disable-libssp
10044         documentation.
10046 2017-10-31  Daniel Santos  <daniel.santos@pobox.com>
10048         config/i386/i386.c (ix86_expand_epilogue): Correct stack
10049         calculation.
10051 2017-10-31  Martin Jambor  <mjambor@suse.cz>
10053         PR c++/81702
10054         * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
10056 2017-10-31  David Malcolm  <dmalcolm@redhat.com>
10058         * auto-profile.c (autofdo_source_profile::read): Use
10059         UNKNOWN_LOCATION rather than 0.
10060         * diagnostic-core.h (warning_at_rich_loc): Rename to...
10061         (warning_at): ...this overload.
10062         (warning_at_rich_loc_n): Rename to...
10063         (warning_n): ...this overload.
10064         (error_at_rich_loc): Rename to...
10065         (error_at): ...this overload.
10066         (pedwarn_at_rich_loc): Rename to...
10067         (pedwarn): ...this overload.
10068         (permerror_at_rich_loc): Rename to...
10069         (permerror): ...this overload.
10070         (inform_at_rich_loc): Rename to...
10071         (inform): ...this overload.
10072         * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
10073         (diagnostic_n_impl_richloc): Rename to...
10074         (diagnostic_n_impl): ...this rich_location *-based decl.
10075         (inform_at_rich_loc): Rename to...
10076         (inform): ...this, and add an assertion.
10077         (inform_n): Update for removal of location_t-based diagnostic_n_impl.
10078         (warning_at_rich_loc): Rename to...
10079         (warning_at): ...this, and add an assertion.
10080         (warning_at_rich_loc_n): Rename to...
10081         (warning_n): ...this, and add an assertion.
10082         (warning_n): Update location_t-based implementation for removal of
10083         location_t-based diagnostic_n_impl.
10084         (pedwarn_at_rich_loc): Rename to...
10085         (pedwarn): ...this, and add an assertion.
10086         (permerror_at_rich_loc): Rename to...
10087         (permerror): ...this, and add an assertion.
10088         (error_n): Update for removal of location_t-based diagnostic_n_impl.
10089         (error_at_rich_loc): Rename to...
10090         (error_at): ...this, and add an assertion.
10091         * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
10092         (driver::do_spec_on_infiles): Likewise.
10093         * substring-locations.c (format_warning_va): Update for renaming
10094         of inform_at_rich_loc.
10096 2017-10-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
10098         * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
10099         _Float<N>X built-in functions so that the variant without the
10100         "__builtin_" prefix is only enabled for the GNU C and Objective C
10101         languages when they are in non-strict ANSI/ISO mode.
10102         (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
10103         * target.def (floatn_builtin_p): Add a target hook to control
10104         whether _Float<N> and _Float<N>X built-in functions without the
10105         "__builtin_" prefix are enabled, and return true for C and
10106         Objective C in the default hook.  Include langhooks.h in
10107         targhooks.c.
10108         * targhooks.h (default_floatn_builtin_p): Likewise.
10109         * targhooks.c (default_floatn_builtin_p): Likewise.
10110         * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
10111         floatn_builtin_p target hook.
10112         * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
10114 2017-10-31  Matthew Fortune  <matthew.fortune@imgtec.com>
10115             Eric Botcazou  <ebotcazou@adacore.com>
10117         PR rtl-optimization/81803
10118         * lra-constraints.c (curr_insn_transform): Also reload the whole
10119         register for a strict subreg no wider than a word if this is for
10120         a WORD_REGISTER_OPERATIONS target.
10122 2017-10-31  Jason Merrill  <jason@redhat.com>
10124         * gdbinit.in: Skip over inlines from timevar.h.
10126 2017-10-31  Martin Liska  <mliska@suse.cz>
10128         * doc/gcov.texi: Document new option.
10129         * gcov.c (print_usage): Likewise print it.
10130         (process_args): Support the argument.
10131         (format_count): New function.
10132         (format_gcov): Use the function.
10134 2017-10-31  Martin Liska  <mliska@suse.cz>
10136         * gcov.c (struct name_map): do not use typedef.
10137         Define operator== and operator<.
10138         (name_search): Remove.
10139         (name_sort): Remove.
10140         (main): Do not allocate names.
10141         (process_file): Add vertical space.
10142         (generate_results): Use std::find.
10143         (release_structures): Do not release memory.
10144         (find_source): Use std::find.
10146 2017-10-31  Martin Liska  <mliska@suse.cz>
10148         * gcov.c (struct line_info): Remove it's typedef.
10149         (line_info::line_info): Add proper ctor.
10150         (line_info::has_block): Do not use a typedef.
10151         (struct source_info): Do not use typedef.
10152         (circuit): Likewise.
10153         (get_cycles_count): Likewise.
10154         (output_intermediate_file): Iterate via vector iterator.
10155         (add_line_counts): Use std::vector methods.
10156         (accumulate_line_counts): Likewise.
10157         (output_lines): Likewise.
10159 2017-10-31  Martin Liska  <mliska@suse.cz>
10161         * gcov.c (struct source_info): Remove typedef.
10162         (source_info::source_info): Add proper ctor.
10163         (accumulate_line_counts): Use struct, not it's typedef.
10164         (output_gcov_file): Likewise.
10165         (output_lines): Likewise.
10166         (main): Do not allocate an array.
10167         (output_intermediate_file): Use size of vector container.
10168         (process_file): Resize the vector.
10169         (generate_results): Do not preallocate, use newly added vector
10170         lines.
10171         (release_structures): Do not release sources.
10172         (find_source): Use vector methods.
10173         (add_line_counts): Do not use typedef.
10175 2017-10-31  Martin Liska  <mliska@suse.cz>
10177         * doc/gcov.texi: Document that.
10178         * gcov.c (add_line_counts): Mark lines with a non-executed
10179         statement.
10180         (output_line_beginning): Handle such lines.
10181         (output_lines): Pass new argument.
10182         (output_intermediate_file): Print it in intermediate format.
10184 2017-10-31  Martin Liska  <mliska@suse.cz>
10186         * color-macros.h: New file.
10187         * diagnostic-color.c: Factor out color related to macros to
10188         color-macros.h.
10189         * doc/gcov.texi: Document -k option.
10190         * gcov.c (INCLUDE_STRING): Include string.h.
10191         (print_usage): Add -k option.
10192         (process_args): Parse it.
10193         (pad_count_string): New function.
10194         (output_line_beginning): Likewise.
10195         (DEFAULT_LINE_START): New macro.
10196         (output_lines): Support color output.
10198 2017-10-31  Martin Liska  <mliska@suse.cz>
10200         PR gcov-profile/82633
10201         * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
10202         their interaction with GCOV infrastructure.
10203         * configure.ac: Add -fkeep-{inline,static}-functions to
10204         coverage_flags.
10205         * configure: Regenerate.
10207 2017-10-31  Uros Bizjak  <ubizjak@gmail.com>
10209         PR target/82772
10210         * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
10212 2017-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
10214         PR target/82674
10215         * config/rs6000/rs6000.md (allocate_stack): Force update interval
10216         into a register if it does not fit into an immediate offset field.
10218 2017-10-31  Olivier Hainque  <hainque@adacore.com>
10220         * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
10222 2017-10-31  Julia Koval  <julia.koval@intel.com>
10224         * config.gcc: Add gfniintrin.h.
10225         * config/i386/gfniintrin.h: New.
10226         * config/i386/i386-builtin-types.def
10227         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10228         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
10229         __builtin_ia32_vgf2p8affineinvqb_v32qi,
10230         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
10231         __builtin_ia32_vgf2p8affineinvqb_v16qi,
10232         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
10233         * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
10234         V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
10235         V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
10236         V64QI_FTYPE_V64QI_V64QI_INT): New types.
10237         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
10238         * config/i386/immintrin.h: Include gfniintrin.h.
10239         * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
10241 2017-10-30  Eric Botcazou  <ebotcazou@adacore.com>
10243         * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10245 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10247         * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10248         (arm_ashldi3_1bit): Remove pattern.
10249         (ashrdi3): Remove shift by 1 expansion.
10250         (arm_ashrdi3_1bit): Remove pattern.
10251         (lshrdi3): Remove shift by 1 expansion.
10252         (arm_lshrdi3_1bit): Remove pattern.
10253         * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10254         cost of ashldi3 by 1.
10255         * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10256         (<shift>di3_neon): Likewise.
10258 2017-10-30  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
10260         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10261         both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10262         and (*aarch64_simd_mov<VQ:mode>).
10263         (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10264         pattern alternative.
10265         (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10266         attributes to match pattern alternative.
10268 2017-10-30  Steven Munroe  <munroesj@gcc.gnu.org>
10270         * config.gcc (powerpc*-*-*): Add emmintrin.h.
10271         * config/rs6000/emmintrin.h: New file.
10272         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10274 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10276         * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10277         * (movdi_vfp_cortexa8): Remove pattern.
10279 2017-10-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10281         * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10282         etc. reference.
10283         (Specific, alpha*-dec-osf5.1): Remove.
10284         (Specific, mips-sgi-irix5): Remove.
10285         (Specific, mips-sgi-irix6): Remove.
10287 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10289         PR middle-end/22141
10290         * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10291         arguments to clear_bit_region_be.
10293 2017-10-30  Jim Wilson  <wilson@tuliptree.org>
10295         * gimplify.c: Include memmodel.h.
10297 2017-10-30  Martin Jambor  <mjambor@suse.cz>
10299         * omp-grid.c (grid_attempt_target_gridification): Also insert a
10300         condition whether loop should be executed at all.
10302 2017-10-30  Will Schmidt  <will_schmidt@vnet.ibm.com>
10304         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10305         gimple folding of vec_madd() intrinsics.
10306         * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10307         fmav8hi4.  (altivec_vmladduhm): Rename to fmav8hi4.
10308         * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10310 2017-10-30  Richard Biener  <rguenther@suse.de>
10312         PR tree-optimization/82762
10313         Revert
10314         2017-10-23  Richard Biener  <rguenther@suse.de>
10316         PR tree-optimization/82129
10317         Revert
10318         2017-08-01  Richard Biener  <rguenther@suse.de>
10320         PR tree-optimization/81181
10321         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10322         (compute_antic): ... end of iteration here.
10324 2017-10-30  Joseph Myers  <joseph@codesourcery.com>
10326         * doc/invoke.texi (C Dialect Options): Document -std=c17,
10327         -std=iso9899:2017 and -std=gnu17.
10328         * doc/standards.texi (C Language): Document C17 support.
10329         * doc/cpp.texi (Overview): Mention -std=c17.
10330         (Standard Predefined Macros): Document C11 and C17 values of
10331         __STDC_VERSION__.  Do not refer to C99 support as incomplete.
10332         * doc/extend.texi (Inline): Do not list individual options for
10333         standards newer than C99.
10334         * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10335         "GNU C17".
10336         * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10337         language name.
10339 2017-10-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
10341         * asan.c (asan_finish_file): Align asan globals array by shadow
10342         granularity.
10344 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10346         PR middle-end/22141
10347         * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10348         (struct store_immediate_info): Add bitregion_start and bitregion_end
10349         fields.
10350         (store_immediate_info::store_immediate_info): Add brs and bre
10351         arguments and initialize bitregion_{start,end} from those.
10352         (struct merged_store_group): Add bitregion_start, bitregion_end,
10353         align_base and mask fields.  Drop unnecessary struct keyword from
10354         struct store_immediate_info.  Add do_merge method.
10355         (clear_bit_region_be): Use memset instead of loop storing zeros.
10356         (merged_store_group::do_merge): New method.
10357         (merged_store_group::merge_into): Use do_merge.  Allow gaps in between
10358         stores as long as the surrounding bitregions have no gaps.
10359         (merged_store_group::merge_overlapping): Use do_merge.
10360         (merged_store_group::apply_stores): Test that bitregion_{start,end}
10361         is byte aligned, rather than requiring that start and width are
10362         byte aligned.  Drop unnecessary struct keyword from
10363         struct store_immediate_info.  Allocate and populate also mask array.
10364         Make start of the arrays relative to bitregion_start rather than
10365         start and size them according to bitregion_{end,start} difference.
10366         (struct imm_store_chain_info): Drop unnecessary struct keyword from
10367         struct store_immediate_info.
10368         (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10369         (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10370         struct keyword from struct store_immediate_info.
10371         (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10372         between stores as long as the surrounding bitregions have no gaps.
10373         Formatting fixes.
10374         (struct split_store): Add orig non-static data member.
10375         (split_store::split_store): Initialize orig to false.
10376         (find_constituent_stmts): Return store_immediate_info *, non-NULL
10377         if there is exactly a single original stmt.  Change stmts argument
10378         to pointer from reference, if NULL, don't push anything to it.  Add
10379         first argument, use it to optimize skipping over orig stmts that
10380         are known to be before bitpos already.  Simplify.
10381         (split_group): Return unsigned int count how many stores are or
10382         would be needed rather than a bool.  Add allow_unaligned argument.
10383         Change split_stores argument from reference to pointer, if NULL,
10384         only do a dry run computing how many stores would be produced.
10385         Rewritten algorithm to use both alignment and misalign if
10386         !allow_unaligned and handle bitfield stores with gaps.
10387         (imm_store_chain_info::output_merged_store): Set start_byte_pos
10388         from bitregion_start instead of start.  Compute allow_unaligned
10389         here, if true, do 2 split_group dry runs to compute which one
10390         produces fewer stores and prefer aligned if equal.  Punt if
10391         new count is bigger or equal than original before emitting any
10392         statements, rather than during that.  Remove no longer needed
10393         new_ssa_names tracking.  Replace num_stmts with
10394         split_stores.length ().  Use 32-bit stack allocated entries
10395         in split_stores auto_vec.  Try to reuse original store lhs/rhs1
10396         if possible.  Handle bitfields with gaps.
10397         (pass_store_merging::execute): Ignore bitsize == 0 stores.
10398         Compute bitregion_{start,end} for the stores and construct
10399         store_immediate_info with that.  Formatting fixes.
10401 2017-10-30  Uros Bizjak  <ubizjak@gmail.com>
10403         PR target/82725
10404         * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10405         UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10407 2017-10-29  Jim Wilson  <wilson@tuliptree.org>
10409         * gimplify.c: Include tm_p.h.
10411         * common.opt (gcoff): Re-add as ignored option.
10412         (gcoff1, gcoff2, gcoff3): Likewise.
10414         * Makefile.in (OBJS): Delete sdbout.o.
10415         (GTFILES): Delete $(srcdir)/sdbout.c.
10416         * debug.h: Delete sdb_debug_hooks.
10417         * final.c: Delete sdbout.h include.
10418         (final_scan_insn): Delete SDB_DEBUG check.
10419         (rest_of_clean_state): Likewise.
10420         * output.h: Delete sdb_begin_function_line.
10421         * sdbout.c: Delete.
10422         * sdbout.h: Delete.
10423         * toplev.c: Delete sdbout.h include.
10424         (process_options): Delete SDB_DEBUG check.
10425         * tree-core.h (tree_type_common): Delete pointer field of
10426         tree_type_symtab.
10427         * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10428         TYPE_SYMTAB_POINTER.
10429         * tree.h (TYPE_SYMTAB_POINTER): Delete.
10430         (TYPE_SYMTAB_IS_POINTER): Delete.
10431         (TYPE_SYMTAB_IS_DIE): Renumber.
10432         * xcoffout.c: Refer to former sdbout.c file.
10433         (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10435         * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10436         * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10437         (Debugging Options): Delete -gcoff.
10438         (-gstabs): Delete SDB reference.
10439         (-gcoff): Delete.
10440         (-gcoff@var{level}): Delete.
10441         * doc/passes.texi (Debugging information output): Delete SDB and
10442         sdbout.c references.
10443         * doc/tm.texi: Regenerate.
10444         * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10445         (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10446         references.
10447         (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10448         (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10449         (SDB_DEBUGGING_INFO): Delete.
10450         (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10451         SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10452         * target.def (output_source_filename): Delete COFF reference.
10454         * common.opt (gcoff): Delete.
10455         (gxcoff+): Update Negative chain.
10456         * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10457         SDB_DEBUG.
10458         * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10459         * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10460         * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10461         and SDB references.
10462         (expand_function_start): Change sdb reference to past tense.
10463         (expand_function_end): Change sdb reference to past tense.
10464         * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10465         * opts.c (debug_type_names): Delete coff entry.
10466         (common_handle_option): Delete OPT_gcoff case.
10467         * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10469         * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10470         * config/cris/cris.h: Delete SDB reference in comment.
10471         * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10472         (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10473         * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10474         * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10475         to past tense.
10476         (ix86_expand_prologue): Likewise.
10477         * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10478         * config/ia64/ia64.h: Likewise.
10479         * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10480         * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10481         support.
10482         * config/mmix/mmix.h: Likewise.
10483         * config/nds32/nds32.c: Likewise.
10484         * config/stormy/storym16.h: Likewise.
10485         * config/visium/visium.h: Likewise.
10486         * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10488 2017-10-28  Sandra Loosemore  <sandra@codesourcery.com>
10490         * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10491         * config/nios2/nios2.c (nios2_initial_elimination_offset):  Make
10492         FRAME_POINTER_REGNUM point at high end of local var area.
10494 2017-10-27  Eric Botcazou  <ebotcazou@adacore.com>
10496         * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10497         Move comment around.  Do not reset best_edge for a copiable
10498         destination if the copy would cause a partition change.
10499         (better_edge_p): Remove redundant check.
10501 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10503         * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10505 2017-10-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
10507         * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10508         for math functions that have _Float<N> and _Float<N>X variants.
10509         (mathfn_built_in_2): Add support for math functions that have
10510         _Float<N> and _Float<N>X variants.
10511         (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10512         (expand_builtin_mathfn_ternary): Add support for fma with
10513         _Float<N> and _Float<N>X variants.
10514         (expand_builtin): Likewise.
10515         (fold_builtin_3): Likewise.
10516         * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10517         create math function _Float<N> and _Float<N>X variants as external
10518         library builtins.
10519         (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10520         DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10521         the __builtin_ prefix and if not strict ansi, without the prefix.
10522         (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10523         (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10524         (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10525         (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10526         (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10527         (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10528         * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10529         function signatures for fma _Float<N> and _Float<N>X variants.
10530         (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10531         (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10532         (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10533         (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10534         (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10535         (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10536         * gencfn-macros.c (print_case_cfn): Add support for math functions
10537         that have _Float<N> and _Float<N>X variants.
10538         (print_define_operator_list): Likewise.
10539         (fltfn_suffixes): Likewise.
10540         (main): Likewise.
10541         * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10542         for math functions that have _Float<N> and _Float<N>X variants.
10543         (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10544         and _Float<N>X variants.
10545         (COPYSIGN): Likewise.
10546         (FMIN): Likewise.
10547         (FMAX): Likewise.
10548         * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10549         copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10550         variants.
10551         (integer_valued_read_call_p): Likewise.
10552         * fold-const-call.c (fold_const_call_ss): Likewise.
10553         (fold_const_call_sss): Add support for copysign, fmin, and fmax
10554         _Float<N> and _Float<N>X variants.
10555         (fold_const_call_ssss): Add support for fma _Float<N> and
10556         _Float<N>X variants.
10557         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10558         support for copysign and fma _Float<N> and _Float<N>X variants.
10559         (backprop::process_builtin_call_use): Likewise.
10560         * tree-call-cdce.c (can_test_argument_range); Add support for
10561         sqrt _Float<N> and _Float<N>X variants.
10562         (edom_only_function): Likewise.
10563         (get_no_error_domain): Likewise.
10564         * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10565         * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10566         copysign _Float<N> and _Float<N>X variants.
10567         * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10568         handled by machine independent code.
10569         (FMAF128): Likewise.
10570         * doc/cpp.texi (Common Predefined Macros): Document defining
10571         __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10572         fma _Float<N> and _Float<N>X variants.
10574 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10576         PR target/82692
10577         * config/i386/i386-modes.def (CCFPU): Remove definition.
10578         * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10579         (ix86_cc_modes_compatible): Ditto.
10580         (ix86_expand_carry_flag_compare): Ditto.
10581         (ix86_expand_int_movcc): Ditto.
10582         (ix86_expand_int_addcc): Ditto.
10583         (ix86_reverse_condition): Ditto.
10584         (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10585         Return true/false for unordered/ordered fp comparisons.
10586         (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10587         (ix86_prepare_fp_compare_args): Update for rename.
10588         (ix86_expand_fp_compare): Update for rename.  Generate unordered
10589         compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10590         (ix86_expand_sse_compare_and_jump): Ditto.
10591         * config/i386/predicates.md (fcmov_comparison_operator):
10592         Remove CCFPU mode handling.
10593         (ix86_comparison_operator): Ditto.
10594         (ix86_carry_flag_operator): Ditto.
10595         * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10596         (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10597         (*cmpu<mode>_cc_i387): Ditto.
10598         (FPCMP): Remove mode iterator.
10599         (unord): Remove mode attribute.
10600         (unord_subst): New define_subst transformation
10601         (unord): New define_subst attribute.
10602         (unordered): Ditto.
10603         (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10604         (*cmpi<unord>xf_i387): Ditto.
10605         * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10606         from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10607         using unord_subst transformation.
10608         * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10609         (round_saeonly): Also handle CCFP mode.
10610         * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10611         Remove UNSPEC_SAHF unspec handling.
10613 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10615         * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10617 2017-10-27  Jeff Law  <law@redhat.com>
10619         * gimple-ssa-sprintf.c: Include domwalk.h.
10620         (class sprintf_dom_walker): New class, derived from dom_walker.
10621         (sprintf_dom_walker::before_dom_children): New function.
10622         (struct call_info): Moved into sprintf_dom_walker class
10623         (compute_formath_length, handle_gimple_call): Likewise.
10624         (sprintf_length::execute): Call the dominator walker rather
10625         than walking the statements.
10627         * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10628         gimple statement locations.
10629         (check_array_bounds): Corresponding changes.  Get the statement's
10630         location directly from wi->stmt.
10632 2017-10-27  Palmer Dabbelt  <palmer@dabbelt.com>
10634         PR target/82717
10635         * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10637 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10639         * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10640         X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10642 2017-10-27  Jakub Jelinek  <jakub@redhat.com>
10644         PR target/82703
10645         * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10646         * config/i386/i386.c (maybe_get_pool_constant): Removed.
10647         (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10648         maybe_get_pool_constant.
10649         * config/i386/predicates.md (zero_extended_scalar_load_operand):
10650         Likewise.
10652 2017-10-27  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10654         * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10655         2.26 caveat.  Update gas and gld versions.
10656         (Specific, *-*-solaris2*): Update binutils version.  Remove caveat
10657         reference.
10659 2017-10-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10661         * cgraph.h (set_malloc_flag): Declare.
10662         * cgraph.c (set_malloc_flag_1): New function.
10663         (set_malloc_flag): Likewise.
10664         * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10665         * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10666         false.
10667         (read_ipa_call_summary): Add support for reading is_return_callee.
10668         (write_ipa_call_summary): Stream is_return_callee.
10669         * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10670         * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10671         ipa-prop.h, ipa-fnsummary.h.
10672         (pure_const_names): Change to static.
10673         (malloc_state_e): Define.
10674         (malloc_state_names): Define.
10675         (funct_state_d): Add field malloc_state.
10676         (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10677         (check_retval_uses): New function.
10678         (malloc_candidate_p): Likewise.
10679         (analyze_function): Add support for malloc attribute.
10680         (pure_const_write_summary): Stream malloc_state.
10681         (pure_const_read_summary): Add support for reading malloc_state.
10682         (dump_malloc_lattice): New function.
10683         (propagate_malloc): New function.
10684         (warn_function_malloc): New function.
10685         (ipa_pure_const::execute): Call propagate_malloc and
10686         ipa_free_fn_summary.
10687         (pass_local_pure_const::execute): Add support for malloc attribute.
10688         * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10689         * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10691 2017-10-27  Martin Liska  <mliska@suse.cz>
10693         PR gcov-profile/82457
10694         * doc/invoke.texi: Document that one needs a non-strict ISO mode
10695         for fork-like functions to be properly instrumented.
10697 2017-10-27  Richard Biener  <rguenther@suse.de>
10699         PR middle-end/81659
10700         * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10701         info when we redirected EH.
10703 2017-10-26  Michael Collison  <michael.collison@arm.com>
10705         * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10706         New pattern.
10707         (<optab>_trunchf<GPI:mode>2: New pattern.
10708         (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10709         * config/aarch64/iterators.md (wv): New mode attribute.
10710         (vf, VF): New mode attributes.
10711         (vgp, VGP): New mode attributes.
10712         (s): Update attribute with SImode and DImode prefixes.
10714 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10716         * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10717         * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10718         * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10719         (nios2_option_overide): Initialize it.  Don't allow R0-relative
10720         addressing with PIC.
10721         (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10722         (nios2_symbolic_constant_p): Likewise.
10723         (nios2_legitimate_address_p): Likewise.
10724         (nios2_r0rel_section_name_p): New.
10725         (nios2_symbol_ref_in_r0rel_data_p): New.
10726         (nios2_emit_move_sequence): Handle r0rel_constant_p.
10727         (r0rel_constant_p): New.
10728         (nios2_print_operand_address): Handle r0rel_constant_p.
10729         (nios2_cdx_narrow_form_p): Likewise.
10730         * config/nios2/nios2.opt (mr0rel-sec=): New option.
10731         * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10732         (Nios II Options): Document -mr0rel-sec.
10734 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10736         * config/nios2/nios2.c: Include xregex.h.
10737         (nios2_gprel_sec_regex): New.
10738         (nios2_option_overide): Initialize it.  Don't allow GP-relative
10739         addressing with PIC.
10740         (nios2_small_section_name_p): Check for regex match.
10741         * config/nios2/nios2.opt (mgprel-sec=): New option.
10742         * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10743         (Nios II Options): Document -mgprel-sec.
10745 2017-10-26  Jim Wilson  <wilson@tuliptree.org>
10747         * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10749 2017-10-26  Tom de Vries  <tom@codesourcery.com>
10751         PR tree-optimization/82707
10752         * gimple.c (gimple_copy): Fix unsharing of
10753         GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10755 2017-10-26  Olga Makhotina  <olga.makhotina@intel.com>
10757         * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10758         _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10759         _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10760         _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10761         _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10762         _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10763         _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10764         _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10765         _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10766         _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10767         _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10768         _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10769         _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10770         _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10771         _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10772         _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10773         _mm512_mask_cmpunord_ps_mask): New intrinsics.
10775 2017-10-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
10777         * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10778         default to IBM.
10779         * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10780         * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10781         warning to rs6000.c.  Remove the Undocumented flag, since it has
10782         been documented.
10783         (-mabi=ibmlongdouble): Likewise.
10784         * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10785         already set, set the default format for long double.
10786         (rs6000_debug_reg_global): Print whether long double is IBM or
10787         IEEE.
10788         (rs6000_option_override_internal): Rework setting long double
10789         format.  Only warn if the user is changing the long double default
10790         and they did not use -Wno-psabi.
10791         * doc/invoke.texi (PowerPC options): Update the documentation for
10792         -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10794 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10795             Alan Hayward  <alan.hayward@arm.com>
10796             David Sherwood  <david.sherwood@arm.com>
10798         * rtl.h (wider_subreg_mode): New function.
10799         * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10800         rather than an unsigned int *.
10801         * ira-color.c (regno_max_ref_width): Replace with...
10802         (regno_max_ref_mode): ...this new variable.
10803         (coalesced_pseudo_reg_slot_compare): Update accordingly.
10804         Use wider_subreg_mode.
10805         (ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
10806         rather than an unsigned int *.
10807         * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10808         (process_alt_operands): Likewise.
10809         (invariant_p): Likewise.
10810         * lra-spills.c (assign_mem_slot): Likewise.
10811         (add_pseudo_to_slot): Likewise.
10812         * lra.c (collect_non_operand_hard_regs): Likewise.
10813         (add_regs_to_insn_regno_info): Likewise.
10814         * reload1.c (regno_max_ref_width): Replace with...
10815         (regno_max_ref_mode): ...this new variable.
10816         (reload): Update accordingly.  Update call to
10817         ira_sort_regnos_for_alter_reg.
10818         (alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
10819         (init_eliminable_invariants): Update to use regno_max_ref_mode.
10820         (scan_paradoxical_subregs): Likewise.
10822 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10824         * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10825         (aarch64_frame): Add emit_frame_chain boolean.
10826         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10827         Move eh_return case to aarch64_layout_frame.
10828         (aarch64_layout_frame): Initialize emit_frame_chain.
10829         (aarch64_expand_prologue): Use emit_frame_chain.
10831 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10833         * config/aarch64/aarch64.c (aarch64_layout_frame):
10834         Ensure LR is always stored at the bottom of the callee-saves.
10835         Remove rarely used frame layout which saves callee-saves at top of
10836         frame, so the store of LR can be used as a valid probe in all cases.
10838 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10840         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10841         Improve unaligned TImode/TFmode base/offset split.
10843 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10844             Alan Hayward  <alan.hayward@arm.com>
10845             David Sherwood  <david.sherwood@arm.com>
10847         * caller-save.c (mark_referenced_regs):  Use read_modify_subreg_p.
10848         * combine.c (find_single_use_1): Likewise.
10849         (expand_field_assignment): Likewise.
10850         (move_deaths): Likewise.
10851         * lra-constraints.c (simplify_operand_subreg): Likewise.
10852         (curr_insn_transform): Likewise.
10853         * lra.c (collect_non_operand_hard_regs): Likewise.
10854         (add_regs_to_insn_regno_info): Likewise.
10855         * rtlanal.c (reg_referenced_p): Likewise.
10856         (covers_regno_no_parallel_p): Likewise.
10858 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10860         * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10861         Don't print any bits outside the precision of the value.
10862         * wide-int.cc (test_printing): Add some new tests.
10864 2017-10-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10866         * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10867         supports -xbrace_comment option.
10868         * configure: Regenerate.
10869         * config.in: Regenerate.
10870         * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10871         (ASM_CPU_SPEC): Use it.
10873 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10875         * target.def (static_rtx_alignment): New hook.
10876         * targhooks.h (default_static_rtx_alignment): Declare.
10877         * targhooks.c (default_static_rtx_alignment): New function.
10878         * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10879         * doc/tm.texi: Regenerate.
10880         * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10881         instead of targetm.constant_alignment.  Remove call to
10882         set_mem_attributes.
10883         * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10884         (cris_preferred_mininum_alignment): New function, split out from...
10885         (cris_constant_alignment): ...here.
10886         (cris_static_rtx_alignment): New function.
10887         * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10888         split out from...
10889         (ix86_constant_alignment): ...here.
10890         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10891         * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10892         (mmix_static_rtx_alignment): New function.
10893         * config/spu/spu.c (spu_static_rtx_alignment): New function.
10894         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10896 2017-10-26  Tamar Christina  <tamar.christina@arm.com>
10898         PR target/81800
10899         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10900         Add flag_trapping_math and flag_fp_int_builtin_inexact.
10902 2017-10-25  Palmer Dabbelt  <palmer@dabbelt.com>
10904         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10905         * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10906         mark as a sign-extending load.
10907         (local_pic_load_u): Define.
10909 2017-10-25  Eric Botcazou  <ebotcazou@adacore.com>
10911         PR middle-end/82062
10912         * fold-const.c (operand_equal_for_comparison_p): Also return true
10913         if ARG0 is a simple variant of ARG1 with narrower precision.
10914         (fold_ternary_loc): Always pass unstripped operands to the predicate.
10916 2017-10-25  Jan Hubicka  <hubicka@ucw.cz>
10918         * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
10919         cost correctly.
10920         * i386.h (processor_costs): Add gather_static, gather_per_elt,
10921         scatter_static, scatter_per_elt.
10922         * x86-tune-costs.h: Add new cost entries.
10924 2017-10-25  Richard Biener  <rguenther@suse.de>
10926         * tree-ssa-sccvn.h (vn_eliminate): Declare.
10927         * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
10928         class pass_fre): Move to ...
10929         * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
10930         class pass_fre): ... here and adjust for statistics.
10932 2017-10-25  Jakub Jelinek  <jakub@redhat.com>
10934         PR libstdc++/81706
10935         * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
10936         for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
10937         (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
10938         * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
10939         declarations.
10941 2017-10-25  Richard Biener  <rguenther@suse.de>
10943         * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
10944         el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
10945         eliminate_push_avail, eliminate_insert): Move inside...
10946         (class eliminate_dom_walker): ... this class in preparation
10947         of move.
10948         (fini_eliminate): Remove by merging with ...
10949         (eliminate): ... this function.  Adjust for class changes.
10950         (pass_pre::execute): Remove fini_eliminate call.
10951         (pass_fre::execute): Likewise.
10953 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10955         PR target/82460
10956         * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
10957         (VPERMI2, VPERMI2I): New mode iterators.
10958         (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
10959         (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
10960         patterns.
10961         (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
10962         mode iterator.  Remove 3 old define_insn patterns.
10963         (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
10964         (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
10965         VPERMI2 mode iterator, remove the other two expanders.
10966         (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
10967         to use VPERMI2 mode iterator, add another alternative for vpermi2*
10968         instructions, remove the other two patterns.
10969         (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
10970         mode iterator, remove the other two patterns.
10971         * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
10972         (ix86_expand_vec_perm_vpermt2): ... this.  Swap mask and op0
10973         arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
10974         and adjust argument order accordingly.
10975         (ix86_expand_vec_perm): Adjust caller.
10976         (expand_vec_perm_1): Likewise.
10977         (expand_vec_perm_vpermi2_vpshub2): Rename to ...
10978         (expand_vec_perm_vpermt2_vpshub2): ... this.
10979         (ix86_expand_vec_perm_const_1): Adjust caller.
10980         (ix86_vectorize_vec_perm_const_ok): Adjust comments.
10982         PR target/82370
10983         * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
10984         (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
10985         (vec_shl_<mode>): Remove unused expander.
10986         (avx512bw_<shift_insn><mode>3): New define_insn.
10987         (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
10988         (<sse2_avx2>_<shift_insn><mode>3): ... this.  New define_insn.
10990 2017-10-24  Paolo Carlini  <paolo.carlini@oracle.com>
10992         PR c++/82466
10993         * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
10994         description.
10996 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10998         PR rtl-optimization/82396
10999         * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
11000         (autopref_multipass_init): Simplify initialization.
11001         (autopref_rank_data): Simplify sort order.
11002         * gcc/sched-int.h (autopref_multipass_data_): Remove
11003         multi_mem_insn_p, min_offset and max_offset.
11005 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
11007         PR middle-end/60580
11008         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
11009         Check special value of flag_omit_frame_pointer.
11010         (aarch64_can_eliminate): Likewise.
11011         (aarch64_override_options_after_change_1): Simplify handling of
11012         -fomit-frame-pointer and -fomit-leaf-frame-pointer.
11014 2017-10-24  Richard Biener  <rguenther@suse.de>
11016         PR tree-optimization/82697
11017         * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
11018         zero for conditional load and unconditional store.
11020 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
11022         * doc/install.texi: Document bootstrap-cet.
11024 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
11026         PR target/82659
11027         * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
11028         ENDBR instruction at function entrance if function is only
11029         called directly.
11031 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
11033         PR target/82628
11034         * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
11035         patterns to better describe from which operation the CF is computed.
11036         (addcarry<mode>_0, subborrow<mode>_0): New patterns.
11037         * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
11038         one LTU with [DT]Imode and another one with [SD]Imode.  If arg0
11039         is 0, use _0 suffixed expanders instead of emitting a comparison
11040         before it.
11042 2017-10-06  Sergey Shalnov  <Sergey.Shalnov@intel.com>
11044         * config/i386/i386.md(*movsf_internal, *movdf_internal):
11045         Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
11047 2017-10-24  Eric Botcazou  <ebotcazou@adacore.com>
11049         PR middle-end/82569
11050         * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
11051         * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
11052         * loop-iv.c (iv_get_reaching_def): Likewise.
11053         * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
11054         variable is promoted and the partition contains undefined values.
11056 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11058         * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
11059         reflect reality.
11060         (nios2_address_cost): Define.
11061         (nios2_legitimize_address): Recognize (exp + constant) directly.
11062         (TARGET_ADDRESS_COST): Define.
11064 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11066         * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
11067         (nios2_symbolic_memory_operand_p): Declare.
11068         (nios2_split_large_constant): Declare.
11069         (nios2_split_symbolic_memory_operand): Declare.
11070         * config/nios2/nios2.c: Adjust includes.
11071         (nios2_symbolic_constant_allowed): New.
11072         (nios2_symbolic_constant_p): New.
11073         (nios2_plus_symbolic_constant_p): New.
11074         (nios2_valid_addr_expr_p): Recognize addresses involving
11075         symbolic constants.
11076         (nios2_legitimate_address_p): Likewise, also LO_SUM.
11077         (nios2_symbolic_memory_operand_p): New.
11078         (nios2_large_constant_p): New.
11079         (nios2_split_large_constant): New.
11080         (nios2_split_plus_large_constant): New.
11081         (nios2_split_symbolic_memory_operand): New.
11082         (nios2_legitimize_address): Code refactoring.  Handle addresses
11083         involving symbolic constants.
11084         (nios2_emit_move_sequence): Likewise.
11085         (nios2_print_operand): Improve error output.
11086         (nios2_print_operand_address): Handle LO_SUM.
11087         (nios2_cdx_narrow_form_p): Likewise.
11088         * config/nios2/nios2.md (movqi_internal): Add splitter for memory
11089         operands involving symbolic constants.
11090         (movhi_internal, movsi_internal): Likewise.
11091         (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
11092         (extendhisi2, extendqi<mode>2): Likewise.
11094 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11096         * tree-pass.h (PROP_rtl_split_insns): Define.
11097         * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
11099 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11101         * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
11103 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11105         PR debug/82630
11106         * target.def (const_not_ok_for_debug_p): Default to
11107         default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
11108         * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
11109         * targhooks.c (default_const_not_ok_for_debug_p): New function.
11110         * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
11111         which targetm.const_not_ok_for_debug_p returned true.
11112         * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
11113         for UNSPECs.
11114         * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
11115         Likewise.
11116         * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
11117         * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
11118         UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
11119         if !base_term_p.
11120         (ix86_const_not_ok_for_debug_p): New function.
11121         (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
11122         (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
11124 2017-10-23  David Malcolm  <dmalcolm@redhat.com>
11126         PR bootstrap/82610
11127         * system.h: Conditionally include "unique-ptr.h" if
11128         INCLUDE_UNIQUE_PTR is defined.
11129         * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
11130         of defining INCLUDE_UNIQUE_PTR before including "system.h".
11132 2017-10-23  Sebastian Perta  <sebastian.perta@renesas.com>
11134         * config/rl78/rl78.md: New define_expand "subdi3".
11136 2017-10-23  H.J. Lu  <hongjiu.lu@intel.com>
11138         PR target/82673
11139         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
11140         DF_REF_INSN if DF_REF_INSN_INFO is false.
11142 2017-10-23  Jan Hubicka  <hubicka@ucw.cz>
11144         * i386.c (dimode_scalar_chain::compute_convert_gain): Use
11145         xmm_move instead of sse_move.
11146         (sse_store_index): New function.
11147         (ix86_register_move_cost): Be more sensible about mismatch stall;
11148         model AVX moves correctly; make difference between sse->integer and
11149         integer->sse.
11150         (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
11151         moves; make difference between SSE and AVX.
11152         * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
11153         and zmm_move. Increase size of sse load and store tables;
11154         add unaligned load and store tables; add ssemmx_to_integer.
11155         * x86-tune-costs.h: Update all entries according to real
11156         move latencies from Agner Fog's manual and chip documentation.
11158 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11160         PR target/82628
11161         * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
11162         * config/i386/constraints.md (Wf): New constraint.
11163         * config/i386/i386.md (UNSPEC_SBB): New unspec.
11164         (cmp<dwi>_doubleword): Removed.
11165         (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
11166         (sub<mode>3_carry_ccgz): Use unspec instead of compare.
11167         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
11168         expand with cmp<dwi>_doubleword.  For LTU and GEU use
11169         sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
11171         * common.opt (gcolumn-info): Enable by default.
11172         * doc/invoke.texi (gcolumn-info): Document new default.
11174 2017-10-23  Richard Biener  <rguenther@suse.de>
11176         PR tree-optimization/82672
11177         * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
11178         Fold the stmt if we propagated into it.
11180 2017-10-23  Richard Biener  <rguenther@suse.de>
11182         * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
11183         (bitmap_remove_expr_from_set): ... this.  All callers call this
11184         for non-constant values.
11185         (bitmap_set_subtract): Rename to...
11186         (bitmap_set_subtract_expressions): ... this.  Adjust and
11187         optimize.
11188         (bitmap_set_contains_value): Remove superfluous check.
11189         (bitmap_set_replace_value): Inline into single caller ...
11190         (bitmap_value_replace_in_set): ... here and simplify.
11191         (dependent_clean): Merge into ...
11192         (clean): ... this using an overload.  Adjust.
11193         (prune_clobbered_mems): Adjust.
11194         (compute_antic_aux): Likewise.
11195         (compute_partial_antic_aux): Likewise.
11197 2017-10-23  Richard Biener  <rguenther@suse.de>
11199         PR tree-optimization/82129
11200         Revert
11201         2017-08-01  Richard Biener  <rguenther@suse.de>
11203         PR tree-optimization/81181
11204         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
11205         (compute_antic): ... end of iteration here.
11207 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11209         * target.def (starting_frame_offset): New hook.
11210         * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
11211         (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
11212         * doc/tm.texi.in: Regenerate.
11213         * hooks.h (hook_hwi_void_0): Declare.
11214         * hooks.c (hook_hwi_void_0): New function.
11215         * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
11216         STARTING_FRAME_OFFSET.
11217         * builtins.c (expand_builtin_setjmp_receiver): Likewise.
11218         * reload1.c (reload): Likewise.
11219         * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
11220         instead of STARTING_FRAME_OFFSET.
11221         * function.c (try_fit_stack_local): Likewise.
11222         (assign_stack_local_1): Likewise
11223         (instantiate_virtual_regs): Likewise.
11224         * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
11225         * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
11226         * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
11227         * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
11228         * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
11229         * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
11230         * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
11231         * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
11232         * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
11233         * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
11234         * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
11235         * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
11236         * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
11237         * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
11238         * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
11239         * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
11240         * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
11241         * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
11242         * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
11243         * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11244         * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11245         * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11246         * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11247         * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11248         * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11249         * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11250         * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11251         * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11252         * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11253         * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11254         * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11255         * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11256         * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11257         * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11258         * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11259         * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11260         * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11261         * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11262         * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11263         * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11264         * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11265         * config/avr/avr.c (avr_starting_frame_offset): Make static and
11266         return a HOST_WIDE_INT.
11267         (avr_builtin_setjmp_frame_value): Use it instead of
11268         STARTING_FRAME_OFFSET.
11269         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11270         * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11271         * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11272         New function.
11273         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11274         * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11275         * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11276         (TARGET_CONSTANT_ALIGNMENT): Redefine.
11277         * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11278         * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11279         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11280         * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11281         * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11282         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11283         * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11284         * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11285         New function.
11286         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11287         * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11288         * config/mips/mips.c (mips_compute_frame_info): Refer to
11289         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11290         (mips_starting_frame_offset): New function.
11291         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11292         * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11293         * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11294         * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11295         and return a HOST_WIDE_INT.
11296         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11297         (mmix_initial_elimination_offset): Refer to
11298         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11299         * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11300         * config/pa/pa.c (pa_starting_frame_offset): New function.
11301         (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11302         (pa_expand_prologue): Likewise.
11303         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11304         * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11305         !FRAME_GROWS_DOWNWARD handling to...
11306         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11307         * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11308         !FRAME_GROWS_DOWNWARD handling to...
11309         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11310         * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11311         !FRAME_GROWS_DOWNWARD handling to...
11312         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11313         * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11314         Redefine.
11315         (rs6000_starting_frame_offset): New function.
11316         * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11317         !FRAME_GROWS_DOWNWARD handling to...
11318         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11319         * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11320         !FRAME_GROWS_DOWNWARD handling to...
11321         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11322         * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11323         !FRAME_GROWS_DOWNWARD handling to...
11324         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11325         * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11326         (rs6000_starting_frame_offset): New function.
11327         * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11328         * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11329         * config/vax/vax.c (vax_starting_frame_offset): New function.
11330         (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11331         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11332         * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11333         * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11334         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11335         * system.h (STARTING_FRAME_OFFSET): Poison.
11337 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11339         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11340         SCALAR_TYPE_MODE instead of TYPE_MODE.
11342 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11343             Alan Hayward  <alan.hayward@arm.com>
11344             David Sherwood  <david.sherwood@arm.com>
11346         * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_TYPE_MODE
11348 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11349             Alan Hayward  <alan.hayward@arm.com>
11350             David Sherwood  <david.sherwood@arm.com>
11352         * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11354 2017-10-23  Richard Biener  <rguenther@suse.de>
11356         PR tree-optimization/82129
11357         * tree-ssa-pre.c (bitmap_set_and): Remove.
11358         (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11359         canonicalizing expressions in the set to those with lowest
11360         ID rather than taking that from the first edge.
11362 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11364         * combine.c (rtx_equal_for_field_assignment_p): Use
11365         byte_lowpart_offset.
11367 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11368             Alan Hayward  <alan.hayward@arm.com>
11369             David Sherwood  <david.sherwood@arm.com>
11371         * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11372         to a SUBREG_PROMOTED_VAR.
11374 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11375             Alan Hayward  <alan.hayward@arm.com>
11376             David Sherwood  <david.sherwood@arm.com>
11378         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11379         (expand_debug_source_expr): Likewise.
11380         * combine.c (combine_simplify_rtx): Likewise.
11381         * cse.c (fold_rtx): Likewise.
11382         * optabs.c (expand_float): Likewise.
11383         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11384         (simplify_binary_operation_1): Likewise.
11386 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11387             Alan Hayward  <alan.hayward@arm.com>
11388             David Sherwood  <david.sherwood@arm.com>
11390         * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11391         (record_promoted_value): Likewise.
11392         * expr.c (expand_expr_real_2): Likewise.
11393         * ree.c (update_reg_equal_equiv_notes): Likewise.
11394         (combine_set_extension): Likewise.
11395         * rtlanal.c (low_bitmask_len): Likewise.
11396         * simplify-rtx.c (neg_const_int): Likewise.
11397         (simplify_binary_operation_1): Likewise.
11399 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11400             Alan Hayward  <alan.hayward@arm.com>
11401             David Sherwood  <david.sherwood@arm.com>
11403         * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11404         * regcprop.c (maybe_mode_change): Likewise.
11405         * reload1.c (alter_reg): Likewise.
11407 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11409         * inchash.h (inchash::hash::add_wide_int): New function.
11410         * lto-streamer-out.c (hash_tree): Use it.
11412 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11414         * inchash.h (inchash::hash::add_wide_int): Rename to...
11415         (inchash::hash::add_hwi): ...this.
11416         * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11417         (polymorphic_call_target_hasher::hash): Likewise.
11418         * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11419         (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11420         (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11421         * lto-streamer-out.c (hash_tree): Likewise.
11422         * optc-save-gen.awk: Likewise.
11423         * tree.c (add_expr): Likewise.
11425 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11427         PR target/52451
11428         * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11429         for ordered inequality comparisons even with TARGET_IEEE_FP.
11431 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11433         PR target/82628
11434         * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11435         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11436         Expand with cmp<dwi>_doubleword.
11438 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11440         * extend.texi: Add x86 specific to 'nocf_check' attribute.
11441         List CET intrinsics.
11442         * invoke.texi: Add -mcet, -mibt, -mshstk options.  Add x86
11443         specific to -fcf-protection option.
11445 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11447         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11448         (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11449         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11450         (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11451         (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11452         * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11453         (extra_objs): Add cet.o for Linux/x86 targets.
11454         (tmake_file): Add i386/t-cet for Linux/x86 targets.
11455         * config/i386/cet.c: New file.
11456         * config/i386/cetintrin.h: Likewise.
11457         * config/i386/t-cet: Likewise.
11458         * config/i386/cpuid.h (bit_SHSTK): New.
11459         (bit_IBT): Likewise.
11460         * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11461         pass IBT and SHSTK bits.
11462         * config/i386/i386-builtin-types.def
11463         (VOID_FTYPE_UNSIGNED_PVOID): New.
11464         (VOID_FTYPE_UINT64_PVOID): Likewise.
11465         * config/i386/i386-builtin.def: Add CET intrinsics.
11466         * config/i386/i386-c.c (ix86_target_macros_internal): Add
11467         OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11468         * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11469         * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11470         prototype.
11471         * config/i386/i386.c (rest_of_insert_endbranch): New.
11472         (pass_data_insert_endbranch): Likewise.
11473         (pass_insert_endbranch): Likewise.
11474         (make_pass_insert_endbranch): Likewise.
11475         (ix86_notrack_prefixed_insn_p): Likewise.
11476         (ix86_target_string): Add -mibt, -mshstk flags.
11477         (ix86_option_override_internal): Add flag_cf_protection
11478         processing.
11479         (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11480         (ix86_print_operand): Add 'notrack' prefix output.
11481         (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11482         (ix86_expand_builtin): Expand CET intrinsics.
11483         (x86_output_mi_thunk): Add 'endbranch' instruction.
11484         * config/i386/i386.h (TARGET_IBT): New.
11485         (TARGET_IBT_P): Likewise.
11486         (TARGET_SHSTK): Likewise.
11487         (TARGET_SHSTK_P): Likewise.
11488         * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11489         UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11490         UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11491         (builtin_setjmp_setup): New pattern.
11492         (builtin_longjmp): Likewise.
11493         (rdssp<mode>): Likewise.
11494         (incssp<mode>): Likewise.
11495         (saveprevssp): Likewise.
11496         (rstorssp): Likewise.
11497         (wrss<mode>): Likewise.
11498         (wruss<mode>): Likewise.
11499         (setssbsy): Likewise.
11500         (clrssbsy): Likewise.
11501         (nop_endbr): Likewise.
11502         * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11503         options.
11504         * config/i386/immintrin.h: Include <cetintrin.h>.
11505         * config/i386/linux-common.h
11506         (file_end_indicate_exec_stack_and_cet): New prototype.
11507         (TARGET_ASM_FILE_END): New.
11509 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11511         * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11512         latencies instead of having separate table; make difference between
11513         integer and float costs.
11514         * i386.h (processor_costs): Remove scalar_stmt_cost,
11515         scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11516         scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11517         vec_store_cost.
11518         * x86-tune-costs.h: Remove entries which has been removed in
11519         procesor_costs from all tables; make cond_taken_branch_cost
11520         and cond_not_taken_branch_cost COST_N_INSNS based.
11522 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11524         * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11526 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11528         * config/i386/i386.md (isa): Remove fma_avx512f.
11529         * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11530         <avx512>_fmadd_<mode>_mask3<round_name>,
11531         <avx512>_fmsub_<mode>_mask<round_name>,
11532         <avx512>_fmsub_<mode>_mask3<round_name>,
11533         <avx512>_fnmadd_<mode>_mask<round_name>,
11534         <avx512>_fnmadd_<mode>_mask3<round_name>,
11535         <avx512>_fnmsub_<mode>_mask<round_name>,
11536         <avx512>_fnmsub_<mode>_mask3<round_name>,
11537         <avx512>_fmaddsub_<mode>_mask<round_name>,
11538         <avx512>_fmaddsub_<mode>_mask3<round_name>,
11539         <avx512>_fmsubadd_<mode>_mask<round_name>,
11540         <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11541         (*vec_widen_umult_even_v16si<mask_name>,
11542         *vec_widen_smult_even_v16si<mask_name>): Likewise.
11543         (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11545 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11547         * extend.texi: Add 'nocf_check' documentation.
11548         * gimple.texi: Add second parameter to
11549         gimple_build_call_from_tree.
11550         * invoke.texi: Add -fcf-protection documentation.
11551         * rtl.texi: Add REG_CALL_NOTRACK documenation.
11553 2017-10-20  Richard Biener  <rguenther@suse.de>
11555         PR tree-optimization/82473
11556         * tree-vect-loop.c (vectorizable_reduction): Properly get at
11557         the largest input type.
11559 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11561         * c-attribs.c (handle_nocf_check_attribute): New function.
11562         (c_common_attribute_table): Add 'nocf_check' handling.
11563         * gimple-parser.c: Add second argument NULL to
11564         gimple_build_call_from_tree.
11565         * attrib.c (comp_type_attributes): Check nocf_check attribute.
11566         * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11567         call insn.
11568         * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11569         * common.opt: Add fcf-protection flag.
11570         * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11571         * flag-types.h: Add enum cf_protection_level.
11572         * gimple.c (gimple_build_call_from_tree): Add second parameter.
11573         Add 'nocf_check' attribute propagation to gimple call.
11574         * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11575         (gimple_build_call_from_tree): Update prototype.
11576         (gimple_call_nocf_check_p): New function.
11577         (gimple_call_set_nocf_check): Likewise.
11578         * gimplify.c: Add second argument to gimple_build_call_from_tree.
11579         * ipa-icf.c: Add nocf_check attribute in statement hash.
11580         * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11581         * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11582         * toplev.c (process_options): Add flag_cf_protection handling.
11584 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11586         * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11588 2017-10-20  Richard Biener  <rguenther@suse.de>
11590         PR tree-optimization/82603
11591         * tree-if-conv.c (predicate_mem_writes): Make sure to only
11592         remove false predicated stores.
11594 2017-10-20  Richard Biener  <rguenther@suse.de>
11596         * graphite-isl-ast-to-gimple.c
11597         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11598         Remove return value and simplify, dump copied stmt after lhs
11599         adjustment.
11600         (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11601         Reduce dump verbosity.
11602         (gsi_insert_earliest): Likewise.
11603         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11604         * graphite.c (print_global_statistics): Adjust dumping.
11605         (print_graphite_scop_statistics): Likewise.
11606         (print_graphite_statistics): Do not dump loops here.
11607         (graphite_transform_loops): But here.
11609 2017-10-20  Nicolas Roche  <roche@adacore.com>
11611         * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11612         * configure: Regenerate.
11614 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11616         PR target/82158
11617         * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11618         functions when optimizing replace GIMPLE_RETURN stmts with
11619         calls to __builtin_unreachable ().
11621         PR sanitizer/82595
11622         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11623         for -fsanitize=thread link of executables.
11624         (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11625         link of executables.
11627         PR target/82370
11628         * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11629         New mode iterators.
11630         (<shift_insn><mode>3<mask_name>): Change the last of the 3
11631         define_insns for logical vector shifts to use VI248_AVX512BW
11632         iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11633         condition, useless isa and prefix attributes.  Change the first
11634         2 of these define_insns to ...
11635         (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11636         define_insn for avx512vl.
11637         (<shift_insn><mode>3): ... and this, new define_insn without
11638         masking for non-avx512vl.
11640         PR target/82370
11641         * config/i386/sse.md (*andnot<mode>3,
11642         <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11643         (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11644         and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11645         not applied use empty suffix even for TARGET_AVX512VL.
11646         * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11647         is applied, supply evex,evex or evex,evex,evex instead of just
11648         evex.
11650 2017-10-20  Julia Koval  <julia.koval@intel.com>
11652         * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11653         (OPTION_MASK_ISA_GFNI_UNSET): New.
11654         (ix86_handle_option): Handle OPT_mgfni.
11655         * config/i386/cpuid.h (bit_GFNI): New.
11656         * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11657         * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11658         * config/i386/i386.c (ix86_target_string): Add -mgfni.
11659         (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11660         * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11661         * config/i386/i386.opt: Add mgfni.
11663 2017-10-20  Orlando Arias  <oarias@knights.ucf.edu>
11665         * config/msp430/msp430.c (msp430_option_override): Disable
11666         -fdelete-null-pointer-checks.
11667         * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11669 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11671         * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11672         of x87 and SSE instructions.
11674 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11676         * asan.c (create_cond_insert_point): Do not update edge count.
11677         * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11678         (afdo_propagate_circuit): Likewise.
11679         (afdo_calculate_branch_prob): Likewise.
11680         (afdo_annotate_cfg): Likewise.
11681         * basic-block.h (struct edge_def): Remove count.
11682         (edge_def::count): New accessor.
11683         * bb-reorder.c (rotate_loop): Update.
11684         (find_traces_1_round): Update.
11685         (connect_traces): Update.
11686         (sanitize_hot_paths): Update.
11687         * cfg.c (unchecked_make_edge): Update.
11688         (make_single_succ_edge): Update.
11689         (check_bb_profile): Update.
11690         (dump_edge_info): Update.
11691         (update_bb_profile_for_threading): Update.
11692         (scale_bbs_frequencies_int): Update.
11693         (scale_bbs_frequencies_gcov_type): Update.
11694         (scale_bbs_frequencies_profile_count): Update.
11695         (scale_bbs_frequencies): Update.
11696         * cfganal.c (connect_infinite_loops_to_exit): Update.
11697         * cfgbuild.c (compute_outgoing_frequencies): Update.
11698         (find_many_sub_basic_blocks): Update.
11699         * cfgcleanup.c (try_forward_edges): Update.
11700         (try_crossjump_to_edge): Update
11701         * cfgexpand.c (expand_gimple_cond): Update
11702         (expand_gimple_tailcall): Update
11703         (construct_exit_block): Update
11704         * cfghooks.c (verify_flow_info): Update
11705         (redirect_edge_succ_nodup): Update
11706         (split_edge): Update
11707         (make_forwarder_block): Update
11708         (duplicate_block): Update
11709         (account_profile_record): Update
11710         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11711         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11712         * cfgloopmanip.c (scale_loop_profile): Update.
11713         (loopify): Update.
11714         (lv_adjust_loop_entry_edge): Update.
11715         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11716         (force_nonfallthru_and_redirect): Update.
11717         (purge_dead_edges): Update.
11718         (rtl_flow_call_edges_add): Update.
11719         * cgraphunit.c (init_lowered_empty_function): Update.
11720         (cgraph_node::expand_thunk): Update.
11721         * gimple-pretty-print.c (dump_probability): Update.
11722         (dump_edge_probability): Update.
11723         * gimple-ssa-isolate-paths.c (isolate_path): Update.
11724         * haifa-sched.c (sched_create_recovery_edges): Update.
11725         * hsa-gen.c (convert_switch_statements): Update.
11726         * ifcvt.c (dead_or_predicable): Update.
11727         * ipa-inline-transform.c (inline_transform): Update.
11728         * ipa-split.c (split_function): Update.
11729         * ipa-utils.c (ipa_merge_profiles): Update.
11730         * loop-doloop.c (add_test): Update.
11731         * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11732         * lto-streamer-in.c (input_cfg): Update.
11733         (input_function): Update.
11734         * lto-streamer-out.c (output_cfg): Update.
11735         * modulo-sched.c (sms_schedule): Update.
11736         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11737         * predict.c (maybe_hot_edge_p): Update.
11738         (unlikely_executed_edge_p): Update.
11739         (probably_never_executed_edge_p): Update.
11740         (dump_prediction): Update.
11741         (drop_profile): Update.
11742         (propagate_unlikely_bbs_forward): Update.
11743         (determine_unlikely_bbs): Update.
11744         (force_edge_cold): Update.
11745         * profile.c (compute_branch_probabilities): Update.
11746         * reg-stack.c (better_edge): Update.
11747         * shrink-wrap.c (handle_simple_exit): Update.
11748         * tracer.c (better_p): Update.
11749         * trans-mem.c (expand_transaction): Update.
11750         (split_bb_make_tm_edge): Update.
11751         * tree-call-cdce.c: Update.
11752         * tree-cfg.c (gimple_find_sub_bbs): Update.
11753         (gimple_split_edge): Update.
11754         (gimple_duplicate_sese_region): Update.
11755         (gimple_duplicate_sese_tail): Update.
11756         (gimple_flow_call_edges_add): Update.
11757         (insert_cond_bb): Update.
11758         (execute_fixup_cfg): Update.
11759         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11760         * tree-complex.c (expand_complex_div_wide): Update.
11761         * tree-eh.c (lower_resx): Update.
11762         (unsplit_eh): Update.
11763         (cleanup_empty_eh_move_lp): Update.
11764         * tree-inline.c (copy_edges_for_bb): Update.
11765         (freqs_to_counts): Update.
11766         (copy_cfg_body): Update.
11767         * tree-ssa-dce.c (remove_dead_stmt): Update.
11768         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11769         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11770         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11771         (unloop_loops): Update.
11772         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11773         * tree-ssa-loop-split.c (connect_loops): Update.
11774         (split_loop): Update.
11775         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11776         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11777         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11778         * tree-ssa-reassoc.c (branch_fixup): Update.
11779         * tree-ssa-tail-merge.c (replace_block_by): Update.
11780         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11781         (compute_path_counts): Update.
11782         (update_profile): Update.
11783         (recompute_probabilities): Update.
11784         (update_joiner_offpath_counts): Update.
11785         (estimated_freqs_path): Update.
11786         (freqs_to_counts_path): Update.
11787         (clear_counts_path): Update.
11788         (ssa_fix_duplicate_block_edges): Update.
11789         (duplicate_thread_path): Update.
11790         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11791         (case_bit_test_cmp): Update.
11792         (collect_switch_conv_info): Update.
11793         (gen_inbound_check): Update.
11794         (do_jump_if_equal): Update.
11795         (emit_cmp_and_jump_insns): Update.
11796         * tree-tailcall.c (decrease_profile): Update.
11797         (eliminate_tail_call): Update.
11798         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11799         (vect_do_peeling): Update.
11800         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11801         * ubsan.c (ubsan_expand_null_ifn): Update.
11802         (ubsan_expand_ptr_ifn): Update.
11803         * value-prof.c (gimple_divmod_fixed_value): Update.
11804         (gimple_mod_pow2): Update.
11805         (gimple_mod_subtract): Update.
11806         (gimple_ic): Update.
11807         (gimple_stringop_fixed_value): Update.
11809 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11811         PR target/82618
11812         * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11814 2017-10-19  Alexander Monakov  <amonakov@ispras.ru>
11816         PR rtl-optimization/82395
11817         * ira-color.c (allocno_priority_compare_func): Fix comparison step
11818         based on non_spilled_static_chain_regno_p.
11820 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11822         * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11823         (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11824         (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11826 2017-10-19  Martin Sebor  <msebor@redhat.com>
11828         PR tree-optimization/82596
11829         * tree.c (array_at_struct_end_p): Handle STRING_CST.
11831 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11833         * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11834         (get_mem_refs_of_builtin_call): Likewise.
11835         * builtins.c (expand_builtin_apply): Adjust call to
11836         allocate_dynamic_stack_space.
11837         (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11838         the third argument to allocate_dynamic_stack_space, otherwise -1.
11839         (expand_builtin): Deal with all alloca variants.
11840         (is_inexpensive_builtin): Likewise.
11841         * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11842         * calls.c (special_function_p): Deal with all alloca variants.
11843         (initialize_argument_information): Adjust call to
11844         allocate_dynamic_stack_space.
11845         (expand_call): Likewise.
11846         * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11847         * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11848         * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11849         use it for the stack usage computation.
11850         * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11851         * function.c (gimplify_parameters): Call build_alloca_call_expr.
11852         * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11853         Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11854         (in_loop_p): Remove first argument and useless check.
11855         (pass_walloca::execute): Remove useless test and adjust call to above.
11856         * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11857         * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11858         (gimplify_call_expr): Deal with all alloca variants.
11859         * hsa-gen.c (gen_hsa_alloca): Likewise.
11860         (gen_hsa_insns_for_call): Likewise.
11861         * ipa-pure-const.c (special_builtin_state): Likewise.
11862         * tree-chkp.c (chkp_build_returned_bound): Likewise.
11863         * tree-object-size.c (alloc_object_size): Likewise.
11864         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11865         (call_may_clobber_ref_p_1): Likewise.
11866         * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11867         (ccp_fold_stmt): Likewise.
11868         (optimize_stack_restore): Likewise.
11869         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11870         (mark_all_reaching_defs_necessary_1): Likewise.
11871         (propagate_necessity): Likewise.
11872         (eliminate_unnecessary_stmts): Likewise.
11873         * tree.c (build_common_builtin_nodes): Build
11874         BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11875         (build_alloca_call_expr): New function.
11876         * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11877         (CASE_BUILT_IN_ALLOCA): Likewise.
11878         (build_alloca_call_expr): Declare.
11879         * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11881 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11883         PR debug/82509
11884         * dwarf2out.c (new_die_raw): New static inline function.
11885         (new_die): Use it to create the DIE.
11886         (add_AT_external_die_ref): Likewise.
11887         (clone_die): Likewise.
11888         (clone_as_declaration): Likewise.
11889         (dwarf2out_vms_debug_main_pointer): Likewise.
11890         (base_type_die): Likewise.  Remove early return for corner cases.
11891         Do not call add_pubtype on the DIE here.
11892         (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11893         (modified_type_die): Adjust the lookup for reverse order DIEs.  Skip
11894         typedefs for base types with DW_AT_endianity.  Make sure a DIE with
11895         native order exists for base types, attach the DIE manually and call
11896         add_pubtype on it.  Do not equate a reverse order DIE to the type.
11898 2017-10-19  Richard Earnshaw  <rearnsha@arm.com>
11900         * config/arm/arm.c (align_ok_ldrd_strd): New function.
11901         (mem_ok_for_ldrd_strd): New parameter align.  Extract the alignment of
11902         the mem into it.
11903         (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11905 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11907         * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN.  Or
11908         SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11909         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11910         BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11911         * opts.c (sanitizer_opts): Add builtin.
11912         * ubsan.c (instrument_builtin): New function.
11913         (pass_ubsan::execute): Call it.
11914         (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11915         * doc/invoke.texi: Document -fsanitize=builtin.
11917         * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
11918         builtins, store max (log2 (align), 0) into uchar field instead of
11919         align into uptr field.
11920         (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
11921         store uchar 0 field instead of uptr 0 field.
11922         (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
11923         instead of passing one address of struct with 2 locations pass
11924         two addresses of structs with 1 location each.
11925         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
11926         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
11927         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
11928         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
11929         (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
11930         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
11931         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
11932         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
11934 2017-10-19  Martin Liska  <mliska@suse.cz>
11936         PR driver/81829
11937         * file-find.c (remove_prefix): Remove.
11938         * file-find.h (remove_prefix): Likewise.
11939         * gcc-ar.c: Remove smartness of lookup.
11941 2017-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
11943         * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
11944         *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
11945         *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
11947 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11949         PR target/82580
11950         * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
11951         (setcc + and to xor + setcc): New peephole2.
11953 2017-10-19  Tom de Vries  <tom@codesourcery.com>
11955         * doc/sourcebuild.texi (Test Directives, Variants of
11956         dg-require-support): Add dg-require-stack-size.
11958 2017-10-19  Martin Liska  <mliska@suse.cz>
11960         PR sanitizer/82517
11961         * gimplify.c (gimplify_decl_expr): Do not instrument variables
11962         that have a large alignment.
11963         (gimplify_target_expr): Likewise.
11965 2017-10-18  Segher Boessenkool  <segher@kernel.crashing.org>
11967         PR rtl-optimization/82602
11968         * ira.c (rtx_moveable_p): Return false for volatile asm.
11970 2017-10-18  Uros Bizjak  <ubizjak@gmail.com>
11972         PR target/82580
11973         * config/i386/i386-modes.def (CCGZ): New CC mode.
11974         * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
11975         * config/i386/predicates.md (ix86_comparison_operator):
11976         Handle CCGZmode.
11977         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11978         Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
11979         with double-word subtraction.
11980         (put_condition_code): Handle CCGZmode.
11982 2017-10-18  Aldy Hernandez  <aldyh@redhat.com>
11984         * wide-int.cc (debug (const wide_int &)): New.
11985         (debug (const wide_int *)): New.
11986         (debug (const widest_int &)): New.
11987         (debug (const widest_int *)): New.
11989 2017-10-18  Vladimir Makarov  <vmakarov@redhat.com>
11991         PR middle-end/82556
11992         * lra-constraints.c (curr_insn_transform): Use non-input operand
11993         instead of output one for matched reload.
11995 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11997         * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
11998         (tree-ssa-loop-ivopts.h): New header file.
11999         (struct builtin_info): New fields.
12000         (classify_builtin_1): Compute and record base and offset parts for
12001         memset builtin partition by calling strip_offset.
12002         (offset_cmp, fuse_memset_builtins): New functions.
12003         (finalize_partitions): Fuse adjacent memset partitions by calling
12004         above function.
12005         * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
12006         Expose the interface.
12007         * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
12009 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
12011         PR tree-optimization/82574
12012         * tree-loop-distribution.c (find_single_drs): New parameter.  Check
12013         that data reference must be executed exactly once per iteration
12014         against the outermost loop in nest.
12015         (classify_partition): Update call to above function.
12017 2017-10-18  Richard Biener  <rguenther@suse.de>
12019         PR tree-optimization/82591
12020         * graphite.c (graphite_transform_loops): Move code gen message
12021         printing ...
12022         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12023         Here.  Handle scop_to_isl_ast failing.
12024         (scop_to_isl_ast): Limit the number of ISL operations.
12026 2017-10-18  Richard Biener  <rguenther@suse.de>
12028         * graphite-isl-ast-to-gimple.c
12029         (translate_isl_ast_to_gimple::set_rename): Simplify.
12030         (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
12031         (graphite_copy_stmts_from_block): ... here.
12032         (copy_bb_and_scalar_dependences): Simplify.
12033         (add_parameters_to_ivs_params): Canonicalize.
12034         (generate_entry_out_of_ssa_copies): Simplify.
12035         * graphite-sese-to-poly.c (extract_affine_name): Simplify
12036         by passing in ISL dimension.
12037         (parameter_index_in_region_1): Rename to ...
12038         (parameter_index_in_region): ... this.
12039         (extract_affine): Adjust assert, pass down parameter index.
12040         (add_param_constraints): Use range-info when available.
12041         (build_scop_context): Adjust.
12042         * sese.c (new_sese_info): Adjust.
12043         (free_sese_info): Likewise.
12044         * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
12045         Remove unused typedefs.
12046         (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
12048 2017-10-18  Martin Liska  <mliska@suse.cz>
12050         * combine.c (simplify_compare_const): Add gcc_fallthrough.
12052 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12054         * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
12055         (s390_sched_init): Do not reset s390_sched_state if we entered the
12056         current basic block via a fallthru edge and all others are unlikely.
12058 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12060         * config/s390/s390.c (NUM_SIDES): New variable.
12061         (LONGRUNNING_THRESHOLD): New variable.
12062         (LATENCY_FACTOR): New variable.
12063         (s390_sched_score): Decrease score for long-running instructions on
12064         wrong side.
12065         (s390_sched_variable_issue): Perform bookkeeping for long-running
12066         instructions.
12068 2017-10-18  Richard Biener  <rguenther@suse.de>
12070         * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
12071         Simplify with removal of the parameter rename map.
12072         (set_rename): Likewise.
12073         (should_copy_to_new_region): Likewise.
12074         (graphite_copy_stmts_from_block): Likewise.
12075         (copy_bb_and_scalar_dependences): Remove initialization of
12076         unused copied_bb_map.
12077         (copy_def): Remove.
12078         (copy_internal_parameters): Likewise.
12079         (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
12080         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12081         Use INTEGRAL_TYPE_P.
12082         (parameter_index_in_region_1): Rename to ...
12083         (assign_parameter_index_in_region): ... this.  Assert we have
12084         a parameter we handle.
12085         (scan_tree_for_params): Adjust.
12086         * sese.h (parameter_rename_map_t): Remove.
12087         (struct sese_info_t): Remove unused parameter_rename_map and
12088         copied_bb_map members.
12089         * sese.c (new_sese_info): Adjust.
12090         (free_sese_info): Likewise.
12092 2017-10-18  Martin Liska  <mliska@suse.cz>
12094         PR sanitizer/82545
12095         * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
12096         on an abnormal edge.
12098 2017-10-18  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12100         * doc/invoke.texi (ffunction-sections and fdata-sections):
12101         Update.
12103 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12105         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
12106         the use statement can throw internally.
12108 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12110         * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
12111         any RTX present on the RHS of a SET.
12112         * compare-elim.c (try_eliminate_compare): Restore comment.
12114 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12116         * langhooks.h (struct lang_hooks): Document that tree_size langhook
12117         may be also called on tcc_type nodes.
12118         * langhooks.c (lhd_tree_size): Likewise.
12120 2017-10-17  David Malcolm  <dmalcolm@redhat.com>
12122         * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
12123         format_warning_at_substring.
12124         (maybe_warn): Convert source_range * param to a location_t.  Pass
12125         UNKNOWN_LOCATION rather than NULL to fmtwarn.
12126         (format_directive): Remove code to extract source_ranges and
12127         source_range * in favor of just a location_t.
12128         (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
12129         fmtwarn.
12130         * substring-locations.c (format_warning_va): Convert
12131         source_range * param to a location_t.
12132         (format_warning_at_substring): Likewise.
12133         * substring-locations.h (format_warning_va): Likewise.
12134         (format_warning_at_substring): Likewise.
12136 2017-10-17  Jan Hubicka  <hubicka@ucw.cz>
12138         * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
12139         vec_scatter_store
12140         * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
12141         and scatter/gather ops.
12143         * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
12144         vec_gather_load and vec_scatter_store.
12145         * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
12146         * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
12147         * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
12148         * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
12149         * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
12150         * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
12152 2017-10-17  Uros Bizjak  <ubizjak@gmail.com>
12154         * reg-stack.c (compare_for_stack_reg): Add bool argument.
12155         Detect FTST instruction and handle its register pops.  Only pop
12156         second operand if can_pop_second_op is true.
12157         (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
12158         set can_pop_second_op to false in the compare_for_stack_reg call.
12160         * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
12161         output_fp_compare for stack register operands.
12162         * config/i386/i386.c (output_fp_compare): Do not output SSE compare
12163         instructions here.  Do not emit stack register pops here.  Assert
12164         that FCOMPP pops next to top stack register.  Rewrite function.
12166 2017-10-17  Nathan Sidwell  <nathan@acm.org>
12168         PR middle-end/82577
12169         * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
12170         use DECL_ASSEMBLER_NAME_RAW.
12172         PR middle-end/82546
12173         * tree.c (tree_code_size): Reformat.  Punt to lang hook for unknown
12174         TYPE nodes.
12176 2017-10-17  Qing Zhao <qing.zhao@oracle.com>
12177             Wilco Dijkstra <wilco.dijkstra@arm.com>
12179         * builtins.c (expand_builtin_update_setjmp_buf): Add a
12180         converstion to Pmode from the buf_addr.
12182 2017-10-17  Richard Biener  <rguenther@suse.de>
12184         * graphite-dependences.c (scop_get_reads_and_writes): Change
12185         output parameters to references.
12187 2017-10-17  Jackson Woodruff  <jackson.woodruff@arm.com>
12189         PR 71026/tree-optimization
12190         * fold-const.c (distribute_real_division): Removed.
12191         (fold_binary_loc): Remove calls to distribute_real_divison.
12193 2017-10-17  Richard Biener  <rguenther@suse.de>
12195         * graphite-scop-detection.c
12196         (scop_detection::stmt_has_simple_data_refs_p): Always use
12197         the full nest as region.
12198         (try_generate_gimple_bb): Likewise.
12199         * sese.c (scalar_evolution_in_region): Simplify now that
12200         SCEV can handle instantiation in regions.
12201         * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
12202         in the non-loop part of a function if requested.
12204 2017-10-17  Richard Biener  <rguenther@suse.de>
12206         PR tree-optimization/82563
12207         * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
12208         New function.
12209         (graphite_regenerate_ast_isl): Call it.
12210         * graphite-scop-detection.c (build_scops): Remove entry edge split.
12212 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12214         PR tree-optimization/82549
12215         * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
12216         Formatting fixes.  Instead of calling make_bit_field_ref with negative
12217         bitpos return 0.
12219 2017-10-17  Olga Makhotina  <olga.makhotina@intel.com>
12221         * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
12222         _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
12223         _mm_maskz_reduce_ss): New.
12224         * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
12225         __builtin_ia32_reducess_mask): Ditto..
12226         (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
12227         * config/i386/sse.md (reduces<mode>): Renamed to ...
12228         (reduces<mode><mask_scalar_name>): ... this.
12229         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
12230         Changed to ...
12231         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
12232         %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
12234 2017-10-16  David Malcolm  <dmalcolm@redhat.com>
12236         * Makefile.in (OBJS): Add unique-ptr-tests.o.
12237         * selftest-run-tests.c (selftest::run_tests): Call
12238         selftest::unique_ptr_tests_cc_tests.
12239         * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
12240         * unique-ptr-tests.cc: New file.
12242 2017-10-16  Vladimir Makarov  <vmakarov@redhat.com>
12244         PR sanitizer/82353
12245         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12246         locations.
12247         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12248         (make_hard_regno_born, make_hard_regno_dead): Update
12249         bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12251 2017-10-16  Jeff Law  <law@redhat.com>
12253         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12255 2017-10-16  Jan Hubicka  <hubicka@ucw.cz>
12257         * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12259 2017-10-16  Olivier Hainque  <hainque@adacore.com>
12261         * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12262         with_cpu if we were configured for an e500v2 target cpu name.
12264 2017-10-16  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12266         * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12267         * doc/invoke.texi: Document +nodsp as a valid extension for
12268         -mcpu=cortex-m33.
12270 2017-10-16  Martin Liska  <mliska@suse.cz>
12272         * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12273         (test_set_range): Likewise.
12274         (test_range_functions): Rename to ...
12275         (test_bit_in_range): ... this.
12276         (sbitmap_c_tests): Add new test.
12278 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12280         * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12281         New.
12282         (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12283         (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12285 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12287         * config/aarch64/aarch64-builtins.c
12288         (aarch64_types_quadopu_lane_qualifiers): New.
12289         (TYPES_QUADOPU_LANE): New.
12290         * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12291         (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12292         (aarch64_<sur>dot_laneq<vsi2qi>): New.
12293         * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12294         (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12295         * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12296         (Vdottype, DOTPROD): New.
12297         (sur): Add SDOT and UDOT.
12299 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12301         * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12302         (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12303         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12304         Add TARGET_DOTPROD.
12305         * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12306         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12307         Enable TARGET_DOTPROD.
12308         (cortex-a75.cortex-a55): Likewise.
12309         * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12311 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12313         * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12314         (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12315         New.
12316         * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12317         New.
12318         * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12319         (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12320         * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12321         (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12322         * config/arm/types.md (neon_dot, neon_dot_q): New.
12323         * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12325 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12327         * config/arm/arm.h (TARGET_DOTPROD): New.
12328         * config/arm/arm.c (arm_arch_dotprod): New.
12329         (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12330         * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12331         * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12332         (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12333         (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12334         * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12335         * doc/invoke.texi (armv8.2-a): Document dotprod
12337 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12339         * i386.c (ix86_vec_cost): New function.
12340         (ix86_rtx_costs): Handle vector operations better.
12341         * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12342         * x86-tune-costs.h: Add new costs to all tables.
12344 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12346         * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12347         operations.
12348         * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12349         divsd, sqrtss and sqrtsd
12350         * x86-tune-costs.h: Add new entries to all costs.
12351         (znver1_cost): Fix to match real instruction latencies.
12353 2017-10-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12354             Michael Collison <michael.collison@arm.com>
12356         * compare-elim.c: Include emit-rtl.h.
12357         (can_merge_compare_into_arith): New function.
12358         (try_validate_parallel): Likewise.
12359         (try_merge_compare): Likewise.
12360         (try_eliminate_compare): Call the above when no previous clobber
12361         is available.
12362         (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12363         dataflow problems.
12365 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12367         PR middle-end/62263
12368         PR middle-end/82498
12369         * tree-ssa-phiopt.c (value_replacement): Comment fix.  Handle
12370         up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12372         PR middle-end/62263
12373         PR middle-end/82498
12374         * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12375         to be any operand_equal_p operands.  For & (B - 1) require
12376         B to be power of 2.  Recognize
12377         (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12379 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12381         PR bootstrap/82553
12382         * optabs.c (expand_memory_blockage): Fix call of
12383         targetm.have_memory_blockage.
12385 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12387         PR bootstrap/82548
12388         * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12389         x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12390         objects to extra_objs instead of overwriting it.
12392 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12394         * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12395         Use any_fp_register_operand as operand[3] predicate.  Simplify
12396         equality test for operands[2] and operands[4] memory location.
12397         (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12398         (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12399         (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12400         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12401         any_fp_register_operand as operand[1] predicate.  Simplify
12402         equality test for operands[0] and operands[3] memory location.
12403         (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12404         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12405         (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12407 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12409         * target-insns.def: Add memory_blockage.
12410         * optabs.c (expand_memory_blockage): New function.
12411         (expand_asm_memory_barrier): Rename ...
12412         (expand_asm_memory_blockage): ... to this.
12413         (expand_mem_thread_fence): Call expand_memory_blockage
12414         instead of expand_asm_memory_barrier.
12415         (expand_mem_singnal_fence): Ditto.
12416         (expand_atomic_load): Ditto.
12417         (expand_atomic_store): Ditto.
12418         * doc/md.texi (Standard Pattern Names For Generation):
12419         Document memory_blockage instruction pattern.
12421 2017-10-13  Sebastian Perta  <sebastian.perta@renesas.com>
12423         * config/rl78/rl78.c (rl78_emit_libcall): New function.
12424         * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12425         * config/rl78/rl78.md: New define_expand "adddi3".
12427 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12429         * cfghooks.c (verify_flow_info): Disable check that all probabilities
12430         are set correctly.
12432 2017-10-13  Jeff Law  <law@redhat.com>
12434         * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12436 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12438         PR target/82274
12439         * internal-fn.c (expand_mul_overflow): If both operands have
12440         the same highpart of -1 or 0 and the topmost bit of lowpart
12441         is different, overflow is if res <= 0 rather than res < 0.
12443 2017-10-13  Pat Haugen  <pthaugen@us.ibm.com>
12445         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12446         TARGET_P9_VECTOR code for unaligned_load case.
12448 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12450         * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12452 2017-10-13  Nathan Sidwell  <nathan@acm.org>
12454         * tree-core.h (tree_contains_struct): Make bool.
12455         * tree.c (tree_contains_struct): Likewise.
12456         * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12457         (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12458         MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12459         MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12460         MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12462 2017-10-13  Richard Biener  <rguenther@suse.de>
12464         * graphite-isl-ast-to-gimple.c
12465         (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12466         parameters and dominance check.
12467         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12468         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12469         (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12470         Do not update SSA form here or do intermediate IL verification.
12471         * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12472         (graphite_initialize): Remove check on the number of loops in
12473         the function and inline into graphite_transform_loops.
12474         (graphite_finalize): Inline into graphite_transform_loops.
12475         (graphite_transform_loops): Perform SSA update and IL verification
12476         here.
12477         * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12479 2017-10-13  Richard Biener  <rguenther@suse.de>
12481         * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12482         graphite_expression_type_precision): Avoid global constructor
12483         by moving ...
12484         (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12485         (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12486         (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12487         (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12488         (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12489         * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12491 2017-10-13  H.J. Lu  <hongjiu.lu@intel.com>
12493         PR target/82499
12494         * config/i386/i386.h (ix86_red_zone_size): New.
12495         * config/i386/i386.md (push peephole2s): Replace
12496         "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12498 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12499             Alan Hayward  <alan.hayward@arm.com>
12500             David Sherwood  <david.sherwood@arm.com>
12502         * combine.c (can_change_dest_mode): Reject changes in
12503         REGMODE_NATURAL_SIZE.
12505 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12506             Alan Hayward  <alan.hayward@arm.com>
12507             David Sherwood  <david.sherwood@arm.com>
12509         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12510         (expand_debug_source_expr): Likewise.
12511         * combine.c (combine_simplify_rtx): Likewise.
12512         * cse.c (fold_rtx): Likewise.
12513         * fwprop.c (canonicalize_address): Likewise.
12514         * targhooks.c (default_shift_truncation_mask): Likewise.
12516 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12517             Alan Hayward  <alan.hayward@arm.com>
12518             David Sherwood  <david.sherwood@arm.com>
12520         * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12521         (widened_mode): Likewise.
12522         (expand_unop): Likewise.
12523         * ree.c (transform_ifelse): Likewise.
12524         (merge_def_and_ext): Likewise.
12525         (combine_reaching_defs): Likewise.
12526         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12528 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12529             Alan Hayward  <alan.hayward@arm.com>
12530             David Sherwood  <david.sherwood@arm.com>
12532         * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12533         * combine.c (gen_lowpart_for_combine): Likewise.
12534         * dwarf2out.c (rtl_for_decl_location): Likewise.
12535         * final.c (alter_subreg): Likewise.
12536         * rtlhooks.c (gen_lowpart_general): Likewise.
12537         (gen_lowpart_if_possible): Likewise.
12539 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12540             Alan Hayward  <alan.hayward@arm.com>
12541             David Sherwood  <david.sherwood@arm.com>
12543         * calls.c (expand_call): Use subreg_lowpart_offset.
12544         * cse.c (cse_insn): Likewise.
12545         * regcprop.c (copy_value): Likewise.
12546         (copyprop_hardreg_forward_1): Likewise.
12548 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12550         PR target/82524
12551         * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12552         *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12553         =Q constraints to +Q and into insn condition add check
12554         that operands[0] and operands[1] are equal.
12555         (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12556         =Q constraints to +Q and into insn condition add check
12557         that operands[0] is equal to either operands[1] or operands[2].
12559         PR target/82498
12560         * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12561         instead of handling MINUS_EXPR twice (once for each argument),
12562         canonicalize operand order and handle just once, use rtype where
12563         possible.  Handle (A << B) | (A >> (-B & (Z - 1))).
12565         PR target/82498
12566         * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12567         any values of __C while still being pattern recognizable as a simple
12568         rotate instruction.
12570 2017-10-13  Richard Biener  <rguenther@suse.de>
12572         PR tree-optimization/82451
12573         Revert
12574         2017-10-02  Richard Biener  <rguenther@suse.de>
12576         PR tree-optimization/82355
12577         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12578         a mapping for the enclosing loop but avoid generating one for
12579         the loop tree root.
12580         (copy_bb_and_scalar_dependences): Remove premature codegen
12581         error on PHIs in blocks duplicated into multiple places.
12582         * graphite-scop-detection.c
12583         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12584         in the region use it as loop and nest to analyze the DR in.
12585         (try_generate_gimple_bb): Likewise.
12586         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12587         (add_loop_constraints): For blocks in a loop not in the region
12588         create a dimension with a single iteration.
12589         * sese.h (gbb_loop_at_index): Remove assert.
12591         * cfgloop.c (loop_preheader_edge): For the loop tree root
12592         return the single successor of the entry block.
12593         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12594         Reset the SCEV hashtable and niters.
12595         * graphite-scop-detection.c
12596         (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12597         assert that we only have POLYNOMIAL_CHREC that vary in loops
12598         contained in the region.
12599         (scop_detection::graphite_can_represent_expr): Adjust.
12600         (scop_detection::stmt_has_simple_data_refs_p): For loops
12601         not in the region set loop to NULL.  The nest is now the
12602         entry edge to the region.
12603         (try_generate_gimple_bb): Likewise.
12604         * sese.c (scalar_evolution_in_region): Adjust for
12605         instantiate_scev change.
12606         * tree-data-ref.h (graphite_find_data_references_in_stmt):
12607         Make nest parameter the edge into the region.
12608         (create_data_ref): Likewise.
12609         * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12610         entry edge into a region and adjust instantiate_scev calls.
12611         (create_data_ref): Likewise.
12612         (graphite_find_data_references_in_stmt): Likewise.
12613         (find_data_references_in_stmt): Pass the loop preheader edge
12614         from the nest argument.
12615         * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12616         parameter the edge into the region.
12617         (instantiate_parameters): Use the loop preheader edge as entry.
12618         * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12619         NULL loop.
12620         (get_instantiated_value_entry): Make instantiate_below parameter
12621         the edge into the region.
12622         (instantiate_scev_name): Likewise.  Adjust dominance checks,
12623         when we cannot use loop-based instantiation instantiate by
12624         walking use-def chains.
12625         (instantiate_scev_poly): Adjust.
12626         (instantiate_scev_binary): Likewise.
12627         (instantiate_scev_convert): Likewise.
12628         (instantiate_scev_not): Likewise.
12629         (instantiate_array_ref): Remove.
12630         (instantiate_scev_3): Likewise.
12631         (instantiate_scev_2): Likewise.
12632         (instantiate_scev_1): Likewise.
12633         (instantiate_scev_r): Do not blindly handle N-operand trees.
12634         Do not instantiate array-refs.  Handle all constants and invariants.
12635         (instantiate_scev): Make instantiate_below parameter
12636         the edge into the region.
12637         (resolve_mixers): Use the loop preheader edge for the region
12638         parameter to instantiate_scev_r.
12639         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12641 2017-10-13  Richard Biener  <rguenther@suse.de>
12643         PR tree-optimization/82525
12644         * graphite-isl-ast-to-gimple.c
12645         (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12646         out from ...
12647         (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12648         Fail code generation when we cannot represent the isl integer.
12649         (binary_op_to_tree): Elide modulo operations that are no-ops
12650         in the type we code generate.  Remove now superfluous code
12651         generation errors.
12653 2017-10-13  Richard Biener  <rguenther@suse.de>
12655         * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12656         (scop_detection::harmful_loop_in_region): Remove premature
12657         IV type restriction.
12658         (scop_detection::graphite_can_represent_scev): We can handle
12659         pointer IVs just fine.
12661 2017-10-13  Alan Modra  <amodra@gmail.com>
12663         * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12664         "Clobbers and Scratch Registers".  Add paragraph on
12665         alternative to clobbers for scratch registers and OpenBLAS
12666         example.
12668 2017-10-13  Alan Modra  <amodra@gmail.com>
12670         * doc/extend.texi (Clobbers): Correct vax example.  Delete old
12671         example of a memory input for a string of known length.  Move
12672         commentary out of table.  Add a number of new examples
12673         covering array memory inputs.
12675 2017-10-12  Martin Liska  <mliska@suse.cz>
12677         PR tree-optimization/82493
12678         * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12679         (test_range_functions): New function.
12680         (sbitmap_c_tests): Likewise.
12681         * selftest-run-tests.c (selftest::run_tests): Run new tests.
12682         * selftest.h (sbitmap_c_tests): New function.
12684         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12686 2017-10-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
12688         * config/rs6000/amo.h: Fix spacing issue.
12690 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12692         PR target/82498
12693         * config/i386/i386.md (*ashl<mode>3_mask_1,
12694         *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12695         *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12696         patterns.
12698 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12700         * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12701         (profile_probability): Set max_probability
12702         to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12703         in temporaries.
12704         * profile-count.c (profile_probability::differs_from_p): Do not
12705         rely on max_probaiblity == 10000
12707 2017-10-12  Jeff Law  <law@redhat.com>
12709         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12710         negative offsets.
12712 2017-10-12  Martin Sebor  <msebor@redhat.com>
12714         PR other/82301
12715         PR c/82435
12716         * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12717         (handle_alias_pairs): Call it.
12718         * common.opt (-Wattribute-alias): New option.
12719         * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12720         * doc/invoke.texi (-Wattribute-alias): Document.
12722 2017-10-12  Vladimir Makarov  <vmakarov@redhat.com>
12724         Revert
12725         2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12726         PR sanitizer/82353
12727         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12728         locations.
12729         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12730         (make_hard_regno_born, make_hard_regno_dead): Update
12731         bb_killed_pseudos and bb_gen_pseudos.
12733 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12735         * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12737 2017-10-12  Uros Bizjak  <ubizjak@gmail.com>
12739         * config/alpha/alpha.c (alpha_split_conditional_move):
12740         Use std::swap instead of manually swapping.
12741         (alpha_stdarg_optimize_hook): Ditto.
12742         (alpha_canonicalize_comparison): Ditto.
12744 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12746         * tree-loop-distribution.c (struct builtin_info): New struct.
12747         (struct partition): Refactor fields into struct builtin_info.
12748         (partition_free): Free struct builtin_info.
12749         (build_size_arg_loc, build_addr_arg_loc): Delete.
12750         (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12751         information from struct builtin_info.
12752         (find_single_drs): New function refactored from classify_partition.
12753         Also moved builtin validity checks to this function.
12754         (compute_access_range, alloc_builtin): New functions.
12755         (classify_builtin_st, classify_builtin_ldst): New functions.
12756         (classify_partition): Refactor code into functions find_single_drs,
12757         classify_builtin_st and classify_builtin_ldst.
12758         (distribute_loop): Don't do runtime alias check when distributing
12759         loop nest.
12760         (find_seed_stmts_for_distribution): New function.
12761         (pass_loop_distribution::execute): Refactor code finding seed
12762         stmts into above function.  Support distribution for the innermost
12763         two-level loop nest.  Adjust dump information.
12765 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12767         * tree-loop-distribution.c: Adjust the general comment.
12768         (NUM_PARTITION_THRESHOLD): New macro.
12769         (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12770         (classify_partition): Skip builtin pattern of loop nest's inner loop.
12771         (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12772         in call to build_partition_graph.
12773         (finalize_partitions): New parameter.  Make loop distribution more
12774         conservative by fusing more partitions.
12775         (distribute_loop): Don't do runtime alias check in case of loop nest
12776         distribution.
12777         (find_seed_stmts_for_distribution): New function.
12778         (prepare_perfect_loop_nest): New function.
12779         (pass_loop_distribution::execute): Refactor code finding seed stmts
12780         and loop nest into above functions.  Support loop nest distribution.
12781         Adjust dump information accordingly.
12783 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12785         * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12786         and set PTYPE_SEQUENTIAL for merged partition.
12788 2017-10-12  Richard Biener  <rguenther@suse.de>
12790         PR tree-optimization/69728
12791         Revert
12792         2017-09-19  Richard Biener  <rguenther@suse.de>
12794         PR tree-optimization/69728
12795         * graphite-sese-to-poly.c (schedule_error): New global.
12796         (add_loop_schedule): Handle empty domain by failing the
12797         schedule.
12798         (build_original_schedule): Handle schedule_error.
12800         * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12801         domain by returning an unchanged schedule.
12803 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12805         * genrecog.c (validate_pattern): For VEC_SELECT verify that
12806         CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12808 2017-10-12  Aldy Hernandez  <aldyh@redhat.com>
12810         * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12811         Handle params.def.
12813 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12815         PR c++/82159
12816         * expr.c (store_field): Don't optimize away bitsize == 0 store
12817         from CALL_EXPR with addressable return type.
12819 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12821         * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12822         * config/rs6000/rs6000.md (sel): Delete mode attribute.
12823         (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12824         *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12825         TARGET_ISEL instead of TARGET_ISEL<sel>.
12827 2017-10-11  David Edelsohn  <dje.gcc@gmail.com>
12829         * config/rs6000/rs6000.c
12830         (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12832 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12834         * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12835         Move up in file.
12836         (reg_or_cint_operand): Fix comment.
12837         (reg_or_zero_operand): New predicate.
12838         * config/rs6000/rs6000-protos.h (output_isel): Delete.
12839         * config/rs6000/rs6000.c (output_isel): Delete.
12840         * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12841         instead of reg_or_cint_operand.  Output instruction directly (not via
12842         output_isel).
12843         (isel_unsigned_<mode>): Ditto.
12844         (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12845         gpc_reg_operand.  Add an instruction alternative for this.  Output
12846         instruction directly.
12847         (*isel_reversed_unsigned_<mode>): Ditto.
12849 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12851         * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12852         (TARGET_CANONICALIZE_COMPARISON): Define.
12854 2017-10-11  Qing Zhao  <qing.zhao@oracle.com>
12856         PR target/81422
12857         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12858         Check whether the dest is REG before adding REG_EQUIV note.
12860 2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12862         PR sanitizer/82353
12863         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12864         locations.
12865         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12866         (make_hard_regno_born, make_hard_regno_dead): Update
12867         bb_killed_pseudos and bb_gen_pseudos.
12869 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12871         * incpath.h (enum incpath_kind): Name enum, prefix values.
12872         (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12873         * incpath.c (heads, tails): Use INC_MAX.
12874         (add_env_var_paths, add_standard_paths): Use incpath_kind.
12875         (merge_include_chains, split_quote_chain,
12876         register_include_chains): Update incpath_kind names.
12877         (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12878         * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12879         names.
12880         (add_framework_path, darwin_register_objc_includes): Likewise.
12881         * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12883 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12885         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12886         Do not use float_operator operator predicate.
12887         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12888         * config/i386/predicates.md (float_operator): Remove predicate.
12890 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12892         * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12893         (*jccxf_i387): Ditto.
12894         (*jcc<mode>_i387): Ditto.
12895         (*jccu<mode>_i387): Ditto.
12896         (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12897         (*jcc_*_i387 splitters): Remove.
12898         * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12899         * config/i386/i386.c (ix86_split_fp_branch): Remove.
12900         * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12901         Remove predicate.
12903 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12905         * profile-count.h (slow_safe_scale_64bit): New function.
12906         (safe_scale_64bit): New inline.
12907         (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12908         * profile-count.c: Include wide-int.h
12909         (slow_safe_scale_64bit): New.
12911 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12913         * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12914         HAS_DECL_ASSEMBLER_NAME_P.
12915         * gimple-expr.c (gimple_decl_printable_name: Check
12916         HAS_DECL_ASSEMBLER_NAME_P too.
12917         * ipa-utils.h (type_in_anonymous_namespace_p): Check
12918         DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
12919         (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
12920         * passes.c (rest_of_decl_compilation): Check
12921         HAS_DECL_ASSEMBLER_NAME_P too.
12922         * recog.c (verify_changes): Likewise.
12923         * tree-pretty-print.c (dump_decl_name): Likewise.
12924         * tree-ssa-structalias.c (alias_get_name): Likewise.  Reimplement.
12926         * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
12927         (SET_DECL_ASSEMBLER_NAME): Use it.
12928         (DECL_ASSEMBLER_NAME_SET_P): Likewise.
12929         (COPY_DECL_ASSEMBLER_NAME): Likewise.
12930         * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
12932 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12934         * config.gcc (i386, x86_64): Add extra objects.
12935         * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
12936         (ix86_min_insn_size): Declare.
12937         (ix86_issue_rate): Declare.
12938         (ix86_adjust_cost): Declare.
12939         (ia32_multipass_dfa_lookahead): Declare.
12940         (ix86_macro_fusion_p): Declare.
12941         (ix86_macro_fusion_pair_p): Declare.
12942         (ix86_bd_has_dispatch): Declare.
12943         (ix86_bd_do_dispatch): Declare.
12944         (ix86_core2i7_init_hooks): Declare.
12945         (ix86_atom_sched_reorder): Declare.
12946         * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
12947         (COSTS_N_BYTES): Move to x86-tune-costs.h.
12948         (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
12949         (rip_relative_addr_p): Rename to ...
12950         (ix86_rip_relative_addr_p): ... this one; export.
12951         (memory_address_length): Update.
12952         (ix86_issue_rate): Move to x86-tune-sched.c.
12953         (ix86_flags_dependent): Move to x86-tune-sched.c.
12954         (ix86_agi_dependent): Move to x86-tune-sched.c.
12955         (exact_dependency_1): Move to x86-tune-sched.c.
12956         (exact_store_load_dependency): Move to x86-tune-sched.c.
12957         (ix86_adjust_cost): Move to x86-tune-sched.c.
12958         (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
12959         (ix86_macro_fusion_p): Move to x86-tune-sched.c.
12960         (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
12961         (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
12962         (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
12963         (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
12964         (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
12965         (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
12966         (min_insn_size): Rename to ...
12967         (ix86_min_insn_size): ... this one; export.
12968         (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
12969         (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
12970         (core2i7_first_cycle_multipass_backtrack): Move to
12971         x86-tune-sched-core.c.
12972         (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
12973         (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
12974         (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
12975         (ix86_avoid_jump_mispredicts): Update.
12976         (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
12977         (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
12978         (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
12979         (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
12980         (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
12981         (MAX_INSN): Move to ix86-tune-sched-bd.c.
12982         (MAX_IMM): Move to ix86-tune-sched-bd.c.
12983         (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
12984         (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
12985         (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
12986         (MAX_LOAD): Move to ix86-tune-sched-bd.c.
12987         (MAX_STORE): Move to ix86-tune-sched-bd.c.
12988         (BIG): Move to ix86-tune-sched-bd.c.
12989         (enum dispatch_group): Move to ix86-tune-sched-bd.c.
12990         (enum insn_path): Move to ix86-tune-sched-bd.c.
12991         (get_mem_group): Move to ix86-tune-sched-bd.c.
12992         (is_cmp): Move to ix86-tune-sched-bd.c.
12993         (dispatch_violation): Move to ix86-tune-sched-bd.c.
12994         (is_branch): Move to ix86-tune-sched-bd.c.
12995         (is_prefetch): Move to ix86-tune-sched-bd.c.
12996         (init_window): Move to ix86-tune-sched-bd.c.
12997         (allocate_window): Move to ix86-tune-sched-bd.c.
12998         (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
12999         (is_end_basic_block): Move to ix86-tune-sched-bd.c.
13000         (process_end_window): Move to ix86-tune-sched-bd.c.
13001         (allocate_next_window): Move to ix86-tune-sched-bd.c.
13002         (find_constant): Move to ix86-tune-sched-bd.c.
13003         (get_num_immediates): Move to ix86-tune-sched-bd.c.
13004         (has_immediate): Move to ix86-tune-sched-bd.c.
13005         (get_insn_path): Move to ix86-tune-sched-bd.c.
13006         (get_insn_group): Move to ix86-tune-sched-bd.c.
13007         (count_num_restricted): Move to ix86-tune-sched-bd.c.
13008         (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
13009         (add_insn_window): Move to ix86-tune-sched-bd.c.
13010         (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
13011         (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
13012         (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
13013         (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
13014         (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
13015         (do_dispatch): Move to ix86-tune-sched-bd.c.
13016         (has_dispatch): Move to ix86-tune-sched-bd.c.
13017         * config/i386/t-i386: Add new object files.
13018         * config/i386/x86-tune-costs.h: New file.
13019         * config/i386/x86-tune-sched-atom.c: New file.
13020         * config/i386/x86-tune-sched-bd.c: New file.
13021         * config/i386/x86-tune-sched-core.c: New file.
13022         * config/i386/x86-tune-sched.c: New file.
13024 2017-10-11  Liu Hao  <lh_mouse@126.com>
13026         * pretty-print.c [_WIN32] (colorize_init): Remove.  Use
13027         the generic version below instead.
13028         (should_colorize): Recognize Windows consoles as terminals
13029         for MinGW targets.
13030         * pretty-print.c [__MINGW32__] (write_all): New function.
13031         [__MINGW32__] (find_esc_head): Likewise.
13032         [__MINGW32__] (find_esc_terminator): Likewise.
13033         [__MINGW32__] (eat_esc_sequence): Likewise.
13034         [__MINGW32__] (mingw_ansi_fputs): New function that handles
13035         ANSI escape codes.
13036         (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
13037         for MinGW targets.
13039 2017-10-11  Richard Biener  <rguenther@suse.de>
13041         * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
13042         Properly call analyze_scalar_evolution with the loop of the stmt.
13044 2017-10-11  Richard Biener  <rguenther@suse.de>
13046         * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
13047         * tree-core.h (tree_base): Add chrec_var union member.
13048         * tree.h (CHREC_VAR): Remove.
13049         (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
13050         * tree-chrec.h (build_polynomial_chrec): Adjust.
13051         * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
13052         * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
13054 2017-10-11  Marc Glisse  <marc.glisse@inria.fr>
13056         * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
13057         * match.pd: ... here.
13058         ((T) X == (T) Y): Relax condition.
13060 2017-10-11  Bin Cheng  <bin.cheng@arm.com>
13062         PR tree-optimization/82472
13063         * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
13064         comment.
13065         (break_alias_scc_partitions): Update postorder number.
13067 2017-10-11  Martin Liska  <mliska@suse.cz>
13069         PR sanitizer/82490
13070         * opts.c (parse_no_sanitize_attribute): Do not use error_value
13071         variable.
13072         * opts.h (parse_no_sanitize_attribute): Remove last argument.
13074 2017-10-11  Martin Liska  <mliska@suse.cz>
13076         * print-rtl.c (print_insn): Move declaration of idbuf
13077         to same scope as name.
13079 2017-10-11  Martin Liska  <mliska@suse.cz>
13081         Revert r253637:
13083         PR sanitizer/82484
13084         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13085         volatile arguments.
13087 2017-10-11  Martin Liska  <mliska@suse.cz>
13089         PR sanitizer/82484
13090         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13091         volatile arguments.
13093 2017-10-11  Adhemerval Zanella  <adhemerval.zanella@linaro.org>
13095         * config.gcc (default_gnu_indirect_function): Default to yes for
13096         arm*-*-linux* with glibc.
13098 2017-10-11  Richard Biener  <rguenther@suse.de>
13100         * tree-scalar-evolution.c (get_scalar_evolution): Handle
13101         default-defs and types we do not want to analyze.
13102         (interpret_loop_phi): Replace unreachable code with an assert.
13103         (compute_scalar_evolution_in_loop): Remove and inline ...
13104         (analyze_scalar_evolution_1): ... here, replacing condition with
13105         what makes the intent clearer.  Remove handling of cases
13106         get_scalar_evolution now handles.
13108 2017-10-10  Jim Wilson  <wilson@tuliptree.org>
13110         PR rtl-optimization/81434
13111         * haifa-sched.c (prune_ready_list): Init min_cost_group to 0.  Update
13112         comment for main loop.  In sched_group_found if, also add checks for
13113         pass and min_cost_group.
13115 2017-10-10  Segher Boessenkool  <segher@kernel.crashing.org>
13117         * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
13118         (rs6000_insn_cost): New function.
13119         * config/rs6000/rs6000.md (cost): New attribute.
13121 2017-10-10  Jakub Jelinek  <jakub@redhat.com>
13122             H.J. Lu  <hongjiu.lu@intel.com>
13124         PR target/79565
13125         PR target/82483
13126         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
13127         OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
13128         __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
13129         (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
13130         to OPTION_MASK_ISA_AVX512VL - builtins that have both
13131         OPTION_MASK_ISA_MMX and some other bit set require both
13132         mmx and the ISAs without the mmx bit.
13133         * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
13134         __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
13135         __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
13136         __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
13137         __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
13138         __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
13139         __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
13140         __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
13141         __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
13142         __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
13143         __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
13144         __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
13145         __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
13146         __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
13147         Add OPTION_MASK_ISA_MMX.
13149 2017-10-10  Andreas Tobler  <andreast@gcc.gnu.org>
13151         * config.gcc (armv7*-*-freebsd*): New target.
13152         (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
13154 2017-10-10  Jan Hubicka  <hubicka@ucw.cz>
13156         * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
13157         X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
13158         spot in the file.
13160 2017-10-10  Richard Sandiford  <richard.sandiford@linaro.org>
13162         * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
13163         a template parameter.
13164         (WIDE_INT_REF_FOR): Update accordingly.
13165         * tree.h (wi::int_traits <const_tree>): Delete.
13166         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
13167         (wi::to_widest, wi::to_offset): Use them.  Expand commentary.
13168         (wi::tree_to_wide_ref): New typedef.
13169         (wi::to_wide): New function.
13170         * calls.c (get_size_range): Use wi::to_wide when operating on
13171         trees as wide_ints.
13172         * cgraph.c (cgraph_node::create_thunk): Likewise.
13173         * config/i386/i386.c (ix86_data_alignment): Likewise.
13174         (ix86_local_alignment): Likewise.
13175         * dbxout.c (stabstr_O): Likewise.
13176         * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
13177         * expr.c (const_vector_from_tree): Likewise.
13178         * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
13179         * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
13180         (fold_negate_expr_1, int_const_binop_1, const_binop)
13181         (fold_convert_const_int_from_real, optimize_bit_field_compare)
13182         (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
13183         (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
13184         (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
13185         (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
13186         (fold_not_const, round_up_loc): Likewise.
13187         * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
13188         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
13189         (alloca_call_type): Likewise.
13190         * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
13191         * godump.c (go_output_typedef): Likewise.
13192         * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
13193         * internal-fn.c (get_min_precision): Likewise.
13194         * ipa-cp.c (ipcp_store_vr_results): Likewise.
13195         * ipa-polymorphic-call.c
13196         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
13197         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
13198         (ipa_modify_call_arguments): Likewise.
13199         * match.pd: Likewise.
13200         * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
13201         * print-tree.c (print_node_brief, print_node): Likewise.
13202         * stmt.c (expand_case): Likewise.
13203         * stor-layout.c (layout_type): Likewise.
13204         * tree-affine.c (tree_to_aff_combination): Likewise.
13205         * tree-cfg.c (group_case_labels_stmt): Likewise.
13206         * tree-data-ref.c (dr_analyze_indices): Likewise.
13207         (prune_runtime_alias_test_list): Likewise.
13208         * tree-dump.c (dequeue_and_dump): Likewise.
13209         * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
13210         * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
13211         * tree-pretty-print.c (dump_generic_node): Likewise.
13212         * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
13213         (simple_iv_with_niters): Likewise.
13214         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
13215         * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
13216         * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
13217         * tree-ssa-loop-niter.c (split_to_var_and_offset)
13218         (refine_value_range_using_guard, number_of_iterations_ne_max)
13219         (number_of_iterations_lt_to_ne, number_of_iterations_lt)
13220         (get_cst_init_from_scev, record_nonwrapping_iv)
13221         (scev_var_range_cant_overflow): Likewise.
13222         * tree-ssa-phiopt.c (minmax_replacement): Likewise.
13223         * tree-ssa-pre.c (compute_avail): Likewise.
13224         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
13225         (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
13226         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
13227         * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
13228         * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
13229         * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
13230         (dump_case_nodes, try_switch_expansion): Likewise.
13231         * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
13232         (vect_do_peeling): Likewise.
13233         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13234         * tree-vect-stmts.c (vectorizable_load): Likewise.
13235         * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
13236         (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
13237         (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
13238         (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
13239         (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
13240         (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
13241         (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
13242         (evrp_dom_walker::before_dom_children): Likewise.
13243         * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13244         (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13245         (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13246         (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13247         (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13248         (get_range_pos_neg): Likewise.
13249         * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13250         * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13251         * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13252         * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13253         * config/avr/avr.c (avr_fold_builtin): Likewise.
13254         * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13255         * config/msp430/msp430.c (msp430_attr): Likewise.
13256         * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13257         * config/powerpcspe/powerpcspe-c.c
13258         (altivec_resolve_overloaded_builtin): Likewise.
13259         * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13260         (rs6000_expand_ternop_builtin): Likewise.
13261         * config/rs6000/rs6000-c.c
13262         (altivec_resolve_overloaded_builtin): Likewise.
13263         * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13264         (rs6000_expand_ternop_builtin): Likewise.
13265         * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13267 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13269         * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13270         when copying loop nest with only one inner loop.
13272 2017-10-10  Richard Biener  <rguenther@suse.de>
13274         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13275         blocks if SCEV is active.
13276         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13277         dead code.
13278         (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13279         (scev_initialize): Assert we are not yet initialized.
13281 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13283         * tree-loop-distribution.c (generate_loops_for_partition): Remove
13284         inner loop's exit stmt by making it always exit the loop, otherwise
13285         we would generate an infinite empty loop.
13287 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13289         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13290         renaming variables in new preheader if it's deleted.
13292 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13294         * tree-loop-distribution.c (struct partition): Remove unused field
13295         loops of the structure.
13296         (partition_alloc, partition_free): Ditto.
13297         (build_rdg_partition_for_vertex): Ditto.
13299 2017-10-09  Jeff Law  <law@redhat.com>
13301         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13302         return type to match prototype and documentation.
13304 2010-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13306         * config/rs6000/rs6000.c (processor_costs): Move to ...
13307         * config/rs6000/rs6000.h: ... here.
13308         (rs6000_cost): Declare.
13310 2017-10-09  Eric Botcazou  <ebotcazou@adacore.com>
13312         * except.c (setjmp_fn): New global variable.
13313         (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13314         (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13315         if DONT_USE_BUILTIN_SETJMP is defined.
13317 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13319         * target.def (insn_cost): New hook.
13320         * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13321         * doc/tm.texi: Regenerate.
13322         * rtlanal.c (insn_cost): Use the new hook.
13324 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13326         * combine.c (combine_validate_cost): Compute the new insn_cost,
13327         not just pattern_cost.
13328         (try_combine): Adjust comment.
13330 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13332         * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13333         insn_cost.
13334         * combine.c (uid_insn_cost): Adjust comment.
13335         (combine_validate_cost): Adjust comment.  Use pattern_cost instead
13336         of insn_rtx_cost
13337         (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13338         * dse.c (find_shift_sequence): Ditto.
13339         * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13340         (bb_valid_for_noce_process_p): Use pattern_cost.
13341         * rtl.h (insn_rtx_cost): Delete.
13342         (pattern_cost): New prototype.
13343         (insn_cost): New prototype.
13344         * rtlanal.c (insn_rtx_cost): Rename to...
13345         (pattern_cost): ... this.
13346         (insn_cost): New.
13348 2017-10-09  Uros Bizjak  <ubizjak@gmail.com>
13350         * config/i386/i386.md (*jcc_2): Remove insn pattern.
13351         (*jcc<mode>_0_r_i387): Ditto.
13352         (*jccxf_r_i387): Ditto.
13353         (*jcc<mode>_r_i387): Ditto.
13354         (*jccu<mode>_r_i387): Ditto.
13355         (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13356         (*jcc): Rename from *jcc_1.
13358 2017-10-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13360         * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13361         deferred rescans after the lvx/stvx recombination pre-pass.
13363 2017-10-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
13365         * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13366         memory operation instruction support.
13367         * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13368         (rs6000-ibm-aix[789]*): Likewise.
13369         * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13370         Document new functions.
13372 2017-10-09  Richard Biener  <rguenther@suse.de>
13374         PR tree-optimization/82397
13375         * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13376         equality only for semantically equal trees.
13378 2017-10-09  Richard Biener  <rguenther@suse.de>
13380         PR tree-optimization/82449
13381         * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13382         * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13383         allow constant addresses.
13384         * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13385         are linear.
13387 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13389         * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13390         flags.
13392 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13394         PR target/82463
13395         * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13396         definitions.
13398 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13400         PR target/82465
13401         * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13403 2017-10-09  Jakub Jelinek  <jakub@redhat.com>
13405         PR target/82464
13406         * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13407         s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13409 2017-10-09  Richard Sandiford  <richard.sandiford@linaro.org>
13411         * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13412         (WI_BINARY_PREDICATE_RESULT): Likewise.
13413         (wi::binary_traits::operator_result): New type.
13414         (wi::binary_traits::predicate_result): Likewise.
13415         (generic_wide_int::operator~, unary generic_wide_int::operator-)
13416         (generic_wide_int::operator==, generic_wide_int::operator!=)
13417         (generic_wide_int::operator&, generic_wide_int::and_not)
13418         (generic_wide_int::operator|, generic_wide_int::or_not)
13419         (generic_wide_int::operator^, generic_wide_int::operator+
13420         (binary generic_wide_int::operator-, generic_wide_int::operator*):
13421         Delete.
13422         (operator~, unary operator-, operator==, operator!=, operator&)
13423         (operator|, operator^, operator+, binary operator-, operator*): New
13424         functions.
13425         * expr.c (get_inner_reference): Use wi::bit_and_not.
13426         * fold-const.c (fold_binary_loc): Likewise.
13427         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13428         * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13429         (bit_value_binop): Likewise.
13430         * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13431         * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13432         (extract_range_from_binary_expr_1): Likewise.
13433         (masked_increment): Likewise.
13434         (simplify_bit_ops_using_ranges): Likewise.
13436 2017-10-09  Martin Jambor  <mjambor@suse.cz>
13438         PR hsa/82416
13439         * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13440         * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13441         (hsa_type_for_scalar_tree_type): Use it.  Always force min32int for
13442         COMPLEX types.
13443         (hsa_fixup_mov_insn_type): New function.
13444         (hsa_op_with_type::get_in_type): Use it.
13445         (hsa_build_append_simple_mov): Likewise.  Allow sub-32bit
13446         immediates in an assert.
13447         (hsa_op_with_type::extend_int_to_32bit): New method.
13448         (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13449         types.  Convert to dest type if necessary.
13450         (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13451         (reg_for_gimple_ssa): Pass false as min32int to
13452         hsa_type_for_scalar_tree_type.
13453         (gen_hsa_addr): Fixup type when creating addresable temporary.
13454         (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13455         (gen_hsa_unary_operation): Extend operands and convert to dest type if
13456         necessary.  Call hsa_fixup_mov_insn_type.
13457         (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13458         extend operands and convert to dest type if necessary.
13459         (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13460         to dest type if necessary.
13461         (set_output_in_type): Call hsa_fixup_mov_insn_type.  Just ude dest
13462         if conversion nt necessary and size matches.
13463         (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13464         to dest type if necessary.
13465         (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13466         (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13467         necessary.
13468         (gen_hsa_clrsb): Likewise.
13469         (gen_hsa_ffs): Likewise.
13470         (gen_hsa_divmod): Extend operands and convert to dest type if
13471         necessary.
13472         (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13474 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13476         * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13477         Remove empty default arguments.  Use a brace block as output
13478         statement.
13479         (conditional return): Ditto.
13480         (jump): Ditto.
13481         (indirect_jump): Ditto.  Use b%T0 instead of bctr/blr.
13482         (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13483         Ditto.
13484         (group_ending_nop): Ditto.
13485         (doloop_end): Ditto.
13486         (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13487         (splitters for those): Ditto.
13489 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13491         * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13492         a conditional jump (and the compare for it) so that pc_rtx is the
13493         last operand.
13494         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13495         for the deleted and renamed ctr<mode>_internal[234] patterns.
13496         * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13497         Delete second conditional return pattern.
13498         (ctr<mode>_internal2): Delete this second bdnz pattern.
13499         (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13500         (ctr<mode>_internal4): Delete this second bdz pattern.
13502 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13504         * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13505         (always_initialized_rtx_for_ssa_name_p): New predicate.
13506         * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13507         (finish_out_of_ssa): Free new field of SA.
13508         * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13509         * tree-ssa-coalesce.c: Include tree-ssa.h.
13510         (get_parm_default_def_partitions): Remove extern keyword.
13511         (get_undefined_value_partitions): New function.
13512         * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13513         not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13514         uninitialized bits.
13515         * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13517 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13519         * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13521 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13523         * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13524         for targets that preffer 128bit.
13526 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13528         * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13530 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13532         * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13533         on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13534         prevent DSE.
13535         (thumb_set_return_address): Likewise.
13537 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13539         * common/config/arm/arm-common.c (arm_except_unwind_info):
13540         Handle DWARF2_UNWIND_INFO.
13542 2017-10-07  Michael Collison <michael.collison@arm.com>
13544         * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13545         New pattern.
13547 2017-10-07  Eric Botcazou  <ebotcazou@adacore.com>
13549         * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13550         of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13551         * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13552         defined, force the creation of a new block for a dispatch label.
13554 2017-10-07  Jan Hubicka  <hubicka@ucw.cz>
13556         * invoke.texi (Wsuggest-attribute=cold): Document.
13557         * common.opt (Wsuggest-attribute=cold): New
13558         * ipa-pure-const.c (warn_function_cold): New function.
13559         * predict.c (compute_function_frequency): Use it.
13560         * predict.h (warn_function_cold): Declare.
13562 2017-10-06  Jan Hubicka  <hubicka@ucw.cz>
13564         * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13565         Update profile.
13567 2017-10-06  Martin Liska  <mliska@suse.cz>
13569         * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13570         keyword for member functions.
13571         (struct sanopt_tree_couple): New struct.
13572         (struct sanopt_tree_couple_hash): New function.
13573         (struct sanopt_ctx): Add new hash_map.
13574         (has_dominating_ubsan_ptr_check): New function.
13575         (record_ubsan_ptr_check_stmt): Likewise.
13576         (maybe_optimize_ubsan_ptr_ifn): Likewise.
13577         (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13578         (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13580 2017-10-06  Sudakshina Das  <sudi.das@arm.com>
13582         PR target/82440
13583         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13584         aarch64_simd_valid_immediate on CONST_VECTORs.
13585         (aarch64_reg_or_bic_imm): Likewise.
13587 2017-10-06  Wilco Dijkstra  <wdijkstr@arm.com>
13589         PR rtl-optimization/82396
13590         * haifa-sched.c (ready_sort_real): Disable qsort checking.
13592 2017-10-06  Sebastian Pop  <sebpop@gmail.com>
13594         * graphite-dependences.c (scop_get_reads): Move code to...
13595         (scop_get_must_writes): Move code to...
13596         (scop_get_may_writes): Move code to...
13597         (scop_get_reads_and_writes): ... here.
13598         (scop_get_dependences): Call scop_get_reads_and_writes.
13600 2017-10-06  Jakub Jelinek  <jakub@redhat.com>
13602         PR tree-optimization/82434
13603         * fold-const.h (can_native_encode_type_p,
13604         can_native_encode_string_p): Remove.
13605         * fold-const.c (native_encode_int): Formatting fixes.  If ptr is NULL,
13606         don't encode anything, just return what would be otherwise returned.
13607         (native_encode_fixed, native_encode_complex, native_encode_vector):
13608         Likewise.
13609         (native_encode_string): Likewise.  Inline by hand
13610         can_native_encode_string_p.
13611         (can_native_encode_type_p): Remove.
13612         (can_native_encode_string_p): Remove.
13613         * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13614         STRING_CSTs using can_native_encode_string_p, test all
13615         CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13616         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13617         argument from native_encode_expr.
13618         (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13619         (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13620         but just 2.
13622 2017-10-06  Richard Biener  <rguenther@suse.de>
13624         PR tree-optimization/82397
13625         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13626         operand_equal_p but rely on data_ref_compare_tree for detecting
13627         equalities.
13628         (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13629         to match up with dr_group_sort_cmp.
13631 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13633         PR target/82322
13634         * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13635         builtin.
13636         * config/s390/s390-builtin-types.def: Regenerate.
13638 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13640         PR target/82317
13641         * config/s390/s390-builtin-types.def: Regenerate.
13642         * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13643         Change flag from B_VXE to B_VX.
13644         (s390_vec_min_dbl): Remove B_VXE flag.
13646 2017-10-06  Richard Biener  <rguenther@suse.de>
13648         * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13649         (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13650         translate_isl_ast_to_gimple::is_valid_rename,
13651         translate_isl_ast_to_gimple::get_rename,
13652         translate_isl_ast_to_gimple::get_def_bb_for_const,
13653         translate_isl_ast_to_gimple::get_new_name,
13654         translate_isl_ast_to_gimple::collect_all_ssa_names,
13655         translate_isl_ast_to_gimple::copy_loop_phi_args,
13656         translate_isl_ast_to_gimple::collect_all_ssa_names,
13657         translate_isl_ast_to_gimple::copy_loop_phi_args,
13658         translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13659         translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13660         translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13661         translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13662         translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13663         translate_isl_ast_to_gimple::copy_cond_phi_args,
13664         translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13665         translate_isl_ast_to_gimple::edge_for_new_close_phis,
13666         translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13667         translate_isl_ast_to_gimple::rename_uses,
13668         translate_isl_ast_to_gimple::rename_all_uses): Remove.
13669         (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13670         (set_rename_for_each_def): Likewise.
13671         (graphite_copy_stmts_from_block): Handle debug stmt resetting
13672         here.  Handle rewriting SCEV analyzable uses here.
13673         (copy_bb_and_scalar_dependences): Generate code for PHI
13674         copy-in/outs.
13675         (graphite_regenerate_ast_isl): Adjust.
13676         * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13677         (add_write, add_read): New functions.
13678         (build_cross_bb_scalars_def): Use it and simplify.
13679         (build_cross_bb_scalars_use): Likewise.
13680         (graphite_find_cross_bb_scalar_vars): Inline into...
13681         (try_generate_gimple_bb): ...here.  Add dependences for PHIs,
13682         simulating out-of-SSA.  Compute liveout and add dependencies.
13683         (build_scops): Force an empty entry block.
13684         * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13685         members.
13686         (sese_build_liveouts): Declare.
13687         (sese_trivially_empty_bb_p): Likewise.
13688         * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13689         compute liveout and debug_liveout.
13690         (sese_bad_liveouts_use): Remove.
13691         (sese_reset_debug_liveouts_bb): Likewise.
13692         (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13693         (sese_build_liveouts): Build liveout and debug_liveout and store
13694         it in region.
13695         (new_sese_info): Adjust.
13696         (free_sese_info): Likewise.
13697         (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13698         do not build liveout here.
13699         (move_sese_in_condition): Adjust region entry.
13700         (scev_analyzable_p): Match up with chrec_apply requirements.
13701         (sese_trivially_empty_bb_p): New.
13702         * tree-into-ssa.c (get_reaching_def): Properly support generating
13703         default-defs for incremental rewrite of anonymous names.
13705 2017-10-06  Richard Biener  <rguenther@suse.de>
13707         * graphite-sese-to-poly.c (extract_affine): For casts increasing
13708         precision do not perform modulo reduction.
13710 2017-10-06  Richard Biener  <rguenther@suse.de>
13712         PR tree-optimization/82436
13713         * tree-vect-slp.c (vect_supported_load_permutation_p): More
13714         conservatively choose the vectorization factor when checking
13715         whether we can perform the required load permutation.
13716         (vect_transform_slp_perm_load): Assert when we may not fail.
13718 2017-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
13720         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13721         message for incompatible -msdata=* and -mcall-* options.
13723 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13725         * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13726         rate for post-reload scheduling.
13728 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13730         * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13732 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13734         * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13735         to improve monte carlo in scimark.
13737 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13739         * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13740         pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13741         athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13742         pentium4_cost, nocona_cost): Set reassociation width to 1.
13743         (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13744         width to 2 for fp operations and 1 otherwise.
13745         (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13746         for int and fp.
13747         (atom_cost): Set reassociation width to 2.
13748         (slm_cost, generic_cost): Set fp reassociation width
13749         to 2 and 1 otherwise.
13750         (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13751         (core_cost): Set fp reassociation width to 4 and vector to 2.
13752         (ix86_reassociation_width): Rewrite using cost table; special case
13753         plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13754         and TARGET_AVX128_OPTIMAL.
13755         * config/i386/i386.h (processor_costs): Add
13756         reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13757         (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13758         TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13759         * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13760         (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13761         (X86_TUNE_VECTOR_PARALLEL_EXECUTION):  Remove.
13763 2017-10-05  Nathan Sidwell  <nathan@acm.org>
13765         * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13767 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13769         * config/arm/arm.c (arm_test_fpu_data): New.
13770         (arm_run_selftests): Call arm_test_fpu_data.
13772 2017-10-04  Nathan Sidwell  <nathan@acm.org>
13774         * toplev.c (toplev::main): Remove excess parens on pretty_printer
13775         decl.
13776         * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13778 2017-10-04  Sudakshina Das  <sudi.das@arm.com>
13780         * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13781         check type for aarch64_simd_valid_immediate.
13782         (aarch64_output_simd_mov_immediate): Update prototype.
13783         (aarch64_simd_valid_immediate): Update prototype.
13784         * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13785         support for ORR-immediate.
13786         (and<mode>3): modified pattern to add support for BIC-immediate.
13787         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13788         now checks for valid immediate for BIC and ORR based on new enum
13789         argument.
13790         (aarch64_output_simd_mov_immediate): Function now used to output
13791         BIC/ORR imm as well based on new enum argument.
13792         * config/aarch64/constraints.md (Do): New vector immediate constraint.
13793         (Db) : Likewise.
13794         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13795         (aarch64_reg_or_bic_imm): Likewise.
13797 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13799         * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13800         ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13802 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13804         Revert r253399:
13806         PR rtl-optimization/82396
13807         * haifa-sched.c (autopref_multipass_init): Simplify
13808         initialization.
13809         (autopref_rank_data): Simplify sort order.
13810         * sched-int.h (autopref_multipass_data_): Remove
13811         multi_mem_insn_p, min_offset and max_offset.
13813 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13815         * doc/sourcebuild.texi: Document vect_peeling_profitable.
13817 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13819         * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13820         vect_doubleint_cvt.
13822 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13824         * doc/sourcebuild.texi: Document vect_long_mult.
13826 2017-10-04  Richard Sandiford  <richard.sandiford@linaro.org>
13828         PR tree-optimization/82413
13829         * fold-const.c (build_range_check): Use widest_int when comparing
13830         the maximum ETYPE value with HIGH.
13832 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13834         PR rtl-optimization/82396
13835         * haifa-sched.c (autopref_multipass_init): Simplify
13836         initialization.
13837         (autopref_rank_data): Simplify sort order.
13838         * sched-int.h (autopref_multipass_data_): Remove
13839         multi_mem_insn_p, min_offset and max_offset.
13841 2017-10-04  Jakub Jelinek  <jakub@redhat.com>
13843         PR tree-optimization/82381
13844         * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13845         oeN->rank first.  Return 1 or -1 if one op is SSA_NAME and the other
13846         is not.
13848         PR tree-optimization/82374
13849         * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13850         DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13851         DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13852         current_function_decl to the new decl.
13854 2017-10-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
13856         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13857         helper macro for IEEE float128 hardware built-in functions.
13858         (SQRTF128_ODD): Add built-in functions with the round-to-odd
13859         semantics.
13860         (TRUNCF128_ODD): Likewise.
13861         (ADDF128_ODD): Likewise.
13862         (SUBF128_ODD): Likewise.
13863         (MULF128_ODD): Likewise.
13864         (DIVF128_ODD): Likewise.
13865         (FMAF128_ODD): Likewise.
13866         * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13867         UNSPEC_TRUNC_ROUND_TO_ODD.
13868         (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13869         (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13870         floating point round to odd instructions.
13871         (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13872         (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13873         (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13874         (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13875         (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13876         (trunc<mode>sf2_hw): Change the truncate with round to odd
13877         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13878         (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13879         to odd hardware instructions.
13880         (sub<mode>3_odd): Likewise.
13881         (mul<mode>3_odd): Likewise.
13882         (div<mode>3_odd): Likewise.
13883         (sqrt<mode>2_odd): Likewise.
13884         (fma<mode>4_odd): Likewise.
13885         (fms<mode>4_odd): Likewise.
13886         (nfma<mode>4_odd): Likewise.
13887         (nfms<mode>4_odd): Likewise.
13888         (trunc<mode>df2_odd): Change the truncate with round to odd
13889         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.  Add a generator
13890         function.
13891         * doc/extend.texi (PowerPC built-in functions): Update documentation
13892         for existing IEEE float128-bit built-in functions.  Add built-in
13893         functions that generate the IEEE 128-bit floating point round to
13894         odd instructions.
13896 2017-10-03  Segher Boessenkool  <segher@kernel.crashing.org>
13898         PR rtl-optimization/77729
13899         * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13900         to (X&(C1&~C2))|C2 transformations.
13902 2017-10-03  Martin Jambor  <mjambor@suse.cz>
13904         PR tree-optimization/82363
13905         * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13906         mismatch, mark lacc written regardless of racc.
13908 2017-10-03  Jakub Jelinek  <jakub@redhat.com>
13910         PR tree-optimization/82381
13911         * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13912         stmt_to_insert nor wheather SSA_NAMEs are default defs.
13913         Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13914         fallthrough into reassoc_stmt_dominates_stmt_p.
13916         PR target/82386
13917         * combine.c (combine_instructions): Don't combine in unreachable
13918         basic blocks.
13920 2017-10-02  Peter Bergner  <bergner@vnet.ibm.com>
13922         PR target/80210
13923         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
13924         function to not use the have_cpu variable.  Do not set cpu_index,
13925         rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
13926         or the default cpu.
13927         (rs6000_valid_attribute_p): Remove duplicate initializations of
13928         old_optimize and func_optimize.
13929         (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
13930         (rs6000_activate_target_options): Make global.
13931         * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
13932         prototype.
13934 2017-10-02  Jakub Jelinek  <jakub@redhat.com>
13936         * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
13937         if *poffset + *pmax_size overflows in HOST_WIDE_INT.
13938         Set *poffset to 0 and *psize and *pmax_size to -1 if
13939         *poffset + *psize overflows in HOST_WIDE_INT.
13941         PR tree-optimization/82387
13942         PR tree-optimization/82388
13943         PR tree-optimization/82389
13944         * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
13945         instead of live_bytes non-NULL.
13947 2017-10-02  Georg-Johann Lay  <avr@gjlay.de>
13949         PR target/41076
13950         * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
13951         alternative.
13953 2017-10-02  Richard Biener  <rguenther@suse.de>
13955         * graphite-isl-ast-to-gimple.c (set_codegen_error): With
13956         -fchecking and --param graphite-allow-codegen-errors=0 ICE.
13957         * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
13959 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13961         * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
13962         requested precision matches the type's.
13963         * calls.c (alloc_max_size): Calculate the new candidate size as
13964         a widest_int and use wi::to_widest when comparing it with the
13965         current candidate size.
13966         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
13967         zero rather than integer_zero_node.
13968         * match.pd: Check for a no-op conversion before using wi::add
13969         rather than after.  Use tree_to_uhwi when summing small shift
13970         counts into an unsigned int.
13972 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13973             Alan Hayward  <alan.hayward@arm.com>
13974             David Sherwood  <david.sherwood@arm.com>
13976         PR target/71307
13977         * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
13978         (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
13979         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13980         POINTER_AND_FP_REGS.
13982 2017-10-02  Richard Biener  <rguenther@suse.de>
13984         PR tree-optimization/82355
13985         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
13986         a mapping for the enclosing loop but avoid generating one for
13987         the loop tree root.
13988         (copy_bb_and_scalar_dependences): Remove premature codegen
13989         error on PHIs in blocks duplicated into multiple places.
13990         * graphite-scop-detection.c
13991         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
13992         in the region use it as loop and nest to analyze the DR in.
13993         (try_generate_gimple_bb): Likewise.
13994         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
13995         (add_loop_constraints): For blocks in a loop not in the region
13996         create a dimension with a single iteration.
13997         * sese.h (gbb_loop_at_index): Remove assert.
13999 2017-10-01  Kevin Buettner  <kevinb@redhat.com>
14001         * omp-expand.c (adjust_context_scope): New function.
14002         (expand_parallel_call): Call adjust_context_scope.
14004 2017-10-01  Jeff Law  <law@redhat.com>
14006         * tree-ssa-dom.c (optimize_stmt): Make this a method within the
14007         dom_opt_dom_walker class with direct access to private members.
14008         Add comments.  Call test_for_singularity.
14009         (dom_opt_dom_walker::before_dom_children): Corresponding changes.
14010         (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
14011         m_dummy_cond anymore.
14012         (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
14013         class ctor.
14014         (pass_dominator:execute): Build the dummy_cond here and pass it
14015         to the dom_opt_dom_walker ctor.
14016         (test_for_singularity): New function.
14018 2017-09-30  Krister Walfridsson  <krister.walfridsson@gmail.com>
14019             Maya Rashish  <coypu@sdf.org>
14021         * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
14022         netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
14023         (alpha*-*-netbsd*) Use nbsd_tm_file.
14024         (arm*-*-netbsdelf*) Likewise.
14025         (i[34567]86-*-netbsdelf*) Likewise.
14026         (x86_64-*-netbsd*) Likewise.
14027         (mips*-*-netbsd*) Likewise.
14028         (powerpc-*-netbsd*) Likewise.
14029         (sh*-*-netbsd*) Likewise.
14030         (sparc-*-netbsdelf*) Likewise.
14031         (sparc64-*-netbsd*) Likewise.
14032         (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
14033         to tm_defines.
14034         (vax-*-netbsdelf*) Likewise.
14035         * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
14036         (UINT_FAST8_TYPE) Likewise.
14037         (INT_FAST16_TYPE) Check CHAR_FAST16.
14038         (UINT_FAST16_TYPE) Likewise.
14040 2017-09-30  Jakub Jelinek  <jakub@redhat.com>
14042         PR target/82361
14043         * config/i386/i386.md
14044         (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
14045         (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
14046         *divmodsi4_zext_2): New define_insn_and_split.
14047         (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
14048         (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
14049         (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
14050         *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
14051         New define_insn_and_split.
14052         (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
14053         * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
14054         operands[1] having DImode when mode is SImode.
14056         * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
14057         always SImode for DIV and MOD in REG_EQUAL notes.
14059 2017-09-29  Yury Gribov  <tetra2005@gmail.com>
14061         PR middle-end/82319
14062         * match.pd: Fix handling of NaNs in pattern.
14064 2017-09-29  Jeff Law  <law@redhat.com>
14066         * sbitmap.c (bitmap_bit_in_range_p): New function.
14067         * sbitmap.h (bitmap_bit_in_range_p): Prototype.
14068         * tree-ssa-dse.c (live_bytes_read): New function.
14069         (dse_classify_store): Ignore reads of dead bytes.
14071         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
14072         typos and whitespace errors.
14073         * config/i386/predicates.md (address_no_seg_operand): Likewise.
14074         * config/s390/s390.c (s390_emit_prologue): Likewise.
14076 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14078         PR target/81481
14079         * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
14080         with a symbol for LRA.
14082 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14084         PR rtl-optimization/82338
14085         * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
14087 2017-09-29  Alexander Monakov  <amonakov@ispras.ru>
14089         * genmodes.c (calc_wider_mode): Suppress qsort macro.
14090         * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
14091         (qsort_chk): Declare.
14092         * vec.c [CHECKING_P] (qsort_chk_error): New static function.
14093         (qsort_chk): New function.
14095 2017-09-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14097         PR tree-optimization/82337
14098         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
14099         phi definition if the PHI result appears in an abnormal PHI.
14100         (find_basis_for_base_expr): Don't record a basis if the LHS of the
14101         basis appears in an abnormal PHI.
14103 2017-09-29  Richard Biener  <rguenther@suse.de>
14105         * graphite-isl-ast-to-gimple.c
14106         (translate_isl_ast_to_gimple::set_codegen_error): New function.
14107         (binary_op_to_tree): Use it.
14108         (get_rename_from_scev): Likewise.
14109         (copy_loop_phi_nodes): Likewise.
14110         (copy_bb_and_scalar_dependences): Likewise.
14111         (translate_pending_phi_nodes): Likewise.
14113 2017-09-29  Jakub Jelinek  <jakub@redhat.com>
14115         PR target/82339
14116         * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
14117         for movabsq $(i32 << shift), r64.
14119 2017-09-28  Uros Bizjak  <ubizjak@gmail.com>
14121         * config/i386/i386.c (ix86_print_operand_address_as): Do not check
14122         index when encoding %esp as %rsp to avoid 0x67 prefix.
14124 2017-09-28  Sergey Shalnov  <Sergey.Shalnov@intel.com>
14126         * config/i386/i386.md (*movsf_internal, *movdf_internal):
14127         Return 256-bit AVX modes for TARGET_PREFER_AVX256.
14129 2017-09-28  Thomas Preud'homme  <thomas.preudhomme@arm.com>
14131         * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
14132         Extensions with more than 16 double VFP registers.
14133         (cmse_nonsecure_entry_clear_before_return): Remove second entry of
14134         to_clear_mask and all code related to it.  Replace the remaining
14135         entry by a sbitmap and adapt code accordingly.
14137 2017-09-28  Henry Linjamäki  <henry.linjamaki@parmance.com>
14139         * brig-builtins.def: Change pure attributes to const.
14141 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14143         * config.gcc (default_gnu_indirect_function): Default to yes for
14144         sparc*-*-linux* with glibc.
14146 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14148         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
14149         (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
14150         when creating .init_array and .fini_array sections with priority
14151         specified.
14153 2017-09-27  Christophe Lyon  <christophe.lyon@linaro.org>
14155         PR target/71727
14156         * config/aarch64/aarch64.c
14157         (aarch64_builtin_support_vector_misalignment): Always return false
14158         when misalignment is unknown.
14160 2017-09-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14162         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
14163         this function to return false if the definition used by the swap
14164         instruction is artificial, or if the memory address from which the
14165         constant value is loaded is not represented by a base address held
14166         in a register or if the base address register is a frame or stack
14167         pointer.  Additionally, return false if the base address of the
14168         loaded constant is a SYMBOL_REF but is not considered to be a
14169         constant.
14170         (replace_swapped_load_constant): New function.
14171         (rs6000_analyze_swaps): Add a new pass to replace a swap of a
14172         loaded constant vector with a load of a swapped constant vector.
14174 2017-09-27  Carl Love  <cel@us.ibm.com>
14176         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14177         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14178         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14179         fctiw instruction.
14181 2017-09-27  Alexander Monakov  <amonakov@ispras.ru>
14183         * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
14184         first, always call autopref_rank_data otherwise.
14186 2017-09-27  Richard Biener  <rguenther@suse.de>
14188         * graphite-scop-detection.c (find_scop_parameters): Move
14189         loop bound handling ...
14190         (gather_bbs::before_dom_children): ... here, avoiding the need
14191         to build scop_info->loop_nest.
14192         (record_loop_in_sese): Remove.
14193         * sese.h (sese_info_t::loop_nest): Remove.
14194         * sese.c (new_sese_info): Do not allocate loop_nest.
14195         (free_sese_info): Do not free loop_nest.
14197 2017-09-27  Jakub Jelinek  <jakub@redhat.com>
14199         PR c++/82159
14200         * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
14201         lhs from calls if the lhs has addressable type.
14203 2017-09-27  Richard Biener  <rguenther@suse.de>
14205         * graphite.h (scop::max_alias_set): New member.
14206         * graphite-scop-detection.c: Remove references to non-existing
14207         --param in comments.
14208         (build_alias_sets): Record the maximum alias set used for drs.
14209         (build_scops): Support zero as unlimited for
14210         --param graphite-max-arrays-per-scop.
14211         * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
14212         and inline into ...
14213         (build_poly_sr_1): ... here.  Compute alias set based on the
14214         maximum alias set used for drs rather than
14215         PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
14217 2017-09-27  Richard Biener  <rguenther@suse.de>
14219         * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
14220         --param loop-block-tile-size=0 to disable tiling.
14222 2017-09-27  Richard Biener  <rguenther@suse.de>
14224         * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
14225         (graphite-max-nb-scop-params): Document special value zero.
14226         * domwalk.h (dom_walker::STOP): New symbolical constant.
14227         (dom_walker::dom_walker): Add optional parameter for bb to
14228         RPO mapping.
14229         (dom_walker::~dom_walker): Declare.
14230         (dom_walker::before_dom_children): Document STOP return value.
14231         (dom_walker::m_user_bb_to_rpo): New member.
14232         (dom_walker::m_bb_to_rpo): Likewise.
14233         * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
14234         mapping here if not provided by the user.
14235         (dom_walker::~dom_walker): Free bb to RPO mapping if not
14236         provided by the user.
14237         (dom_walker::STOP): Define.
14238         (dom_walker::walk): Do not compute bb to RPO mapping here.
14239         Support STOP return value from before_dom_children to stop
14240         walking.
14241         * graphite-optimize-isl.c (optimize_isl): If the schedule
14242         is the same still generate code if -fgraphite-identity
14243         or -floop-parallelize-all are given.
14244         * graphite-scop-detection.c: Include cfganal.h.
14245         (gather_bbs::gather_bbs): Get and pass through bb to RPO
14246         mapping.
14247         (gather_bbs::before_dom_children): Return STOP for BBs
14248         not in the region.
14249         (build_scops): Compute bb to RPO mapping and pass it to
14250         the domwalk.  Treat --param graphite-max-nb-scop-params=0
14251         as not limiting the number of params.
14252         * graphite.c (graphite_initialize): Remove limit on the
14253         number of basic-blocks in a function.
14254         * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14255         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14256         default value of 10.
14258 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14260         * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14261         Adjust code to eliminate needing to do the shift right 32-bits
14262         operation after XSCVDPSPN.
14264 2017-09-26  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14266         * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14267         ((X / Y) != 0 -> X >= Y): Likewise.
14269 2017-09-26  Carl Love  <cel@us.ibm.com>
14271         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14272         P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14273         vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14274         void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14275         * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14276         * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14277         definitions and overloading.
14278         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14279         statement for P9V_BUILTIN_XST_LEN_R.
14280         (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14281         * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14282         define_expand and define_insn for the instructions and builtins.
14283         * doc/extend.texi: Update the built-in documentation file for the new
14284         built-in functions.
14285         * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14286         define_insn for the instructions
14288 2017-09-26  Krister Walfridsson  <krister.walfridsson@gmail.com>
14290         PR target/39570
14291         * gcc/config/netbsd-protos.h: New file.
14292         * gcc/config/netbsd.c: New file.
14293         * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14294         * gcc/config/t-netbsd: New file.
14295         * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14296         (tmake_file) Add t-netbsd.
14297         (extra_objs) Add netbsd.o.
14299 2017-09-26  Janus Weil  <janus@gcc.gnu.org>
14301         PR fortran/82143
14302         PR fortran/82324
14303         * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14305 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14307         * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14308         sign extension from a vector register to a GPR by doing a 32-bit
14309         direct move and then an EXTSW.
14310         (extendsi<mode>2 splitter): Likewise.
14311         (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14312         right or vector extract after doing XSCVDPSPN.  Use
14313         zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14314         the GPRs.
14315         (movdi_from_sf_zero_ext): Likewise.
14316         (reload_gpr_from_vsxsf): Likewise.
14317         (p8_mfvsrd_4_disf): Delete, no longer used.
14318         (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14319         then needing to move the SFmode to a GPR to use the XSCVDPSP
14320         instruction instead of FRSP and XSCVDPSPN.
14321         * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14322         it is adjacent to the other XSCVSPDP insns.
14323         (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14324         SFmode to be in traditional Altivec registers.
14325         (vsx_xscvdpspn): Eliminate useless alternative constraint.
14326         (vsx_xscvspdpn): Likewise.
14327         (vsx_xscvspdpn_scalar): Likewise.
14329 2017-09-26  Martin Jambor  <mjambor@suse.cz>
14331         * tree-sra.c (compare_access_positions): Put integral types first,
14332         stabilize sorting of integral types, remove conditions putting
14333         non-full-precision integers last.
14334         (sort_and_splice_var_accesses): Disable scalarization if a
14335         non-integert would be represented by a non-full-precision integer.
14337 2017-09-26  Joseph Myers  <joseph@codesourcery.com>
14339         * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14340         * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14341         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14342         * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14343         Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14344         conditionals inside the function instead of around it.  Call
14345         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14346         (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14348 2017-09-26  Richard Biener  <rguenther@suse.de>
14350         * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14351         fold in ...
14352         (scop_detection::build_scop_breadth): ... this.  Removed.
14353         (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14354         (scop_detection::harmful_stmt_in_bb): Likewise.
14355         (scop_detection::graphite_can_represent_stmt): Likewise.
14356         (scop_detection::loop_body_is_valid_scop): Likewise.  Remove recursion.
14357         (scop_detection::can_represent_loop): Remove recursion, fold in ...
14358         (scop_detection::can_represent_loop_1): ... this.  Removed.
14359         (scop_detection::harmful_loop_in_region): Simplify after inlining
14360         the above and remove more quadraticness.
14361         (build_scops): Adjust.
14362         * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14363         quadraticness.
14365 2017-09-26  Jakub Jelinek  <jakub@redhat.com>
14367         PR target/82267
14368         * config/i386/i386.c (ix86_print_operand_address_as): Only test
14369         REGNO (base) == SP_REG if base is a REG.
14371         PR middle-end/35691
14372         * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14373         if it is different SSA_NAME.
14374         (optimize_range_tests_cmp_bitwise): New function.
14375         (optimize_range_tests): Call it.
14377 2017-09-26  Richard Biener  <rguenther@suse.de>
14379         PR tree-optimization/82321
14380         * graphite.c (canonicalize_loop_closed_ssa): Properly check
14381         for the def being inside the loop.
14383 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14385         * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14386         assembler output.
14387         * config/s390/s390-builtins.def: Fix constraint on op4.
14389 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14391         * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14392         independent expanders.
14393         * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14394         ("vec_ordered", "vec_unordered"): New expanders.
14396 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14398         * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14399         for SFmode.
14401 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14403         * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14404         vec_unpacks_lo_v16qi.
14405         ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14407 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14409         * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14410         ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14411         ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14413 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14415         * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14416         predicate.
14417         * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14418         and V16QI.
14419         ("*vec_slb<mode>"): New insn pattern.
14420         ("vec_shr_<mode>"): New expander.
14421         * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14422         and force the shift count operand to V16QImode.
14423         ("vec_srb<mode>"): Set shift count mode to V16QI.
14425 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14427         * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14428         ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14429         ("vec_widen_smult_hi_<mode>"): New expander definitions.
14431 2017-09-26  Richard Earnshaw  <rearnsha@arm.com>
14433         PR target/82175
14434         * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14436 2017-09-26  Richard Biener  <rguenther@suse.de>
14438         PR tree-optimization/82320
14439         * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14440         isn't a change.
14442 2017-09-25  Jeff Law  <law@redhat.com>
14444         * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14445         prototype for new argument.
14446         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14447         mostly extracted from rs6000_emit_allocate_stack.
14448         (rs6000_emit_probe_stack_range_stack_clash): New function.
14449         (rs6000_emit_allocate_stack): Call
14450         rs6000_emit_probe_stack_range_stack_clash as needed.
14451         (rs6000_emit_probe_stack_range): Add additional argument
14452         to call to gen_probe_stack_range{si,di}.
14453         (output_probe_stack_range): New.
14454         (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14455         (output_probe_stack_range_stack_clash): New.
14456         (rs6000_emit_prologue): Emit notes into dump file as requested.
14457         * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14458         (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14459         Add additional operand and pass it to output_probe_stack_range.
14461 2017-09-25  Bin Cheng  <bin.cheng@arm.com>
14463         PR tree-optimization/82163
14464         * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14465         (checking_verify_loop_closed_ssa): New parameter.
14466         * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14467         (check_loop_closed_ssa_stmt): Delete.
14468         (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14469         (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14470         (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14471         changed loops.
14473 2017-09-25  Pekka Jaaskelainen <pekka@parmance.com>
14475         * brig-builtins.def: Treat HSAIL barrier builtins as
14476         setjmp/longjump style functions.
14478 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14480         * target.def (constant_alignment): New hook.
14481         * defaults.h (CONSTANT_ALIGNMENT): Delete.
14482         * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14483         (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14484         * doc/tm.texi: Regenerate.
14485         * targhooks.h (default_constant_alignment): Declare.
14486         (constant_alignment_word_strings): Likewise.
14487         * targhooks.c (default_constant_alignment): New function.
14488         (constant_alignment_word_strings): Likewise.
14489         * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14490         instead of CONSTANT_ALIGNMENT.
14491         * varasm.c (align_variable, get_variable_align, build_constant_desc)
14492         (force_const_mem): Likewise.
14493         * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14494         * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14495         (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14496         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14497         * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14498         definition.
14499         * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14500         * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14501         constant_alignment_word_strings.
14502         * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14503         (CONSTANT_ALIGNMENT): Likewise.
14504         * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14505         (arm_constant_alignment): New function.
14506         * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14507         * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14508         constant_alignment_word_strings.
14509         * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14510         * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14511         constant_alignment_word_strings.
14512         * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14513         * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14514         (cris_constant_alignment): New function.
14515         * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14516         * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14517         (epiphany_constant_alignment): New function.
14518         * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14519         * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14520         constant_alignment_word_strings.
14521         * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14522         * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14523         constant_alignment_word_strings.
14524         * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14525         * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14526         constant_alignment_word_strings.
14527         * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14528         * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14529         * config/i386/i386.c (ix86_constant_alignment): Make static.
14530         Use the same interface as the target hook.
14531         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14532         * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14533         * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14534         constant_alignment_word_strings.
14535         * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14536         * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14537         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14538         * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14539         * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14540         constant_alignment_word_strings.
14541         * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14542         * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14543         constant_alignment_word_strings.
14544         * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14545         * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14546         constant_alignment_word_strings.
14547         * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14548         * config/microblaze/microblaze.c (microblaze_constant_alignment):
14549         New function.
14550         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14551         * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14552         * config/mips/mips.c (mips_constant_alignment): New function.
14553         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14554         * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14555         * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14556         * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14557         (mmix_constant_alignment): Make static.  Use the same interface
14558         as the target hook.
14559         * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14560         * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14561         constant_alignment_word_strings.
14562         * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14563         * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14564         constant_alignment_word_strings.
14565         * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14566         * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14567         constant_alignment_word_strings.
14568         * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14569         * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14570         (rs6000_constant_alignment): New function.
14571         * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14572         * config/riscv/riscv.c (riscv_constant_alignment): New function.
14573         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14574         * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14575         * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14576         (rs6000_constant_alignment): New function.
14577         * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14578         * config/s390/s390.c (s390_constant_alignment): New function.
14579         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14580         * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14581         * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14582         constant_alignment_word_strings.
14583         * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14584         * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14585         (sparc_constant_alignment): New function.
14586         * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14587         * config/spu/spu.c (spu_constant_alignment): New function.
14588         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14589         * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14590         * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14591         constant_alignment_word_strings.
14592         * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14593         * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14594         constant_alignment_word_strings.
14595         * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14596         * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14597         constant_alignment_word_strings.
14598         * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14599         * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14600         (visium_constant_alignment): New function.
14601         * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14602         * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14603         (xtensa_constant_alignment): New function.
14604         * system.h (CONSTANT_ALIGNMENT): Poison.
14606 2017-09-25  Will Schmidt  <will_schmidt@vnet.ibm.com>
14608         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14609         for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14610         (rs6000_builtin_valid_without_lhs): New helper function.
14611         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14612         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14614 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14616         * target.h (vec_perm_indices): Use unsigned short rather than
14617         unsigned char.
14618         (auto_vec_perm_indices): Likewise.
14619         * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14620         Use unsigned int rather than unsigned char.
14621         * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14623 2017-09-25  Richard Biener  <rguenther@suse.de>
14625         * cfgloop.h (sort_sibling_loops): Declare.
14626         * cfgloop.c (sort_sibling_loops_cmp): New helper.
14627         (sort_sibling_loops): New function sorting the sibling loop list
14628         in RPO order.
14629         * graphite.c (graphite_transform_loops): Sort sibling loops.
14631 2017-09-25  Richard Sandiford  <richard.sandifird@linaro.org>
14633         * target.def (vec_perm_const_ok): Change sel parameter to
14634         vec_perm_indices.
14635         * optabs-query.c (can_vec_perm_p): Update accordingly.
14636         * doc/tm.texi: Regenerate.
14637         * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14638         auto_vec_perm_indices and remove separate nelt field.
14639         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14640         (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14641         (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14642         (aarch64_expand_vec_perm_const): Update accordingly.
14643         (aarch64_vectorize_vec_perm_const_ok): Likewise.  Change sel
14644         to vec_perm_indices.
14645         * config/arm/arm.c (expand_vec_perm_d): Change perm to
14646         auto_vec_perm_indices and remove separate nelt field.
14647         (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14648         (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14649         (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14650         accordingly.
14651         (arm_vectorize_vec_perm_const_ok): Likewise.  Change sel
14652         to vec_perm_indices.
14653         * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14654         sel to vec_perm_indices.
14655         * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14656         * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14657         * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14658         Likewise.
14659         * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14660         Likewise.
14662 2017-09-25  Pierre-Marie de Rodat  <derodat@adacore.com>
14664         PR debug/82155
14665         * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14666         on the FUNCTION_DECL function context if it has a DIE that is a
14667         declaration.
14669 2017-09-25  Richard Biener  <rguenther@suse.de>
14671         PR tree-optimization/82285
14672         * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14673         enumeral types.
14675 2017-09-25  Tom de Vries  <tom@codesourcery.com>
14677         PR target/80035
14678         PR target/81069
14679         * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14680         noreturn function.
14682 2017-09-25  Richard Biener  <rguenther@suse.de>
14684         * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14685         ISL errors other than isl_error_quota happen.  Dump if the
14686         schedule is the same.
14687         * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14688         errors instead of aborting inside ISL.
14690 2017-09-25  Iain Sandoe  <iain@codesourcery.com>
14692         PR target/80556
14693         * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14694         of libgcc_eh for m64.
14695         * config/i386/darwin64.h: Likewise.
14697 2017-09-25  Richard Biener  <rguenther@suse.de>
14699         PR middle-end/82144
14700         * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14701         attribute for incomplete types nor twice for complete ones.
14703 2017-09-24  Uros Bizjak  <ubizjak@gmail.com>
14705         PR target/82267
14706         * config/i386/i386.c (ix86_print_operand_address_as): Encode
14707         %esp as %rsp to avoid 0x67 prefix if there is no index or base
14708         register.
14710 2017-09-23  Uros Bizjak  <ubizjak@gmail.com>
14712         PR bootstrap/82306
14713         * config/i386/i386.opt (mprefer-avx256): Use
14714         ix86_target_flags variable.
14715         * config/i386/i386.c (ix86_target_string): Move
14716         -mprefer-avx256 to flag2_opts.
14718 2017-09-22  Jakub Jelinek  <jakub@redhat.com>
14720         PR middle-end/35691
14721         * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14722         and x != -1 | y != -1 into (x & y) != -1.
14724 2017-09-22  Steve Ellcey  <sellcey@cavium.com>
14726         * config.gcc: Add new case statement to set
14727         default_gnu_indirect_function.  Remove it from x86_64-*-linux*,
14728         i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14729         s390x-*-linux* case statements.   Added aarch64 to the list of
14730         supported architectures.
14732 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14734         PR tree-optimization/82289
14735         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14736         STMT_VINFO_RELEVANT_P.
14738 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14739             Alan Hayward  <alan.hayward@arm.com>
14740             David Sherwood  <david.sherwood@arm.com>
14742         * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14743         for VR_RANGE only; don't allow VR_ANTI_RANGE.
14744         (extract_range_from_binary_expr_1): Don't call
14745         extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14747 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14748             Alan Hayward  <alan.hayward@arm.com>
14749             David Sherwood  <david.sherwood@arm.com>
14751         * target.def (preferred_vector_alignment): New hook.
14752         * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14753         hook.
14754         * doc/tm.texi: Regenerate.
14755         * targhooks.h (default_preferred_vector_alignment): Declare.
14756         * targhooks.c (default_preferred_vector_alignment): New function.
14757         * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14758         Expand commentary.
14759         (DR_TARGET_ALIGNMENT): New macro.
14760         (aligned_access_p): Update commentary.
14761         (vect_known_alignment_in_bytes): New function.
14762         * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14763         function.
14764         (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14765         Calculate the misalignment based on the target alignment rather than
14766         the vector size.
14767         (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14768         rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14769         (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14770         the target alignment, rather than masking the element misalignment
14771         with the number of elements in a vector.  Also use the target
14772         alignment when calculating the maximum number of peels.
14773         (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14774         instead of TYPE_ALIGN_UNIT.
14775         (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14776         Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14777         (vect_create_addr_base_for_vector_ref): Update call accordingly.
14778         (vect_create_data_ref_ptr): Likewise.
14779         (vect_setup_realignment): Realign by ANDing with
14780         -DR_TARGET_MISALIGNMENT.
14781         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14782         the number of peels based on DR_TARGET_ALIGNMENT.
14783         * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14784         with the guaranteed alignment boundary when deciding whether
14785         overrun is OK.
14786         (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14787         relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14788         (ensure_base_align): Remove stmt_info parameter.  Get the
14789         target base alignment from DR_TARGET_ALIGNMENT.
14790         (vectorizable_store): Update call accordingly.   Interpret
14791         DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14792         TYPE_ALIGN_UNIT.
14793         (vectorizable_load): Likewise.
14795 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14796             Alan Hayward  <alan.hayward@arm.com>
14797             David Sherwood  <david.sherwood@arm.com>
14799         * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14800         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14801         (vect_enhance_data_refs_alignment): Likewise.
14803 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14805         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14806         error.  Only quit immediately if parsing is complete.
14807         (BEGIN): Initialize fatal_err and parse_done.
14808         (begin fpu, end fpu): Check number of arguments.
14809         (begin arch, end arch): Likewise.
14810         (begin cpu, end cpu): Likewise.
14811         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14812         (optalias): Likewise.
14814 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14816         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14817         * config/arm/arm-isa.h: Delete.  Move definitions to ...
14818         * arm-cpus.in: ... here.  Use new feature and fgroup values.
14819         * config/arm/arm.c (arm_option_override): Use lower case for feature
14820         bit names.
14821         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14822         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14823         * config/arm/parsecpu.awk (END): Add new command 'isa'.
14824         (isa_pfx): Delete.
14825         (print_isa_bits_for): New function.
14826         (gen_isa): New function.
14827         (gen_comm_data): Use print_isa_bits_for.
14828         (define feature): New keyword.
14829         (define fgroup): New keyword.
14830         * config/arm/t-arm (TM_H): Remove.
14831         (GTM_H): Add arm-isa.h.
14832         (arm-isa.h): Add rule to generate file.
14833         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14834         case for feature bit names.
14836 2017-09-22  Richard Biener  <rguenther@suse.de>
14838         * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14839         single caller.
14840         (graphite_regenerate_ast_isl): Do not reset SCEV.  Move debug
14841         print of no dependency loops ...
14842         * graphite.c (graphite_transform_loops): ... here.
14843         (canonicalize_loop_closed_ssa_form): Work from inner to outer
14844         loops.
14845         (same_close_phi_node, remove_duplicate_close_phi,
14846         make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14847         (canonicalize_loop_closed_ssa): ... here and simplify.
14848         * graphite-optimize-isl.c: Include tree-vectorizer.h.
14849         (optimize_isl): Use dump_printf_loc to tell when we stopped
14850         optimizing because of an ISL timeout.
14852 2017-09-22  Richard Biener  <rguenther@suse.de>
14854         PR tree-optimization/82291
14855         * tree-if-conv.c (predicate_mem_writes): Make sure to
14856         remove writes in blocks predicated with false.
14858 2017-09-22  Richard Biener  <rguenther@suse.de>
14860         * sese.c: Include cfganal.h.
14861         (if_region_set_false_region): Remove.
14862         (create_if_region_on_edge): Likewise.
14863         (move_sese_in_condition): Re-implement without destroying
14864         dominators.
14866 2017-09-22  Richard Biener  <rguenther@suse.de>
14868         * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14869         Verify both BBs contain loop PHI nodes before dispatching to
14870         copy_loop_phi_args.
14871         (graphite_regenerate_ast_isl): Do not recompute dominators,
14872         do not verify three times.  Restructure for clarity.
14873         * graphite-scop-detection.c (same_close_phi_node,
14874         remove_duplicate_close_phi, make_close_phi_nodes_unique,
14875         defined_in_loop_p, canonicalize_loop_closed_ssa,
14876         canonicalize_loop_closed_ssa_form): Simplify, remove excess
14877         checking and SSA rewrite, move to ...
14878         * graphite.c: ... here.  Include ssa.h and tree-ssa-loop-manip.h.
14879         (graphite_initialize): Do not pass in ctx, do not reset the
14880         SCEV cache, compute only dominators.
14881         (graphite_transform_loops): Allocate ISL ctx after
14882         graphite_initialize.  Call canonicalize_loop_closed_ssa_form.
14883         Maintain post-dominators only around build_scops.
14884         * sese.c (if_region_set_false_region): Make static.  Free
14885         and recompute dominators.
14886         (move_sese_in_condition): Assert we don't get called with
14887         post-dominators computed.
14888         * sese.h (if_region_set_false_region): Remove.
14890 2017-09-22  Sergey Shalnov  <sergey.shalnov@intel.com>
14892         * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14893         mode attribute for TARGET_AVX512VL.
14895 2017-09-21  Sergey Shalnov  <sergey.shalnov@intel.com>
14897         * config/i386/i386.opt (mprefer-avx256): New option.
14898         * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14899         to flag_opts.
14900         (ix86_preferred_simd_mode): Return 256-bit AVX modes
14901         for TARGET_PREFER_AVX256.
14902         * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14904 2017-09-21  Jeff Law  <law@redhat.com>
14906         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14907         Fix dump output if the only stack space is for pushed registers.
14909 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14911         * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14912         of insn_cost.
14914 2017-09-21  Martin Sebor  <msebor@redhat.com>
14916         PR c/81882
14917         * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
14918         code (in C++) or code that triggers warnings.
14920 2017-09-21  Eric Botcazou  <ebotcazou@adacore.com>
14922         * stor-layout.c (bit_from_pos): Do not distribute the conversion.
14924 2017-09-21  Segher Boessenkool  <segher@kernel.crashing.org>
14926         * haifa-sched.c: Rename insn_cost to insn_sched_cost.
14927         * sched-rgn.c: Ditto.
14928         * sel-sched-ir.c: Ditto.
14930 2017-09-21  Alexander Monakov  <amonakov@ispras.ru>
14932         * toplev.h (set_random_seed): Adjust return type.
14933         * toplev.c (init_local_tick): Move eager initialization of random_seed
14934         to get_random_seed.  Adjust comment.
14935         (init_random_seed): Inline to get_random_seed, delete.
14936         (get_random_seed): Initialize random_seed lazily.
14937         (set_random_seed): Do not return previous value.
14938         (print_switch_value): Do not call get_random_seed.
14940 2017-09-21  Evgeny Kudryashov  <kudryashov@ispras.ru>
14942         * cgraph.c (delete_function_version): New, broken out from...
14943         (cgraph_node::delete_function_version): ...here.  Rename to
14944         cgraph_node::delete_function_version_by_decl.  Update all uses.
14945         (cgraph_node::remove): Call delete_function_version.
14947 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14949         PR sanitizer/81715
14950         * tree-inline.c (expand_call_inline): Emit clobber stmts for
14951         VAR_DECLs to which addressable non-volatile parameters are mapped
14952         and for id->retvar after the return value assignment.  Clear
14953         id->retval and id->retbnd after inlining.
14955 2017-09-21  Richard Biener  <rguenther@suse.de>
14957         PR tree-optimization/82276
14958         PR tree-optimization/82244
14959         * tree-vrp.c (build_assert_expr_for): Set
14960         SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
14961         has it set.
14962         (remove_range_assertions): Revert earlier change.
14964 2017-09-21  Wilco Dijkstra  <wdijkstr@arm.com>
14966         PR target/71951
14967         * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
14969 2017-09-21  Richard Biener  <rguenther@suse.de>
14971         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
14972         Restore valid IL after code generation errors.
14973         * graphite.c (graphite_transform_loops): Diagnose code
14974         generation issues as MSG_MISSED_OPTIMIZATION and continue
14975         with processing SCOPs.
14977 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14978             Alan Hayward  <alan.hayward@arm.com>
14979             David Sherwood  <david.sherwood@arm.com>
14981         * calls.c (compute_argument_addresses): Use simplify_gen_binary
14982         rather than choosing between plus_constant and gen_rtx_<CODE>.
14983         * expr.c (emit_push_insn): Likewise.
14984         (expand_expr_real_2): Likewise.
14986 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14987             Alan Hayward  <alan.hayward@arm.com>
14988             David Sherwood  <david.sherwood@arm.com>
14990         * loop-unroll.c (split_iv): Call copy_rtx on the step.
14992 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14993             Alan Hayward  <alan.hayward@arm.com>
14994             David Sherwood  <david.sherwood@arm.com>
14996         * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
14997         calling tree_to_uhwi.
14999 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15000             Alan Hayward  <alan.hayward@arm.com>
15001             David Sherwood  <david.sherwood@arm.com>
15003         * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
15004         INTEGER_CST rather than a negative test for ADDR_EXPR.
15006 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15007             Alan Hayward  <alan.hayward@arm.com>
15008             David Sherwood  <david.sherwood@arm.com>
15010         * tree-vrp.c (extract_range_from_binary_expr_1): Check
15011         int_cst_rangeN before calling value_range_constant_singleton (&vrN).
15013 2017-09-21  Richard Biener  <rguenther@suse.de>
15015         PR tree-optimization/71351
15016         * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
15017         graphite_create_new_loop_guard): Remove, fold remaining parts
15018         into caller ...
15019         (translate_isl_ast_node_for): ... here and simplify.
15021 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
15023         PR target/82260
15024         * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
15025         with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
15026         latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
15027         alternative always use QI mode, for -Os imov (=R,R) alternative
15028         always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
15029         ignore -Os.
15031 2017-09-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
15032             Jeff Law  <law@redhat.com>
15034         * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
15035         (allocate_stack_space): New function, partially extracted from
15036         s390_emit_prologue.
15037         (s390_emit_prologue): Track offset to most recent stack probe.
15038         Code to allocate space moved into allocate_stack_space.
15039         Dump actions when no stack is allocated.
15040         (s390_prologue_plus_offset): New function.
15041         (s390_emit_stack_probe): Likewise.
15043 2017-09-20  Alexandre Oliva  <aoliva@redhat.com>
15045         * common.opt (Wa, Wl, Wp, g, gz=): Add
15046         RejectNegative.
15047         (gno-column-info): Remove.
15048         (gcolumn-info): Drop RejectNegative.
15049         (gno-): New prefix.
15050         (gno-record-gcc-switches): Remove.
15051         (grecord-gcc-switches): Drop RejectNegative.
15052         (gno-split-dwarf): Remove.
15053         (gsplit-dwarf): Drop RejectNegative.
15054         (gno-strict-dwarf): Remove.
15055         (gstrict-dwarf): Drop RejectNegative.
15056         * config/darwin.opt (gfull, gused): Add RejectNegative.
15057         * dwarf2out.c (gen_producer_string): Drop
15058         gno-record-gcc-switches handler.
15059         * optc-gen.awk: Add g to prefixes with negative forms.
15060         * opts-common.c (remapping_prefix_p): New.
15061         (find_opt): Check it.
15062         (generate_canonical_option): Test g prefix.
15063         (option_map): Add -gno- mapping.
15064         (add_misspelling_candidates): Check remapping_prefix_p.
15066 2017-09-20  Jeff Law  <law@redhat.com>
15068         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
15069         thinko in stack clash protection support.
15071         * explow.c (compute_stack_clash_protection_loop_data): Use
15072         CONST_INT_P instead of explicit test.  Verify object is a
15073         CONST_INT_P before looking at INTVAL.
15074         (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
15075         instead of explicit test.
15077 2017-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
15079         PR target/77687
15080         * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
15081         address instead of to r1 and r11.
15083 2017-09-20  Sebastian Peryt  <sebastian.peryt@intel.com>
15085         * config.gcc: Support "knm".
15086         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
15087         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15088         PROCESSOR_KNM.
15089         * config/i386/i386.c (m_KNM): Define.
15090         (processor_target_table): Add "knm".
15091         (PTA_KNM): Define.
15092         (ix86_option_override_internal): Add "knm".
15093         (ix86_issue_rate): Add PROCESSOR_KNM.
15094         (ix86_adjust_cost): Ditto.
15095         (ia32_multipass_dfa_lookahead): Ditto.
15096         (get_builtin_code_for_version): Handle PROCESSOR_KNM.
15097         (fold_builtin_cpu): Add M_INTEL_KNM.
15098         * config/i386/i386.h (processor_costs): Define TARGET_KNM.
15099         (processor_type): Add PROCESSOR_KNM.
15100         * config/i386/x86-tune.def: Add m_KNM.
15101         * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
15103 2017-09-20  Richard Biener  <rguenther@suse.de>
15105         PR tree-optimization/80213
15106         * graphite-scop-detection.c (trivially_empty_bb_p): Labels
15107         are allowed in empty BBs as well.
15108         (canonicalize_loop_closed_ssa): Also look for other complex
15109         edges.
15110         (scop_detection::get_sese): Include the loop-closed PHI block
15111         in loop SESEs.
15112         (scop_detection::merge_sese): Remove code adding extra blocks.
15113         (scop_detection::region_has_one_loop): Adjust for get_sese changes.
15114         (build_scops): Assert the final returned scop is invalid.
15116 2017-09-20  Richard Biener  <rguenther@suse.de>
15118         PR tree-optimization/82264
15119         * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
15120         for GIMPLE_CONDs.
15121         (vn_phi_lookup): Likewise.
15122         (vn_phi_insert): Likewise.
15124 2017-09-20  Jakub Jelinek  <jakub@redhat.com>
15126         * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
15127         that fits into uhwi or shwi, add DW_AT_const_value regardless
15128         of early_dwarf without going through RTL, using add_AT_unsigned
15129         or add_AT_int.
15131         * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
15132         (DEBUG_LTO_ABBREV_SECTION): Likewise.
15133         (DEBUG_LTO_MACINFO_SECTION): Likewise.
15134         (DEBUG_MACRO_SECTION): Likewise.
15135         (DEBUG_LTO_MACRO_SECTION): Likewise.
15136         (DEBUG_STR_DWO_SECTION): Likewise.
15137         (DEBUG_LTO_STR_DWO_SECTION): Likewise.
15138         (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
15139         (DEBUG_LTO_DWO_LINE_SECTION): Define.
15140         (DEBUG_LTO_LINE_STR_SECTION): Define.
15141         (init_sections_and_labels): Initialize debug_line_str_section
15142         variable.  Initialize debug_loc_section for -gdwarf-5 to
15143         DEBUG_LOCLISTS_SECTION.  Formatting fixes.
15145 2017-09-20  Richard Biener  <rguenther@suse.de>
15147         * graphite-sese-to-poly.c (extract_affine): Properly handle
15148         POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
15150 2017-09-20  Richard Biener  <rguenther@suse.de>
15152         PR tree-optimization/81373
15153         * graphite-scop-detection.c (build_cross_bb_scalars_def):
15154         Force SESE live-out defs to be handled even if they are
15155         scev_analyzable_p.
15157 2017-09-19  Jeff Law  <law@redhat.com>
15159         * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
15160         nothing for stack adjustments with REG_STACK_CHECK.
15161         * sched-deps.c (parse_add_or_inc): Reject insns with
15162         REG_STACK_CHECK from dependency breaking.
15163         * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
15164         (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
15165         * reg-notes.def (STACK_CHECK): New note.
15167         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
15168         (ix86_expand_prologue): Dump stack clash info as needed.
15169         Call ix86_adjust_stack_and_probe_stack_clash as needed.
15171         * function.c (dump_stack_clash_frame_info): New function.
15172         * function.h (dump_stack_clash_frame_info): Prototype.
15173         (enum stack_clash_probes): New enum.
15175         * config/alpha/alpha.c (alpha_expand_prologue): Also check
15176         flag_stack_clash_protection.
15177         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
15178         (arm_expand_prologue, thumb1_expand_prologue): Likewise.
15179         (arm_frame_pointer_required): Likewise.
15180         * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
15181         (ia64_expand_prologue): Likewise.
15182         * config/mips/mips.c (mips_expand_prologue): Likewise.
15183         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
15184         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15185         (sparc_flat_expand_prologue): Likewise.
15186         * config/spu/spu.c (spu_expand_prologue): Likewise.
15188         * explow.c: Include "params.h".
15189         (anti_adjust_stack_and_probe_stack_clash): New function.
15190         (get_stack_check_protect): Likewise.
15191         (compute_stack_clash_protection_loop_data): Likewise.
15192         (emit_stack_clash_protection_loop_start): Likewise.
15193         (emit_stack_clash_protection_loop_end): Likewise.
15194         (allocate_dynamic_stack_space): Use get_stack_check_protect.
15195         Use anti_adjust_stack_and_probe_stack_clash.
15196         * explow.h (compute_stack_clash_protection_loop_data): Prototype.
15197         (emit_stack_clash_protection_loop_start): Likewise.
15198         (emit_stack_clash_protection_loop_end): Likewise.
15199         * rtl.h (get_stack_check_protect): Prototype.
15200         * target.def (stack_clash_protection_final_dynamic_probe): New hook.
15201         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
15202         * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
15203         Prototype.
15204         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
15205         Add @hook.
15206         * doc/tm.texi: Rebuilt.
15207         * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
15208         get_stack_check_protect.
15209         * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
15210         * config/arm/arm.c (arm_expand_prologue): Likewise.
15211         (arm_frame_pointer_required): Likewise.
15212         * config/i386/i386.c (ix86_expand_prologue): Likewise.
15213         * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
15214         * config/mips/mips.c (mips_expand_prologue): Likewise.
15215         * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
15216         * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
15217         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15218         (sparc_flat_expand_prologue): Likewise.
15220         * common.opt (-fstack-clash-protection): New option.
15221         * flag-types.h (enum stack_check_type): Note difference between
15222         -fstack-check= and -fstack-clash-protection.
15223         * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
15224         (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
15225         * toplev.c (process_options): Issue warnings/errors for cases
15226         not handled with -fstack-clash-protection.
15227         * doc/invoke.texi (-fstack-clash-protection): Document new option.
15228         (-fstack-check): Note additional problem with -fstack-check=generic.
15229         Note that -fstack-check is primarily for Ada and refer users
15230         to -fstack-clash-protection for stack-clash-protection.
15231         Document new params for stack clash protection.
15233 2017-09-19  Uros Bizjak  <ubizjak@gmail.com>
15235         * config/i386/i386.c (ix86_split_long_move): Do not handle
15236         address used for LEA in a special way.
15238 2017-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
15240         * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
15242 2017-09-19  Martin Sebor  <msebor@redhat.com>
15244         PR c/81854
15245         * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15246         of incompatible types.
15248 2017-09-19  Will Schmidt  <will_schmidt@vnet.ibm.com>
15250         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15251         for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15252         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15253         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15255 2017-09-19  Richard Biener  <rguenther@suse.de>
15257         PR tree-optimization/82244
15258         * tree-vrp.c (remove_range_assertions): Do not propagate
15259         a constant to abnormals but replace the assert with a copy.
15261 2017-09-19  Alexander Monakov  <amonakov@ispras.ru>
15263         PR rtl-optimization/57878
15264         PR rtl-optimization/68988
15265         * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15266         avoidance test involving non_reload_pseudos.  Move frequency test
15267         below the general fragmentation avoidance test.
15269 2017-09-19  Richard Biener  <rguenther@suse.de>
15271         PR tree-optimization/69728
15272         * graphite-sese-to-poly.c (schedule_error): New global.
15273         (add_loop_schedule): Handle empty domain by failing the
15274         schedule.
15275         (build_original_schedule): Handle schedule_error.
15277 2017-09-19  Richard Biener  <rguenther@suse.de>
15279         * graphite-scop-detection.c (scop_detection::can_represent_loop):
15280         Do not iterate to sibling loops but only to siblings of inner
15281         loops.
15283 2017-09-18  Andreas Schwab  <schwab@linux-m68k.org>
15285         PR target/81613
15286         * config/m68k/m68k.md (moveq feeding equality comparison): Check
15287         that the registers are different.
15289 2017-09-18  Uros Bizjak  <ubizjak@gmail.com>
15291         * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15292         to processor_model and "amdfam17h" to arch_names_table.
15293         * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15295 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15297         PR c/82234
15298         * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15300 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15301             Alan Hayward  <alan.hayward@arm.com>
15302             David Sherwood  <david.sherwood@arm.com>
15304         * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15305         with a vec_info *.
15306         * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15307         accordingly.
15308         * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15309         parameter.  Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15310         vect_schedule_slp_instance.
15311         (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15312         Update call to vect_slp_analyze_node_operations.  Simplify return
15313         value.
15314         (vect_slp_analyze_bb_1): Update call accordingly.
15315         (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15316         Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15317         (vect_schedule_slp): Update call accordingly.
15319 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15320             Alan Hayward  <alan.hayward@arm.com>
15321             David Sherwood  <david.sherwood@arm.com>
15323         * tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
15324         with types that aren't in fact scalar.
15326 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15328         * tree-vect-slp.c (vect_record_max_nunits): New function,
15329         split out from...
15330         (vect_build_slp_tree_1): ...here.
15331         (vect_build_slp_tree_2): Call it for phis too.
15333 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15335         * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15336         to vect_get_vec_def_for_operand when getting the mask operand.
15338 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15339             Alan Hayward  <alan.hayward@arm.com>
15340             David Sherwood  <david.sherwood@arm.com>
15342         * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15343         bitstart.
15345 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15346             Alan Hayward  <alan.hayward@arm.com>
15347             David Sherwood  <david.sherwood@arm.com>
15349         * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15350         calculation for vector booleans.
15352 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15353             Alan Hayward  <alan.hayward@arm.com>
15354             David Sherwood  <david.sherwood@arm.com>
15356         * tree-vect-stmts.c (can_vectorize_live_stmts): New function,
15357         split out from...
15358         (vect_transform_stmt): ...here.
15359         (vect_analyze_stmt): Use it instead of calling
15360         vectorizable_live_operation directly.
15362 2017-09-18  Cesar Philippidis  <cesar@codesourcery.com>
15364         * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15365         non-SIMT targets in acc vector loops.
15367 2017-09-18  Claudiu Zissulescu  <claziss@synopsys.com>
15369         * configure.ac: Add arc and check if assembler supports gdwarf2.
15370         * configure: Regenerate.
15372 2017-09-18  Richard Biener  <rguenther@suse.de>
15374         PR tree-optimization/82220
15375         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15376         epilogue niters from the min_profitable_iters compute.
15378 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15380         PR target/82145
15381         * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15382         changes.  Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15383         (ix86_init_pic_reg): Revert 2017-09-01 changes.
15385 2017-09-18  Eric Botcazou  <ebotcazou@adacore.com>
15387         PR target/81361
15388         * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15389         switching to a new text section.
15391 2017-09-18  Richard Biener  <rguenther@suse.de>
15393         * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15394         Simplify.
15395         (build_alias_set): Reject aliases with no access function.
15397 2017-09-18  Richard Biener  <rguenther@suse.de>
15399         PR tree-optimization/79622
15400         * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15401         handle PHIs.
15402         (build_cross_bb_scalars_use): Likewise.
15404 2017-09-18  Pierre-Marie de Rodat  <derodat@adacore.com>
15406         * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15408 2017-09-18  Alan Modra  <amodra@gmail.com>
15410         PR target/81996
15411         * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15412         stack_pointer_rtx for count 0.  Update comments.  Break up
15413         large rtl expression.
15415 2017-09-17  Daniel Santos  <daniel.santos@pobox.com>
15417         * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15418         Increase to 20 bytes.
15419         (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15420         (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15421         or avx version of the stub.
15423 2017-09-17  H.J. Lu  <hongjiu.lu@intel.com>
15425         PR target/82166
15426         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15427         compute the minimum stack alignment.  Also update preferred stack
15428         boundary for leaf functions.
15430 2017-09-16  Richard Sandiford  <richard.sandiford@linaro.org>
15432         PR tree-optimization/82228
15433         * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15434         of ncopies.
15436 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15438         * common/config/nds32/nds32-common.c
15439         (nds32_option_optimization_table): Refine formatting.
15440         (nds32_option_optimization_table): Use -fsched-pressure and
15441         -fomit-frame-pointer for specific optimization level.
15443 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15445         * config/nds32/nds32.c: Refine formatting and comments.
15446         * config/nds32/nds32.h: Likewise.
15447         * config/nds32/nds32.md: Likewise.
15448         * config/nds32/nds32-cost.c: Likewise.
15449         * config/nds32/nds32-isr.c: Likewise.
15450         * config/nds32/nds32-md-auxiliary.c: Likewise.
15451         * config/nds32/nds32-multiple.md: Likewise.
15452         * config/nds32/nds32-predicates.c: Likewise.
15454 2017-09-15  Andrew Sutton  <andrew.n.sutton@gmail.com>
15455             Jakub Jelinek  <jakub@redhat.com>
15457         Add support for -std=c++2a.
15458         * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15459         or -std=gnu+2a.
15460         * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15462 2017-09-15  Steve Ellcey  <sellcey@cavium.com>
15464         PR target/82066
15465         * doc/extend.texi (Common Function Attributes): Add 
15466         references to ARM, AArch64, and S/390 specific attributes.
15467         (Function Specific Option Pragmas): Add AArch64 and S/390
15468         to list of back ends that support the target pragma.
15470 2017-09-15  Nathan Sidwell  <nathan@acm.org>
15472         * doc/standards.texi: Fix C++17 description.  Update URLs for
15473         C++11 & 14.
15475 2017-09-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15477         * common.opt (Wcast-align=strict): New warning option.
15478         * doc/invoke.texi: Document -Wcast-align=strict. 
15480 2017-09-15  Pierre-Marie de Rodat  <derodat@adacore.com>
15482         * cgraph.h (cgraph_thunk_info): Add comments.
15483         * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15484         assert for VIRTUAL_* arguments stricter.
15486 2017-09-15  Jackson Woodruff  <jackson.woodruff@arm.com>
15488         PR tree-optimization/71026
15489         * match.pd: Move RDIV patterns from fold-const.c
15490         * fold-const.c (distribute_real_division): Removed.
15491         (fold_binary_loc): Remove calls to distribute_real_divison.
15493 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15495         * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15496         c++1z and gnu++1z as deprecated.  Change other references to
15497         -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15498         Change -Wc++1z-compat to -Wc++17-compat.
15499         * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15500         * dwarf2out.c (highest_c_language): Handle C++17.
15501         (gen_compile_unit_die): Likewise.
15503 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15505         PR rtl-optimization/82192
15506         * combine.c (make_extraction): Don't look through non-paradoxical
15507         SUBREGs or TRUNCATE if pos + len is or might be bigger than
15508         inner's mode.
15510 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15511             Alan Hayward  <alan.hayward@arm.com>
15512             David Sherwood  <david.sherwood@arm.com>
15514         * target.def (function_arg_offset): New hook.
15515         * targhooks.h (default_function_arg_offset): Declare.
15516         * targhooks.c (default_function_arg_offset): New function.
15517         * function.c (locate_and_pad_parm): Use
15518         targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15519         * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15520         (TARGET_FUNCTION_ARG_OFFSET): ...this.
15521         * doc/tm.texi: Regenerate.
15522         * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15523         * config/spu/spu.c (spu_function_arg_offset): New function.
15524         (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15525         * system.h (FUNCTION_ARG_OFFSET): Poison.
15527 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15528             Alan Hayard  <alan.hayward@arm.com>
15529             David Sherwood  <david.sherwood@arm.com>
15531         * target.def (truly_noop_truncation): New hook.
15532         (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15533         than TRULY_NOOP_TRUNCATION.
15534         * hooks.h (hook_bool_uint_uint_true): Declare.
15535         * hooks.c (hook_bool_uint_uint_true): New function.
15536         * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15537         (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15538         * doc/tm.texi: Regenerate.
15539         * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15540         rather than TRULY_NOOP_TRUNCATION in comments.
15541         (simplify_comparison): Likewise.
15542         (record_truncated_value): Likewise.
15543         * expmed.c (extract_bit_field_1): Likewise.
15544         (extract_split_bit_field): Likewise.
15545         * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15546         instead of TRULY_NOOP_TRUNCATION.
15547         * function.c (assign_parm_setup_block): Likewise.
15548         * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15549         * rtlhooks.c: Include target.h.
15550         * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15551         * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15552         * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15553         * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15554         * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15555         * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15556         * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15557         * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15558         * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15559         * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15560         * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15561         * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15562         * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15563         * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15564         * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15565         * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15566         * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15567         * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15568         * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15569         * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15570         * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15571         * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15572         * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15573         * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15574         * config/mips/mips.c (mips_truly_noop_truncation): New function.
15575         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15576         * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15577         rather than TRULY_NOOP_TRUNCATION in comments.
15578         * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15579         * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15580         * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15581         * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15582         * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15583         * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15584         * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15585         * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15586         * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15587         * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15588         * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15589         * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15590         rather than TRULY_NOOP_TRUNCATION in comments.
15591         * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15592         * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15593         * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15594         * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15595         * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15596         TRULY_NOOP_TRUNCATION condition.
15597         (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15598         (TRULY_NOOP_TRUNCATION): Delete.
15599         * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15600         * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15601         * config/spu/spu.c (spu_truly_noop_truncation): New function.
15602         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15603         * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15604         * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15605         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15606         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15607         * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15608         rather than TRULY_NOOP_TRUNCATION in comments.
15609         * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15610         * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15611         * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15612         * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15613         * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15614         * system.h (TRULY_NOOP_TRUNCATION): Poison.
15616 2017-09-15  Christophe Lyon  <christophe.lyon@linaro.org>
15618         PR target/67591
15619         * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15620         (*cmp_ior): Likewise.
15621         (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15622         (*ior_scc_scc_cmp): Likewise.
15623         (*and_scc_scc): Likewise.
15624         (*and_scc_scc_cmp): Likewise.
15626 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15627             Alan Hayard  <alan.hayward@arm.com>
15628             David Sherwood  <david.sherwood@arm.com>
15630         * target.def (can_change_mode_class): New hook.
15631         (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15632         (hard_regno_nregs): Likewise.
15633         * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15634         * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15635         * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15636         (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15637         (LOAD_EXTEND_OP): Update accordingly.
15638         * doc/tm.texi: Regenerate.
15639         * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15640         CANNOT_CHANGE_MODE_CLASS.
15641         * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15642         (REG_CAN_CHANGE_MODE_P): ...this new macro.
15643         * combine.c (simplify_set): Update accordingly.
15644         * emit-rtl.c (validate_subreg): Likewise.
15645         * recog.c (general_operand): Likewise.
15646         * regcprop.c (mode_change_ok): Likewise.
15647         * reload1.c (choose_reload_regs): Likewise.
15648         (inherit_piecemeal_p): Likewise.
15649         * rtlanal.c (simplify_subreg_regno): Likewise.
15650         * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15651         instead of CANNOT_CHANGE_MODE_CLASS.
15652         (reload_cse_simplify_operands): Likewise.
15653         * reload.c (push_reload): Use targetm.can_change_mode_class
15654         instead of CANNOT_CHANGE_MODE_CLASS.
15655         (push_reload): Likewise.  Also use REG_CAN_CHANGE_MODE_P instead of
15656         REG_CANNOT_CHANGE_MODE_P.
15657         * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15658         * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15659         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15660         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15661         * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15662         (arm_can_change_mode_class): New function.
15663         * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15664         than CANNOT_CHANGE_MODE_CLASS in comments.
15665         * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15666         * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15667         * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15668         (ix86_can_change_mode_class): ...this new function, inverting the
15669         sense of the return value.
15670         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15671         * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15672         * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15673         (ia64_can_change_mode_class): New function.
15674         * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15675         * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15676         * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15677         (m32c_can_change_mode_class): ...this new function, inverting the
15678         sense of the return value.
15679         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15680         * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15681         * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15682         * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15683         (mips_can_change_mode_class): ...this new function, inverting the
15684         sense of the return value.
15685         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15686         * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15687         * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15688         (msp430_can_change_mode_class): New function.
15689         * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15690         * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15691         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15692         * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15693         * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15694         * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15695         * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15696         (pa_cannot_change_mode_class): Replace with...
15697         (pa_can_change_mode_class): ...this new function, inverting the
15698         sense of the return value.
15699         (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15700         than CANNOT_CHANGE_MODE_CLASS in comments.
15701         * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15702         * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15703         * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15704         (pdp11_cannot_change_mode_class): Replace with...
15705         (pdp11_can_change_mode_class): ...this new function, inverting the
15706         sense of the return value.
15707         * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15708         * config/powerpcspe/powerpcspe-protos.h
15709         (rs6000_cannot_change_mode_class_ptr): Delete.
15710         * config/powerpcspe/powerpcspe.c
15711         (rs6000_cannot_change_mode_class_ptr): Delete.
15712         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15713         (rs6000_option_override_internal): Assign to
15714         targetm.can_change_mode_class instead of
15715         rs6000_cannot_change_mode_class_ptr.
15716         (rs6000_cannot_change_mode_class): Replace with...
15717         (rs6000_can_change_mode_class): ...this new function, inverting the
15718         sense of the return value.
15719         (rs6000_debug_cannot_change_mode_class): Replace with...
15720         (rs6000_debug_can_change_mode_class): ...this new function.
15721         * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15722         * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15723         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15724         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15725         * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15726         Delete.
15727         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15728         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15729         (rs6000_option_override_internal): Assign to
15730         targetm.can_change_mode_class instead of
15731         rs6000_cannot_change_mode_class_ptr.
15732         (rs6000_cannot_change_mode_class): Replace with...
15733         (rs6000_can_change_mode_class): ...this new function, inverting the
15734         sense of the return value.
15735         (rs6000_debug_cannot_change_mode_class): Replace with...
15736         (rs6000_debug_can_change_mode_class): ...this new function.
15737         * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15738         * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15739         * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15740         (s390_can_change_mode_class): ...this new function, inverting the
15741         sense of the return value.
15742         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15743         * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15744         * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15745         * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15746         (sh_cannot_change_mode_class): Replace with...
15747         (sh_can_change_mode_class): ...this new function, inverting the
15748         sense of the return value.
15749         * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15750         * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15751         (sparc_can_change_mode_class): New function.
15752         * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15753         * config/spu/spu.c (spu_can_change_mode_class): New function.
15754         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15755         * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15756         * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15757         (visium_can_change_mode_class): New function.
15758         * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15760 2017-09-15  Richard Biener  <rguenther@suse.de>
15762         PR tree-optimization/82217
15763         * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15764         but not undefined case.
15766 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15768         PR target/82145
15769         * postreload.c (reload_cse_simplify_operands): Skip
15770         NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15772 2017-09-15  Richard Biener  <rguenther@suse.de>
15774         PR tree-optimization/68823
15775         * graphite-scop-detection.c (build_alias_set): If we have a
15776         possible dependence check whether we can handle them by just
15777         looking at the DRs DR_ACCESS_FNs.
15778         (build_scops): If build_alias_set fails, fail the SCOP.
15780 2017-09-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
15782         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15783         to support float128 built-in functions that require the ISA 3.0
15784         hardware.
15785         (BU_FLOAT128_3_HW): Likewise.
15786         (SQRTF128): Add support for the IEEE 128-bit square root and fma
15787         built-in functions.
15788         (FMAF128): Likewise.
15789         (FMAQ): Likewise.
15790         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15791         support for built-in functions that need the ISA 3.0 IEEE 128-bit
15792         floating point instructions.
15793         (rs6000_invalid_builtin): Likewise.
15794         (rs6000_builtin_mask_names): Likewise.
15795         * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15796         (RS6000_BTM_FLOAT128_HW): Likewise.
15797         (RS6000_BTM_COMMON): Likewise.
15798         * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15799         function.
15800         * doc/extend.texi (RS/6000 built-in functions): Document the
15801         IEEE 128-bit floating point square root and fused multiply-add
15802         built-in functions.
15804 2017-09-14  Pat Haugen  <pthaugen@us.ibm.com>
15806         * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15807         reg (r2) isn't in the set of registers defined in the prologue.
15809 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15810             Alan Hayward  <alan.hayward@arm.com>
15811             David Sherwood  <david.sherwood@arm.com>
15813         * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15814         (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15815         (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15816         (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15817         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15818         accordingly.
15819         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15820         max_vectorization_factor.
15821         (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
15823 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15824             Alan Hayward  <alan.hayward@arm.com>
15825             David Sherwood  <david.sherwood@arm.com>
15827         * tree-vectorizer.h (vect_min_worthwhile_factor): Delete.
15828         (vect_worthwhile_without_simd_p): Declare.
15829         * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15830         (vectorizable_reduction): Use it.
15831         * tree-vect-stmts.c (vectorizable_shift): Likewise.
15832         (vectorizable_operation): Likewise.
15834 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15835             Alan Hayward  <alan.hayward@arm.com>
15836             David Sherwood  <david.sherwood@arm.com>
15838         * tree-vectorizer.h (vect_get_num_copies): New function.
15839         * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15840         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15841         (vectorizable_induction): Likewise.
15842         (vectorizable_live_operation): Likewise.
15843         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15844         (vectorizable_bswap): Likewise.
15845         (vectorizable_call): Likewise.
15846         (vectorizable_conversion): Likewise.
15847         (vectorizable_assignment): Likewise.
15848         (vectorizable_shift): Likewise.
15849         (vectorizable_operation): Likewise.
15850         (vectorizable_store): Likewise.
15851         (vectorizable_load): Likewise.
15852         (vectorizable_condition): Likewise.
15853         (vectorizable_comparison): Likewise.
15854         (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15856 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15857             Alan Hayward  <alan.hayward@arm.com>
15858             David Sherwood  <david.sherwood@arm.com>
15860         * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15861         of vect_init_vector.
15863 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15864             Alan Hayward  <alan.hayward@arm.com>
15865             David Sherwood  <david.sherwood@arm.com>
15867         * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15868         an inline wrapper that provides a location.
15869         (gimple_build_vector): Likewise.
15870         * gimple-fold.c (gimple_build_vector_from_val): New function.
15871         (gimple_build_vector): Likewise.
15872         * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15873         functions to build the initial value.  Always return a gimple value.
15874         (get_initial_defs_for_reduction): Likewise.  Only compute
15875         neutral_vec once.
15876         (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15877         vect_init_vector on the results from get_initial_def(s)_for_reduction.
15878         (vectorizable_induction): Use gimple_build_vector rather than
15879         vect_init_vector.
15881 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15882             Alan Hayward  <alan.hayward@arm.com>
15883             David Sherwood  <david.sherwood@arm.com>
15885         * target.h (vec_perm_indices): New typedef.
15886         (auto_vec_perm_indices): Likewise.
15887         * optabs-query.h: Include target.h
15888         (can_vec_perm_p): Take a vec_perm_indices *.
15889         * optabs-query.c (can_vec_perm_p): Likewise.
15890         (can_mult_highpart_p): Update accordingly.  Use auto_vec_perm_indices.
15891         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15892         * tree-vect-generic.c (lower_vec_perm): Likewise.
15893         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15894         (vect_grouped_load_supported): Likewise.
15895         (vect_shift_permute_load_chain): Likewise.
15896         (vect_permute_store_chain): Use auto_vec_perm_indices.
15897         (vect_permute_load_chain): Likewise.
15898         * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15899         (fold_ternary_loc): Update accordingly.  Use auto_vec_perm_indices.
15900         Update uses of can_vec_perm_p.
15901         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15902         mode with a number of elements.  Take a vec_perm_indices *.
15903         (vect_create_epilog_for_reduction): Update accordingly.
15904         Use auto_vec_perm_indices.
15905         (have_whole_vector_shift): Likewise.  Update call to can_vec_perm_p.
15906         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15907         (vect_transform_slp_perm_load): Likewise.
15908         (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15909         * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15910         (vect_gen_perm_mask_checked): Likewise.
15911         * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15912         (vect_gen_perm_mask_checked): Likewise.
15913         (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15914         (vectorizable_store): Likewise.
15915         (vectorizable_load): Likewise.
15916         (perm_mask_for_reverse): Likewise.  Update call to can_vec_perm_p.
15917         (vectorizable_bswap): Likewise.
15919 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15920             Alan Hayward  <alan.hayward@arm.com>
15921             David Sherwood  <david.sherwood@arm.com>
15923         * tree.h (build_vector): Take a vec<tree> instead of a tree *.
15924         * tree.c (build_vector): Likewise.
15925         (build_vector_from_ctor): Update accordingly.
15926         (build_vector_from_val): Likewise.
15927         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
15928         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15929         * tree-vect-generic.c (add_rshift): Likewise.
15930         (expand_vector_divmod): Likewise.
15931         (optimize_vector_constructor): Likewise.
15932         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
15933         (vect_transform_slp_perm_load): Likewise.
15934         (vect_schedule_slp_instance): Likewise.
15935         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15936         (vectorizable_call): Likewise.
15937         (vect_gen_perm_mask_any): Likewise.  Add elements in order.
15938         * expmed.c (make_tree): Likewise.
15939         * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
15940         a vector passed to build_vector.
15941         (fold_convert_const): Likewise.
15942         (exact_inverse): Likewise.
15943         (fold_ternary_loc): Likewise.
15944         (fold_relational_const): Likewise.
15945         (const_binop): Likewise.  Use VECTOR_CST_ELT directly when operating
15946         on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
15947         (const_unop): Likewise.  Store the reduction accumulator in a
15948         variable rather than an array.
15949         (vec_cst_ctor_to_array): Take the number of elements as a parameter.
15950         (fold_vec_perm): Update calls accordingly.  Use auto_vec<tree> for
15951         the new vector, rather than constructing it after the input arrays.
15952         (native_interpret_vector): Use auto_vec<tree> when building
15953         a vector passed to build_vector.  Add elements in order.
15954         * tree-vect-loop.c (get_initial_defs_for_reduction): Use
15955         auto_vec<tree> when building a vector passed to build_vector.
15956         (vect_create_epilog_for_reduction): Likewise.
15957         (vectorizable_induction): Likewise.
15958         (get_initial_def_for_reduction): Likewise.  Fix indentation of
15959         case statements.
15960         * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
15961         to a vec<tree> *.
15962         (sparc_fold_builtin): Use auto_vec<tree> when building a vector
15963         passed to build_vector.
15965 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15966             Alan Hayward  <alan.hayward@arm.com>
15967             David Sherwood  <david.sherwood@arm.com>
15969         * tree-core.h (tree_base::u): Add an "nelts" field.
15970         (tree_vector): Use VECTOR_CST_NELTS as the length.
15971         * tree.c (tree_size): Likewise.
15972         (make_vector): Initialize VECTOR_CST_NELTS.
15973         * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
15974         * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
15975         TYPE_VECTOR_SUBPARTS.
15976         * expr.c (const_vector_mask_from_tree): Consistently use "units"
15977         as the number of units, setting it from VECTOR_CST_NELTS.
15978         (const_vector_from_tree): Likewise.
15979         * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
15980         TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
15981         (fold_negate_expr_1): Likewise.
15982         (fold_convert_const): Likewise.
15983         (const_binop): Likewise.  Differentiate the number of output and
15984         input elements.
15985         (const_unop): Likewise.
15986         (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
15987         in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
15988         in cases that did the opposite.
15990 2017-09-14  Richard Biener  <rguenther@suse.de>
15992         * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
15993         to VN_TOP.
15995 2017-09-14  Eric Botcazou  <ebotcazou@adacore.com>
15997         * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
15999 2017-09-14  Jakub Jelinek  <jakub@redhat.com>
16001         PR target/81325
16002         * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
16003         if and where to split a bb, except for splitting before debug insn
16004         sequences followed by non-label real insn.  Delete debug insns
16005         in between basic blocks.
16007         * combine.c (make_compound_operation_int): Formatting fixes.
16009         * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
16010         * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16011         * config/netbsd.h (LINK_EH_SPEC): Likewise.
16012         * config/sol2.h (LINK_EH_SPEC): Likewise.
16013         * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16014         * config/s390/linux.h (LINK_SPEC): Likewise.
16015         * config/freebsd.h (LINK_EH_SPEC): Likewise.
16016         * config/openbsd.h (LINK_EH_SPEC): Likewise.
16017         * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16018         * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
16019         * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
16020         * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16021         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
16022         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
16024 2017-09-13  Jakub Jelinek  <jakub@redhat.com>
16026         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
16027         support.
16028         (ENDFILE_LINUX_SPEC): Likewise.
16029         (LINK_EH_SPEC): Likewise.
16030         * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
16031         (LINK_OS_LINUX_SPEC32): Likewise.
16032         (LINK_OS_LINUX_SPEC64): Likewise.
16033         * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
16034         (LINK_OS_LINUX_SPEC): Likewise.
16036 2017-09-13  Martin Liska  <mliska@suse.cz>
16038         PR middle-end/82154
16039         * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
16040         CASE_HIGH is NULL_TREE.
16042 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16043             Alan Hayward  <alan.hayward@arm.com>
16044             David Sherwood  <david.sherwood@arm.com>
16046         * target.def (secondary_memory_needed): New hook.
16047         (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
16048         instead of SECONDARY_MEMORY_NEEDED.
16049         (secondary_memory_needed_mode): Likewise.
16050         * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
16051         * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
16052         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
16053         (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
16054         (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
16055         * doc/tm.texi: Regenerate.
16056         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
16057         * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
16058         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16059         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
16060         * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
16061         * config/i386/i386.c (inline_secondary_memory_needed): Put the
16062         mode argument first and change the reg_class arguments to reg_class_t.
16063         (ix86_secondary_memory_needed): Likewise.  Remove the strict parameter.
16064         Make static.  Update the call to inline_secondary_memory_needed.
16065         (ix86_register_move_cost): Update the call to
16066         inline_secondary_memory_needed.
16067         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16068         * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
16069         definition.
16070         * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
16071         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16072         in comment.
16073         * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
16074         * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
16075         * config/mips/mips.c (mips_secondary_memory_needed): Make static
16076         and match hook interface.  Add comment from mips.h.
16077         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16078         * config/mmix/mmix.md (truncdfsf2): Refer to
16079         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16080         in comment.
16081         * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
16082         (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
16083         * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16084         (pa_secondary_memory_needed): New function.
16085         * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
16086         * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
16087         * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16088         (pdp11_secondary_memory_needed): Make static and match hook interface.
16089         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
16090         * config/powerpcspe/powerpcspe-protos.h
16091         (rs6000_secondary_memory_needed_ptr): Delete.
16092         * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
16093         Delete.
16094         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16095         (rs6000_option_override_internal): Assign to
16096         targetm.secondary_memory_needed rather than
16097         rs6000_secondary_memory_needed_ptr.
16098         (rs6000_secondary_memory_needed): Match hook interface.
16099         (rs6000_debug_secondary_memory_needed): Likewise.
16100         * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
16101         * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
16102         (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
16103         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16104         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
16105         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
16106         Delete.
16107         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): Delete.
16108         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16109         (rs6000_option_override_internal): Assign to
16110         targetm.secondary_memory_needed rather than
16111         rs6000_secondary_memory_needed_ptr.
16112         (rs6000_secondary_memory_needed): Match hook interface.
16113         (rs6000_debug_secondary_memory_needed): Likewise.
16114         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
16115         * config/s390/s390.c (s390_secondary_memory_needed): New function.
16116         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16117         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
16118         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16119         (sparc_secondary_memory_needed): New function.
16120         * lra-constraints.c (check_and_process_move): Refer to
16121         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16122         in comment.
16123         (curr_insn_transform): Likewise.
16124         (process_alt_operands): Use targetm.secondary_memory_needed
16125         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16126         (check_secondary_memory_needed_p): Likewise.
16127         (choose_split_class): Likewise.
16128         * reload.c: Unconditionally include code that was previously
16129         conditional on SECONDARY_MEMORY_NEEDED.
16130         (push_secondary_reload): Use targetm.secondary_memory_needed
16131         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16132         (push_reload): Likewise.
16133         * reload1.c: Unconditionally include code that was previously
16134         conditional on SECONDARY_MEMORY_NEEDED.
16135         (choose_reload_regs): Use targetm.secondary_memory_needed
16136         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16137         (gen_reload): Likewise.
16138         * system.h (SECONDARY_MEMORY_NEEDED): Poison.
16140 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16141             Alan Hayward  <alan.hayward@arm.com>
16142             David Sherwood  <david.sherwood@arm.com>
16144         * target.def (secondary_memory_needed_mode): New hook:
16145         * targhooks.c (default_secondary_memory_needed_mode): Declare.
16146         * targhooks.h (default_secondary_memory_needed_mode): New function.
16147         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
16148         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
16149         * doc/tm.texi: Regenerate.
16150         * lra-constraints.c (check_and_process_move): Use
16151         targetm.secondary_memory_needed_mode instead of
16152         TARGET_SECONDARY_MEMORY_NEEDED_MODE.
16153         (curr_insn_transform): Likewise.
16154         * reload.c (get_secondary_mem): Likewise.
16155         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16156         * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
16157         function.
16158         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16159         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16160         * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
16161         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16162         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
16163         Delete.
16164         * config/powerpcspe/powerpcspe-protos.h
16165         (rs6000_secondary_memory_needed_mode): Delete.
16166         * config/powerpcspe/powerpcspe.c
16167         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16168         (rs6000_secondary_memory_needed_mode): Make static.
16169         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16170         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
16171         Delete.
16172         * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16173         Redefine.
16174         (rs6000_secondary_memory_needed_mode): Make static.
16175         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16176         * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
16177         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16178         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16179         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16180         Redefine.
16181         (sparc_secondary_memory_needed_mode): New function.
16182         * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
16184 2017-09-13  Jackson Woodruff  <jackson.woodruff@arm.com>
16186         * config/aarch64/constraints.md (Umq): New constraint.
16187         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
16188         Change to use Umq.
16189         (mov<mode>): Update condition.
16191 2017-09-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16193         * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
16194         when bitposition is the same.
16196 2017-09-13  Richard Biener  <rguenther@suse.de>
16198         * dwarf2out.c (output_die_symbol): Remove.
16199         (output_die): Do not output a DIEs symbol.
16201 2017-09-13  Richard Biener  <rguenther@suse.de>
16203         PR middle-end/82128
16204         * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
16205         default-def to avoid breaking iterator update with the weird
16206         interaction with cgraph_update_edges_for_call_stmt_node.
16208 2017-09-13  Richard Biener  <rguenther@suse.de>
16210         * tree-cfg.c (verify_gimple_assign_binary): Add verification
16211         for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
16212         VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
16213         (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
16215 2017-09-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
16217         * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
16218         Disable pc relative literal load irrespective of
16219         TARGET_FIX_ERR_A53_84341 for default.
16221 2017-09-12  Eric Botcazou  <ebotcazou@adacore.com>
16223         * config/sparc/sparc.c (output_return): Output the source location of
16224         the insn in the delay slot, if any.
16225         (output_sibcall): Likewise.
16227 2017-09-12  H.J. Lu  <hongjiu.lu@intel.com>
16229         PR driver/81498
16230         * common.opt (-static-pie): New alias.
16231         (shared): Negate static-pie.
16232         (-no-pie): Update help text.
16233         (-pie): Likewise.
16234         (static-pie): New option.
16235         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
16236         -static-pie support.
16237         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
16238         (LINK_EH_SPEC): Likewise.
16239         (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16240         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16241         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16242         * gcc.c (LINK_COMMAND_SPEC): Likewise.
16243         (init_gcc_specs): Likewise.
16244         (init_spec): Likewise.
16245         (display_help): Update help message for -pie.
16246         * doc/invoke.texi: Update -pie, -no-pie and -static.  Document
16247         -static-pie.
16249 2017-09-12  Wilco Dijkstra  <wdijkstr@arm.com>
16251         * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16252         (movdi_aarch64): Likewise.
16253         (movti_aarch64): Likewise.
16255 2017-09-12 Simon Wright <simon@pushface.org>
16257         PR target/80204
16258         * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16259         calculation of the minor version, always output as 0.
16261 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16263         PR target/82112
16264         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16265         ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16266         on it early, rather than manual conversion late.  For
16267         ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16268         instead of performing manual conversion.
16270 2017-09-12  Carl Love  <cel@us.ibm.com>
16272         * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16273         vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16274         vmulouw, vmulosw.
16275         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16276         VMULOSW): Add definitions.
16277         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16278         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16279         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16280         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16281         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16283 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16285         * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16286         types correctly.
16287         (movti_aarch64): Likewise.
16288         (movdf_aarch64): Likewise.
16289         (movtf_aarch64): Likewise.
16290         (load_pairdi): Likewise.
16291         (store_pairdi): Likewise.
16292         (load_pairdf): Likewise.
16293         (store_pairdf): Likewise.
16294         (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16295         (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16296         (ldr_got_small_<mode>): Likewise.
16297         (ldr_got_small_28k_<mode>): Likewise.
16298         (ldr_got_tiny): Likewise.
16299         * config/aarch64/iterators.md (ldst_sz): New.
16300         (ldpstp_sz): Likewise.
16301         * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16302         to store_16.
16303         (thunderx_load): Split load_8 to load_16.
16304         * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16305         load_8 to load_16.
16306         (thunderx2t99_storepair_basic): Split store_8 to store_16.
16307         * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16308         (xgene1_store_pair): Split store_8 to store_16.
16309         * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16310         (falkor_st_0_st_sd): Split store_8 to store_16.
16312 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16314         * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16315         and store1/2/3/4 to store_4/8/12/16.
16316         * config/aarch64/aarch64.md: Update for rename.
16317         * config/arm/arm.md: Likewise.: Likewise.
16318         * config/arm/arm.c: Likewise.
16319         * config/arm/thumb1.md: Likewise.
16320         * config/arm/thumb2.md: Likewise.
16321         * config/arm/vfp.md: Likewise.
16322         * config/arm/arm-generic.md: Likewise.
16323         * config/arm/arm1020e.md: Likewise.
16324         * config/arm/arm1026ejs.md: Likewise.
16325         * config/arm/arm1136jfs.md: Likewise.
16326         * config/arm/arm926ejs.md: Likewise.
16327         * config/arm/cortex-a15.md: Likewise.
16328         * config/arm/cortex-a17.md: Likewise.
16329         * config/arm/cortex-a5.md: Likewise.
16330         * config/arm/cortex-a53.md: Likewise.
16331         * config/arm/cortex-a57.md: Likewise.
16332         * config/arm/cortex-a7.md: Likewise.
16333         * config/arm/cortex-a8.md: Likewise.
16334         * config/arm/cortex-a9.md: Likewise.
16335         * config/arm/cortex-m4.md: Likewise.
16336         * config/arm/cortex-m7.md: Likewise.
16337         * config/arm/cortex-r4.md: Likewise.
16338         * config/arm/exynos-m1.md: Likewise.
16339         * config/arm/fa526.md: Likewise.
16340         * config/arm/fa606te.md: Likewise.
16341         * config/arm/fa626te.md: Likewise.
16342         * config/arm/fa726te.md: Likewise.
16343         * config/arm/fmp626.md: Likewise.
16344         * config/arm/iwmmxt.md: Likewise.
16345         * config/arm/ldmstm.md: Likewise.
16346         * config/arm/marvell-pj4.md: Likewise.
16347         * config/arm/xgene1.md: Likewise.
16348         * config/aarch64/thunderx.md: Likewise.
16349         * config/aarch64/thunderx2t99.md: Likewise.
16350         * config/aarch64/falkor.md: Likewise.
16352 2017-09-12  Martin Liska  <mliska@suse.cz>
16354         * attribs.c (private_lookup_attribute): New function.
16355         * attribs.h (private_lookup_attribute): Declared here.
16356         (lookup_attribute): Called from this place.
16358 2017-09-12  Richard Biener  <rguenther@suse.de>
16360         PR tree-optimization/82157
16361         * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16362         stmts with side-effects.
16364 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16365             Alan Hayward  <alan.hayward@arm.com>
16366             David Sherwood <david.sherwood@arm.com>
16368         * target.def (hard_regno_nregs): New hook.
16369         (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16370         * targhooks.h (default_hard_regno_nregs): Declare.
16371         * targhooks.c (default_hard_regno_nregs): New function.
16372         * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16373         (TARGET_HARD_REGNO_NREGS): ...this hook.
16374         (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16375         (CLASS_MAX_NREGS): Likewise.
16376         * doc/tm.texi: Regenerate.
16377         * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16378         instead of HARD_REGNO_NREGS.
16379         * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16380         HARD_REGNO_NREGS in the comment.
16381         * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16382         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16383         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16384         Return an unsigned int.
16385         (TARGET_HARD_REGNO_NREGS): Redefine.
16386         * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16387         * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16388         * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16389         (arc_hard_regno_nregs): New function.
16390         * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16391         * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16392         (arm_hard_regno_nregs): New function.
16393         * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16394         * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16395         * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16396         (TARGET_HARD_REGNO_NREGS): Redefine.
16397         * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16398         * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16399         (HARD_REGNO_NREGS): Delete.
16400         * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16401         (cr16_hard_regno_nregs): New function.
16402         (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16403         * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16404         * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16405         (cris_hard_regno_nregs): New function.
16406         * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16407         * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16408         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16409         * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16410         (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16411         * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16412         * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16413         (frv_hard_regno_nregs): Make static.  Take and return an
16414         unsigned int.
16415         (frv_class_max_nregs): Remove outdated copy of documentation.
16416         * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16417         * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16418         * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16419         * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16420         * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16421         * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16422         (TARGET_HARD_REGNO_NREGS): Redefine.
16423         * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16424         (CLASS_MAX_NREGS): Update comment.
16425         * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16426         (ia64_hard_regno_nregs): New function.
16427         * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16428         * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16429         * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16430         * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16431         * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16432         an unsigned int.
16433         (m32c_hard_regno_nregs): Likewise.  Make static.
16434         (TARGET_HARD_REGNO_NREGS): Redefine.
16435         * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16436         * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16437         * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16438         (m68k_hard_regno_nregs): New function.
16439         * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16440         * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16441         * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16442         * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16443         * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16444         Take and return an unsigned int.
16445         (TARGET_HARD_REGNO_NREGS): Redefine.
16446         * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16447         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16448         * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16449         * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16450         * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16451         * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16452         * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16453         (msp430_hard_regno_nregs): Make static.  Take and return an
16454         unsigned int.
16455         * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16456         * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16457         * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16458         (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16459         * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16460         * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16461         * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16462         (TARGET_HARD_REGNO_NREGS): Redefine.
16463         * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16464         (PA_HARD_REGNO_NREGS): ...this.
16465         * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16466         (PA_HARD_REGNO_NREGS): ...this.
16467         * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16468         (pa_hard_regno_nregs): New function.
16469         * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16470         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16471         (pdp11_hard_regno_nregs): New function.
16472         * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16473         * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16474         (rs6000_hard_regno_nregs_hook): New function.
16475         * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16476         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16477         * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16478         Take and return an unsigned int.  Move earlier in file.
16479         (TARGET_HARD_REGNO_NREGS): Redefine.
16480         * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16481         * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16482         * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16483         (rl78_hard_regno_nregs): Make static.  Take and return an
16484         unsigned int.
16485         * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16486         * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16487         (rs6000_hard_regno_nregs_hook): New function.
16488         * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16489         * config/rx/rx.c (rx_hard_regno_nregs): New function.
16490         (TARGET_HARD_REGNO_NREGS): Redefine.
16491         * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16492         * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16493         instead of HARD_REGNO_NREGS.
16494         (s390_hard_regno_nregs): New function.
16495         (s390_hard_regno_mode_ok): Add comment from s390.h.
16496         (TARGET_HARD_REGNO_NREGS): Redefine.
16497         * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16498         * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16499         (sh_hard_regno_nregs): New function.
16500         (sh_pass_in_reg_p): Use it.
16501         * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16502         * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16503         (sparc_hard_regno_nregs): New function.
16504         * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16505         * config/spu/spu.c (spu_hard_regno_nregs): New function.
16506         (spu_function_arg_advance): Use it, supplying a valid register number.
16507         (TARGET_HARD_REGNO_NREGS): Redefine.
16508         * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16509         * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16510         * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16511         * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16512         * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16513         * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16514         (CLASS_MAX_NREGS): Remove copy of old documentation.
16515         * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16516         (visium_hard_regno_nregs): New function.
16517         (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16518         * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16519         * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16520         xtensa_hard_regno_nregs): New function.
16521         * system.h (HARD_REGNO_NREGS): Poison.
16523 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16525         * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16526         hard_regno_nregs instead of HARD_REGNO_NREGS.
16527         (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16528         * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16529         (c6x_expand_epilogue): Likewise.
16530         * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16531         (frv_read_iacc_argument): Likewise.
16532         * config/sh/sh.c: Include regs.h.
16533         (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16534         (regs_used): Likewise.
16535         (output_stack_adjust): Likewise.
16536         * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16537         * expmed.c: Include regs.h.
16538         (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16539         * ree.c: Include regs.h.
16540         (combine_reaching_defs): Use hard_regno_nregs instead of
16541         HARD_REGNO_NREGS.
16542         (add_removable_extension): Likewise.
16544 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16546         * regs.h (hard_regno_nregs): Turn into a function.
16547         (end_hard_regno): Update accordingly.
16548         * caller-save.c (setup_save_areas): Likewise.
16549         (save_call_clobbered_regs): Likewise.
16550         (replace_reg_with_saved_mem): Likewise.
16551         (insert_restore): Likewise.
16552         (insert_save): Likewise.
16553         * combine.c (can_change_dest_mode): Likewise.
16554         (move_deaths): Likewise.
16555         (distribute_notes): Likewise.
16556         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16557         * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16558         (rs6000_split_multireg_move): Likewise.
16559         (rs6000_register_move_cost): Likewise.
16560         (rs6000_memory_move_cost): Likewise.
16561         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16562         (rs6000_split_multireg_move): Likewise.
16563         (rs6000_register_move_cost): Likewise.
16564         (rs6000_memory_move_cost): Likewise.
16565         * cselib.c (cselib_reset_table): Likewise.
16566         (cselib_lookup_1): Likewise.
16567         * emit-rtl.c (set_mode_and_regno): Likewise.
16568         * function.c (aggregate_value_p): Likewise.
16569         * ira-color.c (setup_profitable_hard_regs): Likewise.
16570         (check_hard_reg_p): Likewise.
16571         (calculate_saved_nregs): Likewise.
16572         (assign_hard_reg): Likewise.
16573         (improve_allocation): Likewise.
16574         (calculate_spill_cost): Likewise.
16575         * ira-emit.c (modify_move_list): Likewise.
16576         * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16577         (ira_hard_reg_in_set_p): Likewise.
16578         * ira.c (setup_reg_mode_hard_regset): Likewise.
16579         (clarify_prohibited_class_mode_regs): Likewise.
16580         (check_allocation): Likewise.
16581         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16582         (lra_setup_reg_renumber): Likewise.
16583         (setup_try_hard_regno_pseudos): Likewise.
16584         (spill_for): Likewise.
16585         (assign_hard_regno): Likewise.
16586         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16587         * lra-constraints.c (in_class_p): Likewise.
16588         (lra_constraint_offset): Likewise.
16589         (simplify_operand_subreg): Likewise.
16590         (lra_constraints): Likewise.
16591         (split_reg): Likewise.
16592         (split_if_necessary): Likewise.
16593         (invariant_p): Likewise.
16594         (inherit_in_ebb): Likewise.
16595         * lra-lives.c (process_bb_lives): Likewise.
16596         * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16597         (get_hard_regs): Likewise.
16598         (do_remat): Likewise.
16599         * lra-spills.c (assign_spill_hard_regs): Likewise.
16600         * mode-switching.c (create_pre_exit): Likewise.
16601         * postreload.c (reload_combine_recognize_pattern): Likewise.
16602         * recog.c (peep2_find_free_register): Likewise.
16603         * regcprop.c (kill_value_regno): Likewise.
16604         (set_value_regno): Likewise.
16605         (copy_value): Likewise.
16606         (maybe_mode_change): Likewise.
16607         (find_oldest_value_reg): Likewise.
16608         (copyprop_hardreg_forward_1): Likewise.
16609         * regrename.c (check_new_reg_p): Likewise.
16610         (regrename_do_replace): Likewise.
16611         * reload.c (push_reload): Likewise.
16612         (combine_reloads): Likewise.
16613         (find_dummy_reload): Likewise.
16614         (operands_match_p): Likewise.
16615         (find_reloads): Likewise.
16616         (find_equiv_reg): Likewise.
16617         (reload_adjust_reg_for_mode): Likewise.
16618         * reload1.c (count_pseudo): Likewise.
16619         (count_spilled_pseudo): Likewise.
16620         (find_reg): Likewise.
16621         (clear_reload_reg_in_use): Likewise.
16622         (free_for_value_p): Likewise.
16623         (allocate_reload_reg): Likewise.
16624         (choose_reload_regs): Likewise.
16625         (reload_adjust_reg_for_temp): Likewise.
16626         (emit_reload_insns): Likewise.
16627         (delete_output_reload): Likewise.
16628         * rtlanal.c (subreg_get_info): Likewise.
16629         * sched-deps.c (sched_analyze_reg): Likewise.
16630         * sel-sched.c (init_regs_for_mode): Likewise.
16631         (mark_unavailable_hard_regs): Likewise.
16632         (choose_best_reg_1): Likewise.
16633         (verify_target_availability): Likewise.
16634         * valtrack.c (dead_debug_insert_temp): Likewise.
16635         * var-tracking.c (track_loc_p): Likewise.
16636         (emit_note_insn_var_location): Likewise.
16637         * varasm.c (make_decl_rtl): Likewise.
16638         * reginfo.c (choose_hard_reg_mode): Likewise.
16639         (init_reg_modes_target): Refer directly to
16640         this_target_regs->x_hard_regno_nregs.
16642 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16644         * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16645         instead of hard_regno_nregs.
16647 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16649         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16650         end_hard_regno instead of hard_regno_nregs.
16651         * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16652         * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16653         * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16654         * ira-color.c (improve_allocation): Likewise.
16655         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16656         * lra-lives.c (mark_regno_live): Likewise.
16657         (mark_regno_dead): Likewise.
16658         * lra-remat.c (operand_to_remat): Likewise.
16659         * lra.c (collect_non_operand_hard_regs): Likewise.
16660         * postreload.c (reload_combine_note_store): Likewise.
16661         (move2add_valid_value_p): Likewise.
16662         * reload.c (regno_clobbered_p): Likewise.
16664 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16666         * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16667         hard_regno_nregs.
16668         * config/v850/v850.c (v850_reorg): Likewise.
16669         * reload.c (refers_to_regno_for_reload_p): Likewise.
16670         (find_equiv_reg): Likewise.
16671         * reload1.c (reload_reg_reaches_end_p): Likewise.
16673 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16675         * caller-save.c (add_used_regs): Use REG_NREGS instead of
16676         hard_regno_nregs.
16677         * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16678         * config/arm/arm.c (output_move_neon): Likewise.
16679         (arm_attr_length_move_neon): Likewise.
16680         (neon_split_vcombine): Likewise.
16681         * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16682         (c6x_mark_reg_written): Likewise.
16683         (c6x_dwarf_register_span): Likewise.
16684         * config/i386/i386.c (ix86_save_reg): Likewise.
16685         * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16686         (rws_access_reg): Likewise.
16687         * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16688         * mode-switching.c (create_pre_exit): Likewise.
16689         * ree.c (combine_reaching_defs): Likewise.
16690         (add_removable_extension): Likewise.
16691         * regcprop.c (find_oldest_value_reg): Likewise.
16692         (copyprop_hardreg_forward_1): Likewise.
16693         * reload.c (reload_inner_reg_of_subreg): Likewise.
16694         (push_reload): Likewise.
16695         (combine_reloads): Likewise.
16696         (find_dummy_reload): Likewise.
16697         (reload_adjust_reg_for_mode): Likewise.
16698         * reload1.c (find_reload_regs): Likewise.
16699         (forget_old_reloads_1): Likewise.
16700         (reload_reg_free_for_value_p): Likewise.
16701         (reload_adjust_reg_for_temp): Likewise.
16702         (emit_reload_insns): Likewise.
16703         (delete_output_reload): Likewise.
16704         * sel-sched.c (choose_best_reg_1): Likewise.
16705         (choose_best_pseudo_reg): Likewise.
16707 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16708             Alan Hayward  <alan.hayward@arm.com>
16709             David Sherwood <david.sherwood@arm.com>
16711         * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16712         * target.def (slow_unaligned_access): New hook.
16713         * targhooks.h (default_slow_unaligned_access): Declare.
16714         * targhooks.c (default_slow_unaligned_access): New function.
16715         * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16716         (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16717         * doc/tm.texi: Regenerate.
16718         * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16719         * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16720         * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16721         definition.
16722         * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16723         * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16724         Redefine.
16725         (rs6000_slow_unaligned_access): New function.
16726         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16727         (expand_block_compare): Likewise.
16728         (expand_strn_compare): Likewise.
16729         (rs6000_rtx_costs): Likewise.
16730         * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16731         (riscv_slow_unaligned_access): Likewise.
16732         * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16733         (riscv_slow_unaligned_access_p): ...this and make static.
16734         (riscv_option_override): Update accordingly.
16735         (riscv_slow_unaligned_access): New function.
16736         (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16737         * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16738         * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16739         (rs6000_slow_unaligned_access): New function.
16740         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16741         (rs6000_rtx_costs): Likewise.
16742         * config/rs6000/rs6000-string.c (expand_block_compare)
16743         (expand_strn_compare): Use targetm.slow_unaligned_access instead
16744         of SLOW_UNALIGNED_ACCESS.
16745         * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16746         * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16747         * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16748         of SLOW_UNALIGNED_ACCESS.
16749         * expmed.c (simple_mem_bitfield_p): Likewise.
16750         * expr.c (alignment_for_piecewise_move): Likewise.
16751         (emit_group_load_1): Likewise.
16752         (emit_group_store): Likewise.
16753         (copy_blkmode_from_reg): Likewise.
16754         (emit_push_insn): Likewise.
16755         (expand_assignment): Likewise.
16756         (store_field): Likewise.
16757         (expand_expr_real_1): Likewise.
16758         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16759         * lra-constraints.c (simplify_operand_subreg): Likewise.
16760         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16761         * gimple-ssa-store-merging.c: Likewise in block comment at start
16762         of file.
16763         * tree-ssa-strlen.c: Include target.h.
16764         (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16765         of SLOW_UNALIGNED_ACCESS.
16766         * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16768 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16770         PR rtl-optimization/82185
16771         * expmed.c (emit_store_flag_int): Only test tem if it has been
16772         initialized.
16774 2017-09-12  Richard Biener  <rguenther@suse.de>
16776         PR middle-end/82149
16777         * match.pd ((FTYPE) N CMP CST): Fix typo.
16779 2017-09-12  Simon Atanasyan  <simon.atanasyan@imgtec.com>
16781         * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16782         attribute.
16783         (mips_near_type_p): Add 'short_call' attribute as a synonym
16784         for 'near'.
16785         * doc/extend.texi (short_call): Document new function attribute.
16787 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16789         PR target/82112
16790         * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16791         assertion check that in the condition.
16792         (get_atomic_generic_size): Likewise.  Before testing if parameter
16793         has pointer type, if it has array type, call for C++
16794         default_conversion to perform array-to-pointer conversion.
16796 2017-09-12  Richard Biener  <rguenther@suse.de>
16798         * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16799         for operations we cannot scalarize.
16801 2017-09-12  Aldy Hernandez  <aldyh@redhat.com>
16803         * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16804         vectors heap vectors.  Clean up comments.
16805         Make visited_bbs a reference.
16806         (profitable_jump_thread_path): Make GC
16807         vectors heap vectors.  Clean up comments.
16808         Misc cleanups.
16809         (convert_and_register_jump_thread_path): Make GC vectors heap
16810         vectors.
16811         (check_subpath_and_update_thread_path): Same.  Clean up comments.
16812         Make visited_bbs a reference.
16813         (handle_phi): Abstract common code to to
16814         register_jump_thread_path_if_profitable.
16815         Rename VAR_BB to DEF_BB.
16816         Update comments.
16817         Make GC vectors heap vectors.
16818         Make visited_bbs a reference.
16819         (handle_assignment): Same.
16820         (register_jump_thread_path_if_profitable): New.
16821         (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16822         DEF_BB.
16823         Make GC vectors heap vectors.  Clean up comments.
16824         Make visited_bbs a reference.
16825         (find_jump_threads_backwards): Make visited_bbs live in the stack.
16826         * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16827         comment.
16829 2017-09-11  Max Filippov  <jcmvbkbc@gmail.com>
16831         PR target/82181
16832         * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16833         words of E_DImode object are reachable by xtensa_uimm8x4 access.
16835 2017-09-11  Vidya Praveen  <vidyapraveen@arm.com>
16837         Revert r251800 and r251799.
16839 2017-09-11  Martin Jambor  <mjambor@suse.cz>
16841         PR hsa/82119
16842         * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16843         arguments in advance.
16844         * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16845         use it to find predecessor edges.
16846         (naive_outof_ssa): Collect vector of predecessors.
16848 2017-09-08  Jason Merrill  <jason@redhat.com>
16850         PR c++/70029 - ICE with ref-qualifier and -flto
16851         * langhooks.h (struct lang_hooks_for_types): Add
16852         copy_lang_qualifiers.
16853         * attribs.c (build_type_attribute_qual_variant): Use it.
16854         * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16855         NULL.
16856         (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16857         * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16859 2017-09-08  Eric Botcazou  <ebotcazou@adacore.com>
16861         PR target/81988
16862         * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16863         (*mulsi3_sp64): New instruction.
16864         (mulsi3): New expander.
16866 2017-09-08  Uros Bizjak  <ubizjak@gmail.com>
16868         * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16870 2017-09-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16872         * sancov.c: Include memmodel.h.
16874 2017-09-07  Eric Botcazou  <ebotcazou@adacore.com>
16876         PR target/80897
16877         * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16878         large offsets.
16880 2017-09-07  Carl Love  <cel@us.ibm.com>
16882         * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16883         the sldi instruction.
16885 2017-09-07  David Edelsohn  <dje.gcc@gmail.com>
16887         * sancov.c: Include tm_p.h.
16889 2017-09-07  Jakub Jelinek  <jakub@redhat.com>
16891         PR target/81979
16892         * output.h (switch_to_other_text_partition): New declaration.
16893         * varasm.c (switch_to_other_text_partition): New function.
16894         * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16895         NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16896         (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16897         to the other text partition before emitting LCL label and switch back
16898         after emitting the word after it.
16900 2017-09-07  Richard Biener  <rguenther@suse.de>
16902         * passes.def (pass_split_crit_edges): Remove instance before PRE.
16903         * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16904         critical edges here, after loop init.
16905         (pass_data_pre): Remove PROP_no_crit_edges flags.
16906         * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16907         for valueization of call args to avoid leaking VN_TOP.
16908         (visit_use): Assert we do not visit default defs.
16909         (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16910         Use error_mark_node to more easily detect leaking VN_TOP.
16911         All default-defs are varying, not VN_TOP.  Mark them visited.
16912         (run_scc_vn): Make code match comment.
16914 2017-09-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
16916         * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16917         OPTION_MASK_FLOAT128_KEYWORD.
16918         (POWERPC_MASKS): Likewise.
16919         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
16920         support for the -mfloat128-type option, and make -mfloat128
16921         default on PowerPC Linux systems.  Define or undefine
16922         __FLOAT128__ and  __FLOAT128_HARDWARE__ for the current options.
16923         Define __float128 to be __ieee128 if IEEE 128-bit support is
16924         enabled, or undefine it.
16925         (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
16926         Delete defining __FLOAT128_TYPE__.
16927         * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
16928         -mfloat128-type option and make -mfloat128 default on PowerPC
16929         Linux systems.
16930         (TARGET_FLOAT128_TYPE): Likewise.
16931         (-mfloat128-type): Likewise.
16932         * config/rs6000/rs6000.c (rs6000_option_override_internal):
16933         Delete the -mfloat128-type option and make -mfloat128 default on
16934         PowerPC Linux systems.  Always use __ieee128 to be the keyword for
16935         the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
16936         128-bit floating point is enabled.  Change tests from using
16937         -mfloat128-type to -mfloat128.
16938         (rs6000_mangle_type): Use the correct mangling for the __float128
16939         type even if normal long double is restricted to 64-bits.
16940         (floatn_mode): Enable the _Float128 type by default on VSX Linux
16941         systems.
16942         * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
16943         (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
16944         (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
16945         -mfloat128-type.
16946         * doc/invoke.texi (RS/6000 and PowerPC Options): Update
16947         documentation for -mfloat128.
16949 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16951         * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
16953 2017-09-06  Wish Wu  <wishwu007@gmail.com>
16954             Jakub Jelinek  <jakub@redhat.com>
16956         * asan.c (initialize_sanitizer_builtins): Add
16957         BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
16958         BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
16959         BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
16960         BT_FN_VOID_UINT64_PTR variables.
16961         * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
16962         (BT_FN_VOID_UINT16_UINT16): Likewise.
16963         (BT_FN_VOID_UINT32_UINT32): Likewise.
16964         (BT_FN_VOID_FLOAT_FLOAT): Likewise.
16965         (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
16966         (BT_FN_VOID_UINT64_PTR): Likewise.
16967         * common.opt (flag_sanitize_coverage): New variable.
16968         (fsanitize-coverage=trace-pc): Remove.
16969         (fsanitize-coverage=): Add.
16970         * flag-types.h (enum sanitize_coverage_code): New enum.
16971         * fold-const.c (fold_range_test): Disable non-short-circuit
16972         optimization if flag_sanitize_coverage.
16973         (fold_truth_andor): Likewise.
16974         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16975         * opts.c (COVERAGE_SANITIZER_OPT): Define.
16976         (coverage_sanitizer_opts): New array.
16977         (get_closest_sanitizer_option): Add OPTS argument, handle also
16978         OPT_fsanitize_coverage_.
16979         (parse_sanitizer_options): Adjusted to also handle
16980         OPT_fsanitize_coverage_.
16981         (common_handle_option): Add OPT_fsanitize_coverage_.
16982         * sancov.c (instrument_comparison, instrument_switch): New function.
16983         (sancov_pass): Add trace-cmp support.
16984         * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
16985         BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
16986         BUILT_IN_SANITIZER_COV_TRACE_CMP8,
16987         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
16988         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
16989         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
16990         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
16991         BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
16992         BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
16993         * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
16995 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16997         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
16998         error.  Only quit immediately if parsing is complete.
16999         (BEGIN): Initialize fatal_err and parse_done.
17000         (begin fpu, end fpu): Check number of arguments.
17001         (begin arch, end arch): Likewise.
17002         (begin cpu, end cpu): Likewise.
17003         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
17004         (optalias): Likewise.
17006 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
17008         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
17009         * config/arm/arm-isa.h: Delete.  Move definitions to ...
17010         * arm-cpus.in: ... here.  Use new feature and fgroup values.
17011         * config/arm/arm.c (arm_option_override): Use lower case for feature
17012         bit names.
17013         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
17014         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
17015         * config/arm/parsecpu.awk (END): Add new command 'isa'.
17016         (isa_pfx): Delete.
17017         (print_isa_bits_for): New function.
17018         (gen_isa): New function.
17019         (gen_comm_data): Use print_isa_bits_for.
17020         (define feature): New keyword.
17021         (define fgroup): New keyword.
17022         * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
17023         (arm-isa.h): Add rule to generate file.
17024         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
17025         case for feature bit names.
17027 2017-09-06  Richard Biener  <rguenther@suse.de>
17029         * tree-ssa-pre.c (NECESSARY): Remove.
17030         (create_expression_by_pieces): Do not touch pass-local flags.
17031         (insert_into_preds_of_block): Likewise.
17032         (do_pre_regular_insertion): Likewise.
17033         (eliminate_insert): Likewise.
17034         (eliminate_dom_walker::before_dom_children): Likewise.
17035         (fini_eliminate): Do not look at inserted_exprs.
17036         (mark_operand_necessary): Remove.
17037         (remove_dead_inserted_code): Replace with simple work-list
17038         algorithm based on inserted_exprs and SSA uses.
17039         (pass_pre::execute): Re-order fini_eliminate and
17040         remove_dead_inserted_code.
17042 2017-09-06  Olivier Hainque  <hainque@adacore.com>
17044         * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
17045         for VxWorks 7.  Adjust surrounding comments.
17047 2017-09-06  Richard Biener  <rguenther@suse.de>
17049         * gimple-ssa-strength-reduction.c
17050         (find_candidates_dom_walker::before_dom_children): Also allow
17051         pointer types.
17053 2017-09-06  Richard Biener  <rguenther@suse.de>
17055         PR tree-optimization/82108
17056         * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
17057         for gap in the non-permutation SLP case.
17059 2017-09-06  Martin Jambor  <mjambor@suse.cz>
17061         PR tree-optimization/82078
17062         * tree-sra.c (sort_and_splice_var_accesses): Move call to
17063         add_access_to_work_queue...
17064         (build_accesses_from_assign): ...here.
17065         (propagate_all_subaccesses): Make sure racc is the group
17066         representative, if there is one.
17068 2017-09-06  Jakub Jelinek  <jakub@redhat.com>
17070         PR middle-end/82095
17071         * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
17072         NULL DECL_INITIAL.
17074 2017-09-06  Richard Biener  <rguenther@suse.de>
17076         * gimple-ssa-strength-reduction.c
17077         (find_candidates_dom_walker::before_doom_children): Use a
17078         type and not a mode check.
17080 2017-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17082         PR target/77308
17083         * config/arm/predicates.md (arm_general_adddi_operand): Create new
17084         non-vfp predicate.
17085         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
17087 2017-09-05  Jeff Law  <law@redhat.com>
17089         PR tree-optimization/64910
17090         * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
17091         cases where we have 3 or more operands.
17093 2017-09-05  Jakub Jelinek  <jakub@redhat.com>
17095         PR middle-end/81768
17096         * omp-low.c (lower_omp_for): Recompute tree invariant if
17097         gimple_omp_for_initial/final is ADDR_EXPR.
17099         PR middle-end/81768
17100         * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
17101         into gimple val before gimplification fo the COND_EXPR.
17103 2017-09-05  Aldy Hernandez  <aldyh@redhat.com>
17105         * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
17106         REGION_COPY argument.
17107         (thread_through_all_blocks): Remove unused argument to
17108         duplicate_thread_path.
17110 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17111             Alan Hayward  <alan.hayward@arm.com>
17112             David Sherwood  <david.sherwood@arm.com>
17114         * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
17115         Take a scalar_mode rather than a machine_mode.
17116         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17117         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
17118         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17119         (aarch64_gen_adjusted_ldpstp): Likewise.
17120         (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
17122 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17123             Alan Hayward  <alan.hayward@arm.com>
17124             David Sherwood  <david.sherwood@arm.com>
17126         * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
17127         Take a scalar_int_mode instead of a machine_mode.
17128         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17129         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17130         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17131         (aarch64_simd_attr_length_rglist): Delete.
17132         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
17133         a scalar_int_mode instead of a machine_mode.
17134         (aarch64_add_offset): Likewise.
17135         (aarch64_internal_mov_immediate): Likewise
17136         (aarch64_add_constant_internal): Likewise.
17137         (aarch64_add_constant): Likewise.
17138         (aarch64_movw_imm): Likewise.
17139         (aarch64_rtx_arith_op_extract_p): Likewise.
17140         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17141         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17142         Remove assert that the mode isn't a vector.
17143         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17144         (aarch64_expand_mov_immediate): Update calls after above changes.
17145         (aarch64_output_casesi): Use as_a <scalar_int_mode>.
17146         (aarch64_and_bitmask_imm): Check for scalar integer modes.
17147         (aarch64_move_imm): Likewise.
17148         (aarch64_can_const_movi_rtx_p): Likewise.
17149         (aarch64_strip_extend): Likewise.
17150         (aarch64_extr_rtx_p): Likewise.
17151         (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
17152         a CONST_INT when the mode parameter is VOIDmode.
17153         (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
17155 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17157         * machmode.h (bitwise_mode_for_mode): Return opt_mode.
17158         * stor-layout.c (bitwise_mode_for_mode): Likewise.
17159         (bitwise_type_for_mode): Update accordingly.
17161 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17163         * stor-layout.h (mode_for_size_tree): Return an opt_mode.
17164         * stor-layout.c (mode_for_size_tree): Likewise.
17165         (mode_for_array): Update accordingly.
17166         (layout_decl): Likewise.
17167         (compute_record_mode): Likewise.  Only set the mode once.
17169 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17171         * target.def (get_mask_mode): Change return type to opt_mode.
17172         Expand commentary.
17173         * doc/tm.texi: Regenerate.
17174         * targhooks.h (default_get_mask_mode): Return an opt_mode.
17175         * targhooks.c (default_get_mask_mode): Likewise.
17176         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17177         * optabs-query.c (can_vec_mask_load_store_p): Update use of
17178         targetm.get_mask_mode.
17179         * tree.c (build_truth_vector_type): Likewise.
17181 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17183         * machmode.h (mode_for_vector): Return an opt_mode.
17184         * stor-layout.c (mode_for_vector): Likewise.
17185         (mode_for_int_vector): Update accordingly.
17186         (layout_type): Likewise.
17187         * config/i386/i386.c (emit_memmov): Likewise.
17188         (ix86_expand_set_or_movmem): Likewise.
17189         (ix86_expand_vector_init): Likewise.
17190         (ix86_get_mask_mode): Likewise.
17191         * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
17192         Likewise.
17193         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
17194         * expmed.c (extract_bit_field_1): Likewise.
17195         * expr.c (expand_expr_real_2): Likewise.
17196         * optabs-query.c (can_vec_perm_p): Likewise.
17197         (can_vec_mask_load_store_p): Likewise.
17198         * optabs.c (expand_vec_perm): Likewise.
17199         * targhooks.c (default_get_mask_mode): Likewise.
17200         * tree-vect-stmts.c (vectorizable_store): Likewise.
17201         (vectorizable_load): Likewise.
17202         (get_vectype_for_scalar_type_and_size): Likewise.
17204 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17206         * machmode.h (mode_for_int_vector): New function.
17207         * stor-layout.c (mode_for_int_vector): Likewise.
17208         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
17209         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
17210         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
17211         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
17212         (s390_expand_vcond): Likewise.
17214 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17216         * machmode.h (opt_machine_mode): New type.
17217         (opt_mode<T>): Allow construction from anything that can be
17218         converted to a T.
17219         (is_a, as_a, dyn_cast): Add overloads for opt_mode.
17220         (mode_for_size): Return an opt_machine_mode.
17221         * stor-layout.c (mode_for_size): Likewise.
17222         (mode_for_size_tree): Update call accordingly.
17223         (bitwise_mode_for_mode): Likewise.
17224         (make_fract_type): Likewise.
17225         (make_accum_type): Likewise.
17226         * caller-save.c (replace_reg_with_saved_mem): Update call
17227         accordingly.
17228         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17229         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17230         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17231         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17232         * expmed.c (extract_bit_field_1): Likewise.
17233         * reload.c (get_secondary_mem): Likewise.
17234         * varasm.c (assemble_integer): Likewise.
17235         * lower-subreg.c (simplify_subreg_concatn): Likewise.  Move
17236         early-out.
17238 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17240         * machmode.h (decimal_float_mode_for_size): New function.
17241         * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
17242         (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
17243         (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17244         (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17245         (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17246         (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17248 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17250         * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17251         (get_builtin_sync_mode): Likewise.
17252         (expand_ifn_atomic_compare_exchange): Likewise.
17253         (expand_builtin_atomic_clear): Likewise.
17254         (expand_builtin_atomic_test_and_set): Likewise.
17255         (fold_builtin_atomic_always_lock_free): Likewise.
17256         * calls.c (compute_argument_addresses): Likewise.
17257         (emit_library_call_value_1): Likewise.
17258         (store_one_arg): Likewise.
17259         * combine.c (combine_instructions): Likewise.
17260         * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17261         * config/arm/arm.c (arm_function_value): Likewise.
17262         (aapcs_allocate_return_reg): Likewise.
17263         * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17264         * config/i386/i386.c (construct_container): Likewise.
17265         (ix86_gimplify_va_arg): Likewise.
17266         (ix86_expand_sse_cmp): Likewise.
17267         (emit_memmov): Likewise.
17268         (emit_memset): Likewise.
17269         (expand_small_movmem_or_setmem): Likewise.
17270         (ix86_expand_pextr): Likewise.
17271         (ix86_expand_pinsr): Likewise.
17272         * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17273         * config/microblaze/microblaze.c (microblaze_block_move_straight):
17274         Likewise.
17275         * config/mips/mips.c (mips_function_value_1) Likewise.
17276         (mips_block_move_straight): Likewise.
17277         (mips_expand_ins_as_unaligned_store): Likewise.
17278         * config/powerpcspe/powerpcspe.c
17279         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17280         (rs6000_darwin64_record_arg_flush): Likewise.
17281         * config/rs6000/rs6000.c
17282         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17283         (rs6000_darwin64_record_arg_flush): Likewise.
17284         * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17285         (sparc_function_value_1): Likewise.
17286         * config/spu/spu.c (adjust_operand): Likewise.
17287         (spu_emit_branch_or_set): Likewise.
17288         (arith_immediate_p): Likewise.
17289         * emit-rtl.c (gen_lowpart_common): Likewise.
17290         * expr.c (expand_expr_real_1): Likewise.
17291         * function.c (assign_parm_setup_block): Likewise.
17292         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17293         * reload1.c (alter_reg): Likewise.
17294         * stor-layout.c (mode_for_vector): Likewise.
17295         (layout_type): Likewise.
17297 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17299         * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17300         (spu_convert_move): Likewise.
17301         * lower-subreg.c (resolve_simple_move): Likewise.
17303 2017-09-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17305         PR target/81833
17306         * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17307         define_insn to a define_expand.
17308         (altivec_vsum2sws_direct): New define_insn.
17309         (altivec_vsumsws): Convert from a define_insn to a define_expand.
17311 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17313         * config/arm/arm.c (arm_option_params_internal): Improve setting of
17314         max_insns_skipped.
17316 2017-09-05  H.J. Lu  <hongjiu.lu@intel.com>
17318         PR target/59501
17319         PR target/81624
17320         PR target/81769
17321         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17322         realign stack if stack alignment needed is less than incoming
17323         stack boundary.
17325 2017-09-05  Marek Polacek  <polacek@redhat.com>
17327         PR sanitizer/82072
17328         * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17329         check earlier.
17331 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17333         * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17335 2017-09-05  Richard Biener  <rguenther@suse.de>
17337         PR tree-optimization/82084
17338         * fold-const.c (can_native_encode_string_p): Handle wide characters.
17340 2017-09-05  Richard Biener  <rguenther@suse.de>
17342         PR tree-optimization/82102
17343         * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17345 2017-09-05  Martin Liska  <mliska@suse.cz>
17347         PR tree-optimization/82032
17348         * tree-cfg.c (generate_range_test): New function.
17349         * tree-cfg.h (generate_range_test): Declared here.
17350         * tree-cfgcleanup.c (convert_single_case_switch): New function.
17351         (cleanup_control_expr_graph): Use it.
17352         * tree-switch-conversion.c (try_switch_expansion): Remove
17353         assert.
17354         (emit_case_nodes): Use generate_range_test.
17356 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17358         PR target/82098
17359         * config/i386/i386.md (*<btsc><mode>_mask): Add
17360         TARGET_USE_BT to insn constraint.
17361         (*btr<mode>_mask): Ditto.
17363 2017-09-04  Wilco Dijkstra  <wdijkstr@arm.com>
17365         * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17366         (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17368 2017-09-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17370         PR target/77308
17371         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17372         TARGET_NEON and TARGET_IWMMXT.
17373         (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17374         TARGET_NEON and TARGET_IWMMXT.
17375         (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17377 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17379         * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17380         (ix86_rewrite_tls_address): Ditto.
17381         * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17382         (ix86_rewrite_tls_address_1): Ditto.
17383         (ix86_rewrite_tls_address): Ditto.
17384         * config/i386/predicates.md (tls_address_pattern): New predicate.
17385         * config/i386/i386.md (TLS address splitter): New splitter.
17387 2017-09-04  Richard Biener  <rguenther@suse.de>
17389         PR tree-optimization/82084
17390         * fold-const.h (can_native_encode_string_p): Declare.
17391         * fold-const.c (can_native_encode_string_p): Factor out from ...
17392         (native_encode_string): ... here.
17393         * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17394         vectorizing stores from constants we later cannot handle.
17396 2017-09-04  Marek Polacek  <polacek@redhat.com>
17398         PR c/81783
17399         * doc/invoke.texi: Update -Wtautological-compare documentation.
17401 2017-09-04  Jeff Law  <law@redhat.com>
17403         PR tree-optimization/64910
17404         * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17405         swap the first and last operand if the last is a constant.
17407 2017-09-04  Marek Polacek  <polacek@redhat.com>
17409         PR sanitizer/82072
17410         * convert.c (do_narrow): When sanitizing signed integer overflows,
17411         bail out for signed types.
17412         (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17414 2017-09-04  Richard Biener  <rguenther@suse.de>
17416         PR tree-optimization/82060
17417         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17418         Move devirtualization after stmt folding and before EH/AB/noreturn
17419         cleanup to get the stmt refs canonicalized.  Use a bool instead
17420         of gimple_modified_p since that doesn't work for NOPs.  Schedule
17421         NOPs generated by folding for removal.
17423 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17424             Alan Hayward  <alan.hayward@arm.com>
17425             David Sherwood  <david.sherwood@arm.com>
17427         * coretypes.h (pad_direction): New enum.
17428         * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17429         (FUNCTION_ARG_PADDING): Likewise.
17430         * target.def (function_arg_padding): New hook.
17431         * targhooks.h (default_function_arg_padding): Declare.
17432         * targhooks.c (default_function_arg_padding): New function.
17433         * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17434         (TARGET_FUNCTION_ARG_PADDING): ...this.
17435         * doc/tm.texi: Regenerate.
17436         * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17437         instead of direction.
17438         (compute_argument_addresses): Likewise.
17439         (load_register_parameters): Likewise.
17440         (emit_library_call_value_1): Likewise.
17441         (store_one_arg): Use targetm.calls.function_arg_padding instead
17442         of FUNCTION_ARG_PADDING.
17443         (must_pass_in_stack_var_size_or_pad): Likewise.
17444         * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17445         (emit_group_store): Likewise.
17446         (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17447         instead of FUNCTION_ARG_PADDING.
17448         (emit_push_insn): Likewise, and propagate enum change throughout
17449         function.
17450         * function.h (direction): Delete.
17451         (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17452         of direction.
17453         * function.c (assign_parm_find_stack_rtl): Likewise.
17454         (assign_parm_setup_block_p): Likewise.
17455         (assign_parm_setup_block): Likewise.
17456         (gimplify_parameters): Likewise.
17457         (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17458         instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17459         function.
17460         * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17461         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17462         * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17463         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17464         (aarch64_function_arg_padding): ...this new function.
17465         (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17466         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17467         * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17468         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17469         * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17470         * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17471         (arm_pad_arg_upward): Replace with...
17472         (arm_function_arg_padding): ...this new function.
17473         * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17474         of direction.
17475         * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17476         * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17477         * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17478         (ia64_hpux_function_arg_padding): Replace with...
17479         (ia64_function_arg_padding): ...this new function.  Use pad_direction
17480         instead of direction.  Check for TARGET_HPUX.
17481         * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17482         * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17483         (iq2000_function_arg_padding): New function.
17484         * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17485         * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17486         (mips_function_arg_padding): ...this new function.
17487         (mips_pad_reg_upward): Update accordingly.
17488         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17489         * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17490         targetm.calls.function_arg_padding.
17491         (FUNCTION_ARG_PADDING): Delete.
17492         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17493         * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17494         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17495         * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17496         (nios2_block_reg_padding): Return pad_direction instead of direction.
17497         * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17498         instead of direction.
17499         (nios2_function_arg_padding): Likewise.  Make static.
17500         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17501         * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17502         (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17503         * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17504         * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17505         (pa_function_arg_padding): Make static.  Return pad_direction instead
17506         of direction.
17507         * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17508         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17509         * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17510         instead of direction.  Use targetm.calls.function_arg_padding.
17511         * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17512         * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17513         * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17514         * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17515         * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17516         Redefine.
17517         (function_arg_padding): Rename to...
17518         (rs6000_function_arg_padding): ...this.  Make static.  Return
17519         pad_direction instead of direction.
17520         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17521         * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17522         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17523         * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17524         instead of direction.  Use targetm.calls.function_arg_padding.
17525         * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17526         * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17527         * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17528         * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17529         * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17530         (function_arg_padding): Rename to...
17531         (rs6000_function_arg_padding): ...this.  Make static.  Return
17532         pad_direction instead of direction.
17533         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17534         * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17535         * config/s390/s390.c (s390_function_arg_padding): New function.
17536         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17537         * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17538         * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17539         * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17540         (function_arg_padding): Rename to...
17541         (sparc_function_arg_padding): ...this.  Make static.  Return
17542         pad_direction instead of direction.
17543         * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17544         * config/spu/spu.c (spu_function_arg_padding): New function.
17545         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17546         * system.h (FUNCTION_ARG_PADDING): Poison.
17548 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17549             Alan Hayward  <alan.hayward@arm.com>
17550             David Sherwood  <david.sherwood@arm.com>
17552         * target.def (modes_tieable_p): New hook.
17553         * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17554         (TARGET_MODES_TIEABLE_P): ...this.
17555         * doc/tm.texi.in: Regenerate.
17556         * hooks.h (hook_bool_mode_mode_true): Declare.
17557         * hooks.c (hook_bool_mode_mode_true): New function.
17558         * combine.c (subst): Use targetm.modes_tieable_p instead of
17559         MODES_TIEABLE_P.
17560         * dse.c (find_shift_sequence): Likewise.
17561         * expmed.c (extract_low_bits): Likewise.
17562         * lower-subreg.c: Include target.h.
17563         (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17564         MODES_TIEABLE_P.
17565         * rtlanal.c (rtx_cost): Likewise.
17566         * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17567         * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17568         * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17569         (TARGET_MODES_TIEABLE_P): Redefine.
17570         * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17571         * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17572         (TARGET_MODES_TIEABLE_P): Redefine.
17573         * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17574         * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17575         (arc_modes_tieable_p): New function.
17576         * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17577         * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17578         * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17579         (arm_modes_tieable_p): Make static.
17580         * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17581         * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17582         * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17583         (TARGET_MODES_TIEABLE_P): Redefine.
17584         * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17585         * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17586         (TARGET_MODES_TIEABLE_P): Redefine.
17587         * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17588         * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17589         (cr16_modes_tieable_p): New function.
17590         * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17591         * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17592         * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17593         (TRULY_NOOP_TRUNCATION): Update comment.
17594         * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17595         (TRULY_NOOP_TRUNCATION): Update comment.
17596         * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17597         (frv_modes_tieable_p): New function.
17598         * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17599         * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17600         * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17601         (TARGET_MODES_TIEABLE_P): Redefine.
17602         * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17603         * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17604         * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17605         (TARGET_MODES_TIEABLE_P): Redefine.
17606         * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17607         * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17608         (ia64_modes_tieable_p): New function.
17609         * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17610         * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17611         (iq2000_modes_tieable_p): New function.
17612         * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17613         * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17614         (lm32_modes_tieable_p): New function.
17615         * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17616         * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17617         * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17618         (TARGET_MODES_TIEABLE_P): Redefine.
17619         * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17620         * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17621         (m32r_modes_tieable_p): New function.
17622         * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17623         * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17624         (m68k_modes_tieable_p): New function.
17625         * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17626         * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17627         (mcore_modes_tieable_p): New function.
17628         * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17629         * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17630         function.
17631         (TARGET_MODES_TIEABLE_P): Redefine.
17632         * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17633         * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17634         * config/mips/mips.c (mips_modes_tieable_p): Make static.
17635         (TARGET_MODES_TIEABLE_P): Redefine.
17636         * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17637         * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17638         * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17639         * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17640         (mn10300_modes_tieable_p): ...this and make static.
17641         (TARGET_MODES_TIEABLE_P): Redefine.
17642         * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17643         * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17644         * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17645         * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17646         (msp430_modes_tieable_p): Make static.
17647         * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17648         * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17649         (TARGET_MODES_TIEABLE_P): Redefine.
17650         * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17651         * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17652         * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17653         (TARGET_MODES_TIEABLE_P): Redefine.
17654         * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17655         * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17656         * config/pa/pa.c (pa_modes_tieable_p): Make static.
17657         (TARGET_MODES_TIEABLE_P): Redefine.
17658         * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17659         * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17660         (pdp11_modes_tieable_p): New function.
17661         * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17662         * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17663         (rs6000_modes_tieable_p): New function.
17664         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17665         * config/powerpcspe/powerpcspe.md: Update comment.
17666         * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17667         * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17668         (TARGET_MODES_TIEABLE_P): Redefine.
17669         * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17670         * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17671         (rl78_modes_tieable_p): New function.
17672         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17673         * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17674         (rs6000_modes_tieable_p): New function.
17675         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17676         * config/rs6000/rs6000.md: Update comment.
17677         * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17678         * config/rx/rx.c (rx_modes_tieable_p): New function.
17679         (TARGET_MODES_TIEABLE_P): Redefine.
17680         * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17681         * config/s390/s390.c (s390_modes_tieable_p): New function.
17682         (TARGET_MODES_TIEABLE_P): Redefine.
17683         * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17684         * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17685         (sh_modes_tieable_p): New function.
17686         * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17687         * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17688         * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17689         (sparc_modes_tieable_p): Make static.
17690         * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17691         * config/spu/spu.c (spu_modes_tieable_p): New function.
17692         (TARGET_MODES_TIEABLE_P): Redefine.
17693         * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17694         * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17695         (TARGET_MODES_TIEABLE_P): Redefine.
17696         * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17697         * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17698         * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17699         * config/v850/v850.c (v850_modes_tieable_p): New function.
17700         (TARGET_MODES_TIEABLE_P): Redefine.
17701         * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17702         * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17703         * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17704         (visium_modes_tieable_p): New function.
17705         * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17706         * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17707         (xtensa_modes_tieable_p): New function.
17708         * system.h (MODES_TIEABLE_P): Poison.
17710 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17711             Alan Hayward  <alan.hayward@arm.com>
17712             David Sherwood  <david.sherwood@arm.com>
17714         * target.def (hard_regno_mode_ok): New hook.
17715         * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17716         (TARGET_HARD_REGNO_MODE_OK): ...this.
17717         * doc/tm.texi.in: Regenerate.
17718         * hooks.h (hook_bool_uint_mode_true): Declare.
17719         * hooks.c (hook_bool_uint_mode_true): New function.
17720         * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17721         HARD_REGNO_MODE_OK.
17722         * genpreds.c (write_insn_preds_c): Add an include of target.h.
17723         * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17724         instead of HARD_REGNO_MODE_OK.
17725         * caller-save.c: Include target.h.
17726         (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17727         HARD_REGNO_MODE_OK.
17728         * combine.c (can_combine_p): Likewise.
17729         (combinable_i3pat): Likewise.
17730         (can_change_dest_mode): Likewise.
17731         * expr.c (init_expr_target): Likewise.
17732         (convert_move): Likewise.
17733         (convert_modes): Likewise.
17734         * ira.c (setup_prohibited_class_mode_regs): Likewise.
17735         (setup_prohibited_mode_move_regs): Likewise.
17736         * ira.h (target_ira): Likewise.
17737         * lra-assigns.c (find_hard_regno_for_1): Likewise.
17738         * lra-constraints.c (process_alt_operands): Likewise.
17739         (split_reg): Likewise.
17740         * recog.c (peep2_find_free_register): Likewise.
17741         * ree.c (combine_reaching_defs): Likewise.
17742         * regcprop.c (maybe_mode_change): Likewise.
17743         * reginfo.c (init_reg_sets_1): Likewise.
17744         (choose_hard_reg_mode): Likewise.
17745         (simplifiable_subregs): Likewise.
17746         * regrename.c (check_new_reg_p): Likewise.
17747         * reload.c (find_valid_class): Likewise.
17748         (find_valid_class_1): Likewise.
17749         (reload_inner_reg_of_subreg): Likewise.
17750         (push_reload): Likewise.
17751         (combine_reloads): Likewise.
17752         (find_dummy_reload): Likewise.
17753         (find_reloads): Likewise.
17754         * reload1.c (find_reg): Likewise.
17755         (set_reload_reg): Likewise.
17756         (allocate_reload_reg): Likewise.
17757         (choose_reload_regs): Likewise.
17758         (reload_adjust_reg_for_temp): Likewise.
17759         * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17760         (simplify_subreg_regno): Likewise.
17761         * sel-sched.c (init_regs_for_mode): Likewise.
17762         * varasm.c (make_decl_rtl): Likewise.
17763         * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17764         (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17765         HARD_REGNO_MODE_OK.
17766         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17767         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17768         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17769         * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17770         * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17771         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17772         * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17773         (arc_mode_class): Delete.
17774         (HARD_REGNO_MODE_OK): Delete.
17775         * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17776         (arc_hard_regno_mode_ok): Rename old array to...
17777         (arc_hard_regno_mode_ok_modes): ...this.
17778         (arc_conditional_register_usage): Update accordingly.
17779         (arc_mode_class): Make static.
17780         (arc_hard_regno_mode_ok): New function.
17781         * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17782         * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17783         * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17784         (arm_hard_regno_mode_ok): Make static.
17785         * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17786         HARD_REGNO_MODE_OK.
17787         * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17788         * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17789         * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17790         return a bool.
17791         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17792         * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17793         * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17794         * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17795         (bfin_hard_regno_mode_ok): ...this.  Make static and return a bool.
17796         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17797         * config/bfin/predicates.md (valid_reg_operand): Use
17798         targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17799         * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17800         * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17801         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17802         * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17803         * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17804         * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17805         (cr16_hard_regno_mode_ok): Make static and return a bool.
17806         * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17807         * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17808         (cris_hard_regno_mode_ok): New function.
17809         * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17810         (epiphany_mode_class): Delete.
17811         (HARD_REGNO_MODE_OK): Delete.
17812         * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17813         * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17814         (hard_regno_mode_ok): Rename to...
17815         (epiphany_hard_regno_mode_ok): ...this.  Make static and return a bool.
17816         * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17817         * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17818         HARD_REGNO_MODE_OK.
17819         * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17820         * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17821         * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17822         (frv_hard_regno_mode_ok): Make static and return a bool.
17823         * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17824         HARD_REGNO_MODE_OK.
17825         * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17826         * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17827         * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17828         * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17829         and return a bool.
17830         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17831         * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17832         * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17833         * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17834         return a bool.
17835         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17836         * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17837         * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17838         (ia64_hard_regno_mode_ok): New function.
17839         * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17840         * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17841         (iq2000_hard_regno_mode_ok): New function.
17842         * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17843         * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17844         (lm32_hard_regno_mode_ok): New function.
17845         * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17846         * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17847         * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17848         instead of HARD_REGNO_MODE_OK.
17849         (m32c_hard_regno_ok): Rename to...
17850         (m32c_hard_regno_mode_ok): ...this.  Make static and return a bool.
17851         (m32c_cannot_change_mode_class): Update accordingly.
17852         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17853         * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17854         (m32r_mode_class): Delete.
17855         (HARD_REGNO_MODE_OK): Delete.
17856         * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17857         (m32r_hard_regno_mode_ok): Rename to...
17858         (m32r_hard_regno_modes): ...this.
17859         (m32r_mode_class): Make static.
17860         (m32r_hard_regno_mode_ok): New function.
17861         * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17862         * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17863         * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17864         (m68k_hard_regno_mode_ok): Make static.
17865         * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17866         * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17867         (mcore_hard_regno_mode_ok): New function.
17868         * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17869         (HARD_REGNO_MODE_OK): Delete.
17870         * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17871         Rename to...
17872         (microblaze_hard_regno_mode_ok_p): ...this and make static.
17873         (microblaze_hard_regno_mode_ok): New function.
17874         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17875         * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17876         (mips_hard_regno_mode_ok): Delete.
17877         * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17878         (mips_hard_regno_mode_ok_p): ...this and make static.
17879         (mips_hard_regno_mode_ok_p): Rename to...
17880         (mips_hard_regno_mode_ok_uncached): ...this.
17881         (mips_hard_regno_mode_ok): New function.
17882         (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17883         of HARD_REGNO_MODE_OK.
17884         (mips_option_override): Update after above name changes.
17885         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17886         * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17887         * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17888         * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17889         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17890         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17891         * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17892         * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17893         * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17894         * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17895         (msp430_hard_regno_mode_ok): Make static and return a bool.
17896         * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17897         * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17898         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17899         and return a bool.
17900         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17901         * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17902         * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17903         * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17904         * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17905         (PA_HARD_REGNO_MODE_OK): ...this
17906         * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17907         (PA_HARD_REGNO_MODE_OK): ...this.
17908         * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17909         (pa_hard_regno_mode_ok): New function.
17910         * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17911         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17912         (pdp11_hard_regno_mode_ok): New function.
17913         * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17914         * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17915         Delete.
17916         * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17917         Make static.
17918         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17919         (rs6000_hard_regno_mode_ok): Rename to...
17920         (rs6000_hard_regno_mode_ok_uncached): ...this.
17921         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17922         (rs6000_hard_regno_mode_ok): New function.
17923         * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
17924         * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
17925         * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
17926         (riscv_hard_regno_mode_ok): ...this and make static.
17927         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17928         * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
17929         * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
17930         * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17931         (rl78_hard_regno_mode_ok): Make static and return bool.
17932         * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
17933         * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
17934         Delete.
17935         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
17936         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17937         (rs6000_hard_regno_mode_ok): Rename to...
17938         (rs6000_hard_regno_mode_ok_uncached): ...this.
17939         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17940         (rs6000_hard_regno_mode_ok): New function.
17941         * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
17942         * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
17943         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17944         * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
17945         * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
17946         * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
17947         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17948         * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
17949         * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
17950         * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17951         (sh_hard_regno_mode_ok): Make static.
17952         * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
17953         instead of HARD_REGNO_MODE_OK.
17954         * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
17955         (sparc_mode_class): Delete.
17956         (HARD_REGNO_MODE_OK): Delete.
17957         * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17958         (hard_regno_mode_classes): Make static.
17959         (sparc_mode_class): Likewise.
17960         (sparc_hard_regno_mode_ok): New function.
17961         * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
17962         * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
17963         * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
17964         function.
17965         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17966         * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
17967         * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
17968         * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
17969         * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
17970         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17971         * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
17972         * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
17973         * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17974         (visium_hard_regno_mode_ok): New function.
17975         * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
17976         instead of HARD_REGNO_MODE_OK.
17977         * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
17978         (HARD_REGNO_MODE_OK): Delete.
17979         * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
17980         (xtensa_hard_regno_mode_ok_p): ...this and make static.
17981         (xtensa_option_override): Update accordingly.
17982         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17983         (xtensa_hard_regno_mode_ok): New function.
17984         * system.h (HARD_REGNO_MODE_OK): Poison.
17986 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17987             Alan Hayward  <alan.hayward@arm.com>
17988             David Sherwood  <david.sherwood@arm.com>
17990         * target.def (hard_regno_call_part_clobbered): New hook.
17991         * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
17992         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
17993         * doc/tm.texi: Regenerate.
17994         * hooks.h (hook_bool_uint_mode_false): Declare.
17995         * hooks.c (hook_bool_uint_mode_false): New function.
17996         * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17997         * cselib.c (cselib_process_insn): Use
17998         targetm.hard_regno_call_part_clobbered instead of
17999         HARD_REGNO_CALL_PART_CLOBBERED.
18000         * ira-conflicts.c (ira_build_conflicts): Likewise.
18001         * ira-costs.c (ira_tune_allocno_costs): Likewise.
18002         * lra-constraints.c (need_for_call_save_p): Likewise.
18003         * lra-lives.c: Include target.h.
18004         (check_pseudos_live_through_calls): Use
18005         targetm.hard_regno_call_part_clobbered instead of
18006         HARD_REGNO_CALL_PART_CLOBBERED.
18007         * regcprop.c: Include target.h.
18008         (copyprop_hardreg_forward_1): Use
18009         targetm.hard_regno_call_part_clobbered instead of
18010         HARD_REGNO_CALL_PART_CLOBBERED.
18011         * reginfo.c (choose_hard_reg_mode): Likewise.
18012         * regrename.c (check_new_reg_p): Likewise.
18013         * reload.c (find_equiv_reg): Likewise.
18014         * reload1.c (emit_reload_insns): Likewise.
18015         * sched-deps.c (deps_analyze_insn): Likewise.
18016         * sel-sched.c (init_regs_for_mode): Likewise.
18017         (mark_unavailable_hard_regs): Likewise.
18018         * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
18019         * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18020         * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
18021         New function.
18022         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18023         * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18024         * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
18025         Delete.
18026         * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
18027         and return a bool.
18028         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18029         * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18030         * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
18031         function.
18032         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18033         * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18034         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
18035         function.
18036         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18037         * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
18038         Delete.
18039         * config/powerpcspe/powerpcspe.c
18040         (rs6000_hard_regno_call_part_clobbered): New function.
18041         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18042         * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18043         * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
18044         New function.
18045         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18046         * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18047         * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
18048         function.
18049         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18050         * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18051         * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
18053 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18054             Alan Hayward  <alan.hayward@arm.com>
18055             David Sherwood  <david.sherwood@arm.com>
18057         * rtl.h (subreg_memory_offset): Declare.
18058         * emit-rtl.c (subreg_memory_offset): New function.
18059         * expmed.c (store_bit_field_1): Use it.
18060         * expr.c (undefined_operand_subword_p): Likewise.
18061         * simplify-rtx.c (simplify_subreg): Likewise.
18063 2017-09-04  Alexander Monakov  <amonakov@ispras.ru>
18065         PR rtl-optimization/57448
18066         PR target/67458
18067         PR target/81316
18068         * optabs.c (expand_atomic_load): Place compiler memory barriers if
18069         using atomic_load pattern.
18070         (expand_atomic_store): Likewise.
18072 2017-09-04  Jakub Jelinek  <jakub@redhat.com>
18074         PR sanitizer/81981
18075         * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
18076         and UBSAN_BOUNDS internal calls.  Clean up IFN_UBSAN_OBJECT_SIZE
18077         handling.  Use replace_call_with_value with NULL instead of
18078         gsi_replace, unlink_stmt_vdef and release_defs.
18080         * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
18081         instead of tab.
18083         * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
18085 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18087         PR bootstrap/82045
18088         * rtl.h (emit_library_call_value_1): Declare.
18089         (emit_library_call): Replace declaration with a series of overloads.
18090         Remove the parameter count argument.
18091         (emit_library_call_value): Likewise.
18092         * calls.c (emit_library_call_value_1): Make global.  Replace varargs
18093         with an "rtx_mode_t *".
18094         (emit_library_call_value): Delete.
18095         (emit_library_call): Likewise.
18096         * asan.c (asan_emit_stack_protection): Update calls accordingly.
18097         (asan_emit_allocas_unpoison): Likewise.
18098         * builtins.c (expand_builtin_powi): Likewise.
18099         (expand_asan_emit_allocas_unpoison): Likewise.
18100         * cfgexpand.c (expand_main_function): Likewise.
18101         * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
18102         * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
18103         * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
18104         * config/arm/arm.c (arm_trampoline_init): Likewise.
18105         (arm_call_tls_get_addr): Likewise.
18106         (arm_expand_divmod_libfunc): Likewise.
18107         * config/bfin/bfin.md (umulsi3_highpart): Likewise.
18108         (smulsi3_highpart): Likewise.
18109         * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
18110         (c6x_expand_compare): Likewise.
18111         (c6x_expand_movmem): Likewise.
18112         * config/frv/frv.c (frv_trampoline_init): Likewise.
18113         * config/i386/i386.c (ix86_trampoline_init): Likewise.
18114         (ix86_expand_divmod_libfunc): Likewise.
18115         * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
18116         (ia64_expand_compare): Likewise.
18117         (ia64_profile_hook): Likewise.
18118         * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
18119         (nonlocal_goto): Likewise.
18120         (restore_stack_nonlocal): Likewise.
18121         * config/m32r/m32r.c (block_move_call): Likewise.
18122         (m32r_trampoline_init): Likewise.
18123         * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
18124         * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
18125         (m68k_call_m68k_read_tp): Likewise.
18126         * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
18127         (microblaze_expand_divide): Likewise.
18128         * config/mips/mips.h (mips_args): Likewise.
18129         * config/mips/sdemtk.h (mips_sync_icache): Likewise.
18130         (MIPS_ICACHE_SYNC): Likewise.
18131         * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
18132         (nios2_trampoline_init): Likewise.
18133         * config/pa/pa.c (hppa_tls_call): Likewise.
18134         (pa_trampoline_init): Likewise.
18135         * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
18136         * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
18137         (expand_strn_compare): Likewise.
18138         (rs6000_generate_compare): Likewise.
18139         (rs6000_expand_float128_convert): Likewise.
18140         (output_profile_hook): Likewise.
18141         (rs6000_trampoline_init): Likewise.
18142         * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
18143         * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
18144         * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
18145         * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
18146         (rs6000_generate_compare): Likewise.
18147         (rs6000_expand_float128_convert): Likewise.
18148         (output_profile_hook): Likewise.
18149         (rs6000_trampoline_init): Likewise.
18150         * config/rs6000/rs6000.md (neg<mode>2): Likewise.
18151         * config/sh/sh.c (sh_trampoline_init): Likewise.
18152         * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
18153         (sparc_emit_float_lib_cmp): Likewise.
18154         (sparc32_initialize_trampoline): Likewise.
18155         (sparc64_initialize_trampoline): Likewise.
18156         (sparc_profile_hook): Likewise.
18157         * config/spu/spu.c (ea_load_store): Likewise.
18158         * config/spu/spu.md (floatunssidf2): Likewise.
18159         * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
18160         * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
18161         * config/visium/visium.c (expand_block_move_4): Likewise.
18162         (expand_block_move_2): Likewise.
18163         (expand_block_move_1): Likewise.
18164         (expand_block_set_4): Likewise.
18165         (expand_block_set_2): Likewise.
18166         (expand_block_set_1): Likewise.
18167         (visium_trampoline_init): Likewise.
18168         (visium_profile_hook): Likewise.
18169         * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
18170         (xtensa_setup_frame_addresses): Likewise.
18171         (xtensa_trampoline_init): Likewise.
18172         * except.c (sjlj_emit_function_enter): Likewise.
18173         (sjlj_emit_function_exit): Likewise.
18174         * explow.c (allocate_dynamic_stack_space): Likewise.
18175         (probe_stack_range): Likewise.
18176         * expr.c (convert_mode_scalar): Likewise.
18177         * optabs.c (expand_binop): Likewise.
18178         (expand_twoval_binop_libfunc): Likewise.
18179         (expand_unop): Likewise.
18180         (prepare_cmp_insn): Likewise.
18181         (prepare_float_lib_cmp): Likewise.
18182         (expand_float): Likewise.
18183         (expand_fix): Likewise.
18184         (expand_fixed_convert): Likewise.
18185         (maybe_emit_sync_lock_test_and_set): Likewise.
18186         (expand_atomic_compare_and_swap): Likewise.
18187         (expand_mem_thread_fence): Likewise.
18188         (expand_atomic_fetch_op): Likewise.
18190 2017-09-03  Gerald Pfeifer  <gerald@pfeifer.com>
18192         * doc/generic.texi (OpenACC): Adjust URL.
18193         * doc/invoke.texi (C Dialect Options): Ditto.
18195 2017-09-03  Uros Bizjak  <ubizjak@gmail.com>
18197         * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
18198         predicate for operand 1.  Add (m,<S>) constraint.
18199         (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
18200         Prevent memory operand 1 with register operand 2.
18202 2017-09-01  Segher Boessenkool  <segher@kernel.crashing.org>
18204         PR rtl-optimization/82024
18205         * combine.c (try_combine): If the combination result is a PARALLEL,
18206         and we only need to retain the SET in there that would be placed
18207         at I2, check that we can place that at I3 instead, before doing so.
18209 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18211         PR target/81766
18212         * config/i386/i386.c (ix86_init_large_pic_reg): Return label
18213         instead of void.
18214         (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
18215         if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
18216         and label.
18218 2017-09-01  Joerg Sonnenberger  <joerg@bec.de>
18219             Jeff Law  <law@redhat.com>
18221         * varasm.c (bss_initializer_p): Do not put constants into .bss
18222         (categorize_decl_for_section): Handle bss_initializer_p returning
18223         false when DECL_INITIAL is NULL.
18225 2017-09-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18227         PR target/82012
18228         * config/s390/s390.c (s390_can_inline_p): New function.
18230 2017-09-01  Jeff Law  <law@redhat.com>
18232         PR tree-optimization/82052
18233         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
18234         Always initialize the returned slot after a hash table miss
18235         when INSERT is true.
18237 2017-09-01  Alexander Monakov  <amonakov@ispras.ru>
18239         * config/s390/s390.md (mem_signal_fence): Remove.
18240         * doc/md.texi (mem_signal_fence): Remove.
18241         * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
18242         Update comments.
18243         * target-insns.def (mem_signal_fence): Remove.
18245 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18247         PR sanitizer/81902
18248         * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18250         PR sanitizer/81923
18251         * asan.c (create_odr_indicator): Strip name encoding from assembler
18252         name before appending it after __odr_asan_.
18254 2017-09-01  Martin Liska  <mliska@suse.cz>
18256         PR tree-optimization/82059
18257         * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18258         frequency only when an edge is redirected.
18260 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18262         * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18263         * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18264         (arc_conditional_register_usage): Remove ARC600 lp_count
18265         exception.
18266         (arc_file_start): Emit Tag_ARC_CPU_variation.
18267         (arc_can_use_doloop_p): New conditions to use ZOLs.
18268         (hwloop_fail): New function.
18269         (hwloop_optimize): Likewise.
18270         (hwloop_pattern_reg): Likewise.
18271         (arc_doloop_hooks): New struct, to be used with reorg_loops.
18272         (arc_reorg_loops): New function, calls reorg_loops.
18273         (arc_reorg): Call arc_reorg_loops.  Remove old ZOL handling.
18274         (arc600_corereg_hazard): Remove ZOL checking, case handled by
18275         hwloop_optimize.
18276         (arc_loop_hazard): Remove function, functionality moved into
18277         hwloop_optimize.
18278         (arc_hazard): Remove arc_loop_hazard call.
18279         (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18280         into hwloop_optimize.
18281         (arc_label_align): Remove ZOL handling.
18282         * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18283         * config/arc/arc.md (doloop_begin): Remove pattern.
18284         (doloop_begin_i): Likewise.
18285         (doloop_end_i): Likewise.
18286         (doloop_fallback): Likewise.
18287         (doloop_fallback_m): Likewise.
18288         (doloop_end): Reimplement expand.
18289         (arc_lp): New pattern for LP instruction.
18290         (loop_end): New pattern.
18291         (loop_fail): Likewise.
18292         (decrement_and_branch_until_zero): Likewise.
18293         * config/arc/arc.opt (mlpc-width): New option.
18294         * doc/invoke.texi (mlpc-width): Document option.
18296 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18298         * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18299         (arc_ccfsm_advance): Fix checking for delay slots.
18300         (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18302 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18304         * config/arc/arc.md (movqi_insn): Add stores to save constant long
18305         immediates.
18306         (movhi_insn): Update store instruction constraint which are saving
18307         6-bit short immediates.
18308         (movsi_insn): Consider also short scaled load operations.
18309         (zero_extendhisi2_i): Use Usd constraint instead of T.
18310         (extendhisi2_i): Add q constraint.
18311         (arc_clzsi2): Add type and length attributes.
18312         (arc_ctzsi2): Likewise.
18313         * config/arc/constraints.md (Usc): Update constraint, the
18314         assembler can parse two relocations for a single instruction.
18316 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18318         * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18319         (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18321 2017-08-31  Olivier Hainque  <hainque@adacore.com>
18323         * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18324         match as powerpc-wrs-vxworks*.
18326 2017-08-31  James Greenhalgh  <james.greenhalgh@arm.com>
18328         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18329         register constraint for by-element operand.
18330         (aarch64_mls_elt_merge<mode>): Likewise.
18332 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18334         * config/arc/arc.c (arc_can_follow_jump): Check for short
18335         branches.
18337 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18339         * config.gcc: Use g.opt for arc.
18340         * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18341         functionality moved to ...
18342         (legitimate_scaled_address_p): New function, ...here.
18343         (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18344         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18345         (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18346         condition.
18347         (arc_override_options): Handle G option.
18348         (arc_output_pic_addr_const): Correct function definition.
18349         (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18350         (arc_decl_anon_ns_mem_p): Delete.
18351         (arc_in_small_data_p): Overhaul this function to take into
18352         consideration the value given via G option.
18353         (arc_rewrite_small_data_1): Renamed and corrected old
18354         arc_rewrite_small_data function.
18355         (arc_rewrite_small_data): New function.
18356         (small_data_pattern): Don't use pic_offset_table_rtx.
18357         * config/arc/arc.h (CC1_SPEC): Recognize G option.
18358         * config/arc/simdext.md (movmisalignv2hi): Use
18359         prepare_move_operands function.
18360         (mov*): Likewise.
18361         (movmisalign*): Likewise.
18362         * doc/invoke.texi (ARC options): Document -G option.
18364 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18366         * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18367         prototype.
18368         * config/arc/arc.c (arc_print_operand): Output scalled address for
18369         sdata whenever is possible.
18370         (arc_in_small_data_p): Allow sdata for 64bit datum when double
18371         load/stores are available.
18372         (compact_sda_memory_operand): Check for the alignment required by
18373         code density instructions.
18374         * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18375         constraint.
18376         * config/arc/constraints.md (Usd): Update constraint.
18377         (Us0): New constraint.
18378         (Usc): Update constraint.
18380 2017-08-31  Richard Biener  <rguenther@suse.de>
18382         PR middle-end/82054
18383         * dwarf2out.c (dwarf2out_early_global_decl): Process each
18384         function only once.
18386 2017-08-31  Tamar Christina  <tamar.christina@arm.com>
18388         * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18389         Resize type_signature.
18391 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18392             Alan Hayward  <alan.hayward@arm.com>
18393             David Sherwood  <david.sherwood@arm.com>
18395         * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18396         subregs whose inner modes can be stored in GPRs.
18397         (aarch64_classify_index): Likewise.
18399 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18400             Alan Hayward  <alan.hayward@arm.com>
18401             David Sherwood  <david.sherwood@arm.com>
18403         * config/aarch64/iterators.md (V_cmp_result): Rename to...
18404         (V_INT_EQUIV): ...this.
18405         (v_cmp_result): Rename to...
18406         (v_int_equiv): ...this.
18407         * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18408         * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18409         (copysign<mode>3): Likewise.
18410         (aarch64_simd_bsl<mode>_internal): Likewise.
18411         (aarch64_simd_bsl<mode>): Likewise.
18412         (vec_cmp<mode><mode>): Likewise.
18413         (vcond<mode><mode>): Likewise.
18414         (vcond<v_cmp_mixed><mode>): Likewise.
18415         (vcondu<mode><v_cmp_mixed>): Likewise.
18416         (aarch64_cm<optab><mode>): Likewise.
18417         (aarch64_cmtst<mode>): Likewise.
18418         (aarch64_fac<optab><mode>): Likewise.
18419         (vec_perm_const<mode>): Likewise.
18420         (vcond_mask_<mode><v_cmp_result>): Rename to...
18421         (vcond_mask_<mode><v_int_equiv>): ...this.
18422         (vec_cmp<mode><v_cmp_result>): Rename to...
18423         (vec_cmp<mode><v_int_equiv>): ...this.
18425 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18426             Alan Hayward  <alan.hayward@arm.com>
18427             David Sherwood  <david.sherwood@arm.com>
18429         * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18430         vector modes.
18431         * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18432         * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18433         (UNSPEC_LD4_DREG): New unspecs.
18434         * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18435         (aarch64_ld2<mode>_dreg_be): Replace with...
18436         (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18437         unspec.
18438         (aarch64_ld3<mode>_dreg_le)
18439         (aarch64_ld3<mode>_dreg_be): Replace with...
18440         (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18441         unspec.
18442         (aarch64_ld4<mode>_dreg_le)
18443         (aarch64_ld4<mode>_dreg_be): Replace with...
18444         (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18445         unspec.
18447 2017-08-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18449         PR tree-optimization/81987
18450         * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18451         insert an initializer in a location not dominated by the stride
18452         definition.
18454 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18456         * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18457         on the entire header of the finally block in the fallthru case.
18459 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18461         * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18463 2017-08-30  Pat Haugen  <pthaugen@us.ibm.com>
18465         * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18466         rs6000_emit_move_from_cr and call renamed function.
18467         (rs6000_emit_prologue): Call renamed functions.
18468         * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18469         movesi_from_cr, remove volatile CRs.
18471 2017-08-30  Jon Beniston  <jon@beniston.com>
18472             Richard Biener  <rguenther@suse.de>
18474         * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18475         of VECTOR_MODE_P check.
18476         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18477         element vector types.
18479 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18481         * df.h (df_read_modify_subreg_p): Remove in favor of...
18482         * rtl.h (read_modify_subreg_p): ...this new function.  Take a
18483         const_rtx instead of an rtx.
18484         * cprop.c (local_cprop_find_used_regs): Update accordingly.
18485         * df-problems.c (df_word_lr_mark_ref): Likewise.
18486         * ira-lives.c (mark_pseudo_reg_live): Likewise.
18487         (mark_pseudo_reg_dead): Likewise.
18488         (mark_ref_dead): Likewise.
18489         * reginfo.c (init_subregs_of_mode): Likewise.
18490         * sched-deps.c (sched_analyze_1): Likewise.
18491         * df-scan.c (df_def_record_1): Likewise.
18492         (df_uses_record): Likewise.
18493         (df_read_modify_subreg_p): Remove in favor of...
18494         * rtlanal.c (read_modify_subreg_p): ...this new function.  Take a
18495         const_rtx instead of an rtx.
18497 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18498             Alan Hayward  <alan.hayward@arm.com>
18499             David Sherwood  <david.sherwood@arm.com>
18501         * rtl.h (partial_subreg_p): New function.
18502         * caller-save.c (save_call_clobbered_regs): Use it.
18503         * calls.c (expand_call): Likewise.
18504         * combine.c (combinable_i3pat): Likewise.
18505         (simplify_set): Likewise.
18506         (make_extraction): Likewise.
18507         (make_compound_operation_int): Likewise.
18508         (gen_lowpart_or_truncate): Likewise.
18509         (force_to_mode): Likewise.
18510         (make_field_assignment): Likewise.
18511         (reg_truncated_to_mode): Likewise.
18512         (record_truncated_value): Likewise.
18513         (move_deaths): Likewise.
18514         * cse.c (record_jump_cond): Likewise.
18515         (cse_insn): Likewise.
18516         * cselib.c (cselib_lookup_1): Likewise.
18517         * expmed.c (extract_bit_field_using_extv): Likewise.
18518         * function.c (assign_parm_setup_reg): Likewise.
18519         * ifcvt.c (noce_convert_multiple_sets): Likewise.
18520         * ira-build.c (create_insn_allocnos): Likewise.
18521         * lra-coalesce.c (merge_pseudos): Likewise.
18522         * lra-constraints.c (match_reload): Likewise.
18523         (simplify_operand_subreg): Likewise.
18524         (curr_insn_transform): Likewise.
18525         * lra-lives.c (process_bb_lives): Likewise.
18526         * lra.c (new_insn_reg): Likewise.
18527         (lra_substitute_pseudo): Likewise.
18528         * regcprop.c (mode_change_ok): Likewise.
18529         (maybe_mode_change): Likewise.
18530         (copyprop_hardreg_forward_1): Likewise.
18531         * reload.c (push_reload): Likewise.
18532         (find_reloads): Likewise.
18533         (find_reloads_subreg_address): Likewise.
18534         * reload1.c (alter_reg): Likewise.
18535         (eliminate_regs_1): Likewise.
18536         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18538 2017-08-30  David Edelsohn  <dje.gcc@gmail.com>
18540         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18541         back to if statements, including unpack.
18543 2017-08-30  Martin Liska  <mliska@suse.cz>
18545         PR inline-asm/82001
18546         * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18547         Rename to ...
18548         (func_checker::compare_asm_inputs_outputs): ... this function.
18549         (func_checker::compare_gimple_asm): Use the function to compare
18550         also ASM constrains.
18551         * ipa-icf-gimple.h: Rename the function.
18553 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18554             Alan Hayward  <alan.hayward@arm.com>
18555             David Sherwood  <david.sherwood@arm.com>
18557         * coretypes.h (complex_mode): New type.
18558         * gdbhooks.py (build_pretty_printer): Handle it.
18559         * machmode.h (complex_mode): New class.
18560         (complex_mode::includes_p): New function.
18561         (is_complex_int_mode): Likewise.
18562         (is_complex_float_mode): Likewise.
18563         * genmodes.c (get_mode_class): Handle complex mode classes.
18564         * function.c (expand_function_end): Use is_complex_int_mode.
18566 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18567             Alan Hayward  <alan.hayward@arm.com>
18568             David Sherwood  <david.sherwood@arm.com>
18570         * coretypes.h (scalar_mode_pod): New typedef.
18571         * gdbhooks.py (build_pretty_printer): Handle it.
18572         * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18573         * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18574         * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18575         * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18576         as_a <scalar_mode>.
18578 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18579             Alan Hayward  <alan.hayward@arm.com>
18580             David Sherwood  <david.sherwood@arm.com>
18582         * machmode.h (mode_for_vector): Take a scalar_mode instead
18583         of a machine_mode.
18584         * stor-layout.c (mode_for_vector): Likewise.
18585         * explow.c (promote_mode): Use as_a <scalar_mode>.
18586         * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
18588 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18589             Alan Hayward  <alan.hayward@arm.com>
18590             David Sherwood  <david.sherwood@arm.com>
18592         * target.def (preferred_simd_mode): Take a scalar_mode
18593         instead of a machine_mode.
18594         * targhooks.h (default_preferred_simd_mode): Likewise.
18595         * targhooks.c (default_preferred_simd_mode): Likewise.
18596         * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18597         * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18598         * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18599         * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18600         * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18601         * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18602         * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18603         * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18604         Likewise.
18605         * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18606         * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18607         * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18608         * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18609         (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18610         * doc/tm.texi: Regenerate.
18611         * optabs-query.c (can_vec_mask_load_store_p): Return false for
18612         non-scalar modes.
18614 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18615             Alan Hayward  <alan.hayward@arm.com>
18616             David Sherwood  <david.sherwood@arm.com>
18618         * target.def (scalar_mode_supported_p): Take a scalar_mode
18619         instead of a machine_mode.
18620         * targhooks.h (default_scalar_mode_supported_p): Likewise.
18621         * targhooks.c (default_scalar_mode_supported_p): Likewise.
18622         * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18623         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18624         * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18625         * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18626         * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18627         * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18628         * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18629         * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18630         * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18631         * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18632         * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18633         * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18634         Likewise.
18635         * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18636         * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18637         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18638         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18639         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18640         Likewise.
18641         * doc/tm.texi: Regenerate.
18643 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18644             Alan Hayward  <alan.hayward@arm.com>
18645             David Sherwood  <david.sherwood@arm.com>
18647         * coretypes.h (opt_scalar_mode): New typedef.
18648         * gdbhooks.py (build_pretty_printers): Handle it.
18649         * machmode.h (mode_iterator::get_2xwider): Add overload for
18650         opt_mode<T>.
18651         * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18652         over scalar modes.
18653         * expr.c (convert_mode_scalar): Likewise.
18654         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18655         * optabs.c (expand_float): Likewise.
18656         (expand_fix): Likewise.
18657         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18659 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18660             Alan Hayward  <alan.hayward@arm.com>
18661             David Sherwood  <david.sherwood@arm.com>
18663         * optabs.c (expand_float): Explicitly check for scalars before
18664         using a branching expansion.
18665         (expand_fix): Likewise.
18667 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18668             Alan Hayward  <alan.hayward@arm.com>
18669             David Sherwood  <david.sherwood@arm.com>
18671         * expr.c (convert_mode): Split scalar handling out into...
18672         (convert_mode_scalar): ...this new function.  Treat the modes
18673         as scalar_modes.
18675 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18676             Alan Hayward  <alan.hayward@arm.com>
18677             David Sherwood  <david.sherwood@arm.com>
18679         * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18680         and scalar_mode.
18681         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18683 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18684             Alan Hayward  <alan.hayward@arm.com>
18685             David Sherwood  <david.sherwood@arm.com>
18687         * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18688         rather than a machine_mode.
18689         (fixed_from_string): Likewise.
18690         (fixed_convert): Likewise.
18691         (fixed_convert_from_int): Likewise.
18692         (fixed_convert_from_real): Likewise.
18693         (real_convert_from_fixed): Likewise.
18694         * fixed-value.c (fixed_from_double_int): Likewise.
18695         (fixed_from_string): Likewise.
18696         (fixed_convert): Likewise.
18697         (fixed_convert_from_int): Likewise.
18698         (fixed_convert_from_real): Likewise.
18699         (real_convert_from_fixed): Likewise.
18700         * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
18702 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18703             Alan Hayward  <alan.hayward@arm.com>
18704             David Sherwood  <david.sherwood@arm.com>
18706         * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18707         of separate mode class checks.  Do not allow vector modes here.
18708         (immed_wide_int_const): Use as_a <scalar_mode>.
18709         * explow.c (trunc_int_for_mode): Likewise.
18710         * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18711         (wi::shwi): Likewise.
18712         (wi::min_value): Likewise.
18713         (wi::max_value): Likewise.
18714         * dwarf2out.c (loc_descriptor): Likewise.
18715         * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18716         for CONST_WIDE_INT.
18718 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18719             Alan Hayward  <alan.hayward@arm.com>
18720             David Sherwood  <david.sherwood@arm.com>
18722         * tree.h (SCALAR_TYPE_MODE): New macro.
18723         * expr.c (expand_expr_addr_expr_1): Use it.
18724         (expand_expr_real_2): Likewise.
18725         * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18726         (fold_convert_const_fixed_from_int): Likewise.
18727         (fold_convert_const_fixed_from_real): Likewise.
18728         (native_encode_fixed): Likewise
18729         (native_encode_complex): Likewise
18730         (native_encode_vector): Likewise.
18731         (native_interpret_fixed): Likewise.
18732         (native_interpret_real): Likewise.
18733         (native_interpret_complex): Likewise.
18734         (native_interpret_vector): Likewise.
18735         * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18736         (simd_clone_adjust_argument_types): Likewise.
18737         (simd_clone_init_simd_arrays): Likewise.
18738         (simd_clone_adjust): Likewise.
18739         * stor-layout.c (layout_type): Likewise.
18740         * tree.c (build_minus_one_cst): Likewise.
18741         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18742         * tree-inline.c (estimate_move_cost): Likewise.
18743         * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18744         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18745         (vectorizable_reduction): Likewise.
18746         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18747         (vect_recog_mixed_size_cond_pattern): Likewise.
18748         (check_bool_pattern): Likewise.
18749         (adjust_bool_pattern): Likewise.
18750         (search_type_for_mask_1): Likewise.
18751         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18752         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18753         (vectorizable_load): Likewise.
18754         (vectorizable_store): Likewise.
18755         * ubsan.c (ubsan_encode_value): Likewise.
18756         * varasm.c (output_constant): Likewise.
18758 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18759             Alan Hayward  <alan.hayward@arm.com>
18760             David Sherwood  <david.sherwood@arm.com>
18762         * coretypes.h (scalar_mode): New class.
18763         * machmode.h (scalar_mode): Likewise.
18764         (scalar_mode::includes_p): New function.
18765         (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18766         * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18767         * genmodes.c (get_mode_class): Handle remaining scalar modes.
18768         * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18769         * expmed.c (store_bit_field_1): Likewise.
18770         (extract_bit_field_1): Likewise.
18771         * expr.c (write_complex_part): Likewise.
18772         (read_complex_part): Likewise.
18773         (emit_move_complex_push): Likewise.
18774         (expand_expr_real_2): Likewise.
18775         * function.c (assign_parm_setup_reg): Likewise.
18776         (assign_parms_unsplit_complex): Likewise.
18777         * optabs.c (expand_binop): Likewise.
18778         * rtlanal.c (subreg_get_info): Likewise.
18779         * simplify-rtx.c (simplify_immed_subreg): Likewise.
18780         * varasm.c (output_constant_pool_2): Likewise.
18782 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18783             Alan Hayward  <alan.hayward@arm.com>
18784             David Sherwood  <david.sherwood@arm.com>
18786         * expmed.c (extract_high_half): Use scalar_int_mode and remove
18787         assertion.
18788         (expmed_mult_highpart_optab): Likewise.
18789         (expmed_mult_highpart): Likewise.
18791 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18792             Alan Hayward  <alan.hayward@arm.com>
18793             David Sherwood  <david.sherwood@arm.com>
18795         * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18796         instead of a machine_mode.
18797         (builtin_memset_read_str): Likewise.
18798         * builtins.c (c_readstr): Likewise.
18799         (builtin_memcpy_read_str): Likewise.
18800         (builtin_strncpy_read_str): Likewise.
18801         (builtin_memset_read_str): Likewise.
18802         (builtin_memset_gen_str): Likewise.
18803         (expand_builtin_signbit): Use scalar_int_mode for local variables.
18804         * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18805         instead of a machine_mode.
18806         * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18807         variables.
18808         (make_extraction): Likewise.
18809         (try_widen_shift_mode): Take and return scalar_int_modes instead
18810         of machine_modes.
18811         * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18812         a scalar_int_mode instead of a machine_mode.
18813         * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18814         (avr_addr_space_pointer_mode): Likewise.
18815         * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18816         * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18817         (msp430_unwind_word_mode): Likewise.
18818         * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18819         (spu_addr_space_pointer_mode): Likewise.
18820         (spu_addr_space_address_mode): Likewise.
18821         (spu_libgcc_cmp_return_mode): Likewise.
18822         (spu_libgcc_shift_count_mode): Likewise.
18823         * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18824         (rl78_addr_space_pointer_mode): Likewise.
18825         (fl78_unwind_word_mode): Likewise.
18826         (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18827         machine_mode.
18828         * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18829         * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18830         * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18831         (mips_valid_pointer_mode): Likewise.
18832         * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18833         * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18834         (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18835         of a machine_mode.
18836         (ft32_addr_space_address_mode): Likewise.
18837         * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18838         scalar_int_mode instead of a machine_mode.
18839         (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18840         of a machine_mode.
18841         (m32c_addr_space_address_mode): Likewise.
18842         * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18843         (rs6000_eh_return_filter_mode): Likewise.
18844         * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18845         (rs6000_eh_return_filter_mode): Likewise.
18846         * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18847         (s390_libgcc_shift_count_mode): Likewise.
18848         (s390_unwind_word_mode): Likewise.
18849         (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18850         machine_mode.
18851         * target.def (mode_rep_extended): Likewise.
18852         (valid_pointer_mode): Likewise.
18853         (addr_space.valid_pointer_mode): Likewise.
18854         (eh_return_filter_mode): Return a scalar_int_mode rather than
18855         a machine_mode.
18856         (libgcc_cmp_return_mode): Likewise.
18857         (libgcc_shift_count_mode): Likewise.
18858         (unwind_word_mode): Likewise.
18859         (addr_space.pointer_mode): Likewise.
18860         (addr_space.address_mode): Likewise.
18861         * doc/tm.texi: Regenerate.
18862         * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18863         a machine_mode.
18864         (do_jump): Use scalar_int_mode for local variables.
18865         * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18866         rather than a machine_mode.
18867         * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18868         (scompare_loc_descriptor_wide): Likewise.
18869         (scompare_loc_descriptor_narrow): Likewise.
18870         * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18871         variables.
18872         * except.c (sjlj_emit_dispatch_table): Likewise.
18873         (expand_builtin_eh_copy_values): Likewise.
18874         * explow.c (convert_memory_address_addr_space_1): Likewise.
18875         Take a scalar_int_mode rather than a machine_mode.
18876         (convert_memory_address_addr_space): Take a scalar_int_mode rather
18877         than a machine_mode.
18878         (memory_address_addr_space): Use scalar_int_mode for local variables.
18879         * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18880         rather than a machine_mode.
18881         * expmed.c (mask_rtx): Likewise.
18882         (init_expmed_one_conv): Likewise.
18883         (expand_mult_highpart_adjust): Likewise.
18884         (extract_high_half): Likewise.
18885         (expmed_mult_highpart_optab): Likewise.
18886         (expmed_mult_highpart): Likewise.
18887         (expand_smod_pow2): Likewise.
18888         (expand_sdiv_pow2): Likewise.
18889         (emit_store_flag_int): Likewise.
18890         (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18891         variables.
18892         (extract_low_bits): Likewise.
18893         * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18894         a machine_mode.
18895         * expr.c (pieces_addr::adjust):  Likewise.
18896         (can_store_by_pieces): Likewise.
18897         (store_by_pieces): Likewise.
18898         (clear_by_pieces_1): Likewise.
18899         (expand_expr_addr_expr_1): Likewise.
18900         (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18901         (expand_expr_real_1): Likewise.
18902         (try_casesi): Likewise.
18903         * final.c (shorten_branches): Likewise.
18904         * fold-const.c (fold_convert_const_int_from_fixed): Change the
18905         type of "mode" to machine_mode.
18906         * internal-fn.c (expand_arith_overflow_result_store): Take a
18907         scalar_int_mode rather than a machine_mode.
18908         (expand_mul_overflow): Use scalar_int_mode for local variables.
18909         * loop-doloop.c (doloop_modify): Likewise.
18910         (doloop_optimize): Likewise.
18911         * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18912         than a machine_mode.
18913         (expand_doubleword_shift_condmove): Likewise.
18914         (expand_doubleword_shift): Likewise.
18915         (expand_doubleword_clz): Likewise.
18916         (expand_doubleword_popcount): Likewise.
18917         (expand_doubleword_parity): Likewise.
18918         (expand_absneg_bit): Use scalar_int_mode for local variables.
18919         (prepare_float_lib_cmp): Likewise.
18920         * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
18921         rather than a machine_mode.
18922         (convert_memory_address_addr_space): Likewise.
18923         (get_mode_bounds): Likewise.
18924         (get_address_mode): Return a scalar_int_mode rather than a
18925         machine_mode.
18926         * rtlanal.c (get_address_mode): Likewise.
18927         * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
18928         than a machine_mode.
18929         * targhooks.c (default_mode_rep_extended): Likewise.
18930         (default_valid_pointer_mode): Likewise.
18931         (default_addr_space_valid_pointer_mode): Likewise.
18932         (default_eh_return_filter_mode): Return a scalar_int_mode rather
18933         than a machine_mode.
18934         (default_libgcc_cmp_return_mode): Likewise.
18935         (default_libgcc_shift_count_mode): Likewise.
18936         (default_unwind_word_mode): Likewise.
18937         (default_addr_space_pointer_mode): Likewise.
18938         (default_addr_space_address_mode): Likewise.
18939         * targhooks.h (default_eh_return_filter_mode): Likewise.
18940         (default_libgcc_cmp_return_mode): Likewise.
18941         (default_libgcc_shift_count_mode): Likewise.
18942         (default_unwind_word_mode): Likewise.
18943         (default_addr_space_pointer_mode): Likewise.
18944         (default_addr_space_address_mode): Likewise.
18945         (default_mode_rep_extended): Take a scalar_int_mode rather than
18946         a machine_mode.
18947         (default_valid_pointer_mode): Likewise.
18948         (default_addr_space_valid_pointer_mode): Likewise.
18949         * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
18950         local variables.
18951         * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
18952         rather than a machine_mode.
18953         * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
18954         for local variables.
18955         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
18956         * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
18957         than a machine_mode.
18959 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18960             Alan Hayward  <alan.hayward@arm.com>
18961             David Sherwood  <david.sherwood@arm.com>
18963         * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
18964         the mode argument to scalar_int_mode.
18965         (do_jump_by_parts_zero_rtx): Likewise.
18966         (do_jump_by_parts_equality_rtx): Likewise.
18967         (do_jump_by_parts_greater): Take a mode argument.
18968         (do_jump_by_parts_equality): Likewise.
18969         (do_jump_1): Update calls accordingly.
18971 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18972             Alan Hayward  <alan.hayward@arm.com>
18973             David Sherwood  <david.sherwood@arm.com>
18975         * is-a.h (safe_dyn_cast): New function.
18976         * rtl.h (rtx_jump_table_data::get_data_mode): New function.
18977         (jump_table_for_label): Likewise.
18978         * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
18979         instead of an rtx_insn *.
18980         (shorten_branches): Use dyn_cast instead of LABEL_P and
18981         JUMP_TABLE_DATA_P.  Use jump_table_for_label and
18982         rtx_jump_table_data::get_data_mode.
18983         (final_scan_insn): Likewise.
18985 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18986             Alan Hayward  <alan.hayward@arm.com>
18987             David Sherwood  <david.sherwood@arm.com>
18989         * combine.c (try_combine): Use is_a <scalar_int_mode> when
18990         trying to combine a full-register integer set with a subreg
18991         integer set.
18993 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18994             Alan Hayward  <alan.hayward@arm.com>
18995             David Sherwood  <david.sherwood@arm.com>
18997         * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
18998         that is always either address_mode or pointer_mode.
19000 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19001             Alan Hayward  <alan.hayward@arm.com>
19002             David Sherwood  <david.sherwood@arm.com>
19004         * expr.c (expand_expr_real_2): Use word_mode instead of innermode
19005         when the two are known to be equal.
19007 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19008             Alan Hayward  <alan.hayward@arm.com>
19009             David Sherwood  <david.sherwood@arm.com>
19011         * simplify-rtx.c (simplify_const_unary_operation): Use
19012         is_a <scalar_int_mode> instead of checking for a nonzero
19013         precision.  Forcibly convert op_mode to a scalar_int_mode
19014         in that case.  More clearly differentiate the operand and
19015         result modes and use the former when deciding what the value
19016         of a count-bits operation should be.  Use is_int_mode instead
19017         of checking for a MODE_INT.  Remove redundant check for whether
19018         this mode has a zero precision.
19020 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19021             Alan Hayward  <alan.hayward@arm.com>
19022             David Sherwood  <david.sherwood@arm.com>
19024         * optabs.c (widen_leading): Change the type of the mode argument
19025         to scalar_int_mode.  Use opt_scalar_int_mode for the mode iterator.
19026         (widen_bswap): Likewise.
19027         (expand_parity): Likewise.
19028         (expand_ctz): Change the type of the mode argument to scalar_int_mode.
19029         (expand_ffs): Likewise.
19030         (epand_unop): Check for scalar integer modes before calling the
19031         above routines.
19033 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19034             Alan Hayward  <alan.hayward@arm.com>
19035             David Sherwood  <david.sherwood@arm.com>
19037         * expr.c (const_scalar_mask_from_tree): Add a mode argument.
19038         Expand commentary.
19039         (expand_expr_real_1): Update call accordingly.
19041 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19042             Alan Hayward  <alan.hayward@arm.com>
19043             David Sherwood  <david.sherwood@arm.com>
19045         * expmed.c (store_bit_field_using_insv): Add op0_mode and
19046         value_mode arguments.  Use scalar_int_mode internally.
19047         (store_bit_field_1): Rename the new integer mode from imode
19048         to op0_mode and use it instead of GET_MODE (op0).  Update calls
19049         to store_split_bit_field, store_bit_field_using_insv and
19050         store_fixed_bit_field.
19051         (store_fixed_bit_field): Add op0_mode and value_mode arguments.
19052         Use scalar_int_mode internally.  Use a bit count rather than a mode
19053         when calculating the largest bit size for get_best_mode.
19054         Update calls to store_split_bit_field and store_fixed_bit_field_1.
19055         (store_fixed_bit_field_1): Add mode and value_mode arguments.
19056         Remove assertion that OP0 has a scalar integer mode.
19057         (store_split_bit_field): Add op0_mode and value_mode arguments.
19058         Update calls to extract_fixed_bit_field.
19059         (extract_bit_field_using_extv): Add an op0_mode argument.
19060         Use scalar_int_mode internally.
19061         (extract_bit_field_1): Rename the new integer mode from imode to
19062         op0_mode and use it instead of GET_MODE (op0).  Update calls to
19063         extract_split_bit_field, extract_bit_field_using_extv and
19064         extract_fixed_bit_field.
19065         (extract_fixed_bit_field): Add an op0_mode argument.  Update calls
19066         to extract_split_bit_field and extract_fixed_bit_field_1.
19067         (extract_fixed_bit_field_1): Add a mode argument.  Remove assertion
19068         that OP0 has a scalar integer mode.  Use as_a <scalar_int_mode>
19069         on the target mode.
19070         (extract_split_bit_field): Add an op0_mode argument.  Update call
19071         to extract_fixed_bit_field.
19073 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19074             Alan Hayward  <alan.hayward@arm.com>
19075             David Sherwood  <david.sherwood@arm.com>
19077         * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
19078         * explow.c (hard_function_value): Likewise.
19079         * expmed.c (extract_fixed_bit_field_1): Likewise.  Move the
19080         convert_to_mode call outside the loop.
19081         * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
19082         for the mode iterator.  Require the mode specified by max_pieces
19083         to exist.
19084         (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
19085         mode iterator.
19086         (copy_blkmode_to_reg): Likewise.
19087         (set_storage_via_setmem): Likewise.
19088         * optabs.c (prepare_cmp_insn): Likewise.
19089         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19090         * stor-layout.c (finish_bitfield_representative): Likewise.
19092 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19093             Alan Hayward  <alan.hayward@arm.com>
19094             David Sherwood  <david.sherwood@arm.com>
19096         * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
19097         * expr.c (convert_move): Use them.
19098         (convert_modes): Likewise.
19099         (store_expr_with_bounds): Likewise.
19101 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19102             Alan Hayward  <alan.hayward@arm.com>
19103             David Sherwood  <david.sherwood@arm.com>
19105         * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
19106         parameter for the mode of "x".  Remove the "known_x", "known_mode"
19107         and "known_ret" arguments.  Change the type of the mode argument
19108         to scalar_int_mode.
19109         (rtl_hooks:reg_num_sign_bit_copies): Likewise.
19110         * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
19111         (reg_num_sign_bit_copies_for_combine): Likewise.
19112         * rtlanal.c (nonzero_bits1): Likewise.
19113         (num_sign_bit_copies1): Likewise.
19114         * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
19115         (reg_num_sign_bit_copies_general): Likewise.
19116         * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
19117         (reg_nonzero_bits_general): Likewise.
19119 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19120             Alan Hayward  <alan.hayward@arm.com>
19121             David Sherwood  <david.sherwood@arm.com>
19123         * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
19124         than in subroutines.  Return 1 for non-integer modes.
19125         (cached_num_sign_bit_copies): Change the type of the mode parameter
19126         to scalar_int_mode.
19127         (num_sign_bit_copies1): Likewise.  Remove early exit for other mode
19128         classes.  Handle CONST_INT_P first and then check whether X also
19129         has a scalar integer mode.  Check the same thing for inner registers
19130         of a SUBREG and for values that are being extended or truncated.
19132 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19133             Alan Hayward  <alan.hayward@arm.com>
19134             David Sherwood  <david.sherwood@arm.com>
19136         * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
19137         in subroutines.  Return the mode mask for non-integer modes.
19138         (cached_nonzero_bits): Change the type of the mode parameter
19139         to scalar_int_mode.
19140         (nonzero_bits1): Likewise.  Remove early exit for other mode
19141         classes.  Handle CONST_INT_P first and then check whether X
19142         also has a scalar integer mode.
19144 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19145             Alan Hayward  <alan.hayward@arm.com>
19146             David Sherwood  <david.sherwood@arm.com>
19148         * expr.c (widest_int_mode_for_size): Make the comment match the code.
19149         Return a scalar_int_mode and assert that the size is greater than
19150         one byte.
19151         (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
19152         (op_by_pieces_d::op_by_pieces_d): Likewise.
19153         (op_by_pieces_d::run): Likewise.
19154         (can_store_by_pieces): Likewise.
19156 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19157             Alan Hayward  <alan.hayward@arm.com>
19158             David Sherwood  <david.sherwood@arm.com>
19160         * combine.c (extract_left_shift): Add a mode argument and update
19161         recursive calls.
19162         (make_compound_operation_int): Change the type of the mode parameter
19163         to scalar_int_mode and update the call to extract_left_shift.
19165 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19166             Alan Hayward  <alan.hayward@arm.com>
19167             David Sherwood  <david.sherwood@arm.com>
19169         * combine.c (simplify_and_const_int): Change the type of the mode
19170         parameter to scalar_int_mode.
19171         (simplify_and_const_int_1): Likewise.  Update recursive call.
19173 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19174             Alan Hayward  <alan.hayward@arm.com>
19175             David Sherwood  <david.sherwood@arm.com>
19177         * combine.c (simplify_compare_const): Check that the mode is a
19178         scalar_int_mode (rather than VOIDmode) before testing its
19179         precision.
19180         (simplify_comparison): Move COMPARISON_P handling out of the
19181         loop and restrict the latter part of the loop to scalar_int_modes.
19182         Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
19183         and when considering SUBREG_REGs.  Use is_int_mode instead of
19184         checking GET_MODE_CLASS against MODE_INT.
19186 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19187             Alan Hayward  <alan.hayward@arm.com>
19188             David Sherwood  <david.sherwood@arm.com>
19190         * combine.c (try_widen_shift_mode): Move check for equal modes to...
19191         (simplify_shift_const_1): ...here.  Use scalar_int_mode for
19192         shift_unit_mode and for modes involved in scalar shifts.
19194 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19195             Alan Hayward  <alan.hayward@arm.com>
19196             David Sherwood  <david.sherwood@arm.com>
19198         * combine.c (force_int_to_mode): New function, split out from...
19199         (force_to_mode): ...here.  Keep xmode up-to-date and use it
19200         instead of GET_MODE (x).
19202 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19203             Alan Hayward  <alan.hayward@arm.com>
19204             David Sherwood  <david.sherwood@arm.com>
19206         * optabs-query.h (extraction_insn::struct_mode): Change type to
19207         opt_scalar_int_mode and update comment.
19208         (extraction_insn::field_mode): Change type to scalar_int_mode.
19209         (extraction_insn::pos_mode): Likewise.
19210         * combine.c (make_extraction): Update accordingly.
19211         * optabs-query.c (get_traditional_extraction_insn): Likewise.
19212         (get_optab_extraction_insn): Likewise.
19213         * recog.c (simplify_while_replacing): Likewise.
19214         * expmed.c (narrow_bit_field_mem): Change the type of the mode
19215         parameter to opt_scalar_int_mode.
19217 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19218             Alan Hayward  <alan.hayward@arm.com>
19219             David Sherwood  <david.sherwood@arm.com>
19221         * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
19222         to a scalar_int_mode instead of a machine_mode.
19223         (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
19224         (get_best_mode): Return a boolean and use a pointer argument to store
19225         the selected mode.  Replace the limit mode parameter with a bit limit.
19226         * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
19227         for the values returned by bit_field_mode_iterator::next_mode.
19228         (store_bit_field): Update call to get_best_mode.
19229         (store_fixed_bit_field): Likewise.
19230         (extract_fixed_bit_field): Likewise.
19231         * expr.c (optimize_bitfield_assignment_op): Likewise.
19232         * fold-const.c (optimize_bit_field_compare): Likewise.
19233         (fold_truth_andor_1): Likewise.
19234         * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
19235         Update for new type of m_mode.
19236         (get_best_mode): As above.
19238 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19239             Alan Hayward  <alan.hayward@arm.com>
19240             David Sherwood  <david.sherwood@arm.com>
19242         * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
19243         to scalar_int_mode.  Remove check for SCALAR_INT_MODE_P.
19244         (store_bit_field): Check is_a <scalar_int_mode> before calling
19245         strict_volatile_bitfield_p.
19246         (extract_bit_field): Likewise.
19248 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19249             Alan Hayward  <alan.hayward@arm.com>
19250             David Sherwood  <david.sherwood@arm.com>
19252         * target.def (cstore_mode): Return a scalar_int_mode.
19253         * doc/tm.texi: Regenerate.
19254         * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19255         * targhooks.h (default_cstore_mode): Likewise.
19256         * targhooks.c (default_cstore_mode): Likewise, using a forced
19257         conversion.
19258         * expmed.c (emit_cstore): Expect the target of the cstore to be
19259         a scalar_int_mode.
19261 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19262             Alan Hayward  <alan.hayward@arm.com>
19263             David Sherwood  <david.sherwood@arm.com>
19265         * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19266         scalar_int_mode.
19267         (niter_desc): Likewise mode.
19268         (iv_analyze): Add a mode parameter.
19269         (biv_p): Likewise.
19270         (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19271         and change its type to scalar_int_mode.
19272         * loop-iv.c: Update commentary at head of file.
19273         (iv_constant): Pass the mode paraeter before the rtx it describes
19274         and change its type to scalar_int_mode.  Remove VOIDmode handling.
19275         (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19276         (iv_extend): Likewise.
19277         (shorten_into_mode): Likewise.
19278         (iv_add): Use scalar_int_mode.
19279         (iv_mult): Likewise.
19280         (iv_shift): Likewise.
19281         (canonicalize_iv_subregs): Likewise.
19282         (get_biv_step_1): Pass the outer_mode parameter before the rtx
19283         it describes and change its mode to scalar_int_mode.   Also change
19284         the type of the returned inner_mode to scalar_int_mode.
19285         (get_biv_step): Likewise, turning outer_mode from a pointer
19286         into a direct parameter.  Update call to get_biv_step_1.
19287         (iv_analyze_biv): Add an outer_mode parameter.  Update calls to
19288         iv_constant and get_biv_step.
19289         (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19290         and change its type to scalar_int_mode.  Don't initialise iv->mode
19291         to VOIDmode and remove later checks for its still being VOIDmode.
19292         Update calls to iv_analyze_op and iv_analyze_expr.  Check
19293         is_a <scalar_int_mode> when changing the mode under consideration.
19294         (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19295         Update call to iv_analyze_expr.
19296         (iv_analyze_op): Add a mode parameter.  Reject subregs whose
19297         inner register is not also a scalar_int_mode.  Update call to
19298         iv_analyze_biv.
19299         (iv_analyze): Add a mode parameter.  Update call to iv_analyze_op.
19300         (biv_p): Add a mode parameter.  Update call to iv_analyze_biv.
19301         (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19302         separate mode class checks.  Update calls to iv_analyze.  Remove
19303         fix-up of VOIDmodes after iv_analyze_biv.
19304         * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19305         don't have a scalar_int_mode.  Update call to biv_p.
19307 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19308             Alan Hayward  <alan.hayward@arm.com>
19309             David Sherwood  <david.sherwood@arm.com>
19311         * cfgexpand.c (convert_debug_memory_address): Use
19312         as_a <scalar_int_mode>.
19313         * combine.c (expand_compound_operation): Likewise.
19314         (make_extraction): Likewise.
19315         (change_zero_ext): Likewise.
19316         (simplify_comparison): Likewise.
19317         * cse.c (cse_insn): Likewise.
19318         * dwarf2out.c (minmax_loc_descriptor): Likewise.
19319         (mem_loc_descriptor): Likewise.
19320         (loc_descriptor): Likewise.
19321         * expmed.c (init_expmed_one_mode): Likewise.
19322         (synth_mult): Likewise.
19323         (emit_store_flag_1): Likewise.
19324         (expand_divmod): Likewise.  Use HWI_COMPUTABLE_MODE_P instead
19325         of a comparison with size.
19326         * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19327         (reduce_to_bit_field_precision): Likewise.
19328         * function.c (expand_function_end): Likewise.
19329         * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19330         * loop-doloop.c (doloop_modify): Likewise.
19331         * optabs.c (expand_binop): Likewise.
19332         (expand_unop): Likewise.
19333         (expand_copysign_absneg): Likewise.
19334         (prepare_cmp_insn): Likewise.
19335         (maybe_legitimize_operand): Likewise.
19336         * recog.c (const_scalar_int_operand): Likewise.
19337         * rtlanal.c (get_address_mode): Likewise.
19338         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19339         (simplify_cond_clz_ctz): Likewise.
19340         * tree-nested.c (get_nl_goto_field): Likewise.
19341         * tree.c (build_vector_type_for_mode): Likewise.
19342         * var-tracking.c (use_narrower_mode): Likewise.
19344 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19345             Alan Hayward  <alan.hayward@arm.com>
19346             David Sherwood  <david.sherwood@arm.com>
19348         * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19349         * builtins.c (expand_builtin_signbit): Use it.
19350         * cfgexpand.c (expand_debug_expr): Likewise.
19351         * dojump.c (do_jump): Likewise.
19352         (do_compare_and_jump): Likewise.
19353         * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19354         * expmed.c (make_tree): Likewise.
19355         * expr.c (expand_expr_real_2): Likewise.
19356         (expand_expr_real_1): Likewise.
19357         (try_casesi): Likewise.
19358         * fold-const-call.c (fold_const_call_ss): Likewise.
19359         * fold-const.c (unextend): Likewise.
19360         (extract_muldiv_1): Likewise.
19361         (fold_single_bit_test): Likewise.
19362         (native_encode_int): Likewise.
19363         (native_encode_string): Likewise.
19364         (native_interpret_int): Likewise.
19365         * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19366         * internal-fn.c (expand_addsub_overflow): Likewise.
19367         (expand_neg_overflow): Likewise.
19368         (expand_mul_overflow): Likewise.
19369         (expand_arith_overflow): Likewise.
19370         * match.pd: Likewise.
19371         * stor-layout.c (layout_type): Likewise.
19372         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19373         * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19374         * tree-ssanames.c (get_range_info): Likewise.
19375         * tree-switch-conversion.c (array_value_type) Likewise.
19376         * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19377         (vect_recog_divmod_pattern): Likewise.
19378         (vect_recog_mixed_size_cond_pattern): Likewise.
19379         * tree-vrp.c (extract_range_basic): Likewise.
19380         (simplify_float_conversion_using_ranges): Likewise.
19381         * tree.c (int_fits_type_p): Likewise.
19382         * ubsan.c (instrument_bool_enum_load): Likewise.
19383         * varasm.c (mergeable_string_section): Likewise.
19384         (narrowing_initializer_constant_valid_p): Likewise.
19385         (output_constant): Likewise.
19387 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19388             Alan Hayward  <alan.hayward@arm.com>
19389             David Sherwood  <david.sherwood@arm.com>
19391         * machmode.h (NARROWEST_INT_MODE): New macro.
19392         * expr.c (alignment_for_piecewise_move): Use it instead of
19393         GET_CLASS_NARROWEST_MODE (MODE_INT).
19394         (push_block): Likewise.
19395         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19396         Likewise.
19397         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19399 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19400             Alan Hayward  <alan.hayward@arm.com>
19401             David Sherwood  <david.sherwood@arm.com>
19403         * postreload.c (move2add_valid_value_p): Change the type of the
19404         mode parameter to scalar_int_mode.
19405         (move2add_use_add2_insn): Add a mode parameter and use it instead
19406         of GET_MODE (reg).
19407         (move2add_use_add3_insn): Likewise.
19408         (reload_cse_move2add): Update accordingly.
19410 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19411             Alan Hayward  <alan.hayward@arm.com>
19412             David Sherwood  <david.sherwood@arm.com>
19414         * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19415         double-word mode.
19416         * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19417         * optabs.c (expand_unop): Likewise.
19419 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19420             Alan Hayward  <alan.hayward@arm.com>
19421             David Sherwood  <david.sherwood@arm.com>
19423         * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19424         (clz_loc_descriptor): Likewise.  Remove SCALAR_INT_MODE_P check.
19425         (popcount_loc_descriptor): Likewise.
19426         (bswap_loc_descriptor): Likewise.
19427         (rotate_loc_descriptor): Likewise.
19428         (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19429         calling the functions above.
19431 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19432             Alan Hayward  <alan.hayward@arm.com>
19433             David Sherwood  <david.sherwood@arm.com>
19435         * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19436         checks.
19437         (try_combine): Likewise.
19438         (simplify_if_then_else): Likewise.
19439         * cse.c (cse_insn): Likewise.
19440         * dwarf2out.c (mem_loc_descriptor): Likewise.
19441         * emit-rtl.c (gen_lowpart_common): Likewise.
19442         * simplify-rtx.c (simplify_truncation): Likewise.
19443         (simplify_binary_operation_1): Likewise.
19444         (simplify_const_relational_operation): Likewise.
19445         (simplify_ternary_operation): Likewise.
19446         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
19448 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19449             Alan Hayward  <alan.hayward@arm.com>
19450             David Sherwood  <david.sherwood@arm.com>
19452         * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19453         * reload.c (push_reload): Likewise.
19454         (find_reloads): Likewise.
19456 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19457             Alan Hayward  <alan.hayward@arm.com>
19458             David Sherwood  <david.sherwood@arm.com>
19460         * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19461         (make_compound_operation_int): Likewise.
19462         (change_zero_ext): Likewise.
19463         * expr.c (convert_move): Likewise.
19464         (convert_modes): Likewise.
19465         * fwprop.c (forward_propagate_subreg): Likewise.
19466         * loop-iv.c (get_biv_step_1): Likewise.
19467         * optabs.c (widen_operand): Likewise.
19468         * postreload.c (move2add_valid_value_p): Likewise.
19469         * recog.c (simplify_while_replacing): Likewise.
19470         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19471         (simplify_binary_operation_1): Likewise.  Remove redundant
19472         mode equality check.
19474 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19475             Alan Hayward  <alan.hayward@arm.com>
19476             David Sherwood  <david.sherwood@arm.com>
19478         * combine.c (combine_simplify_rtx): Add checks for
19479         is_a <scalar_int_mode>.
19480         (simplify_if_then_else): Likewise.
19481         (make_field_assignment): Likewise.
19482         (simplify_comparison): Likewise.
19483         * ifcvt.c (noce_try_bitop): Likewise.
19484         * loop-invariant.c (canonicalize_address_mult): Likewise.
19485         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19487 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19488             Alan Hayward  <alan.hayward@arm.com>
19489             David Sherwood  <david.sherwood@arm.com>
19491         * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19492         is_a <scalar_int_mode> instead of != BLKmode.
19494 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19495             Alan Hayward  <alan.hayward@arm.com>
19496             David Sherwood  <david.sherwood@arm.com>
19498         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19499         instead of != VOIDmode.
19500         * combine.c (if_then_else_cond): Likewise.
19501         (change_zero_ext): Likewise.
19502         * dwarf2out.c (mem_loc_descriptor): Likewise.
19503         (loc_descriptor): Likewise.
19504         * rtlanal.c (canonicalize_condition): Likewise.
19505         * simplify-rtx.c (simplify_relational_operation_1): Likewise.
19507 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19508             Alan Hayward  <alan.hayward@arm.com>
19509             David Sherwood  <david.sherwood@arm.com>
19511         * simplify-rtx.c (simplify_binary_operation_1): Use
19512         is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
19514 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19515             Alan Hayward  <alan.hayward@arm.com>
19516             David Sherwood  <david.sherwood@arm.com>
19518         * wide-int.h (int_traits<unsigned char>) New class.
19519         (int_traits<unsigned short>) Likewise.
19520         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19521         Use GET_MODE_UNIT_PRECISION and remove redundant test for
19522         SCALAR_INT_MODE_P.
19523         * combine.c (set_nonzero_bits_and_sign_copies): Use
19524         is_a <scalar_int_mode>.
19525         (find_split_point): Likewise.
19526         (combine_simplify_rtx): Likewise.
19527         (simplify_logical): Likewise.
19528         (expand_compound_operation): Likewise.
19529         (expand_field_assignment): Likewise.
19530         (make_compound_operation): Likewise.
19531         (extended_count): Likewise.
19532         (change_zero_ext): Likewise.
19533         (simplify_comparison): Likewise.
19534         * dwarf2out.c (scompare_loc_descriptor): Likewise.
19535         (ucompare_loc_descriptor): Likewise.
19536         (minmax_loc_descriptor): Likewise.
19537         (mem_loc_descriptor): Likewise.
19538         (loc_descriptor): Likewise.
19539         * expmed.c (init_expmed_one_mode): Likewise.
19540         * lra-constraints.c (lra_constraint_offset): Likewise.
19541         * optabs.c (prepare_libcall_arg): Likewise.
19542         * postreload.c (move2add_note_store): Likewise.
19543         * reload.c (operands_match_p): Likewise.
19544         * rtl.h (load_extend_op): Likewise.
19545         * rtlhooks.c (gen_lowpart_general): Likewise.
19546         * simplify-rtx.c (simplify_truncation): Likewise.
19547         (simplify_unary_operation_1): Likewise.
19548         (simplify_binary_operation_1): Likewise.
19549         (simplify_const_binary_operation): Likewise.
19550         (simplify_const_relational_operation): Likewise.
19551         (simplify_subreg): Likewise.
19552         * stor-layout.c (bitwise_mode_for_mode): Likewise.
19553         * var-tracking.c (adjust_mems): Likewise.
19554         (prepare_call_arguments): Likewise.
19556 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19557             Alan Hayward  <alan.hayward@arm.com>
19558             David Sherwood  <david.sherwood@arm.com>
19560         * machmode.h (is_int_mode): New fuction.
19561         * combine.c (find_split_point): Use it.
19562         (combine_simplify_rtx): Likewise.
19563         (simplify_if_then_else): Likewise.
19564         (simplify_set): Likewise.
19565         (simplify_shift_const_1): Likewise.
19566         (simplify_comparison): Likewise.
19567         * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19568         * cse.c (notreg_cost): Likewise.
19569         (cse_insn): Likewise.
19570         * cselib.c (cselib_lookup_1): Likewise.
19571         * dojump.c (do_jump_1): Likewise.
19572         (do_compare_rtx_and_jump): Likewise.
19573         * dse.c (get_call_args): Likewise.
19574         * dwarf2out.c (rtl_for_decl_init): Likewise.
19575         (native_encode_initializer): Likewise.
19576         * expmed.c (emit_store_flag_1): Likewise.
19577         (emit_store_flag): Likewise.
19578         * expr.c (convert_modes): Likewise.
19579         (store_field): Likewise.
19580         (expand_expr_real_1): Likewise.
19581         * fold-const.c (fold_read_from_constant_string): Likewise.
19582         * gimple-ssa-sprintf.c (get_format_string): Likewise.
19583         * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19584         * optabs.c (expand_binop): Likewise.
19585         (expand_unop): Likewise.
19586         (expand_abs_nojump): Likewise.
19587         (expand_one_cmpl_abs_nojump): Likewise.
19588         * simplify-rtx.c (mode_signbit_p): Likewise.
19589         (val_signbit_p): Likewise.
19590         (val_signbit_known_set_p): Likewise.
19591         (val_signbit_known_clear_p): Likewise.
19592         (simplify_relational_operation_1): Likewise.
19593         * tree.c (vector_type_mode): Likewise.
19595 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19596             Alan Hayward  <alan.hayward@arm.com>
19597             David Sherwood  <david.sherwood@arm.com>
19599         * machmode.h (smallest_mode_for_size): Fix formatting.
19600         (smallest_int_mode_for_size): New function.
19601         * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19602         instead of smallest_mode_for_size.
19603         * combine.c (make_extraction): Likewise.
19604         * config/arc/arc.c (arc_expand_movmem): Likewise.
19605         * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19606         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19607         * config/s390/s390.c (s390_expand_insv): Likewise.
19608         * config/sparc/sparc.c (assign_int_registers): Likewise.
19609         * config/spu/spu.c (spu_function_value): Likewise.
19610         (spu_function_arg): Likewise.
19611         * coverage.c (get_gcov_type): Likewise.
19612         (get_gcov_unsigned_t): Likewise.
19613         * dse.c (find_shift_sequence): Likewise.
19614         * expmed.c (store_bit_field_1): Likewise.
19615         * expr.c (convert_move): Likewise.
19616         (store_field): Likewise.
19617         * internal-fn.c (expand_arith_overflow): Likewise.
19618         * optabs-query.c (get_best_extraction_insn): Likewise.
19619         * optabs.c (expand_twoval_binop_libfunc): Likewise.
19620         * stor-layout.c (layout_type): Likewise.
19621         (initialize_sizetypes): Likewise.
19622         * targhooks.c (default_get_mask_mode): Likewise.
19623         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19625 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19626             Alan Hayward  <alan.hayward@arm.com>
19627             David Sherwood  <david.sherwood@arm.com>
19629         * machmode.h (opt_mode::else_blk): New function.
19630         (int_mode_for_mode): Declare.
19631         * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19632         * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19633         return type.
19634         * cfgexpand.c (expand_debug_expr): Likewise.
19635         * combine.c (gen_lowpart_or_truncate): Likewise.
19636         (gen_lowpart_for_combine): Likewise.
19637         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19638         * config/avr/avr.c (avr_to_int_mode): Likewise.
19639         (avr_out_plus_1): Likewise.
19640         (avr_out_plus): Likewise.
19641         (avr_out_round): Likewise.
19642         * config/i386/i386.c (ix86_split_to_parts): Likewise.
19643         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19644         (s390_expand_vcond): Likewise.
19645         * config/spu/spu.c (spu_split_immediate): Likewise.
19646         (spu_expand_mov): Likewise.
19647         * dse.c (get_stored_val): Likewise.
19648         * expmed.c (store_bit_field_1): Likewise.
19649         (convert_extracted_bit_field): Use int_mode_for_mode instead of
19650         int_mode_for_size.
19651         (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19652         (extract_low_bits): Likewise.
19653         * expr.c (emit_group_load_1): Likewise.  Separate out the BLKmode
19654         handling rather than repeating the check.
19655         (emit_group_store): Likewise.
19656         (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19657         * optabs.c (expand_absneg_bit): Likewise.
19658         (expand_copysign_absneg): Likewise.
19659         (expand_copysign_bit): Likewise.
19660         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19661         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19662         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19663         * var-tracking.c (prepare_call_arguments):  Likewise.
19664         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19665         int_mode_for_mode instead of mode_for_size.
19666         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19668 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19669             Alan Hayward  <alan.hayward@arm.com>
19670             David Sherwood  <david.sherwood@arm.com>
19672         * machmode.h (int_mode_for_size): New function.
19673         * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19674         instead of mode_for_size.
19675         * calls.c (save_fixed_argument_area): Likewise.  Make use of BLKmode
19676         explicit.
19677         * combine.c (expand_field_assignment): Use int_mode_for_size
19678         instead of mode_for_size.
19679         (make_extraction): Likewise.
19680         (simplify_shift_const_1): Likewise.
19681         (simplify_comparison): Likewise.
19682         * dojump.c (do_jump): Likewise.
19683         * dwarf2out.c (mem_loc_descriptor): Likewise.
19684         * emit-rtl.c (init_derived_machine_modes): Likewise.
19685         * expmed.c (flip_storage_order): Likewise.
19686         (convert_extracted_bit_field): Likewise.
19687         * expr.c (copy_blkmode_from_reg): Likewise.
19688         * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19689         * internal-fn.c (expand_mul_overflow): Likewise.
19690         * lower-subreg.c (simple_move): Likewise.
19691         * optabs-libfuncs.c (init_optabs): Likewise.
19692         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19693         * tree.c (vector_type_mode): Likewise.
19694         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19695         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19696         * tree-vect-generic.c (expand_vector_parallel): Likewise.
19697         * tree-vect-stmts.c (vectorizable_load): Likewise.
19698         (vectorizable_store): Likewise.
19700 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19701             Alan Hayward  <alan.hayward@arm.com>
19702             David Sherwood  <david.sherwood@arm.com>
19704         * coretypes.h (pod_mode): New type.
19705         (scalar_int_mode_pod): New typedef.
19706         * machmode.h (pod_mode): New class.
19707         (int_n_data_t::m): Change type to scalar_int_mode_pod.
19708         * genmodes.c (emit_mode_int_n): Update accordingly.
19709         * lower-subreg.h (target_lower_subreg): Change type to
19710         scalar_int_mode_pod.
19711         * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19712         scalar_int_mode_pod.
19714 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19715             Alan Hayward  <alan.hayward@arm.com>
19716             David Sherwood  <david.sherwood@arm.com>
19718         * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19719         machine_mode to scalar_int_mode.
19720         * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19721         (rs6000_option_override_internal): Remove cast to int.
19722         * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19723         machine_mode to scalar_int_mode.
19724         * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19725         (rs6000_option_override_internal): Remove cast to int.
19726         * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19727         * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19728         to machine_mode.
19729         * config/s390/s390.c (s390_expand_builtin): Likewise.
19730         * coretypes.h (scalar_int_mode): New type.
19731         (opt_scalar_int_mode): New typedef.
19732         * machmode.h (scalar_int_mode): New class.
19733         (scalar_int_mode::includes_p): New function.
19734         (byte_mode): Change type to scalar_int_mode.
19735         (word_mode): Likewise.
19736         (ptr_mode): Likewise.
19737         * emit-rtl.c (byte_mode): Likewise.
19738         (word_mode): Likewise.
19739         (ptr_mode): Likewise.
19740         (init_derived_machine_modes): Update accordingly.
19741         * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19742         and MODE_PARTIAL_INT.
19743         * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19744         opt_scalar_int_mode.
19746 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19747             Alan Hayward  <alan.hayward@arm.com>
19748             David Sherwood  <david.sherwood@arm.com>
19750         * target.def (libgcc_floating_mode_supported_p): Take a
19751         scalar_float_mode.
19752         * doc/tm.texi: Regenerate.
19753         * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19754         scalar_float_mode.
19755         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19756         * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19757         Likewise.
19759 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19760             Alan Hayward  <alan.hayward@arm.com>
19761             David Sherwood  <david.sherwood@arm.com>
19763         * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19764         * doc/tm.texi: Regenerate.
19765         * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19766         * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19767         * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19768         * targhooks.h (default_floatn_mode): Likewise.
19769         * targhooks.c (default_floatn_mode): Likewise.
19770         * tree.c (build_common_tree_nodes): Update accordingly.
19772 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19773             Alan Hayward  <alan.hayward@arm.com>
19774             David Sherwood  <david.sherwood@arm.com>
19776         * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19777         (mode_iterator::iterate_p): Likewise.
19778         (mode_iterator::get_wider): Likewise.
19779         * expr.c (init_expr_target): Use opt_scalar_float_mode.
19781 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19782             Alan Hayward  <alan.hayward@arm.com>
19783             David Sherwood  <david.sherwood@arm.com>
19785         * coretypes.h (opt_scalar_float_mode): New typedef.
19786         * machmode.h (float_mode_for_size): New function.
19787         * emit-rtl.c (double_mode): Delete.
19788         (init_emit_once): Use float_mode_for_size.
19789         * stor-layout.c (layout_type): Likewise.
19790         * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
19792 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19793             Alan Hayward  <alan.hayward@arm.com>
19794             David Sherwood  <david.sherwood@arm.com>
19796         * output.h (assemble_real): Take a scalar_float_mode.
19797         * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19798         * config/arm/arm.md (consttable_4): Likewise.
19799         (consttable_8): Likewise.
19800         (consttable_16): Likewise.
19801         * config/mips/mips.md (consttable_float): Likewise.
19802         * config/s390/s390.c (s390_output_pool_entry): Likewise.
19803         * varasm.c (assemble_real): Take a scalar_float_mode.
19804         (output_constant_pool_2): Update accordingly.
19805         (output_constant): Likewise.
19807 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19808             Alan Hayward  <alan.hayward@arm.com>
19809             David Sherwood  <david.sherwood@arm.com>
19811         * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19812         * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19813         * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19814         (native_encode_real): Likewise.
19815         (native_interpret_real): Likewise.
19816         * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19817         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19819 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19820             Alan Hayward  <alan.hayward@arm.com>
19821             David Sherwood  <david.sherwood@arm.com>
19823         * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19824         <scalar_float_mode>.  Simplify.
19825         (gen_extend_conv_libfunc): Likewise.
19827 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19828             Alan Hayward  <alan.hayward@arm.com>
19829             David Sherwood  <david.sherwood@arm.com>
19831         * coretypes.h (scalar_float_mode): New type.
19832         * machmode.h (mode_traits::from_int): Use machine_mode if
19833         USE_ENUM_MODES is defined.
19834         (is_a): New function.
19835         (as_a): Likewise.
19836         (dyn_cast): Likewise.
19837         (scalar_float_mode): New class.
19838         (scalar_float_mode::includes_p): New function.
19839         (is_float_mode): Likewise.
19840         * gdbhooks.py (MachineModePrinter): New class.
19841         (build_pretty_printer): Use it for scalar_float_mode.
19842         * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19843         (format_helper::format_helper): Turn into a template.
19844         * genmodes.c (get_mode_class): New function.
19845         (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19846         or machine_mode if none.
19847         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19848         as_a <scalar_float_mode>.
19849         * dwarf2out.c (mem_loc_descriptor): Likewise.
19850         (insert_float): Likewise.
19851         (add_const_value_attribute): Likewise.
19852         * simplify-rtx.c (simplify_immed_subreg): Likewise.
19853         * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19854         (expand_unop): Update accordingly.
19855         (expand_abs_nojump): Likewise.
19856         (expand_copysign_absneg): Take a scalar_float_mode.
19857         (expand_copysign_bit): Likewise.
19858         (expand_copysign): Update accordingly.
19860 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19861             Alan Hayward  <alan.hayward@arm.com>
19862             David Sherwood  <david.sherwood@arm.com>
19864         * coretypes.h (opt_mode): New class.
19865         * machmode.h (opt_mode): Likewise.
19866         (opt_mode::else_void): New function.
19867         (opt_mode::require): Likewise.
19868         (opt_mode::exists): Likewise.
19869         (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19870         (GET_MODE_2XWIDER_MODE): Likewise.
19871         (mode_iterator::get_wider): Update accordingly.
19872         (mode_iterator::get_2xwider): Likewise.
19873         (mode_iterator::get_known_wider): Likewise, turning into a template.
19874         * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19875         forcing a wider mode to exist.
19876         * config/cr16/cr16.h (LONG_REG_P): Likewise.
19877         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19878         * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19879         GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19880         * lower-subreg.c (init_lower_subreg): Likewise.
19881         * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19882         on the final iteration.
19883         * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19884         a wider mode exists before asking for a move pattern.
19885         (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19886         forcing a wider mode to exist.
19887         (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19888         returning false if no such mode exists.
19889         * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19890         * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19891         * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19892         Avoid checking for a MODE_INT if we already know the mode is not a
19893         SCALAR_INT_MODE_P.
19894         (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19895         forcing a wider mode to exist.
19896         (expmed_mult_highpart_optab): Likewise.
19897         (expmed_mult_highpart): Likewise.
19898         * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19899         using else_void.
19900         * lto-streamer-in.c (lto_input_mode_table): Likewise.
19901         * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19902         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19903         * internal-fn.c (expand_mul_overflow): Update use of
19904         GET_MODE_2XWIDER_MODE.
19905         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19906         * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19907         GET_MODE_WIDER_MODE.
19908         (convert_plusminus_to_widen): Likewise.
19909         * tree-switch-conversion.c (array_value_type): Likewise.
19910         * var-tracking.c (emit_note_insn_var_location): Likewise.
19911         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19912         Return false inside rather than outside the loop if no wider mode
19913         exists
19914         * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19915         and GET_MODE_2XWIDER_MODE
19916         (can_compare_p): Use else_void.
19917         * gdbhooks.py (OptMachineModePrinter): New class.
19918         (build_pretty_printer): Use it for opt_mode.
19920 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19921             Alan Hayward  <alan.hayward@arm.com>
19922             David Sherwood  <david.sherwood@arm.com>
19924         * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
19925         once.  Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
19927 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19928             Alan Hayward  <alan.hayward@arm.com>
19929             David Sherwood  <david.sherwood@arm.com>
19931         * machmode.h (mode_traits): New structure.
19932         (get_narrowest_mode): New function.
19933         (mode_iterator::start): Likewise.
19934         (mode_iterator::iterate_p): Likewise.
19935         (mode_iterator::get_wider): Likewise.
19936         (mode_iterator::get_known_wider): Likewise.
19937         (mode_iterator::get_2xwider): Likewise.
19938         (FOR_EACH_MODE_IN_CLASS): New mode iterator.
19939         (FOR_EACH_MODE): Likewise.
19940         (FOR_EACH_MODE_FROM): Likewise.
19941         (FOR_EACH_MODE_UNTIL): Likewise.
19942         (FOR_EACH_WIDER_MODE): Likewise.
19943         (FOR_EACH_2XWIDER_MODE): Likewise.
19944         * builtins.c (expand_builtin_strlen): Use new mode iterators.
19945         * combine.c (simplify_comparison): Likewise
19946         * config/i386/i386.c (type_natural_mode): Likewise.
19947         * cse.c (cse_insn): Likewise.
19948         * dse.c (find_shift_sequence): Likewise.
19949         * emit-rtl.c (init_derived_machine_modes): Likewise.
19950         (init_emit_once): Likewise.
19951         * explow.c (hard_function_value): Likewise.
19952         * expmed.c (extract_fixed_bit_field_1): Likewise.
19953         (extract_bit_field_1): Likewise.
19954         (expand_divmod): Likewise.
19955         (emit_store_flag_1): Likewise.
19956         * expr.c (init_expr_target): Likewise.
19957         (convert_move): Likewise.
19958         (alignment_for_piecewise_move): Likewise.
19959         (widest_int_mode_for_size): Likewise.
19960         (emit_block_move_via_movmem): Likewise.
19961         (copy_blkmode_to_reg): Likewise.
19962         (set_storage_via_setmem): Likewise.
19963         (compress_float_constant): Likewise.
19964         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19965         * optabs-query.c (get_best_extraction_insn): Likewise.
19966         * optabs.c (expand_binop): Likewise.
19967         (expand_twoval_unop): Likewise.
19968         (expand_twoval_binop): Likewise.
19969         (widen_leading): Likewise.
19970         (widen_bswap): Likewise.
19971         (expand_parity): Likewise.
19972         (expand_unop): Likewise.
19973         (prepare_cmp_insn): Likewise.
19974         (prepare_float_lib_cmp): Likewise.
19975         (expand_float): Likewise.
19976         (expand_fix): Likewise.
19977         (expand_sfix_optab): Likewise.
19978         * postreload.c (move2add_use_add2_insn): Likewise.
19979         * reg-stack.c (reg_to_stack): Likewise.
19980         * reginfo.c (choose_hard_reg_mode): Likewise.
19981         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19982         * stor-layout.c (mode_for_size): Likewise.
19983         (smallest_mode_for_size): Likewise.
19984         (mode_for_vector): Likewise.
19985         (finish_bitfield_representative): Likewise.
19986         * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
19987         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19988         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
19989         * var-tracking.c (prepare_call_arguments): Likewise.
19991 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19992             Alan Hayward  <alan.hayward@arm.com>
19993             David Sherwood  <david.sherwood@arm.com>
19995         * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
19996         * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
19997         USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
19998         * machmode.h (mode_size): Move earlier in file.
19999         (mode_precision): Likewise.
20000         (mode_inner): Likewise.
20001         (mode_nunits): Likewise.
20002         (mode_unit_size): Likewise.
20003         (unit_unit_precision): Likewise.
20004         (mode_wider): Likewise.
20005         (mode_2xwider): Likewise.
20006         (machine_mode): New class.
20007         (mode_to_bytes): New function.
20008         (mode_to_bits): Likewise.
20009         (mode_to_precision): Likewise.
20010         (mode_to_inner): Likewise.
20011         (mode_to_unit_size): Likewise.
20012         (mode_to_unit_precision): Likewise.
20013         (mode_to_nunits): Likewise.
20014         (GET_MODE_SIZE): Use mode_to_bytes.
20015         (GET_MODE_BITSIZE): Use mode_to_bits.
20016         (GET_MODE_PRECISION): Use mode_to_precision.
20017         (GET_MODE_INNER): Use mode_to_inner.
20018         (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
20019         (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
20020         (GET_MODE_NUNITS): Use mode_to_nunits.
20021         * system.h (ALWAYS_INLINE): New macro.
20022         * config/powerpcspe/powerpcspe-c.c
20023         (altivec_resolve_overloaded_builtin): Use machine_mode instead of
20024         int for arg1_mode and arg2_mode.
20026 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20027             Alan Hayward  <alan.hayward@arm.com>
20028             David Sherwood  <david.sherwood@arm.com>
20030         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
20031         Prefix mode names with E_ in case statements.
20032         * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20033         * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
20034         (aarch64_split_simd_move): Likewise.
20035         (aarch64_gen_storewb_pair): Likewise.
20036         (aarch64_gen_loadwb_pair): Likewise.
20037         (aarch64_gen_store_pair): Likewise.
20038         (aarch64_gen_load_pair): Likewise.
20039         (aarch64_get_condition_code_1): Likewise.
20040         (aarch64_constant_pool_reload_icode): Likewise.
20041         (get_rsqrte_type): Likewise.
20042         (get_rsqrts_type): Likewise.
20043         (get_recpe_type): Likewise.
20044         (get_recps_type): Likewise.
20045         (aarch64_gimplify_va_arg_expr): Likewise.
20046         (aarch64_simd_container_mode): Likewise.
20047         (aarch64_emit_load_exclusive): Likewise.
20048         (aarch64_emit_store_exclusive): Likewise.
20049         (aarch64_expand_compare_and_swap): Likewise.
20050         (aarch64_gen_atomic_cas): Likewise.
20051         (aarch64_emit_bic): Likewise.
20052         (aarch64_emit_atomic_swap): Likewise.
20053         (aarch64_emit_atomic_load_op): Likewise.
20054         (aarch64_evpc_trn): Likewise.
20055         (aarch64_evpc_uzp): Likewise.
20056         (aarch64_evpc_zip): Likewise.
20057         (aarch64_evpc_ext): Likewise.
20058         (aarch64_evpc_rev): Likewise.
20059         (aarch64_evpc_dup): Likewise.
20060         (aarch64_gen_ccmp_first): Likewise.
20061         (aarch64_gen_ccmp_next): Likewise.
20062         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
20063         (alpha_emit_xfloating_libcall): Likewise.
20064         (emit_insxl): Likewise.
20065         (alpha_arg_type): Likewise.
20066         * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
20067         (arc_preferred_simd_mode): Likewise.
20068         (arc_secondary_reload): Likewise.
20069         (get_arc_condition_code): Likewise.
20070         (arc_print_operand): Likewise.
20071         (arc_legitimate_constant_p): Likewise.
20072         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20073         * config/arc/arc.md (casesi_load): Likewise.
20074         (casesi_compact_jump): Likewise.
20075         * config/arc/predicates.md (proper_comparison_operator): Likewise.
20076         (cc_use_register): Likewise.
20077         * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20078         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
20079         (arm_init_iwmmxt_builtins): Likewise.
20080         * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
20081         (neon_expand_vector_init): Likewise.
20082         (arm_attr_length_move_neon): Likewise.
20083         (maybe_get_arm_condition_code): Likewise.
20084         (arm_emit_vector_const): Likewise.
20085         (arm_preferred_simd_mode): Likewise.
20086         (arm_output_iwmmxt_tinsr): Likewise.
20087         (thumb1_output_casesi): Likewise.
20088         (thumb2_output_casesi): Likewise.
20089         (arm_emit_load_exclusive): Likewise.
20090         (arm_emit_store_exclusive): Likewise.
20091         (arm_expand_compare_and_swap): Likewise.
20092         (arm_evpc_neon_vuzp): Likewise.
20093         (arm_evpc_neon_vzip): Likewise.
20094         (arm_evpc_neon_vrev): Likewise.
20095         (arm_evpc_neon_vtrn): Likewise.
20096         (arm_evpc_neon_vext): Likewise.
20097         (arm_validize_comparison): Likewise.
20098         * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
20099         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
20100         * config/avr/avr.c (avr_rtx_costs_1): Likewise.
20101         * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
20102         (c6x_preferred_simd_mode): Likewise.
20103         * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
20104         (epiphany_rtx_costs): Likewise.
20105         * config/epiphany/predicates.md (proper_comparison_operator):
20106         Likewise.
20107         * config/frv/frv.c (condexec_memory_operand): Likewise.
20108         (frv_emit_move): Likewise.
20109         (output_move_single): Likewise.
20110         (output_condmove_single): Likewise.
20111         (frv_hard_regno_mode_ok): Likewise.
20112         (frv_matching_accg_mode): Likewise.
20113         * config/h8300/h8300.c (split_adds_subs): Likewise.
20114         (h8300_rtx_costs): Likewise.
20115         (h8300_print_operand): Likewise.
20116         (compute_mov_length): Likewise.
20117         (output_logical_op): Likewise.
20118         (compute_logical_op_length): Likewise.
20119         (compute_logical_op_cc): Likewise.
20120         (h8300_shift_needs_scratch_p): Likewise.
20121         (output_a_shift): Likewise.
20122         (compute_a_shift_length): Likewise.
20123         (compute_a_shift_cc): Likewise.
20124         (expand_a_rotate): Likewise.
20125         (output_a_rotate): Likewise.
20126         * config/i386/i386.c (classify_argument): Likewise.
20127         (function_arg_advance_32): Likewise.
20128         (function_arg_32): Likewise.
20129         (function_arg_64): Likewise.
20130         (function_value_64): Likewise.
20131         (ix86_gimplify_va_arg): Likewise.
20132         (ix86_legitimate_constant_p): Likewise.
20133         (put_condition_code): Likewise.
20134         (split_double_mode): Likewise.
20135         (ix86_avx256_split_vector_move_misalign): Likewise.
20136         (ix86_expand_vector_logical_operator): Likewise.
20137         (ix86_split_idivmod): Likewise.
20138         (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
20139         (ix86_build_const_vector): Likewise.
20140         (ix86_build_signbit_mask): Likewise.
20141         (ix86_match_ccmode): Likewise.
20142         (ix86_cc_modes_compatible): Likewise.
20143         (ix86_expand_branch): Likewise.
20144         (ix86_expand_sse_cmp): Likewise.
20145         (ix86_expand_sse_movcc): Likewise.
20146         (ix86_expand_int_sse_cmp): Likewise.
20147         (ix86_expand_vec_perm_vpermi2): Likewise.
20148         (ix86_expand_vec_perm): Likewise.
20149         (ix86_expand_sse_unpack): Likewise.
20150         (ix86_expand_int_addcc): Likewise.
20151         (ix86_split_to_parts): Likewise.
20152         (ix86_vectorize_builtin_gather): Likewise.
20153         (ix86_vectorize_builtin_scatter): Likewise.
20154         (avx_vpermilp_parallel): Likewise.
20155         (inline_memory_move_cost): Likewise.
20156         (ix86_tieable_integer_mode_p): Likewise.
20157         (x86_maybe_negate_const_int): Likewise.
20158         (ix86_expand_vector_init_duplicate): Likewise.
20159         (ix86_expand_vector_init_one_nonzero): Likewise.
20160         (ix86_expand_vector_init_one_var): Likewise.
20161         (ix86_expand_vector_init_concat): Likewise.
20162         (ix86_expand_vector_init_interleave): Likewise.
20163         (ix86_expand_vector_init_general): Likewise.
20164         (ix86_expand_vector_set): Likewise.
20165         (ix86_expand_vector_extract): Likewise.
20166         (emit_reduc_half): Likewise.
20167         (ix86_emit_i387_round): Likewise.
20168         (ix86_mangle_type): Likewise.
20169         (ix86_expand_round_sse4): Likewise.
20170         (expand_vec_perm_blend): Likewise.
20171         (canonicalize_vector_int_perm): Likewise.
20172         (ix86_expand_vec_one_operand_perm_avx512): Likewise.
20173         (expand_vec_perm_1): Likewise.
20174         (expand_vec_perm_interleave3): Likewise.
20175         (expand_vec_perm_even_odd_pack): Likewise.
20176         (expand_vec_perm_even_odd_1): Likewise.
20177         (expand_vec_perm_broadcast_1): Likewise.
20178         (ix86_vectorize_vec_perm_const_ok): Likewise.
20179         (ix86_expand_vecop_qihi): Likewise.
20180         (ix86_expand_mul_widen_hilo): Likewise.
20181         (ix86_expand_sse2_abs): Likewise.
20182         (ix86_expand_pextr): Likewise.
20183         (ix86_expand_pinsr): Likewise.
20184         (ix86_preferred_simd_mode): Likewise.
20185         (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
20186         * config/i386/sse.md (*andnot<mode>3): Likewise.
20187         (<mask_codefor><code><mode>3<mask_name>): Likewise.
20188         (*<code><mode>3): Likewise.
20189         * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
20190         (ia64_expand_atomic_op): Likewise.
20191         (ia64_arg_type): Likewise.
20192         (ia64_mode_to_int): Likewise.
20193         (ia64_scalar_mode_supported_p): Likewise.
20194         (ia64_vector_mode_supported_p): Likewise.
20195         (expand_vec_perm_broadcast): Likewise.
20196         * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
20197         (iq2000_function_arg_advance): Likewise.
20198         (iq2000_function_arg): Likewise.
20199         * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
20200         * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
20201         (m68k_libcall_value): Likewise.
20202         (m68k_function_value): Likewise.
20203         (sched_attr_op_type): Likewise.
20204         * config/mcore/mcore.c (mcore_output_move): Likewise.
20205         * config/microblaze/microblaze.c (microblaze_function_arg_advance):
20206         Likewise.
20207         (microblaze_function_arg): Likewise.
20208         * config/mips/mips.c (mips16_build_call_stub): Likewise.
20209         (mips_print_operand): Likewise.
20210         (mips_mode_ok_for_mov_fmt_p): Likewise.
20211         (mips_vector_mode_supported_p): Likewise.
20212         (mips_preferred_simd_mode): Likewise.
20213         (mips_expand_vpc_loongson_even_odd): Likewise.
20214         (mips_expand_vec_unpack): Likewise.
20215         (mips_expand_vi_broadcast): Likewise.
20216         (mips_expand_vector_init): Likewise.
20217         (mips_expand_vec_reduc): Likewise.
20218         (mips_expand_msa_cmp): Likewise.
20219         * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
20220         * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
20221         (cc_flags_for_mode): Likewise.
20222         * config/msp430/msp430.c (msp430_print_operand): Likewise.
20223         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
20224         (nds32_output_casesi_pc_relative): Likewise.
20225         * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20226         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
20227         (nvptx_gen_unpack): Likewise.
20228         (nvptx_gen_pack): Likewise.
20229         (nvptx_gen_shuffle): Likewise.
20230         (nvptx_gen_wcast): Likewise.
20231         (nvptx_preferred_simd_mode): Likewise.
20232         * config/pa/pa.c (pa_secondary_reload): Likewise.
20233         * config/pa/predicates.md (base14_operand): Likewise.
20234         * config/powerpcspe/powerpcspe-c.c
20235         (altivec_resolve_overloaded_builtin): Likewise.
20236         * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
20237         Likewise.
20238         (rs6000_preferred_simd_mode): Likewise.
20239         (output_vec_const_move): Likewise.
20240         (rs6000_expand_vector_extract): Likewise.
20241         (rs6000_split_vec_extract_var): Likewise.
20242         (reg_offset_addressing_ok_p): Likewise.
20243         (rs6000_legitimate_offset_address_p): Likewise.
20244         (rs6000_legitimize_address): Likewise.
20245         (rs6000_emit_set_const): Likewise.
20246         (rs6000_const_vec): Likewise.
20247         (rs6000_emit_move): Likewise.
20248         (spe_build_register_parallel): Likewise.
20249         (rs6000_darwin64_record_arg_recurse): Likewise.
20250         (swap_selector_for_mode): Likewise.
20251         (spe_init_builtins): Likewise.
20252         (paired_init_builtins): Likewise.
20253         (altivec_init_builtins): Likewise.
20254         (do_load_for_compare): Likewise.
20255         (rs6000_generate_compare): Likewise.
20256         (rs6000_expand_float128_convert): Likewise.
20257         (emit_load_locked): Likewise.
20258         (emit_store_conditional): Likewise.
20259         (rs6000_output_function_epilogue): Likewise.
20260         (rs6000_handle_altivec_attribute): Likewise.
20261         (rs6000_function_value): Likewise.
20262         (emit_fusion_gpr_load): Likewise.
20263         (emit_fusion_p9_load): Likewise.
20264         (emit_fusion_p9_store): Likewise.
20265         * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20266         (fusion_gpr_mem_load): Likewise.
20267         (fusion_addis_mem_combo_load): Likewise.
20268         (fusion_addis_mem_combo_store): Likewise.
20269         * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20270         (fusion_gpr_mem_load): Likewise.
20271         (fusion_addis_mem_combo_load): Likewise.
20272         (fusion_addis_mem_combo_store): Likewise.
20273         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20274         Likewise.
20275         * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20276         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20277         (rs6000_preferred_simd_mode): Likewise.
20278         (output_vec_const_move): Likewise.
20279         (rs6000_expand_vector_extract): Likewise.
20280         (rs6000_split_vec_extract_var): Likewise.
20281         (reg_offset_addressing_ok_p): Likewise.
20282         (rs6000_legitimate_offset_address_p): Likewise.
20283         (rs6000_legitimize_address): Likewise.
20284         (rs6000_emit_set_const): Likewise.
20285         (rs6000_const_vec): Likewise.
20286         (rs6000_emit_move): Likewise.
20287         (rs6000_darwin64_record_arg_recurse): Likewise.
20288         (swap_selector_for_mode): Likewise.
20289         (paired_init_builtins): Likewise.
20290         (altivec_init_builtins): Likewise.
20291         (rs6000_expand_float128_convert): Likewise.
20292         (emit_load_locked): Likewise.
20293         (emit_store_conditional): Likewise.
20294         (rs6000_output_function_epilogue): Likewise.
20295         (rs6000_handle_altivec_attribute): Likewise.
20296         (rs6000_function_value): Likewise.
20297         (emit_fusion_gpr_load): Likewise.
20298         (emit_fusion_p9_load): Likewise.
20299         (emit_fusion_p9_store): Likewise.
20300         * config/rx/rx.c (rx_gen_move_template): Likewise.
20301         (flags_from_mode): Likewise.
20302         * config/s390/predicates.md (s390_alc_comparison): Likewise.
20303         (s390_slb_comparison): Likewise.
20304         * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20305         (s390_vector_mode_supported_p): Likewise.
20306         (s390_cc_modes_compatible): Likewise.
20307         (s390_match_ccmode_set): Likewise.
20308         (s390_canonicalize_comparison): Likewise.
20309         (s390_emit_compare_and_swap): Likewise.
20310         (s390_branch_condition_mask): Likewise.
20311         (s390_rtx_costs): Likewise.
20312         (s390_secondary_reload): Likewise.
20313         (__SECONDARY_RELOAD_CASE): Likewise.
20314         (s390_expand_cs): Likewise.
20315         (s390_preferred_simd_mode): Likewise.
20316         * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20317         * config/sh/sh.c (sh_print_operand): Likewise.
20318         (dump_table): Likewise.
20319         (sh_secondary_reload): Likewise.
20320         * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20321         * config/sh/sh.md (casesi_worker_1): Likewise.
20322         (casesi_worker_2): Likewise.
20323         * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20324         (fcc_comparison_operator): Likewise.
20325         * config/sparc/sparc.c (sparc_expand_move): Likewise.
20326         (emit_soft_tfmode_cvt): Likewise.
20327         (sparc_preferred_simd_mode): Likewise.
20328         (output_cbranch): Likewise.
20329         (sparc_print_operand): Likewise.
20330         (sparc_expand_vec_perm_bmask): Likewise.
20331         (vector_init_bshuffle): Likewise.
20332         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20333         (spu_vector_mode_supported_p): Likewise.
20334         (spu_expand_insv): Likewise.
20335         (spu_emit_branch_or_set): Likewise.
20336         (spu_handle_vector_attribute): Likewise.
20337         (spu_builtin_splats): Likewise.
20338         (spu_builtin_extract): Likewise.
20339         (spu_builtin_promote): Likewise.
20340         (spu_expand_sign_extend): Likewise.
20341         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20342         (tilegx_simd_int): Likewise.
20343         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20344         (tilepro_simd_int): Likewise.
20345         * config/v850/v850.c (const_double_split): Likewise.
20346         (v850_print_operand): Likewise.
20347         (ep_memory_offset): Likewise.
20348         * config/vax/vax.c (vax_rtx_costs): Likewise.
20349         (vax_output_int_move): Likewise.
20350         (vax_output_int_add): Likewise.
20351         (vax_output_int_subtract): Likewise.
20352         * config/visium/predicates.md (visium_branch_operator): Likewise.
20353         * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20354         (visium_print_operand_address): Likewise.
20355         * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20356         * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20357         (xtensa_expand_conditional_branch): Likewise.
20358         (xtensa_copy_incoming_a7): Likewise.
20359         (xtensa_output_literal): Likewise.
20360         * dfp.c (decimal_real_maxval): Likewise.
20361         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20363 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20364             Alan Hayward  <alan.hayward@arm.com>
20365             David Sherwood  <david.sherwood@arm.com>
20367         * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20368         (mode_nunits_inline): Likewise.
20369         (mode_inner_inline): Likewise.
20370         (mode_unit_size_inline): Likewise.
20371         (mode_unit_precision_inline): Likewise.
20372         (emit_insn_modes_h): Likewise.  Also emit a #define of the
20373         unprefixed name.
20374         (emit_mode_wider): Add an E_ prefix to mode names.
20375         (emit_mode_complex): Likewise.
20376         (emit_mode_inner): Likewise.
20377         (emit_mode_adjustments): Likewise.
20378         (emit_mode_int_n): Likewise.
20379         * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20380         (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20381         (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20382         (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20383         (CRC32_BUILTIN, ENTRY): Likewise.
20384         * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20385         (aarch64_pop_regs): Likewise.
20386         (aarch64_process_components): Likewise.
20387         * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20388         * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20389         (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20390         (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20391         * config/arm/arm.c (arm_init_libfuncs): Likewise.
20392         * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20393         Likewise.
20394         * config/i386/i386-builtin.def (pcmpestr): Likewise.
20395         (pcmpistr): Likewise.
20396         * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20397         * config/mmix/mmix.c (mmix_output_condition): Likewise.
20398         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20399         Likewise.
20400         * config/rl78/rl78.c (mduc_regs): Likewise.
20401         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20402         (htm_expand_builtin): Likewise.
20403         * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20404         * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20405         * config/xtensa/xtensa.c (print_operand): Likewise.
20406         * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20407         (NUM_MODE_VECTOR_INT): Likewise.
20408         * genoutput.c (null_operand): Likewise.
20409         (output_operand_data): Likewise.
20410         * genrecog.c (print_parameter_value): Likewise.
20411         * lra.c (debug_operand_data): Likewise.
20413 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20415         * dwarf2out.c (scompare_loc_descriptor_wide)
20416         (scompare_loc_descriptor_narrow): New functions, split out from...
20417         (scompare_loc_descriptor): ...here.
20418         * expmed.c (emit_store_flag_int): New function, split out from...
20419         (emit_store_flag): ...here.
20421 2017-08-30  Richard Biener  <rguenther@suse.de>
20423         * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20424         (dwarf2out_early_finish): Move setting of AT_pubnames from
20425         early debug output to early finish.
20427 2017-08-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20429         * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20430         and -mdata-region to the assembler.
20432 2017-08-30  Richard Biener  <rguenther@suse.de>
20434         * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20435         attributes.
20436         (gen_subprogram_die): Add DW_AT_object_pointer only early.
20437         (dwarf2out_early_global_decl): Only generate a DIE for the
20438         abstract origin if it doesn't already exist or is a declaration DIE.
20439         (resolve_addr): Do not add the linkage name twice when
20440         generating a stub DIE for the DW_TAG_GNU_call_site target.
20442 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20444         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20445         Use machine_mode rather than int for arg1_mode.
20447 2017-08-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20449         PR target/82015
20450         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20451         that the second argument of the built-in functions to unpack
20452         128-bit scalar types to 64-bit values is 0 or 1.  Change to use a
20453         switch statement instead a lot of if statements.
20454         * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20455         Allow 64-bit values to be in Altivec registers as well as
20456         traditional floating point registers.
20457         (pack<mode>, FMOVE128_VSX iterator): Likewise.
20459 2017-08-29  Alexander Monakov  <amonakov@ispras.ru>
20461         * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20462         MAX_REGS_PER_ADDRESS == 1.
20464 2017-08-29  Uros Bizjak  <ubizjak@gmail.com>
20466         * config/i386/i386.opt (flag_fentry): Do not init to -1.
20467         * config/i386/i386.c (ix86_option_override_internal): Simplify
20468         setting of opts->x_flag_entry.
20470 2017-08-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20471             Jakub Jelinek  <jakub@redhat.com>
20472             Richard Biener  <rguenther@suse.de>
20474         PR tree-optimization/81503
20475         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20476         folded constant fits in the target type; reorder tests for clarity.
20478 2017-08-29  Martin Liska  <mliska@suse.cz>
20480         * passes.def: Include pass_lower_switch.
20481         * stmt.c (dump_case_nodes): Remove and move to
20482         tree-switch-conversion.
20483         (case_values_threshold): Likewise.
20484         (expand_switch_as_decision_tree_p): Likewise.
20485         (emit_case_decision_tree): Likewise.
20486         (expand_case): Likewise.
20487         (balance_case_nodes): Likewise.
20488         (node_has_low_bound): Likewise.
20489         (node_has_high_bound): Likewise.
20490         (node_is_bounded): Likewise.
20491         (emit_case_nodes): Likewise.
20492         (struct simple_case_node): New struct.
20493         (add_case_node): Remove.
20494         (emit_case_dispatch_table): Use vector instead of case_list.
20495         (reset_out_edges_aux): Remove.
20496         (compute_cases_per_edge): Likewise.
20497         (expand_case): Build list of simple_case_node.
20498         (expand_sjlj_dispatch_table): Use it.
20499         * tree-switch-conversion.c (struct case_node): Moved from
20500         stmt.c and adjusted.
20501         (emit_case_nodes): Likewise.
20502         (node_has_low_bound): Likewise.
20503         (node_has_high_bound): Likewise.
20504         (node_is_bounded): Likewise.
20505         (case_values_threshold): Likewise.
20506         (reset_out_edges_aux): Likewise.
20507         (compute_cases_per_edge): Likewise.
20508         (add_case_node): Likewise.
20509         (dump_case_nodes): Likewise.
20510         (balance_case_nodes): Likewise.
20511         (expand_switch_as_decision_tree_p): Likewise.
20512         (emit_jump): Likewise.
20513         (emit_case_decision_tree): Likewise.
20514         (try_switch_expansion): Likewise.
20515         (do_jump_if_equal): Likewise.
20516         (emit_cmp_and_jump_insns): Likewise.
20517         (fix_phi_operands_for_edge): New function.
20518         (record_phi_operand_mapping): Likewise.
20519         (class pass_lower_switch): New pass.
20520         (pass_lower_switch::execute): New function.
20521         (make_pass_lower_switch): Likewise.
20522         (conditional_probability):
20523         * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20524         * tree-pass.h: Add make_pass_lower_switch.
20526 2017-08-29  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20528         PR target/80993
20529         * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20530         handlers as used.
20532 2017-08-29  Richard Biener  <rguenther@suse.de>
20534         * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20535         we do not add a DW_AT_inline attribute twice.
20536         (gen_subprogram_die): Remove code setting DW_AT_inline on
20537         DECL_ABSTRACT_P nodes.
20539 2017-08-29  Richard Sandiford  <richard.sandiford@linaro.org>
20541         * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20542         calls to internal functions.
20543         (gimplify_modify_expr): Likewise.
20544         * tree-call-cdce.c (use_internal_fn): Likewise.
20545         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20546         (convert_to_divmod): Set the nothrow flag.
20547         * tree-if-conv.c (predicate_mem_writes):  Likewise.
20548         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20549         (vectorizable_call): Likewise.
20550         (vectorizable_store): Likewise.
20551         (vectorizable_load): Likewise.
20552         * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20553         (vect_recog_mask_conversion_pattern): Likewise.
20555 2017-08-29  Martin Liska  <mliska@suse.cz>
20557         PR other/39851
20558         * gcc.c (driver_handle_option): Add new argument.
20559         * opts-common.c (handle_option): Pass
20560         target_option_override_hook.
20561         * opts-global.c (lang_handle_option): Add new option.
20562         (set_default_handlers):  Add new argument.
20563         (decode_options): Likewise.
20564         * opts.c (target_handle_option): Likewise.
20565         (common_handle_option): Call target_option_override_hook.
20566         * opts.h (struct cl_option_handler_func): Add hook for
20567         target option override.
20568         (struct cl_option_handlers): Likewise.
20569         (set_default_handlers): Add new argument.
20570         (decode_options): Likewise.
20571         (common_handle_option): Likewise.
20572         (target_handle_option): Likewise.
20573         * toplev.c (toplev::main): Pass targetm.target_option.override
20574         hook.
20576 2017-08-29  Richard Biener  <rguenther@suse.de>
20577         Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20579         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20580         life to the active subtree.
20582 2017-08-28  Jeff Law  <law@redhat.com>
20584         * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20585         derive_equivalences.
20586         (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20587         Code moved into....
20588         (edge_info::derive_equivalences): New private member function
20590         * tree-ssa-dom.c (class edge_info): Changed from a struct
20591         to a class.  Add ctor/dtor, methods and data members.
20592         (edge_info::edge_info): Renamed from allocate_edge_info.
20593         Initialize additional members.
20594         (edge_info::~edge_info): New.
20595         (free_dom_edge_info): Delete the edge info.
20596         (record_edge_info): Use new class & associated member functions.
20597         Tighten forms for testing for edge equivalences.
20598         (record_temporary_equivalences): Iterate over the simple
20599         equivalences rather than assuming there's only one per edge.
20600         (cprop_into_successor_phis): Iterate over the simple
20601         equivalences rather than assuming there's only one per edge.
20602         (optimize_stmt): Use operand_equal_p rather than pointer
20603         equality for mini-DSE code.
20605 2017-08-28  Nathan Sidwell  <nathan@acm.org>
20607         * gcc.c (execute): Fold SIGPIPE handling into switch
20608         statement.  Adjust internal error message.
20610 2017-08-28  Richard Biener  <rguenther@suse.de>
20612         PR debug/81993
20613         * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20614         Do nothing for removed DIEs.
20616 2017-08-28  Richard Biener  <rguenther@suse.de>
20618         PR tree-optimization/81977
20619         * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20620         memcpy.
20622 2017-08-28  Alexander Monakov  <amonakov@ispras.ru>
20624         PR target/80640
20625         * doc/md.texi (mem_thread_fence): Remove mention of mode.  Rewrite.
20626         * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20627         using targetm.gen_mem_thread_fence.
20629 2017-08-27  Uros Bizjak  <ubizjak@gmail.com>
20631         PR target/81995
20632         * config/i386/i386.md (*<btsc><mode>): Change operand 2
20633         predicate to register_operand.  Reorder operands.
20634         (*btr<mode>): Ditto.
20635         (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20636         (*btr<mode>_mask): Ditto.
20638 2017-08-25  Steven Munroe  <munroesj@gcc.gnu.org>
20640         * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20641         * config/rs6000/xmmintrin.h: New file.
20642         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20644 2017-08-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20646         PR target/81504
20647         * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20648         parameter and_insn and return it.
20649         (recombine_lvx_pattern): Insert a copy to ensure availability of
20650         the base register of the copied masking operation at the point of
20651         the instruction replacement.
20652         (recombine_stvx_pattern): Likewise.
20654 2017-08-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
20656         * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20657         undocumented switches.
20658         (-mpower9-dform-vector): Likewise.
20659         (-mpower9-dform): Likewise.
20660         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20661         comments to delete references to -mpower9-dform* switches.
20662         * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20663         Delete reference to -mpower9-dform* switches, test for
20664         -mpower9-vector instead.
20665         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20666         (OTHER_P9_VECTOR_MASKS): Likewise.
20667         (POWERPC_MASKS): Likewise.
20668         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20669         tests against -mpower9-dform* to -mpower9-vector.  Delete code
20670         that checked for -mpower9-dform* consistancy with other options.
20671         Add test for -mpower9-misc to enable other power9 switches.
20672         (rs6000_init_hard_regno_mode_ok): Likewise.
20673         (rs6000_option_override_internal): Likewise.
20674         (rs6000_emit_prologue): Likewise.
20675         (rs6000_emit_epilogue): Likewise.
20676         (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20677         (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20678         (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20679         -mpower9-vector.
20680         (emit_fusion_p9_store): Likewise.
20681         * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20682         resetting these macros if the assembler does not support ISA 3.0
20683         instructions.
20684         (TARGET_P9_DFORM_VECTOR): Likewise.
20685         * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20686         Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20688 2017-08-25  Alan Modra  <amodra@gmail.com>
20690         PR rtl-optimization/81747
20691         * cse.c (cse_extended_basic_block): Don't attempt to record
20692         equivalences for degenerate conditional branches that branch
20693         to their fall-through.
20695 2017-08-24  Martin Sebor  <msebor@redhat.com>
20697         PR middle-end/81908
20698         * gimple-fold.c (size_must_be_zero_p): New function.
20699         (gimple_fold_builtin_memory_op): Call it.
20701 2017-08-24  Steven Munroe  <munroesj@gcc.gnu.org>
20703         * config/rs6000/mm_malloc.h: New file.
20705 2017-08-24  Bin Cheng  <bin.cheng@arm.com>
20707         PR tree-optimization/81913
20708         * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20709         analysis when either IVs in condition can wrap.
20711 2017-08-24  Uros Bizjak  <ubizjak@gmail.com>
20713         * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20714         * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20716 2017-08-24  Richard Biener  <rguenther@suse.de>
20718         PR target/81921
20719         * targhooks.c (default_target_can_inline_p): Properly
20720         use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20721         is present and always compare.
20722         * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20723         infer -mfpmath=sse from TARGET_SSE_P.
20724         (ix86_can_inline_p): Properly use target_option_default_node when
20725         no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20727 2017-08-24  Richard Biener  <rguenther@suse.de>
20729         PR debug/81936
20730         * dwarf2out.c (output_die): Handle flag_generate_offload like
20731         flag_generate_lto.
20732         (output_comp_unit): Likewise.
20733         (gen_array_type_die): Likewise.
20734         (dwarf2out_early_finish): Likewise.
20735         (note_variable_value_in_expr): Likewise.
20736         (dwarf2out_finish): Likewise.  Adjust assert.
20737         * cgraphunit.c (symbol_table::compile): Move setting of
20738         flag_generate_offload earlier ...
20739         (symbol_table::finalize_compilation_unit): ... here, before
20740         early debug finalization.
20742 2017-08-24  Richard Biener  <rguenther@suse.de>
20744         * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20745         and ipa-fnsummary.h.
20746         (ix86_can_inline_p): When ix86_fpmath flags do not match
20747         check whether the callee uses FP math at all.
20749 2017-08-24  Aldy Hernandez  <aldyh@redhat.com>
20751         PR middle-end/81931
20752         * tree-ssanames.c (get_nonzero_bits): Use element_precision
20753         instead of TYPE_PRECISION.
20755 2017-08-24  Richard Sandiford  <richard.sandiford@linaro.org>
20756             Alan Hayward  <alan.hayward@arm.com>
20757             David Sherwood  <david.sherwood@arm.com>
20759         * combine.c (make_extraction): Use subreg_offset_from_lsb.
20761 2017-08-23  Daniel Santos  <daniel.santos@pobox.com>
20763         * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20764         Remove field.
20765         (ix86_frame::stack_realign_allocate): New field.
20766         (struct machine_frame_state): Modify comments.
20767         (machine_frame_state::sp_realigned_fp_end): New field.
20768         * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20769         layout calculation.
20770         (sp_valid_at): Add assertion to assure no attempt to access invalid
20771         offset of a realigned stack.
20772         (fp_valid_at): Likewise.
20773         (choose_baseaddr): Modify comments.
20774         (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20775         ix86_expand_prologue.
20776         (ix86_expand_prologue): Modify stack realignment and allocation.
20777         (ix86_expand_epilogue): Modify comments.
20778         * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20779         avx2_runtime, avx512f, and avx512f_runtime.
20781 2017-08-23  Uros Bizjak  <ubizjak@gmail.com>
20783         * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20784         (mstackrealign): Do not init to -1.
20785         * config/i386/i386.c (ix86_option_override_internal):
20786         Check opts_set, not opts when setting default value of
20787         opts->x_ix86_force_align_arg_pointer.
20789 2017-08-23  Richard Biener  <rguenther@suse.de>
20791         * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20792         lang_hooks.decl_printable_name.
20793         * print-rtl-function.c (print_rtx_function): Likewise.
20794         * tree-pretty-print.c (dump_function_header): Likewise.
20796 2017-08-23  Richard Biener  <rguenther@suse.de>
20798         PR lto/81940
20799         * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20800         -g0 at compile-time.
20802 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20804         PR middle-end/19706
20805         * doc/sourcebuild.texi (Other hardware attributes):
20806         Document xorsign.
20808 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20810         PR middle-end/19706
20811         * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20812         Fix single-use check.
20814 2017-08-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20816         * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20818 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20820         * doc/install.texi: Modify to add more details on running selected
20821         tests.
20823 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20825         * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20826         is combined with -mabi=ms.
20827         (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20828         ms_abi.
20830 2017-08-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20832         PR tree-optimization/81488
20833         * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20834         and cached_basis fields.
20835         (MAX_SPREAD): New constant.
20836         (alloc_cand_and_find_basis): Initialize new fields.
20837         (clear_visited): New function.
20838         (create_phi_basis_1): Rename from create_phi_basis, set visited
20839         and cached_basis fields.
20840         (create_phi_basis): New wrapper function.
20841         (phi_add_costs_1): Rename from phi_add_costs, add spread
20842         parameter, set visited field, short-circuit when limits reached.
20843         (phi_add_costs): New wrapper function.
20844         (record_phi_increments_1): Rename from record_phi_increments, set
20845         visited field.
20846         (record_phi_increments): New wrapper function.
20847         (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20848         (phi_incr_cost): New wrapper function.
20849         (all_phi_incrs_profitable_1): Rename from
20850         all_phi_incrs_profitable, set visited field.
20851         (all_phi_incrs_profitable): New wrapper function.
20853 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20854             Alan Hayward  <alan.hayward@arm.com>
20855             David Sherwood  <david.sherwood@arm.com>
20857         * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20858         that takes the outer and inner modes.
20859         * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20860         comparison as the canonical test for a paradoxical subreg.
20861         * combine.c (simplify_set): Use paradoxical_subreg_p.
20862         (make_extraction): Likewise.
20863         (force_to_mode): Likewise.
20864         (rtx_equal_for_field_assignment_p): Likewise.
20865         (gen_lowpart_for_combine): Likewise.
20866         (simplify_comparison): Likewise.
20867         * cse.c (equiv_constant): Likewise.
20868         * expmed.c (store_bit_field_1): Likewise.
20869         * final.c (alter_subreg): Likewise.
20870         * fwprop.c (propagate_rtx): Likewise.
20871         (forward_propagate_subreg): Likewise.
20872         * ira-conflicts.c (ira_build_conflicts): Likewise.
20873         * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20874         * lra-constraints.c (curr_insn_transform): Likewise.
20875         (split_reg): Likewise.
20876         * lra-eliminations.c (move_plus_up): Likewise.
20877         (lra_eliminate_regs_1): Likewise.
20878         * recog.c (general_operand): Likewise.
20879         * ree.c (combine_reaching_defs): Likewise.
20880         * reload.c (push_reload): Likewise.
20881         (find_reloads): Likewise.
20882         * reload1.c (elimination_effects): Likewise.
20883         (compute_reload_subreg_offset): Likewise.
20884         (choose_reload_regs): Likewise.
20885         * rtlanal.c (subreg_lsb_1): Likewise.
20886         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20887         (simplify_subreg): Likewise.
20888         * var-tracking.c (track_loc_p): Likewise.
20889         * emit-rtl.c (byte_lowpart_offset): Likewise.
20890         (paradoxical_subreg_p): Delete out-of-line definition.
20892 2017-08-22  Jeff Law  <law@redhat.com>
20894         PR tree-optimization/81741
20895         PR tree-optimization/71947
20896         * tree-ssa-dom.c: Include tree-inline.h.
20897         (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20898         equivalences if one is more expensive to compute than the other.
20899         * tree-ssa-scopedtables.h (class const_or_copies): Make
20900         record_const_or_copy_raw method private.
20901         (class avail_exprs_stack): New method simplify_binary_operation.
20902         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20903         avail_exprs_stack::simplify_binary_operation as needed.
20904         (avail_exprs_stack::simplify_binary_operation): New function.
20906 2017-08-22  Sebastian Huber  <sebastian.huber@embedded-brains.de>
20908         * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20909         * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20910         (DOT_SYMBOLS): Likewise.
20911         (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20912         (RELOCATABLE_NEEDS_FIXUP): Likewise.
20913         (RS6000_ABI_NAME): Likewise.
20914         (TARGET_CMODEL): Likewise.
20915         (TOC_SECTION_ASM_OP): Likewise.
20916         (SET_CMODEL): New macro.
20917         (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
20919 2017-08-22  Richard Biener  <rguenther@suse.de>
20921         * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
20922         to allow for free-lang-data replacements similar to verify_type_variant.
20924 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20925             Alan Hayward  <alan.hayward@arm.com>
20926             David Sherwood  <david.sherwood@arm.com>
20928         * config/aarch64/aarch64.md (casesi): Use DImode rather than
20929         VOIDmode for the LABEL_REF.
20931 2017-08-22  Richard Biener  <rguenther@suse.de>
20933         * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
20935 2017-08-22  Richard Biener  <rguenther@suse.de>
20937         * common.opt (feliminate-dwarf2-dups): Ignore.
20938         * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
20939         * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
20940         same_die_p_wrap, compute_section_prefix,
20941         is_symbol_die, assign_symbol_names, break_out_includes): Remove.
20942         (comdat_symbol_id, comdat_symbol_number): Likewise.
20943         (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
20944         Likewise.
20945         (check_duplicate_cu, record_comdat_symbol_number): Likewise.
20946         (output_die): Mark unreachable path unreachable.
20947         (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
20948         (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
20949         (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
20950         (dwarf2out_early_finish): Likewise.
20952 2017-08-22  Aldy Hernandez  <aldyh@redhat.com>
20954         * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
20956 2017-08-22  Georg-Johann Lay  <avr@gjlay.de>
20958         PR target/81910
20959         * config/avr/avr.c (avr_handle_addr_attribute): Early return if
20960         not VAR_P. Filter attribute warnings with OPT_Wattributes.
20961         (avr_attribute_table) <io, io_low, address>: Initialize
20962         .decl_required with true.
20964 2017-08-21  Michael Meissner  <meissner@linux.vnet.ibm.com>
20966         * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
20967         undocumented debugging options.
20968         (-mvsx-scalar-double): Likewise.
20969         (-mallow-df-permute): Likewise.
20970         (-mvectorize-builtins): Likewise.
20971         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20972         (rs6000_builtin_vectorized_function): Likewise.
20973         (rs6000_builtin_md_vectorized_function): Likewise.
20974         (rs6000_opt_vars): Likewise.
20976 2017-08-21  Uros Bizjak  <ubizjak@gmail.com>
20978         PR target/46091
20979         * config/i386/i386.md (*btsq_imm): Rename from *btsq.
20980         (*btrq_imm): Rename from *btrq.
20981         (*btcq_imm): Rename from *btcq.
20982         (btsc): New code attribute.
20983         (*<btsc><mode>): New insn pattern.
20984         (*btr<mode>): Ditto.
20985         (*<btsc><mode>_mask): New insn_and_split pattern.
20986         (*btr<mode>_mask): Ditto.
20988 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20989             Alan Hayward  <alan.hayward@arm.com>
20990             David Sherwood  <david.sherwood@arm.com>
20992         * function.c (pad_below): Simplify padding calculation.
20994 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20995             Alan Hayward  <alan.hayward@arm.com>
20996             David Sherwood  <david.sherwood@arm.com>
20998         * target.def (function_prologue): Remove frame size argument.
20999         (function_epilogue): Likewise.
21000         * doc/tm.texi: Regenerate.
21001         * final.c (final_start_function): Update call to function_prologue.
21002         (final_end_function): Update call to function_epilogue.
21003         (default_function_pro_epilogue): Remove frame size argument.
21004         * output.h (default_function_pro_epilogue): Likewise.
21005         * config/arm/arm.c (arm_output_function_epilogue): Likewise.
21006         (arm_output_function_prologue): Likewise.
21007         * config/frv/frv.c (frv_function_prologue): Likewise.
21008         (frv_function_epilogue): Likewise.
21009         * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
21010         * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
21011         (ia64_output_function_epilogue): Likewise.
21012         * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
21013         (m32r_output_function_epilogue): Likewise.
21014         * config/microblaze/microblaze.c (microblaze_function_prologue)
21015         (microblaze_function_epilogue): Likewise.
21016         * config/mips/mips.c (mips_output_function_prologue): Likewise.
21017         (mips_output_function_epilogue): Likewise.
21018         * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
21019         (mmix_target_asm_function_epilogue): Likewise.
21020         * config/msp430/msp430.c (msp430_start_function): Likewise.
21021         * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
21022         (nds32_asm_function_epilogue): Likewise.
21023         * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
21024         * config/pa/pa.c (pa_output_function_prologue): Likewise.
21025         (pa_output_function_epilogue): Likewise.
21026         * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
21027         (rs6000_output_function_epilogue): Likewise.
21028         * config/rl78/rl78.c (rl78_start_function): Likewise.
21029         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
21030         (rs6000_output_function_epilogue): Likewise.
21031         * config/rx/rx.c (rx_output_function_prologue): Likewise.
21032         * config/sh/sh.c (sh_output_function_epilogue): Likewise.
21033         * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
21034         (sparc_asm_function_epilogue): Likewise.
21036 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21038         * tree.h (type_has_mode_precision_p): New function.
21039         * convert.c (convert_to_integer_1): Use it.
21040         * expr.c (expand_expr_real_2): Likewise.
21041         (expand_expr_real_1): Likewise.
21042         * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
21043         * match.pd: Likewise.
21044         * tree-ssa-forwprop.c (simplify_rotate): Likewise.
21045         * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
21046         * tree-tailcall.c (process_assignment): Likewise.
21047         * tree-vect-loop.c (vectorizable_reduction): Likewise.
21048         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
21049         (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
21050         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
21051         (vectorizable_assignment): Likewise.
21052         (vectorizable_shift): Likewise.
21053         (vectorizable_operation): Likewise.
21054         * tree-vrp.c (register_edge_assert_for_2): Likewise.
21056 2017-08-21  Wilco Dijkstra  <wdijkstr@arm.com>
21058         * match.pd: Add pow (C, x) simplification.
21060 2017-08-21  Richard Biener  <rguenther@suse.de>
21062         PR tree-optimization/81900
21063         * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
21064         for blocks with abnormal predecessors.
21065         (compute_antic): Do not set visited flag prematurely.
21067 2017-08-21  Georg-Johann Lay  <avr@gjlay.de>
21069         PR target/79883
21070         * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
21072 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21074         * stor-layout.h (vector_type_mode): Move to...
21075         * tree.h (vector_type_mode): ...here.
21076         * stor-layout.c (vector_type_mode): Move to...
21077         * tree.c (vector_type_mode): ...here.  Include rtl.h and regs.h.
21079 2017-08-21  Richard Biener  <rguenther@suse.de>
21081         * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
21082         register_external_die hooks.
21083         (debug_false_tree_charstarstar_uhwistar): Declare.
21084         (debug_nothing_tree_charstar_uhwi): Likewise.
21085         * debug.c (do_nothing_debug_hooks): Adjust.
21086         (debug_false_tree_charstarstar_uhwistar): New do nothing.
21087         (debug_nothing_tree_charstar_uhwi): Likewise.
21088         * dbxout.c (dbx_debug_hooks): Adjust.
21089         (xcoff_debug_hooks): Likewise.
21090         * sdbout.c (sdb_debug_hooks): Likewise.
21091         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
21092         * dwarf2out.c (macinfo_label_base): New global.
21093         (dwarf2out_register_external_die): New function for the
21094         register_external_die hook.
21095         (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
21096         (dwarf2_debug_hooks): Use them.
21097         (dwarf2_lineno_debug_hooks): Adjust.
21098         (struct die_struct): Add with_offset flag.
21099         (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
21100         DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
21101         DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
21102         DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
21103         DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
21104         DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
21105         defining section names for the early LTO debug variants.
21106         (reset_indirect_string): New helper.
21107         (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
21108         (print_dw_val): Add support for offsetted symbol references.
21109         (get_ultimate_context): Split out from is_cxx.
21110         (is_cxx): Use get_ultimate_context.
21111         (is_fortran): Add decl overload.
21112         (compute_comp_unit_symbol): Split out worker from
21113         compute_section_prefix.
21114         (compute_section_prefix): Call compute_comp_unit_symbol and
21115         set comdat_type_p here.
21116         (output_die): Skip DIE symbol output for the LTO added one.
21117         Handle DIE symbol references with offset.
21118         (output_comp_unit): Guard section name mangling properly.
21119         For LTO debug sections emit a symbol at the section beginning
21120         which we use to refer to its DIEs.
21121         (add_abstract_origin_attribute): For DIEs registered via
21122         dwarf2out_register_external_die directly refer to the early
21123         DIE rather than indirectly through the shadow one we created.
21124         Remove obsolete call to dwarf2out_abstract_function for
21125         non-function/block origins.
21126         (gen_array_type_die): When generating early LTO debug do
21127         not emit DW_AT_string_length.
21128         (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
21129         late when in LTO.  As suggested place a gcc_unreachable for
21130         the DECL_ABSTRACT_P case.
21131         (gen_subprogram_die): Avoid another specification DIE
21132         for early built declarations/definitions for the late LTO case.
21133         (gen_variable_die): Add type references for late duplicated VLA dies
21134         when in late LTO.
21135         (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
21136         we have the abstract instance already.
21137         (process_scope_var): Adjust decl DIE contexts in LTO which
21138         first puts them in limbo.
21139         (gen_decl_die): Do not generate type DIEs late apart from
21140         types for VLAs or for decls we do not yet have a DIE.  Do not
21141         call dwarf2out_abstract_function late.
21142         (dwarf2out_early_global_decl): Make sure to create DIEs
21143         for abstract instances of a decl first.
21144         (dwarf2out_late_global_decl): Adjust comment.
21145         (output_macinfo_op): With multiple macro sections use
21146         macinfo_label_base to distinguish labels.
21147         (output_macinfo): Likewise.  Update macinfo_label_base.
21148         Pass in the line info label.
21149         (note_variable_value_in_expr): When generating LTO resolve
21150         all variable values here by generating DIEs as needed.
21151         (init_sections_and_labels): Add early LTO debug flag parameter
21152         and generate different sections and names if set.  Add generation
21153         counter for the labels so we can have multiple of them.
21154         (reset_dies): Helper to allow DIEs to be output multiple times.
21155         (dwarf2out_finish): When outputting DIEs to the fat part of an
21156         LTO object first reset DIEs.
21157         (dwarf2out_early_finish): Output early DIEs when generating LTO.
21158         (modified_type_die): Check for decl_ultimate_origin being self
21159         before recursing.
21160         (gen_type_die_with_usage): Likewise.
21161         (gen_typedef_die): Allow decl_ultimate_origin being self.
21162         (set_decl_abstract_flags): Remove.
21163         (set_block_abstract_flags): Likewise.
21164         (dwarf2out_abstract_function): Treat the early generated DIEs
21165         as the abstract copy and only add DW_AT_inline and
21166         DW_AT_artificial here and call set_decl_origin_self.
21167         If the DIE has an abstract origin don't do anything.
21168         * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
21169         if we have none yet (Go fails to build one, PR78628).
21170         (variably_modified_type_p): Prevent endless recursion for Ada
21171         cyclic pointer types.
21172         * lto-streamer-in.c: Include debug.h.
21173         (dref_queue): New global.
21174         (lto_read_tree_1): Stream in DIE references.
21175         (lto_input_tree): Register DIE references.
21176         (input_function): Stream DECL_DEBUG_ARGS.
21177         * lto-streamer-out.c: Include debug.h.
21178         (lto_write_tree_1): Output DIE references.
21179         (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
21180         Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
21181         (output_function): Stream DECL_DEBUG_ARGS.
21182         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21183         Stream DECL_ABSTRACT_ORIGIN.
21184         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21185         (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
21186         DECL_CONTEXT for file-scope decls.
21187         * lto-streamer.h (struct dref_entry): Declare.
21188         (dref_queue): Likewise.
21189         * cfgexpand.c (pass_expand::execute): Do not call the
21190         outlining_inline_function hook here.
21191         * lto-wrapper.c (debug_obj): New global.
21192         (tool_cleanup): Unlink it if required.
21193         (debug_objcopy): New function.
21194         (run_gcc): Handle early debug sections in the IL files by
21195         extracting them to separate files, partially linkin them and
21196         feeding the result back as result to the linker.
21197         * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
21198         DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
21199         DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
21200         sections into a separate segment.
21201         * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
21202         segments.
21203         (darwin_asm_dwarf_section): Likewise.
21204         (darwin_asm_output_dwarf_offset): Likewise.
21205         * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
21207 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21208             Alan Hayward  <alan.hayward@arm.com>
21209             David Sherwood  <david.sherwood@arm.com>
21211         * read-md.h (md_reader::record_potential_iterator_use): Replace
21212         pointer argument with an rtx and an index.
21213         * read-rtl.c (iterator_group::apply_iterator): Likewise.
21214         (apply_mode_iterator): Likewise.
21215         (apply_code_iterator): Likewise.
21216         (apply_int_iterator): Likewise.
21217         (apply_subst_iterator): Likewise.
21218         (record_iterator_use): Likewise.
21219         (record_attribute_use): Likewise.
21220         (md_reader::record_potential_iterator_use): Likewise.  Update calls
21221         to record_iterator_use and apply_iterator.
21222         (iterator_use): Replace ptr with x and index.
21223         (attribute_use): Likewise.
21224         (apply_attribute_uses): Update calls to apply_iterator.
21225         (apply_iterators): Likewise.  Update initialization of iterator_use.
21226         (rtx_reader::read_rtx_code): Update calls to record_iterator_use
21227         and record_potential_iterator_use.
21228         (rtx_reader::read_rtx_operand): Likewise.
21230 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21231             Alan Hayward  <alan.hayward@arm.com>
21232             David Sherwood  <david.sherwood@arm.com>
21234         * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
21235         CONST_WIDE_INT.
21237 2017-08-21  Richard Biener  <rguenther@suse.de>
21239         PR middle-end/81884
21240         * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
21241         at struct end conservatively when comparing common bases.
21243 2017-08-21  Richard Biener  <rguenther@suse.de>
21245         * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21246         (mem_ref_in_stmt): Remove.
21247         (determine_max_movement): Use ref index to get at the reference.
21248         (invariantness_dom_walker::before_dom_children): Deal with
21249         lim data already initialized.
21250         (gather_mem_refs_stmt): Initialize lim data and record ref index.
21252 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21254         * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21255         (TARGET_ISA_ROUND): Ditto.
21256         (TARGET_ROUND): Ditto.
21257         * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21258         * config/i386/i386.md: Ditto.
21259         * config/i386/sse.md: Ditto.
21260         * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21261         with OPTION_MASK_ISA_SSE4_1.
21263 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21265         PR target/81894
21266         * doc/extend.texi (x86 Built-in Functions): Correct the name of
21267         __builtin_ia32_lzcnt_u16.
21269 2017-08-18  Peter Bergner  <bergner@vnet.ibm.com>
21271         PR target/80210
21272         * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21273         (rs6000_set_current_function): Rewrite function to use it.
21275 2017-08-18  H.J. Lu  <hongjiu.lu@intel.com>
21277         PR c/53037
21278         * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21279         and TYPE_WARN_IF_NOT_ALIGN.
21280         * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21281         (handle_warn_if_not_align): New.
21282         (place_union_field): Call handle_warn_if_not_align.
21283         (place_field): Call handle_warn_if_not_align.
21284         Copy TYPE_WARN_IF_NOT_ALIGN.
21285         (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21286         (layout_type): Likewise.
21287         * tree-core.h (tree_type_common): Add warn_if_not_align.  Set
21288         spare to 18.
21289         (tree_decl_common): Add warn_if_not_align.
21290         * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21291         * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21292         (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21293         (DECL_WARN_IF_NOT_ALIGN): Likewise.
21294         (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21295         * doc/extend.texi: Document warn_if_not_aligned attribute.
21296         * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21298 2017-08-17  Martin Liska  <mliska@suse.cz>
21300         PR bootstrap/81864
21301         * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21302         (get_data_dependence): Use it as pointer type.
21303         (distribute_loop): Likewise.
21305 2017-08-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21307         * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21308         (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21309         (p8_vmrgow_<mode>_direct): New define_insn.
21310         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21311         handle endianness for vmrgew and vmrgow permute patterns.
21313 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21315         * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21316         * config/rs6000/rs6000-cpus.def: Remove comment.
21317         (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21318         (POWERPC_MASKS): Likewise.
21319         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21320         use of TARGET_VSX_TIMODE.
21321         (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21322         (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21323         TARGET_VSX_TIMODE.  Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21324         (rs6000_option_override_internal): Remove dead code.
21325         (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21326         (rs6000_legitimize_reload_address): Likewise.
21327         (rs6000_legitimate_address_p): Likewise.
21328         (rs6000_opt_masks): Delete "vsx-timode".
21329         (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21330         from function comment.
21331         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21332         * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21333         (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21334         condition.
21335         * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21336         * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21337         * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21338         (VSX_TI): Likewise.
21339         (VSX_M): Likewise.
21340         (define_peephole2): Likewise.
21342 2017-08-17  Martin Sebor  <msebor@redhat.com>
21344         PR c/81859
21345         * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21346         past the end of an array.
21347         (test_pp_format): Add test cases.
21349 2017-08-17  Richard Sandiford  <richard.sandiford@linaro.org>
21351         * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21352         missing ECF_NOTHROW flags.
21354 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21356         PR target/72804
21357         * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21358         operands residing in integer registers.
21359         (*vsx_le_perm_load_<mode>): Likewise.
21360         (*vsx_le_perm_store_<mode>): Likewise.
21361         (define_peephole2): Add peepholes to optimize the above.
21363 2017-08-17  Marek Polacek  <polacek@redhat.com>
21365         PR middle-end/81814
21366         * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21367         to mimic what shorten_compare did.  Change the return type to bool.
21368         (fold_cond_expr_with_comparison): Update call to
21369         operand_equal_for_comparison_p.
21370         (fold_ternary_loc): Likewise.
21372 2017-08-17  Jackson Woodruff  <jackson.woodruff@arm.com>
21374         * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21375         register.
21376         (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21378 2017-08-17  Richard Biener  <rguenther@suse.de>
21380         * tree-ssa-structalias.c (solve_graph): When propagating
21381         to successors update the graphs succ edges and avoid duplicate work.
21383 2017-08-17  Maxim Ostapenko  <m.ostapenko@samsung.com>
21385         PR target/81861
21386         * config/i386/i386.c (ix86_option_override_internal): Save target
21387         specific options after ix86_stack_protector_guard_reg was changed.
21389 2017-08-17  Richard Biener  <rguenther@suse.de>
21391         PR tree-optimization/81827
21392         * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21393         (new_var_info): Initialize it conservatively.
21394         (get_call_vi): Mark register vars.
21395         (new_scalar_tmp_constraint_exp): Likewise.
21396         (handle_rhs_call): Likewise.
21397         (handle_const_call): Likewise.
21398         (create_function_info_for): Likewise.
21399         (solve_constraints): Sort varinfos to separate register from
21400         non-register vars to pack points-to solution bitmaps during
21401         iteration.
21403 2017-08-17  Marek Polacek  <polacek@redhat.com>
21405         * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21407 2017-08-17  Richard Biener  <rguenther@suse.de>
21409         * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21410         to true when overflow is undefined and we saturated the result.
21412 2017-08-17  Alan Modra  <amodra@gmail.com>
21414         PR target/80938
21415         * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21416         Don't use store multiple if only one reg needs saving.
21417         (interesting_frame_related_regno): New function.
21418         (rs6000_frame_related): Don't emit frame info for regs that
21419         don't need saving.
21420         (rs6000_emit_epilogue): Likewise.
21422 2017-08-16  Nathan Sidwell  <nathan@acm.org>
21424         * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21425         * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21426         (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21427         * tree.c (free_lang_data_in_type): Use else-if chain.  Always
21428         clear TYPE_LANG_1.  Remove obsolete member-function stripping.
21429         (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21430         (verify_type): Adjust for TYPE_BINFO move.
21431         * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21432         process TYPE_BINFO directly.
21433         (hash_tree): Likewise.
21434         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21435         Likewise.
21436         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21437         Likewise.
21439 2017-08-16  David Malcolm  <dmalcolm@redhat.com>
21441         * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21443 2017-08-16  Uros Bizjak  <ubizjak@gmail.com>
21445         PR target/46091
21446         * config/i386/i386.md (*anddi_1_btr): Change predicates of
21447         operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21448         Add ix86_binary_operator_ok to insn constraint.
21449         (*iordi_1_bts): Ditto.
21450         (*xordi_1_btc): Ditto.
21451         (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21452         Update corresponding peephole2 pattern.
21453         (*btrq): Ditto.
21454         (*btcq): Ditto.
21456 2017-08-16  Bin Cheng  <bin.cheng@arm.com>
21458         PR tree-optimization/81832
21459         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21460         copy loop header which has IFN_LOOP_DIST_ALIAS call.
21462 2017-08-16  Marek Polacek  <polacek@redhat.com>
21464         PR middle/81695
21465         * fold-const.c (fold_indirect_ref_1): Restore original behavior
21466         regarding size_zero_node.
21468 2017-08-16  Martin Liska  <mliska@suse.cz>
21470         PR target/81753
21471         * config.gcc: Respect previously set extra_objs in case
21472         of darwin target.
21474 2017-08-16  Richard Sandiford  <richard.sandiford@linaro.org>
21476         PR tree-optimization/81835
21477         * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21478         the phi SSA_NAME.  Check that the condition in a COND_EXPR does
21479         not depend on the phi.
21481 2017-08-16  Alan Modra  <amodra@gmail.com>
21483         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21484         dead code.
21486 2017-08-16  Alan Modra  <amodra@gmail.com>
21488         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21489         (save_reg_p): ..into this.  Update all callers.
21490         (first_reg_to_save): Simplify.
21492 2017-08-16  Alan Modra  <amodra@gmail.com>
21494         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21495         fixed regs.
21497 2017-08-15  Joseph Myers  <joseph@codesourcery.com>
21499         PR target/78460
21500         PR target/67712
21501         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21502         constant count if that count is less than 32.
21504 2017-08-15  Nathan Sidwell  <nathan@acm.org>
21506         * gcc.c (execute): Emit friendlier message if inferior is killed
21507         by an external cause.
21509 2017-08-15  Richard Biener  <rguenther@suse.de>
21511         PR tree-optimization/81790
21512         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21513         CONSTRUCTORs from simplifying and VN.
21515 2017-08-14  Martin Sebor  <msebor@redhat.com>
21517         * builtin-attrs.def: Add comments.
21519 2017-08-14  Martin Sebor  <msebor@redhat.com>
21521         PR c/81117
21522         * doc/extend.texi (attribute nonstring): Document new attribute.
21524 2017-08-14  Martin Sebor  <msebor@redhat.com>
21526         PR c/81117
21527         * tree-diagnostic.c (default_tree_printer): Handle %G.
21528         * gimple-pretty-print.h (percent_G_format): Declare new function.
21529         * gimple-pretty-print.c (percent_G_format): Define.
21530         * tree-pretty-print.c (percent_K_format): Add argument.
21532 2017-08-14  Martin Sebor  <msebor@redhat.com>
21534         PR translation/79998
21535         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21536         Remove a stray space.
21538 2017-08-14  Uros Bizjak  <ubizjak@gmail.com>
21540         PR target/46091
21541         * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21542         (*iordi_1_bts): Ditto.
21543         (*xordi_1_btc): Ditto.
21545 2017-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21547         PR target/79845
21548         * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21549         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21550         Likewise.
21551         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21552         quoted strings, and make more translator-friendly.
21553         (darwin_rs6000_override_options): Likewise.
21554         (rs6000_option_override_internal): Likewise.
21555         (rs6000_return_in_memory): Fix overlong line.
21556         (init_cmulative_args): Use quoted strings, and make more
21557         translator-friendly.
21558         (rs6000_pass_by_reference): Fix overlong line.
21559         (def_builtin): Use quoted strings.
21560         (altivec_expand_predicate_builtin): Use quoted strings, and make
21561         more translator-friendly.
21562         (htm_expand_builtin): Use quoted strings.
21563         (cpu_expand_builtin): Use quoted strings, and make more
21564         translator-friendly.
21565         (altivec_expand_builtin): Likewise.
21566         (paired_expand_predicate_builtin): Likewise.
21567         (rs6000_invalid_builtin): Likewise.
21568         (builtin_function_type): Use quoted strings.
21569         (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21570         more translator-friendly.
21571         (rs6000_trampoline_init): Likewise.
21572         (rs6000_handle_altivec_attribute): Likewise.
21573         (rs6000_inner_target_options): Use quoted strings.
21574         (rs6000_disable_incompatible_switches): Likewise.
21575         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21576         strings, and make more translator-friendly.
21577         (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21579 2017-08-14  Bin Cheng  <bin.cheng@arm.com>
21581         PR tree-optimization/81799
21582         * tree-loop-distribution.c (version_loop_by_alias_check): Force
21583         cond_expr to simple gimple operand.
21585 2017-08-14  Wilco Dijkstra  <wdijkstr@arm.com>
21587         PR middle-end/46932
21588         * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21590 2017-08-14  Georg-Johann Lay  <avr@gjlay.de>
21592         PR target/81754
21593         PR target/81268
21594         * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21595         * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21596         TARGET_GASISR_PROLOGUES.
21597         * config/avr/avr.c (avr_option_override): Same.
21598         (avr_pass_pre_proep::execute): Same.
21600 2017-08-13  H.J. Lu  <hongjiu.lu@intel.com>
21602         PR target/81820
21603         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21604         frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21606 2017-08-13  Uros Bizjak  <ubizjak@gmail.com>
21608         * config/i386/i386.md (*load_tp_<mode>): Redefine as
21609         define_insn_and_split.  Split to a memory load from 0 in
21610         DEFAULT_TLS_SEG_REG address space.  Merge with *load_tp_x32
21611         using PTR mode iterator.
21612         (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21613         Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21614         (*add_tp_<mode>): Redefine as define_insn_and_split.
21615         Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21616         address space.  Merge with *add_tp_x32 using PTR mode iterator.
21617         (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21618         Split to an add with a  memory load from 0 in
21619         DEFAULT_TLS_SEG_REG address space.
21621 2017-08-12  Andrew Pinski  <apinski@cavium.com>
21623         * config/aarch64/aarch64-option-extensions.def (rdma):
21624         Fix feature string to what Linux prints out in /proc/cpuinfo.
21626 2017-08-12  Pierre-Marie de Rodat  <derodat@adacore.com>
21628         PR ada/79542
21629         * dwarf2out.c (modified_type_die): For C typedef types that have
21630         an ultimate origin, process the ultimate origin instead of the
21631         input type.
21632         (gen_typedef_die): Assert that input DECLs have no ultimate
21633         origin.
21634         (gen_type_die_with_usage): For typedef variants that have an
21635         ultimate origin, just call gen_decl_die on the original DECL.
21636         (process_scope_var): Avoid creating DIEs for local typedefs and
21637         concrete static variables.
21639 2017-08-12  Alan Modra  <amodra@gmail.com>
21641         PR target/81170
21642         PR target/81295
21643         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21644         match gnu-user.h startfile.
21645         (ENDFILE_LINUX_SPEC): Similarly.
21647 2017-08-11  Thomas Schwinge  <thomas@codesourcery.com>
21649         PR lto/81430
21650         * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21651         Remove function.
21652         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21654 2017-08-11  Tamar Christina  <tamar.christina@arm.com>
21655         * config/aarch64/aarch64.md (mov<mode>): Change.
21656         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21657         aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21658         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21660 2017-08-11  Eric Botcazou  <ebotcazou@adacore.com>
21662         * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21663         for storage order barriers.
21665 2017-08-11  Martin Liska  <mliska@suse.cz>
21667         PR tree-opt/79987
21668         * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21669         variables of void type.
21671 2017-08-11  Martin Liska  <mliska@suse.cz>
21673         * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21674         TARGET_SUPPORTS_ALIASES.
21675         * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21676         * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21677         (optimize_weakref): Likewise.
21678         * symtab.c (symtab_node::noninterposable_alias): Likewise.
21679         * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21680         * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21682 2017-08-11  Martin Liska  <mliska@suse.cz>
21684         PR ipa/81213
21685         * config/i386/i386.c (make_resolver_func): Do complete
21686         refactoring of the function.
21688 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21690         PR target/81708
21691         * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21692         * config/i386/i386.c (ix86_stack_protect_guard): Use
21693         ix86_stack_protect_guard_symbol_str to generate varible declaration.
21694         * doc/invoke.texi (x86 Options): Document
21695         -mstack-protector-guard-symbol= option.
21697 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21699         * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21700         * config/i386/i386.c (ix86_split_stack_guard): New function.
21701         (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21702         (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21703         (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21704         (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21705         * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21706         (split_stack_space_check): Call ix86_split_stack_guard.
21708 2017-08-10  Martin Sebor  <msebor@redhat.com>
21710         * print-tree.c (print_node): Print location using the established
21711         format %s:%i%i.
21712         Replace spaces with colons.
21713         (debug_raw, debug): Ditto.
21715 2017-08-10  Martin Sebor  <msebor@redhat.com>
21717         PR c++/81586
21718         * pretty-print.c (pp_format): Correct the handling of %s precision.
21720 2017-08-10  H.J. Lu  <hongjiu.lu@intel.com>
21722         PR target/81736
21723         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21724         to ...
21725         (ix86_finalize_stack_frame_flags): This.  Also clear
21726         frame_pointer_needed if -fno-omit-frame-pointer is used without
21727         stack access.
21728         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21729         with ix86_finalize_stack_frame_flags.
21730         (ix86_expand_epilogue): Likewise.
21731         (ix86_expand_split_stack_prologue): Likewise.
21732         * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21734 2017-08-10  Martin Liska  <mliska@suse.cz>
21736         PR c++/81355
21737         * c-attribs.c (handle_target_attribute):
21738         Report warning for an empty string argument of target attribute.
21740 2017-08-09  Jakub Jelinek  <jakub@redhat.com>
21742         PR c/81687
21743         * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21744         LABEL_DECLs.
21745         * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21746         or DECL_NONLOCAL labels.
21747         (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21748         or DECL_NONLOCAL labels here.
21750 2017-08-09  Will Schmidt  <will_schmidt@vnet.ibm.com>
21752         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21753         to indicate when early gimple folding has been disabled.
21754         (rs6000_gimple_fold_builtin): Add debug content.
21755         (rs6000_invalid_builtin): Fix whitespace.
21756         (rs6000_expand_builtin): Fix whitespace.
21757         * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21759 2017-08-09  Segher Boessenkool  <segher@kernel.crashing.org>
21761         PR target/80938
21762         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21763         SAVE_MULTIPLE if not all the registers that saves, should be saved.
21765 2017-08-09  Jim Wilson  <jim.wilson@linaro.org>
21767         * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21768         (qdf24xx): Likewise.
21769         * config/aarch64/aarch64.md: Include falkor.md.
21770         * config/aarch64/falkor.md: New.
21772 2017-08-09  Marek Polacek  <polacek@redhat.com>
21774         PR c/81233
21775         * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21776         * diagnostic.c (emit_diagnostic): Add a comment.
21777         (emit_diagnostic_valist): New function.
21779 2017-08-09  Marek Polacek  <polacek@redhat.com>
21781         PR c/81417
21782         * input.c (make_location): New overload.
21783         * input.h (make_location): Declare.
21785 2017-08-08  Alan Modra  <amodra@gmail.com>
21786             H.J. Lu  <hongjiu.lu@intel.com>
21788         PR driver/81523
21789         * gcc.c (NO_PIE_SPEC): Delete.
21790         (PIE_SPEC): Define as !no-pie/pie.  Move static|shared|r
21791         exclusion..
21792         (LINK_PIE_SPEC): ..to here.
21793         (LINK_COMMAND_SPEC): Support -no-pie.
21794         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21795         chain of crtbegin*.o selection, update for PIE_SPEC changes and
21796         format.
21797         (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21798         * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21799         (ENDFILE_CRTEND_SPEC): Similarly.
21801 2017-08-08  Uros Bizjak  <ubizjak@gmail.com>
21803         PR target/81708
21804         * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21805         (mstack-protector-guard-offset=): Ditto.
21806         * config/i386/i386.c (ix86_option_override): Handle
21807         -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21808         options.
21809         (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21810         ix86_stack_protect_guard_offset variables.
21811         (TARGET_STACK_PROTECT_GUARD): Always define.
21812         * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21813         and -mstack-protector-guard-offset= options.
21815 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21817         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21818         boundary case for the last candidate.
21820 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21822         * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21823         * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21825 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21827         PR middle-end/19706
21828         * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21829         * config/aarch64/aarch64-builtins.c
21830         (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21831         * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21832         * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21834 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21835             Andrew Pinski <pinskia@gmail.com>
21837         PR middle-end/19706
21838         * internal-fn.def (XORSIGN): New.
21839         * optabs.def (xorsign_optab): New.
21840         * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21841         (convert_expand_mult_copysign): New.
21842         (pass_optimize_widening_mul::execute): Call
21843         convert_expand_mult_copysign.
21845 2017-08-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21847         PR tree-optimization/81354
21848         * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21849         Insert on edges rather than explicitly creating landing pads.
21850         (analyze_candidates_and_replace): Commit edge inserts.
21852 2017-08-08  Richard Biener  <rguenther@suse.de>
21854         PR middle-end/81719
21855         * tree-ssa-loop-niter.c: Include tree-dfa.h.
21856         (expand_simple_operations): Also look through ADDR_EXPRs with
21857         MEM_REF bases treating them as POINTER_PLUS_EXPR.
21859 2017-08-08  Richard Biener  <rguenther@suse.de>
21861         PR tree-optimization/81723
21862         * tree-vect-slp.c (struct bst_traits): New hash traits.
21863         (bst_fail): New global.
21864         (vect_build_slp_tree_2): New worker, split out from ...
21865         (vect_build_slp_tree): ... this now wrapping it with using
21866         bst_fail set to cache SLP tree build fails.  Properly handle
21867         max_tree_size.
21868         (vect_analyze_slp_instance): Allocate and free bst_fail.
21870 2017-08-08  Martin Liska  <mliska@suse.cz>
21872         PR tree-opt/81696
21873         * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21874         LABEL_DECLs that can be from a different function.
21876 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21878         PR tree-optimization/81744
21879         * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21880         loop's number of iterations.
21882 2017-08-08  Martin Liska  <mliska@suse.cz>
21884         * asan.c: Include header files.
21885         * attribs.c (build_decl_attribute_variant): New function moved
21886         from tree.[ch].
21887         (build_type_attribute_qual_variant): Likewise.
21888         (cmp_attrib_identifiers): Likewise.
21889         (simple_cst_list_equal): Likewise.
21890         (omp_declare_simd_clauses_equal): Likewise.
21891         (attribute_value_equal): Likewise.
21892         (comp_type_attributes): Likewise.
21893         (build_type_attribute_variant): Likewise.
21894         (lookup_ident_attribute): Likewise.
21895         (remove_attribute): Likewise.
21896         (merge_attributes): Likewise.
21897         (merge_type_attributes): Likewise.
21898         (merge_decl_attributes): Likewise.
21899         (merge_dllimport_decl_attributes): Likewise.
21900         (handle_dll_attribute): Likewise.
21901         (attribute_list_equal): Likewise.
21902         (attribute_list_contained): Likewise.
21903         * attribs.h (lookup_attribute): New function moved from tree.[ch].
21904         (lookup_attribute_by_prefix): Likewise.
21905         * bb-reorder.c: Include header files.
21906         * builtins.c: Likewise.
21907         * calls.c: Likewise.
21908         * cfgexpand.c: Likewise.
21909         * cgraph.c: Likewise.
21910         * cgraphunit.c: Likewise.
21911         * convert.c: Likewise.
21912         * dwarf2out.c: Likewise.
21913         * final.c: Likewise.
21914         * fold-const.c: Likewise.
21915         * function.c: Likewise.
21916         * gimple-expr.c: Likewise.
21917         * gimple-fold.c: Likewise.
21918         * gimple-pretty-print.c: Likewise.
21919         * gimple.c: Likewise.
21920         * gimplify.c: Likewise.
21921         * hsa-common.c: Likewise.
21922         * hsa-gen.c: Likewise.
21923         * internal-fn.c: Likewise.
21924         * ipa-chkp.c: Likewise.
21925         * ipa-cp.c: Likewise.
21926         * ipa-devirt.c: Likewise.
21927         * ipa-fnsummary.c: Likewise.
21928         * ipa-inline.c: Likewise.
21929         * ipa-visibility.c: Likewise.
21930         * ipa.c: Likewise.
21931         * lto-cgraph.c: Likewise.
21932         * omp-expand.c: Likewise.
21933         * omp-general.c: Likewise.
21934         * omp-low.c: Likewise.
21935         * omp-offload.c: Likewise.
21936         * omp-simd-clone.c: Likewise.
21937         * opts-global.c: Likewise.
21938         * passes.c: Likewise.
21939         * predict.c: Likewise.
21940         * sancov.c: Likewise.
21941         * sanopt.c: Likewise.
21942         * symtab.c: Likewise.
21943         * toplev.c: Likewise.
21944         * trans-mem.c: Likewise.
21945         * tree-chkp.c: Likewise.
21946         * tree-eh.c: Likewise.
21947         * tree-into-ssa.c: Likewise.
21948         * tree-object-size.c: Likewise.
21949         * tree-parloops.c: Likewise.
21950         * tree-profile.c: Likewise.
21951         * tree-ssa-ccp.c: Likewise.
21952         * tree-ssa-live.c: Likewise.
21953         * tree-ssa-loop.c: Likewise.
21954         * tree-ssa-sccvn.c: Likewise.
21955         * tree-ssa-structalias.c: Likewise.
21956         * tree-ssa.c: Likewise.
21957         * tree-streamer-in.c: Likewise.
21958         * tree-vectorizer.c: Likewise.
21959         * tree-vrp.c: Likewise.
21960         * tsan.c: Likewise.
21961         * ubsan.c: Likewise.
21962         * varasm.c: Likewise.
21963         * varpool.c: Likewise.
21964         * tree.c: Remove functions moved to attribs.[ch].
21965         * tree.h: Likewise.
21966         * config/aarch64/aarch64.c: Add attrs.h header file.
21967         * config/alpha/alpha.c: Likewise.
21968         * config/arc/arc.c: Likewise.
21969         * config/arm/arm.c: Likewise.
21970         * config/avr/avr.c: Likewise.
21971         * config/bfin/bfin.c: Likewise.
21972         * config/c6x/c6x.c: Likewise.
21973         * config/cr16/cr16.c: Likewise.
21974         * config/cris/cris.c: Likewise.
21975         * config/darwin.c: Likewise.
21976         * config/epiphany/epiphany.c: Likewise.
21977         * config/fr30/fr30.c: Likewise.
21978         * config/frv/frv.c: Likewise.
21979         * config/ft32/ft32.c: Likewise.
21980         * config/h8300/h8300.c: Likewise.
21981         * config/i386/winnt.c: Likewise.
21982         * config/ia64/ia64.c: Likewise.
21983         * config/iq2000/iq2000.c: Likewise.
21984         * config/lm32/lm32.c: Likewise.
21985         * config/m32c/m32c.c: Likewise.
21986         * config/m32r/m32r.c: Likewise.
21987         * config/m68k/m68k.c: Likewise.
21988         * config/mcore/mcore.c: Likewise.
21989         * config/microblaze/microblaze.c: Likewise.
21990         * config/mips/mips.c: Likewise.
21991         * config/mmix/mmix.c: Likewise.
21992         * config/mn10300/mn10300.c: Likewise.
21993         * config/moxie/moxie.c: Likewise.
21994         * config/msp430/msp430.c: Likewise.
21995         * config/nds32/nds32-isr.c: Likewise.
21996         * config/nds32/nds32.c: Likewise.
21997         * config/nios2/nios2.c: Likewise.
21998         * config/nvptx/nvptx.c: Likewise.
21999         * config/pa/pa.c: Likewise.
22000         * config/pdp11/pdp11.c: Likewise.
22001         * config/powerpcspe/powerpcspe.c: Likewise.
22002         * config/riscv/riscv.c: Likewise.
22003         * config/rl78/rl78.c: Likewise.
22004         * config/rx/rx.c: Likewise.
22005         * config/s390/s390.c: Likewise.
22006         * config/sh/sh.c: Likewise.
22007         * config/sol2.c: Likewise.
22008         * config/sparc/sparc.c: Likewise.
22009         * config/spu/spu.c: Likewise.
22010         * config/stormy16/stormy16.c: Likewise.
22011         * config/tilegx/tilegx.c: Likewise.
22012         * config/tilepro/tilepro.c: Likewise.
22013         * config/v850/v850.c: Likewise.
22014         * config/vax/vax.c: Likewise.
22015         * config/visium/visium.c: Likewise.
22016         * config/xtensa/xtensa.c: Likewise.
22018 2017-08-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
22020         PR target/81593
22021         * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
22022         constraints since the -mupper-regs-* switches have been
22023         eliminated.
22024         (vsx_concat_<mode>_1): New combiner insns to recognize inserting
22025         into a vector from a double word element that was extracted from
22026         another vector, and eliminate extra XXPERMDI instructions.
22027         (vsx_concat_<mode>_2): Likewise.
22028         (vsx_concat_<mode>_3): Likewise.
22029         (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
22030         concat to allow optimizing inserts from previous extracts.
22032 2017-08-07  Uros Bizjak  <ubizjak@gmail.com>
22034         * config/i386/i386.c (ix86_stack_protect_guard): Generate
22035         memory reference to a SSP offset in TLS address space.
22036         (ix86_print_operand) <case '@'>: Remove.
22037         (ix86_print_operand_punct_valid_p): Remove '@' code.
22038         * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
22039         UNSPEC_SP_TLS_TEST.
22040         (stack_tls_protect_set_<mode>): Remove.
22041         (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
22042         (stack_tls_protect_test_<mode>): Remove.
22043         (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
22045 2017-08-07  Olivier Hainque  <hainque@adacore.com>
22047         PR target/81755
22048         * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
22050 2017-08-07  Douglas Rupp  <rupp@adacore.com>
22052         * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
22053         variable was referenced as multidir in command.
22055 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22057         PR c/69389
22058         * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
22059         BIT_FIELD_REF.
22061 2017-08-07  Martin Liska  <mliska@suse.cz>
22063         * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
22064         * config/rl78/rl78.c: Add include of attribs.h.
22065         * config/sh/sh.c: Likewise.
22066         * config/v850/v850.c: Likewise.
22068 2017-08-07  Tom de Vries  <tom@codesourcery.com>
22070         PR middle-end/78266
22071         * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
22073 2017-08-07  Martin Liska  <mliska@suse.cz>
22075         * config/mips/mips.c: Include attribs.h.
22077 2017-08-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
22079         PR fortran/68829
22080         * doc/invoke.texi: Document change in behvaior for -Ofast for
22081         Fortran.
22083 2017-08-07  Wilco Dijkstra  <wdijkstr@arm.com>
22085         * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
22086         Use gen_frame_mem.
22087         (aarch64_pop_regs): Likewise.
22088         (aarch64_gen_load_pair): Likewise.
22089         (aarch64_save_callee_saves): Likewise.
22090         (aarch64_restore_callee_saves): Likewise.
22092 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22094         * config/i386/i386.c: Revert the last change.
22096 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22098         PR target/81736
22099         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
22100         to ...
22101         (ix86_finalize_stack_frame_flags): This.  Also clear
22102         frame_pointer_needed if -fno-omit-frame-pointer is used without
22103         stack access.
22104         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
22105         with ix86_finalize_stack_frame_flags.
22106         (ix86_expand_epilogue): Likewise.
22107         (ix86_expand_split_stack_prologue): Likewise.
22109 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22111         PR target/81743
22112         * config/i386/i386.c (get_builtin_code_for_version): Set priority
22113         to P_AES for Westmere.
22115 2017-08-07  Jonathan Yong  <10walls@gmail.com>
22117         * config/i386/mingw.opt (fset-stack-executable): Removed.
22118         * config/i386/cygming.opt (fset-stack-executable): Moved
22119         from mingw.opt.
22120         * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
22122 2017-08-07  Segher Boessenkool  <segher@kernel.crashing.org>
22124         * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
22126 2017-08-07  Marek Polacek  <polacek@redhat.com>
22128         PR middle-end/81737
22129         * fold-const.c (fold_indirect_ref_1): Check type_domain.
22131 2017-08-07  Martin Liska  <mliska@suse.cz>
22133         * attribs.h (canonicalize_attr_name): New function.
22134         (cmp_attribs): Move from c-format.c and adjusted.
22135         (is_attribute_p): Moved from tree.h.
22136         * tree-inline.c: Add new includes.
22137         * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
22138         (private_is_attribute_p): Remove.
22139         (private_lookup_attribute): Likewise.
22140         (private_lookup_attribute_by_prefix): Simplify.
22141         (remove_attribute): Use is_attribute_p.
22142         * tree.h: Remove removed declarations.
22144 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22146         PR middle-end/81698
22147         * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
22148         instead of computing it in the function.  Formatting fix.
22149         (expand_case): Don't rely on default_edge being the first edge,
22150         clear it if removing it, pass default_edge to
22151         emit_case_dispatch_table.
22152         (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
22153         fix.
22155 2017-08-06  Uros Bizjak  <ubizjak@gmail.com>
22157         * config/alpha/alpha.c (alpha_reorg): If trap is the last active
22158         insn in the function, emit NOP after the insn.
22160 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22162         * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
22163         and element loops.
22165 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22167         * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
22168         loop.
22170 2017-08-04  Yury Gribov  <tetra2005@gmail.com>
22172         PR tree-optimization/57371
22173         * match.pd: New pattern.
22175 2017-08-04  Marek Polacek  <polacek@redhat.com>
22177         PR middle-end/81695
22178         * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
22179         perform the computation in offset_int.
22181 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22183         PR tree-optimization/81136
22184         * tree-vectorizer.h: Include tree-hash-traits.h.
22185         (vec_base_alignments): New typedef.
22186         (vec_info): Add a base_alignments field.
22187         (vect_record_base_alignments): Declare.
22188         * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
22189         field.
22190         (DR_IS_CONDITIONAL_IN_STMT): New macro.
22191         (create_data_ref): Add an is_conditional_in_stmt argument.
22192         * tree-data-ref.c (create_data_ref): Likewise.  Use it to initialize
22193         the is_conditional_in_stmt field.
22194         (data_ref_loc): Add an is_conditional_in_stmt field.
22195         (get_references_in_stmt): Set the is_conditional_in_stmt field.
22196         (find_data_references_in_stmt): Update call to create_data_ref.
22197         (graphite_find_data_references_in_stmt): Likewise.
22198         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
22199         * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
22200         (vect_record_base_alignment): New function.
22201         (vect_record_base_alignments): Likewise.
22202         (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
22203         for nested statements even if we fail to compute a misalignment.
22204         Use pooled base alignments for unconditional references.
22205         (vect_find_same_alignment_drs): Compare base addresses instead
22206         of base objects.
22207         (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
22208         * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
22210 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22212         * tree-vectorizer.h (vec_info): Add a constructor and destructor.
22213         Add an explicit name for the enum.  Use auto_vec for slp_instances
22214         and grouped_stores.
22215         (_loop_vec_info): Add a constructor and destructor.  Use auto_vec
22216         for all vectors.
22217         (_bb_vec_info): Add a constructor and destructor.
22218         (vinfo_for_stmt): Return NULL for uids of -1 as well.
22219         (destroy_loop_vec_info): Delete.
22220         (vect_destroy_datarefs): Likewise.
22221         * tree-vectorizer.c (vect_destroy_datarefs): Delete.
22222         (vec_info::vec_info): New function.
22223         (vec_info::~vec_info): Likewise.
22224         (vectorize_loops): Use delete instead of destroy_loop_vec_info.
22225         * tree-parloops.c (gather_scalar_reductions): Use delete instead of
22226         destroy_loop_vec_info.
22227         * tree-vect-loop.c (new_loop_vec_info): Replace with...
22228         (_loop_vec_info::_loop_vec_info): ...this.
22229         (destroy_loop_vec_info): Replace with...
22230         (_loop_vec_info::~_loop_vec_info): ...this.  Unconditionally delete
22231         the stmt_vec_infos.  Leave handling of vec_info information to its
22232         destructor.  Remove explicit vector releases.
22233         (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
22234         (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
22235         * tree-vect-slp.c (new_bb_vec_info): Replace with...
22236         (_bb_vec_info::_bb_vec_info): ...this.  Don't reserve space in
22237         BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
22238         (destroy_bb_vec_info): Replace with...
22239         (_bb_vec_info::~_bb_vec_info): ...this.  Leave handling of vec_info
22240         information to its destructor.
22241         (vect_slp_analyze_bb_1): Use new and delete instead of
22242         new_bb_vec_info and destroy_bb_vec_info.
22243         (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22244         single delete.
22246 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22248         * tree-data-ref.h (subscript): Add access_fn field.
22249         (data_dependence_relation): Add could_be_independent_p.
22250         (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22251         (same_access_functions): Move to tree-data-ref.c.
22252         * tree-data-ref.c (ref_contains_union_access_p): New function.
22253         (access_fn_component_p): Likewise.
22254         (access_fn_components_comparable_p): Likewise.
22255         (dr_analyze_indices): Add a reference to access_fn_component_p.
22256         (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22257         DR_ACCESS_FN.
22258         (constant_access_functions): Likewise.
22259         (add_other_self_distances): Likewise.
22260         (same_access_functions): Likewise.  (Moved from tree-data-ref.h.)
22261         (initialize_data_dependence_relation): Use XCNEW and remove
22262         explicit zeroing of DDR_REVERSED_P.  Look for a subsequence
22263         of access functions that have the same type.  Allow the
22264         subsequence to end with different bases in some circumstances.
22265         Record the chosen access functions in SUB_ACCESS_FN.
22266         (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22267         a_index and b_index.  Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22268         (subscript_dependence_tester_1): Likewise dra and drb.
22269         (build_classic_dist_vector): Update calls accordingly.
22270         (subscript_dependence_tester): Likewise.
22271         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22272         DDR_COULD_BE_INDEPENDENT_P.
22273         * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22274         comp_alias_ddrs instead of may_alias_ddrs.
22275         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22276         New function.
22277         (vect_analyze_data_ref_dependence): Use it if
22278         DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22279         distance vectors if that fails.
22280         (dependence_distance_ge_vf): New function.
22281         (vect_prune_runtime_alias_test_list): Use it.  Don't clear
22282         LOOP_VINFO_MAY_ALIAS_DDRS.
22284 2017-08-04  Richard Biener  <rguenther@suse.de>
22286         PR middle-end/81705
22287         * fold-const.c (fold_binary_loc): Properly restrict
22288         minus_var0 && minus_var1 case when associating undefined overflow
22289         entities.
22291 2017-08-04  Tom de Vries  <tom@codesourcery.com>
22293         * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22295 2017-08-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22297         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22298         Don't start diagnostic messages with a capital letter.
22299         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22300         Likewise.
22301         (rs6000_invalid_builtin): Likewise.
22302         (rs6000_trampoline_init): Likewise.
22304 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22306         PR target/81621
22307         * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22308         after setting changeable df flags.
22310 2017-08-03  Richard Biener  <rguenther@suse.de>
22312         * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22313         up if the use is in USE - X.
22315 2017-08-03  Alexander Monakov  <amonakov@ispras.ru>
22317         * toplev.c (dumpfile.h): New include.
22318         (internal_error_reentered): New static function.  Use it...
22319         (internal_error_function): ...here to handle reentered internal_error.
22321 2017-08-03  Richard Biener  <rguenther@suse.de>
22323         PR middle-end/81148
22324         * fold-const.c (split_tree): Add minus_var and minus_con
22325         arguments, remove unused loc arg.  Never generate NEGATE_EXPRs
22326         here but always use minus_*.
22327         (associate_trees): Assert we never associate with MINUS_EXPR
22328         and NULL first operand.  Do not recurse for PLUS_EXPR operands
22329         when associating as MINUS_EXPR either.
22330         (fold_binary_loc): Track minus_var and minus_con.
22332 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22334         PR lto/81430
22335         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22336         ACCEL_COMPILER, apply finish_options on
22337         DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22339 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22341         PR target/81662
22342         * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22343         function_entry_patch_area_size > 0.
22345 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22347         PR driver/81650
22348         * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22349         instead of 10??LU, perform unit multiplication in wide_int,
22350         don't change alloc_object_size_limit if the limit is larger
22351         than SSIZE_MAX.
22353         PR tree-optimization/81655
22354         PR tree-optimization/81588
22355         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22356         the case when ranges[i].low and high are 1 for unsigned type with
22357         precision 1.
22359         PR middle-end/81052
22360         * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22361         (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22363 2017-08-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22365         * tree-vrp.h: Add include guard.
22367 2017-08-02  Uros Bizjak  <ubizjak@gmail.com>
22369         PR target/81644
22370         * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22371         (ud2): New insn pattern.
22372         * config/i386/i386.c (ix86_expand_epilogue):
22373         For naked functions, generate ud2 instead of trap insn.
22375 2017-08-02  Marek Polacek  <polacek@redhat.com>
22377         PR other/81667
22378         * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22380 2017-08-02  Tom de Vries  <tom@codesourcery.com>
22381             Cesar Philippidis  <cesar@codesourcery.com>
22383         * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22384         Add missing edge probabilities.
22386 2017-08-02  Tamar Christina  <tamar.christina@arm.com>
22388         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22389         Correct endianness.
22391 2017-08-02  Jakub Jelinek  <jakub@redhat.com>
22393         PR middle-end/79499
22394         * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22395         for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22396         of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22397         prologue_seq sequences - if any.
22399 2017-08-02  Richard Biener  <rguenther@suse.de>
22401         * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22402         via vectors if supported, integer extracts via punning if supported
22403         or otherwise vector extracts.
22405 2017-08-02  Richard Biener  <rguenther@suse.de>
22407         * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22408         into ...
22409         (bitmap_insert_into_set): ... this.
22411 2017-08-02  Richard Biener  <rguenther@suse.de>
22413         PR tree-optimization/81633
22414         Revert
22415         2015-08-17  Alan Hayward  <alan.hayward@arm.com>
22417         PR tree-optimization/71752
22418         * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22420 2017-08-01  Daniel Santos  <daniel.santos@pobox.com>
22422         * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22423         (machine_function::call_ms2sysv_pad_out): Remove field.
22424         * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22425         (ix86_compute_frame_layout): Likewise.
22427 2017-08-01  H.J. Lu  <hongjiu.lu@intel.com>
22429         PR target/81654
22430         * config/i386/i386.c (ix86_set_func_type): Disallow naked
22431         attribute with interrupt attribute.
22433 2017-08-01  Andrew Pinski  <apinski@cavium.com>
22435         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22436         BIT_INSERT_EXPR's operand 1
22437         to see if the types precision matches.
22439 2017-08-01  Martin Liska  <mliska@suse.cz>
22441         PR middle-end/70140
22442         * builtins.c (expand_builtin_memcpy_args): Remove.
22443         (expand_builtin_memcpy): Call newly added function
22444         expand_builtin_memory_copy_args.
22445         (expand_builtin_memcpy_with_bounds): Likewise.
22446         (expand_builtin_mempcpy): Remove last argument.
22447         (expand_builtin_mempcpy_with_bounds): Likewise.
22448         (expand_builtin_memory_copy_args): New function created from
22449         expand_builtin_mempcpy_args with small modifications.
22450         (expand_builtin_mempcpy_args): Remove.
22451         (expand_builtin_stpcpy): Remove unused argument.
22452         (expand_builtin): Likewise.
22453         (expand_builtin_with_bounds): Likewise.
22455 2017-08-01  Martin Liska  <mliska@suse.cz>
22457         Revert r250771
22458         Make mempcpy more optimal (PR middle-end/70140).
22460 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22462         PR target/81622
22463         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22464         __builtin_vec_cmpne verify both arguments are compatible vectors
22465         before looking at TYPE_MODE on the element type.  For __builtin_vec_ld
22466         verify arg1_type is a pointer or array type.  For __builtin_vec_st,
22467         move computation of aligned to after checking the argument types.
22468         Formatting fixes.
22470         PR target/80846
22471         * config/rs6000/vsx.md (vextract_fp_from_shorth,
22472         vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22473         calls.
22475 2017-08-01  Jerome Lambourg  <lambourg@adacore.com>
22476             Doug Rupp  <rupp@adacore.com>
22477             Olivier Hainque  <hainque@adacore.com>
22479         * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22480         well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22481         arm8 (arch v4).
22482         * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22483         for TARGET_OS_CPP_BUILTIN.
22484         (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22485         refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22486         arm_arch7.
22487         (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22488         passing required abi options to the assembler for EABI configurations.
22489         (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22490         of .text.hot and .text.unlikely sections for kernel modules when
22491         using ARM style exceptions.
22492         (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22493         options. Add EXTRA_CC1_SPEC.
22494         (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22495         toolchain options.
22496         (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22497         transition.
22498         (ARM_TARGET2_DWARF_FORMAT): Define.
22499         * config/arm/t-vxworks: Adjust multilib control to removal of the
22500         Diab command line options.
22502 2017-08-01  Martin Liska  <mliska@suse.cz>
22504         PR gcov-profile/81561
22505         * gcov.c (unblock): Make unblocking safe as we need to preserve
22506         index correspondence of blocks and block_lists.
22508 2017-08-01  Richard Biener  <rguenther@suse.de>
22510         PR tree-optimization/81181
22511         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22512         (compute_antic): ... end of iteration here.
22514 2017-08-01  James Greenhalgh  <james.greenhalgh@arm.com>
22516         * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22517         (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22518         (ftree-slp-vectorize): Likewise.
22519         * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22520         can no longer be set independent of flag_tree_loop_vectorize.
22521         * omp-general.c (emp_max_vf): Likewise.
22522         * opts.c (enable_fdo_optimizations): Remove references to
22523         flag_tree_vectorize, these are now implicit.
22524         (common_handle_option): Remove handling for OPT_ftree_vectorize,
22525         and leave it for the options machinery.
22527 2017-08-01  Martin Liska  <mliska@suse.cz>
22529         PR middle-end/70140
22530         * builtins.c (expand_builtin_memcpy_args): Remove.
22531         (expand_builtin_memcpy): Call newly added function
22532         expand_builtin_memory_copy_args.
22533         (expand_builtin_memcpy_with_bounds): Likewise.
22534         (expand_builtin_mempcpy): Remove last argument.
22535         (expand_builtin_mempcpy_with_bounds): Likewise.
22536         (expand_builtin_memory_copy_args): New function created from
22537         expand_builtin_mempcpy_args with small modifications.
22538         (expand_builtin_mempcpy_args): Remove.
22539         (expand_builtin_stpcpy): Remove unused argument.
22540         (expand_builtin): Likewise.
22541         (expand_builtin_with_bounds): Likewise.
22543 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22545         PR target/81641
22546         * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22547         print "ds:" only for immediates in generic address space.
22549 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22551         PR target/81639
22552         * config/i386/i386.c (ix86_funciton_naked): New prototype.
22553         (ix86_function_ok_for_sibcall): Return false for naked functions.
22555 2017-08-01  Richard Biener  <rguenther@suse.de>
22557         * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22558         (compute_antic): Seed worklist with exit block predecessors.
22559         * cfganal.c (dfs_find_deadend): For a cycle return the source
22560         of the edge closing it.
22562 2017-08-01  Tamar Christina  <tamar.christina@arm.com>
22564         * config/aarch64/aarch64.c
22565         (aarch64_can_const_movi_rtx_p): Move 0 check.
22567 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22569         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22570         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22571         above macro.
22572         * match.pd: Ditto in address comparison pattern.
22574 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22576         PR tree-optimization/81627
22577         * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22578         closed ssa form for store-store chain.
22580 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22582         PR tree-optimization/81620
22583         * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22584         for store-store chain.
22586 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22588         PR tree-optimization/81588
22589         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22590         ranges[i].in_p, invert comparison code ccode.  For >/>=,
22591         swap rhs1 and rhs2 and comparison code unconditionally,
22592         for </<= don't do that.  Don't swap rhs1/rhs2 again if
22593         ranges[i].in_p, instead invert comparison code ccode if
22594         opcode or oe->rank is BIT_IOR_EXPR.
22596         PR target/80846
22597         * optabs.def (vec_extract_optab, vec_init_optab): Change from
22598         a direct optab to conversion optab.
22599         * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22600         with GET_MODE_INNER as last argument instead of optab_handler.
22601         * expmed.c (extract_bit_field_1): Likewise.  Use vector from
22602         vector extraction if possible and optab is available.
22603         * expr.c (store_constructor): Use convert_optab_handler instead
22604         of optab_handler.  Use vector initialization from smaller
22605         vectors if possible and optab is available.
22606         * tree-vect-stmts.c (vectorizable_load): Likewise.
22607         * doc/md.texi (vec_extract, vec_init): Document that the optabs
22608         now have two modes.
22609         * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22610         of vec_init from half-sized vectors with the same element mode.
22611         * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22612         (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22613         (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22614         reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22615         reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22616         after mode in gen_vec_extract* calls.
22617         (vec_extract<mode>): Renamed to ...
22618         (vec_extract<mode><ssescalarmodelower>): ... this.
22619         (vec_extract<mode><ssehalfvecmodelower>): New expander.
22620         (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22621         element mode after mode in gen_vec_init* calls.
22622         (VEC_INIT_HALF_MODE): New mode iterator.
22623         (vec_init<mode>): Renamed to ...
22624         (vec_init<mode><ssescalarmodelower>): ... this.
22625         (vec_init<mode><ssehalfvecmodelower>): New expander.
22626         * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22627         (vec_extractv2sfsf): ... this.
22628         (vec_initv2sf): Renamed to ...
22629         (vec_initv2sfsf): ... this.
22630         (vec_extractv2si): Renamed to ...
22631         (vec_extractv2sisi): ... this.
22632         (vec_initv2si): Renamed to ...
22633         (vec_initv2sisi): ... this.
22634         (vec_extractv4hi): Renamed to ...
22635         (vec_extractv4hihi): ... this.
22636         (vec_initv4hi): Renamed to ...
22637         (vec_initv4hihi): ... this.
22638         (vec_extractv8qi): Renamed to ...
22639         (vec_extractv8qiqi): ... this.
22640         (vec_initv8qi): Renamed to ...
22641         (vec_initv8qiqi): ... this.
22642         * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22643         (vec_init<mode>): Renamed to ...
22644         (vec_init<mode><VEC_base_l>): ... this.
22645         (vec_extract<mode>): Renamed to ...
22646         (vec_extract<mode><VEC_base_l>): ... this.
22647         * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22648         (vec_initv2sfsf): ... this.
22649         * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22650         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22651         vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22652         element mode after mode in gen_vec_init* calls.
22653         * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22654         (vec_init<mode><Vel>): ... this.
22655         (vec_extract<mode>): Renamed to ...
22656         (vec_extract<mode><Vel>): ... this.
22657         * config/aarch64/iterators.md (Vel): New mode attribute.
22658         * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22659         Add element mode after mode in gen_vec_extract* calls.
22660         * config/s390/vector.md (non_vec_l): New mode attribute.
22661         (vec_extract<mode>): Renamed to ...
22662         (vec_extract<mode><non_vec_l>): ... this.
22663         (vec_init<mode>): Renamed to ...
22664         (vec_init<mode><non_vec_l>): ... this.
22665         * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22666         s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22667         vec_extract mode.
22668         * config/arm/iterators.md (V_elem_l): New mode attribute.
22669         * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22670         (vec_extract<mode><V_elem_l>): ... this.
22671         (vec_extractv2di): Renamed to ...
22672         (vec_extractv2didi): ... this.
22673         (vec_init<mode>): Renamed to ...
22674         (vec_init<mode><V_elem_l>): ... this.
22675         (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22676         reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22677         reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22678         Add element mode after gen_vec_extract* calls.
22679         * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22680         (vec_init<mode><unitmode>): ... this.
22681         (vec_extract<mode>): Renamed to ...
22682         (vec_extract<mode><unitmode>): ... this.
22683         * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22684         (vec_init<mode><unitmode>): ... this.
22685         * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22686         (vec_initv2sfsf): ... this.
22687         (vec_extractv2sf): Renamed to ...
22688         (vec_extractv2sfsf): ... this.
22689         (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22690         Add element mode after gen_vec_extract* calls.
22691         * config/mips/mips.md (unitmode): New mode iterator.
22692         * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22693         spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22694         * config/spu/spu.md (inner_l): New mode attribute.
22695         (vec_init<mode>): Renamed to ...
22696         (vec_init<mode><inner_l>): ... this.
22697         (vec_extract<mode>): Renamed to ...
22698         (vec_extract<mode><inner_l>): ... this.
22699         * config/sparc/sparc.md (veltmode): New mode iterator.
22700         (vec_init<VMALL:mode>): Renamed to ...
22701         (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22702         * config/ia64/vect.md (vec_initv2si): Renamed to ...
22703         (vec_initv2sisi): ... this.
22704         (vec_initv2sf): Renamed to ...
22705         (vec_initv2sfsf): ... this.
22706         (vec_extractv2sf): Renamed to ...
22707         (vec_extractv2sfsf): ... this.
22708         * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22709         (vec_init<mode>): Renamed to ...
22710         (vec_init<mode><VEC_base_l>): ... this.
22711         (vec_extract<mode>): Renamed to ...
22712         (vec_extract<mode><VEC_base_l>): ... this.
22713         * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22714         (vec_initv2sfsf): ... this.
22715         * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22716         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22717         vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22718         gen_vec_init* calls.
22720 2017-08-01  Richard Biener  <rguenther@suse.de>
22722         PR tree-optimization/81297
22723         * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22724         TREE_OVERFLOW from INTEGER_CSTs.
22726 2017-07-31  Segher Boessenkool  <segher@kernel.crashing.org>
22728         * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22730 2017-07-31  Carl Love  <cel@us.ibm.com>
22732         * config/rs6000/rs6000-c: Add support for built-in functions
22733         vector signed char vec_xl_be (signed long long, signed char *);
22734         vector unsigned char vec_xl_be (signed long long, unsigned char *);
22735         vector signed int vec_xl_be (signed long long, signed int *);
22736         vector unsigned int vec_xl_be (signed long long, unsigned int *);
22737         vector signed long long vec_xl_be (signed long long, signed long long *);
22738         vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22739         vector signed short vec_xl_be (signed long long, signed short *);
22740         vector unsigned short vec_xl_be (signed long long, unsigned short *);
22741         vector double vec_xl_be (signed long long, double *);
22742         vector float vec_xl_be (signed long long, float *);
22743         * config/rs6000/altivec.h (vec_xl_be): Add #define.
22744         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22745         XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22746         for the builtins.
22747         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22748         (altivec_expand_builtin): Add switch statement to call
22749         altivec_expand_xl_be for each builtin.
22750         (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22751         __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22752         __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22753         __builtin_vsx_le_be_v16qi.
22754         * doc/extend.texi: Update the built-in documentation file for the
22755         new built-in functions.
22757 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22759         PR target/25967
22760         * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22761         New function.
22762         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22764 2017-07-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22766         * config.gcc: Add z14.
22767         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22768         CPU model numbers for z13s and z14.
22769         * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22770         arch12 with z14.
22771         * config/s390/s390-opts.h (enum processor_type): Rename
22772         PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22773         * config/s390/s390.c (processor_table): Add field for CPU name to
22774         be passed to Binutils.
22775         (s390_asm_output_machine_for_arch): Use the new field in
22776         processor_table for Binutils.
22777         (s390_expand_builtin): Replace arch12 with z14.
22778         (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22779         (s390_get_sched_attrmask): Likewise.
22780         (s390_get_unit_mask): Likewise.
22781         * config/s390/s390.opt: Add z14 to processor_type enum.
22783 2017-07-31  Martin Jambor  <mjambor@suse.cz>
22785         PR hsa/81477
22786         * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22787         regardless of optimization level.
22789 2017-07-31  Jan Hubicka <hubicka@ucw.cz>
22790             Martin Liska  <mliska@suse.cz>
22792         * predict.def: Remove old comment and adjust probability.
22793         * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22794         PREDICT statements.
22796 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22798         PR target/25967
22799         * config/i386/i386.c (ix86_function_naked): New function.
22800         (ix86_can_use_return_insn_p): Return false for naked functions.
22801         (ix86_expand_prologue): Skip prologue for naked functions.
22802         (ix86_expand_epilogue): Skip epilogue for naked functions
22803         and emit trap instruction.
22804         (ix86_warn_func_return): New function.
22805         (ix86_attribute_table): Add "naked" attribute specification.
22806         (TARGET_WARN_FUNC_RETURN): Define.
22807         * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22809 2017-07-31  Martin Liska  <mliska@suse.cz>
22811         * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22812         (dump_gimple_bb_header): Always dump BB info.
22813         (pp_cfg_jump): Do not append info about BB when dumping a jump.
22815 2017-07-31  Martin Liska  <mliska@suse.cz>
22817         PR sanitize/81530
22818         * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22819         also with current_function_decl non-null equality.
22821 2017-07-31  Jakub Jelinek  <jakub@redhat.com>
22823         PR sanitizer/81604
22824         * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22825         change type to the element type, instead add eltype variable and
22826         use it where we are interested in the element type.
22828         PR tree-optimization/81603
22829         * ipa-polymorphic-call.c
22830         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22831         offset arithmetic in offset_int, bail out if the resulting bit offset
22832         doesn't fit into shwi.
22834 2017-07-31  Martin Liska  <mliska@suse.cz>
22836         * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22837         (gimplify_save_expr): Fix comment.
22839 2017-07-30  H.J. Lu  <hongjiu.lu@intel.com>
22841         PR target/79793
22842         * config/i386/i386.c (ix86_function_arg): Update arguments for
22843         exception handler.
22844         (ix86_compute_frame_layout): Set the initial stack offset to
22845         INCOMING_FRAME_SP_OFFSET.  Update red-zone offset with
22846         INCOMING_FRAME_SP_OFFSET.
22847         (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22848         stack before exception handler returns.
22849         * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22850         the 'ERROR_CODE' for exception handler.
22852 2017-07-30  Uros Bizjak  <ubizjak@gmail.com>
22854         * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22855         (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22856         (ASM_OUTPUT_REG_POP): Ditto.
22857         * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22858         instead of asm_fprintf to output pure string.
22860 2017-07-29  Jakub Jelinek  <jakub@redhat.com>
22862         * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22863         to imported_module_or_decl hook.
22864         (debug_nothing_tree_tree_tree_bool): Remove.
22865         (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22866         * debug.c (do_nothing_debug_hooks): Use
22867         debug_nothing_tree_tree_tree_bool_bool instead of
22868         debug_nothing_tree_tree_tree_bool.
22869         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22870         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22871         * sdbout.c (sdb_debug_hooks): Likewise.
22872         * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22873         (gen_namespace_die): Add DW_AT_export_symbols attribute if
22874         langhook wants it.
22875         (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22876         if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22877         attribute, don't add anything.
22879 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22881         * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22882         (fold_build2_stat_loc): Likewise.
22883         (fold_build3_stat_loc): Likewise.
22884         * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22885         (fold_build1_loc): Remove macro.
22886         (fold_build2_loc): Likewise.
22887         (fold_build3_loc): Likewise.
22889 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22891         * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22892         (gimple_build_debug_bind_source_stat): Likewise.
22893         * gimple.h (gimple_build_debug_bind): Remove macro.
22894         (gimple_build_debug_bind_source): Likewise.
22896 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22898         * bitmap.c (bitmap_alloc): Adjust.
22899         (bitmap_gc_alloc): Likewise.
22900         * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22902 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22904         * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22905         (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22906         * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22907         (bitmap_gc_alloc_stat): Likewise.
22908         (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22910 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22912         * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22913         * rtl.h (shallow_copy_rtx): Remove macro.
22915 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22917         * emit-rtl.c (gen_raw_REG): Adjust.
22918         * gengenrtl.c (gendef): Likewise.
22919         * rtl.c (rtx_alloc_stat): Remove _stat from name.
22920         * rtl.h (rtx_alloc): Remove macro.
22922 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22924         * tree.c (build_tree_list_vec_stat): Remove _stat from name.
22925         (build_tree_list_stat): Likewise.
22926         * tree.h (build_tree_list): Remove macro.
22927         (build_tree_list_vec): Likewise.
22929 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22931         * tree.c (make_vector_stat): Remove _stat from name.
22932         (build_vector_stat): Likewise.
22933         * tree.h (make_vector_stat): Remove macro.
22934         (build_vector_stat): Likewise.
22936 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22938         * tree.h (build_var_debug_value): Remove prototype.
22940 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22942         * tree.c (tree_cons_stat): Remove _stat from name.
22943         * tree.h (tree_cons): Remove macro.
22945 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22947         * tree.c (build_vl_exp_stat): Remove _stat from name.
22948         * tree.h (build_vl_exp): Remove macro.
22950 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22952         * tree.c (build_decl_stat): Remove _stat from name.
22953         * tree.h (build_decl): Remove macro.
22955 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22957         * gimple.c (gimple_build_with_ops_stat): Adjust.
22958         (gimple_alloc_stat): Remove _stat from name.
22959         * gimple.h (gimple_alloc): Remove macro.
22961 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22963         * tree.c (make_tree_vec_stat): Remove _stat from name.
22964         (grow_tree_vec_stat): Likewise.
22965         * tree.h (make_tree_vec_stat): Adjust prototype.
22966         (grow_tree_vec_stat): Likewise.
22967         (make_tree_vec): Remove macro.
22968         (grow_tree_vec): Likewise.
22970 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22972         * fold-const.c (fold_build1_stat_loc): Adjust.
22973         (fold_build2_stat_loc): Likewise.
22974         (fold_build3_stat_loc): Likewise.
22975         * tree.c (build0_stat): Remove _stat from name.
22976         (build1_stat): Likewise.
22977         (build2_stat): Likewise.
22978         (build3_stat): Likewise.
22979         (build4_stat): Likewise.
22980         (build5_stat): Likewise.
22981         * tree.h (build1_loc): Remove macro, and rename _stat function
22982         to this.
22983         (build2_loc): Likewise.
22984         (build3_loc): Likewise.
22985         (build4_loc): Likewise.
22986         (build5_loc): Likewise.
22988 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22990         * tree.c (make_int_cst_stat): Remove _stat from name.
22991         * tree.h (make_int_cst_stat): Adjust prototype.
22992         (make_int_cst): Remove macro.
22994 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22996         * tree.c (make_tre_binfo_stat): Remove _stat from name.
22997         * tree.h (make_tree_binfo_stat): Adjust prototype.
22998         (make_tree_binfo): Remove.
23000 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23002         * tree.c (copy_node_stat): Rename to copy_node.
23003         (build_distinct_type_copy): Adjust.
23004         * tree.h (copy_node_stat): Adjust prototype.
23005         (copy_node): Remove macro.
23007 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23009         * tree.c (make_node_stat): rename to make_node.
23010         (build_tree_list_stat): Adjust.
23011         (build0_stat): Likewise.
23012         (build2_stat): Likewise.
23013         (build3_stat): Likewise.
23014         (build4_stat): Likewise.
23015         (build5_stat): Likewise.
23016         (build_decl_stat): Likewise.
23017         * tree.h (make_node_stat): Adjust prototype.
23018         (make_node): remove macro.
23020 2017-07-28  Peter Bergner  <bergner@vnet.ibm.com>
23022         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
23023         (PPC_FEATURE2_SCV): Likewise.
23024         * config/rs6000/rs6000.c (cpu_supports_info): Use them.
23026 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23028         * config/aarch64/aarch64.c
23029         (aarch64_internal_mov_immediate): Add new special pattern.
23030         * config/aarch64/aarch64.md (*movdi_aarch64):
23031         Add reg/32bit const mov case.
23033 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23034             Richard Sandiford <richard.sandiford@linaro.org>
23036         * config/aarch64/aarch64.md (mov<mode>): Generalize.
23037         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
23038         Add integer and movi cases.
23039         (movi-split-hf-df-sf split, fp16): New.
23040         (enabled): Added TARGET_FP_F16INST.
23041         * config/aarch64/iterators.md (GPF_HF): New.
23042         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
23044 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23046         * config/aarch64/aarch64.c
23047         (aarch64_simd_container_mode): Add prototype.
23048         (aarch64_expand_mov_immediate): Add HI support.
23049         (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
23050         (aarch64_can_const_movi_rtx_p): New.
23051         (aarch64_preferred_reload_class):
23052         Remove restrictions of using FP registers for certain SIMD operations.
23053         (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
23054         (aarch64_valid_floating_const): Add integer move validation.
23055         (aarch64_simd_imm_scalar_p): Remove.
23056         (aarch64_output_scalar_simd_mov_immediate): Generalize function.
23057         (aarch64_legitimate_constant_p): Expand list of supported cases.
23058         * config/aarch64/aarch64-protos.h
23059         (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
23060         (aarch64_reinterpret_float_as_int): New.
23061         (aarch64_simd_imm_scalar_p): Remove.
23062         * config/aarch64/constraints.md (Uvi): New.
23063         (Dd): Split into Ds and new Dd.
23064         * config/aarch64/aarch64.md (*movsi_aarch64):
23065         Add SIMD mov case.
23066         (*movdi_aarch64): Add SIMD mov case.
23068 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23070         * tree-predcom.c: (struct chain): Handle store-store chain in which
23071         stores for elimination only store loop invariant values.
23072         (execute_pred_commoning_chain): Ditto.
23073         (prepare_initializers_chain_store_elim): Ditto.
23074         (prepare_finalizers): Ditto.
23075         (is_inv_store_elimination_chain): New function.
23076         (initialize_root_vars_store_elim_1): New function.
23078 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23080         * tree-predcom.c: Revise general description of the pass.
23081         (enum chain_type): New enum type for store elimination.
23082         (struct chain): New field supporting store elimination.
23083         (struct component): Ditto.
23084         (dump_chain): Dump store-stores chain.
23085         (release_chain): Release resources.
23086         (split_data_refs_to_components): Compute and create component
23087         contains only stores for elimination.
23088         (get_chain_last_ref_at): New function.
23089         (make_invariant_chain): Initialization.
23090         (make_rooted_chain): Specify chain type in parameter and record it.
23091         (add_looparound_copies): Skip for store-stores chain.
23092         (determine_roots_comp): Compute type of chain and pass it to
23093         make_rooted_chain.
23094         (initialize_root_vars_store_elim_2): New function.
23095         (finalize_eliminated_stores): New function.
23096         (remove_stmt): Handle store for elimination.
23097         (execute_pred_commoning_chain): Execute predictive commoning on
23098         store-store chains.
23099         (determine_unroll_factor): Skip unroll for store-stores chain.
23100         (prepare_initializers_chain_store_elim): New function.
23101         (prepare_initializers_chain): Hanlde store-store chain.
23102         (prepare_finalizers_chain, prepare_finalizers): New function.
23103         (tree_predictive_commoning_loop): Return integer value indicating
23104         if loop is unrolled or lcssa form is corrupted.
23105         (tree_predictive_commoning): Rewrite for lcssa form if necessary.
23107 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23109         * tree-predcom.c (initialize_root): Delete.
23110         (execute_pred_commoning_chain): Initialize root vars and replace
23111         reference of non-combined chain directly, rather than call above
23112         function.
23114 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23116         * tree-predcom.c (ref_at_iteration): Add parameter NITERS.  Compute
23117         memory reference to DR at (NITERS + ITERS)-th iteration of loop.
23119 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23121         * tree-predcom.c (struct chain): New field init_seq.
23122         (release_chain): Release init_seq.
23123         (prepare_initializers_chain): Record intialization stmts in above
23124         field.
23125         (insert_init_seqs): New function.
23126         (tree_predictive_commoning_loop): Call insert_init_seqs.
23128 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23130         * tree-predcom.c (determine_roots_comp): Skip trivial components.
23132 2017-07-28  Richard Biener  <rguenther@suse.de>
23134         * match.pd: Remove superfluous :c.
23135         * genmatch.c (simplify::id): Add member.
23136         (lower_commutative, lower_opt_convert, lower_cond, lower_for):
23137         Copy id.
23138         (current_id): New global.
23139         (dt_node::parent): Move from ...
23140         (dt_operand::parent): ... here.  Add for_id member.
23141         (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
23142         (decision_tree::find_node): Relax order requirement when
23143         merging DT_TRUE nodes to ones inbetween the current simplify
23144         and the one we try to merge with.  Add diagnostic whenever
23145         we need to enforce pattern order by not merging.
23146         (decision_tree::insert): Set current_id.
23147         (decision_tree::print_node): Dump parent node and for_id.
23148         (parser::last_id): Add member.
23149         (parser::push_simplify): Assign unique id.
23150         (parser::parser): Initialize last_id.
23152 2017-07-28  Martin Liska  <mliska@suse.cz>
23154         PR sanitizer/81340
23155         * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
23156         gimple_build_debug_bind.
23158 2017-07-28  Richard Biener  <rguenther@suse.de>
23160         PR tree-optimization/81502
23161         * match.pd: Add pattern combining BIT_INSERT_EXPR with
23162         BIT_FIELD_REF.
23163         * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
23164         size/pos operands.
23165         (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
23166         * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
23167         for BIT_FIELD_REF args.
23168         * fold-const.c (make_bit_field_ref): Likewise.
23169         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23171 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23173         PR sanitizer/80998
23174         * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
23175         * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
23176         * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
23177         Or it into SANITIZER_UNDEFINED.
23178         * ubsan.c: Include gimple-fold.h and varasm.h.
23179         (ubsan_expand_ptr_ifn): New function.
23180         (instrument_pointer_overflow): New function.
23181         (maybe_instrument_pointer_overflow): New function.
23182         (instrument_object_size): Formatting fix.
23183         (pass_ubsan::execute): Call instrument_pointer_overflow
23184         and maybe_instrument_pointer_overflow.
23185         * internal-fn.c (expand_UBSAN_PTR): New function.
23186         * ubsan.h (ubsan_expand_ptr_ifn): Declare.
23187         * sanitizer.def (__ubsan_handle_pointer_overflow,
23188         __ubsan_handle_pointer_overflow_abort): New builtins.
23189         * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
23190         * internal-fn.def (UBSAN_PTR): New internal function.
23191         * opts.c (sanitizer_opts): Add pointer-overflow.
23192         * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
23193         * fold-const.c (build_range_check): Compute pointer range check in
23194         integral type if pointer arithmetics would be needed.  Formatting
23195         fixes.
23197 2017-07-28  Martin Liska  <mliska@suse.cz>
23199         PR sanitizer/81460
23200         * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
23201         parameters that are of a variable-length.
23203 2017-07-28  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23205         * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h.  Add
23206         rs6000/biarch64.h.
23207         * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
23208         (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
23209         (CRT_CALL_STATIC_FUNCTION): Likewise.
23210         (ASM_DEFAULT_SPEC): New define.
23211         (ASM_SPEC32): Likewise.
23212         (ASM_SPEC64): Likewise.
23213         (ASM_SPEC_COMMON): Likewise.
23214         (ASM_SPEC): Likewise.
23215         (INVALID_64BIT): Likewise.
23216         (LINK_OS_DEFAULT_SPEC): Likewise.
23217         (LINK_OS_SPEC32): Likewise.
23218         (LINK_OS_SPEC64): Likewise.
23219         (POWERPC_LINUX): Likewise.
23220         (PTRDIFF_TYPE): Likewise.
23221         (RESTORE_FP_PREFIX): Likewise.
23222         (RESTORE_FP_SUFFIX): Likewise.
23223         (SAVE_FP_PREFIX): Likewise.
23224         (SAVE_FP_SUFFIX): Likewise.
23225         (SIZE_TYPE): Likewise.
23226         (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
23227         (TARGET_64BIT): Likewise.
23228         (TARGET_64BIT): Likewise.
23229         (TARGET_AIX): Likewise.
23230         (WCHAR_TYPE_SIZE): Likewise.
23231         (WCHAR_TYPE): Undefine.
23232         (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
23233         (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
23234         (CPP_OS_RTEMS_SPEC): Delete.
23235         (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems.  Add
23236         asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
23237         link_os_spec64.
23238         * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
23240 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23242         PR tree-optimization/81578
23243         * tree-parloops.c (build_new_reduction): Bail out if
23244         reduction_code isn't one of the standard OpenMP reductions.
23245         Move the details printing after that decision.
23247 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23249         * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23250         related to reload_in_progress.
23251         (splat_input_operand): Likewise.
23252         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23253         Delete prototype.
23254         * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23255         field.
23256         (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23257         (TARGET_INSTANTIATE_DECLS): Likewise.
23258         (legitimate_indexed_address_p): Delete reload_in_progress code.
23259         (rs6000_debug_legitimate_address_p): Likewise.
23260         (rs6000_eliminate_indexed_memrefs): Likewise.
23261         (rs6000_emit_le_vsx_store): Likewise.
23262         (rs6000_emit_move_si_sf_subreg): Likewise.
23263         (rs6000_emit_move): Likewise.
23264         (register_to_reg_type): Likewise.
23265         (rs6000_pre_atomic_barrier): Likewise.
23266         (rs6000_machopic_legitimize_pic_address): Likewise.
23267         (rs6000_allocate_stack_temp): Likewise.
23268         (rs6000_address_for_fpconvert): Likewise.
23269         (rs6000_address_for_altivec): Likewise.
23270         (rs6000_secondary_memory_needed_rtx): Delete function.
23271         (rs6000_check_sdmode): Likewise.
23272         (rs6000_alloc_sdmode_stack_slot): Likewise.
23273         (rs6000_instantiate_decls): Likewise.
23274         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23275         * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23276         Delete reload_in_progress.
23277         (*vec_reload_and_plus_<mptrsize>): Likewise.
23278         * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23279         (vsx_div_v2di): Likewise.
23280         (vsx_udiv_v2di): Likewise.
23282 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23284         * config/rs6000/rs6000.opt (mlra): Replace with stub.
23285         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23286         * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23287         (rs6000_debug_reg_global): Delete print of LRA status.
23288         (rs6000_option_override_internal): Delete dead LRA related code.
23289         (rs6000_lra_p): Delete function.
23290         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23292 2017-07-27  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23294         * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23295         * config/riscv/rtems.h: New file.
23297 2017-07-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23298             Sudakshina Das  <sudi.das@arm.com>
23300         * config/aarch64/aarch64.md
23301         (define_split for and<mode>3nr_compare): Move
23302         non aarch64_logical_operand to a register.
23303         (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23304         register immediate operand to a register.
23305         * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23307 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23309         PR middle-end/81564
23310         * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23312 2017-07-27  Richard Biener  <rguenther@suse.de>
23314         PR tree-optimization/81573
23315         PR tree-optimization/81494
23316         * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23317         multi defuse cycle case.
23319 2017-07-27  Richard Biener  <rguenther@suse.de>
23321         PR tree-optimization/81571
23322         * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23323         PHIs.
23325 2017-07-27  Eric Botcazou  <ebotcazou@adacore.com>
23327         * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23328         earlier and only if MASK_FPU is set.  Adjust formatting.
23330 2017-07-27  Martin Liska  <mliska@suse.cz>
23332         * opt-functions.awk: Add validation of value of Init.
23333         * optc-gen.awk: Pass new argument.
23335 2017-07-27  Martin Liska  <mliska@suse.cz>
23337         * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23338         Fix wrong condition.
23340 2017-07-27  Martin Liska  <mliska@suse.cz>
23342         * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23343         BBs and edges seen by autoFDO.
23345 2017-07-27  Richard Biener  <rguenther@suse.de>
23347         PR tree-optimization/81502
23348         * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23349         with incompatible but same sized type.
23350         (execute_update_addresses_taken): Likewise.
23352 2017-07-27  James Greenhalgh  <james.greenhalgh@arm.com>
23354         * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23355         flag_tree_loop_vectorize rather than flag_tree_vectorize.
23357 2017-07-27  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23359         PR target/81534
23360         * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23361         ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23362         Change s_operand to memory_operand.
23364 2017-07-27  Richard Sandiford  <richard.sandiford@linaro.org>
23366         * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23367         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23368         (rs6000_emit_le_vsx_permute): ...this.  Take the destination as input.
23369         Emit instructions rather than returning an expression.  Handle TFmode
23370         and KFmode by casting to TImode.
23371         (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23372         (rs6000_emit_le_vsx_store): Likewise.
23373         * config/rs6000/vsx.md (VSX_TI): New iterator.
23374         (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23375         (*vsx_le_undo_permute_<mode>): Likewise.
23376         (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23377         emit the split sequence.
23378         (*vsx_le_perm_store_<mode>): Likewise.
23380 2017-07-27  Jakub Jelinek  <jakub@redhat.com>
23382         PR tree-optimization/81555
23383         PR tree-optimization/81556
23384         * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23385         if true, force CHANGED for the recursive invocation.
23386         (reassociate_bb): Remember original length of ops array, pass
23387         len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23389         * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23390         attributes for noipa attribute.  For naked attribute use
23391         lookup_attribute first before lookup_attribute_spec.
23392         * final.c (rest_of_handle_final): Disable IPA RA for functions with
23393         noipa attribute.
23394         * ipa-visibility.c (non_local_p): Fix comment typos.  Return true
23395         for functions with noipa attribute.
23396         (cgraph_externally_visible_p): Return true for functions with noipa
23397         attribute.
23398         * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23399         for functions with noipa attribute.
23400         * doc/extend.texi: Document noipa function attribute.
23401         * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23402         also for functions with noipa attribute.
23403         (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23405 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23407         * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23408         vec_unalign_load_cost and vec_unalign_store_cost.
23410 2017-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
23412         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23413         -mvsx-small-integer option.
23414         (ISA_3_0_MASKS_IEEE): Likewise.
23415         (OTHER_VSX_VECTOR_MASKS): Likewise.
23416         (POWERPC_MASKS): Likewise.
23417         * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23418         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23419         code, only testing for DImode being allowed in non-VSX floating
23420         point registers.
23421         (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23422         to TARGET_P8_VECTOR test.  Remove redundant VSX test inside of
23423         another VSX test.
23424         (rs6000_option_override_internal): Delete -mvsx-small-integer.
23425         (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23426         TARGET_P8_VECTOR test.
23427         (rs6000_secondary_reload_simple_move): Likewise.
23428         (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23429         since TARGET_P9_VECTOR was already tested.
23430         (rs6000_opt_masks): Remove -mvsx-small-integer.
23431         * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23432         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23433         used.
23434         (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23435         test for TARGET_VEXTRACTUB was used, and that uses
23436         TARGET_P9_VECTOR.
23437         (p9 extract splitter): Likewise.
23438         (vsx_extract_<mode>_di_p9): Likewise.
23439         (vsx_extract_<mode>_store_p9): Likewise.
23440         (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23441         for TARGET_P9_VECTOR was used.  Delete code that is now dead with
23442         the elimination of TARGET_VSX_SMALL_INTEGER.
23443         (vsx_extract_<mode>_p8): Likewise.
23444         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23445         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23446         (vsx_set_<mode>_p9): Likewise.
23447         (vsx_set_v4sf_p9): Likewise.
23448         (vsx_set_v4sf_p9_zero): Likewise.
23449         (vsx_insert_extract_v4sf_p9): Likewise.
23450         (vsx_insert_extract_v4sf_p9_2): Likewise.
23451         * config/rs6000/rs6000.md (sign extend splitter): Change
23452         TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23453         (floatsi<mode>2_lfiwax_mem): Likewise.
23454         (floatunssi<mode>2_lfiwzx_mem): Likewise.
23455         (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23456         since a test for TARGET_P9_VECTOR was used.
23457         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23458         (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23459         (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23460         (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23461         TARGET_P8_VECTOR test.
23462         (fix_trunc<mode>si2_stfiwx): Likewise.
23463         (fix_trunc<mode>si2_internal): Likewise.
23464         (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23465         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23466         used.
23467         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23468         (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23469         TARGET_P8_VECTOR test.
23470         (fixuns_trunc<mode>si2_stfiwx): Likewise.
23471         (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23472         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23473         used.
23474         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23475         (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23476         since a test for TARGET_P9_VECTOR was used.
23477         (splitter for loading small constants): Likewise.
23479 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23481         * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23482         vec_fp_stmt_cost.
23484 2017-07-26  H.J. Lu  <hongjiu.lu@intel.com>
23486         PR target/81563
23487         * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23488         (fp_valid_at): Likewise.
23490 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23492         * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23493         (qdf24xx_addrcost_table): Likewise.
23494         (cortexa57_tunings): Update to use generic_branch_cost.
23495         (cortexa72_tunings): Likewise.
23496         (cortexa73_tunings): Likewise.
23497         (qdf24xx_tunings): Likewise.
23499 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23501         * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23502         (thunderx2t99_branch_cost): Likewise.
23503         (cortexa35_tunings): Update to use generic_branch_cost.
23504         (cortexa53_tunings): Likewise.
23505         (cortexa57_tunings): Likewise.
23506         (cortexa72_tunings): Likewise.
23507         (cortexa73_tunings): Likewise.
23508         (thunderx2t99_tunings): Likewise.
23510 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23512         * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23513         (sparc_option_override): Honour MASK_FSMULD.
23514         * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23515         * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23516         * config/sparc/sparc.opt (mfsmuld): New option.
23517         * doc/invoke.texi (mfsmuld): Document option.
23519 2017-07-26  Marek Polacek  <polacek@redhat.com>
23521         PR middle-end/70992
23522         * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23524 2017-07-26  Richard Biener  <rguenther@suse.de>
23526         * gimple-match-head.c (do_valueize): Return OP if valueize
23527         returns NULL_TREE.
23528         (get_def): New helper to get at the def stmt of a SSA name
23529         if valueize allows.
23530         * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23531         do_valueize to get at the def stmt.
23532         (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23534 2017-07-26  Wilco Dijkstra  <wdijkstr@arm.com>
23536         PR middle-end/46932
23537         * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23539 2017-07-26  Martin Liska  <mliska@suse.cz>
23541         PR sanitize/81186
23542         * function.c (expand_function_start): Make expansion of
23543         nonlocal_goto_save_area after parm_birth_insn.
23545 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23547         * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23548         from all CPU target flags enable members.
23550 2017-07-26  Richard Biener  <rguenther@suse.de>
23552         * genmatch.c (dt_simplify::gen): Make iterator vars const.
23553         (decision_tree::gen): Make 'type' const.
23554         (write_predicate): Likewise.
23556 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23558         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23559         Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23560         (rs6000_option_override_internal): Likewise.
23561         (rs6000_expand_vector_set): Likewise.
23562         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23563         (TARGET_UPPER_REGS_SF): Likewise.
23564         (TARGET_UPPER_REGS_DI): Likewise.
23565         (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23566         (TARGET_DIRECT_MOVE_64BIT): Likewise.
23567         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23568         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23569         (Splitters for DI constants in Altivec registers): Likewise.
23570         * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23571         (vsx_set_v4sf_p9): Likewise.
23572         (vsx_set_v4sf_p9_zero): Likewise.
23573         (vsx_insert_extract_v4sf_p9): Likewise.
23574         (vsx_insert_extract_v4sf_p9_2): Likewise.
23576 2017-07-25  Carl Love  <cel@us.ibm.com>
23578         * doc/extend.texi: Update the built-in documentation file for the
23579         existing built-in functions
23580         vector signed char vec_cnttz (vector signed char);
23581         vector unsigned char vec_cnttz (vector unsigned char);
23582         vector signed short vec_cnttz (vector signed short);
23583         vector unsigned short vec_cnttz (vector unsigned short);
23584         vector signed int vec_cnttz (vector signed int);
23585         vector unsigned int vec_cnttz (vector unsigned int);
23586         vector signed long long vec_cnttz (vector signed long long);
23587         vector unsigned long long vec_cnttz (vector unsigned long long);
23589 2017-07-25  Andrew Pinski  <apinski@cavium.com>
23591         * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23592         accesses where the use is for the first operand of a BIT_INSERT.
23594 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23596         PR bootstrap/81521
23597         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23598         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23600 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23602         * config/i386/gstabs.h: Delete.
23603         * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23605 2017-07-25  Uros Bizjak  <ubizjak@gmail.com>
23607         * config/i386/i386.c (ix86_decompose_address): Do not check for
23608         register RTX when looking at index_reg or base_reg.
23609         * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23611 2017-07-25  Eric Botcazou  <ebotcazou@adacore.com>
23613         * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23614         to update EH info here.
23616 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23618         * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23620 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23622         * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23624 2017-07-25  Torsten Duwe  <duwe@suse.de>
23626         * common.opt: Introduce -fpatchable-function-entry
23627         command line option, and its variables function_entry_patch_area_size
23628         and function_entry_patch_area_start.
23629         * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23630         including a two-value parser.
23631         * target.def (print_patchable_function_entry): New target hook.
23632         * targhooks.h (default_print_patchable_function_entry): New function.
23633         * targhooks.c (default_print_patchable_function_entry): Likewise.
23634         * toplev.c (process_options): Switch off IPA-RA if
23635         patchable function entries are being generated.
23636         * varasm.c (assemble_start_function): Look at the
23637         patchable-function-entry command line switch and current
23638         function attributes and maybe generate NOP instructions by
23639         calling the print_patchable_function_entry hook.
23640         * doc/extend.texi: Document patchable_function_entry attribute.
23641         * doc/invoke.texi: Document -fpatchable_function_entry
23642         command line option.
23643         * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23644         New target hook.
23645         * doc/tm.texi: Re-generate.
23647 2017-07-25  Jakub Jelinek  <jakub@redhat.com>
23649         PR target/81532
23650         * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23651         TARGET_AVX512DQ rather than TARGET_AVX512BW.
23653 2017-07-25  Tamar Christina  <tamar.christina@arm.com>
23655         * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23657 2017-07-25  Richard Biener  <rguenther@suse.de>
23659         PR tree-optimization/81455
23660         * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23661         not walk in cycles when looking for guards.
23663 2017-07-25  Richard Biener  <rguenther@suse.de>
23665         PR tree-optimization/81529
23666         * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23667         when optimizing backedge uses.
23669 2017-07-25  David Edelsohn  <dje.gcc@gmail.com>
23671         * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23672         character for AIX.
23673         * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23674         to dl_section_ref.  On AIX, append an expression to subtract
23675         the size of the section length to dl_section_ref.
23677 2017-07-25  Segher Boessenkool  <segher@kernel.crashing.org>
23679         * configure.ac: If any of the config.* scripts fail, exit 1.
23680         * configure: Regenerate.
23682 2017-07-25  Richard Biener  <rguenther@suse.de>
23684         PR middle-end/81546
23685         * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23686         of immediate uses, be more verbose on errors.
23688 2017-07-25  Richard Biener  <rguenther@suse.de>
23690         PR tree-optimization/81510
23691         * tree-vect-loop.c (vect_is_simple_reduction): When the
23692         reduction stmt is not inside the loop bail out.
23694 2017-07-25  Richard Biener  <rguenther@suse.de>
23696         PR tree-optimization/81303
23697         * tree-vect-loop-manip.c (vect_loop_versioning): Build
23698         profitability check against LOOP_VINFO_NITERSM1.
23700 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23702         * domwalk.c (cmp_bb_postorder): Simplify.
23703         (sort_bbs_postorder): New function.  Use it...
23704         (dom_walker::walk): ...here to optimize common cases.
23706 2017-07-25  Martin Liska  <mliska@suse.cz>
23708         PR ipa/81520
23709         * ipa-visibility.c (function_and_variable_visibility): Make the
23710         redirection just on target that supports aliasing.
23711         Fix GNU coding style.
23713 2017-07-25  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23715         PR libgcc/61152
23716         * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23717         Format changes.
23718         * config/arm/rtems.h: Likewise.
23719         * config/bfin/rtems.h: Likewise.
23720         * config/i386/rtemself.h: Likewise.
23721         * config/lm32/rtems.h: Likewise.
23722         * config/m32c/rtems.h: Likewise.
23723         * config/m68k/rtemself.h: Likewise.
23724         * config/microblaze/rtems.h: Likewise.
23725         * config/mips/rtems.h: Likewise.
23726         * config/moxie/rtems.h: Likewise.
23727         * config/nios2/rtems.h: Likewise.
23728         * config/powerpcspe/rtems.h: Likewise.
23729         * config/rs6000/rtems.h: Likewise.
23730         * config/rtems.h: Likewise.
23731         * config/sh/rtems.h: Likewise.
23732         * config/sh/rtemself.h: Likewise.
23733         * config/sparc/rtemself.h: Likewise.
23735 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23737         PR 81487
23738         * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23739         * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23740         * tree-ssa-structalias.c (alias_get_name): Same.
23742 2017-07-25  Bin Cheng  <bin.cheng@arm.com>
23744         PR target/81414
23745         * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23746         instructions if no du chain is found.
23748 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23750         * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23752 2017-07-25  Richard Biener  <rguenther@suse.de>
23754         PR middle-end/81505
23755         * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23756         sticky.
23758 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23760         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23761         upper-regs options.
23762         (ISA_2_7_MASKS_SERVER): Likewise.
23763         (ISA_3_0_MASKS_IEEE): Likewise.
23764         (OTHER_P8_VECTOR_MASKS): Likewise.
23765         (OTHER_VSX_VECTOR_MASKS): Likewise.
23766         (POWERPC_MASKS): Likewise.
23767         (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23768         duplicate list of options.
23769         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23770         explicit -mupper-regs options.
23771         * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23772         -mupper-regs* options.  Delete -mvsx-scalar-memory, which was an
23773         alias for -mupper-regs-df.
23774         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23775         (rs6000_init_hard_regno_mode_ok): Likewise.
23776         (rs6000_option_override_internal): Likewise.
23777         (rs6000_opt_masks): Likewise.
23778         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23779         options in terms of whether -mvsx or -mpower8-vector was used.
23780         (TARGET_UPPER_REGS_DI): Likewise.
23781         (TARGET_UPPER_REGS_SF): Likewise.
23782         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23783         -mupper-regs-* options.
23785 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23787         * passes.c (emergency_dump_function): Print some empty lines and a
23788         header before the RTL dump.
23790 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23792         * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23794 2017-07-24  Wilco Dijkstra  <wdijkstr@arm.com>
23796         PR target/79041
23797         * config/aarch64/aarch64.c (aarch64_classify_symbol):
23798         Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23800 2017-07-24  Carl Love  <cel@us.ibm.com>
23802         * config/rs6000/rs6000-c.c: Add support for built-in functions
23803         vector float vec_extract_fp32_from_shorth (vector unsigned short);
23804         vector float vec_extract_fp32_from_shortl (vector unsigned short);
23805         * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23806         vec_extract_fp_from_shortl): Add defines for the two builtins.
23807         * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23808         VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23809         new builtins.
23810         * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23811         (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23812         * doc/extend.texi: Update the built-in documentation file for the
23813         new built-in function.
23815 2017-07-24  Jakub Jelinek  <jakub@redhat.com>
23817         PR bootstrap/81521
23818         * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23819         documentation.
23820         * doc/generic.texi: Likewise.
23821         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23822         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23824 2017-07-24  Jackson Woodruff  <jackson.woodruff@arm.com>
23826         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23827         (aarch64_mls_elt_merge<mode>): Likewise.
23829 2017-07-23  Krister Walfridsson  <krister.walfridsson@gmail.com>
23831         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23832         having __cxa_atexit.
23834 2017-07-23  Michael Collison  <michael.collison@arm.com>
23836         * config/arm/arm.c (arm_option_override): Deprecate
23837         use of -mstructure-size-boundary.
23838         * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23839         * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23841 2017-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23843         PR target/80695
23844         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23845         Reduce cost estimate for direct moves.
23847 2017-07-23  Uros Bizjak  <ubizjak@gmail.com>
23849         PR target/80569
23850         * config/i386/i386.c (ix86_option_override_internal): Disable
23851         BMI, BMI2 and TBM instructions for -m16.
23853 2017-07-21  Carl Love  <cel@us.ibm.com>
23855         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23856         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23857         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23858         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23859         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23860         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23861         VMULOSW): New enum "unspec" values.
23862         (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23863         altivec_vmulosw): New patterns.
23864         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23865         VMULOSW): Add definitions.
23867 2017-07-21  Jim Wilson  <jim.wilson@linaro.org>
23869         * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23870         (qdf24xx): Likewise.
23871         * config/aarch64/aarch64-options-extensions.def (rdma); New.
23872         * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23873         (AARCH64_FL_V8_1): Renumber.
23874         (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23875         (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23876         * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23877         * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs.  Add
23878         rdma to feature modifiers list.
23880 2017-07-21  Yury Gribov  <tetra2005@gmail.com>
23882         PR middle-end/56727
23883         * ipa-visibility (function_and_variable_visibility): Convert
23884         recursive PLT call to direct call if appropriate.
23886 2017-07-21  Andrew Pinski  <apinski@cavium.com>
23888         * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23889         operand 1 to see if the types precision matches.
23890         * fold-const.c (operand_equal_p): Likewise.
23892 2017-07-21  Richard Biener  <rguenther@suse.de>
23894         PR tree-optimization/81303
23895         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23896         in datarefs vector.  Allow NULL dr0 for no peeling cost estimate.
23897         (vect_peeling_hash_get_lowest_cost): Adjust.
23898         (vect_enhance_data_refs_alignment): Likewise.  Use
23899         vect_get_peeling_costs_all_drs to compute the penalty for no
23900         peeling to match up costs.
23902 2017-07-21  Richard Biener  <rguenther@suse.de>
23904         PR tree-optimization/81500
23905         * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23906         we didn't identify a reduction path.
23908 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23909             Cesar Philippidis  <cesar@codesourcery.com>
23911         PR gcov-profile/81442
23912         * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23913         probabilities.
23915 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23917         PR lto/81430
23918         * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
23919         function.
23920         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
23921         nvptx_override_options_after_change.
23923 2017-07-21  Ulrich Drepper  <drepper@redhat.com>
23925         * dwarf2out.c (output_file_names): Avoid double testing for
23926         dwarf_version >= 5.
23928 2017-07-21  Georg-Johann Lay  <avr@gjlay.de>
23930         * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
23932 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23934         * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
23935         hot/cold regions.
23936         (try_crossjump_to_edge): Do not punt on partitioned functions.
23938 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23940         * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
23941         Put all BBs reachable only via paths crossing cold region to cold
23942         region.
23943         * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
23945 2016-07-21  Richard Biener  <rguenther@suse.de>
23947         PR tree-optimization/81303
23948         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
23949         into account prologue and epilogue iterations when raising
23950         min_profitable_iters to sth at least covering one vector iteration.
23952 2017-07-21  Tamar Christina  <tamar.christina@arm.com>
23954         * config/arm/arm.c (arm_test_cpu_arch_dat):
23955         Check for overlap.
23957 2017-07-20  Nathan Sidwell  <nathan@acm.org>
23959         Remove TYPE_METHODS.
23960         * tree.h (TYPE_METHODS): Delete.
23961         * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
23962         * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
23963         (dbxout_type_methods): Scan TYPE_FIELDS.
23964         (dbxout_type): Don't check TYPE_METHODS here.
23965         * function.c (use_register_for_decl): Always ignore register for
23966         class types when not optimizing.
23967         * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
23968         * tree.c (free_lang_data_in_type): Stitch out member functions and
23969         templates from TYPE_FIELDS.
23970         (build_distinct_type_copy, verify_type_variant,
23971         verify_type): Member fns are on TYPE_FIELDS.
23972         * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
23973         * tree-pretty-print.c (dump_generic_node): Likewise.
23975 2017-07-20  Jakub Jelinek  <jakub@redhat.com>
23977         PR target/80846
23978         * config/i386/i386.c (ix86_expand_vector_init_general): Handle
23979         V2TImode and V4TImode.
23980         (ix86_expand_vector_extract): Likewise.
23981         * config/i386/sse.md (VMOVE): Enable V4TImode even for just
23982         TARGET_AVX512F, instead of only for TARGET_AVX512BW.
23983         (ssescalarmode): Handle V4TImode and V2TImode.
23984         (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
23985         (*vec_extractv2ti, *vec_extractv4ti): New insns.
23986         (VEXTRACTI128_MODE): New mode iterator.
23987         (splitter for *vec_extractv?ti first element): New.
23988         (VEC_INIT_MODE): New mode iterator.
23989         (vec_init<mode>): Consolidate 3 expanders into one using
23990         VEC_INIT_MODE mode iterator.
23992 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23994         * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
23995         non_spilled_static_chain_regno_p.
23997 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23999         * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
24001 2017-07-20  Jan Hubicka  <hubicka@ucw.cz>
24003         * bb-reorder.c (connect_traces): Allow copying of blocks within
24004         single partition.
24006 2017-07-20  Richard Biener  <rguenther@suse.de>
24008         * gimple.h (gimple_phi_result): Add gphi * overload.
24009         (gimple_phi_result_ptr): Likewise.
24010         (gimple_phi_arg): Likewise.  Adjust index assert to only
24011         allow actual argument accesses rather than all slots available
24012         by capacity.
24013         (gimple_phi_arg_def): Add gphi * overload.
24014         * tree-phinodes.c (make_phi_node): Initialize only actual
24015         arguments.
24016         (resize_phi_node): Clear memory not covered by old node,
24017         do not initialize excess argument slots.
24018         (reserve_phi_args_for_new_edge): Initialize new argument slot
24019         completely.
24021 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
24023         PR tree-optimization/81388
24024         Revert r238585:
24025         2016-07-21  Bin Cheng  <bin.cheng@arm.com>
24027         * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
24028         by removing computation of may_be_zero.
24030 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24031             Tom de Vries  <tom@codesourcery.com>
24033         PR middle-end/81030
24034         * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
24035         when gimple level profile disagrees with what RTL expander did.
24037 2017-07-20  Richard Biener  <rguenther@suse.de>
24039         PR tree-optimization/61171
24040         * tree-vectorizer.h (slp_instance): Add reduc_phis member.
24041         (vect_analyze_stmt): Add slp instance parameter.
24042         (vectorizable_reduction): Likewise.
24043         * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
24044         (vect_is_simple_reduction): Deal with chains not detected
24045         as SLP reduction chain, specifically not properly associated
24046         chains containing a mix of plus/minus.
24047         (get_reduction_op): Remove.
24048         (get_initial_defs_for_reduction): Simplify, pass in whether
24049         this is a reduction chain, pass in the SLP node for the PHIs.
24050         (vect_create_epilog_for_reduction): Get the SLP instance as
24051         arg and adjust.
24052         (vectorizable_reduction): Get the SLP instance as arg.
24053         During analysis remember the SLP node with the PHIs in the
24054         instance.  Simplify getting at the vectorized reduction PHIs.
24055         * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
24056         through SLP instance.
24057         (vect_slp_analyze_operations): Likewise.
24058         * tree-vect-stms.c (vect_analyze_stmt): Likewise.
24059         (vect_transform_stmt): Likewise.
24061 2017-07-20  Tom de Vries  <tom@codesourcery.com>
24063         PR tree-optimization/81489
24064         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
24065         read of phi arg location to before loop that modifies phi.
24067 2017-07-20  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
24069         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
24070         New pattern.
24072 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24074         PR middle-end/81331
24075         * except.c (execute): Fix ordering issue.
24077 2018-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24079         PR rtl-optimization/81423
24080         * combine.c (make_compound_operation_int): Don't try to optimize
24081         the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
24083 2017-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24085         PR rtl-optimization/81423
24086         * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
24087         with a constant that is -1 in the truncated to mode.
24089 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24091         * predict.c (propagate_unlikely_bbs_forward): Break out from ...
24092         (determine_unlikely_bbs): ... here.
24093         * predict.h (propagate_unlikely_bbs_forward): Declare.
24094         * cfgexpand.c (pass_expand::execute): Use it.
24095         * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
24096         unlikely edges.
24097         (find_rarely_executed_basic_blocks_and_crossing_edges): Use
24098         propagate_unlikely_bbs_forward.
24100 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24102         PR middle-end/81331
24103         * except.c (maybe_add_nop_after_section_switch): New function.
24104         (execute): Use it.
24106 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24108         * gimple.h (gimple_phi_set_arg): Make assert more strict.
24110 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24112         * gimple.h (gimple_phi_arg): Make assert more strict.
24114 2017-07-19  Steven Munroe  <munroesj@gcc.gnu.org>
24116         * config.gcc (powerpc*-*-*): Add mmintrin.h.
24117         * config/rs6000/mmintrin.h: New file.
24118         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
24120 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24122         PR tree-optimization/81346
24123         * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
24125 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24127         * config/nvptx/nvptx.md (VECIM): Add V2DI.
24129 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24131         * config/nvptx/nvptx-modes.def: Add V2DImode.
24132         * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
24133         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
24134         (nvptx_output_mov_insn): Handle lack of mov.b128.
24135         (nvptx_print_operand): Handle 'H' and 'L' codes.
24136         (nvptx_vector_mode_supported): Allow V2DImode.
24137         (nvptx_preferred_simd_mode): New function.
24138         (nvptx_data_alignment): New function.
24139         (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
24140         nvptx_preferred_simd_mode.
24141         * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
24142         64 to 128 bits.
24143         (DATA_ALIGNMENT): Define.  Set to nvptx_data_alignment.
24145 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24147         * config/nvptx/nvptx-modes.def: New file.  Add V2SImode.
24148         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
24149         (nvptx_vector_mode_supported): New function.  Allow V2SImode.
24150         (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
24151         * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
24152         (mov<VECIM>_insn): New define_insn.
24153         (define_expand "mov<VECIM>): New define_expand.
24155 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24157         * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
24159 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24161         PR tree-optimization/81346
24162         * fold-const.h (fold_div_compare, range_check_type): Declare.
24163         * fold-const.c (range_check_type): New function.
24164         (build_range_check): Use range_check_type.
24165         (fold_div_compare): No longer static, rewritten into
24166         a match.pd helper function.
24167         (fold_comparison): Don't call fold_div_compare here.
24168         * match.pd (X / C1 op C2): New optimization using fold_div_compare
24169         as helper function.
24171 2017-07-19  Nathan Sidwell  <nathan@acm.org>
24173         * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
24174         (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
24175         * tree.c (find_decls_types_r, verify_type): Use
24176         TYPE_{MIN,MAX}_VALUE_RAW.
24177         * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
24178         (hash_tree): Likewise.
24179         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
24180         Likewise.
24181         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
24182         Likewise.
24184 2017-07-18  Tom de Vries  <tom@codesourcery.com>
24186         PR middle-end/81464
24187         * omp-expand.c (expand_omp_for_static_chunk): Handle
24188         equal-argument loop exit phi.
24190 2017-07-18  Uros Bizjak  <ubizjak@gmail.com>
24192         PR target/81471
24193         * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
24194         (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
24195         operand 2 predicate.
24196         (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
24197         operand 2 predicate.
24198         (ror,rol -> rorx splitters): Use const_int_operand as
24199         operand 2 predicate.
24201 2017-06-18  Richard Biener  <rguenther@suse.de>
24203         PR tree-optimization/81410
24204         * tree-vect-stmts.c (vectorizable_load): Properly adjust for
24205         the gap in the ! slp_perm SLP case after each group.
24207 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24209         PR middle-end/81463
24210         * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
24211         again.
24213 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24215         PR middle-end/81462
24216         * predict.c (set_even_probabilities): Cleanup; do not affect
24217         probabilities that are already known.
24218         (combine_predictions_for_bb): Call even when count is set.
24220 2017-07-18  Nathan Sidwell  <nathan@acm.org>
24222         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
24223         TYPE_MAX_VALUE.
24225 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24227         PR target/81408
24228         * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
24229         optimization for loop niter analysis.
24231 2017-07-18  Georg-Johann Lay  <avr@gjlay.de>
24233         PR target/81473
24234         * config/avr/avr.c (avr_optimize_casesi): Don't use
24235         INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
24237 2017-07-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
24239         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
24240         body_cost_vec from _vect_peel_extended_info.
24241         (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
24242         (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24243         npeel.
24245 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24247         * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24249 2017-07-18  Richard Biener  <rguenther@suse.de>
24251         PR tree-optimization/80620
24252         PR tree-optimization/81403
24253         * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24254         info when re-using a VN table entry.
24256 2017-07-18  Richard Biener  <rguenther@suse.de>
24258         PR tree-optimization/81418
24259         * tree-vect-loop.c (vectorizable_reduction): Properly compute
24260         vectype_in.  Verify that with lane-reducing reduction operations
24261         we have a single def-use cycle.
24263 2017-07-17  Carl Love  <cel@us.ibm.com>
24265         Revert commit r249424   2017-06-20  Carl Love  <cel@us.ibm.com>
24267         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24268         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24269         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24270         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24271         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24272         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24273         VMULOSW): New enum "unspec" values.
24274         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24275         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24276         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24277         altivec_vmulosw): New patterns.
24278         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24279         VMULOSW): Add definitions.
24281 2017-07-17  Uros Bizjak  <ubizjak@gmail.com>
24283         * config/alpha/alpha.c: Include predict.h.
24285 2017-07-17  Yury Gribov  <tetra2005@gmail.com>
24287         * tree-vrp.c (compare_assert_loc): Fix comparison function
24288         to return predictable results.
24290 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24292         * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24293         option.
24294         (subdi3): Likewise.
24295         * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24296         * doc/invoke.texi (mexpand-adddi): Update text.
24298 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24300         * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24301         that also clobbers the CC register. The old expand code is moved
24302         to ...
24303         (*arc_clzsi2): ... here.
24304         (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24305         the CC register. The old expand code is moved to ...
24306         (arc_ctzsi2): ... here.
24308 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24310         * config/arc/arc.opt (mindexed-loads): Use initial value
24311         TARGET_INDEXED_LOADS_DEFAULT.
24312         (mauto-modify-reg): Use initial value
24313         TARGET_AUTO_MODIFY_REG_DEFAULT.
24314         * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24315         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24316         * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24317         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24319 2017-07-17  Martin Liska  <mliska@suse.cz>
24321         PR sanitizer/81302
24322         * opts.c (finish_options): Do not allow -fgnu-tm
24323         w/ -fsanitize={kernel-,}address.  Say sorry.
24325 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24327         PR target/81369
24328         * tree-loop-distribution.c (classify_partition): Only assert on
24329         numer of iterations.
24330         (merge_dep_scc_partitions): Delete prameter.  Update function call.
24331         (distribute_loop): Remove code handling loop with unknown niters.
24332         (pass_loop_distribution::execute): Skip loop with unknown niters.
24334 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24336         PR target/81369
24337         * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24338         function sort_partitions_by_post_order.
24340 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24342         PR tree-optimization/81374
24343         * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24344         the max index of basic blocks, rather than number of basic blocks.
24346 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24348         * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24349         proto.
24350         (arc_legitimate_pic_operand_p): Likewise.
24351         * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24352         function.
24353         (arc_needs_pcl_p): Likewise.
24354         (arc_legitimate_pc_offset_p): Likewise.
24355         (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24356         function is also used in constrains.md.
24357         (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24358         validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24359         PLUS.  Only return true/false in known cases, otherwise assert.
24360         (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24361         is already called in arc_legitimate_constant_p.
24362         * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24363         pic addresses.
24364         (LEGITIMATE_PIC_OPERAND_P): Use
24365         arc_raw_symbolic_reference_mentioned_p function.
24366         * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24367         function.
24368         (Cal): Likewise.
24369         (C32): Likewise.
24371 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24372         Andrew Burgess  <andrew.burgess@embecosm.com>
24374         * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24375         (arc_return_address_register): New function.
24376         * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24377         (arc_handle_fndecl_attribute): Add naked attribute.
24378         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24379         (TARGET_WARN_FUNC_RETURN): Likewise.
24380         (arc_allocate_stack_slots_for_args): New function.
24381         (arc_warn_func_return): Likewise.
24382         (machine_function): Change type fn_type.
24383         (arc_compute_function_type): Consider new naked function type,
24384         change function return type.
24385         (arc_must_save_register): Adapt to handle new
24386         arc_compute_function_type's return type.
24387         (arc_expand_prologue): Likewise.
24388         (arc_expand_epilogue): Likewise.
24389         (arc_return_address_regs): Delete.
24390         (arc_return_address_register): New function.
24391         (arc_epilogue_uses): Use above function.
24392         * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24393         (arc_function_type): Change encoding, add naked type.
24394         (ARC_INTERRUPT_P): Change to handle the new encoding.
24395         (ARC_FAST_INTERRUPT_P): Likewise.
24396         (ARC_NORMAL_P): Define.
24397         (ARC_NAKED_P): Likewise.
24398         (arc_compute_function_type): Delete prototype.
24399         * config/arc/arc.md (in_ret_delay_slot): Use
24400         arc_return_address_register function.
24401         (simple_return): Likewise.
24402         (p_return_i): Likewise.
24404 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24406         PR tree-optimization/81428
24407         * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24408         can't be built for those types.
24410 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24412         Remove stuff dead since r239246.
24414         * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24415         * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24416         (avr_inform_devices): Remove dead stuff.
24418 2017-07-17  Tamar Christina  <tamar.christina@arm.com>
24420         * config/arm/arm_neon.h: Fix softp typo.
24422 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24424         PR tree-optimization/81365
24425         * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24426         aggregate moves onto bb predecessor edges, make sure there are no
24427         loads that could alias the lhs in between the start of bb and the
24428         loads from *phi.
24430 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24432         PR 80929
24433         * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24434         (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24435         [LSHIFTRT, outer_code = TRUNCATE]: Same.
24437 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24439         PR tree-optimization/81396
24440         * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24441         (init_symbolic_number): Initialize it to 1.
24442         (perform_symbolic_merge): Add n_ops from both operands into the new
24443         n_ops.
24444         (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24445         without base_addr as useless if they need more than one operation.
24446         (bswap_replace): Handle !bswap case for NULL base_addr.
24448 2017-07-17  Tom de Vries  <tom@codesourcery.com>
24450         PR target/81069
24451         * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24452         as possible.
24454 2017-07-17  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24456         * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24457         conditional builtin define __FIX_LEON3FT_B2BST.
24459 2017-07-17  Daniel Cederman  <cederman@gaisler.com>
24461         * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24462         MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24463         with -mfix-ut700.
24465 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24467         PR rtl-optimization/81424
24468         * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24469         to remove potential trapping from operands if -fnon-call-exceptions.
24471 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24473         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24474         profile_proability for scalling.
24475         * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24477 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24479         * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24481 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24483         * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24484         fixpoint arithmetics.
24486 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24488         * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24489         fixpoint arithmetics.
24491 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24493         * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24494         fixpoint arithmetics.
24496 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24498         * profile-count.h (profile_probability::from_reg_br_prob_note,
24499         profile_probability::to_reg_br_prob_note): New functions.
24500         * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24501         * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24502         * predict.c (probability_reliable_p): Update.
24503         (edge_probability_reliable_p): Update.
24504         (br_prob_note_reliable_p): Update.
24505         (invert_br_probabilities): Update.
24506         (add_reg_br_prob_note): New function.
24507         (combine_predictions_for_insn): Update.
24508         * asan.c (asan_clear_shadow): Update.
24509         * cfgbuild.c (compute_outgoing_frequencies): Update.
24510         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24511         (update_br_prob_note): Update.
24512         (rtl_verify_edges): Update.
24513         (purge_dead_edges): Update.
24514         (fixup_reorder_chain): Update.
24515         * emit-rtl.c (try_split): Update.
24516         * ifcvt.c (cond_exec_process_insns): Update.
24517         (cond_exec_process_if_block): Update.
24518         (dead_or_predicable): Update.
24519         * internal-fn.c (expand_addsub_overflow): Update.
24520         (expand_neg_overflow): Update.
24521         (expand_mul_overflow): Update.
24522         * loop-doloop.c (doloop_modify): Update.
24523         * loop-unroll.c (compare_and_jump_seq): Update.
24524         * optabs.c (emit_cmp_and_jump_insn_1): Update.
24525         * predict.h: Update.
24526         * reorg.c (mostly_true_jump): Update.
24527         * rtl.h: Update.
24528         * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24529         * config/alpha/alpha.c (emit_unlikely_jump): Update.
24530         * config/arc/arc.c: (emit_unlikely_jump): Update.
24531         * config/arm/arm.c: (emit_unlikely_jump): Update.
24532         * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24533         * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24534         * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24535         (ix86_print_operand): Update.
24536         (ix86_split_fp_branch): Update.
24537         (predict_jump): Update.
24538         * config/ia64/ia64.c (ia64_print_operand): Update.
24539         * config/mmix/mmix.c (mmix_print_operand): Update.
24540         * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24541         (rs6000_expand_split_stack_prologue): Update.
24542         * config/rs6000/rs6000.c: Update.
24543         * config/s390/s390.c (s390_expand_vec_strlen): Update.
24544         (s390_expand_vec_movstr): Update.
24545         (s390_expand_cs_tdsi): Update.
24546         (s390_expand_split_stack_prologue): Update.
24547         * config/sh/sh.c (sh_print_operand): Update.
24548         (expand_cbranchsi4): Update.
24549         (expand_cbranchdi4): Update.
24550         * config/sparc/sparc.c (output_v9branch): Update.
24551         * config/spu/spu.c (get_branch_target): Update.
24552         (ea_load_store_inline): Update.
24553         * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24554         * config/tilepro/tilepro.c: Update.
24556 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24558         * gimplify.c (mostly_copy_tree_r): Revert latest change.
24559         (gimplify_save_expr): Likewise.
24561 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24563         * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24565 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24567         * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24568         TV_IPA_FNSUMMARY.
24569         * timevar.def (TV_IPA_FNSUMMARY): Define.
24571 2017-07-16  Daniel Cederman  <cederman@gaisler.com>
24573         * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24574         to back store errata sensitive sequence from being generated.
24575         (sqrtdf2_fix): Likewise.
24577 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24579         * tree-ssa-threadupdate.c (compute_path_counts,
24580         update_joiner_offpath_counts): Use profile_probability.
24582 2017-07-15  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24584         Revert:
24585         2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24587         * config/arm/arm-c.c (arm_cpu_builtins): Define
24588         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24590 2017-07-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24592         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24593         array entries to represent __ieee128 versions of the
24594         scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24595         scalar_extract_sig, and scalar_insert_exp built-in functions.
24596         (altivec_resolve_overloaded_builtin): Add special case handling
24597         for the __builtin_scalar_insert_exp function, as represented by
24598         the P9V_BUILTIN_VEC_VSIEDP constant.
24599         * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24600         exponent support for __ieee128 argument.
24601         (VSESQP): Add scalar extract signature support for __ieee128
24602         argument.
24603         (VSTDCNQP): Add scalar test negative support for __ieee128
24604         argument.
24605         (VSIEQP): Add scalar insert exponent support for __int128 argument
24606         with __ieee128 result.
24607         (VSIEQPF): Add scalar insert exponent support for __ieee128
24608         argument with __ieee128 result.
24609         (VSTDCQP): Add scalar test data class support for __ieee128
24610         argument.
24611         (VSTDCNQP): Add overload support for scalar test negative with
24612         __ieee128 argument.
24613         (VSTDCQP): Add overload support for scalar test data class
24614         __ieee128 argument.
24615         * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24616         UNSPEC_VSX_SXSIGDP.
24617         (UNSPEC_VSX_SIEXPQP): New constant.
24618         (xsxexpqp): New insn for VSX scalar extract exponent quad
24619         precision.
24620         (xsxsigqp): New insn for VSX scalar extract significand quad
24621         precision.
24622         (xsiexpqpf): New insn for VSX scalar insert exponent quad
24623         precision with floating point argument.
24624         (xststdcqp): New expand for VSX scalar test data class quad
24625         precision.
24626         (xststdcnegqp): New expand for VSX scalar test negative quad
24627         precision.
24628         (xststdcqp): New insn to match expansions for VSX scalar test data
24629         class quad precision and VSX scalar test negative quad precision.
24630         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24631         special case operand checking to enforce that second operand of
24632         VSX scalar test data class with quad precision argument is a 7-bit
24633         unsigned literal.
24634         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24635         prototypes and descriptions of __ieee128 versions of
24636         scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24637         scalar_test_data_class, and scalar_test_neg built-in functions.
24639 2016-07-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24641         PR tree-optimization/81162
24642         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24643         replace a negate with an add.
24645 2017-07-14  James Greenhalgh  <james.greenhalgh@arm.com>
24647         * doc/invoke.texi (arm/-mcpu): Document +crypto.
24649 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24651         * config/arm/arm-c.c (arm_cpu_builtins): Define
24652         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24654 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24656         * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24657         (armv8-r): Set ARM Cortex-R52 as default CPU.
24658         * config/arm/arm-tables.opt: Regenerate.
24659         * config/arm/arm-tune.md: Regenerate.
24660         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24661         Cortex-R52.
24662         * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24663         extension for -mcpu=cortex-r52.
24665 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24667         * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24668         (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24669         * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24670         (fp-armv8): Define it as FP_ARMv8 only.
24671         config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24672         (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24673         TARGET_FPU_ARMV8.
24674         config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24675         TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24676         * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24677         first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24678         than TARGET_FPU_ARMV8.
24679         * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24680         __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24681         * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24682         TARGET_FPU_ARMV8.
24683         * config/arm/neon.md (neon_vrint): Likewise.
24684         (neon_vcvt): Likewise.
24685         (neon_<fmaxmin_op><mode>): Likewise.
24686         (<fmaxmin><mode>3): Likewise.
24687         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24688         * config/arm/predicates.md (arm_cond_move_operator): Check against
24689         TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24691 2017-07-14  Jackson Woodruff  <jackson.woodruff@arm.com>
24693         * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24694         to top of function.
24696 2017-07-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24698         * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24699         loop in comment with memset.
24701 2017-07-14  Martin Liska  <mliska@suse.cz>
24703         * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24704         * dwarf2out.c (is_java): Remove the function.
24705         (output_pubname): Remove usage of the function.
24706         (lower_bound_default): Remove usage of DW_LANG_Java.
24707         (gen_compile_unit_die): Likewise.
24708         * gcc.c: Remove compiler defaults for .java and .zip files.
24709         * gimple-expr.c (remove_suffix): Change as there's no longer
24710         extension than 4-letter one.
24711         * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24712         (gimplify_save_expr): Likewise.
24713         * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24714         as it's possible even for other languages than Java.
24715         * langhooks.h (struct lang_hooks): Remove Java from a comment.
24716         * lto-opts.c (lto_write_options): Remove reference to Java.
24717         * opts.c (strip_off_ending): Update file extension handling.
24718         * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24719         * tree-eh.c (lower_resx): Likewise.
24720         * tree.c (free_lang_data_in_type): Remove dead code.
24721         (find_decls_types_r): Likewise.
24722         (build_common_builtin_nodes): Remove Java from a comment.
24723         (verify_type): Remove dead code.
24724         * varasm.c (assemble_external): Remove Java from a comment.
24726 2017-07-14  Martin Liska  <mliska@suse.cz>
24728         * opts.c (finish_options): Add quotes.
24729         (common_handle_option): Likewise.
24731 2017-07-14  Martin Liska  <mliska@suse.cz>
24733         * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24734         * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24735         Remove N_SO_PASCAL.
24736         * dwarf2out.c (lower_bound_default): Do not handle
24737         DW_LANG_Pascal83.
24738         (gen_compile_unit_die): Likewise.
24739         * gcc.c: Remove default extension binding for GNU Pascal.
24740         * stmt.c: Remove Pascal language from a comment.
24741         * xcoffout.c: Likewise.
24743 2017-07-13  David Malcolm  <dmalcolm@redhat.com>
24745         PR c/81405
24746         * diagnostic-show-locus.c (fixit_cmp): New function.
24747         (layout::layout): Sort m_fixit_hints.
24748         (column_range::column_range): Assert that the values are valid.
24749         (struct char_span): New struct.
24750         (correction::overwrite): New method.
24751         (struct source_line): New struct.
24752         (line_corrections::add_hint): Add assertions.  Reimplement memcpy
24753         calls in terms of classes source_line and char_span, and
24754         correction::overwrite.
24755         (selftest::test_overlapped_fixit_printing_2): New function.
24756         (selftest::diagnostic_show_locus_c_tests): Call it.
24758 2017-07-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
24760         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24761         early if there is no lhs.
24763 2017-07-13  Martin Liska  <mliska@suse.cz>
24765         * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24766         (gen_reference_type_die): Likewise.
24767         * stor-layout.c: Remove Pascal-related comment.
24769 2017-07-13  Martin Liska  <mliska@suse.cz>
24771         * opts.c (finish_options): Add quotes to error messages.
24772         (parse_sanitizer_options): Likewise.
24774 2017-07-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24776         * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24778 2017-07-13  Richard Earnshaw  <rearnsha@arm.com>
24780         * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24782 2017-07-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
24784         * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24785         during expansion.
24786         * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24788 2017-07-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
24790         PR target/81193
24791         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24792         provides the hardware capability bits, define the macro
24793         __BUILTIN_CPU_SUPPORTS__.
24794         * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24795         if GLIBC does not provide the hardware capability bits.  Add a
24796         gcc_unreachable call if the built-in cpu function is neither
24797         __builtin_cpu_is nor __builtin_cpu_supports.
24798         (rs6000_get_function_versions_dispatcher): Change the warning
24799         that an old GLIBC is used which does not export the capability
24800         bits to be an error.
24801         * doc/extend.texi (target_clones attribute): Document the
24802         restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24803         (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24804         needed by __builtin_cpu_is and __builtin_cpu_supports.  Document
24805         the macros defined by GCC if the newer GLIBC is available.
24807 2017-07-12  Jeff Law  <law@redhat.com>
24809         * config/riscv/riscv.c: Remove unnecessary includes.  Reorder
24810         remaining includes slightly.
24811         * config/riscv/riscv-builtins.c: Include profile-count.h.
24813 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24815         PR target/79883
24816         * config/avr/avr.c (avr_set_current_function): In diagnostic
24817         messages: Quote keywords and (parts of) identifiers.
24818         [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24819         "INTERUPT".
24821 2017-07-12  Carl Love  <cel@us.ibm.com>
24823         * config/rs6000/rs6000-c.c: Add support for built-in functions
24824         vector bool char vec_revb (vector bool char);
24825         vector bool short vec_revb (vector short char);
24826         vector bool int vec_revb (vector bool int);
24827         vector bool long long vec_revb (vector bool long long);
24828         * doc/extend.texi: Update the built-in documentation file for the
24829         new built-in functions.
24831 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24833         * config/s390/s390.md: Remove movcc splitter.
24835 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24837         * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24838         load/store on condition.
24840 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24842         PR target/81407
24843         * config/avr/avr.c (avr_encode_section_info)
24844         [progmem && !TREE_READONLY]: Error if progmem object needs
24845         constructing.
24847 2017-07-11  Michael Collison  <michael.collison@arm.com>
24849         * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24850         New pattern.
24852 2017-07-11  Carl Love  <cel@us.ibm.com>
24854         * config/rs6000/rs6000-c.c: Add support for builtins
24855         vector unsigned int vec_parity_lsbb (vector signed int);
24856         vector unsigned int vec_parity_lsbb (vector unsigned int);
24857         vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24858         vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24859         vector unsigned long long vec_parity_lsbb (vector signed long long);
24860         vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24861         * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24862         * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24863         * doc/extend.texi: Update the built-in documentation file for the
24864         new built-in functions.
24866 2017-07-11  David Malcolm  <dmalcolm@redhat.com>
24868         * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24869         (layout::m_primary_loc): New field.
24870         (layout::layout): Initialize new field.  Move location filtering
24871         logic from here to...
24872         (layout::maybe_add_location_range): ...this new method.  Add
24873         support for filtering to just the lines already specified by other
24874         locations.
24875         (layout::will_show_line_p): New method.
24876         (gcc_rich_location::add_location_if_nearby): New method.
24877         (selftest::test_add_location_if_nearby): New test function.
24878         (selftest::diagnostic_show_locus_c_tests): Call it.
24879         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24880         New method.
24882 2017-07-11  Tom de Vries  <tom@codesourcery.com>
24884         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24885         (bb_first_real_insn): New function.
24886         (nvptx_single): Add extra initialization of broadcasted condition
24887         variables.
24889 2017-07-11  Nathan Sidwell  <nathan@acm.org>
24891         * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24893 2017-07-11  Georg-Johann Lay  <avr@gjlay.de>
24895         * doc/extend.texi (AVR Function Attributes): Remove weblink to
24896         Binutils doc as TEXI will mess them up.
24897         * doc/invoke.texi (AVR Options): Same here.
24899 2017-07-11  Daniel Cederman  <cederman@gaisler.com>
24901         * config/sparc/sparc.opt (mfix-ut700): New option.
24902         (mfix-gr712rc): Likewise.
24903         (sparc_fix_b2bst): New variable.
24904         * doc/invoke.texi (SPARC options): Document them.
24905         (ARM options): Fix warnings.
24906         * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24907         instructions to prevent sequences that can trigger the store-store
24908         errata for certain LEON3FT processors.
24909         (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24910         (sparc_option_override): Set sparc_fix_b2bst appropriately.
24911         * config/sparc/sparc.md (fix_b2bst): New attribute.
24912         (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24914 2017-07-10  Uros Bizjak  <ubizjak@gmail.com>
24916         PR target/81375
24917         * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
24918         (rcpps): Ditto.
24919         (*rsqrtsf2_sse): Ditto.
24920         (rsqrtsf2): Ditto.
24921         (div<mode>3): Macroize insn from divdf3 and divsf3
24922         using MODEF mode iterator.
24924 2017-07-10  Martin Sebor  <msebor@redhat.com>
24926         PR tree-optimization/80397
24927         * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
24928         instead of testing for equality to INTEGER_TYPE.
24930 2017-07-10  Vineet Gupta <vgupta@synopsys.com>
24932         * config.gcc: Remove uclibc from arc target spec.
24934 2017-07-10  Claudiu Zissulescu  <claziss@synopsys.com>
24936         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
24938 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24940         PR lto/80838
24941         * lto-wrapper.c (remove_option): New function.
24942         (merge_and_complain): Merge PIC/PIE options more realistically.
24944 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24946         Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
24948         PR target/20296
24949         PR target/81268
24950         * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
24951         (HAVE_AS_AVR_MGCCISR_OPTION):  If so, AC_DEFINE it.
24952         * config.in: Regenerate.
24953         * configure: Regenerate.
24954         * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
24955         * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
24956         * config/avr/avr.opt (-mgas-isr-prologues): New option and...
24957         (TARGET_GASISR_PROLOGUES): ...target mask.
24958         * common/config/avr/avr-common.c
24959         (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
24960         Set -mgas-isr-prologues.
24961         * config/avr/avr-passes.def (avr_pass_pre_proep): Add
24962         INSERT_PASS_BEFORE for it.
24963         * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
24964         * config/avr/avr.c (avr_option_override)
24965         [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
24966         (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
24967         (avr_attribute_table) <no_gccisr>: Add new function attribute.
24968         (avr_set_current_function) <is_no_gccisr>: Init machine field.
24969         (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
24970         and rtl_opt_pass.
24971         (make_avr_pass_pre_proep): New function.
24972         (emit_push_sfr) <treg>: Add argument to function and use it
24973         instead of TMP_REG.
24974         (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
24975         and set machine->gasisr.yes.
24976         (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
24977         (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
24978         __gcc_isr.n_pushed to .L__stack_usage.
24979         (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
24980         (avr_asm_final_postscan_insn): ...this new static function.
24981         * config/avr/avr.h (machine_function)
24982         <is_no_gccisr, use_L__stack_usage>: New fields.
24983         <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
24984         * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
24985         (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
24986         (gasisr, *gasisr): New expander and insn.
24987         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
24988         [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
24989         * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
24991 2017-07-10  Richard Earnshaw  <rearnsha@arm.com>
24993         * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
24994         in quoted strings.
24996 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24998         Move jump-tables out of .text again.
25000         PR target/81075
25001         * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
25002         (ASM_OUTPUT_ADDR_VEC): New function.
25003         (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
25004         (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
25005         INSN_ADDRESSes as asm comment.
25006         * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
25007         (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
25008         (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
25009         * config/avr/avr.md (*tablejump): Adjust comment.
25010         * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
25011         * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
25012         New detail.
25013         * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
25014         (avr_output_addr_vec): New proto.
25015         (avr_log_t) <insn_addresses>: New field.
25017 2017-07-09  H.J. Lu  <hongjiu.lu@intel.com>
25019         PR target/81313
25020         * config/i386/i386.c (ix86_function_arg_advance): Set
25021         outgoing_args_on_stack to true if there are outgoing arguments
25022         on stack.
25023         (ix86_function_arg): Likewise.
25024         (ix86_get_drap_rtx): Use DRAP only if there are outgoing
25025         arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
25026         * config/i386/i386.h (machine_function): Add
25027         outgoing_args_on_stack.
25029 2017-07-09  Krister Walfridsson  <krister.walfridsson@gmail.com>
25031         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
25032         supporting pthreds.
25033         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
25035 2017-07-08  Richard Sandiford  <richard.sandiford@linaro.org>
25037         * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
25038         (REAL_H): Remove $(MACHMODE_H).
25039         (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
25040         double-int.h.
25041         (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
25042         $(MACHMODE_H) and double-int.h.
25043         (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
25044         $(MACHMODE_H).
25045         (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
25046         double-int.h.
25048 2017-07-07  Andrew Pinski  <apinski@cavium.com>
25050         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
25051         prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
25053 2017-07-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
25055         * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
25056         Add warning if GCC was not configured to link against a GLIBC that
25057         exports the hardware capability bits.
25058         (make_resolver_func): Make resolver function private and not a
25059         COMDAT function.  Create the name with clone_function_name instead
25060         of make_unique_name.
25062         PR target/81348
25063         * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
25064         correct operand in doing the split.
25066 2017-07-07 Carl Love  <cel@us.ibm.com>
25068         * config/rs6000/rs6000-c: Add support for built-in function
25069         vector unsigned short vec_pack_to_short_fp32 (vector float,
25070                                                       vector float).
25071         * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
25072         BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
25073         * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
25074         * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
25075         (convert_4f32_8i16): Add define_expand.
25076         * doc/extend.texi: Update the built-in documentation file for the
25077         new built-in function.
25079 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25081         * config/sparc/m8.md: New file.
25082         * config/sparc/sparc.md: Include m8.md.
25084 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25086         * config/sparc/sparc.opt: New option -mvis4b.
25087         * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
25088         (sparc_option_override): Handle VIS4B.
25089         (enum sparc_builtins): Define
25090         SPARC_BUILTIN_DICTUNPACK{8,16,32},
25091         SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
25092         SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
25093         SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
25094         SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
25095         (check_constant_argument): New function.
25096         (sparc_vis_init_builtins): Define builtins
25097         __builtin_vis_dictunpack{8,16,32},
25098         __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
25099         __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
25100         __builtin_vis_fpcmpde{8,16,32}shl and
25101         __builtin_vis_fpcmpur{8,16,32}shl.
25102         (sparc_expand_builtin): Check that the constant operands to
25103         __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
25104         constant and in range.
25105         * config/sparc/sparc-c.c (sparc_target_macros): Handle
25106         TARGET_VIS4B.
25107         * config/sparc/sparc.h (SPARC_IMM2_P): Define.
25108         (SPARC_IMM5_P): Likewise.
25109         * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
25110         (enabled): Handle vis4b.
25111         (UNSPEC_DICTUNPACK): New unspec.
25112         (UNSPEC_FPCMPSHL): Likewise.
25113         (UNSPEC_FPUCMPSHL): Likewise.
25114         (UNSPEC_FPCMPDESHL): Likewise.
25115         (UNSPEC_FPCMPURSHL): Likewise.
25116         (cpu_feature): New CPU feature `vis4b'.
25117         (dictunpack{8,16,32}): New insns.
25118         (FPCSMODE): New mode iterator.
25119         (fpcscond): New code iterator.
25120         (fpcsucond): Likewise.
25121         (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
25122         (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
25123         (fpcmpde{8,16,32}{si,di}shl): Likewise.
25124         (fpcmpur{8,16,32}{si,di}shl): Likewise.
25125         * config/sparc/constraints.md: Define constraints `q' for unsigned
25126         2-bit integer constants and `t' for unsigned 5-bit integer
25127         constants.
25128         * config/sparc/predicates.md (imm5_operand_dictunpack8): New
25129         predicate.
25130         (imm5_operand_dictunpack16): Likewise.
25131         (imm5_operand_dictunpack32): Likewise.
25132         (imm2_operand): Likewise.
25133         * doc/invoke.texi (SPARC Options): Document -mvis4b.
25134         * doc/extend.texi (SPARC VIS Built-in Functions): Document the
25135         ditunpack* and fpcmp*shl builtins.
25137 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25139         * config.gcc: Handle m8 in --with-{cpu,tune} options.
25140         * config.in: Add HAVE_AS_SPARC6 define.
25141         * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
25142         M8.
25143         * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
25144         TARGET_CPU_m8.
25145         (ASM_CPU32_DEFAUILT_SPEC): Likewise.
25146         (CPP_CPU_SPEC): Handle m8.
25147         (ASM_CPU_SPEC): Likewise.
25148         * config/sparc/sparc-opts.h (enum processor_type): Add
25149         PROCESSOR_M8.
25150         * config/sparc/sparc.c (m8_costs): New struct.
25151         (sparc_option_override): Handle TARGET_CPU_m8.
25152         (sparc32_initialize_trampoline): Likewise.
25153         (sparc64_initialize_trampoline): Likewise.
25154         (sparc_issue_rate): Likewise.
25155         (sparc_register_move_cost): Likewise.
25156         * config/sparc/sparc.h (TARGET_CPU_m8): Define.
25157         (CPP_CPU64_DEFAULT_SPEC): Define for M8.
25158         (ASM_CPU64_DEFAULT_SPEC): Likewise.
25159         (CPP_CPU_SPEC): Handle M8.
25160         (ASM_CPU_SPEC): Likewise.
25161         (AS_M8_FLAG): Define.
25162         * config/sparc/sparc.md: Add m8 to the cpu attribute.
25163         * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
25164         * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
25165         M8 instructions.
25166         * configure: Regenerate.
25167         * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
25168         -mtune=m8.
25170 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25172         * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
25173         subtypes.
25174         * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
25175         ("*movdi_insn_sp32"): Do not set v3pipe.
25176         ("*movsi_insn"): Likewise.
25177         ("*movdi_insn_sp64"): Likewise.
25178         ("*movsf_insn"): Likewise.
25179         ("*movdf_insn_sp32"): Likewise.
25180         ("*movdf_insn_sp64"): Likewise.
25181         ("*zero_extendsidi2_insn_sp64"): Likewise.
25182         ("*sign_extendsidi2_insn"): Likewise.
25183         ("*mov<VM32:mode>_insn"): Likewise.
25184         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25185         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25186         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25187         ("<vlop:code><VL:mode>3"): Likewise.
25188         ("*not_<vlop:code><VL:mode>3"): Likewise.
25189         ("*nand<VL:mode>_vis"): Likewise.
25190         ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
25191         ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
25192         ("one_cmpl<VL:mode>2"): Likewise.
25193         ("faligndata<VM64:mode>_vis"): Likewise.
25194         ("alignaddrsi_vis"): Likewise.
25195         ("alignaddrdi_vis"): Likweise.
25196         ("alignaddrlsi_vis"): Likewise.
25197         ("alignaddrldi_vis"): Likewise.
25198         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25199         ("bmaskdi_vis"): Likewise.
25200         ("bmasksi_vis"): Likewise.
25201         ("bshuffle<VM64:mode>_vis"): Likewise.
25202         ("cmask8<P:mode>_vis"): Likewise.
25203         ("cmask16<P:mode>_vis"): Likewise.
25204         ("cmask32<P:mode>_vis"): Likewise.
25205         ("pdistn<P:mode>_vis"): Likewise.
25206         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25208 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25210         * config/sparc/sparc.md ("subtype"): New insn attribute.
25211         ("*wrgsr_sp64"): Set insn subtype.
25212         ("*rdgsr_sp64"): Likewise.
25213         ("alignaddrsi_vis"): Likewise.
25214         ("alignaddrdi_vis"): Likewise.
25215         ("alignaddrlsi_vis"): Likewise.
25216         ("alignaddrldi_vis"): Likewise.
25217         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25218         ("fexpand_vis"): Likewise.
25219         ("fpmerge_vis"): Likewise.
25220         ("faligndata<VM64:mode>_vis"): Likewise.
25221         ("bshuffle<VM64:mode>_vis"): Likewise.
25222         ("cmask8<P:mode>_vis"): Likewise.
25223         ("cmask16<P:mode>_vis"): Likewise.
25224         ("cmask32<P:mode>_vis"): Likewise.
25225         ("fchksm16_vis"): Likewise.
25226         ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
25227         ("fmean16_vis"): Likewise.
25228         ("fp<plusminus_insn>64_vis"): Likewise.
25229         ("<plusminus_insn>v8qi3"): Likewise.
25230         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25231         ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
25232         ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
25233         ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
25234         ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
25235         ("*movqi_insn"): Likewise.
25236         ("*movhi_insn"): Likewise.
25237         ("*movsi_insn"): Likewise.
25238         ("movsi_pic_gotdata_op"): Likewise.
25239         ("*movdi_insn_sp32"): Likewise.
25240         ("*movdi_insn_sp64"): Likewise.
25241         ("movdi_pic_gotdata_op"): Likewise.
25242         ("*movsf_insn"): Likewise.
25243         ("*movdf_insn_sp32"): Likewise.
25244         ("*movdf_insn_sp64"): Likewise.
25245         ("*zero_extendhisi2_insn"): Likewise.
25246         ("*zero_extendqihi2_insn"): Likewise.
25247         ("*zero_extendqisi2_insn"): Likewise.
25248         ("*zero_extendqidi2_insn"): Likewise.
25249         ("*zero_extendhidi2_insn"): Likewise.
25250         ("*zero_extendsidi2_insn_sp64"): Likewise.
25251         ("ldfsr"): Likewise.
25252         ("prefetch_64"): Likewise.
25253         ("prefetch_32"): Likewise.
25254         ("tie_ld32"): Likewise.
25255         ("tie_ld64"): Likewise.
25256         ("*tldo_ldub_sp32"): Likewise.
25257         ("*tldo_ldub1_sp32"): Likewise.
25258         ("*tldo_ldub2_sp32"): Likewise.
25259         ("*tldo_ldub_sp64"): Likewise.
25260         ("*tldo_ldub1_sp64"): Likewise.
25261         ("*tldo_ldub2_sp64"): Likewise.
25262         ("*tldo_ldub3_sp64"): Likewise.
25263         ("*tldo_lduh_sp32"): Likewise.
25264         ("*tldo_lduh1_sp32"): Likewise.
25265         ("*tldo_lduh_sp64"): Likewise.
25266         ("*tldo_lduh1_sp64"): Likewise.
25267         ("*tldo_lduh2_sp64"): Likewise.
25268         ("*tldo_lduw_sp32"): Likewise.
25269         ("*tldo_lduw_sp64"): Likewise.
25270         ("*tldo_lduw1_sp64"): Likewise.
25271         ("*tldo_ldx_sp64"): Likewise.
25272         ("*mov<VM32:mode>_insn"): Likewise.
25273         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25274         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25276 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25278         * config/sparc/sparc.md ("type"): New insn type viscmp.
25279         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25280         viscmp.
25281         ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25282         ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25283         ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25284         * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25285         viscmp.
25286         ("n7_vis_logical_11cycle"): Likewise.
25287         * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25288         * config/sparc/niagara2.md ("niag3_vis": Likewise.
25289         * config/sparc/niagara.md ("niag_vis"): Likewise.
25290         * config/sparc/ultra3.md ("us3_fga"): Likewise.
25291         * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25293 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25295         * config/sparc/sparc.md: New instruction type `bmask'.
25296         (bmaskdi_vis): Use the `bmask' type.
25297         (bmasksi_vis): Likewise.
25298         * config/sparc/ultra3.md (us3_array): Likewise.
25299         * config/sparc/niagara7.md (n7_array): Likewise.
25300         * config/sparc/niagara4.md (n4_array): Likewise.
25301         * config/sparc/niagara2.md (niag2_vis): Likewise.
25302         (niag3_vis): Likewise.
25303         * config/sparc/niagara.md (niag_vis): Likewise.
25305 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25307         * ipa-comdats.c: Remove optimize check from gate.
25308         * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25309         for functions not optimized.
25310         (ipa_fn_summary_read): Skip optimize check.
25311         (ipa_fn_summary_write): Likewise.
25312         * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25313         is optimized.
25314         * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25315         uninlinable.
25316         (can_inline_edge_p): Check flag_pcc_struct_return for match.
25317         (check_callers): Give up on caller which is not optimized.
25318         (inline_small_functions): Likewise.
25319         (ipa_inline): Do not give up when not optimizing.
25320         * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25321         away unoptimizes cdtors.
25322         (whole_program_function_and_variable_visibility): Do
25323         ipa_discover_readonly_nonaddressable_vars in LTO mode.
25324         * ipa.c (process_references): Do not check optimize.
25325         (symbol_table::remove_unreachable_nodes): Update optimize check.
25326         (set_writeonly_bit): Update optimize check.
25327         (pass_ipa_cdtor_merge::gate): Do not check optimize.
25328         (pass_ipa_single_use::gate): Remove.
25330 2017-07-06  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25332         * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25333         insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25334         rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25335         mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25336         permute_load, permute_store, adjust_extract, adjust_splat,
25337         adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25338         replace_swap_with_copy, dump_swap_insn_table,
25339         alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25340         recombine_lvx_pattern, recombine_stvx_pattern,
25341         recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25342         make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25343         to file rs6000-p8swap.c.
25344         * config/rs6000/rs6000-p8swap.c: New file.
25345         * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25346         * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25347         and rs6000*-*-* targets.
25349 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25351         * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25353 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25355         * lto-wrapper.c (merge_and_complain): Do not merge
25356         fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25357         fsigned_zeros, ftrapping_math, fwrapv.
25358         (append_compiler_options): Do not track these options.
25359         (append_linker_options): Likewie
25361 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25363         * cgraphunit.c (cgraph_node::finalize_function): When
25364         !flag_toplevel_reorde set no_reorder flag.
25365         (varpool_node::finalize_decl): Likewise.
25366         (symbol_table::compile): Drop no toplevel reorder path.
25368 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25370         * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25371         edges; zero probability is not better than uninitialized.
25373 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25375         * asan.h (asan_sanitize_allocas_p): Declare.
25376         * asan.c (asan_sanitize_allocas_p): New function.
25377         (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25378         (handle_builtin_alloca): Likewise.
25379         * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25380         if !asan_sanitize_allocas_p.
25381         * params.def (asan-instrument-allocas): Add new option.
25382         * params.h (ASAN_PROTECT_ALLOCAS): Define.
25383         * opts.c (common_handle_option): Disable allocas sanitization for
25384         KASan by default.
25386 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25388         * asan.c: Include gimple-fold.h.
25389         (get_last_alloca_addr): New function.
25390         (handle_builtin_stackrestore): Likewise.
25391         (handle_builtin_alloca): Likewise.
25392         (asan_emit_allocas_unpoison): Likewise.
25393         (get_mem_refs_of_builtin_call): Add new parameter, remove const
25394         quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25395         BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25396         (instrument_builtin_call): Pass gimple iterator to
25397         get_mem_refs_of_builtin_call.
25398         (last_alloca_addr): New global.
25399         * asan.h (asan_emit_allocas_unpoison): Declare.
25400         * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25401         (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25402         * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25403         if function calls alloca.
25404         * gimple-fold.c (replace_call_with_value): Remove static keyword.
25405         * gimple-fold.h (replace_call_with_value): Declare.
25406         * internal-fn.c: Include asan.h.
25407         * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25408         BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25410 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25412         * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25413         (C_SELFTEST_FLAGS): New.
25414         (CPP_SELFTEST_FLAGS): New.
25415         (SELFTEST_DEPS): New, from deps of s-selftest.
25416         (C_SELFTEST_DEPS): New, from deps of s-selftest.
25417         (CPP_SELFTEST_DEPS): New.
25418         (selftest): Add dependency on s-selftest-c++.
25419         (s-selftest): Rename to...
25420         (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25421         and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25422         than SELFTEST_FLAGS.
25423         (selftest-gdb): Rename to...
25424         (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25425         C_SELFTEST_FLAGS.
25426         (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25427         (selftest-valgrind): Rename to...
25428         (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25429         C_SELFTEST_FLAGS.
25430         (selftest-valgrind): Reintroduce as an alias for
25431         selftest-c-valgrind.
25432         (s-selftest-c++): New.
25433         (selftest-c++-gdb): New.
25434         (selftest-c++-valgrind): New.
25436 2017-07-06  Olivier Hainque  <hainque@adacore.com>
25438         * gcc.c (process_command): When deciding if undefined variables
25439         should be ignored when processing specs, accept "gcc -v" as well.
25441 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25443         * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25444         afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25446 2017-07-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25448         * config/arm/arm-cpus.in (armv8-r): Add new entry.
25449         * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25450         * config/arm/arm-tables.opt: Regenerate.
25451         * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25452         enumerator.
25453         * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25455 2017-07-06  Carl Love  <cel@us.ibm.com>
25457         * ChangeLog: Clean up from mid air collision
25459 2017-07-06  Carl Love  <cel@us.ibm.com>
25461         * config/rs6000/rs6000-c.c: Add support for built-in functions
25462         vector signed int vec_subc (vector signed int, vector signed int);
25463         vector signed __int128 vec_subc (vector signed __int128,
25464                                          vector signed __int128);
25465         vector unsigned __int128 vec_subc (vector unsigned __int128,
25466                                            vector unsigned __int128);
25467         vector signed int vec_sube (vector signed int, vector signed int,
25468                                     vector signed int);
25469         vector unsigned int vec_sube (vector unsigned int,
25470                                       vector unsigned int,
25471                                       vector unsigned int);
25472         vector signed __int128 vec_sube (vector signed __int128,
25473                                          vector signed __int128,
25474                                          vector signed__int128);
25475         vector unsigned __int128 vec_sube (vector unsigned __int128,
25476                                            vector unsigned __int128,
25477                                            vector unsigned __int128);
25478         vector signed int vec_subec (vector signed int, vector signed int,
25479                                      vector signed int);
25480         vector unsigned int vec_subec (vector unsigned int,
25481                                        vector unsigned int,
25482                                        vector unsigned int);
25483         vector signed __int128 vec_subec (vector signed __int128,
25484                                           vector signed __int128,
25485                                           vector signed__int128);
25486         vector unsigned __int128 vec_subec (vector unsigned __int128,
25487                                             vector unsigned __int128,
25488                                             vector unsigned __int128);
25489         * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25490         ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25491         * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25492         BU_ALTIVEC_OVERLOAD_X definitions.
25493         * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25494         * doc/extend.texi: Update the built-in documentation file for the new
25495         built-in functions.
25497 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25499         PR c++/79300
25500         * diagnostic-show-locus.c (layout::layout): Use start and finish
25501         spelling location for the start and finish of each range.
25502         * genmatch.c (linemap_client_expand_location_to_spelling_point):
25503         Add unused aspect param.
25504         * input.c (expand_location_1): Add "aspect" param, and use it
25505         to access the correct part of the location.
25506         (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25507         expand_location_1.
25508         (expand_location_to_spelling_point): Likewise.
25509         (linemap_client_expand_location_to_spelling_point): Add "aspect"
25510         param, and pass it to expand_location_1.
25512 2017-07-06  Sebastian Peryt  <sebastian.peryt@intel.com>
25514         * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25515         _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25516         _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25517         _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25518         _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25519         _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25520         _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25521         _mm_maskz_getmant_ss): New intrinsics.
25522         (__builtin_ia32_getexpss128_mask): Changed to ...
25523         __builtin_ia32_getexpss128_round ... this.
25524         (__builtin_ia32_getexpsd128_mask): Changed to ...
25525         __builtin_ia32_getexpsd128_round ... this.
25526         * config/i386/i386-builtin-types.def
25527         ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25528         (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25529         * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25530         __builtin_ia32_getexpss_mask_round,     __builtin_ia32_getmantsd_mask_round,
25531         __builtin_ia32_getmantss_mask_round): New builtins.
25532         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25533         V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25534         (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25535         CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25536         * config/i386/sse.md
25537         (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25538         avx512f_sgetexp<mode><mask_scalar_name>
25539         <round_saeonly_scalar_name> ... this.
25540         (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25541         %0, %1, %2<round_saeonly_op3>}): Changed to ...
25542         vgetexp<ssescalarmodesuffix>
25543         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25544         %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25545         (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25546         avx512f_vgetmant<mode><mask_scalar_name>
25547         <round_saeonly_scalar_name> ... this.
25548         (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25549         %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25550         vgetmant<ssescalarmodesuffix>
25551         \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25552         %0<mask_scalar_operand4>, %1, %2
25553         <round_saeonly_scalar_mask_op4>, %3} ... this.
25554         * config/i386/subst.md (mask_scalar_operand4,
25555         round_saeonly_scalar_mask_operand4,     round_saeonly_scalar_mask_op4,
25556         round_saeonly_scalar_nimm_predicate): New subst attributes.
25558 2017-07-06  Julia Koval  <julia.koval@intel.com>
25560         * config/i386/i386.c (ix86_erase_embedded_rounding):
25561         Remove code for old rounding pattern.
25563 2017-07-06  Richard Earnshaw  <rearnsha@arm.com>
25565         * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25567 2017-07-06  Christophe Lyon  <christophe.lyon@linaro.org>
25569         * doc/sourcebuild.texi (Test Directives, Variants of
25570         dg-require-support): Add documentation for dg-require-stack-check.
25572 2017-07-05  Sebastian Peryt  <sebastian.peryt@intel.com>
25574         * config/i386/subst.md (mask_scalar, round_scalar,
25575         round_saeonly_scalar): New meta-templates.
25576         (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25577         round_scalar_mask_operand3, round_scalar_mask_op3,
25578         round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25579         round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25580         round_saeonly_scalar_constraint,
25581         round_saeonly_scalar_prefix): New subst attribute.
25582         * config/i386/sse.md
25583         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25584         <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25585         <round_scalar_name> ... this.
25586         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25587         <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25588         <round_scalar_name> ... this.
25589         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25590         <sse>_vm<code><mode>3<mask_scalar_name>
25591         <round_saeonly_scalar_name> ... this.
25592         (v<plusminus_mnemonic><ssescalarmodesuffix>
25593         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25594         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25595         v<plusminus_mnemonic><ssescalarmodesuffix>
25596         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25597         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25598         (v<multdiv_mnemonic><ssescalarmodesuffix>
25599         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25600         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25601         v<multdiv_mnemonic><ssescalarmodesuffix>
25602         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25603         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25604         (v<maxmin_float><ssescalarmodesuffix>
25605         \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25606         %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25607         v<maxmin_float><ssescalarmodesuffix>
25608         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25609         %0<mask_scalar_operand3>, %1, %<iptr>2
25610         <round_saeonly_scalar_mask_op3>} ... this.
25612 2017-07-05  Richard Earnshaw  <rearnsha@arm.com>
25614         * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25615         (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25617 2017-07-05  Richard Sandiford  <richard.sandiford@linaro.org>
25618             Alan Hayward  <alan.hayward@arm.com>
25619             David Sherwood  <david.sherwood@arm.com>
25621         * combine.c (simplify_if_then_else): Remove "enum" before
25622         "machine_mode".
25623         * compare-elim.c (can_eliminate_compare): Likewise.
25624         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25625         Likewise.
25626         (aarch64_lookup_simd_builtin_type): Likewise.
25627         (aarch64_simd_builtin_type): Likewise.
25628         (aarch64_init_simd_builtin_types): Likewise.
25629         (aarch64_simd_expand_args): Likewise.
25630         * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25631         Likewise.
25632         (aarch64_reverse_mask): Likewise.
25633         (aarch64_simd_emit_reg_reg_move): Likewise.
25634         (aarch64_gen_adjusted_ldpstp): Likewise.
25635         (aarch64_ccmp_mode_to_code): Likewise.
25636         (aarch64_operands_ok_for_ldpstp): Likewise.
25637         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25638         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25639         Likewise.
25640         (aarch64_min_divisions_for_recip_mul): Likewise.
25641         (aarch64_reassociation_width): Likewise.
25642         (aarch64_get_condition_code_1): Likewise.
25643         (aarch64_simd_emit_reg_reg_move): Likewise.
25644         (aarch64_simd_attr_length_rglist): Likewise.
25645         (aarch64_reverse_mask): Likewise.
25646         (aarch64_operands_ok_for_ldpstp): Likewise.
25647         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25648         (aarch64_gen_adjusted_ldpstp): Likewise.
25649         * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25650         Likewise.
25651         * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25652         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25653         (arm_lookup_simd_builtin_type): Likewise.
25654         (arm_simd_builtin_type): Likewise.
25655         (arm_init_simd_builtin_types): Likewise.
25656         (arm_expand_builtin_args): Likewise.
25657         * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25658         * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25659         (ft32_setup_incoming_varargs): Likewise.
25660         (ft32_function_arg): Likewise.
25661         (ft32_function_arg_advance): Likewise.
25662         (ft32_pass_by_reference): Likewise.
25663         (ft32_arg_partial_bytes): Likewise.
25664         (ft32_valid_pointer_mode): Likewise.
25665         (ft32_addr_space_pointer_mode): Likewise.
25666         (ft32_addr_space_legitimate_address_p): Likewise.
25667         * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25668         Likewise.
25669         * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25670         (ix86_emit_outlined_ms2sysv_restore): Likewise.
25671         (iamcu_alignment): Likewise.
25672         (canonicalize_vector_int_perm): Likewise.
25673         (ix86_noce_conversion_profitable_p): Likewise.
25674         (ix86_mpx_bound_mode): Likewise.
25675         (ix86_operands_ok_for_move_multiple): Likewise.
25676         * config/microblaze/microblaze-protos.h
25677         (microblaze_expand_conditional_branch_reg): Likewise.
25678         * config/microblaze/microblaze.c
25679         (microblaze_expand_conditional_branch_reg): Likewise.
25680         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25681         Likewise.
25682         (rs6000_reassociation_width): Likewise.
25683         (rs6000_invalid_binary_op): Likewise.
25684         (fusion_p9_p): Likewise.
25685         (emit_fusion_p9_load): Likewise.
25686         (emit_fusion_p9_store): Likewise.
25687         * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25688         Likewise.
25689         (riscv_hard_regno_mode_ok_p): Likewise.
25690         (riscv_address_insns): Likewise.
25691         (riscv_split_symbol): Likewise.
25692         (riscv_legitimize_move): Likewise.
25693         (riscv_function_value): Likewise.
25694         (riscv_hard_regno_nregs): Likewise.
25695         (riscv_expand_builtin): Likewise.
25696         * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25697         (riscv_build_integer): Likewise.
25698         (riscv_split_integer): Likewise.
25699         (riscv_legitimate_constant_p): Likewise.
25700         (riscv_cannot_force_const_mem): Likewise.
25701         (riscv_regno_mode_ok_for_base_p): Likewise.
25702         (riscv_valid_base_register_p): Likewise.
25703         (riscv_valid_offset_p): Likewise.
25704         (riscv_valid_lo_sum_p): Likewise.
25705         (riscv_classify_address): Likewise.
25706         (riscv_legitimate_address_p): Likewise.
25707         (riscv_address_insns): Likewise.
25708         (riscv_load_store_insns): Likewise.
25709         (riscv_force_binary): Likewise.
25710         (riscv_split_symbol): Likewise.
25711         (riscv_force_address): Likewise.
25712         (riscv_legitimize_address): Likewise.
25713         (riscv_move_integer): Likewise.
25714         (riscv_legitimize_const_move): Likewise.
25715         (riscv_legitimize_move): Likewise.
25716         (riscv_address_cost): Likewise.
25717         (riscv_subword): Likewise.
25718         (riscv_output_move): Likewise.
25719         (riscv_canonicalize_int_order_test): Likewise.
25720         (riscv_emit_int_order_test): Likewise.
25721         (riscv_function_arg_boundary): Likewise.
25722         (riscv_pass_mode_in_fpr_p): Likewise.
25723         (riscv_pass_fpr_single): Likewise.
25724         (riscv_pass_fpr_pair): Likewise.
25725         (riscv_get_arg_info): Likewise.
25726         (riscv_function_arg): Likewise.
25727         (riscv_function_arg_advance): Likewise.
25728         (riscv_arg_partial_bytes): Likewise.
25729         (riscv_function_value): Likewise.
25730         (riscv_pass_by_reference): Likewise.
25731         (riscv_setup_incoming_varargs): Likewise.
25732         (riscv_print_operand): Likewise.
25733         (riscv_elf_select_rtx_section): Likewise.
25734         (riscv_save_restore_reg): Likewise.
25735         (riscv_for_each_saved_reg): Likewise.
25736         (riscv_register_move_cost): Likewise.
25737         (riscv_hard_regno_mode_ok_p): Likewise.
25738         (riscv_hard_regno_nregs): Likewise.
25739         (riscv_class_max_nregs): Likewise.
25740         (riscv_memory_move_cost): Likewise.
25741         * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25742         * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25743         (rl78_addr_space_address_mode): Likewise.
25744         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25745         Likewise.
25746         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25747         (rs6000_reassociation_width): Likewise.
25748         (rs6000_invalid_binary_op): Likewise.
25749         (fusion_p9_p): Likewise.
25750         (emit_fusion_p9_load): Likewise.
25751         (emit_fusion_p9_store): Likewise.
25752         * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25753         (ok_for_simple_move_operands): Likewise.
25754         (ok_for_simple_move_strict_operands): Likewise.
25755         (ok_for_simple_arith_logic_operands): Likewise.
25756         (visium_legitimize_reload_address): Likewise.
25757         (visium_select_cc_mode): Likewise.
25758         (output_cbranch): Likewise.
25759         (visium_split_double_move): Likewise.
25760         (visium_expand_copysign): Likewise.
25761         (visium_expand_int_cstore): Likewise.
25762         (visium_expand_fp_cstore): Likewise.
25763         * config/visium/visium.c (visium_pass_by_reference): Likewise.
25764         (visium_function_arg): Likewise.
25765         (visium_function_arg_advance): Likewise.
25766         (visium_libcall_value): Likewise.
25767         (visium_setup_incoming_varargs): Likewise.
25768         (visium_legitimate_constant_p): Likewise.
25769         (visium_legitimate_address_p): Likewise.
25770         (visium_legitimize_address): Likewise.
25771         (visium_secondary_reload): Likewise.
25772         (visium_register_move_cost): Likewise.
25773         (visium_memory_move_cost): Likewise.
25774         (prepare_move_operands): Likewise.
25775         (ok_for_simple_move_operands): Likewise.
25776         (ok_for_simple_move_strict_operands): Likewise.
25777         (ok_for_simple_arith_logic_operands): Likewise.
25778         (visium_function_value_1): Likewise.
25779         (rtx_ok_for_offset_p): Likewise.
25780         (visium_legitimize_reload_address): Likewise.
25781         (visium_split_double_move): Likewise.
25782         (visium_expand_copysign): Likewise.
25783         (visium_expand_int_cstore): Likewise.
25784         (visium_expand_fp_cstore): Likewise.
25785         (visium_split_cstore): Likewise.
25786         (visium_select_cc_mode): Likewise.
25787         (visium_split_cbranch): Likewise.
25788         (output_cbranch): Likewise.
25789         (visium_print_operand_address): Likewise.
25790         * expmed.c (flip_storage_order): Likewise.
25791         * expmed.h (emit_cstore): Likewise.
25792         (flip_storage_order): Likewise.
25793         * genrecog.c (validate_pattern): Likewise.
25794         * hsa-gen.c (gen_hsa_addr): Likewise.
25795         * internal-fn.c (expand_arith_overflow): Likewise.
25796         * ira-color.c (allocno_copy_cost_saving): Likewise.
25797         * lra-assigns.c (find_hard_regno_for_1): Likewise.
25798         * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25799         (process_invariant_for_inheritance): Likewise.
25800         * lra-eliminations.c (move_plus_up): Likewise.
25801         * omp-low.c (lower_oacc_reductions): Likewise.
25802         * simplify-rtx.c (simplify_subreg): Likewise.
25803         * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25804         (TARGET_CHKP_BOUND_MODE): Likewise..
25805         * targhooks.c (default_chkp_bound_mode): Likewise.
25806         (default_setup_incoming_vararg_bounds): Likewise.
25807         * targhooks.h (default_chkp_bound_mode): Likewise.
25808         (default_setup_incoming_vararg_bounds): Likewise.
25809         * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25810         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25811         (have_whole_vector_shift): Likewise.
25812         * tree-vect-stmts.c (vectorizable_load): Likewise.
25813         * doc/tm.texi: Regenerate.
25815 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25817         Graceful degrade if Binutils PR21472 is not available.
25819         PR target/81072
25820         * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25821         .rodata in flash test fails.
25822         (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25823         * confgure: Regenerate.
25824         * config.in: Regenerate.
25825         * config/avr/avr.c (avr_asm_named_section)
25826         [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25827         __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25828         (avr_asm_init_sections): Same.
25830 2017-07-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25832         * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25833         (fma<VH:mode>4_intrinsic): Likewise.
25834         (*fmsub<VCVTF:mode>4): Likewise.
25835         (*fmsub<VH:mode>4_intrinsic): Likewise.
25837 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25839         PR target/81305
25840         * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25841         Don't depend on "optimize > 0".
25842         (out_movhi_r_mr, out_movqi_mr_r): Same.
25843         (out_movhi_mr_r, out_movqi_r_mr): Same.
25844         (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25845         io_address_operand on "optimize > 0".
25847 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25849         * tree-loop-distribution.c: Add general explanantion on the pass.
25850         (generate_loops_for_partition): Mark distributed loop.
25851         (pg_add_dependence_edges): New parameter.  Handle alias data
25852         dependence specially and record it in the parameter if asked.
25853         (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25854         (init_partition_graph_vertices, add_partition_graph_edge): New.
25855         (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25856         (free_partition_graph_vdata, build_partition_graph): New.
25857         (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25858         (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25859         (data_ref_segment_size, latch_dominated_by_data_ref): New.
25860         (compute_alias_check_pairs, version_loop_by_alias_check): New.
25861         (version_for_distribution_p, finalize_partitions): New.
25862         (distribute_loop): Handle alias data dependence specially.  Factor
25863         out loop fusion code as functions and call these functions.
25865 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25867         * tree-loop-distribution.c (classify_partition): New parameter and
25868         better handle reduction statement.
25869         (rdg_build_partitions): Revise comment.
25870         (distribute_loop): Compute statements in all partitions and pass it
25871         to classify_partition.
25873 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25875         * tree-loop-distribution.c (enum partition_type): New.
25876         (struct partition): New field type.
25877         (partition_merge_into): Add parameter.  Update partition type.
25878         (data_dep_in_cycle_p, update_type_for_merge): New functions.
25879         (build_rdg_partition_for_vertex): Compute partition type.
25880         (rdg_build_partitions): Dump partition type.
25881         (distribute_loop): Update calls to partition_merge_into.
25883 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25885         * tree-loop-distribution.c (struct ddr_hasher): New.
25886         (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25887         (ddrs_table): New.
25888         (classify_partition): Call get_data_dependence.
25889         (pg_add_dependence_edges): Ditto.
25890         (distribute_loop): Release data dependence hash table.
25892 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25894         * tree-loop-distribution.c (ref_base_address): Delete.
25895         (similar_memory_accesses): Rename ...
25896         (share_memory_accesses): ... to this.  Check if partitions access
25897         the same memory reference.
25898         (distribute_loop): Call share_memory_accesses.
25900 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25902         * tree-loop-distribution.c (struct partition): New field recording
25903         its data reference.
25904         (partition_alloc, partition_free): Init and release data refs.
25905         (partition_merge_into): Merge data refs.
25906         (build_rdg_partition_for_vertex): Collect data refs for partition.
25907         (pg_add_dependence_edges): Change parameters from vector to bitmap.
25908         Update uses.
25909         (distribute_loop): Remve data refs from vertice data of partition
25910         graph.
25912 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25914         * tree-loop-distribution.c (params.h): Include header file.
25915         (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25916         (datarefs_vec): New global var.
25917         (create_rdg_vertices): Use datarefs_vec directly.
25918         (free_rdg): Don't free data references.
25919         (build_rdg): Update use.  Don't free data references.
25920         (distribute_loop): Compute global variable for data references.
25921         Bail out if there are too many data references.
25923 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25925         * tree-loop-distribution.c (loop_nest): New global var.
25926         (build_rdg): Use loop directly, rather than loop nest.
25927         (pg_add_dependence_edges): Remove loop nest parameter.  Use global
25928         variable directly.
25929         (distribute_loop): Compute global variable loop nest.  Update use.
25931 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25933         * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
25934         (partition_merge_into): New parameter.  Dump reason for fusion.
25935         (distribute_loop): Update use of partition_merge_into.
25937 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25939         * tree-loop-distribution.c (bb_top_order_index): New.
25940         (bb_top_order_index_size, bb_top_order_cmp): New.
25941         (stmts_from_loop): Use topological order.
25942         (pass_loop_distribution::execute): Compute and release topological
25943         order for basic blocks.
25945 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25947         * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
25948         if no loops.
25950 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25952         * cfgloop.h (struct loop): Add comment.  New field orig_loop_num.
25953         * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
25954         * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
25955         * internal-fn.def (LOOP_DIST_ALIAS): New.
25956         * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
25957         (fold_loop_internal_call): ... this.
25958         (vect_loop_dist_alias_call): New function.
25959         (set_uid_loop_bbs): Call fold_loop_internal_call.
25960         (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
25961         internal calls.
25963 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25965         PR target/81300
25966         * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
25967         Require dead FLAGS_REG at the beginning of a peephole.
25969 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25971         PR target/81294
25972         * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
25973         arguments in the call to __builtin_ia32_sbb_u32.
25974         (_subborrow_u64): Swap _X and _Y arguments in the call to
25975         __builtin_ia32_sbb_u64.
25977 2017-07-04  Jakub Jelinek  <jakub@redhat.com>
25979         PR debug/81278
25980         * tree-vrp.c (compare_assert_loc): Turn into a function template
25981         with stable template parameter.  Only test if a->e is NULL,
25982         !a->e == !b->e has been verified already.  Use e == NULL or
25983         e != NULL instead of e or ! e tests.  If stable is true, don't use
25984         iterative_hash_expr, on the other side allow a or b or both NULL
25985         and sort the NULLs last.
25986         (process_assert_insertions): Sort using compare_assert_loc<false>
25987         instead of compare_assert_loc, later sort using
25988         compare_assert_loc<true> before calling process_assert_insertions_for
25989         in a loop.  Use break instead of continue once seen NULL pointer.
25991 2017-07-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25993         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
25994         Cortex-R7 and Cortex-R8 processors.
25996 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25998         * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
25999         uninitialized while src is not.
26001 2017-07-04  Richard Earnshaw  <rearnsha@arm.com>
26003         * common/config/arm/arm-common.c: Adjust include path for
26004         arm-cpu-cdata.h
26005         * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
26006         (arm-cpu.h): Create in build directory.  Adjust dependency rules.
26007         (arm-cpu-data.h): Likewise.
26008         (arm-cpu-cdata.h): Likewise.
26009         * config/arm/arm-cpu.h: Delete.
26010         * config/arm/arm-cpu-cdata.h: Delete.
26011         * config/arm/arm-cpu-data.h: Delete.
26013 2017-07-04  James Greenhalgh  <james.greenhalgh@arm.com>
26015         * config/arm/arm-cpus.in (cortex-a55): New.
26016         (cortex-a75): Likewise.
26017         (cortex-a75.cortex-a55): Likewise.
26018         * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
26019         cortex-a75.
26020         * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
26021         * config/arm/arm-cpu-cdata.h: Regenerate.
26022         * config/arm/arm-cpu-data.h: Regenerate.
26023         * config/arm/arm-cpu.h: Regenerate.
26024         * config/arm/arm-tables.opt: Regenerate.
26025         * config/arm/arm-tune.md: Regenerate.
26027 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26029         * haifa-sched.c (sched_create_recovery_edges): Update profile.
26031 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26033         * bb-reorder.c (better_edge_p): Fix handling of uninitialized
26034         probability.
26036 2017-07-04  Richard Sandiford  <richard.sandiford@linaro.org>
26038         PR tree-optimization/81292
26039         * tree-ssa-strlen.c (handle_builtin_strlen): When setting
26040         full_string_p, also call adjust_related_strinfos if the adjustment
26041         is simple, otherwise invalidate related strinfos.
26043 2017-07-04  Martin Liska  <mliska@suse.cz>
26045         PR sanitizer/81040
26046         * sanopt.c (sanitize_rewrite_addressable_params): Mark the
26047         newly created variable as DECL_IGNORED_P.
26049 2017-07-04  Martin Liska  <mliska@suse.cz>
26051         PR ipa/81293
26052         * ipa-inline.c (inline_small_functions):
26053         Use xstrdup_for_dump.
26055 2017-07-04  Tom de Vries  <tom@codesourcery.com>
26057         * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
26059 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
26061         PR target/81033
26062         * config/darwin.c (darwin_function_switched_text_sections):
26063         Fix spaces.
26065 2017-07-03  Jan Hubicka  <hubicka@ucw.cz>
26067         * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
26069 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26071         * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
26073 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26075         * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
26076         min_profitable_iters, and th as inclusive lower bounds.
26077         Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
26078         (vect_estimate_min_profitable_iters): Return inclusive lower bounds
26079         for min_profitable_iters and min_profitable_estimate.
26080         (vect_transform_loop): Treat th as an inclusive lower bound.
26081         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
26083 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
26085         PR target/81033
26086         * config/darwin.c (darwin_function_switched_text_sections):
26087         Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
26088         in two pieces, and suppress the use of buf.
26090 2017-07-03  Nathan Sidwell  <nathan@acm.org>
26092         * hash-table.h (hash_table_mod1): Fix indentation.
26094 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26096         PR middle-end/81290
26097         * predict.c (force_edge_cold): Be more careful about propagation
26098         backward.
26099         * profile-count.h (profile_probability::guessed,
26100         profile_probability::fdo, profile_count::guessed, profile_count::fdo):
26101         New.
26102         * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
26104 2017-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
26106         * doc/invoke.texi (rcpc architecture extension): Document it.
26108 2017-07-03  Richard Biener  <rguenther@suse.de>
26110         PR tree-optimization/60510
26111         * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
26112         the scalar reduction PHI and use it.
26113         (vectorizable_reduction): Properly guard the single_defuse_cycle
26114         path for non-SLP reduction chains where we cannot use it.
26115         Rework reduc_def/index and vector type deduction.  Rework
26116         vector operand gathering during reduction op code-gen.
26117         * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
26118         chains dissolve the chain and leave it to non-SLP reduction
26119         handling.
26121 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26123         * tree-data-ref.h (dr_alignment): Declare.
26124         * tree-data-ref.c (dr_alignment): New function.
26125         * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
26126         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26127         set it.
26128         * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
26130 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26132         * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
26133         and base_misalignment fields.
26134         (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
26135         * tree-data-ref.c: Include builtins.h.
26136         (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
26137         * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
26138         (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
26139         * tree-vect-data-refs.c: Include tree-cfg.h.
26140         (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
26141         fields instead of calculating an alignment here.
26142         (vect_analyze_data_refs): Use dr_analyze_innermost.  Dump the new
26143         innermost_loop_behavior fields.
26145 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26147         * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
26148         field.
26149         (DR_STEP_ALIGNMENT): New macro.
26150         * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
26151         * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
26152         (create_data_ref): Print it.
26153         * tree-vect-stmts.c (vectorizable_load): Use the step alignment
26154         to tell whether the step preserves vector (mis)alignment.
26155         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26156         Move the check for an integer step and generalise to all INTEGER_CST.
26157         (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
26158         Print the outer step alignment.
26160 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26162         * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
26163         with offset_alignment.
26164         (DR_ALIGNED_TO): Delete.
26165         (DR_OFFSET_ALIGNMENT): New macro.
26166         * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
26167         (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
26168         * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
26169         (create_data_ref): Likewise.
26170         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26171         (vect_analyze_data_refs): Likewise.
26172         * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
26173         creating dummy innermost behavior.
26175 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26177         * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
26178         with a "innermost_loop_behavior *" and refeence tree.
26179         * tree-data-ref.c (dr_analyze_innermost): Likewise.
26180         (create_data_ref): Update call accordingly.
26181         * tree-predcom.c (find_looparound_phi): Likewise.
26183 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26185         * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
26186         fields with dr_wrt_vec_loop.
26187         (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
26188         (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
26189         (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
26190         (vect_dr_behavior): New function.
26191         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26192         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26193         vect_dr_behavior.  Use a step_preserves_misalignment_p boolean to
26194         track whether the step preserves the misalignment.
26195         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26196         Use vect_dr_behavior.
26197         (vect_setup_realignment): Update call accordingly.
26198         (vect_create_data_ref_ptr): Likewise.  Use vect_dr_behavior.
26199         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
26200         call to vect_create_addr_base_for_vector_ref.
26201         (vect_create_cond_for_align_checks): Likewise.
26202         * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
26203         STMT_VINFO_DR_WRT_VEC_LOOP as a block.
26204         (vect_recog_mask_conversion_pattern): Likewise.
26205         * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
26206         (new_stmt_vec_info): Remove redundant zeroing.
26208 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26210         * common/config/arm/arm-common.c (arm_be8_option): New function.
26211         * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
26212         (ISA_ARMv6): Add isa_bit_be8.
26213         * config/arm/arm.h (arm_be8_option): Add prototype.
26214         (BE8_SPEC_FUNCTION): New define.
26215         (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
26216         * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
26217         (mlittle-endian): Similarly.
26218         (mbe8, mbe32): New options.
26219         * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
26220         * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
26222 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26224         * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
26226 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26228         * tree-cfgcleanup.c (want_merge_blocks_p): New function.
26229         (cleanup_tree_cfg_bb): Use it.
26230         * profile-count.h (profile_count::of_for_merging, profile_count::merge):
26231         New functions.
26232         * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
26234 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26236         PR bootstrap/81285
26237         * loop-doloop.c (add_test): Update profile.
26239 2017-07-03  Martin Liska  <mliska@suse.cz>
26241         PR sanitize/81040
26242         * sanopt.c (rewrite_usage_of_param): New function.
26243         (sanitize_rewrite_addressable_params): Likewise.
26244         (pass_sanopt::execute): Call rewrite_usage_of_param.
26246 2017-07-03  Richard Biener  <rguenther@suse.de>
26248         * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26249         back to using VIEW_CONVERT_EXPR.
26251 2017-07-03  Martin Liska  <mliska@suse.cz>
26253         PR other/78366
26254         * doc/extend.texi: Document when a resolver function is
26255         generated for target_clones.
26257 2017-07-03  Martin Liska  <mliska@suse.cz>
26259         * asan.c (asan_emit_stack_protection): Unpoison just red zones
26260         and shadow memory of auto variables which are subject of
26261         use-after-scope sanitization.
26262         (asan_expand_mark_ifn): Add do set only when is_poison.
26264 2016-07-03  Richard Biener  <rguenther@suse.de>
26266         * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26267         reduction PHIs.
26268         (vect_force_simple_reduction): Record reduction def -> phi mapping.
26269         (vectorizable_reduction): Perform reduction PHI creation when
26270         visiting a reduction PHI and adjust and simplify code generation
26271         phase of the reduction op.  Cache dts, use fold_binary, not fold_build2.
26272         (vect_transform_loop): Visit reduction PHIs.
26273         * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26274         defs into the SLP tree.
26275         (vect_build_slp_tree): Reduction defs terminate the recursion.
26276         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26277         of reduction defs.
26278         (vect_get_vec_defs_for_stmt_copy): Export.
26279         (vect_get_vec_defs): Likewise.
26280         * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26281         purpose.
26282         (vect_get_vec_defs_for_stmt_copy): Declare.
26283         (vect_get_vec_defs): Likewise.
26285 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26287         * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26288         parameter with a "loop" parameter and use it instead of the
26289         loop containing DR_STMT.  Don't check simple_iv when doing
26290         BB analysis.  Describe the two analysis modes in the comment.
26292 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26294         PR tree-optimization/69468
26295         * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26296         (find_same_succ_bb): Handle ignore_edge_flags.
26298 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26300         PR tree-optimization/81192
26301         * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26302         hash.
26303         (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26304         differs.
26305         (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26307 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26309         PR tree-optimization/81192
26310         * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26311         BB_SAME_SUCC (bb) == NULL.
26313 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26315         * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26316         consistency.
26318 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26320         * dumpfile.c: Include profile-count.h
26321         * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26322         update profile.
26323         (insert_cond_bb): Update profile.
26324         * tree-cfg.h (insert_cond_bb): Update prototype.
26325         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26326         * tree-dump.c: Do not include tree-cfg.
26328 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26330         * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26332 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26334         * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26335         bb.
26337 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26339         * tree-complex.c (expand_complex_div_wide): update profile.
26341 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26342             Alan Hayward  <alan.hayward@arm.com>
26343             David Sherwood  <david.sherwood@arm.com>
26345         * Makefile.in (MACHMODE_H): Remove insn-modes.h
26346         (CORETYPES_H): New define.
26347         (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26348         (insn-modes-inline.h, s-modes-inline-h): New rules.
26349         (generated_files): Add insn-modes-inline.h.
26350         (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26351         (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26352         (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26353         (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26354         (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26355         (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26356         (build/gencodes.o, build/genconditions.o): Likewise.
26357         (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26358         (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26359         (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26360         (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26361         (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26362         (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26363         * coretypes.h: Include everything up to real.h for generators.
26364         Include insn-modes.h first.  Include wide-int-print.h after
26365         wide-int.h.  Include insn-modes-inline.h and then machmode.h.
26366         * machmode.h: Don't include insn-modes.h here.
26367         * function-tests.c: Remove includes of signop.h, machmode.h,
26368         double-int.h and wide-int.h.
26369         * rtl.h: Likewise.
26370         * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26371         and wide-int.h.
26372         * optc-save-gen.awk: Likewise.
26373         * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26374         * godump.c: Remove include of wide-int-print.h.
26375         * pretty-print.h: Likewise.
26376         * wide-int-print.cc: Likewise.
26377         * wide-int.cc: Likewise.
26378         * hash-map-tests.c: Remove include of signop.h.
26379         * hash-set-tests.c: Likewise.
26380         * rtl-tests.c: Likewise.
26381         * mkconfig.sh: Remove include of machmode.h.
26382         * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26383         into...
26384         (emit_insn_modes_inline_h): ...this new function.  Emit the code
26385         into an insn-modes-inline.h header file, adding appropriate
26386         include guards and end comments.
26387         (emit_insn_modes_c_header): Remove include of machmode.h.
26388         (emit_min_insn_modes_c_header): Include coretypes.h rather than
26389         machmode.h.
26390         (main): Handle -i flag and call emit_insn_modes_inline_h when
26391         it is passed.
26393 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26395         * tree-ssa-strlen.c (strinfo): Rename the length field to
26396         nonzero_chars.  Add a full_string_p field.
26397         (compare_nonzero_chars, zero_length_string_p): New functions.
26398         (get_addr_stridx): Add an offset_out parameter.
26399         Use compare_nonzero_chars.
26400         (get_stridx): Update accordingly.  Use compare_nonzero_chars.
26401         (new_strinfo): Update after above changes to strinfo.
26402         (set_endptr_and_length): Set full_string_p.
26403         (get_string_length): Update after above changes to strinfo.
26404         (unshare_strinfo): Update call to new_strinfo.
26405         (maybe_invalidate): Likewise.
26406         (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26407         Use compare_nonzero_chars and zero_string_p.  Treat nonzero_chars
26408         as a uhwi instead of an shwi.  Update after above changes to
26409         strinfo and new_strinfo.
26410         (zero_length_string): Assert that chainsi contains full strings.
26411         Use zero_length_string_p.  Update call to new_strinfo.
26412         (adjust_related_strinfos): Update after above changes to strinfo.
26413         Copy full_string_p from origsi.
26414         (adjust_last_stmt): Use zero_length_string_p.
26415         (handle_builtin_strlen): Update after above changes to strinfo and
26416         new_strinfo.  Install the lhs as the string length if the previous
26417         entry didn't describe a full string.
26418         (handle_builtin_strchr): Update after above changes to strinfo
26419         and new_strinfo.
26420         (handle_builtin_strcpy): Likewise.
26421         (handle_builtin_strcat): Likewise.
26422         (handle_builtin_malloc): Likewise.
26423         (handle_pointer_plus): Likewise.
26424         (handle_builtin_memcpy): Likewise.  Track nonzero characters
26425         that aren't necessarily followed by a nul terminator.
26426         (handle_char_store): Likewise.
26428 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26430         PR tree-optimization/80769
26431         * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26432         for malloc and calloc.  Document the new invariant that all related
26433         strinfos have delayed lengths or none do.
26434         (verify_related_strinfos): Move earlier in file.
26435         (set_endptr_and_length): New function, split out from...
26436         (get_string_length): ...here.  Also set the lengths of related
26437         strinfos.
26438         (zero_length_string): Assert that chainsi has known (rather than
26439         delayed) lengths.
26440         (adjust_related_strinfos): Likewise.
26442 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26444         PR tree-optimization/81136
26445         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26446         assert that two references with the same misalignment have the same
26447         compile-time misalignment if those compile-time misalignments
26448         are known.
26450 2017-07-01  Andi Kleen  <ak@linux.intel.com>
26452         * print-tree.c (print_node): Print all attributes.
26454 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26456         * cfg.c (scale_bbs_frequencies): New function.
26457         * cfg.h (scale_bbs_frequencies): Declare it.
26458         * cfgloopanal.c (single_likely_exit): Cleanup.
26459         * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26460         as parameter.
26461         (scale_loop_profile): Likewise.
26462         (loop_version): Likewise.
26463         (create_empty_loop_on_edge): Update.
26464         * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26465         scale_loop_frequencies, scale_loop_profile, loopify,
26466         loop_version): Update prototypes.
26467         * modulo-sched.c (sms_schedule): Update.
26468         * predict.c (unlikely_executed_edge_p): Also check probability.
26469         (probably_never_executed_edge_p): Fix typo.
26470         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26471         * tree-parloops.c (gen_parallel_loop): Update.
26472         * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26473         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26474         * tree-ssa-loop-split.c (split_loop): Update.
26475         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26476         * tree-vect-loop-manip.c (vect_do_peeling): Update.
26477         (vect_loop_versioning): Update.
26478         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26480 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26482         * trans-mem.c (split_bb_make_tm_edge): Update profile.
26484 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26486         * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26487         to keep profile consistent.
26489 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26491         * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26492         * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26493         * profile-count.h (max_safe_multiplier): Make unsigned.
26494         (profile_count::guessed_zero): New.
26496 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26498         * bb-reorder.c (fix_up_crossing_landing_pad,
26499         fix_crossing_conditional_branches): Use make_single_succ_edge
26500         to keep profile consistent.
26502 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26504         * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26505         to update profile.
26507 2017-07-01  Jakub Jelinek  <jakub@redhat.com>
26509         PR sanitizer/81262
26510         * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26511         the right scopes, make sure cond_jump isn't preserved between multiple
26512         iterations.  Search for fallthru edge whenever there are 3+ edges and
26513         use find_fallthru_edge for it.
26515 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26517         Patch by Alexander Monakov <amonakov@ispras.ru>
26518         * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26519         probabilities consistently.
26521 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26523         * pa.c (pa_expand_compare_and_swap_loop): Update call of
26524         emit_cmp_and_jump_insns.
26526 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26528         PR ipa/81261
26529         * tree-inline.c (expand_call_inline): Combine profile statuses.
26531 2017-06-30  Andrew Pinski  <apinski@cavium.com>
26533         * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26534         fold_stmt returned true.
26536 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26538         * ggc.h (empty_string): Delete.
26539         * cfgexpand.c (expand_asm_stmt): Use plain "".
26540         * optabs.c (expand_asm_memory_barrier): Likewise.
26541         * stringpool.c (empty_string): Delete.
26542         (digit_vector, digit_string): Delete.
26543         (ggc_alloc_string): Use plain "", don't optimize single digit
26544         strings.  Use ggc_alloc_atomic.
26546 2017-06-30  Richard Earnshaw  <rearnsha@arm.com>
26548         * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26549         comparison set and one other set, use the cost of the non-comparison
26550         set.
26552 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26554         * ggc.h: Replace all 'static inline' with plain 'inline'.  Fix
26555         some formatting.
26557 2017-06-30  Peter Bergner  <bergner@vnet.ibm.com>
26559         * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26560         loops.  Remove now unneeded calls to gimple_switch_set_label() that
26561         just set removed labels to NULL_TREE.
26563 2017-06-30  Aldy Hernandez  <aldyh@redhat.com>
26565         * tree-ssanames.c (set_range_info_raw): Abstract from ...
26566         (set_range_info): ...here.  Only call set_range_info_raw if domain
26567         is useful.
26568         (set_nonzero_bits): Call set_range_info_raw.
26569         * tree-ssanames.h (set_range_info_raw): New.
26571 2017-06-30  Jakub Jelinek  <jakub@redhat.com>
26573         PR target/81225
26574         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26575         V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26576         of nonimmediate_operand and <store_mask_constraint> instead of m
26577         for the input operand.  For V8FI iterator, always split if input
26578         is a MEM.  For V16FI and V8SF_256 iterators, don't test if both
26579         operands are MEM if <mask_applied>.  For VI4F_256 iterator, use
26580         <store_mask_predicate> instead of register_operand and
26581         <store_mask_constraint> instead of v for the input operand.  Make
26582         sure both operands aren't MEMs for if not <mask_applied>.
26584 2017-06-30  Sylvestre Ledru  <sylvestre@debian.org>
26586         * lto-wrapper.c (copy_file) Close both file descriptors before
26587         exiting normally.
26589 2017-06-30  Martin Liska  <mliska@suse.cz>
26591         PR ipa/81214
26592         * multiple_target.c (create_dispatcher_calls): Make ifunc
26593         also for function that don't have calls or are not referenced.
26595 2017-06-30  Richard Biener  <rguenther@suse.de>
26597         * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26598         analyze the first scalar stmt.  Move vector type computation
26599         for the BB case here from ...
26600         * tree-vect-stmts.c (vect_analyze_stmt): ... here.  Guard
26601         live operation processing in the SLP case properly.
26603 2017-06-30  Richard Biener  <rguenther@suse.de>
26605         * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26607 2017-06-30  Martin Liska  <mliska@suse.cz>
26609         PR sanitizer/81021
26610         * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26611         before BUILT_IN_UNWIND_RESUME when ASAN is used.
26613 2017-06-30  Yvan Roux  <yvan.roux@linaro.org>
26615         * doc/invoke.texi (AArch64): Add missing options and remove redundant
26616         ones.
26618 2017-06-30  Richard Biener  <rguenther@suse.de>
26620         PR tree-optimization/81249
26621         * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26622         condition reduction result to original scalar type.
26624 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26626         * profile-count.h (enum profile_quality): Fix typos and whitespace
26627         issues.
26629 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26631         * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26632         type for branch probabilities.
26634 2017-06-29  Julian Brown  <julian@codesourcery.com>
26635             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26637         * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26638         * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26639         (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26640         (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26642 2017-06-29  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26644         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26645         check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26646         * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26647         CC usage from generic code to here.
26648         * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26649         CC usage into the target macros.
26651 2017-06-29  Maya Rashish  <coypu@sdf.org>
26653         * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26654         objects.
26656 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26658         * arm/arm-builtins.c: Include profile-count.h
26659         * except.c (sjlj_emit_function_enter): Use
26660         profile_probability::unlikely.
26662 2017-06-29  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26664         * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26665         and tocrel_offset be pointer args rather than implicitly using
26666         static versions.
26667         (legitimate_constant_pool_address_p, rs6000_emit_move,
26668         const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26669         tocrel_offset and use in toc_relative_expr_p call.
26670         (print_operand, print_operand_address): Use static tocrel_base_oac
26671         and tocrel_offset_oac.
26672         (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26673         tocrel_offset_oac.
26675 2017-06-29  Maya Rashish  <coypu@sdf.org>
26677         * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26679 2017-06-29  Eric Botcazou  <ebotcazou@adacore.com>
26681         * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26682         objects, take into account only the alignment of 'op0' and 'mode1' if
26683         'op0' is a MEM.
26685 2017-06-29  Steve Ellcey  <sellcey@cavium.com>
26687         * ccmp.c (ccmp_tree_comparison_p): New function.
26688         (ccmp_candidate_p): Update to use above function.
26689         (get_compare_parts): New function.
26690         (expand_ccmp_next): Update to use new functions.
26691         (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26692         new functions.
26693         (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26694         take mode as argument.
26695         * ccmp.h (expand_ccmp_expr): Add mode as argument.
26696         * expr.c (expand_expr_real_1): Pass mode as argument.
26698 2017-06-29  Segher Boessenkool  <segher@kernel.crashing.org>
26700         * combine.c (combine_instructions): Print insns to dump_file, together
26701         with their costs.
26703 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26705         * asan.c (asan_emit_stack_protection): Update.
26706         (create_cond_insert_point): Update.
26707         * auto-profile.c (afdo_propagate_circuit): Update.
26708         * basic-block.h (struct edge_def): Turn probability to
26709         profile_probability.
26710         (EDGE_FREQUENCY): Update.
26711         * bb-reorder.c (find_traces_1_round): Update.
26712         (better_edge_p): Update.
26713         (sanitize_hot_paths): Update.
26714         * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26715         (make_single_succ_edge): Update.
26716         (check_bb_profile): Update.
26717         (dump_edge_info): Update.
26718         (update_bb_profile_for_threading): Update.
26719         * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26720         probabilitycount to 0.
26721         * cfgbuild.c (compute_outgoing_frequencies): Update.
26722         * cfgcleanup.c (try_forward_edges): Update.
26723         (outgoing_edges_match): Update.
26724         (try_crossjump_to_edge): Update.
26725         * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26726         (expand_gimple_tailcall): Update.
26727         (construct_init_block): Use make_single_succ_edge.
26728         (construct_exit_block): Use make_single_succ_edge.
26729         * cfghooks.c (verify_flow_info): Update.
26730         (redirect_edge_succ_nodup): Update.
26731         (split_edge): Update.
26732         (account_profile_record): Update.
26733         * cfgloopanal.c (single_likely_exit): Update.
26734         * cfgloopmanip.c (scale_loop_profile): Update.
26735         (set_zero_probability): Remove.
26736         (duplicate_loop_to_header_edge): Update.
26737         * cfgloopmanip.h (loop_version): Update prototype.
26738         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26739         (force_nonfallthru_and_redirect): Update.
26740         (update_br_prob_note): Update.
26741         (rtl_verify_edges): Update.
26742         (purge_dead_edges): Update.
26743         (rtl_lv_add_condition_to_bb): Update.
26744         * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26745         * cgraphunit.c (init_lowered_empty_function): Update.
26746         (cgraph_node::expand_thunk): Update.
26747         * cilk-common.c: Include profile-count.h
26748         * dojump.c (inv): Remove.
26749         (jumpifnot): Update.
26750         (jumpifnot_1): Update.
26751         (do_jump_1): Update.
26752         (do_jump): Update.
26753         (do_jump_by_parts_greater_rtx): Update.
26754         (do_compare_rtx_and_jump): Update.
26755         * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26756         do_jump_1. do_compare_rtx_and_jump): Update prototype.
26757         * dwarf2cfi.c: Include profile-count.h
26758         * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26759         (sjlj_emit_dispatch_table): Likewise.
26760         * explow.c: Include profile-count.h
26761         * expmed.c (emit_store_flag_force): Update.
26762         (do_cmp_and_jump): Update.
26763         * expr.c (compare_by_pieces_d::generate): Update.
26764         (compare_by_pieces_d::finish_mode): Update.
26765         (emit_block_move_via_loop): Update.
26766         (store_expr_with_bounds): Update.
26767         (store_constructor): Update.
26768         (expand_expr_real_2): Update.
26769         (expand_expr_real_1): Update.
26770         * expr.h (try_casesi, try_tablejump): Update prototypes.
26771         * gimple-pretty-print.c (dump_probability): Update.
26772         (dump_profile): New.
26773         (dump_gimple_label): Update.
26774         (dump_gimple_bb_header): Update.
26775         * graph.c (draw_cfg_node_succ_edges): Update.
26776         * hsa-gen.c (convert_switch_statements): Update.
26777         * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26778         (find_if_case_1): Update.
26779         (find_if_case_2): Update.
26780         * internal-fn.c (expand_arith_overflow_result_store): Update.
26781         (expand_addsub_overflow): Update.
26782         (expand_neg_overflow): Update.
26783         (expand_mul_overflow): Update.
26784         (expand_vector_ubsan_overflow): Update.
26785         * ipa-cp.c (good_cloning_opportunity_p): Update.
26786         * ipa-split.c (split_function): Use make_single_succ_edge.
26787         * ipa-utils.c (ipa_merge_profiles): Update.
26788         * loop-doloop.c (add_test): Update.
26789         (doloop_modify): Update.
26790         * loop-unroll.c (compare_and_jump_seq): Update.
26791         (unroll_loop_runtime_iterations): Update.
26792         * lra-constraints.c (lra_inheritance): Update.
26793         * lto-streamer-in.c (input_cfg): Update.
26794         * lto-streamer-out.c (output_cfg): Update.
26795         * mcf.c (adjust_cfg_counts): Update.
26796         * modulo-sched.c (sms_schedule): Update.
26797         * omp-expand.c (expand_omp_for_init_counts): Update.
26798         (extract_omp_for_update_vars): Update.
26799         (expand_omp_ordered_sink): Update.
26800         (expand_omp_for_ordered_loops): Update.
26801         (expand_omp_for_generic): Update.
26802         (expand_omp_for_static_nochunk): Update.
26803         (expand_omp_for_static_chunk): Update.
26804         (expand_cilk_for): Update.
26805         (expand_omp_simd): Update.
26806         (expand_omp_taskloop_for_outer): Update.
26807         (expand_omp_taskloop_for_inner): Update.
26808         * omp-simd-clone.c (simd_clone_adjust): Update.
26809         * optabs.c (expand_doubleword_shift): Update.
26810         (expand_abs): Update.
26811         (emit_cmp_and_jump_insn_1): Update.
26812         (expand_compare_and_swap_loop): Update.
26813         * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26814         * predict.c (predictable_edge_p): Update.
26815         (edge_probability_reliable_p): Update.
26816         (set_even_probabilities): Update.
26817         (combine_predictions_for_insn): Update.
26818         (combine_predictions_for_bb): Update.
26819         (propagate_freq): Update.
26820         (estimate_bb_frequencies): Update.
26821         (force_edge_cold): Update.
26822         * profile-count.c (profile_count::dump): Add missing space into dump.
26823         (profile_count::debug): Add newline.
26824         (profile_count::differs_from_p): Explicitly convert to unsigned.
26825         (profile_count::stream_in): Update.
26826         (profile_probability::dump): New member function.
26827         (profile_probability::debug): New member function.
26828         (profile_probability::differs_from_p): New member function.
26829         (profile_probability::differs_lot_from_p): New member function.
26830         (profile_probability::stream_in): New member function.
26831         (profile_probability::stream_out): New member function.
26832         * profile-count.h (profile_count_quality): Rename to ...
26833         (profile_quality): ... this one.
26834         (profile_probability): New.
26835         (profile_count): Update.
26836         * profile.c (compute_branch_probabilities): Update.
26837         * recog.c (peep2_attempt): Update.
26838         * sched-ebb.c (schedule_ebbs): Update.
26839         * sched-rgn.c (find_single_block_region): Update.
26840         (compute_dom_prob_ps): Update.
26841         (schedule_region): Update.
26842         * sel-sched-ir.c (compute_succs_info): Update.
26843         * stmt.c (struct case_node): Update.
26844         (do_jump_if_equal): Update.
26845         (get_outgoing_edge_probs): Update.
26846         (conditional_probability): Update.
26847         (emit_case_dispatch_table): Update.
26848         (expand_case): Update.
26849         (expand_sjlj_dispatch_table): Update.
26850         (emit_case_nodes): Update.
26851         * targhooks.c: Update.
26852         * tracer.c (better_p): Update.
26853         (find_best_successor): Update.
26854         * trans-mem.c (expand_transaction): Update.
26855         * tree-call-cdce.c: Update.
26856         * tree-cfg.c (gimple_split_edge): Upate.
26857         (move_sese_region_to_fn): Upate.
26858         * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26859         * tree-eh.c (lower_resx): Upate.
26860         (cleanup_empty_eh_move_lp): Upate.
26861         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26862         * tree-inline.c (copy_edges_for_bb): Update.
26863         (copy_cfg_body): Update.
26864         * tree-parloops.c (gen_parallel_loop): Update.
26865         * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26866         (gimple_gen_time_profiler): Update.
26867         * tree-ssa-dce.c (remove_dead_stmt): Update.
26868         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26869         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26870         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26871         (unloop_loops): Update.
26872         (try_peel_loop): Update.
26873         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26874         * tree-ssa-loop-split.c (connect_loops): Update.
26875         (split_loop): Update.
26876         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26877         (hoist_guard): Update.
26878         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26879         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26880         (value_replacement): Update.
26881         * tree-ssa-reassoc.c (branch_fixup): Update.
26882         * tree-ssa-tail-merge.c (replace_block_by): Update.
26883         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26884         (create_edge_and_update_destination_phis): Update.
26885         (compute_path_counts): Update.
26886         (recompute_probabilities): Update.
26887         (update_joiner_offpath_counts): Update.
26888         (freqs_to_counts_path): Update.
26889         (duplicate_thread_path): Update.
26890         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26891         (struct switch_conv_info): Update.
26892         (gen_inbound_check): Update.
26893         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26894         (vect_do_peeling): Update.
26895         (vect_loop_versioning): Update.
26896         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26897         (optimize_mask_stores): Update.
26898         * ubsan.c (ubsan_expand_null_ifn): Update.
26899         * value-prof.c (gimple_divmod_fixed_value): Update.
26900         (gimple_divmod_fixed_value_transform): Update.
26901         (gimple_mod_pow2): Update.
26902         (gimple_mod_pow2_value_transform): Update.
26903         (gimple_mod_subtract): Update.
26904         (gimple_mod_subtract_transform): Update.
26905         (gimple_ic): Update.
26906         (gimple_stringop_fixed_value): Update.
26907         (gimple_stringops_transform): Update.
26908         * value-prof.h: Update.
26910 2017-06-29  Carl Love  <cel@us.ibm.com>
26912         * config/rs6000/rs6000-c.c: Add support for built-in functions
26913         vector signed int vec_signed (vector float);
26914         vector signed long long vec_signed (vector double);
26915         vector signed int vec_signed2 (vector double, vector double);
26916         vector signed int vec_signede (vector double);
26917         vector signed int vec_signedo (vector double);
26918         * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
26919         instruction generator.
26920         * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26921         UNSPEC_VSX_VSIGNED2): Add UNSPECS.
26922         (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
26923         Add define_insn.
26924         (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
26925         vunsignede_v2df): Add define_expands.
26926         * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
26927         VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
26928         VEC_UNSIGNEDO): Add definitions.
26929         * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26930         UNSPEC_VSX_VSIGNED2): Add UNSPECs.
26931         (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
26932         (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
26933         vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
26934         * config/rs6000/altivec.h (vec_signed, vec_signed2,
26935         vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
26936         vec_unsignede, vec_unsignedo): Add builtin defines.
26937         * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
26938         declaration.
26939         * doc/extend.texi: Update the built-in documentation file for the
26940         new built-in functions.
26942 2017-06-29  Richard Biener  <rguenther@suse.de>
26944         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
26945         reduction chains to LOOP_VINFO_REDUCTIONS.
26946         * tree-vect-slp.c (vect_analyze_slp): Continue looking for
26947         SLP reductions after processing reduction chains.
26949 2017-06-29  Nathan Sidwell  <nathan@acm.org>
26951         * builtins.c (fold_builtin_FUNCTION): Use
26952         lang_hooks.decl_printable_name.
26954 2017-06-29  Peter Bergner  <bergner@vnet.ibm.com>
26956         PR middle-end/81194
26957         * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
26958         with only one label.
26959         * stmt.c (expand_case): Assert NCASES is greater than one.
26961 2017-06-29  Richard Biener  <rguenther@suse.de>
26963         * tree-cfg.c (group_case_labels_stmt): Return whether we changed
26964         anything.
26965         (group_case_labels): Likewise.
26966         (find_taken_edge): Push sanity checking on val to workers...
26967         (find_taken_edge_cond_expr): ... here
26968         (find_taken_edge_switch_expr): ... and here, handle cases
26969         with just a default label.
26970         * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
26971         (group_case_labels): Likewise.
26972         * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
26973         group_case_labels does anything cleanup the CFG again.
26975 2017-06-29  Bin Cheng  <bin.cheng@arm.com>
26977         PR tree-optimization/81196
26978         * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
26979         exit condition comparing two IVs.
26981 2017-06-29  Richard Earnshaw  <rearnsha@arm.com>
26983         * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
26984         profile to the dummy entry at the end of the list of architectures.
26985         * config/arm/arm-cpu-cdata.h: Regenerated.
26987 2017-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26988             Michael Collison <michael.collison@arm.com>
26990         PR target/70119
26991         * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
26992         New pattern.
26993         (*aarch64_reg_<mode>3_neg_mask2): New pattern.
26994         (*aarch64_reg_<mode>3_minus_mask): New pattern.
26995         (*aarch64_<optab>_reg_di3_mask2): New pattern.
26996         * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
26997         of shift when the shift amount is masked with constant equal to
26998         the size of the mode.
26999         * config/aarch64/predicates.md (subreg_lowpart_operator): New
27000         predicate.
27002 2017-06-29  Martin Liska  <mliska@suse.cz>
27004         * config/i386/i386.opt: Change range from [1,5] to [0,5].
27006 2017-06-29  Yury Gribov  <tetra2005@gmail.com>
27008         PR bootstrap/80565
27009         * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
27010         code.
27011         * ipa-inline.h
27012         (edge_growth_cache_entry::edge_growth_cache_entry): New
27013         function.
27014         (reset_edge_growth_cache): Update to use constructor.
27016 2017-06-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
27018         * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
27019         (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
27020         (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
27022 2017-06-28  Sebastian Peryt  <sebastian.peryt@intel.com>
27024         * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
27025         (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
27027 2017-06-28  Szabolcs Nagy  <szabolcs.nagy@arm.com>
27029         * config.gcc (*-linux-musl*): Add t-musl tmake_file.
27030         (*-linux-uclibc*): Add t-uclibc tmake_file.
27031         * config/t-musl: New.
27032         * config/t-uclibc: New.
27034 2017-06-28  Richard Earnshaw  <rearnsha@arm.com>
27036         * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
27037         context.
27038         (gen_comm_data): Emit architectural setting of arch_prof.
27039         * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
27040         profile.
27041         (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
27042         (armv8-m.base, armv8-m.main): Likewise.
27043         * arm-protos.h (arm_build_target): Add profile field.
27044         (arch_option): Likewise.
27045         * config/arm/arm.c (arm_configure_build_target): Copy the profile to
27046         the active target.
27047         * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
27048         arm_active_target.profile.
27050 2017-06-28  Richard Biener  <rguenther@suse.de>
27052         PR middle-end/81227
27053         * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
27054         TYPE_OVERFLOW_WRAPS.
27055         * match.pd (negate_expr_p): Likewise.
27056         * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
27057         fold_build2, not fold_binary.
27059 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
27061         * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
27062         Convert memory address to Pmode.
27063         (aarch64_print_operand): Assert MEM operands are always Pmode.
27065 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
27067         PR target/79665
27068         * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
27069         Remove redundant if.
27070         (aarch_forward_to_shift_is_not_shifted_reg): Remove.
27071         * config/arm/aarch-common-protos.h
27072         (aarch_forward_to_shift_is_not_shifted_re): Remove.
27073         * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
27075 2017-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
27077         PR ipa/81238
27078         * multiple_target.c (create_dispatcher_calls): Set the default
27079         clone to be static, not public.
27081 2017-06-28  Richard Biener  <rguenther@suse.de>
27083         * tree-vect-loop.c (vectorizable_reduction): Move special
27084         cond reduction IV var creation ...
27085         (vect_create_epilog_for_reduction): ... here.  Remove induction_index
27086         parameter.  Use STMT_VINFO_VECTYPE.
27087         * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
27088         constant_p.
27090 2017-06-28  Martin Liska  <mliska@suse.cz>
27092         PR ipa/81128
27093         * ipa-visibility.c (non_local_p): Handle visibility.
27095 2017-06-28  Martin Liska  <mliska@suse.cz>
27097         PR driver/79659
27098         * common.opt: Add IntegerRange to various options.
27099         * opt-functions.awk (integer_range_info): New function.
27100         * optc-gen.awk: Add integer_range_info to cl_options struct.
27101         * opts-common.c (decode_cmdline_option): Handle
27102         CL_ERR_INT_RANGE_ARG.
27103         (cmdline_handle_error): Likewise.
27104         * opts.c (print_filtered_help): Show valid interval in
27105         when --help is provided.
27106         * opts.h (struct cl_option): Add range_min and range_max fields.
27107         * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
27109 2017-06-28  Marc Glisse  <marc.glisse@inria.fr>
27111         * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
27112         (x * C EQ/NE y * C): New transformation.
27114 2017-06-28  Christophe Lyon  <christophe.lyon@linaro.org>
27116         * genmultilib (combination_space): Accept '+' in option names.
27118 2017-06-28  Martin Liska  <mliska@suse.cz>
27120         PR sanitizer/81224
27121         * asan.c (instrument_derefs): Bail out inner references
27122         that are hard register variables.
27124 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
27126         PR target/81175
27127         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
27128         rather than def_builtin_pure for __builtin_ia32_gatherpf*.
27130 2017-06-28  Richard Biener  <rguenther@suse.de>
27132         * tree-vectorizer.h (vect_get_vec_defs): Remove.
27133         (vect_get_slp_defs): Adjust.
27134         * tree-vect-loop.c (get_initial_defs_for_reduction): Split
27135         out from ...
27136         * tree-vect-slp.c (vect_get_constant_vectors): ... here and
27137         simplify.
27138         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
27139         get_initial_defs_for_reduction instead of vect_get_vec_defs.
27140         (vectorizable_reduction): Adjust.
27141         * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
27142         handling.
27143         (vect_get_slp_defs): Likewise.
27144         * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
27145         (vectorizable_bswap): Adjust.
27146         (vectorizable_call): Likewise.
27147         (vectorizable_conversion): Likewise.
27148         (vectorizable_assignment): Likewise.
27149         (vectorizable_shift): Likewise.
27150         (vectorizable_operation): Likewise.
27151         (vectorizable_store): Likewise.
27152         (vectorizable_condition): Likewise.
27153         (vectorizable_comparison): Likewise.
27155 2017-06-28  Michael Collison  <michael.collison@arm.com>
27157         PR target/68535
27158         * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
27159         set of base_reg
27160         (arm_gen_movmemqi): Removed unused variable 'i'.
27161         Convert 'for' loop into 'while' loop.
27162         (arm_expand_prologue): Remove last unnecessary set of insn.
27163         (thumb_pop): Remove unused variable 'pushed_words'.
27164         (thumb_exit): Remove last unnecessary set of regs_to_pop.
27166 2017-06-28  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27168         * config/s390/predicates.md: Use s390_rel_address_ok_p.
27169         * config/s390/s390-protos.h: Add prototype of
27170         s390_rel_address_ok_p.
27171         * config/s390/s390.c (s390_got_symbol): New function.
27172         (s390_rel_address_ok_p): New function.
27173         (legitimize_pic_address): Use s390_rel_address_ok_p.
27174         (s390_load_got): Use s390_got_symbol.
27175         (s390_option_override): Issue error if
27176         -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
27177         * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
27178         New macro.
27179         * config/s390/s390.opt: New option mpic-data-is-text-relative.
27181 2017-06-27  Andrew Pinski  <apinski@cavium.com>
27183         * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
27184         (X * copysign (1.0, X)): New pattern.
27185         (X * copysign (1.0, -X)): New pattern.
27186         (copysign (-1.0, CST)): New pattern.
27188 2017-06-27  Joseph Myers  <joseph@codesourcery.com>
27190         * genmultilib (combination_space): Remove variable.
27191         Validate reuse rules against regular expression for any sequence
27192         of multilib options in any order.
27194 2017-06-27  Michael Collison  <michael.collison@arm.com>
27196         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
27197         call aarch64_split_simd_combine.
27198         * (aarch64_combine_internal<mode>): Delete pattern.
27199         * config/aarch64/aarch64.c (aarch64_split_simd_combine):
27200         Allow register and subreg operands.
27202 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27204         * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
27205         specific need, just fallback on defaults.
27206         (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
27208 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27209             Olivier Hainque  <hainque@adacore.com>
27211         * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
27212         map for 64bits.
27213         (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
27214         targets. Pick a default if no particular attempt applied.
27215         (STACK_CHECK_PROTECT): Double for 64bit targets, which have
27216         larger contexts.
27218 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27220         * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
27221         (x86_64-wrs-vxworks7): Likewise.
27223 2017-06-27  Marek Polacek  <polacek@redhat.com>
27225         PR sanitizer/81223
27226         * ubsan.c (instrument_null): Check get_base_address's result for null.
27228 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27230         * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
27232 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27234         * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
27235         BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
27236         (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
27237         BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
27238         New function types.
27239         * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
27240         BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
27241         BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
27242         BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
27243         BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27244         BUILT_IN_FEUPDATEENV): New builtins.
27245         * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27246         TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27247         * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27248         fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27249         macros.
27250         (builtin_structptr_types): Adjust size.
27251         * tree.c (builtin_structptr_types): Add four entries.
27253 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27254             Olivier Hainque  <hainque@adacore.com>
27256         * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27257         (TLS_SYM): New local macro, forcing reference to __tls__ on
27258         link command lines for VxWorks 7 RTPs, triggering initialization
27259         of tlsLib.
27260         (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27261         OS features TLS support, true for RTPs on VxWorks 7.
27262         * config/vxworks.c (vxworks_override_options): Setup emutls
27263         accordingly.
27265 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27267         * predict.c (test_prediction_value_range): Use -1U instead of -1
27268         to avoid narrowing conversion warning.
27269         * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27270         to avoid narrowing conversion warning.
27271         * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27272         -1.
27273         * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27275 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27277         * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27278         64bit configurations.
27279         (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27280         (SIZE_TYPE): Likewise.
27281         * config/vxworks.c (vxworks_emutls_var_fields): Use
27282         long_unsigned_type_node instead of unsigned_type_node as the offset
27283         field type, which is "pointer" mode in emutls.c.
27285 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27287         PR sanitizer/81209
27288         * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27290         PR middle-end/81207
27291         * gimple-fold.c (replace_call_with_call_and_fold): Handle
27292         gimple_vuse copying separately from gimple_vdef copying.
27294 2017-06-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27296         * value-prof.c (free_hist): Remove call to memset and the enclosing if
27297         condition.
27299 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27300             Olivier Hainque  <hainque@adacore.com>
27302         * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27303         for all vxworks7 targets.
27304         * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27305         (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27306         (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27307         variations for VX6/VX7 and 32/64bits later on in ...
27308         (VXWORKS_LIB_SPEC): Leverage new macros.
27309         (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27310         as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27312 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27314         * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27315         _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27317 2017-06-26  Carl Love  <cel@us.ibm.com>
27319         * config/rs6000/rs6000-c.c: Add support for built-in functions
27320         vector bool char vec_reve (vector bool char);
27321         vector signed char vec_reve (vector signed char);
27322         vector unsigned char vec_reve (vector unsigned char);
27323         vector bool int vec_reve (vector bool int);
27324         vector signed int vec_reve (vector signed int);
27325         vector unsigned int vec_reve (vector unsigned int);
27326         vector bool long long vec_reve (vector bool long long);
27327         vector signed long long vec_reve (vector signed long long);
27328         vector unsigned long long vec_reve (vector unsigned long long);
27329         vector bool short vec_reve (vector bool short);
27330         vector signed short vec_reve (vector signed short);
27331         vector double vec_reve (vector double);
27332         vector float vec_reve (vector float);
27333         * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27334         VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27335         * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27336         (altivec_vreve): New pattern.
27337         * config/rs6000/altivec.h (vec_reve): New define.
27338         * doc/extend.texi (vec_rev): Update the built-in documentation file
27339         for the new built-in functions.
27341 2016-06-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27343         PR tree-optimization/71815
27344         * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27345         function.
27346         (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27347         has_single_use.
27348         (slsr_process_phi): Likewise.
27349         (replace_uncond_cands_and_profitable_phis): Don't replace a
27350         multiply candidate with a stride of 1 (copy or cast).
27351         (phi_incr_cost): Call uses_consumed_by_stmt rather than
27352         has_single_use.
27353         (lowest_cost_path): Likewise.
27354         (total_savings): Likewise.
27356 2017-06-26  Richard Biener  <rguenther@suse.de>
27358         PR target/81175
27359         * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27360         Use def_builtin_pure for all gather builtins.
27362 2017-06-26  Richard Biener  <rguenther@suse.de>
27364         PR tree-optimization/81203
27365         * tree-tailcall.c (find_tail_calls): Do not move stmts into
27366         non-dominating BBs.
27368 2017-06-26  Marek Polacek  <polacek@redhat.com>
27370         PR c/80116
27371         * doc/invoke.texi: Document -Wmultistatement-macros.
27373 2017-06-26  Christophe Lyon  <christophe.lyon@linaro.org>
27375         * doc/sourcebuild.texi (ARM-specific attributes): Document new
27376         arm_neon_ok_no_float_abi effective target.
27378 2017-06-26  Richard Biener  <rguenther@suse.de>
27380         PR tree-optimization/80928
27381         * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27382         (copy_bbs): Set BB_DUPLICATED flag early.
27383         (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27384         marked blocks.
27385         (execute_on_shrinking_pred): Likewise.
27386         * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27387         BB_DUPLICATED blocks.
27388         * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27389         iterate over all PHIs considering removal of *gsi.
27391 2017-06-23  Jim Wilson  <jim.wilson@linaro.org>
27393         * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27394         qdf24xx.
27396 2017-06-23  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27398         * config/rs6000/rs6000-string.c: (expand_block_clear,
27399         do_load_for_compare, select_block_compare_mode,
27400         compute_current_alignment, expand_block_compare,
27401         expand_strncmp_align_check, expand_strn_compare,
27402         expand_block_move, rs6000_output_load_multiple)
27403         Move functions related to string/block move/compare
27404         to a separate file.
27405         * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27406         * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27407         for this function which is now used in two files.
27408         * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27409         * config.gcc: Add rs6000-string.o to extra_objs for
27410         targets powerpc*-*-* and rs6000*-*-*.
27412 2017-06-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
27414         PR target/80510
27415         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27416         32-bit, since indexed is not valid for DImode.
27417         (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27418         3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27419         (define_peephole2 for Altivec d-form load): Add 32-bit support.
27420         (define_peephole2 for Altivec d-form store): Likewise.
27422         PR ipa/81185
27423         * multiple_target.c (create_dispatcher_calls): Only create the
27424         dispatcher call if the function is the default clone of a
27425         versioned function.
27427 2017-06-23  Segher Boessenkool  <segher@kernel.crashing.org>
27429         PR middle-end/80902
27430         * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27431         a call, force the call to not be a tail call.
27433 2017-06-23  Jeff Law  <law@redhat.com>
27435         * doc/contrib.texi: Add entry for Steven Pemberton's work on
27436         enquire.
27438 2017-06-23  Will Schmidt  <will_schmidt@vnet.ibm.com>
27440         * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27441         update_call_from_tree().  (rs6000_gimple_fold_builtin): Add
27442         handling for early expansion of vector shifts (sl,sr,sra,rl).
27443         (builtin_function_type): Add vector shift right instructions
27444         to the unsigned argument list.
27446 2017-06-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27448         rtl-optimizatoin/79286
27449         * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27450         * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27451         trap.  PIC register plus a const unspec without offset can never trap.
27453 2017-06-23  Marc Glisse  <marc.glisse@inria.fr>
27455         * tree.h (builtin_structptr_type): New type.
27456         (builtin_structptr_types): Declare new array.
27457         * tree.c (builtin_structptr_types): New array.
27458         (free_lang_data, build_common_tree_nodes): Use it.
27460 2017-06-23  Jonathan Wakely  <jwakely@redhat.com>
27462         PR c++/81187
27463         * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27464         -Wnoexcept.
27466 2017-06-22  Matt Turner  <mattst88@gmail.com>
27468         * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27469         Lake models to skylake case.  Assume skylake for unknown
27470         models with clflushopt.
27472 2017-06-22  Jeff Law  <law@redhat.com>
27474         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27475         frame sizes that do not satisfy aarch64_uimm12_shift.
27477 2017-06-22  Jan Hubicka <hubicka@ucw.cz>
27479         * profile-count.h (apply_probability,
27480         apply_scale, probability_in): Fix checks for zero.
27482 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27484         * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27485         * doc/cppdiropts.texi (-I @var{dir}): Document it.
27487 2016-06-22  Richard Biener  <rguenther@suse.de>
27489         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27490         COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27491         REDUC_MAX_EXPR support.
27492         (vectorizable_reduction): Likewise.
27493         (vect_create_epilog_for_reduction): Likewise.
27495 2017-06-22  James Greenhalgh  <james.greenhalgh@arm.com>
27497         * match.pd (A / (1 << B) -> A >> B): New.
27498         * generic-match-head.c: Include optabs-tree.h.
27499         * gimple-match-head.c: Likewise.
27500         * optabs-tree.h (target_supports_op_p): New.
27501         * optabs-tree.c (target_supports_op_p): New.
27503 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27505         * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27506         $gcc_cv_ld --help output.
27507         (gcc_cv_ld_demangle): Likewise.
27508         (gcc_cv_ld_eh_frame_hdr): Likewise.
27509         (gcc_cv_ld_pie): Likewise.
27510         (gcc_cv_ld_as_needed): Likewise.  Prefer native forms unless $gnu_ld.
27511         (gcc_cv_ld_buildid): Likewise.
27512         (gcc_cv_ld_sysroot): Likewise.
27513         (ld_bndplt_support): Likewise.
27514         (ld_pushpopstate_support): Likewise.
27515         * configure: Regenerate.
27516         * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27518 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27520         PR target/81151
27521         * config/i386/sse.md (round<mode>2): Renumber match_dup and
27522         operands indexes to avoid gap between operands and match_dups.
27524 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27526         * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27527         Increment Arith_shift and Arith_shift_reg by 1.
27528         * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27529         New tuning flag.
27530         * config/aarch64/aarch64.c (thunderx_tunings): Enable
27531         AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27532         (aarch64_strip_extend): Add new argument and test for it.
27533         (aarch64_cheap_mult_shift_p): New function.
27534         (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27535         add a cost if it is true.
27536         Update calls to aarch64_strip_extend.
27537         (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27539 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27541         * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27542         tunings.
27543         (thunderxt88): Likewise.
27544         * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27545         (thunderx_prefetch_tune): New variable.
27546         (thunderx2t99_prefetch_tune): Update for the correct values.
27547         (thunderxt88_tunings): New variable.
27548         (thunderx_tunings): Use thunderx_prefetch_tune instead of
27549         generic_prefetch_tune.
27550         (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27552 2017-06-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27554         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27555         SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27556         (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27557         (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27558         (aarch64_atomic_cas<mode>, GPI): Likewise.
27560 2017-06-21  Martin Liska  <mliska@suse.cz>
27562         * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27563         statements on cold and hot labels.
27564         * predict.c (tree_estimate_probability_bb): Remove the
27565         prediction from this place.
27567 2017-06-21  Martin Liska  <mliska@suse.cz>
27569         PR tree-optimization/79489
27570         * gimplify.c (maybe_add_early_return_predict_stmt): New
27571         function.
27572         (gimplify_return_expr): Call the function.
27573         * predict.c (tree_estimate_probability_bb): Remove handling
27574         of early return.
27575         * predict.def: Update comment about early return predictor.
27576         * gimple-predict.h (is_gimple_predict): New function.
27577         * predict.def: Change default value of early return to 66.
27578         * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27579         statements.
27580         * passes.def: Put pass_strip_predict_hints to the beginning of
27581         IPA passes.
27583 2017-06-21  Pierre-Marie de Rodat  <derodat@adacore.com>
27585         * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27586         FUNCTION_DECL declarations.
27587         (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27588         declarations.
27589         (dwaf2out_decl): Likewise.
27590         * godump.c (go_early_global_decl): Skip call to the real debug hook
27591         for FUNCTION_DECL declarations.
27592         * passes.c (rest_of_decl_compilation): Skip call to the
27593         early_global_decl debug hook for FUNCTION_DECL declarations, unless
27594         -fdump-go-spec is passed.
27596 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27598         * config/i386/i386.c (struct builtin_isa): New field pure_p.
27599         Reorder for compactness.
27600         (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27601         (def_builtin_pure, def_builtin_pure2): New functions.
27602         (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27604 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27606         * match.pd (nop_convert): New predicate.
27607         ((A +- CST1) +- CST2): Allow some NOP conversions.
27609 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27611         PR c++/81130
27612         * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27613         with ctors/dtors if GOVD_SHARED is set.
27615 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27617         * config/aarch64/aarch64.md (movti_aarch64):
27618         Emit mov rather than orr.
27619         (movtf_aarch64): Likewise.
27620         * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27621         Emit mov rather than orr.
27623 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27625         * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27626         Swap alternatives, make integer dup more expensive.
27628 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27630         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27631         Return true for non-tls symbols.
27633 2017-06-21  James Greenhalgh  <james.greenhalgh@arm.com>
27635         * config/aarch64/aarch64-cores.def (cortex-a55): New.
27636         (cortex-a75): Likewise.
27637         (cortex-a75.cortex-a55): Likewise.
27638         * config/aarch64/aarch64-tune.md: Regenerate.
27639         * doc/invoke.texi (-mtune): Document new values for -mtune.
27641 2017-06-21  Tom de Vries  <tom@codesourcery.com>
27643         * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27644         stack_size feature.
27645         (Effective-Target Keywords, Other attributes): Suggest using
27646         dg-add-options stack_size feature to get stack limit in stack_size
27647         effective target documentation.
27649 2017-06-21  Julian Brown  <julian@codesourcery.com>
27650             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
27652         * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27653         (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27654         * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27655         reservation.
27656         * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27657         attribute type list for neon_multiply.
27658         * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27659         attribute type list for neon_multiply.
27660         * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27661         * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27662         attribute type list for neon_multiply.
27663         * config/arm/types.md (crypto_pmull): Add.
27664         * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27665         attribute type list.
27667 2017-06-20  Andreas Tobler  <andreast@gcc.gnu.org>
27669         * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27670         arm1176jzf-s.
27672 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27674         * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27675         to make sure not to dereference a NULL cost_classes_ptr pointer.
27677 2017-06-20  Carl Love  <cel@us.ibm.com>
27679         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27680         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27681         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27682         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27683         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27684         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27685         VMULOSW): New enum "unspec" values.
27686         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27687         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27688         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27689         altivec_vmulosw): New patterns.
27690         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27691         VMULOSW): Add definitions.
27693 2017-06-20  Julia Koval  <julia.koval@intel.com>
27695         * config/i386/i386.c: Fix rounding expand for new pattern.
27696         * config/i386/subst.md: Fix pattern (parallel -> unspec).
27698 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27700         * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27701         * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27703 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27705         * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27706         feature string.
27708 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27710         * config/aarch64/aarch64-cores.def: Rearrange to sort by
27711         architecture, then by implementer ID.
27712         * config/aarch64/aarch64-tune.md: Regenerate.
27714 2017-06-20  Richard Biener  <rguenther@suse.de>
27716         PR middle-end/81097
27717         * fold-const.c (split_tree): Fold to type before negating.
27719 2017-06-20  David Malcolm  <dmalcolm@redhat.com>
27721         * diagnostic-show-locus.c
27722         (selftest::test_fixit_deletion_affecting_newline): New function.
27723         (selftest::diagnostic_show_locus_c_tests): Call it.
27725 2017-06-20  Andreas Schwab  <schwab@suse.de>
27727         PR target/80970
27728         * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27729         instead of "+d".
27731 2017-06-20  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
27733         * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27734         __ARM_FEATURE_COPROC according to support.
27736 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27738         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27739         Rewritten to avoid overflow for > 32-bit pointers.
27741         PR sanitizer/81125
27742         * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27743         by removing enum keyword.
27744         (ubsan_type_descriptor): Likewise.  Formatting fix.
27746         PR target/81121
27747         * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27748         splitter): Require TARGET_SSE2 in the condition.
27750 2017-06-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
27752         PR target/79799
27753         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27754         for doing vector set of SFmode on ISA 3.0.
27755         * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27756         (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27757         element.
27758         (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27759         SFmode value into a V4SF variable that was extracted from another
27760         V4SF variable without converting the element to double precision
27761         and back to single precision vector format.
27762         (vsx_insert_extract_v4sf_p9_2): Likewise.
27764 2017-06-19  Jakub Jelinek  <jakub@redhat.com>
27766         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27767         in UWHI to avoid undefined overflow.
27769         PR sanitizer/81125
27770         * ubsan.h (enum ubsan_encode_value_phase): New.
27771         (ubsan_encode_value): Change second argument to
27772         enum ubsan_encode_value_phase with default value of
27773         UBSAN_ENCODE_VALUE_GENERIC.
27774         * ubsan.c (ubsan_encode_value): Change second argument to
27775         enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27776         adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27777         create_tmp_var_raw instead of create_tmp_var and use a
27778         TARGET_EXPR.
27779         (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27780         instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27781         ubsan_encode_value callers.
27783         PR sanitizer/81111
27784         * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27785         use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27786         just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27788 2017-06-19  Richard Biener  <rguenther@suse.de>
27790         PR middle-end/81118
27791         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27792         estimates if we changed anything.
27794 2017-06-19  Richard Biener  <rguenther@suse.de>
27796         PR tree-optimization/80887
27797         * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27798         (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27799         simplified lookups, then reset mprts_hook.
27800         (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27801         simplifying.
27802         (try_to_simplify): Likewise.
27804 2017-06-19  Martin Liska  <mliska@suse.cz>
27806         PR sanitizer/80879
27807         * gimplify.c (gimplify_switch_expr):
27808         Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27810 2017-06-19  Martin Liska  <mliska@suse.cz>
27812         * doc/install.texi: Document that PGO runs in 4 stages.
27814 2017-06-19  Martin Liska  <mliska@suse.cz>
27816         PR ipa/80732
27817         * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27818         to dispatcher function name.
27819         * multiple_target.c (replace_function_decl): New function.
27820         (create_dispatcher_calls): Redirect both edges and references.
27822 2017-06-19  Jan Hubicka <hubicka@ucw.cz>
27824         * profile-count.c (profile_count::dump): Dump quality.
27825         (profile_count::differs_from_p): Update for unsigned val.
27826         * profile-count.h (profile_count_quality): New enum.
27827         (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27829 2017-06-19  Richard Biener  <rguenther@suse.de>
27831         * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27832         struct function as arg.
27833         (estimate_numbers_of_iterations): Export overload with loop arg.
27834         (free_numbers_of_iterations_estimates_loop): Use an overload of
27835         free_numbers_of_iterations_estimates instead.
27836         * tree-cfg.c (remove_bb): Adjust.
27837         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27838         * tree-parloops.c (gen_parallel_loop): Likewise.
27839         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27840         Likewise.
27841         (tree_unroll_loops_completely): Likewise.
27842         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27843         Use an overload instead and export.
27844         (estimated_loop_iterations): Adjust.
27845         (max_loop_iterations): Likewise.
27846         (likely_max_loop_iterations): Likewise.
27847         (estimate_numbers_of_iterations): Take struct function as arg
27848         and adjust.
27849         (loop_exits_before_overflow): Adjust.
27850         (free_numbers_of_iterations_estimates_loop): Use an overload.
27851         * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27852         * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27854 2017-06-19  Richard Biener  <rguenther@suse.de>
27856         PR ipa/81112
27857         * ipa-prop.c (find_constructor_constant_at_offset): Handle
27858         RANGE_EXPR conservatively.
27860 2017-06-16  Carl Love  <cel@us.ibm.com>
27862         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27863         definitions for vec_float, vec_float2, vec_floato,
27864         vec_floate built-ins.
27865         * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27866         for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27867         floate.
27868         * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27869         FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27870         UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27871         * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27872         define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27873         * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27874         vec_floato): Add builtin defines.
27875         * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27876         Update the built-in documentation file for the new built-in
27877         functions.
27879 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27881         * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27882         (mthumb): Mark as the negative of -marm.
27884 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27886         * doc/invoke.texi (ARM Options, -mcpu): Document supported
27887         extension options.
27888         (ARM Options, -mtune): Document that this accepts the same
27889         extension options as -mcpu.
27890         (ARM Options, -mfpu): Document addition of -mfpu=auto.
27892 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27894         * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27895         permitted extensions.
27897 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27899         * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27900         (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27901         (armv8-m.main): Add option +nodsp.
27902         * config/arm/arm-cpu-cdata.h: Regenerated.
27904 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27906         * config/arm/t-fuchsia: New file.
27907         * config.gcc (arm*-*-fuchsia*): Use it.
27909 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27911         * config/arm/t-symbian: Rewrite for new option infrastructure.
27913 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27915         * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27916         (MULTILIB_REQUIRED): Likewise.
27918 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27920         * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
27921         (MULTILIB_RESUE): Likewise.
27922         (MULTILIB_MATCHES): Likewise.
27923         (MULTLIB_REQUIRED): Likewise.
27925 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27927         * config/arm/t-rtems: Rewrite for new option framework.
27929 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27931         * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
27932         (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
27933         (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
27934         (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
27935         * config/arm/t-multilib: ... here.
27936         (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
27937         (MULTILIB_MATCHES): Use armv7 libraries for armv7-r.  Also use for
27938         armv7-a and armv8*-a when A-profile libraries have not been built.
27939         * config/arm/t-rmprofile: Rewrite.
27941 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27943         * genmultilib (multilib_reuse): Allow an explicit period to be escaped
27944         with a backslash.  Remove the backslash after substituting unescaped
27945         periods.
27946         * doc/fragments.texi (MULTILIB_REUSE): Document it.
27948 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27950         * config.gcc: (arm*-*-*): When building a-profile libraries, force
27951         the driver to pass through the default setting of -mfloat-abi.
27952         * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
27953         rather than NULL.
27954         * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
27955         (all_feat_combs): New rule.
27956         (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories.  Rework
27957         default libraries.
27958         * config/arm/t-aprofile: Rewrite.
27960 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27962         * config/arm/arm.h (FPUTYPE_AUTO): Define.
27963         * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
27964         fpu is not specified by the user/command-line.
27965         * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
27966         * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
27967         * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
27968         * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
27969         * common/config/arm/arm-common.c (arm_canon_arch_option): Use
27970         FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
27972 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27974         * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
27975         * config/arm/t-arm-elf: Rewritten.
27977 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27979         * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
27980         have some floating-point instructions.
27981         (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
27982         (TARGET_MAYBE_HARD_FLOAT): New macro.
27983         * config/arm/arm-builtins.c (arm_init_builtins): Use
27984         TARGET_MAYBE_HARD_FLOAT.
27985         * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
27987 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27989         * common/config/arm/arm-common.c: Define INCLUDE_LIST.
27990         (configargs.h): Include it.
27991         (arm_print_hint_for_fpu_option): New function.
27992         (arm_parse_fpu_option): New function.
27993         (candidate_extension): New class.
27994         (arm_canon_for_multilib): New function.
27995         * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
27996         (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
27997         (ARCH_CANONICAL_SPECS): New macro.
27998         (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
28000 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28002         * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
28003         are set after handling multilib fragments.  Set target_cpu_default2
28004         from with_cpu.
28006 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28008         * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
28009         cpu name.
28010         (arm*-*-*): Set target_cpu_default2 to a quoted string.
28011         * config/arm/parsecpu.awk (check_cpu): Validate any extension
28012         options.
28013         (check_arch): Likewise.
28014         * config/arm/arm.c (arm_configure_build_target): Handle
28015         TARGET_CPU_DEFAULT being a string constant.  Scan any feature
28016         options in the default.
28018 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28020         * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
28021         when an option is an alias of another.
28022         * config/arm/parsecpu.awk (optalias): New parser token.
28023         (gen_comm_data): Mark non-alias options as such.  Emit entries
28024         for extension aliases.
28025         * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
28026         (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
28027         (armv6kz, armv6zk, armv6t2): Likewise.
28028         (armv7): Make vfpv3-d16 an alias.
28029         (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases.  Sort in
28030         canonical order.
28031         (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
28032         Sort in canonical order.
28033         (armv8-a): Sort in canonical order.
28034         (armv8.1-a, armv8.2-a):  Likewise.
28035         (generic-armv7-a): Make neon and neon-vfpv3 aliases.  Sort in
28036         canonical order.
28037         (cortex-a9): Sort in canonical order.
28038         * config/arm/arm.c (selftests.h): Include it.
28039         (arm_test_cpu_arch_data): New function.
28040         (arm_run_self_tests): New function.
28041         (TARGET_RUN_TARGET_SELFTESTS): Redefine.
28042         (targetm): Move declaration to the end of the file.
28043         * arm-cpu-cdata.h: Regenerated.
28045 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28047         * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
28048         call to target_mode_check describing the type of option passed.
28049         * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
28050         (arm_target_thumb_only): Use arm_parse_arch_option_name or
28051         arm_parse_cpu_option_name to match parameters against list of
28052         available targets.
28053         * config/arm/parsecpu.awk (gen_comm_data): Don't generate
28054         arm_arch_core_flags data structure.
28055         * config/arm/arm-cpu_cdata.h: Regenerated.
28057 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28059         * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
28060         config/arm/arm.c.
28061         (arm_print_hint_for_cpu_option): Likewise.
28062         (arm_print_hint_for_arch_option): Likewise.
28063         (arm_parse_cpu_option_name): Likewise.
28064         (arm_parse_arch_option_name): Likewise.
28065         * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
28066         of entries in the all_fpus list.
28067         * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
28068         (arm_parse_cpu_option_name): Declare.
28069         (arm_parse_arch_option_name): Declare.
28070         (arm_parse_option_features): Declare.
28071         (arm_intialize_isa): Declare.
28072         * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
28073         data tables to ...
28074         (gen_comm_data): ... here.  Make definitions non-static.
28075         * config/arm/arm-cpu-data.h: Regenerated.
28076         * config/arm/arm-cpu-cdata.h: Regenerated.
28078 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28080         * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
28081         (cpu_arch_extension): New structure.
28082         (cpu_arch_option, arch_option, cpu_option): New structures.
28083         * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
28084         architecture types.
28085         (gen_data): Generate new format data tables.
28086         * config/arm/arm.c (cpu_tune): New structure.
28087         (cpu_option, processors): Delete.
28088         (arm_print_hint_for_core_or_arch): Delete.  Replace with ...
28089         (arm_print_hint_for_cpu_option): ... this and ...
28090         (arm_print_hint_for_arch_option): ... this.
28091         (arm_parse_arch_cpu_name): Delete.  Replace with ...
28092         (arm_parse_cpu_option_name): ... this and ...
28093         (arm_parse_arch_option_name): ... this.
28094         (arm_unrecognized_feature): Change type of target parameter to
28095         cpu_arch_option.
28096         (arm_parse_arch_cpu_features): Delete.  Replace with ...
28097         (arm_parse_option_features): ... this.
28098         (arm_configure_build_target): Rework to use new configuration data
28099         tables.
28100         (arm_print_tune_info): Rework for new configuration data tables.
28101         * config/arm/arm-cpu-data.h: Regenerated.
28102         * config/arm/arm-cpu.h: Regenerated.
28104 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28106         * Makefile.in (OBJS): Move sbitmap.o from here ...
28107         (OBJS-libcommon): ... to here.
28109 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28111         * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
28112         (ISA_ALL_CRYPTO): New macro.
28113         (ISA_ALL_SIMD): New macro
28114         (ISA_ALL_FP): New macro.
28115         * config/arm/arm.c (fpu_bitlist): Update initializer.
28116         * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
28117         simd or fp.
28118         (arm9e): Add fpu.  Add option for nofp
28119         (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
28120         (arm926ej-s, arm1026ej-s): Likewise.
28121         (generic-armv7-a): Add fpu.  Add options for simd, vfpv3, vfpv3-d16,
28122         vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
28123         neon-fp16, neon-vfpv4, nofp and nosimd.
28124         (cortex-a5, cortex-a7): Add fpu.  Add options for nosimd and nofp.
28125         (cortex-a8): Add fpu.  Add option for nofp.
28126         (cortex-a9): Add fpu.  Add options for nosimd and nofp.
28127         (cortex-a12, cortex-a15, cortex-a17): Add fpu.  Add option for nofp.
28128         (cortex-r4f): Add fpu.
28129         (cortex-r5): Add fpu.  Add options for nofp.dp and nofp.
28130         (cortex-r7): Use idiv option from architecture.  Add fpu.  Add option
28131         for nofp.
28132         (cortex-r8): Likewise.
28133         (cortex-m4): Add fpu.  Add option for nofp.
28134         (cortex-a15.cortex-a7): Add fpu.  Add option for nofp.
28135         (cortex-a17.cortex-a7): Likewise.
28136         (cortex-a32): Add fpu.  Add options for crypto and nofp.
28137         (cortex-a35, cortex-a53): Likewise.
28138         (cortex-a57): Add fpu.  Add option for crypto.
28139         (cortex-a72, cortex-a73): Likewise.
28140         (exynos-m1): Likewise.
28141         (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
28142         (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
28143         (cortex-m33): Add fpu.  Add option for nofp.
28144         * config/arm/arm-cpu-cdata.h: Regenerated
28145         * config/arm/arm-cpu-data.h: Regenerated.
28147 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28149         * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
28150         (armv5te, armv5tej): Likewise.
28151         (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
28152         (armv7): Add options fp and vfpv3-d16.
28153         (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
28154         vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
28155         nofp and nosimd.
28156         (armv7ve): Likewise.
28157         (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
28158         (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
28159         (armv8-a): Add nocrypto option.
28160         (armv8.1-a, armv8.2-a): Likewise.
28161         (armv8-m.main): add options fp, fp.dp and nofp.
28163 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28165         * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
28166         nofp.
28167         (armv8-a+crc): Delete.
28168         (armv8.1-a): Add options simd, crypto and nofp.
28169         (armv8.2-a): Add options fp16, simd, crypto and nofp.
28170         (armv8.2-a+fp16): Delete.
28171         (armv8-m.main): Add option dsp.
28172         (armv8-m.main+dsp): Delete.
28173         (cortex-a8): Add fpu.  Add nofp option.
28174         (cortex-a9): Add fpu.  Add nofp and nosimd options.
28175         * config/arm/parsecpu.awk (gen_data): Generate option tables and
28176         link to main cpu and architecture data structures.
28177         (gen_comm_data): Only put isa attributes from the main architecture
28178         in common tables.
28179         (option): New statement for architecture and CPU entries.
28180         * arm.c (struct cpu_option): New structure.
28181         (struct processors): Add entry for options.
28182         (arm_unrecognized_feature): New function.
28183         (arm_parse_arch_cpu_name): Ignore any characters after the first
28184         '+' character.
28185         (arm_parse_arch_cpu_feature): New function.
28186         (arm_configure_build_target): Separate out any CPU and architecture
28187         features and parse separately.  Don't error out if -mfpu=auto is
28188         used with only an architecture string.
28189         (arm_print_asm_arch_directives): New function.
28190         (arm_file_start): Call it.
28191         * config/arm/arm-cpu-cdata.h: Regenerated.
28192         * config/arm/arm-cpu-data.h: Likewise.
28193         * config/arm/arm-tables.opt: Likewise.
28195 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28197         * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
28198         assembler when it is not -mfpu=auto.
28200 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28202         * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
28203         (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
28204         (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
28205         (ASM_CPU_SPEC): Rewrite.
28206         (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
28207         (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef.  Use
28208         MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS.  Remove
28209         reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
28210         * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
28211         copied string is NUL-terminated.  Also strip any characters prefixed
28212         by '+'.
28213         (arm_rewrite_selected_arch): New function.
28214         (arm_rewrite_march): New function.
28216 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28218         * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
28219         (x_arm_cpu_string, x_arm_tune_string): Likewise.
28220         (march, mcpu, mtune): Convert to string-based options.
28221         * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
28222         (arm_parse_arch_cpu_name): New function.
28223         (arm_configure_build_target): Use arm_parse_arch_cpu_name to
28224         identify selected architecture or CPU.
28225         (arm_option_save): New function.
28226         (TARGET_OPTION_SAVE): Redefine.
28227         (arm_option_restore): Restore string options.
28228         (arm_option_print): Print string options.
28230 2017-06-16  Martin Sebor  <msebor@redhat.com>
28232         PR tree-optimization/80933
28233         PR tree-optimization/80934
28234         * builtins.c (fold_builtin_3): Do not handle bcmp here.
28235         * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
28236         (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
28237         (gimple_fold_builtin): Call them.
28239 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28241         * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
28242         as unlikely; update profile.
28244 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28246         * predict.c (force_edge_cold): Handle declaring edges impossible
28247         more aggresively.
28249 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28251         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28252         profile.
28253         (try_unroll_loop_completely): Fix reporting.
28255 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28257         * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28259 2017-06-16  James Greenhalgh  <james.greenhalgh@arm.com>
28261         PR target/71778
28262         * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28263         if given a non-constant argument for an intrinsic which requires a
28264         constant.
28266 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28268         * profile.c (compare_freqs): New function.
28269         (branch_prob): Sort edge list.
28270         (find_spanning_tree): Assume that the list is priority sorted.
28272 2017-06-16  Richard Biener  <rguenther@suse.de>
28274         PR tree-optimization/81090
28275         * passes.def (pass_record_bounds): Remove.
28276         * tree-pass.h (make_pass_record_bounds): Likewise.
28277         * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28278         make_pass_record_bounds): Likewise.
28279         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28280         not free niter estimates at the beginning but at the end.
28281         * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28283 2017-06-16  Richard Biener  <rguenther@suse.de>
28285         * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28286         initializer to workaround ICE in host GCC 4.8.
28288 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28290         * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28291         counts.
28292         (clone_inlined_nodes): Update.
28294 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28296         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28297         prefetch settings, and enable prefetching by default at -O3.
28299 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28301         * config/aarch64/aarch64.c (aarch64_override_options_internal):
28302         Set flag_prefetch_loop_arrays according to tuning data.
28304 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28306         * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28307         New tune structure.
28308         (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28309         [Unrelated to main purpose of the patch] Place the pointer field last
28310         to enable type checking errors when tune structure are wrongly merged.
28311         * config/aarch64/aarch64.c (generic_prefetch_tune,)
28312         (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28313         (thunderx2t99_prefetch_tune): New tune constants.
28314         (tune_params *_tunings): Update all tunings (no functional change).
28315         (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28316         PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28317         from tunings structures.
28319 2017-06-16  Jakub Jelinek  <jakub@redhat.com>
28321         PR sanitizer/81094
28322         * ubsan.c (instrument_null): Add T argument, use it instead
28323         of computing it based on IS_LHS.
28324         (instrument_object_size): Likewise.
28325         (pass_ubsan::execute): Adjust instrument_null and
28326         instrument_object_size callers to pass gimple_get_lhs or
28327         gimple_assign_rhs1 result to it.  Use instrument_null instead of
28328         calling get_base_address and instrument_mem_ref.  Handle
28329         aggregate call arguments for object-size sanitization.
28331 2017-06-16  Yury Gribov  <tetra2005@gmail.com>
28333         PR tree-optimization/81089
28334         * tree-vrp.c (is_masked_range_test): Validate operands of
28335         subexpression.
28337 2017-06-15  Martin Sebor  <msebor@redhat.com>
28339         PR c++/80560
28340         * dumpfile.c (dump_register): Avoid calling memset to initialize
28341         a class with a default ctor.
28342         * gcc.c (struct compiler): Remove const qualification.
28343         * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28344         * hash-table.h: Ditto.
28345         * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28346           assignment.
28347         * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28348         * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28349         default ctor.
28350         * params.h (struct param_info): Make struct members non-const.
28351         * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28352         with default initialization.
28353         * vec.h (vec_copy_construct, vec_default_construct): New helper
28354         functions.
28355         (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28356         with vec_copy_construct.
28357         (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28358         (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28359         * doc/invoke.texi (-Wclass-memaccess): Document.
28361 2017-06-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28363         * emit-rtl.h (is_leaf): Update comment about local
28364         register allocator.
28366 2017-06-15  Jozef Lawrynowicz   <jozef.l@somniumtech.com>
28368         PR target/78818
28369         * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28370         for a variable to have a section before checking if the section has a
28371         name.
28372         Set section to.persistent if persistent attribute is set.
28373         Warn if .persistent attribute is used on an automatic variable.
28375 2017-06-15  Eric Botcazou  <ebotcazou@adacore.com>
28377         PR rtl-optimization/80474
28378         * reorg.c (update_block): Do not ignore instructions in a delay slot.
28380 2017-06-15  Segher Boessenkool  <segher@kernel.crashing.org>
28382         * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28383         of REGNO.
28385 2017-06-14  Maciej W. Rozycki  <macro@imgtec.com>
28387         * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28388         (casesi): Emit bounds checking as RTL.
28389         (casesi_internal_mips16_<mode>): Remove bounds checking.
28391 2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
28393         * config/xtensa/xtensa.c (xtensa_option_override): Append
28394         MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28395         (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28396          xtensa_doloop_hooks): Define unconditionally.
28397         (xtensa_reorg_loops): Only call reorg_loops in the presence of
28398         TARGET_LOOPS.
28399         * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28400         (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28401         for it in xtensa_option_override.
28402         (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28403          HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28405 2017-06-14  Boris Kolpackov  <boris@codesynthesis.com>
28407         * doc/cppopts.texi: Document '-' special value to -MF.
28409 2017-06-14  Wilco Dijkstra  <wdijkstr@arm.com>
28411         * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28412         (cortex_a53_fconst): Likewise.
28413         (cortex_a53_fpmul): Likewise.
28414         (cortex_a53_f_load_64): Likewise.
28415         (cortex_a53_f_load_many): Likewise.
28416         (cortex_a53_advsimd_alu): Likewise.
28417         (cortex_a53_advsimd_alu_q): Likewise.
28418         (cortex_a53_advsimd_mul): Likewise.
28419         (cortex_a53_advsimd_mul_q): Likewise.
28420         (fpmac bypass): Add new bypass for fpmac-fpmac case.
28421         Add missing fmul, r2f_cvt and fconst cases.
28423 2017-06-14  Richard Biener  <rguenther@suse.de>
28425         PR middle-end/81088
28426         * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28427         literal constants.
28428         (fold_binary_loc): When associating do not treat pre-existing
28429         TREE_OVERFLOW on literal constants as a reason to allow
28430         TREE_OVERFLOW on associated literal constants.
28432 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28434         * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28435         (MASK_FEATURES): New macro.
28436         * config/sparc/sparc.c (sparc_option_override): Remove the special
28437         handling of -mfpu and generalize it to all MASK_FEATURES switches.
28439 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28441         * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28442         a division of 0 if non-call exceptions are enabled.
28444 2017-06-14  Andrew Pinski  <apinski@cavium.com>
28445             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
28447         PR target/71663
28448         * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28449         Improve vector initialization code gen for only variable case.
28451 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28453         * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28455 2017-06-14  Richard Biener  <rguenther@suse.de>
28457         PR tree-optimization/81083
28458         * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28459         as values.
28461 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28463         * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28464         (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28465         * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28466         * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete.  Adjust former use.
28467         * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28468         * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28470 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28472         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28473         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28475 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28477         * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28479 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28481         * config/rs6000/t-rtems: Don't handle SPE.
28483 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28485         * config/rs6000/t-linux: Don't handle SPE.
28487 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28489         * config/rs6000/eabispe.h: Delete file.
28491 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28493         * config/rs6000/t-spe: Delete file.
28495 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28497         * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28498         (rs6000_legitimate_offset_address_p): Return false for anything in
28499         V2SImode or V2SFmode.
28501 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28503         * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28504         except V2SF and V2SI.  Rearrange the vector modes, and add comments.
28505         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28506         and V4HImode.
28507         (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28508         (rs6000_legitimate_offset_address_p): Ditto.
28509         (rs6000_emit_move): Ditto.
28510         (rs6000_init_builtins): Remove V4HI_type_node.
28512 2017-06-13  Martin Liska  <mliska@suse.cz>
28514         PR sanitize/78204
28515         * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28516         (gate_asan): Likewise.
28517         * asan.h (asan_no_sanitize_address_p): Remove the function.
28518         (sanitize_flags_p): New function.
28519         * builtins.def: Fix coding style.
28520         * common.opt: Use renamed enum value.
28521         * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28522         * doc/extend.texi: Document no_sanitize attribute.
28523         * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28524         to SANITIZE_UNDEFINED_NONDEFAULT.
28525         * gcc.c (sanitize_spec_function): Use the renamed enum value.
28526         * gimple-fold.c (optimize_atomic_compare_exchange_p):
28527         Use sanitize_flags_p.
28528         * gimplify.c (gimplify_function_tree): Likewise.
28529         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28530         * opts.c (parse_no_sanitize_attribute): New function.
28531         (common_handle_option): Use renamed enum value.
28532         * opts.h (parse_no_sanitize_attribute): Declare.
28533         * tree.c (sanitize_flags_p): New function.
28534         * tree.h: Declared here.
28535         * tsan.c: Use sanitize_flags_p.
28536         * ubsan.c (ubsan_expand_null_ifn): Likewise.
28537         (instrument_mem_ref): Likewise.
28538         (instrument_bool_enum_load): Likewise.
28539         (do_ubsan_in_current_function): Remove the function.
28540         (pass_ubsan::execute): Use sanitize_flags_p.
28541         * ubsan.h: Remove do_ubsan_in_current_function
28542         * tree-cfg.c (print_no_sanitize_attr_value): New function.
28543         (dump_function_to_file): Use it here.
28545 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28547         PR tree-optimization/80803
28548         PR tree-optimization/81063
28549         * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28550         (propagate_subaccesses_across_link): Enqueue subtree whenever
28551         necessary instead of relying on the caller.
28553 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28555         * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28556         that have a first_link.
28557         (sort_and_splice_var_accesses): Do not check first_link before
28558         enquing.
28559         (subtree_mark_written_and_enqueue): Likewise.
28560         (propagate_all_subaccesses): Likewise and do not stop at first
28561         parent with a first_link.
28563 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28565         * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28566         instead of f.
28568 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28570         * match.pd: New pattern.
28572 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28574         * tree-vrp.c (is_masked_range_test): New function.
28575         (register_edge_assert_for): Determine ranges for
28576         some bit tests.
28578 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28580         PR tree-optimization/67328
28581         * fold-const.c (maskable_range_p): New function.
28582         (build_range_check): Generate bittests if possible.
28584 2017-06-13  Martin Liska  <mliska@suse.cz>
28586         * gimple-pretty-print.c (dump_probability): Add new argument.
28587         (dump_edge_probability): Dump both probability and count.
28588         (dump_gimple_label): Likewise.
28589         (dump_gimple_bb_header): Likewise.
28591 2017-06-13  Georg-Johann Lay  <avr@gjlay.de>
28593         PR target/81072
28594         * config/avr/avr-devices.c: Fix indentation.
28595         * config/avr/gen-avr-mmcu-specs.c: Dito.
28597 2017-06-13  Richard Biener  <rguenther@suse.de>
28599         * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28600         instead get vector type from stmt_info.
28601         (vectorizable_reduction): Adjust.  Remove dead code.
28603 2017-06-13  Richard Biener  <rguenther@suse.de>
28605         PR middle-end/81065
28606         * fold-const.c (extract_muldiv_1): Remove bogus distribution
28607         case of C * (x * C2 + C3).
28608         (fold_addr_of_array_ref_difference): Properly fold index difference.
28610 2017-06-12  David S. Miller  <davem@davemloft.net>
28612         PR target/80968
28613         * config/sparc/sparc.md (return expander): Emit frame blockage if
28614         function uses alloca.
28616 2017-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
28618         * combine.c (make_field_assignment): Check len rather than the mode
28619         precision when calling force_to_mode.
28621 2017-06-12  Georg-Johann Lay  <avr@gjlay.de>
28623         Support multilibs and devices that see flash in RAM address range.
28625         PR target/81072
28626         * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28627         (avr_mcu_t) <flash_pm_offset>: New field.
28628         (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28629         * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28630         (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28631         (AVR_TINY_PM_OFFSET): Remove macro.
28632         * config/avr/avr.opt (-mshort-calls): New option.
28633         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28634         [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28635         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28636         <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28637         <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28638         instead of avr_arch->have_jmp_call.
28639         <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28640         [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28641         avr_arch->flash_pm_offset to define.
28642         * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28643         new field flash_pm_offset.  Add entry for avrxmega3.
28644         (avr_texinfo): Add entry for avrxmega3.
28645         * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28646         attiny212, attiny214,
28647         attiny412, attiny414, attiny416, attiny417,
28648         attiny814, attiny816, attiny817,
28649         attiny1614, attiny1616, attiny1617,
28650         attiny3214, attiny3216, attiny3217.
28651         * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28652         avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28653         (avr_print_operand_address) [AVR_TINY]: Same.
28654         (avr_asm_init_sections) <readonly_data_section>: Only patch
28655         callback if avr_arch->flash_pm_offset = 0.
28656         (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28657         for rodata if avr_arch->flash_pm_offset != 0.
28658         (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28659         * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28660         (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28661         (m_options): Append opt_rcall.
28662         (m_dirnames): Append dir_rcall.
28663         * config/avr/t-multilib: Regenerate.
28665         * configure.ac [target=avr]: Check whether avrxmega3 default
28666         linker description file works as needed.
28667         * configure: Regenerate.
28668         * doc/avr-mmcu.texi: Regenerate.
28669         * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28670         <__AVR_ARCH__>: Document avrxmega3 and 103.
28671         <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28672         <__AVR_SHORT_CALLS__>: Document it.
28673         <__AVR_PM_BASE_ADDRESS__>: Document it.
28674         * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28675         (AVR Variable Attributes) <progmem>: Document this is
28676         not needed for avrxmega3.
28677         (AVR Named Address Spaces) <__flash>: Dito.
28679 2017-06-12  Jan Hubicka  <hubicka@ucw.cz>
28681         * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28683 2017-06-12  Doug Rupp  <rupp@adacore.com>
28685         * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28686         Append vxworks-stdint.h to the tm_file list.
28687         * config/vxworks-stdint.h: New file.
28689 2017-06-12  Martin Liska  <mliska@suse.cz>
28691         PR tree-optimization/81041
28692         * tree-profile.c (gimple_gen_ic_func_profiler):
28693         Create an extra BB in profile-generate
28694         (gimple_gen_time_profiler): Likewise.
28696 2017-06-12  Jakub Jelinek  <jakub@redhat.com>
28698         PR tree-optimization/81003
28699         * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28700         (update_range_test): Use it instead of force_gimple_operand_gsi.
28702 2017-06-12  Richard Biener  <rguenther@suse.de>
28704         PR tree-optimization/81053
28705         * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28706         with backedge value not defined in loop.  Simplify def stmt
28707         compute.
28709 2017-06-11  Tom de Vries  <tom@codesourcery.com>
28711         PR target/79939
28712         * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28713         Return true.
28714         (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28715         nvptx_cannot_force_const_mem.
28717 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28719         * opts.c (finish_options): Move test for flag_split_stack after
28720         it has been initialized.
28722 2017-06-11  Jason Merrill  <jason@redhat.com>
28724         * tree.h (id_equal): New.
28725         * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28726         omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28727         instead of strcmp of IDENTIFIER_POINTER.
28729 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28731         * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28732         (mark_all_inlined_calls_cdtor): Fix formating.
28733         (inline_transform): Rescale profile before inlining.
28735 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28737         * cgraph.h (cgraph_edge::clone): Update prototype.
28738         * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28739         (cgraph_node::create_clone): Update.
28740         (cgraph_node::create_version_clone): Update.
28741         * tree-inline.c (copy_bb): Update.
28742         (expand_call_inline): Update.
28744 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28746         * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28747         factored out from ...
28748         (rs6000_emit_prologue): ... here.
28750 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28752         * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28753         factored out from ...
28754         (rs6000_emit_prologue): ... here.
28756 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28758         * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28759         edge counts.
28760         (handle_missing_profiles): Fix computation of tp_first_run.
28761         (counts_to_freqs): Do not touch freqs when count is 0.
28763 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28765         * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28766         profile.
28768 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28770         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28771         attributes): Document signal effective target.
28773 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28775         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28776         Document effective target stack_size.
28778 2017-06-09  David Malcolm  <dmalcolm@redhat.com>
28780         * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28781         to the edit_context if they can be auto-applied.
28783 2017-06-9  Ian Lance Taylor  <iant@golang.org>
28785         * opts.c (finish_options): If -fsplit-stack, disable implicit
28786         -forder-blocks-and-partition.
28787         * doc/invoke.texi (Optimize Options): Document that when using
28788         -fsplit-stack -forder-blocks-and-partition is not implicitly
28789         enabled.
28791 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28793         * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28794         ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28795         ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28796         * builtins.def (abort, trap, unreachable): Declare cold.
28797         * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28798         * tree-core.h (ECF_COLD): New.
28799         * tree.c (set_call_expr_flags): Handle ECF_COLD.
28800         (build_common_builtin_nodes): Mark unreachable and abort as cold.
28802 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28804         * predict.c (unlikely_executed_stmt_p): Cleanup.
28806 2017-06-09  Richard Biener  <rguenther@suse.de>
28808         * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28809         model if the ref is always written to.
28811 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28813         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28815 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28817         * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28818         than udiv.
28820 2017-06-09  Tom de Vries  <tom@codesourcery.com>
28822         PR target/80855
28823         * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28824         "target cannot support label values" when encountering LABEL_REF.
28826 2017-06-09  Martin Liska  <mliska@suse.cz>
28828         * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28829         (gimple_gen_ic_func_profiler): Emit direct comparison
28830         of __gcov_indirect_call_callee with NULL.
28831         (gimple_gen_time_profiler): Change probability from
28832         PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28834 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28836         * profile.c (edge_gcov_counts): Turn to pointer.
28837         (compute_branch_probabilities, compute_branch_probabilities): Update.
28838         (branch_prob): Do not clear edge_gcov_count.
28839         * profile.h (edge_gcov_counts): Turn to pointer.
28840         (edge_gcov_count): Update.
28842 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28844         * gimple.h (gimple_check_failed): Mark cold.
28846 2017-06-09  Richard Biener  <rguenther@suse.de>
28848         PR tree-optimization/66623
28849         * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28850         refactor check_reduction into two parts, properly computing
28851         whether we have to check reduction validity for outer loop
28852         vectorization.
28854 2017-06-09  Richard Biener  <rguenther@suse.de>
28856         PR tree-optimization/79483
28857         * graphite-scop-detection.c (order): New global.
28858         (get_order): Compute bb to order mapping that satisfies code
28859         generation constraints.
28860         (cmp_pbbs): New helper.
28861         (build_scops): Start domwalk at entry block, sort generated
28862         pbbs.
28864 2017-06-09  Richard Biener  <rguenther@suse.de>
28866         PR middle-end/81007
28867         * ipa-polymorphic-call.c
28868         (ipa_polymorphic_call_context::restrict_to_inner_class):
28869         Skip FIELD_DECLs with error_mark_node type.
28870         * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28871         last again.
28873 2017-06-09  Martin Liska  <mliska@suse.cz>
28875         * predict.c (struct branch_predictor): New struct.
28876         (test_prediction_value_range): New test.
28877         (predict_c_tests): New function.
28878         * selftest-run-tests.c (selftest::run_tests): Run the function.
28879         * selftest.h: Declare new tests.
28881 2017-06-09  Segher Boessenkool  <segher@kernel.crashing.org>
28883         PR target/80966
28884         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28885         gen_add3_insn did not fail.
28886         * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28887         r0, construct that number in a temporary reg and add that reg to r0.
28888         If asked to put the result in r0 as well, fail.
28890 2017-06-08  Will Schmidt  <will_schmidt@vnet.ibm.com>
28892         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28893         for early expansion of vec_eqv.
28895 2017-06-08  Jakub Jelinek  <jakub@redhat.com>
28897         PR middle-end/81005
28898         * ubsan.c (instrument_null): Avoid pointless code temporary.
28899         (pass_ubsan::execute): Instrument aggregate arguments of calls.
28901 2017-06-08  Uros Bizjak  <ubizjak@gmail.com>
28903         PR target/81015
28904         Revert:
28905         2016-12-14  Uros Bizjak  <ubizjak@gmail.com>
28907         PR target/59874
28908         * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28909         (*clzhi2): Ditto.
28911 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28913         * predict.c (unlikely_executed_edge_p): Move ahead.
28914         (probably_never_executed_edge_p): Use it.
28916 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28918         PR middle-end/79988
28919         * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
28920         gimple_call_builtin_p call.
28922 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28924         * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
28925         * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
28926         rtl_check_failed_type2, rtl_check_failed_code1,
28927         rtl_check_failed_code2, rtl_check_failed_code_mode,
28928         rtl_check_failed_block_symbol, cwi_check_failed_bounds,
28929         rtvec_check_failed_bounds, rtl_check_failed_flag,
28930         _fatal_insn_not_found, _fatal_insn): Likewise.
28931         * tree.h (tree_contains_struct_check_failed,
28932         tree_check_failed, tree_not_check_failed,
28933         tree_class_check_failed, tree_range_check_failed,
28934         tree_not_class_check_failed, tree_int_cst_elt_check_failed,
28935         tree_vec_elt_check_failed, phi_node_elt_check_failed,
28936         tree_operand_check_failed, omp_clause_check_failed,
28937         omp_clause_operand_check_failed, omp_clause_range_check_failed):
28938         Likewise.
28940 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28942         * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
28943         flag_branch_probabilities.
28944         * ipa-inline.c (edge_badness): Likewise.
28945         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
28946         * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
28947         * predict.c (maybe_hot_frequency_p): Likewise.
28948         (probably_never_executed): Likewise.
28949         * sched-ebb.c (schedule_ebbs): Likewise.
28950         * sched-rgn.c (find_single_block_region): Likewise.
28951         * tracer.c (tail_duplicate): Likewise.
28953 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28955         * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
28956         longer requires x_flag_profile_use.
28958 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28960         * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
28961         instead of flag_reorder_blocks_and_partition.
28962         * dbxout.c (dbxout_function_end): Likewise.
28963         * dwarf2out.c (gen_subprogram_die): Likewise.
28964         * haifa-sched.c (sched_create_recovery_edges): Likewise.
28965         * hw-doloop.c (reorg_loops): Likewise.
28966         * varasm.c (assemble_start_function,
28967         assemble_end_function): Likewise.
28968         (decide_function_section): Do not check for
28969         flag_reorder_blocks_and_partition.
28971 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28973         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
28974         New function.
28975         (chkp_get_hard_register_fake_addr_expr): Ditto.
28976         (chkp_build_addr_expr): Add check for hard reg case.
28977         (chkp_parse_array_and_component_ref): Ditto.
28978         (chkp_find_bounds_1): Ditto.
28979         (chkp_process_stmt): Don't generate bounds store for
28980         hard reg case.
28982 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28984         * predict.c (maybe_hot_bb_p): Do not check profile status.
28985         (maybe_hot_edge_p): Likewise.
28986         (probably_never_executed): Check for zero counts even if profile
28987         is not read.
28988         (unlikely_executed_edge_p): New function.
28989         (unlikely_executed_stmt_p): New function.
28990         (unlikely_executed_bb_p): New function.
28991         (set_even_probabilities): Use unlikely predicates.
28992         (combine_predictions_for_bb): Likewise.
28993         (predict_paths_for_bb): Likewise.
28994         (predict_paths_leading_to_edge): Likewise.
28995         (determine_unlikely_bbs): New function.
28996         (estimate_bb_frequencies): Use it.
28997         (compute_function_frequency): Use zero counts even if profile is
28998         not read.
28999         * profile-count.h: Fix typo.
29001 2017-08-08  Julia Koval  <julia.koval@intel.com>
29003         * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
29004         _mm512_mask_cvtsepi16_storeu_epi8,
29005         _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
29006         * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
29007         _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
29008         _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
29009         _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
29010         * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
29011         (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
29012         VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
29013         * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
29014         __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
29015         __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
29016         __builtin_ia32_pmovuswb256mem_mask,
29017         __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
29018         __builtin_ia32_pmovwb512mem_mask): New builtins.
29020 2017-08-08  Julia Koval  <julia.koval@intel.com>
29022         PR target/73350,80862
29023         * config/i386/subst.md (round): Fix round pattern.
29024         * config/i386/i386.c (ix86_erase_embedded_rounding):
29025         Fix erasing rounding for the fixed pattern.
29027 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29029         * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
29031 2017-06-08  Martin Liska  <mliska@suse.cz>
29033         PR gcov-profile/80911
29034         * gcov.c (block_info::block_info): New constructor.
29036 2017-06-07  Carl Love  <cel@us.ibm.com>
29038         * config/rs6000/rs6000-c: The return type of the following
29039         built-in functions was implemented as int not long long.  Fix sign
29040         of return value for the unsigned version of vec_mulo and vec_mule.
29041         vector unsigned long long vec_bperm (vector unsigned long long,
29042                                              vector unsigned char)
29043         vector signed long long vec_mule (vector signed int,
29044                                           vector signed int)
29045         vector unsigned long long vec_mule (vector unsigned int,
29046                                             vector unsigned int)
29047         vector signed long long vec_mulo (vector signed int,
29048                                           vector signed int)
29049         vector unsigned long long vec_mulo (vector unsigned int,
29050                                             vector unsigned int)
29051         * doc/extend.texi: Fix the documentation for the built-in
29052         functions.
29054 2017-06-07  Carl Love  <cel@us.ibm.com>
29056         PR target/80982
29057         * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
29058         for BE.
29060 2017-06-07  Carl Love  <cel@us.ibm.com>
29062         * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
29063         support, Generate       doublehv for signed int/float for BE case only.
29065 2017-06-07  Alexander Monakov  <amonakov@ispras.ru>
29067         * doc/invoke.texi (mcx16): Rewrite.
29069 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29071         * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
29072         * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
29073         *mov<mode>_softfloat, and an anonymous splitter): Use
29074         nonimmediate_operand instead of rs6000_nonimmediate_operand.
29076 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29078         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
29079         SPEFSCR registers.
29080         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29081         (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
29082         (rs6000_debug_reg_global): Adjust.
29083         (rs6000_init_hard_regno_mode_ok): Adjust.
29084         (rs6000_dbx_register_number): Adjust.
29085         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
29086         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29087         Remove SPE_ACC and SPEFSCR.
29088         (REG_ALLOC_ORDER): Ditto.
29089         (FRAME_POINTER_REGNUM): Change to 111.
29090         (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
29091         (REG_CLASS_NAMES): Ditto.
29092         (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
29093         (REGISTER_NAMES): Ditto.
29094         (ADDITIONAL_REG_NAMES): Ditto.
29095         (rs6000_reg_names): Ditto.
29096         * config/rs6000/rs6000.md: Renumber some register number
29097         define_constants.
29099 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29101         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
29102         registers.
29103         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29104         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
29105         to 117.
29106         (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
29107         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29108         Delete the SPE high registers.
29109         (REG_ALLOC_ORDER): Ditto.
29110         (enum reg_class): Remove SPE_HIGH_REGS.
29111         (REG_CLASS_NAMES): Ditto.
29112         (REG_CLASS_CONTENTS): Delete the SPE high registers.
29113         (REGISTER_NAMES): Ditto.
29114         (rs6000_reg_names): Ditto.
29115         * doc/tm.texi.in: Remove SPE as example.
29116         * doc/tm.texi: Regenerate.
29118 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29120         * config/rs6000/8540.md (ppc8540_brinc): Delete.
29121         * config/rs6000/e500mc.md (e500mc_brinc): Delete.
29122         * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
29123         * config/rs6000/rs6000.md (type): Remove "brinc".
29125 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29127         * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
29128         (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
29129         * config/rs6000/linuxspe.h: Delete file.
29130         * config/rs6000/rs6000.md: Don't include spe.md.
29131         * config/rs6000/spe.h: Delete file.
29132         * config/rs6000/spe.md: Delete file.
29133         * config/rs6000/t-rs6000: Remove spe.md.
29135 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29137         * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
29138         (reg_or_none500mem_operand): Delete.
29139         * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
29140         instead of reg_or_none500mem_operand.
29142 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29144         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29145         handling of SPE flags.
29146         * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
29148 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29150         * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
29151         SPE ABI handling.
29152         * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
29153         (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
29154         paired_divv2sf3): Similar.
29155         * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
29156         SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
29157         * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
29158         RS6000_BUILTIN_S.
29159         Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
29160         Rename the paired_* instruction patterns.
29161         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
29162         define __SPE__.
29163         * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
29164         * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
29165         (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
29166         spe_padding_size, and spe_64bit_regs_used.  Replace TARGET_SPE and
29167         TARGET_SPE_ABI with 0, simplify.  Replace SPE_VECTOR_MODE with
29168         PAIRED_VECTOR_MODE.
29169         (struct machine_function): Delete field spe_insn_chain_scanned_p.
29170         (spe_func_has_64bit_regs_p): Delete.
29171         (spe_expand_predicate_builtin): Delete.
29172         (spe_expand_evsel_builtin): Delete.
29173         (TARGET_DWARF_REGISTER_SPAN): Do not define.
29174         (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
29175         (invalid_e500_subreg): Delete.
29176         (rs6000_legitimize_address): Always force_reg op2 as well, for
29177         paired single memory accesses.
29178         (rs6000_member_type_forces_blk): Delete.
29179         (rs6000_spe_function_arg): Delete.
29180         (rs6000_expand_unop_builtin): Delete SPE handling.
29181         (rs6000_expand_binop_builtin): Ditto.
29182         (spe_expand_stv_builtin): Delete.
29183         (bdesc_2arg_spe): Delete.
29184         (spe_expand_builtin): Delete.
29185         (spe_expand_predicate_builtin): Delete.
29186         (spe_expand_evsel_builtin): Delete.
29187         (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
29188         (spe_init_builtins): Delete.
29189         (spe_func_has_64bit_regs_p): Delete.
29190         (savres_routine_name): Delete "info" parameter.  Adjust callers.
29191         (rs6000_emit_stack_reset): Ditto.
29192         (rs6000_dwarf_register_span): Delete.
29193         * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
29194         UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
29195         SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
29196         Delete.
29197         * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
29198         Delete.
29199         * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
29200         * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
29201         * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
29202         mulv2sf3, divv2sf3): Delete expanders.
29204 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29206         config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
29208 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29210         * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
29211         * config/rs6000/rs6000.c: Ditto.
29213 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29215         * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
29216         * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
29217         comparison_operator.
29219 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29221         * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
29222         * config/rs6000/rs6000.opt: Ditto.
29223         * config/rs6000/t-rtems: Ditto.
29225 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29227         * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
29228         TARGET_E500_SINGLE by 0, simplify.
29229         * config/rs6000/rs6000.c: Ditto.
29230         (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
29231         (spe_build_register_parallel): Delete.
29232         * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
29233         TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
29234         * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
29235         TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
29236         (E500_CONVERT): Delete.
29237         * config/rs6000/spe.md: Remove many patterns and all define_constants.
29239 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29241         * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
29242         * config/rs6000/dfp.md: Ditto.
29243         (negdd2, *negdd2_fpr): Merge.
29244         (absdd2, *absdd2_fpr): Merge.
29245         (negtd2, *negtd2_fpr): Merge.
29246         (abstd2, *abstd2_fpr): Merge.
29247         * config/rs6000/e500.h: Delete file.
29248         * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29249         TARGET_FPRS by 1 and simplify.
29250         * config/rs6000/rs6000-c.c: Ditto.
29251         * config/rs6000/rs6000.c: Ditto.  Also replace TARGET_SF_SPE and
29252         TARGET_DF_SPE by 0.
29253         * config/rs6000/rs6000.h: Ditto.  Delete TARGET_SF_SPE and
29254         TARGET_DF_SPE.
29255         * config/rs6000/rs6000.md: Ditto.
29256         (floatdidf2, *floatdidf2_fpr): Merge.
29257         (move_from_CR_gt_bit): Delete.
29258         * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29259         (E500_CR_IOR_COMPARE): Delete.
29260         (All patterns that require !TARGET_FPRS): Delete.
29261         * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29263 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29265         * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29267 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29269         * graphds.c (add_edge): Intitialize edge's attached data.
29270         (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29271         pointer parameter.  Call pointed function on each edge during
29272         graph traversing.  Skip traversing the edge when the function
29273         returns true.
29274         (graphds_dfs, graphds_scc): Ditto.
29275         (for_each_edge): New parameter.  Pass the new parameter to callback
29276         function.
29277         * graphds.h (skip_edge_callback): New function pointer type.
29278         (graphds_dfs, graphds_scc): New function pointer parameter.
29279         (graphds_edge_callback, for_each_edge): New parameter.
29281 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29283         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29284         out code checking if runtime alias check is possible to below ...
29285         Call the new function.
29286         * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29287         * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29289 2017-06-07  Marek Polacek  <polacek@redhat.com>
29291         PR sanitizer/80932
29292         * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29293         TYPE_OVERFLOW_WRAPS check.
29295 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29297         * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29298         if versioning is required.
29299         * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29300         peeling with the check for versioning.
29302 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29304         * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29305         * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29306         Set true to new parameter if new ssa variable is defined.
29307         (vect_gen_vector_loop_niters): Refactor.  Set range information
29308         for the new vector loop bound variable.
29309         (vect_do_peeling): Ditto.
29311 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29313         * tree-affine.c (ssa.h): Include header file.
29314         (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29315         has wrapping overflow behavior.
29317 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29319         * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29321 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29323         (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29324         (tree_to_aff_combination): ... here.
29326 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29328         * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29329         reg_pressure model function.
29330         (ivopts_global_cost_for_size): Delete.
29331         (determine_set_costs, iv_ca_recount_cost): Call new model function
29332         ivopts_estimate_reg_pressure.
29334 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29336         * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29337         expensive than udiv.  Remove floating point cases from mod.
29339 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29341         * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29342         Increase idiv cost.
29344 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29346         * config/aarch64/aarch64.md
29347         (copysignsf3): Fix mask generation.
29349 2017-06-07  Jakub Jelinek  <jakub@redhat.com>
29351         * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29352         TDI_gimple.
29353         (class dump_manager): Add register_dumps method.
29354         * dumpfile.c: Include langhooks.h.
29355         (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29356         (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29357         (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29358         (dump_manager::dump_register): Start with 512 entries instead of 32.
29359         (dump_manager::register_dumps): New method.
29360         * toplev.c (general_init): Instead of invoking register_dumps
29361         langhook, invoke register_dumps method on the dump manager.
29362         * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29363         TDI_generic.
29365 2017-06-07  Richard Sandiford  <richard.sandiford@linaro.org>
29367         * doc/md.texi: Clarify the restrictions on a define_insn condition.
29368         Say that # requires an associated define_split to exist, and that
29369         the define_split must be suitable for use after register allocation.
29371 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29373         * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29374         (compute_outgoing_frequencies): Also initialize zero counts.
29375         (find_many_sub_basic_blocks): Do not produce uninitialized profile
29376         around loops; preserve more of profile when nothing changes.
29378 2017-06-06  Jim Wilson  <jim.wilson@linaro.org>
29380         * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29381         here.
29382         * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29383         * config/arm/arm-cpu-cdata.h: Regenerate.
29384         * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29385         * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29386         * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29387         * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29388         * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29389         support.
29390         * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29391         support.
29392         * config/arm/t-rmprofile: Likewise.
29393         * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29395 2017-06-06  David S. Miller  <davem@davemloft.net>
29397         PR target/80968
29398         * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29399         blockage if function uses alloca.
29401 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29403         * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29404         New "uid" fields to hold pretty-print IDs of group and ref.
29405         Memory references are now identified as <group_id>:<ref_id>
29406         instead of using [random] addresses.
29407         (dump_mem_details): Simplify, no functional change.
29408         (dump_mem_ref): Simplify and make output more concise.
29409         Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29410         (find_or_create_group): Initialize group uid.
29411         (record_ref): Initialize ref uid.  Improve debug output.
29412         (prune_group_by_reuse, should_issue_prefetch_p,)
29413         (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29414         (mark_nontemporal_store, determine_loop_nest_reuse):
29415         Improve debug output.
29417 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29419         * dbgcnt.def (prefetch): New debug counter.
29420         * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29421         (schedule_prefetches): Stop issueing prefetches if debug counter
29422         tripped.
29424 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29426         * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29427         gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29429 2017-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29431         * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29432         Use aarch64_reg_or_zero predicate for operand 4.
29433         (aarch64_compare_and_swap<mode> define_insn_and_split):
29434         Use aarch64_reg_or_zero predicate for operand 3.  Add 'Z' constraint.
29435         (aarch64_store_exclusive<mode>): Likewise for operand 2.
29437 2017-06-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29439         * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29440         (arm_compute_save_core_reg_mask): This.
29441         (thumb1_compute_save_reg_mask): Rename into ...
29442         (thumb1_compute_save_core_reg_mask): This.
29443         (arm_compute_save_reg0_reg12_mask): Adapt comment.
29444         (arm_compute_frame_layout): Likewise.
29446 2017-06-06  Richard Biener  <rguenther@suse.de>
29448         PR tree-optimization/80974
29449         * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29450         keep or clear leaders SSA info.
29452 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29454         * config/nvptx/nvptx.c (split_mode_p): New function.
29455         (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29457 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29459         * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29461 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29463         PR bootstrap/80978
29464         * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29465         profile.
29467 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29469         * shrink-wrap.c (handle_simple_exit): Update profile.
29470         (try_shrink_wrapping): Upate profile.
29472 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29474         * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29475         (tree_guess_outgoing_edge_probabilities): New.
29476         * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29477         * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29479 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29481         * ipa-split.c (split_function): Initialize return bb profile.
29483 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29485         * profile.c (compute_branch_probabilities): Also initialize
29486         EXIT_BLOCK profile.
29488 2017-06-06  Richard Biener  <rguenther@suse.de>
29490         PR tree-optimization/80928
29491         * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29492         (vect_analyze_loop_operations): Properly guard analysis for
29493         pure SLP case.
29494         (vect_transform_loop): Likewise.
29495         (vect_analyze_loop_2): Also reset SLP type on PHIs.
29496         (vect_model_induction_cost): Do not cost for pure SLP.
29497         (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29498         of induction in inner loop vectorization.
29499         * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29500         (vect_get_and_check_slp_defs): Handle vect_induction_def.
29501         (vect_build_slp_tree): Likewise.  Handle PHIs as terminating the
29502         recursion.
29503         (vect_analyze_slp_cost_1): Cost induction.
29504         (vect_detect_hybrid_slp_stmts): Handle PHIs.
29505         (vect_get_slp_vect_defs): Likewise.
29506         * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29507         (vect_transform_stmt): Handle SLP reductions.
29508         * tree-vectorizer.h (vectorizable_induction): Adjust.
29510 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29512         * config/rs6000/rs6000.c (make_resolver_func): Update
29513         init_lowered_empty_function call.
29515 2017-06-05  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29517         * doc/invoke.texi: Document the -fprofile-abs-path option.
29518         * common.opt (fprofile-abs-path): New option.
29519         * gcov-io.h (gcov_write_filename): Declare.
29520         * gcov-io.c (gcov_write_filename): New function.
29521         * coverage.c (coverage_begin_function): Use gcov_write_filename.
29522         * profile.c (output_location): Likewise.
29524 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29526         * shring-wrap.c: Revert accidental commit.
29528 2017-06-05  Volker Reichelt  <v.reichelt@netcologne.de>
29530         * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29532 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29534         * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29535         new edge.
29536         * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29537         profile in callgraph edge.
29538         * profile-count.h (apply_probability): If THIS is 0, then result is 0
29539         (apply_scale): Likewise.
29540         * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29541         Also scale profile when inlining function with zero profile.
29542         (initialize_cfun): Update exit block profile even when it is zero.
29543         * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29544         when profile is read.
29546 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29548         * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29549         (CLONE_*): New constants to define the processors we can generate
29550         code for with the target_clone attribute.
29551         (rs6000_clone_map): New array to identify which clone processors
29552         the current program is running on.
29553         (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29554         target_clone attribute.
29555         (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29556         (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29557         (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29558         (cpu_expand_builtin): Add support for target_clone attribute.
29559         (rs6000_valid_attribute_p): Allow "default" attribute.
29560         (get_decl_name): New debug function to simplify printing the
29561         current function name in debugging statements.
29562         (rs6000_clone_priority): New functions to support the target_clone
29563         attribute, and be able to generate code to switch between ISA 2.05
29564         through ISA 3.0 (power6 through power9).
29565         (rs6000_compare_version_priority): Likewise.
29566         (rs6000_get_function_versions_dispatcher): Likewise.
29567         (make_resolver_func): Likewise.
29568         (add_condition_to_bb): Likewise.
29569         (dispatch_function_versions): Likewise.
29570         (rs6000_generate_version_dispatcher_body): Likewise.
29571         (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29572         (fusion_gpr_load_p): Fix a spacing issue.
29573         * doc/extend.texi (Common Function Attributes): Document that the
29574         PowerPC supports the target_clone attribute.
29576 2017-06-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29578         * config/arm/arm.h: explain F symbol found in description of ARM
29579         register allocation in its legend.
29581 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29583         * config/mips/frame-header-opt.c: Include profile-count.h.
29584         * config/riscv/riscv.c: Include profile-count.h
29586 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29588         * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29589         update profile.
29590         (sm_set_flag_if_changed): Add bbs field.
29591         (execute_sm_if_changed_flag_set): Pass BBS.
29592         (execute_sm): Update.
29594 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29596         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29597         New pattern.
29599 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29601         * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29602         (peephole2): New peephole2 to emit the above.
29603         * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29605 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29607         * config/aarch64/aarch64.c (define_peephole2 above
29608         *sub_<shift>_<mode>): New peephole.
29610 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
29612         * config/i386/i386.c (make_resolver_func): Update.
29613         * Makefile.in: Add profile-count.h and profile-count.o
29614         * auto-profile.c (afdo_indirect_call): Update to new API.
29615         (afdo_set_bb_count): Update.
29616         (afdo_propagate_edge): Update.
29617         (afdo_propagate_circuit): Update.
29618         (afdo_calculate_branch_prob): Update.
29619         (afdo_annotate_cfg): Update.
29620         * basic-block.h: Include profile-count.h
29621         (struct edge_def): Turn count to profile_count.
29622         (struct basic_block_def): Likewie.
29623         (REG_BR_PROB_BASE): Move to profile-count.h
29624         (RDIV): Move to profile-count.h
29625         * bb-reorder.c (max_entry_count): Turn to profile_count.
29626         (find_traces): Update.
29627         (rotate_loop):Update.
29628         (connect_traces):Update.
29629         (sanitize_hot_paths):Update.
29630         * bt-load.c (migrate_btr_defs): Update.
29631         * cfg.c (RDIV): Remove.
29632         (init_flow): Use alloc_block.
29633         (alloc_block): Uninitialize count.
29634         (unchecked_make_edge): Uninitialize count.
29635         (check_bb_profile): Update.
29636         (dump_edge_info): Update.
29637         (dump_bb_info): Update.
29638         (update_bb_profile_for_threading): Update.
29639         (scale_bbs_frequencies_int): Update.
29640         (scale_bbs_frequencies_gcov_type): Update.
29641         (scale_bbs_frequencies_profile_count): New.
29642         * cfg.h (update_bb_profile_for_threading): Update.
29643         (scale_bbs_frequencies_profile_count): Declare.
29644         * cfgbuild.c (compute_outgoing_frequencies): Update.
29645         (find_many_sub_basic_blocks): Update.
29646         * cfgcleanup.c (try_forward_edges): Update.
29647         (try_crossjump_to_edge): Update.
29648         * cfgexpand.c (expand_gimple_tailcall): Update.
29649         (construct_exit_block): Update.
29650         * cfghooks.c (verify_flow_info): Update.
29651         (dump_bb_for_graph): Update.
29652         (split_edge): Update.
29653         (make_forwarder_block): Update.
29654         (duplicate_block): Update.
29655         (account_profile_record): Update.
29656         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29657         (get_estimated_loop_iterations): Update.
29658         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29659         (single_likely_exit): Update.
29660         * cfgloopmanip.c (scale_loop_profile): Update.
29661         (loopify): Update.
29662         (set_zero_probability): Update.
29663         (lv_adjust_loop_entry_edge): Update.
29664         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29665         (purge_dead_edges): Update.
29666         (rtl_account_profile_record): Update.
29667         * cgraph.c (cgraph_node::create): Uninitialize count.
29668         (symbol_table::create_edge): Uninitialize count.
29669         (cgraph_update_edges_for_call_stmt_node): Update.
29670         (cgraph_edge::dump_edge_flags): Update.
29671         (cgraph_node::dump): Update.
29672         (cgraph_edge::maybe_hot_p): Update.
29673         * cgraph.h: Include profile-count.h
29674         (create_clone), create_edge, create_indirect_edge): Update.
29675         (cgraph_node): Turn count to profile_count.
29676         (cgraph_edge0: Likewise.
29677         (make_speculative, clone): Update.
29678         (create_edge): Update.
29679         (init_lowered_empty_function): Update.
29680         * cgraphclones.c (cgraph_edge::clone): Update.
29681         (duplicate_thunk_for_node): Update.
29682         (cgraph_node::create_clone): Update.
29683         * cgraphunit.c (cgraph_node::analyze): Update.
29684         (cgraph_node::expand_thunk): Update.
29685         * final.c (dump_basic_block_info): Update.
29686         * gimple-streamer-in.c (input_bb): Update.
29687         * gimple-streamer-out.c (output_bb): Update.
29688         * graphite.c (print_global_statistics): Update.
29689         (print_graphite_scop_statistics): Update.
29690         * hsa-brig.c: Include basic-block.h.
29691         * hsa-dump.c: Include basic-block.h.
29692         * hsa-gen.c (T sum_slice): Update.
29693         (convert_switch_statements):Update.
29694         * hsa-regalloc.c: Include basic-block.h.
29695         * ipa-chkp.c (chkp_produce_thunks): Update.
29696         * ipa-cp.c (struct caller_statistics): Update.
29697         (init_caller_stats): Update.
29698         (gather_caller_stats): Update.
29699         (ipcp_cloning_candidate_p): Update.
29700         (good_cloning_opportunity_p): Update.
29701         (get_info_about_necessary_edges): Update.
29702         (dump_profile_updates): Update.
29703         (update_profiling_info): Update.
29704         (update_specialized_profile): Update.
29705         (perhaps_add_new_callers): Update.
29706         (decide_about_value): Update.
29707         (ipa_cp_c_finalize): Update.
29708         * ipa-devirt.c (struct odr_type_warn_count): Update.
29709         (struct decl_warn_count): Update.
29710         (struct final_warning_record): Update.
29711         (possible_polymorphic_call_targets): Update.
29712         (ipa_devirt): Update.
29713         * ipa-fnsummary.c (redirect_to_unreachable): Update.
29714         * ipa-icf.c (sem_function::merge): Update.
29715         * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29716         * ipa-inline.c (compute_uninlined_call_time): Update.
29717         (compute_inlined_call_time): Update.
29718         (want_inline_small_function_p): Update.
29719         (want_inline_self_recursive_call_p): Update.
29720         (edge_badness): Update.
29721         (lookup_recursive_calls): Update.
29722         (recursive_inlining): Update.
29723         (inline_small_functions): Update.
29724         (dump_overall_stats): Update.
29725         (dump_inline_stats): Update.
29726         * ipa-profile.c (ipa_profile_generate_summary): Update.
29727         (ipa_propagate_frequency): Update.
29728         (ipa_profile): Update.
29729         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29730         * ipa-utils.c (ipa_merge_profiles): Update.
29731         * loop-doloop.c (doloop_modify): Update.
29732         * loop-unroll.c (report_unroll): Update.
29733         (unroll_loop_runtime_iterations): Update.
29734         * lto-cgraph.c (lto_output_edge): Update.
29735         (lto_output_node): Update.
29736         (input_node): Update.
29737         (input_edge): Update.
29738         (merge_profile_summaries): Update.
29739         * lto-streamer-in.c (input_cfg): Update.
29740         * lto-streamer-out.c (output_cfg): Update.
29741         * mcf.c (create_fixup_graph): Update.
29742         (adjust_cfg_counts): Update.
29743         (sum_edge_counts): Update.
29744         * modulo-sched.c (sms_schedule): Update.
29745         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29746         * predict.c (maybe_hot_count_p): Update.
29747         (probably_never_executed): Update.
29748         (dump_prediction): Update.
29749         (combine_predictions_for_bb): Update.
29750         (propagate_freq): Update.
29751         (handle_missing_profiles): Update.
29752         (counts_to_freqs): Update.
29753         (rebuild_frequencies): Update.
29754         (force_edge_cold): Update.
29755         * predict.h: Include profile-count.h
29756         (maybe_hot_count_p, counts_to_freqs): UPdate.
29757         * print-rtl-function.c: Do not include cfg.h
29758         * print-rtl.c: Include basic-block.h
29759         * profile-count.c: New file.
29760         * profile-count.h: New file.
29761         * profile.c (is_edge_inconsistent): Update.
29762         (correct_negative_edge_counts): Update.
29763         (is_inconsistent): Update.
29764         (set_bb_counts): Update.
29765         (read_profile_edge_counts): Update.
29766         (compute_frequency_overlap): Update.
29767         (compute_branch_probabilities): Update; Initialize and deinitialize
29768         gcov_count tables.
29769         (branch_prob): Update.
29770         * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29771         (edge_gcov_count): New.
29772         (bb_gcov_count): New.
29773         * shrink-wrap.c (try_shrink_wrapping): Update.
29774         * tracer.c (better_p): Update.
29775         * trans-mem.c (expand_transaction): Update.
29776         (ipa_tm_insert_irr_call): Update.
29777         (ipa_tm_insert_gettmclone_call): Update.
29778         * tree-call-cdce.c: Update.
29779         * tree-cfg.c (gimple_duplicate_sese_region): Update.
29780         (gimple_duplicate_sese_tail): Update.
29781         (gimple_account_profile_record): Update.
29782         (execute_fixup_cfg): Update.
29783         * tree-inline.c (copy_bb): Update.
29784         (copy_edges_for_bb): Update.
29785         (initialize_cfun): Update.
29786         (freqs_to_counts): Update.
29787         (copy_cfg_body): Update.
29788         (expand_call_inline): Update.
29789         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29790         * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29791         (try_unroll_loop_completely): Update.
29792         (try_peel_loop): Update.
29793         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29794         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29795         * tree-ssa-loop-split.c (connect_loops): Update.
29796         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29797         * tree-ssa-reassoc.c (branch_fixup): Update.
29798         * tree-ssa-tail-merge.c (replace_block_by): Update.
29799         * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29800         (compute_path_counts): Update.
29801         (update_profile): Update.
29802         (recompute_probabilities): Update.
29803         (update_joiner_offpath_counts): Update.
29804         (estimated_freqs_path): Update.
29805         (freqs_to_counts_path): Update.
29806         (clear_counts_path): Update.
29807         (ssa_fix_duplicate_block_edges): Update.
29808         (duplicate_thread_path): Update.
29809         * tree-switch-conversion.c (case_bit_test_cmp): Update.
29810         (struct switch_conv_info): Update.
29811         * tree-tailcall.c (decrease_profile): Update.
29812         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29813         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29814         * value-prof.c (check_counter): Update.
29815         (gimple_divmod_fixed_value): Update.
29816         (gimple_mod_pow2): Update.
29817         (gimple_mod_subtract): Update.
29818         (gimple_ic_transform): Update.
29819         (gimple_stringop_fixed_value): Update.
29820         * value-prof.h (gimple_ic): Update.
29822 2017-06-02  Carl Love  <cel@us.ibm.com>
29824         * config/rs6000/rs6000-c: Add support for built-in functions
29825         vector double vec_doublee (vector signed int);
29826         vector double vec_doublee (vector unsigned int);
29827         vector double vec_doublee (vector float);
29828         vector double vec_doubleh (vector signed int);
29829         vector double vec_doubleh (vector unsigned int);
29830         vector double vec_doubleh (vector float);
29831         vector double vec_doublel (vector signed int);
29832         vector double vec_doublel (vector unsigned int);
29833         vector double vec_doublel (vector float);
29834         vector double vec_doubleo (vector signed int);
29835         vector double vec_doubleo (vector unsigned int);
29836         vector double vec_doubleo (vector float);.
29837         * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29838         DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29839         UNS_DOUBLEL.
29840         * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29841         unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29842         unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29843         VS_sxwsp.
29844         * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29845         vec_doublel, vec_doubleh.
29846         * doc/extend.texi: Update the built-in documentation file for the
29847         new built-in functions.
29849 2017-06-02  David Malcolm  <dmalcolm@redhat.com>
29851         PR jit/80954
29852         * ipa-inline-analysis.c (free_growth_caches): Set
29853         edge_removal_hook_holder to NULL after removing it.
29855 2017-06-02  Sudakshina Das  <sudi.das@arm.com>
29857         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29858         comparision with zero.
29860 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29861         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29862         for early expansion of vec_min and vec_max builtins.
29863         (builtin_function_type): Add min/max unsigned variants to those
29864         identified as having unsigned arguments.
29866 2017-06-02  Olivier Hainque  <hainque@adacore.com>
29868         * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29870 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29872         * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29873         Use VALL_F16 iterator rather than VALL.
29875 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29877         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29878         Emit CBNZ inside loop when doing a strong exchange and comparing
29879         against zero.  Generate the CC flags after the loop.
29881 2017-06-02  David Edelsohn  <dje.gcc@gmail.com>
29883         * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29884         (dl_section_ref): New.
29885         (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29886         On AIX, append an expression to subtract the size of the
29887         section length to dl_section_ref.
29889 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29891         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29892         for early expansion of vector absolute builtins.
29894 2017-06-02  Richard Biener  <rguenther@suse.de>
29896         * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29897         what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29899 2017-06-02  Richard Biener  <rguenther@suse.de>
29901         PR tree-optimization/80948
29902         * tree-tailcall.c (find_tail_calls): Track stmts to move in
29903         stmt order as well.
29905 2017-06-02  Richard Biener  <rguenther@suse.de>
29907         * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29908         PHIs are ok.
29909         * tree-vect-stmts.c (process_use): Do not mark backedge defs
29910         for inductions as relevant.
29912 2017-06-02  Richard Biener  <rguenther@suse.de>
29914         * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29915         (vectorizable_induction): ... this.  Remove dead code.
29917 2017-06-02  Eric Botcazou  <ebotcazou@adacore.com>
29919         * builtins. (expand_builtin_alloca): Remove second parameter and
29920         infer its value from the first parameter instead.
29921         (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
29923 2017-06-02  Jakub Jelinek  <jakub@redhat.com>
29925         PR rtl-optimization/80903
29926         * loop-doloop.c (add_test): Unshare sequence.
29928 2017-06-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29930         * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
29932 2017-06-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29934         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
29935         static.
29936         (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
29937         xlogue_layout::get_instance, logue_layout::xlogue_layout,
29938         sp_valid_at, fp_valid_at, choose_basereg): Formatting.
29939         (xlogue_layout::get_stub_rtx): Make static.
29940         (xlogue_layout::get_stub_name): Avoid const-cast, make static.
29941         (xlogue_layout::compute_stub_managed_regs): Rename to...
29942         (xlogue_layout::count_stub_managed_regs): ...this.
29943         (xlogue_layout::is_stub_managed_reg): New function.
29944         (xlogue_layout::m_stub_names): Rename to...
29945         (xlogue_layout::s_stub_names): ...this, make static.
29946         (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
29947         xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
29948         xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
29949         xlogue_layout::s_stub_names): Instantiate statics.
29950         (stub_managed_regs): Remove.
29951         (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
29952         (disable_call_ms2sysv_xlogues): Rename to...
29953         (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
29954         (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
29955         warning logic.
29956         (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
29957         change after reload_completed.
29958         (ix86_can_use_return_insn_p): Use the ix86_frame data structure
29959         directly.
29960         (ix86_expand_prologue): Likewise.
29961         (ix86_expand_epilogue): Likewise.
29962         (ix86_expand_split_stack_prologue): Likewise.
29963         (ix86_compute_frame_layout): Remove frame parameter ...
29964         (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
29965         (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
29966         only if necessary.
29967         (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
29968         (ix86_frame): Move from here ...
29969         * config/i386/i386.h (ix86_frame): ... to here.
29970         (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
29971         complete ix86_frame data structure instead.  Remove some_ld_name.
29973 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29975         * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
29976         symbols that hold a DECL_VALUE_EXPR.
29978 2017-06-01  Martin Jambor  <mjambor@suse.cz>
29980         PR tree-optimization/80898
29981         * tree-sra.c (process_subtree_disqualification): Removed.
29982         (disqualify_candidate): Do not acll
29983         process_subtree_disqualification.
29984         (subtree_mark_written_and_enqueue): New function.
29985         (propagate_all_subaccesses): Set grp_write of LHS subtree if the
29986         RHS has been disqualified and re-queue LHS if necessary.  Apart
29987         from that, ignore disqualified RHS.
29989 2017-06-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29991         * config/s390/s390.c (s390_emit_epilogue): Disable early return
29992         address fetch for z10 or later.
29994 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29996         * config/arc/arc.md (tst_movb): Add guard when splitting.
29998 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30000         * config/arc/arc.c (arc_can_eliminate): Test against
30001         arc_frame_pointer_needed.
30003 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30005         * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
30006         to prevent store reordering.
30007         * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
30008         (type): Add block type.
30009         (stack_tie): Define special instruction to be used in
30010         expand_prologue.
30012 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30014         * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
30015         constraint. It is not valid for the pattern.
30016         (noncommutative_binary_comparison): Likewise.
30018 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30020         * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
30021         scaled addresses.
30023 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30025         * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
30026         be used by the reg-alloc.
30028 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30030         * config/arc/arc.md (mulsi3): Avoid use of hard registers before
30031         reg-alloc when having mul64 or mul32x16 instructions.
30032         (mulsidi3): Likewise.
30033         (umulsidi3): Likewise.
30034         (mulsi32x16): New pattern.
30035         (mulsi64): Likewise.
30036         (mulsidi64): Likewise.
30037         (umulsidi64): Likewise.
30038         (MUL32x16_REG): Define.
30039         (mul64_600): Use MUL32x16_REG.
30040         (mac64_600): Likewise.
30041         (umul64_600): Likewise.
30042         (umac64_600): Likewise.
30044 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30046         * config/arc/arc.md (mulsi3_700): Make it commutative.
30048 2017-06-01  Jose E. Marchesi  <jose.marchesi@oracle.com>
30050         * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
30051         type for movstouw.
30052         (*sign_extendsidi2_insn): Likewise for movstosw.
30054 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
30056         * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
30057         the type of the input discriminant value.  Convert the
30058         discriminant value of signedness vary.
30060 2017-06-01  Volker Reichelt  <v.reichelt@netcologne.de>
30062         * doc/invoke.texi (-Wcatch-value): Document new shortcut.
30063         Add to -Wall section.
30065 2017-06-01  Richard Biener  <rguenther@suse.de>
30067         PR middle-end/66313
30068         * fold-const.c (fold_plusminus_mult_expr): If the factored
30069         factor may be zero use a wrapping type for the inner operation.
30070         * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
30071         and handle moved defs.
30072         (process_assignment): Properly guard the unary op case.  Return a
30073         tri-state indicating that moving the stmt before the call may allow
30074         to continue.  Pass through to_move.
30075         (find_tail_calls): Handle moving unrelated defs before
30076         the call.
30078 2017-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
30080         PR target/80618
30081         * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
30082         splitter result in the canonical way.
30084 2017-05-31  Uros Bizjak  <ubizjak@gmail.com>
30086         * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
30087         also for 32bit target.  Update insn attributes.
30088         (zero-extendsidi2 splitter): Allow all registers for operand 1.
30090 2017-05-31  Sebastian Peryt  <sebastian.peryt@intel.com>
30092         * config/i386/avx512fintrin.h (_mm_mask_max_sd)
30093         (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
30094         (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
30095         (_mm_maskz_min_ss): New intrinsics.
30097 2017-05-31  Martin Liska  <mliska@suse.cz>
30099         * tree-vect-loop.c (vect_create_epilog_for_reduction):
30100         Change comment style to one we normally use.
30101         (vectorizable_reduction): Likewise.
30102         (vectorizable_induction): Likewise.
30103         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
30104         (vectorizable_call): Likewise.
30105         (vectorizable_simd_clone_call): Likewise.
30106         (vectorizable_conversion): Likewise.
30107         (vectorizable_assignment): Likewise.
30108         (vectorizable_shift): Likewise.
30109         (vectorizable_operation): Likewise.
30110         (vectorizable_store): Likewise.
30111         (vectorizable_load): Likewise.
30112         * tree-vectorizer.h: Likewise.
30114 2017-05-31  Alexander Monakov  <amonakov@ispras.ru>
30116         * passes.c (emergency_dump_function): New.
30117         * tree-pass.h (emergency_dump_function): Declare.
30118         * plugin.c (plugins_internal_error_function): Remove.
30119         * plugin.h (plugins_internal_error_function): Remove declaration.
30120         * toplev.c (internal_error_function): New static function.  Use it...
30121         (general_init): ...here.
30123 2017-05-31  Graham Markall  <graham.markall@embecosm.com>
30125         * config/arc/arc.c (arc_print_operand): Handle constant operands.
30126         (arc_rtx_costs): Add costs for new patterns.
30127         * config/arc/arc.md: Additional *add_n and *sub_n patterns.
30128         * config/arc/predicates.md: Add _1_2_3_operand predicate.
30130 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30132         * tree-ssa-strlen.c (get_next_strinfo): New function.
30133         (get_stridx_plus_constant): Use it.
30134         (zero_length_string): Likewise.
30135         (adjust_related_strinfos): Likewise.
30136         (adjust_last_stmt): Likewise.
30138 2017-05-31  Richard Biener  <rguenther@suse.de>
30140         PR target/80880
30141         * config/i386/i386.c (ix86_expand_builtin): Remove assert
30142         for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
30144 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30146         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
30147         loop_vinfo argument and use of dependence distance vectors.
30148         Check instead whether the two references differ only in their
30149         initial value and assume that they have the same alignment if the
30150         difference is a multiple of the vector alignment.
30151         (vect_analyze_data_refs_alignment): Update call accordingly.
30153 2017-05-31  Martin Liska  <mliska@suse.cz>
30155         PR target/79155
30156         * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
30158 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30160         * tree-vect-loop-manip.c (create_intersect_range_checks_index)
30161         (create_intersect_range_checks): Move from ...
30162         * tree-data-ref.c (create_intersect_range_checks_index)
30163         (create_intersect_range_checks): ... to here.
30164         (create_runtime_alias_checks): New function factored from ...
30165         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
30166         here.  Call above function.
30167         * tree-data-ref.h (create_runtime_alias_checks): New function.
30169 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30171         * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
30172         segment length for dr_b and compute it in wide_int.
30174 2017-05-31  Richard Biener  <rguenther@suse.de>
30176         PR tree-optimization/80906
30177         * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
30178         and pass through iv_map.
30179         (copy_bb_and_scalar_dependences): Adjust.
30180         (translate_pending_phi_nodes): Likewise.
30181         (copy_loop_close_phi_args): Handle code-generating IVs instead
30182         of ICEing.
30184 2017-05-30  David Malcolm  <dmalcolm@redhat.com>
30186         * diagnostic-color.c (color_dict): Add "type-diff".
30187         (parse_gcc_colors): Update comment.
30188         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
30189         -fdiagnostics-show-template-tree and -fno-elide-type.
30190         (GCC_COLORS): Add type-diff to example.
30191         (type-diff=): New.
30192         (-fdiagnostics-show-template-tree): New.
30193         (-fno-elide-type): New.
30194         * pretty-print.c (pp_format): Pass quote and formatters[argno] to
30195         the pp_format_decoder callback.  Call any m_format_postprocessor's
30196         "handle" method.
30197         (pretty_printer::pretty_printer): Initialize
30198         m_format_postprocessor.
30199         (pretty_printer::~pretty_printer): Delete any
30200         m_format_postprocessor.
30201         * pretty-print.h (printer_fn): Add bool and const char ** parameters.
30202         (class format_postprocessor): New class.
30203         (struct pretty_printer::format_decoder): Document the new parameters.
30204         (struct pretty_printer::m_format_postprocessor): New field.
30205         * tree-diagnostic.c (default_tree_printer): Update for new
30206         bool and const char ** params.
30207         * tree-diagnostic.h (default_tree_printer): Likewise.
30209 2017-05-30  Segher Boessenkool  <segher@kernel.crashing.org>
30211         * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
30212         (lwa_operand): Delete rs6000_gen_cell_microcode test.
30213         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30214         rs6000_gen_cell_microcode code.
30215         (rs6000_final_prescan_insn): Delete.
30216         (rs6000_opt_vars): Delete the "gen-cell-microcode" and
30217         "warn-cell-microcode" entries.
30218         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
30219         * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
30220         throughout.  Change cc_reg_not_micro_cr0_operand to
30221         cc_reg_not_cr0_operand throughout.
30222         (*extendhi<mode>2_noload): Delete.
30223         * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
30224         (mwarn-cell-microcode): Delete.
30225         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
30226         -mgen-cell-microcode and -mwarn-cell-microcode.
30228 2017-05-30  Uros Bizjak  <ubizjak@gmail.com>
30230         PR target/80833
30231         * config/i386/constraints.md (Yd): New constraint.
30232         (Ye): Ditto.
30233         * config/i386/i386.md (*movti_internal): Add (?r, Ye)
30234         and (?Yd, r) alternatives.  Update insn attributes.
30235         * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
30236         and (?*Yd, r) alternatives.  Update insn attributes.
30237         (double-mode inter-unit splitters): Add new GR<->XMM splitters.
30239 2017-05-30  Pierre-Marie de Rodat  <derodat@adacore.com>
30241         * gimplify.c (gimplify_modify_expr): Don't create a
30242         DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30243         function.
30245 2017-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
30247         * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30249 2017-05-30  Richard Biener  <rguenther@suse.de>
30251         * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30252         and reduc_def fields.
30253         (STMT_VINFO_REDUC_TYPE): New define.
30254         (STMT_VINFO_REDUC_DEF): Likewise.
30255         (vect_force_simple_reduction): Adjust prototype.
30256         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30257         (vect_is_simple_reduction): Remove check_reduction argument.
30258         (vect_force_simple_reduction): Adjust and set
30259         STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30260         (vectorizable_reduction): Do not re-do reduction analysis
30261         but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30262         * tree-parloops.c (gather_scalar_reductions): Adjust.
30264 2017-05-30  Richard Biener  <rguenther@suse.de>
30266         PR middle-end/80901
30267         * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30268         split_edge code.
30270 2017-05-24  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30272         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30273         Introduce unknown_misalignment parameter and remove vf.
30274         (vect_peeling_hash_get_lowest_cost):
30275         Pass unknown_misalignment parameter.
30276         (vect_enhance_data_refs_alignment):
30277         Fix unsupportable data ref treatment.
30279 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30281         * tree-vect-data-refs.c (vect_get_data_access_cost):
30282         Workaround for SLP handling.
30283         (vect_enhance_data_refs_alignment):
30284         Compute costs for doing no peeling at all, compare to the best
30285         peeling costs so far and avoid peeling if cheaper.
30287 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30289         * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30290         Return peeling info and set costs to zero for unlimited cost
30291         model.
30292         (vect_enhance_data_refs_alignment): Also inspect all datarefs
30293         with unknown misalignment. Compute and costs for unknown
30294         misalignment, compare them to the costs for known misalignment
30295         and choose the cheapest for peeling.
30297 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30299         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30300         (vect_get_peeling_costs_all_drs): Create function.
30301         (vect_peeling_hash_get_lowest_cost):
30302         Use vect_get_peeling_costs_all_drs.
30303         (vect_peeling_supportable): Create function.
30304         (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30306 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30308         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30309         DR_HAS_NEGATIVE_STEP.
30310         (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30311         (vect_enhance_data_refs_alignment): Use.
30312         (vect_duplicate_ssa_name_ptr_info): Use.
30313         * tree-vectorizer.h (dr_misalignment): Use.
30314         (known_alignment_for_access_p): Use.
30316 2017-05-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
30318         PR target/78838
30319         * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30320         .lowtext.
30321         (has_section_name): New function.
30323 2017-05-30  Martin Liska  <mliska@suse.cz>
30325         PR other/80909
30326         * auto-profile.c (get_function_decl_from_block): Fix
30327         parenthesis.
30329 2017-05-30  Richard Biener  <rguenther@suse.de>
30331         PR middle-end/80876
30332         * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30334 2017-05-30  Martin Liska  <mliska@suse.cz>
30336         * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30337         * dumpfile.h (struct dump_file_info): Remove ctors.
30339 2017-05-30  Martin Liska  <mliska@suse.cz>
30341         * predict.def: Fix GNU coding style.
30343 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30345         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30346         Mark 'to' argument with ATTRIBUTE_UNUSED.
30348 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30350         * config/xtensa/xtensa.c (xtensa_emit_call): Use
30351         HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30352         (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30353         format string.
30355 2017-05-29  Eric Botcazou  <ebotcazou@adacore.com>
30357         * doc/install.texi (Options specification): Restore entry of
30358         --enable-sjlj-exceptions.
30360 2017-05-27  Michael Eager  <eager@eagercon.com>
30362         Revert:
30363         2016-01-21  Ajit Agarwal  <ajitkum@xilinx.com>
30365         See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30367         * config/microblaze/microblaze.h
30368         (FIXED_REGISTERS): Update in macro.
30369         (CALL_USED_REGISTERS): Update in macro.
30371 2017-05-27  François-Xavier Coudett  <fxcoudert@gcc.gnu.org>
30373         * doc/install.texi: Add links to macOS binary distributions.
30375 2017-05-27  Jakub Jelinek  <jakub@redhat.com>
30377         PR bootstrap/80887
30378         Revert:
30379         2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30381         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30383 2017-05-26  Martin Liska  <mliska@suse.cz>
30385         * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30387 2017-05-26  Martin Liska  <mliska@suse.cz>
30389         * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30390         always leading ';; '.
30391         (dump_bb_info): Likewise.
30392         (brief_dump_cfg): Likewise.
30393         * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30394         * dumpfile.c: Remove usage of TDF_VERBOSE.
30395         * dumpfile.h (enum dump_kind): Likewise.
30396         (dump_gimple_bb_header): Do not use TDF_COMMENT.
30397         * print-tree.c (debug_verbose): Remove.
30398         * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30399         (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30400         * tree-diagnostic.c (default_tree_printer): Replace
30401         TDF_DIAGNOSTIC with TDF_SLIM.
30403 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30405         * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30406         in parameter loop, rather than loop_vinfo.
30407         (create_intersect_range_checks): Ditto.
30408         (vect_create_cond_for_alias_checks): Update call to above functions.
30410 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30412         PR tree-optimization/80815
30413         * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30414         for merging runtime alias checks.  Handle negative DR_STEPs.
30416 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30418         * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30419         Move from ...
30420         * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30421         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30422         out code pruning runtime alias checks.
30423         * tree-data-ref.c (prune_runtime_alias_test_list): New function
30424         factored out from above.
30425         * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30426         Move from ...
30427         * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30428         ... to here.
30429         (prune_runtime_alias_test_list): New decalaration.
30431 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30433         * tree-vect-data-refs.c (compare_tree): Rename and move ...
30434         * tree-data-ref.c (data_ref_compare_tree): ... to here.
30435         * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30436         * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30437         (operator==, comp_dr_with_seg_len_pair): Ditto.
30438         (vect_prune_runtime_alias_test_list): Ditto.
30440 2017-05-26  Martin Liska  <mliska@suse.cz>
30442         PR ipa/80663
30443         * params.def: Bound partial-inlining-entry-probability param.
30445 2017-05-26  Marek Polacek  <polacek@redhat.com>
30447         PR sanitizer/80875
30448         * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30449         can be negated.
30451 2017-05-26  Richard Biener  <rguenther@suse.de>
30453         PR tree-optimization/80842
30454         * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30455         value.
30457 2017-05-26  Richard Biener  <rguenther@suse.de>
30459         PR tree-optimization/80844
30460         * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30462 2017-05-25  Sebastian Peryt  <sebastian.peryt@intel.com>
30464         * doc/md.texi (Machine Constraints): Update x86 family
30465         machine constraints section to match 'config/i386/constraints.md'.
30467 2017-05-25  Volker Reichelt  <v.reichelt@netcologne.de>
30469         * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30471 2017-05-25  Nathan Sidwell  <nathan@acm.org>
30473         * doc/invoke.texi (--enable-languages): Update documentation.
30475 2017-05-25  Martin Liska  <mliska@suse.cz>
30477         * dumpfile.c: Add TDF_FOLDING.
30478         * dumpfile.h (enum dump_kind): Likewise.
30479         * genmatch.c (dt_simplify::gen_1): Use it.
30481 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30483         * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30485 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30487         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30488         * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30490 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30492         * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30493         * match.pd (X == C): Rewrite it here.
30494         (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30495         with_certain_nonzero_bits2): New predicates.
30496         * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30498 2017-05-24  Nathan Sidwell  <nathan@acm.org>
30500         * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30501         avoid warning.
30503         * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30504         warning.
30506 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30508         * config/powerpcspe: New port.  Files are copied from the rs6000
30509         port, with "rs6000" in filenames replaced by "powerpcspe".
30511 2017-05-24  Wilco Dijkstra  <wdijkstr@arm.com>
30513         PR rtl-optimization/80754
30514         * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30516 2017-05-24  Sheldon Lobo  <smlobo@sheldon.us.oracle.com>
30518         * config/sparc/sparc.md (length): Return the correct value for -mflat
30519         sibcalls to match output_sibcall.
30521 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30523         PR bootstrap/80860
30524         PR bootstrap/80843
30525         * config/rs6000/rs6000.c (struct machine_function): Add new field
30526         n_components.
30527         (rs6000_get_separate_components): Init that field, use it.
30528         (rs6000_components_for_bb): Use the field.
30530 2017-05-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30532         * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30534 2017-05-24  Peter Bergner  <bergner@vnet.ibm.com>
30536         PR middle-end/80823
30537         * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30539 2017-05-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30541         PR target/80725
30542         * config/s390/s390.c (s390_check_qrst_address): Check incoming
30543         address against address_operand predicate.
30544         * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30546 2017-05-24  Eric Botcazou  <ebotcazou@adacore.com>
30548         * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30549         parameters passed indirectly.
30551 2017-05-23  Uros Bizjak  <ubizjak@gmail.com>
30553         * config/i386/i386.md (*movdi_internal): Remove SSE4
30554         alternative 18 (?r, *v).  Update insn attributes.
30555         (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30556         Update insn attributes.
30557         (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30558         Update insn attributes.
30559         * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30560         alternative 1 (r, v). Remove isa attribute.
30561         * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30562         Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30563         and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30565 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30567         * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30568         dg-line directive.
30570 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30572         * cgraphunit.c (symbol_table::process_new_functions): Update.
30573         * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30574         (inline_generate_summary): Rename to ...
30575         (ipa_fn_summary_generate): ... this one.
30576         (inline_read_summary): Rename to ...
30577         (ipa_fn_summary_read): ... this one.
30578         (inline_write_summary): Rename to ...
30579         (ipa_fn_summary_write): ... this one.
30580         (inline_free_summary): Rename to ...
30581         (ipa_free_fn_summary): ... this one.
30582         (pass_data_local_fn_summary, pass_local_fn_summary,
30583         make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30584         pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30585         pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30586         make_pass_ipa_fn_summary): New.
30587         * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30588         inline_write_summary, inline_free_summary): Remove.
30589         (ipa_free_fn_summary) : New.
30590         * ipa-inline.c (ipa_inline): Update.
30591         (pass_ipa_inline): Do not generate summaries.
30592         * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30593         Remove.
30594         * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30595         and add pass_ipa_fn_summary.
30596         * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30597         New.
30598         (make_pass_inline_parameters): Remove.
30600 2017-05-23  Thomas Schwinge  <thomas@codesourcery.com>
30602         * omp-low.c (struct omp_context): Remove "default_kind" member.
30603         Adjust all users.
30605         * omp-offload.c (execute_oacc_device_lower): Remove the
30606         parallelism dimensions function attributes for unparallelized
30607         OpenACC kernels constructs.
30609 2017-05-23  Martin Liska  <mliska@suse.cz>
30611         * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30612         functions.
30613         (cgraph_edge::make_speculative): Likewise.
30614         (cgraph_edge::resolve_speculation): Likewise.
30615         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30616         (cgraph_node::dump): Likewise.
30617         * cgraph.h: Likewise.
30618         * cgraphunit.c (analyze_functions): Likewise.
30619         (symbol_table::compile): Likewise.
30620         * ipa-cp.c (print_all_lattices): Likewise.
30621         (determine_versionability): Likewise.
30622         (initialize_node_lattices): Likewise.
30623         (ipcp_verify_propagated_values): Likewise.
30624         (estimate_local_effects): Likewise.
30625         (update_profiling_info): Likewise.
30626         (create_specialized_node): Likewise.
30627         (perhaps_add_new_callers): Likewise.
30628         (decide_about_value): Likewise.
30629         (decide_whether_version_node): Likewise.
30630         (identify_dead_nodes): Likewise.
30631         (ipcp_store_bits_results): Likewise.
30632         * ipa-devirt.c (dump_targets): Likewise.
30633         (ipa_devirt): Likewise.
30634         * ipa-icf.c (sem_item::dump): Likewise.
30635         (sem_function::equals): Likewise.
30636         (sem_variable::equals): Likewise.
30637         (sem_item_optimizer::read_section): Likewise.
30638         (sem_item_optimizer::execute): Likewise.
30639         (congruence_class::dump): Likewise.
30640         * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30641         (dump_inline_summary): Likewise.
30642         (estimate_node_size_and_time): Likewise.
30643         (inline_analyze_function): Likewise.
30644         * ipa-inline-transform.c (inline_call): Likewise.
30645         * ipa-inline.c (report_inline_failed_reason): Likewise.
30646         (want_early_inline_function_p): Likewise.
30647         (edge_badness): Likewise.
30648         (update_edge_key): Likewise.
30649         (inline_small_functions): Likewise.
30650         * ipa-profile.c (ipa_profile): Likewise.
30651         * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30652         (ipa_make_edge_direct_to_target): Likewise.
30653         (remove_described_reference): Likewise.
30654         (ipa_impossible_devirt_target): Likewise.
30655         (propagate_controlled_uses): Likewise.
30656         (ipa_print_node_params): Likewise.
30657         (ipcp_transform_function): Likewise.
30658         * ipa-pure-const.c (pure_const_read_summary): Likewise.
30659         (propagate_pure_const): Likewise.
30660         * ipa-reference.c (generate_summary): Likewise.
30661         (read_write_all_from_decl): Likewise.
30662         (propagate): Likewise.
30663         (ipa_reference_read_optimization_summary): Likewise.
30664         * ipa-utils.c (ipa_merge_profiles): Likewise.
30665         * ipa.c (walk_polymorphic_call_targets): Likewise.
30666         (symbol_table::remove_unreachable_nodes): Likewise.
30667         (ipa_single_use): Likewise.
30668         * passes.c (execute_todo): Likewise.
30669         * predict.c (drop_profile): Likewise.
30670         * symtab.c (symtab_node::get_dump_name): New function.
30671         (symtab_node::dump_name): Likewise.
30672         (symtab_node::dump_asm_name): Likewise.
30673         (symtab_node::dump_references): Likewise.
30674         (symtab_node::dump_referring): Likewise.
30675         (symtab_node::dump_base): Likewise.
30676         (symtab_node::debug_symtab): Likewise.
30677         * tree-sra.c (convert_callers_for_node): Likewise.
30678         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30679         * value-prof.c (init_node_map): Likewise.
30681 2017-05-23  Martin Liska  <mliska@suse.cz>
30683         * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30684         and symtab_node::debug_symtab to symbol_table::debug.
30685         * cgraphunit.c (analyze_functions): Use the renamed function.
30686         (symbol_table::compile): Likewise.
30687         * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30688         * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30689         * passes.c (execute_todo): Likewise.
30690         * symtab.c (symbol_table::dump): New function.
30691         * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30693 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30695         * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30696         that nonconst implies exec.
30698 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30700         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30701         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30702         (inline_edge_summary_vec): Turn into ...
30703         (ipa_call_summaries): ... this one.
30704         (redirect_to_unreachable, edge_set_predicate,
30705         evaluate_properties_for_edge, inline_summary_alloc,
30706         reset_ipa_call_summary, reset_inline_summary,
30707         inline_summary_t::duplicate): Update.
30708         (inline_edge_duplication_hook): Turn to ...
30709         (ipa_call_summary_t::duplicate): ... this one.
30710         (inline_edge_removal_hook): Turn to ...
30711         (ipa_call_summary_t::remove): ... this one.
30712         (dump_inline_edge_summary): Turn to ...
30713         (dump_ipa_call_summary): ... this one.
30714         (estimate_function_body_sizes): Update.
30715         (inline_update_callee_summaries): Update.
30716         (remap_edge_change_prob): Update.
30717         (remap_edge_summaries): Update.
30718         (inline_merge_summary): Update.
30719         (do_estimate_edge_time): Update.
30720         (inline_generate_summary): Update.
30721         (inline_read_section): Update.
30722         (inline_read_summary): Update.
30723         (inline_free_summary): Update.
30724         * ipa-inline.c (can_inline_edge_p): Update.
30725         (compute_inlined_call_time): Update.
30726         (want_inline_small_function_p): Update.
30727         (edge_badness): Update.
30728         (early_inliner): Update.
30729         * ipa-inline.h (inline_edge_summary): Turn to ...
30730         (ipa_call_summary): ... this one.
30731         (ipa_call_summary_t): New class.
30732         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30733         (ipa_call_summaries): New.
30734         (inline_edge_summary): Remove.
30735         (estimate_edge_growth): Update.
30736         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30737         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30738         * ipa-split.c (execute_split_functions): Update.
30739         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30741 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30743         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30744         attributes): Document rdrand effective target.
30746 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30748         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30749         attributes): Sort alphabetically.
30751 2017-05-23  Georg-Johann Lay  <avr@gjlay.de>
30753         * config/avr/genmultilib.awk: Use gsub instead of gensub.
30755 2017-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30757         PR target/80718
30758         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30759         V2DF/V2DI splat into two separate patterns, one that handles
30760         registers, and the other that only handles memory.  Drop support
30761         for splatting from a GPR on ISA 2.07 and then splitting the
30762         splat into direct move and splat.
30763         (vsx_splat_<mode>_reg): Likewise.
30764         (vsx_splat_<mode>_mem): Likewise.
30766 2017-05-22  Segher Boessenkool  <segher@kernel.crashing.org>
30768         * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30770 2017-05-22  Jakub Jelinek  <jakub@redhat.com>
30772         PR middle-end/80809
30773         * omp-low.c (finish_taskreg_remap): New function.
30774         (finish_taskreg_scan): If unit size of ctx->record_type
30775         is non-constant, unshare the size expression and replace
30776         decls in it with possible outer var refs.
30778         PR middle-end/80809
30779         * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30780         GOVD_SHARED rather than GOVD_PRIVATE with it.
30781         (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30782         GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30784         PR middle-end/80853
30785         * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30786         as last argument to build_outer_var_ref for pointer bases of array
30787         section reductions.
30789 2017-05-19  Martin Sebor  <msebor@redhat.com>
30791         * print-tree.c (print_node): Print DECL_READ_P flag.
30793 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30795         * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30796         * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30797         * cgraph.c: Likewise.
30798         * cgraphunit.c: Likewise.
30799         * gengtype.c: Likewise.
30800         * ipa-cp.c: Likewise.
30801         * ipa-devirt.c: Likewise.
30802         * ipa-icf.c: Likewise.
30803         * ipa-predicate.c: Likewise.
30804         * ipa-profile.c: Likewise.
30805         * ipa-prop.c: Likewise.
30806         * ipa-split.c: Likewise.
30807         * ipa.c: Likewise.
30808         * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30809         edge_predicate_pool, dump_inline_hints,
30810         inline_summary::account_size_time, redirect_to_unreachable,
30811         edge_set_predicate, set_hint_predicate,
30812         evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30813         inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30814         inline_summary_t::remove, remap_hint_predicate_after_duplication,
30815         inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30816         ipa_call_summary_t::remove, initialize_growth_caches,
30817         free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30818         debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30819         mark_modified, unmodified_parm_1, unmodified_parm,
30820         unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30821         set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30822         compute_bb_predicates, will_be_nonconstant_expr_predicate,
30823         will_be_nonconstant_predicate, record_modified_bb_info,
30824         get_minimal_bb, record_modified, param_change_prob,
30825         phi_result_unknown_predicate, predicate_for_phi_result,
30826         array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30827         estimate_function_body_sizes, compute_inline_parameters,
30828         compute_inline_parameters_for_curren, pass_data_inline_parameters,
30829         estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30830         inline_update_callee_summaries, remap_edge_change_prob,
30831         remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30832         inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30833         inline_analyze_function, inline_summary_t::insert,
30834         inline_generate_summary, read_ipa_call_summary, inline_read_section,
30835         inline_read_summary, write_ipa_call_summary, inline_write_summary,
30836         inline_free_summary): Move to ipa-fnsummary.h
30837         (predicate_t): Remove.
30838         * ipa-fnsummary.c: New file.
30839         * ipa-inline.h:  Do not include sreal.h and ipa-predicate.h
30840         (enum inline_hints_vals, inline_hints, agg_position_info,
30841         INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30842         inline_summaries, ipa_call_summary, ipa_call_summary_t,
30843         ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30844         dump_inline_summary, dump_inline_hints, inline_generate_summary,
30845         inline_read_summary, inline_write_summary, inline_free_summary,
30846         inline_analyze_function, initialize_inline_failed,
30847         inline_merge_summary, inline_update_overall_summary,
30848         compute_inline_parameters): Move to ipa-fnsummary.h
30849         * ipa-fnsummary.h: New file.
30850         * ipa-inline-transform.h: Include ipa-inline.h.
30851         * ipa-inline.c: LIkewise.
30853 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30855         * ipa-inline.c (edge_badness): Use inlined_time instead of
30856         inline_summaries->get.
30858 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30860         * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30862 2017-05-22  Nathan Sidwell  <nathan@acm.org>
30864         * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30865         (fdump-lang): Document 'raw' option.
30866         * dumpfile.h (TDI_tu): Delete.
30867         * dumpfile.c (dump_files): Remove translation-unit.
30868         (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30870 2017-05-22  Georg-Johann Lay  <avr@gjlay.de>
30872         * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30873         command option from $(AWK) call.
30874         * config/avr/genmultilib.awk: Simplify and rewrite so that it
30875         generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30876         [FORMAT]: Remove handling of variable.
30877         * config/avr/t-multilib: Regenerate.
30879 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30881         * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30882         self_time.
30883         (dump_inline_summary): Do not print self_time.
30884         (estimate_function_body_sizes): Do not set self_time.
30885         (compute_inline_parameters): Likewise.
30886         (inline_read_section, inline_write_summary): Do not stream self_time.
30887         * ipa-inline.h (inline_summary): Drop self_time.
30889 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30891         * ipa-inline-analysis.c (account_size_time): Rename to ...
30892         (inline_summary::account_size_time): ... this one.
30893         (reset_ipa_call_summary): Turn to ...
30894         (ipa_call_summary::reset): ... this one.
30895         (reset_inline_summary): Turn to ...
30896         (inline_summary::reset): ... this one.
30897         (inline_summary_t::remove): Update.
30898         (inline_summary_t::duplicate): Update.
30899         (ipa_call_summary_t::remove): Update.
30900         (dump_inline_summary): Update.
30901         (estimate_function_body_sizes): Update.
30902         (compute_inline_parameters): Update.
30903         (estimate_node_size_and_time): Update.
30904         (inline_merge_summary): Update.
30905         (inline_update_overall_summary): Update.
30906         (inline_read_section): Update.
30907         (inline_write_summary): Update.
30908         * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30909         add account_size_time and reset member functions.
30910         (ipa_call_summary): Add reset function.
30911         * ipa-predicate.h (predicate::operator &): Constify.
30913 2017-05-22  Richard Biener  <rguenther@suse.de>
30915         * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30917 2017-05-19  Jason Merrill  <jason@redhat.com>
30919         * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
30921 2017-05-19  Marek Polacek  <polacek@redhat.com>
30923         PR sanitizer/80800
30924         * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
30925         TYPE_OVERFLOW_WRAPS checks.
30927 2017-05-19  Thomas Schwinge  <thomas@codesourcery.com>
30929         * tree-core.h (enum omp_clause_default_kind): Add
30930         "OMP_CLAUSE_DEFAULT_PRESENT".
30931         * tree-pretty-print.c (dump_omp_clause): Handle it.
30932         * gimplify.c (enum gimplify_omp_var_data): Add
30933         "GOVD_MAP_FORCE_PRESENT".
30934         (gimplify_adjust_omp_clauses_1): Map it to
30935         "GOMP_MAP_FORCE_PRESENT".
30936         (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
30938         * gimplify.c (oacc_default_clause): Clarify.
30940 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30942         LANG_HOOK_REGISTER_DUMPS
30943         * toplev.c (general_init): Call register dump lang hook.
30944         * doc/invoke.texi: Document -fdump-lang option family.
30945         * dumpfile.c (dump_files): Remove class dump here.
30946         (FIRST_AUTO_NUMBERED_DUMP): Adjust.
30947         * dumpfile.h (tree_dump_index): Remove TDI_class.
30948         * langhooks-def.h (lhd_register_dumps): Declare.
30949         (LANG_HOOKS_REGISTER_DUMPS): Define.
30950         (LANG_HOOKS_INITIALIZER): Add it.
30951         * langhooks.c (lhd_register_dumps): Define.
30952         * langhooks.h (struct lang_hooks): Add register_dumps.
30954 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30956         * context.h (context::set_passes): New.
30957         * context.c (context::context): Do not create pass manager.
30958         * toplev.c (general_init): Create pass manager here.
30960 2017-05-19  Segher Boessenkool  <segher@kernel.crashing.org>
30962         * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
30963         use this splitter if two add or or instructions would also work for
30964         the constant we want to generate.
30966 2017-05-19  Richard Biener  <rguenther@suse.de>
30968         PR build/80821
30969         * genmatch.c (dt_node::gen_kids_1): Add missing scope around
30970         predicate evaluation.
30972 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30974         * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
30975         add ctor.
30976         * ipa-inline.c (want_inline_small_function_p): Do not cast to
30977         unsigned.
30979 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30981         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30982         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30983         (inline_edge_summary_vec): Turn into ...
30984         (ipa_call_summaries): ... this one.
30985         (redirect_to_unreachable, edge_set_predicate,
30986         evaluate_properties_for_edge, inline_summary_alloc,
30987         reset_ipa_call_summary, reset_inline_summary,
30988         inline_summary_t::duplicate): Update.
30989         (inline_edge_duplication_hook): Turn to ...
30990         (ipa_call_summary_t::duplicate): ... this one.
30991         (inline_edge_removal_hook): Turn to ...
30992         (ipa_call_summary_t::remove): ... this one.
30993         (dump_inline_edge_summary): Turn to ...
30994         (dump_ipa_call_summary): ... this one.
30995         (estimate_function_body_sizes): Update.
30996         (inline_update_callee_summaries): Update.
30997         (remap_edge_change_prob): Update.
30998         (remap_edge_summaries): Update.
30999         (inline_merge_summary): Update.
31000         (do_estimate_edge_time): Update.
31001         (inline_generate_summary): Update.
31002         (inline_read_section): Update.
31003         (inline_read_summary): Update.
31004         (inline_free_summary): Update.
31005         * ipa-inline.c (can_inline_edge_p): Update.
31006         (compute_inlined_call_time): Update.
31007         (want_inline_small_function_p): Update.
31008         (edge_badness): Update.
31009         (early_inliner): Update.
31010         * ipa-inline.h (inline_edge_summary): Turn to ...
31011         (ipa_call_summary): ... this one.
31012         (ipa_call_summary_t): New class.
31013         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
31014         (ipa_call_summaries): New.
31015         (inline_edge_summary): Remove.
31016         (estimate_edge_growth): Update.
31017         * ipa-profile.c (ipa_propagate_frequency_1): Update.
31018         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
31019         * ipa-split.c (execute_split_functions): Update.
31020         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
31022 2017-05-19  Richard Biener  <rguenther@suse.de>
31024         PR middle-end/80764
31025         * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
31027 2017-05-18  Segher Boessenkool  <segher@kernel.crashing.org>
31029         * config/rs6000/rs6000.c (struct machine_function): Add field
31030         fpr_is_wrapped_separately.
31031         (rs6000_get_separate_components): Use 64 components.  Handle the
31032         new FPR components.
31033         (rs6000_components_for_bb): Handle the FPR components.
31034         (rs6000_emit_prologue_components): Handle the FPR components.
31035         (rs6000_emit_epilogue_components): Handle the FPR components.
31036         (rs6000_set_handled_components): Handle the FPR components.
31037         (rs6000_emit_prologue): Don't output prologue code for those FPRs
31038         that are already separately shrink-wrapped.
31039         (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
31040         that are already separately shrink-wrapped.
31042 2017-05-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
31044         PR target/80510
31045         * config/rs6000/predicates.md (simple_offsettable_mem_operand):
31046         New predicate.
31048         * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
31049         (define_peephole2 for Altivec d-form load): Add peepholes to catch
31050         cases where the register allocator uses a move and an offsettable
31051         memory operation to/from a FPR register on ISA 2.06/2.07.
31052         (define_peephole2 for Altivec d-form store): Likewise.
31054 2017-05-18  Uros Bizjak  <ubizjak@gmail.com>
31056         PR target/80799
31057         * config/i386/mmx.md (*mov<mode>_internal): Enable
31058         alternatives 11, 12, 13 and 14 also for 32bit targets.
31059         Remove alternatives 15, 16, 17 and 18.
31060         * config/i386/sse.md (vec_concatv2di): Change
31061         alternative (!x, *y) to (x, ?!*Yn).
31063 2017-05-18  Paolo Carlini  <paolo.carlini@oracle.com>
31065         * dumpfile.h (enum dump_kind): Remove stray comma.
31067 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31069         * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
31070         * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
31071         predicate::num_conditions
31072         (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
31073         (CHANGED): turn into predicate::changed.
31074         (agg_position_info): Move to ipa-predicate.h
31075         (add_condition, predicate::add_clause, predicate::operator &=,
31076         predicate::or_with, predicate::evaluate, predicate::probability,
31077         dump_condition, dump_clause, predicate::dump,
31078         predicate::remap_after_duplication, predicate::remap_after_inlining,
31079         predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
31080         (evaluate_conditions_for_known_args): Update.
31081         (set_cond_stmt_execution_predicate): Update.
31082         * ipa-inline.h: Include ipa-predicate.h
31083         (condition, inline_param_summary, conditions, agg_position_info,
31084         predicate): Move to ipa-predicate.h
31085         * ipa-predicate.c: New file.
31086         * ipa-predicate.h: New file.
31088 2017-05-18  Wilco Dijkstra  <wdijkstr@arm.com>
31090         * final.c (leaf_function_p): Check we are not in a sequence.
31092 2017-05-18  Martin Liska  <mliska@suse.cz>
31094         * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
31095         * dumpfile.c (dump_register): Use new enum dump_kind.
31096         (get_dump_file_name): Likewise.
31097         (dump_enable_all): Likewise.
31098         (dump_switch_p_1): Likewise.
31099         (enable_rtl_dump_file): Remove usage of TDF_RTL.
31100         * dumpfile.h (enum dump_kind): New enum type.
31101         (struct dump_file_info): Create constructor and
31102         format fields and comments.
31103         * passes.c (pass_manager::register_one_dump_file):
31104         Use num dump_kind.
31105         * statistics.c (statistics_early_init): Likewise.
31106         * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
31107         TDF_TREE with TDF_SLIM.
31108         (gather_memory_references_ref): Likewise.
31110 2017-05-18  Martin Liska  <mliska@suse.cz>
31112         * vec.h (struct vnull): Use it.
31114 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31116         * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
31117         (true_predicate, false_predicate, true_predicate_p,
31118         false_predicate_p): Remove.
31119         (single_cond_predicate, not_inlined_predicate): Turn to member function
31120         in ipa-inline.h
31121         (add_condition): Update.
31122         (add_clause): Turn to...
31123         (predicate::add_clause): ... this one; update; allow passing NULL
31124         as parameter.
31125         (and_predicates): Turn to ...
31126         (predicate::operator &=): ... this one.
31127         (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
31128         (or_predicates): Turn to ...
31129         (predicate::or_with): ... this one.
31130         (evaluate_predicate): Turn to ...
31131         (predicate::evaluate): ... this one.
31132         (predicate_probability): Turn to ...
31133         (predicate::probability): ... this one.
31134         (dump_condition): Update.
31135         (dump_predicate): Turn to ...
31136         (predicate::dump): ... this one.
31137         (account_size_time): Update.
31138         (edge_set_predicate): Update.
31139         (set_hint_predicate): UPdate.
31140         (evaluate_conditions_for_known_args): Update.
31141         (evaluate_properties_for_edge): Update.
31142         (remap_predicate_after_duplication): Turn to...
31143         (predicate::remap_after_duplication): ... this one.
31144         (remap_hint_predicate_after_duplication): Update.
31145         (inline_summary_t::duplicate): UPdate.
31146         (dump_inline_edge_summary): Update.
31147         (dump_inline_summary): Update.
31148         (set_cond_stmt_execution_predicate): Update.
31149         (set_switch_stmt_execution_predicate): Update.
31150         (compute_bb_predicates): Update.
31151         (will_be_nonconstant_expr_predicate): Update.
31152         (will_be_nonconstant_predicate): Update.
31153         (phi_result_unknown_predicate): Update.
31154         (predicate_for_phi_result): Update.
31155         (array_index_predicate): Update.
31156         (estimate_function_body_sizes): Update.
31157         (estimate_node_size_and_time): Update.
31158         (estimate_ipcp_clone_size_and_time): Update.
31159         (remap_predicate): Rename to ...
31160         (predicate::remap_after_inlining): ... this one.
31161         (remap_hint_predicate): Update.
31162         (inline_merge_summary): Update.
31163         (inline_update_overall_summary): Update.
31164         (estimate_size_after_inlining): Update.
31165         (read_predicate): Rename to ...
31166         (predicate::stream_in): ... this one.
31167         (read_inline_edge_summary): Update.
31168         (write_predicate): Rename to ...
31169         (predicate::stream_out): ... this one.
31170         (write_inline_edge_summary): Update.
31171         * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
31172         (clause_t): Turn to uint32_t
31173         (predicate): Turn to class; implement constructor and operators
31174         ==, !=, &
31175         (size_time_entry): Update.
31176         (inline_summary): Update.
31177         (inline_edge_summary): Update.
31179 2017-05-18  Marc Glisse  <marc.glisse@inria.fr>
31181         * fold-const.c (fold_binary_loc): Move transformation...
31182         * match.pd (C - X CMP X): ... here.
31184 2017-05-18  Sheldon Lobo  <sheldon.lobo@oracle.com>
31186         * config/sparc/sparc.c (sparc_option_override): Set function
31187         alignment for -mcpu=niagara7 to 64 to match the I$ line.
31188         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
31189         latency to 1.
31190         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
31191         latency to 2.
31192         * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
31194 2017-05-18  Marek Polacek  <polacek@redhat.com>
31196         PR sanitizer/80797
31197         * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
31198         (pass_ubsan::execute): Call gimple_assign_single_p instead of
31199         gimple_assign_load_p.
31201 2017-05-17  Segher Boessenkool  <segher@kernel.crashing.org>
31203         PR middle-end/80692
31204         * real.c (do_compare): Give decimal_do_compare preference over
31205         comparing just the signs.
31207 2017-05-17  Uros Bizjak  <ubizjak@gmail.com>
31209         * doc/md.texi (Canonicalization of Instructions): Describe the
31210         canonical form of instructions that inherently set a condition
31211         code register.
31213 2017-05-17  Peter Bergner  <bergner@vnet.ibm.com>
31215         PR middle-end/80775
31216         * tree-cfg.c: Move deletion of unreachable case statements to after
31217         the merging of consecutive case labels.
31219 2017-05-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
31221         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
31222         readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
31223         restoring of callee-saved registers.
31225 2017-05-17  Eric Botcazou  <ebotcazou@adacore.com>
31227         * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
31228         * config/visium/visium.c (single_set_and_flags): Likewise.
31229         * config/visium/visium.md (Substitutions): Likewise.
31231 2017-05-17  Martin Liska  <mliska@suse.cz>
31233         * cfg.c: Introduce dump_flags_t type and
31234         use it instead of int type.
31235         * cfg.h: Likewise.
31236         * cfghooks.c: Likewise.
31237         * cfghooks.h (struct cfg_hooks): Likewise.
31238         * cfgrtl.c: Likewise.
31239         * cfgrtl.h: Likewise.
31240         * cgraph.c (cgraph_node::get_body): Likewise.
31241         * coretypes.h: Likewise.
31242         * domwalk.c: Likewise.
31243         * domwalk.h: Likewise.
31244         * dumpfile.c (struct dump_option_value_info): Likewise.
31245         (dump_enable_all): Likewise.
31246         (dump_switch_p_1): Likewise.
31247         (opt_info_switch_p): Likewise.
31248         * dumpfile.h (enum tree_dump_index): Likewise.
31249         (struct dump_file_info): Likewise.
31250         * genemit.c: Likewise.
31251         * generic-match-head.c: Likewise.
31252         * gengtype.c (open_base_files): Likewise.
31253         * gimple-pretty-print.c: Likewise.
31254         * gimple-pretty-print.h: Likewise.
31255         * graph.c (print_graph_cfg): Likewise.
31256         * graphite-scop-detection.c (dot_all_sese): Likewise.
31257         * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31258         * loop-unroll.c (report_unroll): Likewise.
31259         * passes.c (pass_manager::register_one_dump_file): Likewise.
31260         * print-tree.c: Likewise.
31261         * statistics.c: Likewise.
31262         * tree-cfg.c: Likewise.
31263         * tree-cfg.h: Likewise.
31264         * tree-dfa.c: Likewise.
31265         * tree-dfa.h: Likewise.
31266         * tree-dump.c (dump_function): Likewise.
31267         * tree-dump.h (struct dump_info): Likewise.
31268         * tree-pretty-print.c: Likewise.
31269         * tree-pretty-print.h: Likewise.
31270         * tree-ssa-live.c: Likewise.
31271         * tree-ssa-live.h: Likewise.
31272         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31273         * tree-vect-loop.c: Likewise.
31274         * tree-vect-slp.c: Likewise.
31276 2017-05-16  James Greenhalgh  <james.greenhalgh@arm.com>
31277             Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31279         PR tree-optimization/80457
31280         * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31281         of all arguments to a statement as scalar_to_vec operations.
31282         (vectorizable_call): Adjust call to vect_model_simple_cost for
31283         new parameter.
31284         (vectorizable_conversion): Likewise.
31285         (vectorizable_assignment): Likewise.
31286         (vectorizable_shift): Likewise.
31287         (vectorizable_operation): Likewise.
31288         (vectorizable_comparison): Likewise.
31289         (vect_is_simple_cond): Record the def types for operands.
31290         (vectorizable_condition): Likewise, call vect_model_simple_cost.
31291         * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31292         for statement argument count.
31294 2017-05-16  Carl Love  <cel@us.ibm.com>
31296         * config/rs6000/rs6000-c: Add support for built-in functions
31297         vector unsigned long long vec_bperm (vector unsigned long long,
31298                                              vector unsigned char)
31299         vector signed long long vec_mule (vector signed int,
31300                                           vector signed int)
31301         vector unsigned long long vec_mule (vector unsigned int,
31302                                             vector unsigned int)
31303         vector signed long long vec_mulo (vector signed int,
31304                                           vector signed int)
31305         vector unsigned long long vec_mulo (vector unsigned int,
31306                                             vector unsigned int)
31307         vector signed char vec_sldw (vector signed char,
31308                                      vector signed char,
31309                                      const int)
31310         vector unsigned char vec_sldw (vector unsigned char,
31311                                        vector unsigned char,
31312                                        const int)
31313         vector signed short vec_sldw (vector signed short,
31314                                       vector signed short,
31315                                       const int)
31316         vector unsigned short vec_sldw (vector unsigned short,
31317                                         vector unsigned short,
31318                                         const int)
31319         vector signed int vec_sldw (vector signed int,
31320                                     vector signed int,
31321                                     const int)
31322         vector unsigned int vec_sldw (vector unsigned int,
31323                                       vector unsigned int,
31324                                       const int)
31325         vector signed long long vec_sldw (vector signed long long,
31326                                           vector signed long long,
31327                                           const int)
31328         vector unsigned long long vec_sldw (vector unsigned long long,
31329                                             vector unsigned long long,
31330                                             const int)
31331         * config/rs6000/rs6000-c: Add support for built-in functions
31332         * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31333         * config/rs6000/altivec.h: Add defintion for vec_sldw.
31334         * doc/extend.texi: Update the built-in documentation for the
31335         new built-in functions.
31337 2017-05-16  Marek Polacek  <polacek@redhat.com>
31339         PR sanitizer/80536
31340         PR sanitizer/80386
31341         * tree.c (save_expr): Don't fold the expression.
31343 2017-05-16  Uros Bizjak  <ubizjak@gmail.com>
31345         * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31346         to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31347         and (?*y,m).  Update insn attributes.
31349 2017-05-16  Martin Liska  <mliska@suse.cz>
31351         * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31352         flags argument of print_gimple_stmt, print_gimple_expr,
31353         print_generic_stmt and print_generic_expr.
31354         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31355         * coretypes.h: Likewise.
31356         * except.c (dump_eh_tree): Likewise.
31357         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31358         * gimple-pretty-print.h: Likewise.
31359         * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31360         (backprop::push_to_worklist): Likewise.
31361         (backprop::pop_from_worklist): Likewise.
31362         (backprop::process_use): Likewise.
31363         (backprop::intersect_uses): Likewise.
31364         (note_replacement): Likewise.
31365         * gimple-ssa-store-merging.c
31366         (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31367         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31368         (pass_store_merging::execute): Likewise.
31369         * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31370         (ssa_base_cand_dump_callback): Likewise.
31371         (dump_incr_vec): Likewise.
31372         (replace_refs): Likewise.
31373         (replace_mult_candidate): Likewise.
31374         (create_add_on_incoming_edge): Likewise.
31375         (create_phi_basis): Likewise.
31376         (insert_initializers): Likewise.
31377         (all_phi_incrs_profitable): Likewise.
31378         (introduce_cast_before_cand): Likewise.
31379         (replace_one_candidate): Likewise.
31380         * gimplify.c (gimplify_expr): Likewise.
31381         * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31382         (set_rename): Likewise.
31383         (rename_uses): Likewise.
31384         (copy_loop_phi_nodes): Likewise.
31385         (add_close_phis_to_merge_points): Likewise.
31386         (copy_loop_close_phi_args): Likewise.
31387         (copy_cond_phi_args): Likewise.
31388         (graphite_copy_stmts_from_block): Likewise.
31389         (translate_pending_phi_nodes): Likewise.
31390         * graphite-poly.c (print_pdr): Likewise.
31391         (dump_gbb_cases): Likewise.
31392         (dump_gbb_conditions): Likewise.
31393         (print_scop_params): Likewise.
31394         * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31395         (build_cross_bb_scalars_use): Likewise.
31396         (gather_bbs::before_dom_children): Likewise.
31397         * hsa-dump.c (dump_hsa_immed): Likewise.
31398         * ipa-cp.c (print_ipcp_constant_value): Likewise.
31399         (get_replacement_map): Likewise.
31400         * ipa-inline-analysis.c (dump_condition): Likewise.
31401         (estimate_function_body_sizes): Likewise.
31402         * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31403         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31404         * ipa-prop.c (ipa_dump_param): Likewise.
31405         (ipa_print_node_jump_functions_for_edge): Likewise.
31406         (ipa_modify_call_arguments): Likewise.
31407         (ipa_modify_expr): Likewise.
31408         (ipa_dump_param_adjustments): Likewise.
31409         (ipa_dump_agg_replacement_values): Likewise.
31410         (ipcp_modif_dom_walker::before_dom_children): Likewise.
31411         * ipa-pure-const.c (check_stmt): Likewise.
31412         (pass_nothrow::execute): Likewise.
31413         * ipa-split.c (execute_split_functions): Likewise.
31414         * omp-offload.c (dump_oacc_loop_part): Likewise.
31415         (dump_oacc_loop): Likewise.
31416         * trans-mem.c (tm_log_emit): Likewise.
31417         (tm_memopt_accumulate_memops): Likewise.
31418         (dump_tm_memopt_set): Likewise.
31419         (dump_tm_memopt_transform): Likewise.
31420         * tree-cfg.c (gimple_verify_flow_info): Likewise.
31421         (print_loop): Likewise.
31422         * tree-chkp-opt.c (chkp_print_addr): Likewise.
31423         (chkp_gather_checks_info): Likewise.
31424         (chkp_get_check_result): Likewise.
31425         (chkp_remove_check_if_pass): Likewise.
31426         (chkp_use_outer_bounds_if_possible): Likewise.
31427         (chkp_reduce_bounds_lifetime): Likewise.
31428         * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31429         (chkp_mark_completed_bounds): Likewise.
31430         (chkp_register_incomplete_bounds): Likewise.
31431         (chkp_mark_invalid_bounds): Likewise.
31432         (chkp_maybe_copy_and_register_bounds): Likewise.
31433         (chkp_build_returned_bound): Likewise.
31434         (chkp_get_bound_for_parm): Likewise.
31435         (chkp_build_bndldx): Likewise.
31436         (chkp_get_bounds_by_definition): Likewise.
31437         (chkp_generate_extern_var_bounds): Likewise.
31438         (chkp_get_bounds_for_decl_addr): Likewise.
31439         * tree-chrec.c (chrec_apply): Likewise.
31440         * tree-data-ref.c (dump_data_reference): Likewise.
31441         (dump_subscript): Likewise.
31442         (dump_data_dependence_relation): Likewise.
31443         (analyze_overlapping_iterations): Likewise.
31444         * tree-inline.c (expand_call_inline): Likewise.
31445         (tree_function_versioning): Likewise.
31446         * tree-into-ssa.c (dump_defs_stack): Likewise.
31447         (dump_currdefs): Likewise.
31448         (dump_names_replaced_by): Likewise.
31449         (dump_update_ssa): Likewise.
31450         (update_ssa): Likewise.
31451         * tree-object-size.c (pass_object_sizes::execute): Likewise.
31452         * tree-parloops.c (build_new_reduction): Likewise.
31453         (try_create_reduction_list): Likewise.
31454         (ref_conflicts_with_region): Likewise.
31455         (oacc_entry_exit_ok_1): Likewise.
31456         (oacc_entry_exit_single_gang): Likewise.
31457         * tree-pretty-print.h: Likewise.
31458         * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31459         (get_scalar_evolution): Likewise.
31460         (add_to_evolution): Likewise.
31461         (get_loop_exit_condition): Likewise.
31462         (analyze_evolution_in_loop): Likewise.
31463         (analyze_initial_condition): Likewise.
31464         (analyze_scalar_evolution): Likewise.
31465         (instantiate_scev): Likewise.
31466         (number_of_latch_executions): Likewise.
31467         (gather_chrec_stats): Likewise.
31468         (final_value_replacement_loop): Likewise.
31469         (scev_const_prop): Likewise.
31470         * tree-sra.c (dump_access): Likewise.
31471         (disqualify_candidate): Likewise.
31472         (create_access): Likewise.
31473         (reject): Likewise.
31474         (maybe_add_sra_candidate): Likewise.
31475         (create_access_replacement): Likewise.
31476         (analyze_access_subtree): Likewise.
31477         (analyze_all_variable_accesses): Likewise.
31478         (sra_modify_assign): Likewise.
31479         (initialize_constant_pool_replacements): Likewise.
31480         (find_param_candidates): Likewise.
31481         (decide_one_param_reduction): Likewise.
31482         (replace_removed_params_ssa_names): Likewise.
31483         * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31484         * tree-ssa-copy.c (dump_copy_of): Likewise.
31485         (copy_prop_visit_cond_stmt): Likewise.
31486         * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31487         * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31488         (record_equivalences_from_stmt): Likewise.
31489         * tree-ssa-dse.c (compute_trims): Likewise.
31490         (delete_dead_call): Likewise.
31491         (delete_dead_assignment): Likewise.
31492         * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31493         (forward_propagate_into_cond): Likewise.
31494         (pass_forwprop::execute): Likewise.
31495         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31496         * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31497         Likewise.
31498         (move_computations_worker): Likewise.
31499         (execute_sm): Likewise.
31500         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31501         (remove_exits_and_undefined_stmts): Likewise.
31502         (remove_redundant_iv_tests): Likewise.
31503         * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31504         (adjust_iv_update_pos): Likewise.
31505         * tree-ssa-math-opts.c (bswap_replace): Likewise.
31506         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31507         (value_replacement): Likewise.
31508         * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31509         * tree-ssa-pre.c (print_pre_expr): Likewise.
31510         (get_representative_for): Likewise.
31511         (create_expression_by_pieces): Likewise.
31512         (insert_into_preds_of_block): Likewise.
31513         (eliminate_insert): Likewise.
31514         (eliminate_dom_walker::before_dom_children): Likewise.
31515         (eliminate): Likewise.
31516         (remove_dead_inserted_code): Likewise.
31517         * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31518         * tree-ssa-reassoc.c (get_rank): Likewise.
31519         (eliminate_duplicate_pair): Likewise.
31520         (eliminate_plus_minus_pair): Likewise.
31521         (eliminate_not_pairs): Likewise.
31522         (undistribute_ops_list): Likewise.
31523         (eliminate_redundant_comparison): Likewise.
31524         (update_range_test): Likewise.
31525         (optimize_range_tests_var_bound): Likewise.
31526         (optimize_vec_cond_expr): Likewise.
31527         (rewrite_expr_tree): Likewise.
31528         (rewrite_expr_tree_parallel): Likewise.
31529         (linearize_expr): Likewise.
31530         (break_up_subtract): Likewise.
31531         (linearize_expr_tree): Likewise.
31532         (attempt_builtin_powi): Likewise.
31533         (attempt_builtin_copysign): Likewise.
31534         (transform_stmt_to_copy): Likewise.
31535         (transform_stmt_to_multiply): Likewise.
31536         (dump_ops_vector): Likewise.
31537         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31538         (print_scc): Likewise.
31539         (set_ssa_val_to): Likewise.
31540         (visit_reference_op_store): Likewise.
31541         (visit_use): Likewise.
31542         (sccvn_dom_walker::before_dom_children): Likewise.
31543         (run_scc_vn): Likewise.
31544         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31545         Likewise.
31546         (expr_hash_elt::print): Likewise.
31547         (const_and_copies::pop_to_marker): Likewise.
31548         (const_and_copies::record_const_or_copy_raw): Likewise.
31549         * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31550         * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31551         (dump_predicates): Likewise.
31552         (find_uninit_use): Likewise.
31553         (warn_uninitialized_phi): Likewise.
31554         (pass_late_warn_uninitialized::execute): Likewise.
31555         * tree-ssa.c (verify_vssa): Likewise.
31556         (verify_ssa): Likewise.
31557         (maybe_optimize_var): Likewise.
31558         * tree-vrp.c (dump_value_range): Likewise.
31559         (dump_all_value_ranges): Likewise.
31560         (dump_asserts_for): Likewise.
31561         (register_edge_assert_for_2): Likewise.
31562         (vrp_visit_cond_stmt): Likewise.
31563         (vrp_visit_switch_stmt): Likewise.
31564         (vrp_visit_stmt): Likewise.
31565         (vrp_visit_phi_node): Likewise.
31566         (simplify_cond_using_ranges_1): Likewise.
31567         (fold_predicate_in): Likewise.
31568         (evrp_dom_walker::before_dom_children): Likewise.
31569         (evrp_dom_walker::push_value_range): Likewise.
31570         (evrp_dom_walker::pop_value_range): Likewise.
31571         (execute_early_vrp): Likewise.
31573 2017-05-16  Richard Biener  <rguenther@suse.de>
31575         * dwarf2out.c (loc_list_from_tree_1): Do not create
31576         DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31578 2017-05-16  Richard Biener  <rguenther@suse.de>
31580         * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31581         just generated.
31582         (note_variable_value_in_expr): If we resolved the decl ref
31583         do not push to the stack.
31585 2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
31587         * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31588         operations in fast-math mode.
31589         (vaddq_f16): Likewise.
31590         (vmul_f16): Likewise.
31591         (vmulq_f16): Likewise.
31592         (vsub_f16): Likewise.
31593         (vsubq_f16): Likewise.
31594         * config/arm/neon.md (add<mode>3): New.
31595         (sub<mode>3): New.
31596         (fma:<VH:mode>3): New.  Also remove outdated comment.
31597         (mul<mode>3): New.
31599 2017-05-16  Martin Liska  <mliska@suse.cz>
31601         PR ipa/79849.
31602         PR ipa/79850.
31603         * ipa-devirt.c (warn_types_mismatch): Fix typo.
31604         (odr_types_equivalent_p): Likewise.
31606 2017-05-15  Sylvestre Ledru  <sylvestre@debian.org>
31608         * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31610 2017-05-15  Uros Bizjak  <ubizjak@gmail.com>
31612         PR target/80425
31613         * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31614         non-interunit SSE move alternatives with '?'.
31615         (zero-extendsidi peephole2): New peephole to skip intermediate
31616         general register in SSE zero-extend sequence.
31618 2017-05-15  Jeff Law  <law@redhat.com>
31620         * reorg.c (relax_delay_slots): Create a new variable to hold
31621         the temporary target rather than clobbering TARGET_LABEL.
31623         * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31624         missing argument to extract_bit_field call.
31625         * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31627 2017-05-15  Martin Liska  <mliska@suse.cz>
31629         PR driver/31468
31630         * gcc.c (process_command): Do not allow empty argument of -o option.
31632 2017-05-15  Renlin Li  <renlin.li@arm.com>
31634         * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31635         * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31636         * config/aarch64/constraints.md (Usf): Add long call check.
31637         * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31638         (call_value): Likewise.
31639         (sibcall): Likewise.
31640         (sibcall_value): Likewise.
31641         (call_insn): New.
31642         (call_value_insn): New.
31643         (sibcall_insn): Update rtx pattern.
31644         (sibcall_value_insn): Likewise.
31645         (call_internal): Remove.
31646         (call_value_internal): Likewise.
31647         (sibcall_internal): Likewise.
31648         (sibcall_value_internal): Likewise.
31649         (call_reg): Likewise.
31650         (call_symbol): Likewise.
31651         (call_value_reg): Likewise.
31652         (call_value_symbol): Likewise.
31654 2017-05-14  Krister Walfridsson  <krister.walfridsson@gmail.com>
31656         PR target/80600
31657         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31659 2017-05-14  Uros Bizjak  <ubizjak@gmail.com>
31661         * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31662         compatible with CCGOCmode and with CCZmode.
31664 2017-05-14  Martin Sebor  <msebor@redhat.com>
31666         PR middle-end/77671
31667         * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31668         (gimple_fold_builtin_snprintf): Same.
31669         * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31670         (gimple_fold_builtin_snprintf): Same.
31671         * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31672         of character types.
31673         (is_call_safe): New function.
31674         (try_substitute_return_value): Call it.
31675         (try_simplify_call): New function.
31676         (pass_sprintf_length::handle_gimple_call): Call it.
31678 2017-05-14  Martin Sebor  <msebor@redhat.com>
31680         PR middle-end/80669
31681         * builtins.c (expand_builtin_stpncpy): Simplify.
31683 2017-05-14  Daniel Santos  <daniel.santos@pobox.com>
31685         * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31686         * config/i386/i386.h
31687         (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31688         (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31689         (struct machine_function): Add new members call_ms2sysv,
31690         call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31691         (struct machine_frame_state): New fields sp_realigned and
31692         sp_realigned_offset.
31693         * config/i386/i386.c
31694         (enum xlogue_stub): New enum.
31695         (enum xlogue_stub_sets): New enum.
31696         (class xlogue_layout): New class.
31697         (struct ix86_frame): New fields stack_realign_allocate_offset,
31698         stack_realign_offset and outlined_save_offset.  Modify comments to
31699         detail stack layout when using out-of-line stubs.
31700         (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31701         (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31702         -mcall-ms2sysv-xlogues.
31703         (stub_managed_regs): New static variable.
31704         (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31705         registers managed by out-of-line stub.
31706         (disable_call_ms2sysv_xlogues): New function.
31707         (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31708         m->call_ms2sysv when appropriate and compute frame layout for
31709         out-of-line stubs.
31710         (sp_valid_at, fp_valid_at): New inline functions.
31711         (choose_basereg): New function.
31712         (choose_baseaddr): Add align parameter, use choose_basereg and modify
31713         all callers.
31714         (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31715         Use align parameter of choose_baseaddr to generated aligned SSE movs
31716         when possible.
31717         (pro_epilogue_adjust_stack): Modify to track
31718         machine_frame_state::sp_realigned.
31719         (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31720         (ix86_nsaved_sseregs): Likewise.
31721         (ix86_emit_save_regs): Likewise.
31722         (ix86_emit_save_regs_using_mov): Likewise.
31723         (ix86_emit_save_sse_regs_using_mov): Likewise.
31724         (get_scratch_register_on_entry): Likewise.
31725         (gen_frame_set): New function.
31726         (gen_frame_load): Likewise.
31727         (gen_frame_store): Likewise.
31728         (emit_outlined_ms2sysv_save): Likewise.
31729         (emit_outlined_ms2sysv_restore): Likewise.
31730         (ix86_expand_prologue): Modify stack re-alignment code and call
31731         emit_outlined_ms2sysv_save when appropriate.
31732         (ix86_emit_leave): Clear machine_frame_state::sp_realigned.  Add
31733         parameter rtx_insn *insn, which allows the function to be used to only
31734         generate the notes.
31735         (ix86_expand_epilogue): Modify validity checks of frame and stack
31736         pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31737         (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31738         * config/i386/predicates.md
31739         (save_multiple): New predicate.
31740         (restore_multiple): Likewise.
31741         * config/i386/sse.md
31742         (save_multiple<mode>): New pattern.
31743         (save_multiple_realign<mode>): Likewise.
31744         (restore_multiple<mode>): Likewise.
31745         (restore_multiple_and_return<mode>): Likewise.
31746         (restore_multiple_leave_return<mode>): Likewise.
31747         * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31749 2017-05-14  Julia Koval  <julia.koval@intel.com>
31751         * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31752         * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31753         (__builtin_ia32_xsetbv): New builtins.
31754         * config/i386/i386.c (ix86_expand_special_args_builtin):
31755         Process new types.
31756         (ix86_expand_builtin): Special expand for new intrinsics.
31757         * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31758         (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31759         * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31761 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31763         * cfganal.c (inverted_post_order_compute): Change argument type
31764         to vec *.
31765         * cfganal.h (inverted_post_order_compute): Adjust prototype.
31766         * df-core.c (rest_of_handle_df_initialize): Adjust.
31767         (rest_of_handle_df_finish): Likewise.
31768         (df_analyze_1): Likewise.
31769         (df_analyze): Likewise.
31770         (loop_inverted_post_order_compute): Change argument to be a vec *.
31771         (df_analyze_loop): Adjust.
31772         (df_get_n_blocks): Likewise.
31773         (df_get_postorder): Likewise.
31774         * df.h (struct df_d): Change field to be a vec.
31775         * lcm.c (compute_laterin): Adjust.
31776         (compute_available): Likewise.
31777         * lra-lives.c (lra_create_live_ranges_1): Likewise.
31778         * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31779         * tree-ssa-pre.c (compute_antic): Likewise.
31781 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31783         * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31784         (depth_first_search::depth_first_search): Change structure init
31785         function to this constructor.
31786         (depth_first_search::add_bb): Rename function to this member.
31787         (depth_first_search::execute): Likewise.
31788         (flow_dfs_compute_reverse_finish): Adjust.
31790 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31792         * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31793         (longest_simple_path): Likewise.
31794         * shrink-wrap.c (spread_components): Likewise.
31795         (disqualify_problematic_components): Likewise.
31796         (emit_common_heads_for_components): Likewise.
31797         (emit_common_tails_for_components): Likewise.
31798         (insert_prologue_epilogue_for_components): Likewise.
31800 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31802         * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31803         auto_sbitmap.
31805 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31807         * df-core.c (df_set_blocks): Start using auto_bitmap.
31808         (df_compact_blocks): Likewise.
31809         * df-problems.c (df_rd_confluence_n): Likewise.
31810         * df-scan.c (df_insn_rescan_all): Likewise.
31811         (df_process_deferred_rescans): Likewise.
31812         (df_update_entry_block_defs): Likewise.
31813         (df_update_exit_block_uses): Likewise.
31814         (df_entry_block_bitmap_verify): Likewise.
31815         (df_exit_block_bitmap_verify): Likewise.
31816         (df_scan_verify): Likewise.
31817         * lra-constraints.c (lra_constraints): Likewise.
31818         (undo_optional_reloads): Likewise.
31819         (lra_undo_inheritance): Likewise.
31820         * lra-remat.c (calculate_gen_cands): Likewise.
31821         (do_remat): Likewise.
31822         * lra-spills.c (assign_spill_hard_regs): Likewise.
31823         (spill_pseudos): Likewise.
31824         * tree-ssa-pre.c (bitmap_set_and): Likewise.
31825         (bitmap_set_subtract_values): Likewise.
31827 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31829         * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31830         management with auto_bitmap.
31831         (fix_inter_tick): Likewise.
31832         (fix_recovery_deps): Likewise.
31833         * ira.c (add_store_equivs): Likewise.
31834         (find_moveable_pseudos): Likewise.
31835         (split_live_ranges_for_shrink_wrap): Likewise.
31836         * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31837         (rtx_reuse_manager::seen_def_p): Likewise.
31838         (rtx_reuse_manager::set_seen_def): Likewise.
31839         * print-rtl.h (class rtx_reuse_manager): Likewise.
31841 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31843         * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31844         lifetime.
31845         (migrate_btr_def): Likewise.
31846         * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31847         * df-core.c (loop_post_order_compute): Likewise.
31848         (loop_inverted_post_order_compute): Likewise.
31849         * hsa-common.h: Likewise.
31850         * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31851         * init-regs.c (initialize_uninitialized_regs): Likewise.
31852         * ipa-inline.c (resolve_noninline_speculation): Likewise.
31853         (inline_small_functions): Likewise.
31854         * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31855         * ira.c (combine_and_move_insns): Likewise.
31856         (build_insn_chain): Likewise.
31857         * loop-invariant.c (find_invariants): Likewise.
31858         * lower-subreg.c (propagate_pseudo_copies): Likewise.
31859         * predict.c (tree_predict_by_opcode): Likewise.
31860         (predict_paths_leading_to): Likewise.
31861         (predict_paths_leading_to_edge): Likewise.
31862         (estimate_loops_at_level): Likewise.
31863         (estimate_loops): Likewise.
31864         * shrink-wrap.c (try_shrink_wrapping): Likewise.
31865         (spread_components): Likewise.
31866         * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31867         * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31868         * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31869         * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31870         * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31871         * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31872         * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31873         * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31874         (mark_threaded_blocks): Likewise.
31875         (thread_through_all_blocks): Likewise.
31876         * tree-ssa.c (verify_ssa): Likewise.
31877         (execute_update_addresses_taken): Likewise.
31878         * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31880 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31882         * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31883         auto_vec.
31884         (post_order_compute): Likewise.
31885         (inverted_post_order_compute): Likewise.
31886         (pre_and_rev_post_order_compute_fn): Likewise.
31888 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31890         * genrecog.c (int_set::int_set): Explicitly construct our
31891         auto_vec base class.
31892         * vec.h (auto_vec::auto_vec): New constructor.
31894 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31896         * bitmap.h (class auto_bitmap): New constructor taking
31897         bitmap_obstack * argument.
31899 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31901         * bitmap.h (class auto_bitmap): Change type of m_bits to
31902         bitmap_head, and adjust ctor / dtor and member operators.
31904 2017-05-13  Uros Bizjak  <ubizjak@gmail.com>
31906         * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31907         when returned register mode doesn't match original mode.
31909 2017-05-12  Jeff Law  <law@redhat.com>
31910             Jakub Jelinek  <jakub@redhat.com>
31912         * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31913         we look for cc setter after the compare-elim changes.
31914         * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31915         within the vector to match what compare-elim now expects.
31916         (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31917         (xorsi3_flags, one_cmplsi2_flags): Likewise.
31919         * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
31920         after the compare-elim changes.
31921         * config/rx/rx.md (abssi2_flags): Fix order of patterns within
31922         the vector to match what compare-elim now expects.
31923         (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
31924         (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
31925         (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
31926         (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
31927         (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
31929         * config/visium/visium.c (single_set_and_flags): Fix where
31930         we look for cc setter after the compare-elim changes.
31931         * config/visium/visium.md (flags_subst_logic): Fix order of patterns
31932         with the vector to match what compare-elim now expects.
31933         (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
31934         (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
31935         (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
31936         (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
31937         (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
31938         (neg<mode>2_insn_set_overflow): Likewise.
31940 2017-05-12  Jim Wilson  <jim.wilson@linaro.org>
31942         PR middle-end/79794
31943         * expmed.c (extract_bit_field_1): Add alt_rtl argument.  Before
31944         maybe_expand_insn call, set ops[0].target.  If still set after call,
31945         set alt_rtl.  Add extra arg to recursive calls.
31946         (extract_bit_field): Add alt_rtl argument.  Pass to
31947         extract_bit_field.
31948         * expmed.h (extract_bit_field): Fix prototype.
31949         * expr.c (emit_group_load_1, copy_blkmode_from_reg)
31950         (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
31951         to extract_bit_field_calls.
31952         (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
31953         Pass alt_rtl to extract_bit_field calls.
31954         * calls.c (store_unaligned_arguments_into_psuedos)
31955         load_register_parameters): Pass extra NULL to extract_bit_field calls.
31956         * optabs.c (maybe_legitimize_operand): Clear op->target when call
31957         gen_reg_rtx.
31958         * optabs.h (struct expand_operand): Add target bitfield.
31960 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31962         * compare-elim.c (try_eliminate_compare): Canonicalize
31963         operation with embedded compare to
31964         [(set (reg:CCM) (compare:CCM (operation) (immediate)))
31965          (set (reg) (operation)].
31967         * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
31969 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31971         PR target/80723
31972         * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
31973         cost of adding a carry flag for ADC instruction.
31974         [case MINUS]: Ignore the cost of subtracting a carry flag
31975         for SBB instruction.
31977 2017-05-12  Steven Munroe  <munroesj@gcc.gnu.org>
31979         * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
31980         and x86intrin.h
31981         * config/rs6000/bmiintrin.h: New file.
31982         * config/rs6000/bmi2intrin.h: New file.
31983         * config/rs6000/x86intrin.h: New file.
31985 2017-05-12  Jeff Law  <law@redhat.com>
31987         * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
31988         markers.
31990 2017-05-12  Peter Bergner  <bergner@vnet.ibm.com>
31992         PR middle-end/80707
31993         * tree-cfg.c: Remove cfg edges of unreachable case statements.
31995 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31997         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31998         early expansion of vector divide builtins.
31999         (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
32000         builtins identified as having unsigned arguments.
32002 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32004         * config/rs6000/rs6000.c (gimple-fold.h): New #include.
32005         (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
32006         expansion of vector logical operations (and, andc, or, xor,
32007         nor, orc, nand).
32009 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32011         * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
32012         * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
32014 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32016         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
32017         early GIMPLE expansion of vector multiplies.
32019 2017-05-12  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32021         * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
32022         TARGET_HAVE_MOVT conditional.
32023         (movt splitter): Likewise.
32025 2017-05-12  Richard Biener  <rguenther@suse.de>
32027         * tree-ssa-sccvn.h (has_VN_INFO): Declare.
32028         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32029         Fold all stmts not inplace.
32031 2017-05-12  Richard Biener  <rguenther@suse.de>
32033         PR tree-optimization/80713
32034         * tree-ssa-pre.c (remove_dead_inserted_code): Clear
32035         inserted_exprs bit for not removed stmts.
32037 2017-05-12  Thomas Schwinge  <thomas@codesourcery.com>
32039         PR middle-end/69921
32040         * tree-parloops.c (create_parallel_loop): Set "oacc kernels
32041         parallelized" attribute for parallelized OpenACC kernels.
32042         * omp-offload.c (execute_oacc_device_lower): Use it.
32044         * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
32045         Set "oacc kernels" attribute.
32046         * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
32047         parameter.  Adjust all users.
32048         (oacc_fn_attrib_kernels_p): Remove function.
32049         * omp-offload.c (execute_oacc_device_lower): Look for "oacc
32050         kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
32051         * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
32052         * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
32053         assert "oacc kernels" attribute is set.
32055 2017-05-11  Carl Love  <cel@us.ibm.com>
32057         * config/rs6000/rs6000-c: Add support for built-in functions
32058         vector unsigned char vec_popcnt (vector signed char)
32059         vector unsigned char vec_popcnt (vector unsigned char)
32060         vector unsigned short vec_popcnt (vector signed short)
32061         vector unsigned short vec_popcnt (vector unsigned short)
32062         vector unsigned int vec_popcnt (vector signed int)
32063         vector unsigned int vec_popcnt (vector unsigned int)
32064         vector unsigned long long vec_popcnt (vector signed long long)
32065         vector unsigned long long vec_popcnt (vector unsigned long long)
32066         vector signed long long vec_slo (vector signed long long,
32067                                          vector signed char)
32068         vector signed long long vec_slo (vector signed long long,
32069                                          vector unsigned char)
32070         vector unsigned long long vec_slo (vector unsigned long long,
32071                                            vector signed char)
32072         vector unsigned long long vec_slo (vector unsigned long long,
32073                                            vector unsigned char)
32074         * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
32075         VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
32076         * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
32077         vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
32078         * doc/extend.texi: Update the built-in documentation file for the
32079         new built-in functions.
32081 2017-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
32083         * attribs.h (sorted_attr_string): Move machine independent
32084         functions for target clone support from the i386 port to common
32085         code.  Rename ix86_function_versions to common_function_versions.
32086         Rename make_name to make_unique_name.
32087         (common_function_versions): Likewise.
32088         (make_unique_name): Likewise.
32089         (make_dispatcher_decl): Likewise.
32090         (is_function_default_version): Likewise.
32091         * attribs.c (attr_strcmp): Likewise.
32092         (sorted_attr_string): Likewise.
32093         (common_function_versions): Likewise.
32094         (make_unique_name): Likewise.
32095         (make_dispatcher_decl): Likewise.
32096         (is_function_default_version): Likewise.
32097         * config/i386/i386.c (attr_strcmp): Likewise.
32098         (sorted_attr_string): Likewise.
32099         (ix86_function_versions): Likewise.
32100         (make_name): Likewise.
32101         (make_dispatcher_decl): Likewise.
32102         (is_function_default_version): Likewise.
32103         (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
32105 2017-05-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32107         PR target/80695
32108         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
32109         Account for direct move costs for vec_construct of integer
32110         vectors.
32112 2017-05-11  Uros Bizjak  <ubizjak@gmail.com>
32114         PR target/80706
32115         * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
32116         (UNSPEC_STX_ATOMIC): Ditto.
32117         (loaddi_via_sse): New insn.
32118         (storedi_via_sse): Ditto.
32119         (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
32120         Update corresponding peephole2 patterns.
32121         (atomic_storedi_fpu): Ditto.
32123 2017-05-11  Julia Koval  <julia.koval@intel.com>
32125         * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
32126         (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
32127         New intrinsics.
32128         * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
32129         (__builtin_ia32_rsqrt14ss_mask): New builtins.
32130         * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
32132 2017-05-11  Nathan Sidwell  <nathan@acm.org>
32134         * graphite-poly.c: Include dumpfile.h.
32136         * dumpfle.h (dump_function): Declare here ...
32137         * tree-dump.h (dump_function): ... not here.
32138         * dumpfile.c: #include tree-cfg.h.
32139         (dump_function): Move here from ...
32140         * tree-dump.c (dump_function): ... here.
32141         * gimplify.c: #include splay-tree.h, not tree-dump.h.
32142         * graphite-poly.c: Don't include tree-dump.h.
32143         * cgraphclones.c: Include dumpfile.h not tree-dump.h.
32144         * print-tree.c: Likewise.
32145         * stor-layout.c: Likewise.
32146         * tree-nested.c: Likewise.
32148         * dumpfile.c (dump_start): Use TDF_FLAGS.
32149         (dump_enable_all): Fix TDF_KIND check thinko.
32151 2017-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32153         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
32154         array entries to represent two legal parameterizations of the
32155         overloaded __builtin_cmpb function, as represented by the
32156         P6_OV_BUILTIN_CMPB constant.
32157         (altivec_resolve_overloaded_builtin): Add special case handling
32158         for the __builtin_cmpb function, as represented by the
32159         P6_OV_BUILTIN_CMPB constant.
32160         * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
32161         (BU_P6_64BIT_2): New macro.
32162         (BU_P6_OVERLOAD_2): New macro
32163         (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
32164         (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
32165         (CMPB): Add overload support to represent both 32-bit and 64-bit
32166         compare-bytes function.
32167         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
32168         support for TARGET_CMPB.
32169         * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
32170         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
32171         documentation of the __builtin_cmpb overloaded built-in function.
32173 2017-05-11  Richard Biener  <rguenther@suse.de>
32175         PR tree-optimization/80705
32176         * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
32177         bases are not vectorizable.
32179 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32181         * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
32182         when counting register pressure.
32184 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32186         * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
32187         (iv_ca_more_deps): Renamed to ...
32188         (iv_ca_compare_deps): ... this.
32189         (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
32191 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32193         * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
32194         to ...
32195         (determine_group_iv_costs): ... here.
32196         (find_inv_vars_cb): Record inv var if it's not recorded before.
32198 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32200         * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
32201         (get_shiftadd_cost): Ditto.
32203 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32205         * tree-ssa-address.c: Include header file.
32206         (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
32207         address.
32208         (add_to_parts): Refactor.
32209         (addr_to_parts): New parameter.  Update use of move_hint_to_base.
32210         (create_mem_ref): Update use of addr_to_parts.  Re-associate addr
32211         in new order.
32213 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32215         PR tree-optimization/53090
32216         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
32217         COMP_IV_EXPR_2.
32218         (extract_cond_operands): Detect condition with IV on both sides
32219         and return COMP_IV_EXPR_2.
32220         (find_interesting_uses_cond): Add iv_use for both IVs in condition.
32221         (rewrite_use_compare): Simplify by removing call to function
32222         extract_cond_operands.
32224 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32226         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
32227         (extract_cond_operands): Detect condition comparing against non-
32228         invariant bound and return appropriate enum value.
32229         (find_interesting_uses_cond): Update use of extract_cond_operands.
32230         Handle its return value accordingly.
32231         (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
32233 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32235         * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
32236         nonlinear iv_use computation in loop invariant sensitive way.
32238 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32240         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
32241         (find_iv_candidates): Call relate_compare_use_with_all_cands.
32243 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32245         * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32246         (dump_cand): Support iv_cand.inv_exprs.
32247         (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32248         for candidates.
32249         (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32250         iv_cand.inv_exprs.
32252 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32254         * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32255         from ...
32256         * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32257         as local function.  Include necessary header files.
32258         * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32260 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32262         * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32264 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32266         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32267         operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32268         RSHIFT_EXPR and BIT_NOT_EXPR.
32270 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32272         * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32273         (adjust_setup_cost): New parameter supporting round up adjustment.
32274         (struct address_cost_data): Delete.
32275         (force_expr_to_var_cost): Don't bound cost with spill_cost.
32276         (split_address_cost, ptr_difference_cost): Delete.
32277         (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32278         (struct ainc_cost_data): New struct.
32279         (get_address_cost_ainc): New function.
32280         (get_address_cost, get_computation_cost): Reimplement.
32281         (determine_group_iv_cost_address): Record inv_expr for all uses of
32282         a group.
32283         (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32284         (iv_ca_has_deps): Reimplemented to ...
32285         (iv_ca_more_deps): ... this.  Check if NEW_CP introduces more deps
32286         than OLD_CP.
32287         (iv_ca_extend): Call iv_ca_more_deps.
32289 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32291         * tree-ssa-address.c (struct mem_address): Move to header file.
32292         (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32293         * tree-ssa-address.h (struct mem_address): Move from C file.
32294         (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32296 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32298         * tree-affine.h (aff_combination_type): New interface.
32299         (aff_combination_zero_p): Remove static.
32300         (aff_combination_const_p): New interface.
32301         (aff_combination_singleton_var_p): New interfaces.
32303 2017-05-11  Richard Biener  <rguenther@suse.de>
32305         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32306         Skip unreachable blocks and destinations.
32307         (eliminate): Move stmt removal and fixup ...
32308         (fini_eliminate): ... here.  Skip inserted exprs.
32309         (pass_pre::execute): Move fini_pre after fini_eliminate.
32310         * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32311         (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32312         PRE to get rid of dead code that has invalid SSA form and
32313         split critical edges again.
32315 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32317         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32319 2017-05-11  Richard Biener  <rguenther@suse.de>
32321         * passes.c (execute_function_todo): Verify loops if they are
32322         said to be up-to-date.
32323         * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32324         * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32326 2017-05-10  John David Anglin  <danglin@gcc.gnu.org>
32328         PR target/80090
32329         * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32330         handle calling assemble_external ourself.
32332         PR target/79027
32333         * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32334         modes with zero size.  Enhance comment.
32336 2017-05-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32338         * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32339         built-ins for vec_xl and vec_xst with short and char pointer
32340         arguments.
32342 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32344         * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32345         (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32346         (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32347         (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32348         (_mm_maskz_min_round_ss): New intrinsics.
32349         * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32350         (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32351         * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32352         (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32353         (__builtin_ia32_minss_mask_round): New builtins.
32354         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32355         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32356         * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32357         Rename to ...
32358         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32359         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32360         Change to ...
32361         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32362         ... this.
32364 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32366         * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32367         (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32368         (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32369         (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32370         (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32371         (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32372         (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32373         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32374         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32375         * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32376         (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32377         (__builtin_ia32_mulss_mask_round): New builtins.
32378         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32379         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32380         * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32381         Rename to ...
32382         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32383         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32384         Change to ...
32385         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32386         ... this.
32388 2017-05-10  Julia Koval  <julia.koval@intel.com>
32390         * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32391         (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32392         (_mm256_setr_m128i): New intrinsics.
32394 2017-05-10  Julia Koval  <julia.koval@intel.com>
32396         * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32397         (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32398         (_mm_maskz_rcp14_ss): New intrinsics.
32399         * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32400         (__builtin_ia32_rcp14ss_mask): New builtins.
32401         * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32403 2017-05-10  Peter Bergner  <bergner@vnet.ibm.com>
32405         PR tree-optimization/51513
32406         * tree-cfg.c (gimple_seq_unreachable_p): New function.
32407         (assert_unreachable_fallthru_edge_p): Use it.
32408         (group_case_labels_stmt): Likewise.
32409         * tree-cfg.h: Prototype it.
32410         * stmt.c: Include cfghooks.h and tree-cfg.h.
32411         (emit_case_dispatch_table) <gap_label>: New local variable.
32412         Use it to fill dispatch table gaps.
32413         Test for default_label before updating probabilities.
32414         (expand_case) <default_label>: Remove unneeded initialization.
32415         Test for unreachable default case statement and remove its edge.
32416         Set default_label accordingly.
32417         * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32419 2017-05-10  Carl Love  <cel@us.ibm.com>
32421         * config/rs6000/rs6000-c: Add support for built-in functions
32422         vector signed char      vec_neg (vector signed char)
32423         vector signed short int vec_neg (vector short int)
32424         vector signed int       vec_neg (vector signed int)
32425         vector signed long long vec_neg (vector signed long long)
32426         vector float            vec_neg (vector float)
32427         vector double           vec_neg (vector double)
32428         * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32429         overload.
32430         * config/rs6000/altivec.h: Add define for vec_neg
32431         * doc/extend.texi: Update the built-in documentation for the
32432         new built-in functions.
32434 2017-05-10  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32436         PR tree-optimization/77644
32437         * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32439 2017-05-10  Nathan Sidwell  <nathan@acm.org>
32441         * dumpfile.h (TDI_lang_all): New.
32442         (TDF_KIND): New. Renumber others
32443         (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32444         than bits.
32445         * dumpfile.c (dump_files): Mark language dumps as TDF_LANG.  add
32446         lang-all.
32447         (get_dump_file_name): Adjust suffix generation.
32448         (dump_enable_all): Use TDF_KIND.
32449         * doc/invoke.texi (-fdump-lang-all): Document.
32451         * dumpfile.h: Tabify.
32453 2017-05-10  Wilco Dijkstra  <wdijkstr@arm.com>
32455         PR target/80671
32456         * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32457         Move member access before delete.
32459 2017-05-10  Alexandre Oliva  <aoliva@redhat.com>
32461         * tree-inline.c (expand_call_inline): Split block at stmt
32462         before the call.
32464 2017-05-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
32466         PR target/68163
32467         * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32468         are now unused after splitting mov{sf,sd}_hardfloat.
32469         (f32_lr2): Likewise.
32470         (f32_lm): Likewise.
32471         (f32_lm2): Likewise.
32472         (f32_li): Likewise.
32473         (f32_li2): Likewise.
32474         (f32_lv): Likewise.
32475         (f32_sr): Likewise.
32476         (f32_sr2): Likewise.
32477         (f32_sm): Likewise.
32478         (f32_sm2): Likewise.
32479         (f32_si): Likewise.
32480         (f32_si2): Likewise.
32481         (f32_sv): Likewise.
32482         (f32_dm): Likewise.
32483         (f32_vsx): Likewise.
32484         (f32_av): Likewise.
32485         (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32486         For movsf, order stores so the VSX stores occur before the GPR
32487         store which encourages the register allocator to use a traditional
32488         FPR instead of a GPR.  For movsd, order the stores so that the GPR
32489         store comes before the VSX stores to allow the power6 to work.
32490         This is due to the power6 not having a 32-bit integer store
32491         instruction from a FPR.
32492         (movsf_hardfloat): Likewise.
32493         (movsd_hardfloat): Likewise.
32495 2017-05-09  Martin Sebor  <msebor@redhat.com>
32497         PR translation/80280
32498         * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32499         added in r247778.
32501         PR translation/80280
32502         * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32503         data member added in r247778.
32504         (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32506 2017-05-09  Nathan Sidwell  <nathan@acm.org>
32508         * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32510         * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32511         typedefs.
32513 2017-05-09  Marek Polacek  <polacek@redhat.com>
32515         * doc/invoke.texi: Fix typo.
32517 2017-05-09  Richard Biener  <rguenther@suse.de>
32519         * tree-vrp.c (vrp_val_is_max): Adjust comment.
32520         (vrp_val_is_min): Likewise.
32521         (set_value_range_to_value): Likewise.
32522         (set_value_range_to_nonnegative): Likewise.
32523         (gimple_assign_nonzero_p): Likewise.
32524         (gimple_stmt_nonzero_p): Likewise.
32525         (vrp_int_const_binop): Likewise.  Remove unreachable case.
32526         (adjust_range_with_scev): Adjust comments.
32527         (compare_range_with_value): Likewise.
32528         (extract_range_from_phi_node): Likewise.
32529         (test_for_singularity): Likewise.
32531 2017-05-09  Richard Biener  <rguenther@suse.de>
32533         * tree-vrp.c (get_single_symbol): Add assert that we don't
32534         get overflowed constants as invariant part.
32535         (compare_values_warnv): Add comment before the TREE_NO_WARNING
32536         checks.  Use wi::cmp instead of recursing for integer constants.
32537         (compare_values): Just ignore whether we assumed undefined
32538         overflow instead of failing the compare.
32539         (extract_range_for_var_from_comparison_expr): Add comment before the
32540         TREE_NO_WARNING sets.
32541         (test_for_singularity): Likewise.
32542         (extract_range_from_comparison): Do not disable optimization
32543         when we assumed undefined overflow.
32544         (extract_range_basic): Remove init of unused var.
32546 2017-05-09  Richard Biener  <rguenther@suse.de>
32548         * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32549         (extract_range_from_multiplicative_op_1): Adjust.
32550         (extract_range_from_binary_expr_1): Use int_const_binop.
32552 2017-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32554         PR target/80101
32555         * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32556         rs6000_store_data_bypass_p in seven define_bypass directives and
32557         in several comments.
32558         * config/rs6000/rs6000-protos.h: Add prototype for
32559         rs6000_store_data_bypass_p function.
32560         * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32561         function implements slightly different (rs6000-specific) semantics
32562         than store_data_bypass_p, returning false rather than aborting
32563         with assertion error when arguments do not satisfy the
32564         requirements of store data bypass.
32565         (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32566         rs6000_store_data_bypass_p.
32568 2017-05-08  Max Filippov  <jcmvbkbc@gmail.com>
32570         * config/xtensa/xtensa-protos.h
32571         (xtensa_initial_elimination_offset): New declaration.
32572         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32573         New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32574         macro definition, add case for FRAME_POINTER_REGNUM when
32575         FRAME_GROWS_DOWNWARD.
32576         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32577         (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32578         xtensa_initial_elimination_offset.
32580 2017-05-08  Nathan Sidwell  <nathan@acm.org>
32582         * doc/invoke.texi: Alphabetize -fdump options.
32584 2017-05-08  Martin Sebor  <msebor@redhat.com>
32586         PR translation/80280
32587         * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32589 2017-05-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
32591         * target.def (compute_frame_layout): New optional target hook.
32592         * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32593         * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32594         * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32595         target hook.
32596         * reload1.c (verify_initial_elim_offsets): Likewise.
32597         * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32598         (use_simple_return_p): Call arm_compute_frame_layout if needed.
32599         (arm_get_frame_offsets): Split up into this ...
32600         (arm_compute_frame_layout): ... and this function.
32602 2017-05-08  Richard Sandiford  <richard.sandiford@arm.com>
32604         * config/aarch64/constraints.md (Usa): New constraint.
32605         * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32607 2017-05-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32609         * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32610         with_multilib_list after it has been checked.
32612 2017-05-08  Richard Biener  <rguenther@suse.de>
32614         * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32615         (bitmap_set_subtract_values): Likewise.
32617 2017-05-08  Richard Biener  <rguenther@suse.de>
32619         * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32620         (gimple_assign_nonzero): ... this and remove strict_overflow_p
32621         argument.
32622         (gimple_stmt_nonzero_warnv_p): Rename to ...
32623         (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32624         argument.
32625         (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32626         (extract_range_basic): Adjust, do not disable propagation on
32627         strict overflow sensitive simplification.
32628         (vrp_visit_cond_stmt): Likewise.
32630 2017-05-05  Jan Hubicka  <hubicka@ucw.cz>
32632         * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32633         body size unconditionally.
32635 2017-05-07  Jeff Law  <law@redhat.com>
32637         Revert:
32638         2017-05-06  Jeff Law  <law@redhat.com>
32639         PR tree-optimization/78496
32640         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32641         code.
32643         PR tree-optimization/78496
32644         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32645         (simplify_stmt_using_ranges): Call it.
32646         (vrp_dom_walker::before_dom_children): Extract equivalences
32647         from an ASSERT_EXPR with an equality comparison against a
32648         constant.
32650 2017-05-06  Jeff Law  <law@redhat.com>
32652         PR tree-optimization/78496
32653         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32654         code.
32656         PR tree-optimization/78496
32657         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32658         (simplify_stmt_using_ranges): Call it.
32659         (vrp_dom_walker::before_dom_children): Extract equivalences
32660         from an ASSERT_EXPR with an equality comparison against a
32661         constant.
32663 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32665         * lra-constraints.c (lra_copy_reg_equiv): New function.
32666         (split_reg): Use it to copy equivalence information from the
32667         original register to the spill register.
32669 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32671         PR rtl-optimization/75964
32672         * simplify-rtx.c (simplify_const_relational_operation): Remove
32673         invalid handling of comparisons of integer ABS.
32675 2017-05-06  Uros Bizjak  <ubizjak@gmail.com>
32677         * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32678         initialize to zero.
32679         (init_regs): Remove declaration.
32680         (function_arg_advance_32): Initialize error_p as boolean variable.
32682 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32684         * store-motion.c (remove_reachable_equiv_notes): Reformat long
32685         lines.  Use for (;;).
32687 2017-05-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32689         * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32690         (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32691         (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32692         (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32693         VF=2 that require versioning.
32695 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32697         * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32698         int.
32700 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32702         * diagnostic.h (diagnostic_override_option_index): Convert from
32703         macro to inline function.
32705 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32707         * diagnostic.c (last_module_changed_p): New function.
32708         (set_last_module): New function.
32709         (diagnostic_report_current_module): Convert macro usage to
32710         the above functions.
32711         * diagnostic.h (diagnostic_context::last_module): Strengthen
32712         from const line_map * to const line_map_ordinary *.
32713         (diagnostic_last_module_changed): Delete macro.
32714         (diagnostic_set_last_module): Delete macro.
32716 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32718         * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32719         with diagnostic_report_diagnostic.
32720         (diagnostic_n_impl_richloc): Likewise.
32721         * diagnostic.h (report_diagnostic): Delete macro.
32722         * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32723         with diagnostic_report_diagnostic.
32724         * substring-locations.c (format_warning_va): Likewise.
32726 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32728         * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32729         save/restor of format_spec.  Move option-printing code to...
32730         (print_option_information): ...this new function, and
32731         reimplement by simply printing to the pretty_printer,
32732         rather than appending to the format string.
32734 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32736         * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32737         handling logic into...
32738         (update_effective_level_from_pragmas): ...this new function.
32740 2017-05-04  Andrew Waterman  <andrew@sifive.com>
32742         * config/riscv/riscv.opt (mstrict-align): New option.
32743         * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it.  Update comment.
32744         (SLOW_UNALIGNED_ACCESS): Define.
32745         (riscv_slow_unaligned_access): Declare.
32746         * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32747         field.
32748         (riscv_slow_unaligned_access): New variable.
32749         (rocket_tune_info): Set slow_unaligned_access to true.
32750         (optimize_size_tune_info): Set slow_unaligned_access to false.
32751         (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32752         (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32753         (riscv_option_override): Set riscv_slow_unaligned_access.
32754         * doc/invoke.texi: Add -mstrict-align to RISC-V.
32756 2017-05-04  Kito Cheng  <kito.cheng@gmail.com>
32758         * config/riscv/riscv.md: Unify indentation.
32760 2017-05-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
32762         PR target/79038
32763         PR target/79202
32764         PR target/79203
32765         * config/rs6000/rs6000.md (u code attribute): Add FIX and
32766         UNSIGNED_FIX.
32767         (extendsi<mode>2): Add support for doing sign extension via
32768         VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32769         don't have ISA 3.0 instructions.
32770         (extendsi<mode>2 splitter): Likewise.
32771         (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32772         generate the normal insns since SImode can now go in vector
32773         registers.  Disallow the special UNSPECs needed for previous
32774         machines to hide SImode being used.  Add new insns
32775         fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32776         (fix_trunc<mode>si2_stfiwx): Likewise.
32777         (fix_trunc<mode>si2_internal): Likewise.
32778         (fixuns_trunc<mode>si2): Likewise.
32779         (fixuns_trunc<mode>si2_stfiwx): Likewise.
32780         (fctiw<u>z_<mode>_smallint): Likewise.
32781         (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32782         of floating point to 32-bit integer from doing a direct move to
32783         the GPR registers to do a store.
32784         (fctiwz_<mode>): Break long line.
32786 2017-05-05  Bin Cheng  <bin.cheng@arm.com>
32788         * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32789         * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32790         (addr_list, addr_offset_valid_p): New.
32791         (split_address_groups): Check offset validity with above function.
32792         (gt-tree-ssa-loop-ivopts.h): Include header file.
32794 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32796         * config.gcc (arm*-*-*): Add missing 'fi'.
32798 2017-05-05  Steve Ellcey  <sellcey@cavium.com>
32800         * doc/invoke.texi (-fopt-info): Explicitly say order of options
32801         included in -fopt-info does not matter.
32802         * doc/optinfo.texi (-fopt-info): Fix description of default
32803         behavour. Explicitly say order of options included in -fopt-info
32804         does not matter.
32806 2017-05-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32808         * config.gcc: Allow combinations of aprofile and rmprofile values for
32809         --with-multilib-list.
32810         * config/arm/t-multilib: New file.
32811         * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32812         variables.  Remove setting of ISA and floating-point ABI in
32813         MULTILIB_OPTIONS and MULTILIB_DIRNAMES.  Set architecture and FPU in
32814         MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32815         and MULTILIB_DIRNAMES respectively.  Add comment to introduce all
32816         matches.  Add architecture matches for marvel-pj4 and generic-armv7-a
32817         CPU options.
32818         * config/arm/t-rmprofile: Likewise except for the matches changes.
32819         * doc/install.texi (--with-multilib-list): Document the combination of
32820         aprofile and rmprofile values and warn about pitfalls in doing that.
32822 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32824         * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32825         (movdi_aarch64): Likewise.
32827 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32829         PR tree-optimization/80632
32830         * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32831         field.
32832         (build_arrays): Initialize it for virtual phis.
32833         (fix_phi_nodes): Use it for virtual phis.
32835         PR tree-optimization/80558
32836         * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32837         [x, y] op z into [x op, y op z] for op & or | if conditions
32838         are met.
32840 2017-05-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
32841             Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32843         PR target/71607
32844         * config/arm/arm.md (use_literal_pool): Remove.
32845         (64-bit immediate split): No longer takes cost into consideration
32846         if arm_disable_literal_pool is enabled.
32847         * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32848         used when arm_disable_literal_pool is enabled.
32849         (arm_max_const_double_inline_cost): Remove use of
32850         arm_disable_literal_pool.
32851         (push_minipool_fix): Add assert.
32852         (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32853         * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32854         (no_literal_pool_sf_immediate): New.
32856 2017-05-05  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32858         PR tree-optimization/80613
32859         * tree-ssa-dce.c (propagate_necessity): Remove cases for
32860         BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32862 2017-05-05  Richard Biener  <rguenther@suse.de>
32864         * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32866 2017-05-05  Georg-Johann Lay  <avr@gjlay.de>
32868         * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32869         of this flag from insn conditions due to removal from r247495.
32871 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32873         * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32874         New function.
32875         (arm_early_store_addr_dep_ptr): Likewise.
32876         * config/arm/aarch-common-protos.h
32877         (arm_early_load_addr_dep_ptr): Add prototype.
32878         (arm_early_store_addr_dep_ptr): Likewise.
32879         * config/arm/cortex-a53.md: Add new bypasses.
32881 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32883         * tree.c (next_type_uid): Change type to unsigned.
32884         (type_hash_canon): Decrement back next_type_uid if
32885         freeing a type node with the highest TYPE_UID.  For INTEGER_TYPEs
32886         also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32887         if possible.
32889 2017-05-04  Martin Sebor  <msebor@redhat.com>
32891         * builtins.c: Fix a trivial typo in a comment.
32893         PR middle-end/79234
32894         * builtins.c (check_sizes): Adjust to handle reading past the end.
32895         Avoid printing excessive upper bound of ranges.  Use %E to print
32896         tree nodes instead of converting them to %wu.
32897         (expand_builtin_memchr): New function.
32898         (compute_dest_size): Rename...
32899         (compute_objsize): ...to this.
32900         (expand_builtin_memcpy): Adjust.
32901         (expand_builtin_mempcpy): Adjust.
32902         (expand_builtin_strcat): Adjust.
32903         (expand_builtin_strcpy): Adjust.
32904         (check_strncat_sizes): Adjust.
32905         (expand_builtin_strncat): Adjust.
32906         (expand_builtin_strncpy): Adjust and simplify.
32907         (expand_builtin_memset): Adjust.
32908         (expand_builtin_bzero): Adjust.
32909         (expand_builtin_memcmp): Adjust.
32910         (expand_builtin): Handle memcmp.
32911         (maybe_emit_chk_warning): Check strncat just once.
32913 2017-05-04  Martin Sebor  <msebor@redhat.com>
32915         PR preprocessor/79214
32916         PR middle-end/79222
32917         PR middle-end/79223
32918         * builtins.c (check_sizes): Add inlining context and issue
32919         warnings even when -Wno-system-headers is set.
32920         (check_strncat_sizes): Same.
32921         (expand_builtin_strncat): Same.
32922         (expand_builtin_memmove): New function.
32923         (expand_builtin_stpncpy): Same.
32924         (expand_builtin): Handle memmove and stpncpy.
32926 2017-05-04  Bin Cheng  <bin.cheng@arm.com>
32928         * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
32929         which is not used any more.
32931 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32933         * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
32935 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32937         * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
32938         (cortexa53_tunings): Likewise.
32939         (cortexa57_tunings): Likewise.
32940         (cortexa72_tunings): Likewise.
32941         (cortexa73_tunings): Likewise.
32943 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32945         * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
32946         Set loop alignment to 8.
32948 2017-05-04  Martin Sebor  <msebor@redhat.com>
32950         PR translation/80280
32951         * builtins.c (expand_builtin_object_size): Add missing quoting to
32952         %D and like directives.
32953         * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
32954         (hsa_type_for_tree_type): Same.
32955         (verify_function_arguments): Same.
32956         * symtab.c (symbol_table::change_decl_assembler_name): Same.
32957         * varasm.c (get_section): Same.
32958         (mark_weak): Same.
32960 2017-05-04  Martin Sebor  <msebor@redhat.com>
32962         PR translation/80280
32963         * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
32965 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32967         * config/aarch64/aarch64.c (generic_addrcost_table):
32968         Change HI/TI mode setting.
32970 2017-05-04  Martin Jambor  <mjambor@suse.cz>
32972         PR tree-optimization/80622
32973         * tree-sra.c (comes_initialized_p): New function.
32974         (build_accesses_from_assign): Only set write lazily when
32975         comes_initialized_p is false.
32976         (analyze_access_subtree): Use comes_initialized_p.
32977         (propagate_subaccesses_across_link): Assert !comes_initialized_p
32978         instead of testing for PARM_DECL.
32980 2017-05-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32982         * config/aarch64/aarch64.md (prefetch); Adjust predicate and
32983         constraint on operand 0 to allow more general addressing modes.
32984         Adjust output template.
32985         * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
32986         New function.
32987         * config/aarch64/aarch64-protos.h
32988         (aarch64_address_valid_for_prefetch_p): Declare prototype.
32989         * config/aarch64/constraints.md (Dp): New address constraint.
32990         * config/aarch64/predicates.md (aarch64_prefetch_operand): New
32991         predicate.
32993 2017-05-04  Jan Hubicka  <hubicka@ucw.cz>
32995         * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
32996         update use of estimate_ipcp_clone_size_and_time.
32997         (estimate_local_effects): Update use of
32998         estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
32999         * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
33000         * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
33001         Return nonspecialized time.
33003 2017-05-04  Richard Biener  <rguenther@suse.de>
33005         * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
33006         for the last VUSE which def dominates the PHI.  Directly call
33007         maybe_skip_until.
33008         (get_continuation_for_phi_1): Remove.
33010 2017-05-04  Richard Sandiford  <richard.sandiford@linaro.org>
33012         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
33013         to explain the use of truncating division.  Cap the number of
33014         iterations to the maximum given by nb_iterations_upper_bound,
33015         if defined.
33017 2017-05-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33019         * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
33020         * configure: Regenerate.
33021         * config.in: Regenerate.
33022         * config/i386/driver-mingw32.c: new file.
33023         * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
33024         * config.host: Link driver-mingw32.o on MinGW host.
33025         * doc/install.texi: Document new --enable-mingw-wildcard configure
33026         option.
33028 2017-05-04  Marek Polacek  <polacek@redhat.com>
33030         PR tree-optimization/80612
33031         * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
33033 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
33034             Andre Simoes Dias Vieira  <andre.simoesdiasvieira@arm.com>
33036         * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
33037         (movt splitter): Likewise.
33038         * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
33039         to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
33040         (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
33041         block for Thumb-1 with MOVT.
33042         (thumb2_legitimate_address_p): Move code block ...
33043         (can_avoid_literal_pool_for_label_p): ... into this new function.
33044         (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
33045         literal pool.
33046         (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
33047         * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
33048         "M-profile targets with the MOVT instruction".
33050 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
33052         * config/arm/arm-builtins.c (arm_init_builtins): Rename
33053         __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
33054         __builtin_arm_stfscr to __builtin_arm_set_fpscr.
33056 2017-05-04  Martin Liska  <mliska@suse.cz>
33058         * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
33059         variable cond_code.
33061 2017-05-04  Richard Biener  <rguenther@suse.de>
33063         * tree.c (array_at_struct_end_p): Handle arrays at struct
33064         end with flexarrays more conservatively.  Refactor and treat
33065         arrays of arrays or aggregates more strict.  Fix
33066         VIEW_CONVERT_EXPR handling.  Remove allow_compref argument.
33067         * tree.c (array_at_struct_end_p): Adjust prototype.
33068         * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
33069         * gimple-fold.c (get_range_strlen): Likewise.
33070         * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
33072 2017-05-04  Richard Biener  <rguenther@suse.de>
33074         PR tree-optimization/31130
33075         * tree-vrp.c (needs_overflow_infinity): Remove as always returning
33076         false.
33077         (supports_overflow_infinity): Likewise.
33078         (is_negative_overflow_infinity): Likewise.
33079         (is_positive_overflow_infinity): Likewise.
33080         (is_overflow_infinity): Likewise.
33081         (stmt_overflow_infinity): Likewise.
33082         (overflow_infinity_range_p): Likewise.
33083         (usable_range_p): Remove as always returning true.
33084         (make_overflow_infinity): Remove.
33085         (negative_overflow_infinity): Likewise.
33086         (positive_overflow_infinity): Likewise.
33087         (avoid_overflow_infinity): Likewise.
33088         (set_value_range): Adjust accordingly.
33089         (set_value_range_to_nonnegative): Likewise, remove now unused
33090         overflow_infinity arg.
33091         (vrp_operand_equal_p): Adjust.
33092         (update_value_range): Likewise.
33093         (range_int_cst_singleton_p): Likewise.
33094         (operand_less_p): Likewise.
33095         (compare_values_warnv): Likewise.
33096         (extract_range_for_var_from_comparison_expr): Likewise.
33097         (vrp_int_const_binop): Likewise.
33098         (zero_nonzero_bits_from_vr): Likewise.
33099         (extract_range_from_multiplicative_op_1): Likewise.
33100         (extract_range_from_binary_expr_1): Likewise.
33101         (extract_range_from_unary_expr): Likewise.
33102         (extract_range_from_comparison): Likewise.
33103         (extract_range_basic): Likewise.
33104         (adjust_range_with_scev): Likewise.
33105         (compare_ranges): Likewise.
33106         (compare_range_with_value): Likewise.
33107         (dump_value_range): Likewise.
33108         (test_for_singularity): Likewise, remove strict_overflow_p parameter
33109         never used.
33110         (simplify_cond_using_ranges): Adjust.
33112 2017-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
33114         * brig-builtins.def: Added a builtin for class_f64.
33115         * builtin-types.def: Added a builtin type needed by class_f64.
33117 2017-05-03  Jason Merrill  <jason@redhat.com>
33119         * timevar.def: Add TV_CONSTEXPR.
33121 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33123         * common.opt (fdiagnostics-parseable-fixits): Fix typo.
33125 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33127         * ipa-prop.c (ipa_update_after_lto_read): Removed.
33128         * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
33129         * ipa-cp.c (ipcp_propagate_stage): Do not call
33130         ipa_update_after_lto_read.
33131         * ipa-inline.c (ipa_inline): Likewise.
33133 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33135         * ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
33136         tag.  Added a default constructor and a destructor.
33137         (ipa_edge_args_sum_t): New class;
33138         (ipa_edge_args_sum): Declare.
33139         (ipa_edge_args_vector): Remove declaration.
33140         (IPA_EDGE_REF): Use ipa_edge_args_sum.
33141         (ipa_free_edge_args_substructures): Remove declaration.
33142         (ipa_check_create_edge_args): Use ipa_edge_args_sum.
33143         (ipa_edge_args_info_available_for_edge_p): Likewise.
33144         * ipa-prop.c (ipa_edge_args_vector): Removed.
33145         (edge_removal_hook_holder): Likewise.
33146         (edge_duplication_hook_holder): Likewise.
33147         (ipa_edge_args_sum): New variable.
33148         (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
33149         ipa_edge_args_vector.
33150         (ipa_free_edge_args_substructures): Likewise.
33151         (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
33152         ipa_edge_args_vector.
33153         (ipa_edge_removal_hook): Turned into method
33154         ipa_edge_args_sum_t::remove.
33155         (ipa_edge_duplication_hook): Turned into method
33156         ipa_edge_args_sum_t::duplicate.
33157         (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
33158         registering edge hooks.
33159         (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
33160         * ipa-inline-analysis.c (estimate_function_body_sizes): Test
33161         ipa_edge_args_sum instead of ipa_edge_args_vector.
33162         * ipa-profile.c (ipa_profile): Likewise.
33164 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33166         * symbol-summary.h (function_summary): New method exists.
33167         (function_summary::symtab_removal): Deallocate through release.
33168         (call_summary): New class.
33169         (gt_ggc_mx): New overload.
33170         (gt_pch_nx): Likewise.
33171         (gt_pch_nx): Likewise.
33173 2017-05-03  Jeff Law  <law@redhat.com>
33175         PR tree-optimization/78496
33176         * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
33177         from simplify_cond_using_ranges.  Split off code to walk
33178         backwards through casts into ...
33179         (simplify_cond_using_ranges_2): New function.
33180         (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
33181         (execute_vrp): After identifying jump threads, call
33182         simplify_cond_using_ranges_2.
33184 2017-05-03  Jan Hubicka  <hubicka@ucw.cz>
33186         PR bootstrap/80609
33187         * ipa-inline.h (inline_summary): Add ctor.
33188         (create_ggc): Do not use ggc_cleared_alloc.
33190 2017-05-03  Jeff Downs  <heydowns@somuchpressure.net>
33191             Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33193         * gcc.c (handle_braces): Support escaping in switch matching
33194         text.
33195         * doc/invoke.texi (Spec Files): Document it.
33196         Remove superfluous @code markup in items.
33198 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33200         * diagnostic-show-locus.c (struct column_range): New struct.
33201         (get_affected_columns): New function.
33202         (get_printed_columns): New function.
33203         (struct correction): New struct.
33204         (correction::ensure_capacity): New function.
33205         (correction::ensure_terminated): New function.
33206         (struct line_corrections): New struct.
33207         (line_corrections::~line_corrections): New dtor.
33208         (line_corrections::add_hint): New function.
33209         (layout::print_trailing_fixits): Reimplement in terms of the new
33210         classes.
33211         (selftest::test_overlapped_fixit_printing): New function.
33212         (selftest::diagnostic_show_locus_c_tests): Call it.
33214 2017-05-03  Nathan Sidwell  <nathan@acm.org>
33216         Canonicalize canonical type hashing
33217         * tree.h (type_hash_canon_hash): Declare.
33218         * tree.c (type_hash_list, attribute_hash_list): Move into
33219         type_hash_canon_hash.
33220         (build_type_attribute_qual_variant): Break out hash code calc into
33221         type_hash_canon_hash.
33222         (type_hash_canon_hash): New.  Generic type hash computation.
33223         (build_range_type_1, build_array_type_1, build_function_type,
33224         build_method_type_directly, build_offset_type, build_complex_type,
33225         make_vector_type): Call it.
33227 2017-05-03  Richard Biener  <rguenther@suse.de>
33229         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
33230         When all DRs have unknown misaligned do not always peel
33231         when there is a store but apply the same costing model as if
33232         there were only loads.
33234 2017-05-03  Richard Biener  <rguenther@suse.de>
33236         Revert
33237         PR tree-optimization/80492
33238         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33239         compare_base_decls returning dont-know properly.
33241 2017-05-03  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33243         * config/arm/iterators.md (CCSI): New mode iterator.
33244         (arch): New mode attribute.
33245         * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33246         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33247         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This.  Use CCSI
33248         code iterator for success result mode.
33249         * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33250         the corresponding new insn generators.
33252 2017-05-03  Bin Cheng  <bin.cheng@arm.com>
33254         Revert r247509
33255         2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33256         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33258 2017-05-03  Richard Sandiford  <richard.sandiford@linaro.org>
33260         * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33261         (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33262         (DDR_A): Wrap DDR argument in brackets.
33263         (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33264         (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33265         (DDR_REVERSED_P): Likewise.
33267 2017-05-03  Jakub Jelinek  <jakub@redhat.com>
33269         PR tree-optimization/79472
33270         * tree-switch-conversion.c (struct switch_conv_info): Add
33271         contiguous_range and default_case_nonstandard fields.
33272         (collect_switch_conv_info): Compute contiguous_range and
33273         default_case_nonstandard fields, don't clear final_bb if
33274         contiguous_range and only the default case doesn't have the required
33275         structure.
33276         (check_all_empty_except_final): Set default_case_nonstandard instead
33277         of failing if contiguous_range and the default case doesn't have empty
33278         block.
33279         (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33280         and only the default case doesn't have the required constants.  Skip
33281         virtual phis.
33282         (gather_default_values): Skip virtual phis.  Allow non-NULL CASE_LOW
33283         if default_case_nonstandard.
33284         (build_constructors): Build constant 1 just once.  Assert that default
33285         values aren't inserted in between cases if contiguous_range.  Skip
33286         virtual phis.
33287         (build_arrays): Skip virtual phis.
33288         (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33289         (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33290         Handle virtual phis.
33291         (gen_inbound_check): Handle default_case_nonstandard case.
33292         (process_switch): Adjust check_final_bb caller.  Call
33293         gather_default_values with the first non-default case instead of
33294         default case if default_case_nonstandard.
33296 2017-05-02  Nathan Sidwell  <nathan@acm.org>
33298         * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33299         check.  Fix formatting.
33301 2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
33303         * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33304         errors when comparing specialized and unspecialized times.
33306 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
33308         * diagnostic-show-locus.c
33309         (layout::should_print_annotation_line_p): Make private.
33310         (layout::print_annotation_line): Make private.
33311         (layout::annotation_line_showed_range_p): Make private.
33312         (layout::show_ruler): Make private.
33313         (layout::print_source_line): Make private.  Pass in line and
33314         line_width, rather than calling location_get_source_line.  Drop
33315         returned value.
33316         (layout::print_leading_fixits): New method.
33317         (layout::print_any_fixits): Rename to...
33318         (layout::print_trailing_fixits): ...this, and make private.
33319         Don't print newline fixits.
33320         (diagnostic_show_locus): Move logic for printing one row into...
33321         (layout::print_line): ...this new function.  Move the
33322         location_get_source_line call and error-handling from
33323         print_source_line to here.  Call print_leading_fixits, and rename
33324         print_any_fixits to print_trailing_fixits.
33325         (selftest::test_fixit_insert_containing_newline): Update now that
33326         newlines are partially supported.
33327         (selftest::test_fixit_insert_containing_newline_2): New test.
33328         (selftest::test_fixit_replace_containing_newline): Update comments.
33329         (selftest::diagnostic_show_locus_c_tests): Call the new test.
33330         * edit-context.c (class added_line): New class.
33331         (class edited_line): Describe newline handling in comment.
33332         (edited_line::actually_edited_p): New method.
33333         (edited_line::print_content): Delete redundant decl.
33334         (edited_line::m_predecessors): New field.
33335         (edited_file::print_content): Call edited_line::print_content.
33336         (edited_file::print_diff): Update to support newlines.
33337         (edited_file::print_diff_hunk): Likewise.
33338         (edited_file::print_run_of_changed_lines): New function.
33339         (edited_file::print_diff_line): Convert to...
33340         (print_diff_line): ...this.
33341         (edited_file::get_effective_line_count): New function.
33342         (edited_line::edited_line): Initialize new field m_predecessors.
33343         (edited_line::~edited_line): Clean up m_predecessors.
33344         (edited_line::apply_fixit): Handle newlines.
33345         (edited_line::get_effective_line_count): New function.
33346         (edited_line::print_content): New function.
33347         (edited_line::print_diff_lines): New function.
33348         (selftest::test_applying_fixits_insert_containing_newline): New
33349         test.
33350         (selftest::test_applying_fixits_replace_containing_newline): New
33351         test.
33352         (selftest::insert_line): New function.
33353         (selftest::test_applying_fixits_multiple_lines): Add example of
33354         inserting a line.
33355         (selftest::edit_context_c_tests): Call the new tests.
33357 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33359         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33360         parameter cand.  Update dump information.
33361         (get_computation_cost): Update uses.
33363 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33365         * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33366         (get_computation_aff): Reorder parameters.  Use get_computation_aff_1.
33367         (get_computation_at, rewrite_use_address): Update use of
33368         get_computation_aff.
33370 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33372         * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33373         (get_computation): Delete.
33374         (get_computation_cost): Implement like get_computation_cost_at.
33375         Use get_computation_at.
33376         (get_computation_cost_at): Delete.
33377         (rewrite_use_nonlinear_expr): Use get_computation_at.
33378         (rewrite_use_compare, remove_unused_ivs): Ditto.
33380 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33382         * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33384 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33386         * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33387         (ivopts_global_cost_for_size): Rename parameter and update uses.
33388         (iv_ca_recount_cost): Update uses.
33389         (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33390         candidates seperately in n_invs and n_cands.
33391         (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33393 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33395         * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33396         (find_inv_vars_cb): New.
33397         (find_depends): Renamed to ...
33398         (find_inv_vars): ... this.
33399         (add_candidate_1, force_var_cost): Call find_inv_vars.
33400         (split_address_cost, determine_group_iv_cost_cond): Ditto.
33402 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33404         * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33405         inv_vars.  Add inv_exprs.
33406         (struct iv_cand): Rename depends_on to inv_vars.
33407         (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33408         max_inv_var_id/n_inv_var_uses.  Move max_inv_expr_id around.
33409         Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33410         (dump_cand): Dump inv_vars.
33411         (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33412         (record_invariant, find_depends, add_candidate_1): Ditto.
33413         (set_group_iv_cost, force_var_cost): Ditto.
33414         (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33415         (get_computation_cost_at, get_computation_cost): Ditto.
33416         (determine_group_iv_cost_generic): Ditto.
33417         (determine_group_iv_cost_address): Ditto.
33418         (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33419         (determine_group_iv_costs): Ditto.
33420         (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33421         (iv_ca_set_remove_invariants): Renamed to ...
33422         (iv_ca_set_remove_invs): ... this.  Support inv_vars and inv_exprs.
33423         (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33424         (iv_ca_set_add_invariants):  Renamed to ...
33425         (iv_ca_set_add_invs): ... this.  Support inv_vars and inv_exprs.
33426         (iv_ca_set_cp): Use iv_ca_set_add_invs.
33427         (iv_ca_has_deps): Support inv_vars and inv_exprs.
33428         (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33429         (create_new_ivs): Remove useless dump.
33431 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33433         * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33434         iv_cand code.
33435         (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33436         (iv_ca_set_no_cp, create_new_iv): Ditto.
33438 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33440         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33442 2017-05-02  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33444         * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33445         function tree_check2.
33447 2017-05-02  Martin Liska  <mliska@suse.cz>
33449         * doc/gcov.texi: Add missing preposition.
33450         * gcov.c (function_info::function_info): Properly fill up
33451         all member variables.
33453 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33455         * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33457 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33459         * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33461 2017-05-02  Martin Liska  <mliska@suse.cz>
33463         PR lto/77954.
33464         * lto-streamer-in.c (lto_read_tree_1): Remove
33465         LTO_STREAMER_DEBUG.
33466         * lto-streamer.c (struct tree_hash_entry): Likewise.
33467         (struct tree_entry_hasher): Likewise.
33468         (tree_entry_hasher::hash): Likewise.
33469         (tree_entry_hasher::equal): Likewise.
33470         (lto_streamer_init): Likewise.
33471         (lto_orig_address_map): Likewise.
33472         (lto_orig_address_get): Likewise.
33473         (lto_orig_address_remove): Likewise.
33474         * lto-streamer.h: Likewise.
33475         * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33476         * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33478 2017-05-02  Sebastian Peryt  <sebastian.peryt@intel.com>
33480         * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33481         (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33482         (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33483         (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33484         (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33485         (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33486         (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33487         (mm_maskz_sub_ss): New intrinsics.
33488         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33489         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33490         * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33491         (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33492         (__builtin_ia32_subss_mask_round): New builtins.
33493         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33494         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33495         * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33496         Renamed to ...
33497         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33498         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33499         Changed to ...
33500         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33501         ... this.
33503 2017-05-02  Martin Jambor  <mjambor@suse.cz>
33505         PR tree-optimization/78687
33506         * tree-sra.c (access): New field parent.
33507         (process_subtree_disqualification): New function.
33508         (disqualify_candidate): Call it.
33509         (build_accesses_from_assign): Reset write flag if creating an
33510         assighnment link.
33511         (build_access_subtree): Fill in parent field and also prpagate
33512         down grp_write flag.
33513         (create_artificial_child_access): New parameter set_grp_write, set
33514         grp_write to its value.
33515         (propagate_subaccesses_across_link): Also propagate grp_write flag
33516         values.
33517         (propagate_all_subaccesses): Push the closest parent back to work
33518         queue if add_access_to_work_queue returned true.
33520 2017-05-02  Richard Biener  <rguenther@suse.de>
33522         * common.opt (fstrict-overflow): Alias negative to fwrapv.
33523         * doc/invoke.texi (fstrict-overflow): Remove all traces of
33524         -fstrict-overflow documentation.
33525         * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33526         (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33527         flag_strict_overflow.
33528         * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33529         * lto-opts.c (lto_write_options): Do not stream it.
33530         * lto-wrapper.c (merge_and_complain): Do not handle it.
33531         * opts.c (default_options_table): Do not set -fstrict-overflow.
33532         (finish_options): Likewise do not clear it when sanitizing.
33533         * simplify-rtx.c (simplify_const_relational_operation): Do not
33534         test flag_strict_overflow.
33536 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33538         * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33539         using enabled attribute.
33540         (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33541         (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33542         (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33543         (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33544         (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33545         (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33546         (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33547         (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33548         (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33549         (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33551 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33553         * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33555 2017-05-02  Richard Biener  <rguenther@suse.de>
33557         PR tree-optimization/80591
33558         Revert
33559         2017-04-10  Richard Biener  <rguenther@suse.de>
33561         * tree-ssa-structalias.c (find_func_aliases): Properly handle
33562         asm inputs.
33564 2017-05-02  Richard Biener  <rguenther@suse.de>
33566         PR tree-optimization/80549
33567         * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33568         (cleanup_tree_cfg_noloop): Create forwarders to known loop
33569         headers if they do not have a preheader.
33571 2017-05-02  Martin Liska  <mliska@suse.cz>
33573         PR other/80589
33574         * common.opt: Fix typo.
33575         * doc/invoke.texi: Likewise.
33577 2017-05-01  Jan Beulich  <jbeulich@suse.com>
33579         * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33580         swapping, add (x,x,m,x,n) alternative.
33582 2017-05-01  Nathan Sidwell  <nathan@acm.org>
33584         * calls.c (combine_pending_stack_adjustment_and_call): Remove
33585         unnecessary unadjusted_alignment check.
33587 2017-05-01  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33589         PR c++/80038
33590         * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33591         operations here.
33592         * gimplify.c (gimplify_cilk_detach): New function.
33593         (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33594         * tree-core.h: Document EXPR_CILK_SPAWN.
33595         * tree.h (EXPR_CILK_SPAWN): Define.
33597 2017-05-01  David Malcolm  <dmalcolm@redhat.com>
33599         * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33600         to use new fixit_hint representation, using the "replace" logic.
33601         (get_line_span_for_fixit_hint): Likewise.
33602         (layout::print_any_fixits): Likewise.
33603         (selftest::test_one_liner_many_fixits): Rename to...
33604         (selftest::test_one_liner_many_fixits_1): ...this, and update
33605         comment and expected output to reflect that the multiple fix-it
33606         hints are now consolidated into one insertion.
33607         (selftest::test_one_liner_many_fixits_2): New test.
33608         (selftest::test_diagnostic_show_locus_one_liner): Update for
33609         above.
33610         (selftest::test_fixit_consolidation): Update for fix-it API
33611         change.
33612         * diagnostic.c (print_parseable_fixits): Likewise.
33613         * edit-context.c (edited_line::m_line_events): Convert from
33614         auto_vec <line_event *> to auto_vec <line_event>.
33615         (class line_event): Convert from abstract base class to a concrete
33616         class, taking over the role of replace_event.
33617         (class insert_event): Delete.
33618         (class replace_event): Rename to class line_event.  Convert to
33619         half-open range.
33620         (edit_context::add_fixits): Reimplement.
33621         (edit_context::apply_insert): Delete.
33622         (edit_context::apply_replace): Rename to...
33623         (edit_context::apply_fixit): ...this.  Convert to half-open range.
33624         (edited_file::apply_insert): Delete.
33625         (edited_file::apply_replace): Rename to...
33626         (edited_file::apply_fixit): ...this.
33627         (edited_line::~edited_line): Drop deletion of events.
33628         (edited_line::apply_insert): Delete.
33629         (edited_line::apply_replace): Rename to...
33630         (edited_line::apply_fixit): ...this.  Convert to half-open range.
33631         Update for change to type of m_line_events.
33632         * edit-context.h (edit_context::apply_insert): Delete.
33633         (edit_context::apply_replace): Rename to...
33634         (edit_context::apply_fixit): ...this.
33636 2017-05-01  Martin Sebor  <msebor@redhat.com>
33638         * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33639         known.
33641 2017-05-01  Uros Bizjak  <ubizjak@gmail.com>
33643         PR target/68491
33644         * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33645         __get_cpuid_max returns 0.
33646         (__get_cpuid_count): Ditto.
33648 2017-05-01  Eric Botcazou  <ebotcazou@adacore.com>
33650         * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33651         replacement expression is another instance of one of its arguments.
33653 2017-05-01  Jakub Jelinek  <jakub@redhat.com>
33655         PR target/79430
33656         * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33657         check for stack push/pop autoinc.
33658         * config/i386/i386.c (ix86_agi_dependent): Return false
33659         if the only reason why modified_in_p returned true is that
33660         addr is SP based and set_insn is a push or pop.
33662 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33664         * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33665         overflow check.
33667 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33669         PR ipa/79224
33670         * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33671         (account_size_time): Use two predicates - exec_pred and
33672         nonconst_pred_ptr.
33673         (evaluate_conditions_for_known_args): Compute both clause and
33674         nonspec_clause.
33675         (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33676         (inline_summary_t::duplicate): Update.
33677         (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33678         separately.
33679         (compute_inline_parameters): Likewise.
33680         (estimate_edge_size_and_time): Update caluclation of time.
33681         (estimate_node_size_and_time): Compute both time and nonspecialized
33682         time.
33683         (estimate_ipcp_clone_size_and_time): Update.
33684         (inline_merge_summary): Update.
33685         (do_estimate_edge_time): Update.
33686         (do_estimate_edge_size): Update.
33687         (do_estimate_edge_hints): Update.
33688         (inline_read_section, inline_write_summary): Stream both new predicates.
33689         * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33690         as argument.
33691         (compute_inlined_call_time): Cleanup.
33692         (big_speedup_p): Update.
33693         (edge_badness): Update.
33694         * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33695         (size_time_entry): Replace predicate by exec_predicate and
33696         nonconst_predicate.
33697         (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33698         (estimate_edge_time): Return also nonspec_time.
33699         (reset_edge_growth_cache): Update.
33701 2017-04-29  Jakub Jelinek  <jakub@redhat.com>
33703         PR rtl-optimization/80491
33704         * ifcvt.c (noce_process_if_block): When looking for x setter
33705         with missing else_bb, don't check only the insn right before
33706         cond_earliest, but look for the last insn that x is modified in
33707         within the same bb.
33709         PR rtl-optimization/80491
33710         * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33712 2017-04-29  Marc Glisse  <marc.glisse@inria.fr>
33714         PR tree-optimization/80487
33715         * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33717 2017-04-29  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33719         PR tree-optimization/79697
33720         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33721         is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33722         (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33723         BUILT_IN_STRNDUP.
33724         * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33725         (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33727 2017-04-28  Martin Sebor  <msebor@redhat.com>
33729         PR tree-optimization/80523
33730         * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33731         (init_target_to_host_charmap, target_to_host, target_strtol10): New
33732         functions.
33733         (maybe_warn, format_directive, parse_directive): Use new functions.
33734         (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33736 2017-04-28  Marc Glisse  <marc.glisse@inria.fr>
33738         * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33740 2017-04-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
33742         * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33743         target_header_dir): Set correctly.
33744         * configure: Regenerated.
33745         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33746         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33747         instead of SYSTEM_HEADER_DIR.
33749 2017-04-28  Jan Hubicka  <hubicka@ucw.cz>
33751         * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33752         (estimate_local_effects): Likewise.
33753         * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33754         edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33755         * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33756         estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33757         do_estimate_edge_time, estimate_edge_time): Likewise.
33758         * ipa-inline-analysis.c (estimate_node_size_and_time,
33759         estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33760         (estimate_time_after_inlining): Remove.
33762 2017-04-28  Martin Liska  <mliska@suse.cz>
33764         * doc/gcov.texi: Enhance documentation of gcov.
33766 2017-04-28  Martin Liska  <mliska@suse.cz>
33768         * doc/gcov.texi: Sort options in alphabetic order.
33769         * doc/gcov-dump.texi: Likewise.
33770         * doc/gcov-tool.texi: Likewise.
33771         * gcov.c (print_usage): Likewise.
33772         * gcov-dump.c (print_usage): Likewise.
33773         * gcov-tool.c (print_merge_usage_message): Likewise.
33774         (print_rewrite_usage_message): Likewise.
33775         (print_overlap_usage_message): Likewise.
33777 2017-04-28  Martin Liska  <mliska@suse.cz>
33779         PR gcov-profile/53915
33780         * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33782 2017-04-28  Martin Liska  <mliska@suse.cz>
33784         PR gcov-profile/79891
33785         * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33786         is marked by compiler as living on a line.
33787         (get_cycles_count): Remove usage of the union.
33788         (output_intermediate_file): Likewise.
33789         (find_source): Fix GNU coding style.
33790         (accumulate_line_counts): Remove old non-all block mode.
33791         (output_lines): Remove usage of the union.
33792         * profile.c (output_location): Include all BBs, even if
33793         belonging to a same line (and file) as a previous BB.
33795 2017-04-28  Martin Liska  <mliska@suse.cz>
33797         * gcov.c (process_args): Handle new argument 'w'.
33798         (read_graph_file): Assign ID to BBs.
33799         (output_branch_count): Display BB # if verbose flag is set.
33800         (output_lines): Likewise for arcs.
33801         (print_usage): Add '--verbose' option help.
33802         * doc/gcov.texi: Document --verbose (-w) option.
33804 2017-04-28  Martin Liska  <mliska@suse.cz>
33806         * gcov.c (struct block_location_info): New struct.
33807         (process_file): Fill up the new structure.
33808         (read_graph_file): Replace usage of encoding by the newly added
33809         struct.
33810         (add_line_counts): Likewise.
33811         (accumulate_line_counts): Remove usage of the union.
33812         (function_info::function_info): New function.
33813         (function_info::~function_info): Likewise.
33814         (process_file): Call delete instead of release_function.
33815         (release_function): Release the function.
33816         (release_structures): Call delete instead of release_function.
33817         (solve_flow_graph): Replace usage of num_blocks.
33818         (find_exception_blocks): Likewise.
33819         (output_lines): Fix GNU coding style.
33821 2017-04-28  Martin Liska  <mliska@suse.cz>
33823         PR driver/56469
33824         * coverage.c (coverage_remove_note_file): New function.
33825         * coverage.h: Declare the function.
33826         * toplev.c (finalize): Clean if an error has been seen.
33828 2017-04-28  Martin Liska  <mliska@suse.cz>
33830         PR gcov-profile/80031
33831         * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33832         * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33833         * gcov.c (read_graph_file): Read just number of blocks.
33834         * profile.c (branch_prob): Do not stream 0 flags per a basic
33835         block.
33837 2017-04-28  Martin Liska  <mliska@suse.cz>
33839         * gcov-dump.c (tag_*): Add new argument to declarations.
33840         (dump_gcov_file): Likewise.
33841         (tag_blocks): Add and use new argument depth.
33842         (tag_arcs): Likewise.
33843         (tag_lines): Likewise.
33844         (tag_counters): Likewise.
33845         (tag_summary): Likewise.
33846         (dump_working_sets): Use depth to do a proper indentation.
33848 2017-04-28  Jakub Jelinek  <jakub@redhat.com>
33850         PR bootstrap/80531
33851         * cgraph.h (symtab_node::debug_symtab): No longer inline.
33852         * symtab.c (symtab_node::debug_symtab): Move definition here.
33854 2017-04-28  Richard Biener  <rguenther@suse.de>
33856         * lto-streamer.h (LTO_major_version): Bump to 7.
33858 2017-04-28  Richard Biener  <rguenther@suse.de>
33860         * tree-vrp.c (assert_info): New struct.
33861         (add_assert_info): New helper.
33862         (register_edge_assert_for_2): Refactor to add asserts to a vector
33863         of assert_info.
33864         (register_edge_assert_for_1): Likewise.
33865         (register_edge_assert_for): Likewise.
33866         (finish_register_edge_assert_for): New helper actually registering
33867         asserts where live on edge.
33868         (find_conditional_asserts): Adjust.
33869         (find_switch_asserts): Likewise.
33870         (evrp_dom_walker::try_find_new_range): Generalize.
33871         (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33873 2017-04-27  Marek Polacek  <polacek@redhat.com>
33875         PR sanitizer/80349
33876         * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33877         arg10 and arg11 to itype.
33879 2017-04-27  Jonathan Wakely  <jwakely@redhat.com>
33881         * doc/extend.texi (Object Size Checking): Improve grammar.
33883 2017-04-27  Richard Earnshaw  <rearnsha@arm.com>
33885         PR target/80530
33886         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33887         that the logic for permitting reciprocal estimates matches that
33888         in use_rsqrt_p.
33890 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33892         PR c++/80534
33893         * tree.c (type_cache_hasher::equal): Only compare
33894         TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33895         (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33896         non-aggregate element types.
33897         * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33898         about the flag on ARRAY_TYPEs in the comment, formatting fix.
33900 2017-04-27  Richard Biener  <rguenther@suse.de>
33902         PR middle-end/80533
33903         * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33904         stripping ARRAY_REFs from MEM_EXPR make sure we're not
33905         keeping a reference to a trailing array.
33907 2017-04-27  Richard Biener  <rguenther@suse.de>
33909         PR middle-end/80539
33910         * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33911         being in loop-closed SSA form conservatively.
33912         (chrec_fold_multiply_poly_poly): Likewise.
33914 2017-04-27  Tamar Christina  <tamar.christina@arm.com>
33916         PR middle-end/79665
33917         * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
33918         CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
33920 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33922         PR target/77728
33923         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
33924         (aarch64_function_arg_alignment): Return unsigned int again, but still
33925         ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
33926         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33927         Don't emit -Wpsabi note.
33928         (aarch64_function_arg_boundary): Likewise.
33929         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33930         caller.
33932 2017-04-26  Nathan Sidwell  <nathan@acm.org>
33934         * tree.h (crc32_unsigned_n): Declare.
33935         (crc32_unsigned, crc32_unsigned): Make inline.
33936         * tree.c (crc32_unsigned_bits): Replace with ...
33937         (crc32_unsigned_n): ... this.
33938         (crc32_unsigned, crc32_byte): Remove.
33939         (crc32_string): Remove unnecessary braces.
33941 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33943         * ipa-cp.c (estimate_local_effects): Convert sreal to int.
33944         * ipa-inline-analysis.c (MAX_TIME): Remove.
33945         (account_size_time): Use sreal for time.
33946         (dump_inline_summary): Update.
33947         (estimate_function_body_sizes): Update.
33948         (estimate_edge_size_and_time): Update.
33949         (estimate_calls_size_and_time): Update.
33950         (estimate_node_size_and_time): Update.
33951         (inline_merge_summary): Update.
33952         (inline_update_overall_summary): Update.
33953         (estimate_time_after_inlining): Update.
33954         (inline_read_section): Update.
33955         (inline_write_summary): Update.
33956         * ipa-inline.c (compute_uninlined_call_time): Update.
33957         (compute_inlined_call_time): Update.
33958         (recursive_inlining): Update.
33959         (inline_small_functions): Update.
33960         (dump_overall_stats): Update.
33961         * ipa-inline.h: Include sreal.h.
33962         (size_time_entry): Turn time to sreal.
33963         (inline_summary): Turn self_time nad time to sreal.
33965 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33967         * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
33968         data-streamer.h
33969         (sreal::stream_out, sreal::stream_in): New.
33970         * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
33972 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
33974         * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
33975         environment.
33977 2017-04-25  Uros Bizjak  <ubizjak@gmail.com>
33979         PR target/70799
33980         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
33981         Handle ASHIFTRT.
33982         (dimode_scalar_chain::compute_convert_gain): Ditto.
33983         (dimode_scalar_chain::make_vector_copies): Ditto.
33984         (dimode_scalar_chain::convert_reg): Ditto.
33985         (dimode_scalar_chain::convert_insn): Ditto.
33986         * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
33987         (VI248_AVX512BW_1): New mode iterator.
33988         (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
33989         <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
33990         mode iterator.
33992 2017-04-25  Martin Sebor  <msebor@redhat.com>
33994         PR tree-optimization/80497
33995         * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
33996         constants are representable in HOST_WIDE_INT.
33997         (parse_directive): Ditto.
33999 2017-04-25  Martin Sebor  <msebor@redhat.com>
34001         PR bootstrap/80486
34002         * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
34003         (new_zero_array): Adjust signature.
34004         (dom_info::dom_init): Used unsigned rather that size_t.
34005         (dom_info::dom_info): Same.
34007 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
34008             Jakub Jelinek  <jakub@redhat.com>
34010         PR target/77728
34011         * config/arm/arm.c: Include gimple.h.
34012         (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
34013         returns negative, increment ncrn only if it returned positive.
34014         (arm_needs_doubleword_align): Return int instead of bool,
34015         ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
34016         members, but if there is any such non-FIELD_DECL
34017         > PARM_BOUNDARY aligned decl, return -1 instead of false.
34018         (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
34019         returns negative, increment nregs only if it returned positive.
34020         (arm_setup_incoming_varargs): Likewise.
34021         (arm_function_arg_boundary): Emit -Wpsabi note if
34022         arm_needs_doubleword_align returns negative, return
34023         DOUBLEWORD_ALIGNMENT only if it returned positive.
34025 2017-04-25  Marek Polacek  <polacek@redhat.com>
34027         PR sanitizer/80349
34028         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34029         first argument to type.
34031 2017-04-25  Bill Seurer  <seurer@linux.vnet.ibm.com>
34033         PR target/80482
34034         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
34035         type checks to test for compatibility instead of equality.
34037 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
34038             Jakub Jelinek  <jakub@redhat.com>
34040         PR target/77728
34041         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
34042         type.
34043         (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
34044         struct.  Ignore DECL_ALIGN of decls other than FIELD_DECL for
34045         the alignment computation, but return their maximum in warn_alignment.
34046         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
34047         Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
34048         is smaller.
34049         (aarch64_function_arg_boundary): Likewise.  Simplify using MIN/MAX.
34050         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
34051         caller.
34053 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34055         * config/arc/simdext.md (dmpyh): Fix typo.
34057 2017-04-25  Richard Biener  <rguenther@suse.de>
34059         PR tree-optimization/80492
34060         * alias.c (compare_base_decls): Handle registers with asm
34061         specification conservatively.
34062         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
34063         compare_base_decls returning dont-know properly.
34065 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34067         * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
34068         (legitimate_offset_address_p): New function.
34069         (arc_legitimate_address_p): Use above function.
34071 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34073         * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
34075 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34077         * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
34078         ACCH registers whenever they are available.
34080 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34082         * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
34083         double regs fix when not used.
34085 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34087         * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
34088         core registers.
34089         (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
34090         (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
34092 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34094         * config/arc/arc.c (arc_output_addsi): Check for h-register class
34095         when emitting short ADD instructions.
34097 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34099         * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
34100         constraint.
34101         (cmpsi_cc_c_insn): Likewise.
34102         (cbranchsi4_scratch): Compute proper instruction length using
34103         compact_hreg_operand.
34104         * config/arc/predicates.md (compact_hreg_operand): New predicate.
34106 2017-04-25  Richard Biener  <rguenther@suse.de>
34108         PR middle-end/80509
34109         * passes.c (pass_manager::pass_manager): Initialize
34110         m_name_to_pass_map.
34112 2017-04-25  Richard Biener  <rguenther@suse.de>
34114         PR tree-optimization/79201
34115         * tree-ssa-sink.c (statement_sink_location): Handle calls.
34117 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34119         PR target/80464
34120         * config/s390/vector.md: Split MEM->GPR vector moves for
34121         non-s_operand addresses.
34123 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34125         PR target/79895
34126         * config/s390/predicates.md (reload_const_wide_int_operand): New
34127         predicate.
34128         * config/s390/s390.md ("movti"): Remove d/P alternative.
34129         ("movti_bigconst"): New pattern definition.
34131 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34133         PR target/80080
34134         * s390-protos.h (s390_expand_cs_hqi): Removed.
34135         (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
34136         * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
34137         modes as well as CCZ1mode and CCZmode.
34138         (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
34139         signature of s390_emit_compare_and_swap.
34140         (s390_expand_cs_hqi): Likewise, make static.
34141         (s390_expand_cs_tdsi): Generate an explicit compare before trying
34142         compare-and-swap, in some cases.
34143         (s390_expand_cs): Wrapper function.
34144         (s390_expand_atomic_exchange_tdsi): New backend specific expander for
34145         atomic_exchange.
34146         (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
34147         * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
34148         patterns for small and large integers.  Forbid symref memory operands.
34149         Move expander to s390.c.  Require cc register.
34150         ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
34151         ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
34152         ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
34153         ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
34154         symref memory operands.  Remove CC mode and call s390_match_ccmode
34155         instead.
34156         ("atomic_exchange<mode>"): Allow and implement all integer modes.
34158 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34160         * config/s390/s390.md (define_peephole2): New peephole to help
34161         combining the load-and-test pattern with volatile memory.
34163 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34165         * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
34166         with CCZmode for TARGET_Z196.
34168 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34170         PR rtl-optimization/80501
34171         * combine.c (make_compound_operation_int): Set subreg_code to SET
34172         even for AND with mask of the sign bit of mode.
34174         PR rtl-optimization/80500
34175         * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
34176         sum's initial value.
34178 2017-04-25  Julian Brown  <julian@codesourcery.com>
34179             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34181         * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
34183 2017-04-25  Marc Glisse  <marc.glisse@inria.fr>
34185         * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
34187 2017-04-25  Julian Brown  <julian@codesourcery.com>
34188             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34190         * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
34191         (thunderx2t99_sha): New Reservation.
34193 2017-04-25  Julian Brown  <julian@codesourcery.com>
34194             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34196         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
34197         type for 1-element load.
34199 2017-04-24  Marc Glisse  <marc.glisse@inria.fr>
34201         * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
34203 2017-04-24  Martin Jambor  <mjambor@suse.cz>
34205         PR tree-optimization/80293
34206         * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
34207         char arrays not totally scalarizable if it is false.
34208         (analyze_all_variable_accesses): Pass correct value in the new
34209         parameter.  Add a statistics counter.
34211 2017-04-24  Jan Hubicka  <hubicka@ucw.cz>
34213         PR middle-end/79931
34214         * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
34216 2017-04-24  Richard Biener  <rguenther@suse.de>
34218         PR tree-optimization/80494
34219         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
34220         out for complex types.
34222 2017-04-24  Richard Biener  <rguenther@suse.de>
34224         * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
34225         * tree-ssa-sccvn.c (print_scc): Print SCC size.
34226         (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
34227         (DFS): Adjust and never fail.
34228         (sccvn_dom_walker::fail): Remove.
34229         (sccvn_dom_walker::before_dom_children): Adjust.
34230         (run_scc_vn): Likewise and never fail.
34231         * tree-ssa-pre.c (pass_pre::execute): Adjust.
34232         (pass_fre::execute): Likewise.
34234 2017-04-24  Richard Biener  <rguenther@suse.de>
34236         PR tree-optimization/79725
34237         * tree-ssa-sink.c (statement_sink_location): Return whether
34238         failure reason was zero uses.  Move that check later.
34239         (sink_code_in_bb): Deal with zero uses by removing the stmt
34240         if possible.
34242 2017-04-24  Richard Biener  <rguenther@suse.de>
34244         PR c++/2972
34245         * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34246         pointer-based references.
34248 2017-04-24  Richard Biener  <rguenther@suse.de>
34250         PR bootstrap/79814
34251         * pass_manager.h (pass_manager::operator new): Remove.
34252         (pass_manager::operator delete): Likewise.
34253         * passes.c (pass_manager::operator new): Remove.
34254         (pass_manager::operator delete): Likewise.
34255         (pass_manager::pass_manager): Zero individual pass members.
34257 2017-04-23  Uros Bizjak  <ubizjak@gmail.com>
34259         PR target/70799
34260         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34261         <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34262         Check "XEXP (src, 1)" operand here.
34263         <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34264         Check "XEXP (src, 1)" operand here.
34265         (dimode_scalar_chain::make_vector_copies): Detect count register
34266         of a shift instruction.  Zero extend count register from QImode
34267         to DImode to satisfy vector shift pattern count operand predicate.
34268         Substitute vector shift count operand with a DImode copy.
34269         (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34270         vector register.
34272 2017-04-21  Uros Bizjak  <ubizjak@gmail.com>
34274         * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34275         Remove UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34276         (*insvqi_1_mem_rex64): Move above insv<mode>_1.  Remove
34277         UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34278         (UNSPEC_NOREX_MEM): Remove definition.
34280 2017-04-21  Richard Biener  <rguenther@suse.de>
34282         PR tree-optimization/79547
34283         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34284         Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34285         bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34286         without any constraints.
34288 2017-04-21  Richard Biener  <rguenther@suse.de>
34290         PR tree-optimization/78847
34291         * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34293 2017-04-21  Richard Biener  <rguenther@suse.de>
34295         * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34296         (build_distinct_type_copy): Likewise.
34297         (build_variant_type_copy): Likewise.
34298         * tree.c (build_qualified_type): Pass down mem-stat info.
34299         (build_distinct_type_copy): Likewise.
34300         (build_variant_type_copy): Likewise.
34302 2017-04-21  Richard Biener  <rguenther@suse.de>
34304         PR tree-optimization/80237
34305         * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34306         defaulted to NULL.
34307         (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34308         for a simplified result.
34310 2016-04-21  Richard Biener  <rguenther@suse.de>
34312         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34313         sth as strict as a simple_iv but a chrec without symbols and an
34314         operand defined in the loop we are peeling (and not some subloop).
34315         (propagate_constants_for_unrolling): Propagate all constants.
34317 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34319         PR target/79804
34320         * config/i386/i386.c (print_reg): Remove assert for disalowed
34321         regno values, call output_operand_lossage instead.
34323 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34325         PR target/78090
34326         * config/i386/constraints.md (Yc): New register constraint.
34327         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34328         Use Yc constraint for alternative 2 of operand 0.  Remove
34329         preferred_for_speed attribute.
34331 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34333         * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34334         lastprivate clauses in SIMT case.
34336 2017-04-20  Volker Reichelt  <v.reichelt@netcologne.de>
34338         * doc/invoke.texi (-Wextra-semi): Document new warning option.
34340 2017-04-20  Richard Biener  <rguenther@suse.de>
34342         PR tree-optimization/57796
34343         * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34344         as N scalar stores.
34345         (vect_model_load_cost): Cost gathers as N scalar loads.
34347 2017-04-20  Richard Biener  <rguenther@suse.de>
34349         * ggc-page.c (ggc_allocated_p): Rename to ...
34350         (safe_lookup_page_table_entry): ... this and return the lookup
34351         result.
34352         (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34354 2017-04-20  Richard Biener  <rguenther@suse.de>
34356         PR tree-optimization/80453
34357         * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34358         * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34359         from the conditions.
34360         (vn_phi_eq): Pass them down.
34361         (vn_phi_lookup): Record them.
34362         (vn_phi_insert): Likewise.
34364 2017-04-20  Matthew Fortune  <matthew.fortune@imgtec.com>
34366         * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34367         uninitialized variable warning to avoid buffer overrun.
34369 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34371         PR other/71250
34372         * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34373         is suppressed for '{ 0 }' in C.
34375 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34377         * BASE-VER: Set to 8.0.0.
34379 2017-04-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34381         * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34382         priority .init_array and .fini_array section with SECTION_NOTYPE
34383         flag.
34385 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34387         PR middle-end/80423
34388         * tree.h (build_array_type): Add typeless_storage default argument.
34389         * tree.c (type_cache_hasher::equal): Also compare
34390         TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34391         (build_array_type): Add typeless_storage argument, set
34392         TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34393         recursive call.
34394         (build_nonshared_array_type): Adjust build_array_type_1 caller.
34395         (build_array_type): Likewise.  Add typeless_storage argument.
34397 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34398             Jakub Jelinek  <jakub@redhat.com>
34400         PR tree-optimization/80426
34401         * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34402         operation on symbolic operands, also compute the overflow for the
34403         invariant part when the operation degenerates into a negation.
34405 2017-04-19  Jakub Jelinek  <jakub@redhat.com>
34407         PR debug/80461
34408         * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34409         Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34411         PR debug/80436
34412         * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34414 2017-04-19  Georg-Johann Lay  <avr@gjlay.de>
34416         PR target/80462
34417         * config/avr/avr.c (tree.h): Include it.
34418         (cgraph.h): Include it.
34419         (avr_encode_section_info): Don't warn for uninitialized progmem
34420         variable if it's just an alias.
34422 2017-04-19  Richard Biener  <rguenther@suse.de>
34424         PR ipa/65972
34425         * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34426         when needed by AutoPGO.
34428 2017-04-19  Paulo J. Matos  <paulo@matos-sorge.com>
34430         PR lto/50345
34431         * doc/lto.texi: Remove an extra 'that'.
34433 2017-04-19  Segher Boessenkool  <segher@kernel.crashing.org>
34435         PR rtl-optimization/80429
34436         * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34437         are only used in debug insns.
34439 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34440             Vladimir Makarov  <vmakarov@redhat.com>
34442         * config/sparc/predicates.md (input_operand): Add comment.  Return
34443         true for any memory operand when LRA is in progress.
34444         * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34446 2017-04-18  Jeff Law  <law@redhat.com>
34448         PR target/74563
34449         * mips.md ({return,simple_return}_internal): Do not overwrite
34450         operands[0].
34452 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34454         PR tree-optimization/80443
34455         * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34456         instead of adding 1, subtract -1 and similarly instead of subtracting
34457         1 add -1.
34459 2017-04-18  Richard Sandiford  <richard.sandiford@arm.com>
34461         PR rtl-optimization/80357
34462         * haifa-sched.c (tmp_bitmap): New variable.
34463         (model_recompute): Handle duplicate use records.
34464         (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34465         (free_global_sched_pressure_data): Free it.
34467 2017-04-18  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34469         Revert:
34470         2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34471         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34472         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34473         instead of SYSTEM_HEADER_DIR.
34475 2017-04-18  Jeff Law  <law@redhat.com>
34477         PR middle-end/80422
34478         * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34479         predecessors after walking up the insn chain.
34481 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34483         PR debug/80263
34484         * dwarf2out.c (modified_type_die): Try harder not to emit internal
34485         sizetype type into debug info.
34487 2017-04-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
34489         PR target/80099
34490         * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34491         unneeded test for TARGET_UPPER_REGS_SF.
34492         * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34494 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34496         PR sanitizer/80444
34497         * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34498         instead of gsi_after_labels.
34500 2017-04-18  Jeff Law  <law@redhat.com>
34502         * regcprop.c (maybe_mode_change): Avoid creating copies of the
34503         stack pointer.
34505         Revert:
34506         2017-04-13  Jeff Law  <law@redhat.com>
34507         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34508         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34510 2017-04-18  Georg-Johann Lay  <avr@gjlay.de>
34512         PR target/79453
34513         * config/avr/avr.c (intl.h): Include it.
34514         (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34516 2017-04-18  Martin Liska  <mliska@suse.cz>
34518         PR gcov-profile/78783
34519         * gcov-tool.c (gcov_output_files): Validate that destination
34520         file is either removed by the tool or by a user.
34522 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34523             Guy Benyei  <guybe@mellanox.com>
34525         * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34526         block, and do not negate it, the stored id is already negative.
34528 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34530         * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34532 2017-04-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
34534         PR target/80098
34535         * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34536         masks of options that should be turned off if the VSX vector
34537         options are turned off.
34538         (OTHER_P8_VECTOR_MASKS): Likewise.
34539         (OTHER_VSX_VECTOR_MASKS): Likewise.
34540         * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34541         rs6000_disable_incompatible_switches to validate no type switches
34542         like -mvsx.
34543         (rs6000_incompatible_switch): New function to disallow turning on
34544         other vector options if -mno-vsx, -mno-power8-vector, or
34545         -mno-power9-vector are specified.
34547 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34549         * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34551 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34553         * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34554         * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34555         * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34556         (ARG_POINTER_CFA_OFFSET): Likewise.
34558 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34560         * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34561         conditions to take advantage of various optimizations.
34563 2017-04-13  Jeff Law  <law@redhat.com>
34565         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34566         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34567         (zero_extendsidi2_dext): Likewise.
34569 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34571         PR sanitizer/80403
34572         * fold-const.c (fold_ternary_loc): Revert
34573         use op0 instead of fold_convert_loc (loc, type, arg0) part of
34574         2017-04-12 change.
34576 2017-04-13  Vladimir Makarov  <vmakarov@redhat.com>
34578         PR rtl-optimization/80343
34579         * lra-remat.c (update_scratch_ops): Assign original hard reg to
34580         new scratch pseudo.
34582 2017-04-13  Denis Khalikov <d.khalikov@partner.samsung.com>
34584         PR sanitizer/80414
34585         * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34586         to ubsan_encode_value.
34588 2017-04-13  Jeff Law  <law@redhat.com>
34590         * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34591         appearing in DEBUG_INSNs.
34593 2017-04-13  Martin Liska  <mliska@suse.cz>
34595         PR gcov-profile/80413
34596         * gcov-io.c (gcov_write_string): Copy to buffer just when
34597         allocated size is greater than zero.
34599 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34601         PR debug/80321
34602         * dwarf2out.c (decls_for_scope): Ignore declarations of
34603         current_function_decl in BLOCK_NONLOCALIZED_VARS.
34605 2017-04-12  Jan Hubicka  <hubicka@ucw.cz>
34607         PR lto/69953
34608         * ipa-visibility.c (non_local_p): Fix typos.
34609         (localize_node): When localizing symbol in same comdat group,
34610         dissolve the group only when we know external symbols are going
34611         to be privatized.
34612         (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34614 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34616         PR tree-optimization/79390
34617         * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34618         order does not result in usable sequence, retry with reversed operand
34619         order.
34621         PR sanitizer/80403
34622         PR sanitizer/80404
34623         PR sanitizer/80405
34624         * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34625         to fold_build2_loc.  Convert TREE_OPERAND (tem, 0) to type.  Use
34626         op0 instead of fold_convert_loc (loc, type, arg0).
34628 2017-04-12  Jeff Law  <law@redhat.com>
34630         * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34631         has a delay slot in the generated code.
34633         * config/cris/cris.md (cris_preferred_reload_class): Return
34634         GENNONACR_REGS rather than GENERAL_REGS.
34636 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34638         PR c/80163
34639         * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34640         vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34641         signedness of the result type.
34643 2017-04-12  Richard Biener  <rguenther@suse.de>
34644             Jeff Law  <law@redhat.com>
34646         PR tree-optimization/80359
34647         * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34648         trim stores to TARGET_MEM_REFs.
34650 2017-04-12  Richard Biener  <rguenther@suse.de>
34652         PR tree-optimization/79390
34653         * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34654         threading case even more.
34656 2017-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
34658         PR target/80382
34659         * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34660         for quad_address_p for TImode, instead of just not indexed_address.
34662 2017-04-12  Richard Biener  <rguenther@suse.de>
34663             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34665         PR middle-end/79671
34666         * alias.c (component_uses_parent_alias_set_from): Handle
34667         TYPE_TYPELESS_STORAGE.
34668         (get_alias_set): Likewise.
34669         * tree-core.h (tree_type_common): Add typeless_storage flag.
34670         * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34671         * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34672         for types containing members with TYPE_TYPELESS_STORAGE.
34673         (place_field): Likewise.
34674         (layout_type): Likewise for ARRAY_TYPE.
34675         * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34676         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34677         TYPE_TYPELESS_STORAGE.
34678         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34680 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34682         PR sanitizer/80349
34683         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34684         first argument to type.
34686 2017-04-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34688         PR target/80376
34689         PR target/80315
34690         * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34691         CONST0_RTX (mode) rather than const0_rtx where appropriate.
34692         (rs6000_expand_binop_builtin): Likewise.
34693         (rs6000_expand_ternop_builtin): Likewise; also add missing
34694         vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34695         vshasigma built-ins.
34696         * doc/extend.texi: Document that vec_xxpermdi's third argument
34697         must be a constant.
34699 2017-04-11  Uros Bizjak  <ubizjak@gmail.com>
34701         * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34702         Use shift_const cost parameter when calculating gain of STV shifts.
34704 2017-04-11  Vladimir Makarov  <vmakarov@redhat.com>
34706         PR rtl-optimization/70478
34707         * lra-constraints.c (process_alt_operands): Check memory for
34708         disfavoring memory insn operand.
34710 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34712         PR middle-end/80100
34713         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34714         left shift in unsigned HOST_WIDE_INT type.
34716         PR rtl-optimization/80385
34717         * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34718         (not (neg X)) into (plus X -1) for complex or non-integral modes.
34720         PR libgomp/80394
34721         * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34722         if they have any depend clauses.
34724 2017-04-11  Martin Liska  <mliska@suse.cz>
34726         PR ipa/80212
34727         * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34728         * ipa-split.c (split_function): Create a local comdat symbol
34729         if caller is in a comdat group.
34731 2017-04-11  Martin Liska  <mliska@suse.cz>
34733         PR ipa/80212
34734         * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34735         flags.
34737 2017-04-11  Martin Sebor  <msebor@redhat.com>
34739         PR middle-end/80364
34740         * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34741         always use the int type.  Use INTEGRAL_TYPE_P() rather than testing
34742         for INTEGER_TYPE.
34743         (directive::set_width, directive::set_precision, format_character):
34744         Adjust.
34745         (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34746         INTEGER_TYPE.
34748 2017-04-11  Richard Earnshaw  <rearnsha@arm.com>
34750         PR target/80389
34751         * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34752         conflict, set target->arch_name instead of target->cpu_name.
34754 2017-04-11  Richard Biener  <rguenther@suse.de>
34756         PR tree-optimization/80374
34757         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34758         build_zero_cst, remove fold_convertible_p check again.
34760 2017-04-11  Martin Liska  <mliska@suse.cz>
34762         PR sanitizer/70878
34763         * ubsan.c (instrument_object_size): Do not instrument register
34764         variables.
34766 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34768         PR target/80381
34769         * config/i386/i386-builtin-types.def
34770         (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34771         V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34772         V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34773         V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34774         V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34775         V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34776         V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34777         V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34778         V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34779         V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34780         V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34781         V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34782         V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34783         V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34784         V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34785         V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34786         V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34787         V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34788         * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34789         __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34790         __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34791         __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34792         __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34793         __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34794         __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34795         __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34796         __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34797         __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34798         __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34799         __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34800         __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34801         __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34802         __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34803         __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34804         __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34805         __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34806         __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34807         __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34808         __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34809         __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34810         __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34811         __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34812         __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34813         __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34814         __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34815         __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34816         aliases.
34817         * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34818         flag to second_arg_count, handle 4 argument function type _COUNT
34819         aliases, handle second_arg_count on second argument rather than last.
34821 2017-04-10  Jeff Law  <law@redhat.com>
34823         PR tree-optimization/80374
34824         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34825         record anything if we can not convert integer_zero_node to the
34826         desired type.
34828 2017-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
34830         PR target/80108
34831         * config/rs6000/rs6000.c (rs6000_option_override_internal):
34832         Enhance special handling given to the TARGET_P9_MINMAX option in
34833         relation to certain other options.
34835 2017-04-10  Bin Cheng  <bin.cheng@arm.com>
34837         PR tree-optimization/80153
34838         * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34839         remove POINTER_PLUS_EXPR's base part directly, rather than through
34840         aff_tree.
34842 2017-04-10  Richard Biener  <rguenther@suse.de>
34843             Bin Cheng  <bin.cheng@arm.com>
34845         PR tree-optimization/80153
34846         * tree-affine.c (aff_combination_to_tree): Get base pointer from
34847         the first element of pointer type aff_tree.  Build result expr in
34848         aff_tree's type.
34849         (add_elt_to_tree): Convert to type unconditionally.  Remove other
34850         fold_convert calls.
34851         * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34852         (rewrite_use_nonlinear_expr): Check invariant using iv information.
34854 2017-04-10  Richard Biener  <rguenther@suse.de>
34856         * tree-ssa-structalias.c (find_func_aliases): Properly handle
34857         asm inputs.
34859 2017-04-10  Vladimir Makarov  <vmakarov@redhat.com>
34861         PR rtl-optimization/70478
34862         * lra-constraints.c (curr_small_class_check): New.
34863         (update_and_check_small_class_inputs): New.
34864         (process_alt_operands): Update curr_small_class_check.  Disfavor
34865         alternative insn memory operands.  Check available regs for small
34866         class operands.
34868 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
34870         PR target/80057
34871         * config/mips/mips.opt (-mvirt): Update description.
34872         * doc/invoke.texi (-mvirt): Likewise.
34874 2017-04-10  Richard Biener  <rguenther@suse.de>
34876         PR middle-end/80362
34877         * fold-const.c (fold_binary_loc): Look at unstripped ops when
34878         looking for NEGATE_EXPR in -A / -B to A / B folding.
34880 2017-04-10  Martin Liska  <mliska@suse.cz>
34882         PR gcov-profile/80224
34883         * gcov.c (print_usage): Fix usage string.
34884         (get_gcov_intermediate_filename): Remove.
34885         (output_gcov_file): Use both for normal and intermediate format.
34886         (generate_results): Do not initialize special file for
34887         intermediate format.
34889 2017-04-10  Richard Biener  <rguenther@suse.de>
34891         PR tree-optimization/80304
34892         * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34893         for safelen.
34895 2017-04-10  Nathan Sidwell  <nathan@acm.org>
34897         PR target/79905
34898         * config/rs6000/rs6000.c (rs6000_vector_type): New.
34899         (rs6000_init_builtins): Use it.
34901 2016-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34903         * config/arm/arm.md (<mrc>): Add mode to SET source.
34904         (<mrrc>): Likewise.
34906 2017-04-10  Richard Biener  <rguenther@suse.de>
34908         PR middle-end/80344
34909         * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34911 2017-04-10  Jakub Jelinek  <jakub@redhat.com>
34913         PR target/80324
34914         * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34915         _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34916         _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34917         _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
34918         _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
34919         _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
34920         _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
34921         _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
34922         _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
34923         _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
34924         _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
34925         _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
34926         _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
34927         _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
34928         _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
34929         _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
34930         _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
34931         _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
34932         _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
34933         _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
34934         _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
34935         _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
34936         _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
34938 2017-04-08  Vladimir Makarov  <vmakarov@redhat.com>
34940         PR rtl-optimization/70478
34941         * lra-constraints.c: Reverse the last patch.
34943 2017-04-08  Andreas Tobler  <andreast@gcc.gnu.org>
34945         * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
34946         Add comment for WCHAR_T.
34948 2017-04-08  Martin Liska  <mliska@suse.cz>
34950         Revert:
34951         2017-04-07  Martin Liska  <mliska@suse.cz>
34953         PR ipa/80212
34954         * ipa-split.c (split_function): Add function part to a same comdat
34955         group.
34957 2017-04-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34959         PR target/80358
34960         * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
34962 2017-04-07  Pat Haugen  <pthaugen@us.ibm.com>
34964         * rs6000/rs6000.c (vec_load_pendulum): Rename...
34965         (vec_pairing): ...to this.
34966         (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
34967         (rs6000_sched_init): Adjust for name change.
34968         (struct rs6000_sched_context): Likewise.
34969         (rs6000_init_sched_context): Likewise.
34970         (rs6000_set_sched_context): Likewise.
34972 2017-04-07  Jakub Jelinek  <jakub@redhat.com>
34974         PR target/80322
34975         PR target/80323
34976         PR target/80325
34977         PR target/80326
34978         * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
34979         intrinsics.
34980         * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
34981         _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
34982         _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
34984 2017-04-07  Andreas Tobler  <andreast@gcc.gnu.org>
34986         * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
34988 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34990         PR rtl-optimization/70703
34991         * ira-color.c (update_conflict_hard_regno_costs): Use
34992         int64_t instead of HOST_WIDE_INT.
34994 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34996         PR rtl-optimization/70478
34997         * lra-constraints.c (process_alt_operands): Disfavor alternative
34998         insn memory operands.
35000 2017-04-07  Jeff Law  <law@redhat.com>
35002         * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
35003         CALL and NOTE_INSN_CALL_ARG_LOCATION.
35005 2017-04-07  Martin Liska  <mliska@suse.cz>
35007         PR target/79889
35008         * config/aarch64/aarch64.c (aarch64_process_target_attr):
35009         Show error message instead of an ICE.
35011 2017-04-07  Martin Liska  <mliska@suse.cz>
35013         PR ipa/80212
35014         * ipa-split.c (split_function): Add function part to a same comdat
35015         group.
35017 2017-04-07  Richard Biener  <rguenther@suse.de>
35019         PR middle-end/80341
35020         * tree.c (get_unwidened): Also handle ! for_type case for
35021         INTEGER_CSTs.
35022         * convert.c (do_narrow): Split out from ...
35023         (convert_to_integer_1): ... here.  Do not pass final truncation
35024         type to get_unwidened for TRUNC_DIV_EXPR.
35026 2017-04-07  Richard Biener  <rguenther@suse.de>
35028         * tree-affine.c (wide_int_ext_for_comb): Take type rather
35029         than aff_tree.
35030         (aff_combination_const): Adjust.
35031         (aff_combination_scale): Likewise.
35032         (aff_combination_add_elt): Likewise.
35033         (aff_combination_add_cst): Likewise.
35034         (aff_combination_convert): Likewise.
35035         (add_elt_to_tree): Likewise.  Remove unused argument.
35036         (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
35038 2017-04-07  Sebastian Huber  <sebastian.huber@embedded-brains.de>
35040         * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
35041         definition.
35042         * config/arm/arm.c (arm_default_short_enums): Use
35043         ARM_DEFAULT_SHORT_ENUMS.
35044         * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
35046 2017-04-06  Jakub Jelinek  <jakub@redhat.com>
35048         PR debug/80234
35049         * dwarf2out.c (gen_member_die): Handle C++17 inline static data
35050         members with redundant out-of-class redeclaration.
35052 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
35054         PR target/80286
35055         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
35056         * config/i386/i386.md (*zero_extendsidi2):
35057         Add (?*x,*x) and (?*v,*v) alternatives.
35059 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
35061         PR target/79733
35062         * config/i386/i386.c (ix86_expand_builtin)
35063         <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
35064         mode from insn data. Convert operands to insn operand mode.
35065         Copy operands that don't satisfy insn predicate to a register.
35067 2017-04-06  Sam Thursfield  <sam.thursfield@codethink.co.uk>
35069         * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
35070         Update comments.
35072 2017-04-06  Richard Biener  <rguenther@suse.de>
35074         PR tree-optimization/80334
35075         * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
35076         preserve alignment of accesses.
35078 2017-04-06  Richard Biener  <rguenther@suse.de>
35080         PR tree-optimization/80262
35081         * tree-sra.c (build_ref_for_offset): Preserve address-space
35082         information.
35083         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
35084         Drop useless address-space information on MEM_REF offsets.
35086 2017-04-05  Andreas Schwab  <schwab@linux-m68k.org>
35088         * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
35090 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35092         PR rtl-optimization/70703
35093         * ira-color.c (update_conflict_hard_regno_costs): Use
35094         HOST_WIDE_INT instead of long.
35096 2017-04-05  Uros Bizjak  <ubizjak@gmail.com>
35098         PR target/80298
35099         * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
35100         not defined for x86_64 target.  Add -mmmx target option when __SSE2__
35101         is not defined.
35102         * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
35103         for x86_64 target.  Handle -m3dnowa option.
35105 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35107         PR rtl-optimization/70703
35108         * ira-color.c (update_costs_from_allocno): Use the smallest mode.
35109         (update_conflict_hard_regno_costs): Use long instead of unsigned
35110         arithmetic for cost calculation.
35112 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35113             Bernd Edlinger  <bernd.edlinger@hotmail.de>
35115         PR sanitizer/80308
35116         * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
35117         for big endian.
35119 2017-04-05  Eric Botcazou  <ebotcazou@adacore.com>
35121         PR target/78002
35122         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
35123         ptr_mode with Pmode throughout.
35124         * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
35125         into probe_stack_range and use DImode.
35127 2017-04-05  Dominik Vogt  <vogt@linux.vnet.ibm.com>
35129         PR target/79890
35130         * config/s390/s390.c (s390_register_info_gprtofpr): Return if
35131         call_eh_return is true.
35133 2017-04-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35135         * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
35136         Initialize last_match_fntype_index.
35138 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35140         PR target/80310
35141         * tree-nvr.c: Include internal-fn.h.
35142         (pass_return_slot::execute): Ignore internal calls without
35143         direct optab.
35145 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35146             Richard Biener  <rguenther@suse.de>
35148         PR c++/80297
35149         * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
35150         captures used multiple times, except for the last use.
35151         * generic-match-head.c: Include gimplify.h.
35153 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35155         PR tree-optimization/79390
35156         * target.h (struct noce_if_info): Declare.
35157         * targhooks.h (default_noce_conversion_profitable_p): Declare.
35158         * target.def (noce_conversion_profitable_p): New target hook.
35159         * ifcvt.h (struct noce_if_info): New type, moved from ...
35160         * ifcvt.c (struct noce_if_info): ... here.
35161         (noce_conversion_profitable_p): Renamed to ...
35162         (default_noce_conversion_profitable_p): ... this.  No longer
35163         static nor inline.
35164         (noce_try_store_flag_constants, noce_try_addcc,
35165         noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
35166         noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
35167         instead of noce_conversion_profitable_p.
35168         * config/i386/i386.c: Include ifcvt.h.
35169         (ix86_option_override_internal): Don't override
35170         PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
35171         (ix86_noce_conversion_profitable_p): New function.
35172         (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
35173         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
35174         * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
35175         * doc/tm.texi: Regenerated.
35177 2017-04-04  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35179         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
35180         correction.
35182 2017-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35184         PR target/80307
35185         * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
35186         instructions for small multiply cores.
35188 2017-04-04  Jeff Law  <law@redhat.com>
35190         * config/mips/mips.c (mips_multi_add): Zero initialize the newly
35191         added member.
35192         (mips_expand_vec_perm_const): Initialize elements in orig_perm
35193         that are not set by the loop over the elements.
35195 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35197         PR target/80286
35198         * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
35199         int mode, convert_modes it to mode as unsigned, otherwise use
35200         lowpart_subreg to mode rather than SImode.
35201         * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
35202         ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
35203         Use DImode instead of SImode for the shift count operand.
35204         * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
35205         Likewise.
35207 2017-04-04  Richard Biener  <rguenther@suse.de>
35209         PR middle-end/80281
35210         * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
35211         arithmetic done for the negate or the plus.  Simplify.
35212         (A - (-B) -> A + B): Likewise.
35213         * fold-const.c (split_tree): Make sure to not negate pointers.
35215 2017-04-04  Segher Boessenkool  <segher@kernel.crashing.org>
35217         PR rtl-optimization/60818
35218         * simplify-rtx.c (simplify_binary_operation_1): Do not replace
35219         a compare of comparisons with the thing compared if this results
35220         in a different machine mode.
35222 2017-04-03  Jonathan Wakely  <jwakely@redhat.com>
35224         * alias.c (base_alias_check): Fix typo in comment.
35225         * cgraph.h (class ipa_polymorphic_call_context): Likewise.
35226         * cgraphunit.c (symbol_table::compile): Likewise.
35227         * collect2.c (maybe_run_lto_and_relink): Likewise.
35228         * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
35229         * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
35230         * config/avr/avr.c (avr_map_op_t): Likewise.
35231         * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
35232         * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
35233         * config/epiphany/epiphany.md (movcc): Likewise.
35234         * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
35235         * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
35236         Likewise.
35237         * config/mips/mips.c (mips_save_restore_reg): Likewise.
35238         * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
35239         * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
35240         * config/sh/sh.c (sh_rtx_costs): Likewise.
35241         * fold-const.c (fold_truth_andor): Likewise.
35242         * genautomata.c (collapse_flag): Likewise.
35243         * gengtype.h (struct type::u::s): Likewise.
35244         * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35245         * input.c (FORMAT_AMOUNT): Likewise.
35246         * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35247         (known_aggs_to_agg_replacement_list): Likewise.
35248         * ipa-inline-analysis.c: Likewise.
35249         * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35250         * ipa-polymorphic-call.c
35251         (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35252         * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35253         * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35254         Likewise.
35255         * modulo-sched.c (apply_reg_moves): Likewise.
35256         * omp-expand.c (build_omp_regions_1): Likewise.
35257         * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35258         * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35259         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35260         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35261         * value-prof.c: Likewise.
35262         * var-tracking.c (val_reset): Likewise.
35264 2017-04-03  Richard Biener  <rguenther@suse.de>
35266         PR tree-optimization/80275
35267         * fold-const.c (split_address_to_core_and_offset): Handle
35268         POINTER_PLUS_EXPR.
35270 2017-04-03  Eric Botcazou  <ebotcazou@adacore.com>
35272         * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35273         descriptors is at least equal to that of functions.
35275 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35277         * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35279 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35281         PR target/80250
35282         * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35283         (mov<IMOD4:mode>): New expander.
35284         (*mov<IMOD4:mode>_internal): New insn and split pattern.
35286 2017-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
35288         PR rtl-optimization/79405
35289         * fwprop.c (propagations_left): New variable.
35290         (forward_propagate_into): Decrement it.
35291         (fwprop_init): Initialize it.
35292         (fw_prop): If the variable has reached zero, stop propagating.
35293         (fwprop_addr): Ditto.
35295 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35297         PR debug/79255
35298         * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35299         a FUNCTION_DECL, pass it as decl instead of origin to
35300         process_scope_var.
35302 2017-03-31  Alexander Monakov  <amonakov@ispras.ru>
35304         * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35305         string.
35307 2017-03-31  Pat Haugen  <pthaugen@us.ibm.com>
35309         PR target/80107
35310         * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35311         TARGET_VSX_SMALL_INTEGER.
35313 2017-03-31  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35315         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35316         reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35318 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
35320         * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35321         extraction from odd-numbered MSA register.
35323 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35325         PR middle-end/80173
35326         * expmed.c (store_bit_field_1): Don't attempt to create
35327         a word subreg out of hard registers wider than word if they
35328         have HARD_REGNO_NREGS of 1 for their mode.
35330         PR middle-end/80163
35331         * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35332         conversions to integer types wider than word and pointer.
35334         PR debug/80025
35335         * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35336         (rtx_equal_for_cselib_p): Pass 0 to it.
35337         * cselib.c (cselib_hasher::equal): Likewise.
35338         (rtx_equal_for_cselib_1): Add depth argument.  If depth
35339         is 128, don't look up VALUE locs and punt.  Increment
35340         depth in recursive calls when walking VALUE locs.
35342 2017-03-31  Bernd Edlinger  <bernd.edlinger@hotmail.de>
35344         * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35345         (make_gcov_file_name): Use the canonical path name for generating
35346         the MD5 value.
35347         (read_line): Fix handling of files with ascii null bytes.
35349 2017-03-30  Matthew Fortune  <matthew.fortune@imgtec.com>
35351         * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35352         to initialise a vector register instead
35353         of using a const_int.
35355 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35357         PR translation/80189
35358         * gimplify.c (omp_default_clause): Use %qs instead of %s in
35359         diagnostic messages.
35361 2017-03-30  Peter Bergner  <bergner@vnet.ibm.com>
35363         PR target/80246
35364         * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35365         (dfp_diex_<mode>): Update mode of operand 1.
35366         * doc/extend.texi (dxex, dxexq): Document change to return type.
35367         (diex, diexq): Document change to argument type.
35369 2017-03-30  Martin Jambor  <mjambor@suse.cz>
35371         PR ipa/77333
35372         * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35373         * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35374         it reflects the signature changes performed at the callee side.
35375         * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35376         to cgraph_build_function_type_skip_args.
35377         (build_function_decl_skip_args): Adjust call to the above function.
35379 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35381         PR target/80206
35382         * config/i386/sse.md
35383         (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35384         register as dest whenever it is a MEM not rtx_equal_p to the
35385         corresponding dup operand, and when forcing into reg move the
35386         reg into the memory afterwards.
35387         (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35388         Likewise.  Use <ssehalfvecmode> instead of <ssequartermode>
35389         for the force_reg mode.
35390         (avx512vl_vextractf128<mode>): Use register as dest either
35391         always when a MEM, or when it is a MEM not rtx_equal_p to the
35392         corresponding dup operand, or even not when it is a CONST_VECTOR
35393         depending on the mode and lo vs. hi.
35394         (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35395         parens.
35396         (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35397         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35398         Likewise.  Require that operands[2] is even.
35399         (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35400         Remove extraneous parens.  Require that operands[2] is a multiple
35401         of 4.
35402         (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35403         operands[0] is a MEM if <mask_applied>, the predicates/constraints
35404         disallow memory then.
35406 2017-03-30  Richard Biener  <rguenther@suse.de>
35408         PR tree-optimization/77498
35409         * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35410         to non-constants over backedges.
35412 2017-03-29  Segher Boessenkool  <segher@kernel.crashing.org>
35414         PR rtl-optimization/80233
35415         * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35416         as last_combined_insn.  Do not test for BARRIER_P separately.
35418 2017-03-29  Andreas Schwab  <schwab@suse.de>
35420         PR ada/80146
35421         * calls.c (prepare_call_address): Convert funexp to Pmode before
35422         copying to temp reg.
35424 2017-03-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35426         PR tree-optimization/80158
35427         * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35428         Handle possible future case of more than one alternate
35429         interpretation.
35430         (replace_rhs_if_not_dup): Likewise.
35431         (replace_one_candidate): Likewise.
35433 2017-03-28  Vladimir Makarov  <vmakarov@redhat.com>
35435         PR rtl-optimization/80193
35436         * ira.c (ira): Do not check allocation for LRA.
35438 2017-03-28  Alexander Monakov  <amonakov@ispras.ru>
35440         * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35441         (nvptx_output_simt_exit): Declare.
35442         * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35443         cfun->machine->unisimt_location.  Handle NULL unisimt_predicate.
35444         (init_softstack_frame): Move initialization of crtl->is_leaf to...
35445         (nvptx_declare_function_name): ...here.  Emit declaration of local
35446         memory space buffer for omp_simt_enter insn.
35447         (nvptx_output_unisimt_switch): New.
35448         (nvptx_output_softstack_switch): New.
35449         (nvptx_output_simt_enter): New.
35450         (nvptx_output_simt_exit): New.
35451         * config/nvptx/nvptx.h (struct machine_function): New fields
35452         has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35453         * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35454         (UNSPECV_SIMT_EXIT): Ditto.
35455         (omp_simt_enter_insn): New insn.
35456         (omp_simt_enter): New expansion.
35457         (omp_simt_exit): New insn.
35458         * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35460         * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35461         (expand_GOMP_SIMT_ENTER_ALLOC): New.
35462         (expand_GOMP_SIMT_EXIT): New.
35463         * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35464         (GOMP_SIMT_ENTER_ALLOC): Ditto.
35465         (GOMP_SIMT_EXIT): Ditto.
35466         * target-insns.def (omp_simt_enter): New insn.
35467         (omp_simt_exit): Ditto.
35468         * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35469         simt_dlist.
35470         (lower_rec_simd_input_clauses): Implement SIMT privatization.
35471         (lower_rec_input_clauses): Likewise.
35472         (lower_lastprivate_clauses): Handle SIMT privatization.
35474         * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35475         (ompdevlow_adjust_simt_enter): New.
35476         (find_simtpriv_var_op): New.
35477         (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35478         IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35480         * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35481         * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35482         (copy_decl_for_dup_finish): Ditto.
35484         * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35486 2017-03-28  Uros Bizjak  <ubizjak@gmail.com>
35488         PR target/53383
35489         * config/i386/i386.c (ix86_option_override_internal): Always
35490         allow -mpreferred-stack-boundary=3 for 64-bit targets.
35492 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35494         * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35496 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35498         * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35499         mark new edge's irreducible flag accordign to it.
35500         (vect_do_peeling): Check loop preheader edge's irreducible flag
35501         and pass it to function slpeel_add_loop_guard.
35503 2017-03-28  Richard Sandiford  <richard.sandiford@arm.com>
35505         PR tree-optimization/80218
35506         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35507         Update block frequencies and counts.
35509 2017-03-28  Richard Biener  <rguenther@suse.de>
35511         PR tree-optimization/78644
35512         * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35513         of a simplification result we may not use it at all.
35515 2017-03-28  Richard Biener  <rguenther@suse.de>
35517         PR ipa/80205
35518         * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35519         without arguments, generate default definition of a SSA name.
35521 2017-03-28  Richard Biener  <rguenther@suse.de>
35523         PR middle-end/80222
35524         * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35525         TYPE_REF_CAN_ALIAS_ALL references.
35526         * fold-const.c (fold_indirect_ref_1): Likewise.
35528 2017-03-28  Martin Liska  <mliska@suse.cz>
35530         PR ipa/80104
35531         * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35532         thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35534 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35535             Thomas Petazzoni  <thomas.petazzoni@free-electrons.com>
35537         * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35538         (EXTRA_SPECS): Define.
35539         (SUBTARGET_EXTRA_SPECS): Likewise.
35540         (SUBTARGET_CPP_SPEC): Likewise.
35541         * config/arc/elf.h (EXTRA_SPECS): Renamed to
35542         SUBTARGET_EXTRA_SPECS.
35543         * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35545 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35547         * config/arc/simdext.md (vst64_insn): Update pattern.
35548         (vld32wh_insn): Likewise.
35549         (vld32wl_insn): Likewise.
35550         (vld64_insn): Likewise.
35551         (vld32_insn): Likewise.
35553 2017-03-28  Marek Polacek  <polacek@redhat.com>
35555         PR sanitizer/80067
35556         * fold-const.c (fold_comparison): Use protected_set_expr_location
35557         instead of SET_EXPR_LOCATION.
35559 2017-03-28  Markus Trippelsdorf  <markus@trippelsdorf.de>
35561         * tree.c (add_expr): Avoid name lookup warning.
35563 2017-03-27  Jeff Law  <law@redhat.com>
35565         PR tree-optimization/80216
35566         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35567         function name.  Limit recursion depth.
35568         (record_temporary_equivalences): Corresponding changes.
35570 2017-03-27  Jonathan Wakely  <jwakely@redhat.com>
35572         * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35573         covered first.
35575 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35577         PR target/80102
35578         * reg-notes.def (REG_CFA_NOTE): Define.  Use it for CFA related
35579         notes.
35580         * cfgcleanup.c (reg_note_cfa_p): New array.
35581         (insns_have_identical_cfa_notes): New function.
35582         (old_insns_match_p): Don't cross-jump in between /f
35583         and non-/f instructions.  If both i1 and i2 are frame related,
35584         verify all CFA notes, their order and content.
35586 2017-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
35588         PR target/78543
35589         * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35590         HImode and SImode with zero extend to DImode to one insn.
35591         (bswap<mode>2_extenddi): Likewise.
35592         (bswapsi2_extenddi): Likewise.
35593         (bswaphi2_extendsi): Likewise.
35594         (bswaphi2): Combine bswap HImode and SImode into one insn.
35595         Separate memory insns from swapping register.
35596         (bswapsi2): Likewise.
35597         (bswap<mode>2): Likewise.
35598         (bswaphi2_internal): Delete, no longer used.
35599         (bswapsi2_internal): Likewise.
35600         (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35601         store, and gpr<-gpr swap insns.
35602         (bswap<mode>2_store): Likewise.
35603         (bswaphi2_reg): Register only splitter, combine with the splitter.
35604         (bswaphi2 splitter): Likewise.
35605         (bswapsi2_reg): Likewise.
35606         (bswapsi2 splitter): Likewise.
35607         (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35608         the insns into load, store, and register/register insns.
35609         (bswapdi2_ldbrx): Likewise.
35610         (bswapdi2_load): Likewise.
35611         (bswapdi2_store): Likewise.
35612         (bswapdi2_reg): Likewise.
35614 2017-03-27  Gunther Nikl  <gnikl@users.sourceforge.net>
35616         * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35617         (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35619 2017-03-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35621         PR target/80103
35622         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35623         add comments.
35624         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35625         special handling for target option conflicts between dform
35626         options (-mpower9-dform, -mpower9-dform-vector,
35627         -mpower9-dform-scalar) and -mno-direct-move.
35629 2017-03-27  Richard Biener  <rguenther@suse.de>
35631         PR tree-optimization/80181
35632         * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35634 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35636         * config/arc/predicates.md (move_double_src_operand): Replace the
35637         call to move_double_src_operand with a call to address_operand.
35639 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35641         * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35642         * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35643         * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35645 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35647         * config/arc/predicates.md (long_immediate_loadstore_operand):
35648         Consider scaled addresses cases.
35650 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35652         * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35653         restored when in interrupt.
35654         * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35655         doesn't have delay slot.
35657 2017-03-27  Richard Biener  <rguenther@suse.de>
35659         PR ipa/79776
35660         * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35661         inlined thunk clones.
35663 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35665         PR sanitizer/80168
35666         * asan.c (instrument_derefs): Copy over last operand from
35667         original COMPONENT_REF to the new COMPONENT_REF with
35668         DECL_BIT_FIELD_REPRESENTATIVE.
35669         * ubsan.c (instrument_object_size): Likewise.
35671 2017-03-27  Richard Biener  <rguenther@suse.de>
35673         PR tree-optimization/80170
35674         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35675         sure DR/SCEV didnt fold in constants we do not see when looking
35676         at the reference base alignment.
35678 2017-03-27  Richard Biener  <rguenther@suse.de>
35680         PR middle-end/80171
35681         * gimple-fold.c (fold_ctor_reference): Properly guard against
35682         NULL return value from canonicalize_constructor_val.
35684 2017-03-25  Uros Bizjak  <ubizjak@gmail.com>
35686         PR target/80180
35687         * config/i386/i386.c (ix86_expand_builtin)
35688         <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35689         flags reg setting and flags reg using instructions.
35690         <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto.  Use non-flags reg
35691         clobbering instructions to zero extend op2.
35693 2017-03-25  Gerald Pfeifer  <gerald@pfeifer.com>
35695         * doc/install.texi (Configuration) <--with-aix-soname>:
35696         Update link to AIX ld.
35698 2017-03-25  Bernd Schmidt  <bschmidt@redhat.com>
35700         PR rtl-optimization/80160
35701         PR rtl-optimization/80159
35702         * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35703         reg_alternate_class into account.
35705 2017-03-24  Vladimir Makarov  <vmakarov@redhat.com>
35707         PR target/80148
35708         * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35709         to consider in curr_insn_transform.
35711 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35713         * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35714         * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35715         and emit_mode_inner.
35717 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35719         * config/s390/s390-builtins.def: Add VXE builtins.  Add a flags
35720         argument to the overloaded builtin variants.  Use the new flag to
35721         deprecate certain builtin variants.
35722         * config/s390/s390-builtin-types.def: Add new builtin types.
35723         * config/s390/s390-builtins.h: Support new flags field for
35724         overloaded builtins.
35725         * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35726         (s390_macro_to_expand): Enable vector float data type.
35727         (s390_cpu_cpp_builtins_internal): Indicate support of the new
35728         builtins by incrementing the __VEC__ version number.
35729         (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35730         vec_xst.
35731         (s390_resolve_overloaded_builtin): Emit error messages depending
35732         on the builtin flags.
35733         * config/s390/s390.c (s390_expand_builtin): Support additional
35734         flags argument.  Change error message to match the messages
35735         emitted in s390-c.c.
35736         * config/s390/s390.md: New UNSPEC_* constants.
35737         (op_type): Add new instruction types.
35738         * config/s390/vecintrin.h: Add new builtins and test data class
35739         constants.
35740         * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35741         (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35742         (VEC_INEXACT, VEC_NOINEXACT): New constants.
35743         ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35744         ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35745         ("vec_mergel<mode>"): V_HW -> VEC_HW.
35747         ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35748         ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35749         ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35750         ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35752         ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35753         ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35754         ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35755         ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35757         ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35758         ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35759         ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35760         ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35761         ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35762         ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35763         ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35765         ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35766         ("vec_scatter_element<V_HW_4:mode>_DI")
35767         ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35768         ("vec_fpint<mode>", "vflls")
35769         ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35770         ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35771         ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35772         ("*vec_cmphe<mode>_cc"): ... these.
35774         ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35775         mode constant instead of magic value.
35777 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35779         * config/s390/s390.c (s390_expand_vec_compare): Support other
35780         vector floating point modes than just V2DF.
35781         (s390_expand_vcond): Likewise.
35782         (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35783         (s390_cannot_change_mode_class): Prevent mode changes between TF
35784         and V1TF in vector registers.
35785         * config/s390/s390.md (DF, SF): New mode attributes.
35786         ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35787         ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35788         SFmode support for VRs.
35789         * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35790         vector fp modes.
35791         (VFT, VF_HW): New mode iterators.
35792         (vw, sdx): New mode attributes.
35793         ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35794         ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35795         ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35796         ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35797         ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35798         also the new vector floating point modes.  Renaming to ...
35800         ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35801         ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35802         ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35803         ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35804         ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35805         ("vec_unordered<mode>"): ... these.
35807         ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35808         ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35809         ("*vec_extendv2df"): New insn definitions.
35811 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35813         * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35814         ("mulditi3_2", "*muldi3_sign"): New patterns.
35815         ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35816         rename the pattern definition.
35818 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35820         * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35821         expander.
35822         ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35824 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35826         * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35827         instruction if possible.
35828         * config/s390/vector.md (vec_halfnumelts): New mode
35829         attribute.
35830         ("*vec_vllezlf<mode>"): New pattern.
35832 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35834         * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35835         ("popcountv4si2", "popcountv2di2"): Rename to ...
35836         ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35837         ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35838         condition.
35839         ("popcount<mode>2_vxe"): New pattern.
35841 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35843         * common/config/s390/s390-common.c (processor_flags_table): Add
35844         arch12.
35845         * config.gcc: Add arch12.
35846         * config/s390/driver-native.c (s390_host_detect_local_cpu):
35847         Default to arch12 for unknown CPU model numbers.
35848         * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35849         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35850         PROCESSOR_max sanity check.
35851         * config/s390/s390-opts.h (enum processor_type): Add
35852         PROCESSOR_ARCH12.
35853         * config/s390/s390.c (processor_table): Add arch12.
35854         (s390_expand_builtin): Add check for B_VXE flag.
35855         (s390_issue_rate): Add PROCESSOR_ARCH12.
35856         (s390_get_sched_attrmask): Likewise.
35857         (s390_get_unit_mask): Likewise.
35858         (s390_sched_score): Enable z13 scheduling for arch12.
35859         (s390_sched_reorder): Likewise.
35860         (s390_sched_variable_issue): Likewise.
35861         * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35862         PF_VXE.
35863         (s390_tune_attr): Use z13 scheduling also for arch12.
35864         (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35865         (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35866         (TARGET_VXE_P): New macros.
35867         * config/s390/s390.md: Add arch12 to cpu attribute.  Add arch12
35868         and vxe to cpu_facility.  Add arch12 and vxe to enabled attribute.
35869         * config/s390/s390.opt: Add arch12 as processor_type.
35871 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35873         * config/s390/s390.md
35874         ("fixuns_truncdddi2", "fixuns_trunctddi2")
35875         ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35876         ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35878         ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35879         Rename expanders to ...
35881         ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35882         ("fixuns_truncdddi2_emu"): ... these.
35884         ("fixuns_trunc<mode>si2_emu"): New expander.
35886         ("*fixuns_truncdfdi2_z13"): Rename to ...
35887         ("*fixuns_truncdfdi2_vx"): ... this.
35889 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35891         * config/s390/2964.md: Remove the single element vector compare
35892         instructions which are no longer used.
35893         * config/s390/s390.c (s390_select_ccmode): Remove handling of
35894         vector CCmodes.
35895         (s390_canonicalize_comparison): Remove handling of DFmode
35896         compares.
35897         (s390_expand_vec_compare_scalar): Remove function.
35898         (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35899         * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35900         pattern.
35901         ("*cmp<mode>_ccs"): Add wfcdb instruction.
35903 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35905         * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35906         FP zero.
35907         ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35908         will anyway by matched by mov<mode>_64dfp.
35910 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35912         * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35913         vlef/vstef.  Add missing operand to vleif.
35915 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35917         * config/s390/s390.c (s390_expand_vec_init): Enable vector load
35918         pair for all vector types with 64 bit elements.
35919         * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
35920         * config/s390/vector.md (V_HW_64): ... here.
35921         (V_128_NOSINGLE): New mode iterator.
35922         ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
35923         ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
35924         ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
35925         ("*vec_load_pairv2di"): Change to ...
35926         ("*vec_load_pair<mode>"): ... this one.
35928 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35930         * config/s390/constraints.md: Add comments.
35931         (jKK): Reject element sizes > 8 bytes.
35932         * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
35933         s_operands.
35934         * config/s390/s390.md: Add the s_operand checks formerly in
35935         s390_split_ok_p to various splitters where they are still
35936         required.
35937         * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
35938         for 128 bit vectors.  Plus two splitters.
35940 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35942         * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
35943         the file.
35945 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35947         PR target/79893
35948         * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
35949         error if the boundary argument is not constant.
35951 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35953         PR rtl-optimization/80112
35954         * loop-doloop.c (doloop_condition_get): Don't check condition
35955         if cmp isn't SET with IF_THEN_ELSE src.
35957 2017-03-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35959         PR tree-optimization/80158
35960         * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
35961         replacing a candidate statement, also replace it for the
35962         candidate's alternate interpretation.
35963         (replace_rhs_if_not_dup): Likewise.
35964         (replace_one_candidate): Likewise.
35966 2017-03-24  Richard Biener  <rguenther@suse.de>
35968         PR tree-optimization/80167
35969         * graphite-isl-ast-to-gimple.c
35970         (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
35971         properly.
35972         (translate_isl_ast_to_gimple::get_rename): Likewise.
35974 2017-03-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35976         * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
35977         handling of certain combinations of target options, including the
35978         combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
35979         -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
35981 2017-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
35983         PR target/71436
35984         * config/arm/arm.md (*load_multiple): Add reload_completed to
35985         matching condition.
35987 2017-03-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35988             Richard Biener  <rguenth@suse.de>
35990         PR tree-optimization/79908
35991         PR tree-optimization/80136
35992         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
35993         been cast away, gimplify_and_add suffices.
35995 2017-03-23  Markus Trippelsdorf  <markus@trippelsdorf.de>
35997         * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
35999 2017-03-23  Richard Biener  <rguenther@suse.de>
36001         PR tree-optimization/80032
36002         * gimplify.c (gimple_push_cleanup): Forced unconditional
36003         cleanups still have to go to the conditional_cleanups
36004         sequence.
36006 2017-03-22  Jakub Jelinek  <jakub@redhat.com>
36008         PR tree-optimization/80072
36009         * tree-ssa-reassoc.c (struct operand_entry): Change id field type
36010         to unsigned int.
36011         (next_operand_entry_id): Change type to unsigned int.
36012         (sort_by_operand_rank): Make sure to return the right return value
36013         even if unsigned fields are bigger than INT_MAX.
36014         (struct oecount): Change cnt and id type to unsigned int.
36015         (oecount_hasher::equal): Formatting fix.
36016         (oecount_cmp): Make sure to return the right return value
36017         even if unsigned fields are bigger than INT_MAX.
36018         (undistribute_ops_list): Change next_oecount_id type to unsigned int.
36020         PR c++/80129
36021         * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
36022         TREE_READONLY on result if writing it more than once.
36024         PR sanitizer/80110
36025         * doc/invoke.texi (-fsanitize=thread): Document that with
36026         -fnon-call-exceptions atomics are not able to throw
36027         exceptions.
36029         PR sanitizer/80110
36030         * tsan.c: Include tree-eh.h.
36031         (instrument_builtin_call): Call maybe_clean_eh_stmt or
36032         maybe_clean_or_replace_eh_stmt where needed.
36033         (instrument_memory_accesses): Add cfg_changed argument.
36034         Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
36035         if it returned true.
36036         (tsan_pass): Adjust caller.  Return TODO_cleanup_cfg if cfg_changed.
36038         PR rtl-optimization/63191
36039         * config/i386/i386.c (ix86_delegitimize_address): Turn into small
36040         wrapper function, moved the whole old content into ...
36041         (ix86_delegitimize_address_1): ... this.  New inline function.
36042         (ix86_find_base_term): Use ix86_delegitimize_address_1 with
36043         true as last argument instead of ix86_delegitimize_address.
36045 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
36047         * config/aarch64/aarch64.c (generic_branch_cost): Copy
36048         cortexa57_branch_cost.
36050 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
36052         * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
36054 2017-03-21  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36056         PR target/80123
36057         * doc/md.texi (Constraints): Document wA constraint.
36058         * config/rs6000/constraints.md (wA): New.
36059         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
36060         (rs6000_init_hard_regno_mode_ok): Init wA constraint.
36061         * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
36062         * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
36064 2017-03-22  Cesar Philippidis  <cesar@codesourcery.com>
36066         PR c++/80029
36067         * gimplify.c (is_oacc_declared): New function.
36068         (oacc_default_clause): Use it to set default flags for acc declared
36069         variables inside parallel regions.
36070         (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
36071         declared variables.
36072         (gimplify_oacc_declare): Gimplify the declare clauses.  Add the
36073         declare attribute to any decl as necessary.
36075 2017-03-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
36077         PR target/80082
36078         * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
36079         (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
36080         * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
36081         (arm_arch_lpae): This.
36082         * config/arm/arm.c (arm_arch7ve): Rename into ...
36083         (arm_arch_lpae): This.  Define it in term of isa_bit_lpae.
36084         * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
36085         arm_arch_lpae.
36087 2017-03-22  Martin Liska  <mliska@suse.cz>
36089         PR target/79906
36090         * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
36091         error message instead of an ICE.
36093 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36095         * doc/extend.texi (6.11 Additional Floating Types): Revise.
36097 2017-03-21  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36099         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
36100         comments.
36101         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
36102         comments.
36104 2017-03-21  Martin Sebor  <msebor@redhat.com>
36106         * doc/extend.texi: Use "cannot" instead of "can't."
36107         * doc/hostconfig.texi: Same.
36108         * doc/install.texi: Same.
36109         * doc/invoke.texi: Same.
36110         * doc/loop.texi: Same.
36111         * doc/md.texi: Same.
36112         * doc/objc.texi: Same.
36113         * doc/rtl.texi: Same.
36114         * doc/tm.texi: Same.
36115         * doc/tm.texi.in: Same.
36116         * doc/trouble.texi: Same.
36118 2017-03-21  Alexandre Oliva  <aoliva@redhat.com>
36120         PR debug/63238
36121         * dwarf2out.c (struct checksum_attributes): Add at_alignment.
36122         (collect_checksum_attributes): Set it.
36123         (die_checksum_ordered): Use it.
36125 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36127         PR tree-optimization/79908
36128         * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
36129         change: For a VA_ARG whose LHS has been cast away, use
36130         force_gimple_operand to construct the side effects.
36132 2017-03-21  David Malcolm  <dmalcolm@redhat.com>
36134         PR translation/80001
36135         * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
36136         more amenable to translation.
36137         (oacc_loop_auto_partitions): Likewise.
36139 2017-03-21  Marek Polacek  <polacek@redhat.com>
36140             Martin Sebor  <msebor@redhat.com>
36142         PR tree-optimization/80109
36143         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
36144         on INTEGRAL_TYPE_P.
36146 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36147             Segher Boessenkool  <segher@kernel.crashing.org>
36149         PR target/80125
36150         * combine.c (can_combine_p): Revert the 2017-03-20 change, only
36151         check reg_used_between_p between insn and one of succ or succ2
36152         depending on if succ is artificial insn not inserted into insn
36153         stream.
36155 2017-03-21  Martin Liska  <mliska@suse.cz>
36157         PR gcov-profile/80081
36158         * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
36159         * doc/gcc.texi: Include gcov-dump stuff.
36160         * doc/gcov-dump.texi: New file.
36162 2017-03-21  Toma Tabacu  <toma.tabacu@imgtec.com>
36164         PR rtl-optimization/79150
36165         * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
36166         conditional jump, if the jump is the last insn of the loop.
36168 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36169             Richard Biener  <rguenth@suse.de>
36171         PR tree-optimization/79908
36172         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36173         been cast away, use force_gimple_operand to construct the side
36174         effects.
36176 2017-03-21  Martin Liska  <mliska@suse.cz>
36178         PR libfortran/79956
36179         * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
36180         to NULL.
36182 2017-03-21  Brad Spengler <spender@grsecurity.net>
36184         PR plugins/80094
36185         * plugin.c (htab_hash_plugin): New function.
36186         (add_new_plugin): Use it and adjust.
36187         (parse_plugin_arg_opt): Adjust.
36188         (init_one_plugin): Likewise.
36190 2017-03-21  Richard Biener  <rguenther@suse.de>
36192         PR tree-optimization/80032
36193         * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
36194         if set force the cleanup to happen unconditionally.
36195         (gimplify_target_expr): Push inserted clobbers with force_uncond
36196         to avoid them being removed by control-dependent DCE.
36198 2017-03-21  Richard Biener  <rguenther@suse.de>
36200         PR tree-optimization/80122
36201         * tree-inline.c (copy_bb): Do not expans va-arg packs or
36202         va_arg_pack_len when the inlined call stmt requires pack
36203         expansion itself.
36204         * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
36206 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36208         PR sanitizer/78158
36209         * tsan.c (instrument_builtin_call): If the memory model argument
36210         is not a constant, assume it is valid.
36212         PR c/67338
36213         * fold-const.c (round_up_loc): Negate divisor in unsigned type to
36214         avoid UB.
36216 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36218         PR rtl-optimization/79910
36219         * combine.c (can_combine_p): Do not allow combining an I0 or I1
36220         if its dest is used by an insn before I2 (other than the combined
36221         insns themselves, which are properly handled already).
36223 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36225         Revert:
36226         2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36228         * combine.c (record_used_regs): New static function.
36229         (try_combine): Handle situations where there is an additional
36230         instruction between I2 and I3 which needs to have a LOG_LINK
36231         updated.
36233         Revert:
36234         2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36236         * combine.c (try_combine): Delete redundant i1 test.  Call
36237         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36239 2017-03-20  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36241         PR target/80083
36242         * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
36243         alternatives 13/14.
36245 2017-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36247         PR tree-optimization/80054
36248         * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36249         the optimization if a PHI or any of its arguments is not dominated
36250         by the candidate's basis.  Use gphi* rather than gimple* as
36251         appropriate.
36252         (replace_profitable_candidates): Clean up a gimple* variable that
36253         should be a gphi* variable.
36255 2017-03-20  Martin Sebor  <msebor@redhat.com>
36257         PR c++/52477
36258         * doc/extend.texi (attribute constructor): Document present limitation.
36260 2017-03-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36262         PR target/79963
36263         * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36264         __POWER9_VECTOR__ #ifdef control, change template definition to
36265         use Power9-specific built-in function.
36266         (vec_any_eq): Likewise.
36267         * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36268         to control outcomes from this test.
36269         (vector_ae_<mode>p): For VEC_F modes, likewise.
36271 2017-03-20  Ian Lance Taylor  <iant@google.com>
36273         * config/i386/i386.c (ix86_function_regparm): Save an extra
36274         register for -fsplit-stack with DECL_STATIC_CHAIN.
36276 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36278         PR target/79912
36279         * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36280         (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36282 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36284         * config/riscv/riscv.c (riscv_print_operand): Use "fence
36285         iorw,ow".
36286         * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36287         iorw,iorw".
36289 2017-03-20  Marek Polacek  <polacek@redhat.com>
36291         PR sanitizer/80063
36292         * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36294 2017-03-20  Richard Biener  <rguenther@suse.de>
36296         PR tree-optimization/80113
36297         * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36298         allocate extra SSA name for PHI def.
36299         (add_close_phis_to_outer_loops): Likewise.
36300         (add_close_phis_to_merge_points): Likewise.
36301         (copy_loop_close_phi_args): Likewise.
36302         (copy_cond_phi_nodes): Likewise.
36304 2017-03-20  Martin Liska  <mliska@suse.cz>
36306         PR middle-end/79753
36307         * tree-chkp.c (chkp_build_returned_bound): Do not build
36308         returned bounds for a LHS that's not a BOUNDED_P type.
36310 2017-03-20  Martin Liska  <mliska@suse.cz>
36312         PR target/79769
36313         PR target/79770
36314         * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36315         COMPLEX_CST and VECTOR_CST.
36317 2017-03-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36319         PR target/78857
36320         * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36321         target operand.  A new splitter adds the clobber statement in case
36322         the target operand is dead anyway.
36324 2017-03-19  Gerald Pfeifer  <gerald@pfeifer.com>
36326         * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36327         to age-old versions of binutils and glibc.
36329 2017-03-18  Segher Boessenkool  <segher@kernel.crashing.org>
36331         * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36333 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36335         * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36337 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36339         * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36340         requirement for binutils 2.13.
36342 2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36344         * combine.c (try_combine): Delete redundant i1 test.  Call
36345         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36347 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com
36349         * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36350         riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36351         contents.
36352         <riscv64-*-elf>: Re-arrange section
36353         <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36354         <riscv32-*-linux>: Likewise.
36355         <riscv64-*-elf>: Likewise
36356         <riscv64-*-linux>: Likewise.
36358 2017-03-17  Richard Earnshaw  <rearnsha@arm.com>
36360         PR target/80052
36361         * aarch64.opt(verbose-cost-dump): Fix typo.
36363 2017-03-17  Pat Haugen  <pthaugen@us.ibm.com>
36365         PR target/79951
36366         * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36367         for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36369 2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36371         * reload.c (find_reloads): When reloading a nonoffsettable address,
36372         use RELOAD_OTHER for it and its address reloads.
36374         PR rtl-optimization/79910
36375         * combine.c (record_used_regs): New static function.
36376         (try_combine): Handle situations where there is an additional
36377         instruction between I2 and I3 which needs to have a LOG_LINK
36378         updated.
36380 2017-03-17  Jeff Law  <law@redhat.com>
36382         PR tree-optimization/71437
36383         * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36384         conditional in the hash table first.
36385         (vrp_dom_walker::before_dom_children): Extract condition from
36386         ASSERT_EXPR.  Record condition, its inverion and any implied
36387         conditions as well.
36389 2017-03-17  Marek Polacek  <polacek@redhat.com>
36390             Markus Trippelsdorf  <markus@trippelsdorf.de>
36392         PR tree-optimization/80079
36393         * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36394         m_stores_head.
36396 2017-03-17  Richard Biener  <rguenther@suse.de>
36398         PR middle-end/80075
36399         * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36400         Properly verify the LHS before the RHS possibly claims to be
36401         handled.
36402         (stmt_could_throw_p): Hande gimple conds fully here.  Clobbers
36403         do not throw.
36405 2017-03-17  Martin Jambor  <mjambor@suse.cz>
36407         * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36408         (List of -O2 options): Likewise.
36409         (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36410         (-fipa-vrp) New.
36412 2017-03-17  Tom de Vries  <tom@codesourcery.com>
36414         * gcov-dump.c (print_usage): Print bug_report_url.
36416 2017-03-17  Richard Biener  <rguenther@suse.de>
36418         PR middle-end/80050
36419         * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36420         (parser::peek): Likewise.
36422 2017-03-17  Richard Biener  <rguenther@suse.de>
36424         PR tree-optimization/80048
36425         * sese.c (free_sese_info): Properly release rename_map and
36426         copied_bb_map elements.
36428 2017-03-16  Alexandre Oliva  <aoliva@redhat.com>
36430         * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36431         Add linked-list forward and backlinks.  Insert on
36432         construction, remove on destruction.
36433         (class pass_store_merging): Add m_stores_head field.
36434         (pass_store_merging::terminate_and_process_all_chains):
36435         Iterate over m_stores_head list.
36436         (pass_store_merging::terminate_all_aliasing_chains):
36437         Likewise.
36438         (pass_store_merging::execute): Check for debug stmts first.
36439         Push new chains onto the m_stores_head stack.
36441 2017-03-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
36443         PR target/71294
36444         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36445         SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36446         but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36448 2017-03-16  Jeff Law  <law@redhat.com>
36450         PR tree-optimization/71437
36451         * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36452         member function.  Implementation moved into after_dom_children
36453         member function and into the threader's thread_outgoing_edges
36454         function.
36455         (dom_opt_dom_walker::after_dom_children): Simplify by moving
36456         some code into new thread_outgoing_edges.
36457         * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36458         definition.  Simplify marker handling (do it here).   Assume we always
36459         have the available expression and the const/copies tables.
36460         (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36461         and tree-vrp.c
36462         * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36463         * tree-vrp.c (equiv_stack): No longer file scoped.
36464         (vrp_dom_walker): New class.
36465         (vrp_dom_walker::before_dom_children): New member function.
36466         (vrp_dom_walker::after_dom_children): Likewise.
36467         (identify_jump_threads):  Setup domwalker.  Use it rather than
36468         walking edges in a random order by hand.  Simplify setup/finalization.
36469         (finalize_jump_threads): Remove.
36470         (vrp_finalize): Do not call identify_jump_threads here.
36471         (execute_vrp): Do it here instead and call thread_through_all_blocks
36472         here too.
36474         PR tree-optimization/71437
36475         * tree-ssa-dom.c (pfn_simplify): Add basic_block argument.  All
36476         callers changed.
36477         (simplify_stmt_for_jump_threading): Add basic_block argument.  All
36478         callers changed.
36479         (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36480         (dom_opt_dom_walker::thread_across_edge): Remove
36481         handle_dominating_asserts argument.  All callers changed.
36482         (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36483         changes.  Remove calls to lhs_of_dominating_assert.  Other
36484         uses of handle_dominating_asserts turn into unconditional code
36485         (simplify_control_stmt_condition_1): Likewise.
36486         (simplify_control_stmt_condition): Likewise.
36487         (thread_through_normal_block, thread_across_edge): Likewise.
36488         * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36489         * tree-vrp.c (lhs_of_dominating_assert): Move here.  Return original
36490         object if it is not an SSA_NAME.
36491         (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36492         before calling into the VRP specific simplifiers.
36493         (identify_jump_threads): Remove handle_dominating_asserts
36494         argument.
36496 2017-03-16  Jakub Jelinek  <jakub@redhat.com>
36498         PR fortran/79886
36499         * tree-diagnostic.c (default_tree_printer): No longer static.
36500         * tree-diagnostic.h (default_tree_printer): New prototype.
36502 2017-03-16  Tamar Christina  <tamar.christina@arm.com>
36504         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36505         Change ins into fmov.
36507 2017-03-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36509         * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36510         * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36511         Use h_con constraint for operand 1.
36512         (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36513         (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36515 2017-03-15  Jeff Law  <law@redhat.com>
36517         PR tree-optimization/71437
36518         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36519         (record_temporary_equivalences): Use it.
36521         PR tree-optimization/71437
36522         * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36523         tree-ssa-scopedtables.
36524         (lookup_avail_expr, build_and_record_new_cond): Likewise.
36525         (record_conditions, record_cond, vuse_eq): Likewise.
36526         (record_edge_info): Adjust to API tweak of record_conditions.
36527         (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36528         (record_temporary_equivalences, optimize_stmt): Likewise.
36529         (eliminate_redundant_computations): Likewise.
36530         (record_equivalences_from_stmt): Likewise.
36531         * tree-ssa-scopedtables.c: Include options.h and params.h.
36532         (vuse_eq): New function, moved from tree-ssa-dom.c
36533         (build_and_record_new_cond): Likewise.
36534         (record_conditions): Likewise.  Accept vector of conditions rather
36535         than edge_equivalence structure for first argument.
36536         for the first argument.
36537         (avail_exprs_stack::lookup_avail_expr): New member function, moved
36538         from tree-ssa-dom.c.
36539         (avail_exprs_stack::record_cond): Likewise.
36540         * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36541         from tree-ssa-dom.c.
36542         (avail_exprs_stack): Add new member functions lookup_avail_expr
36543         and record_cond.
36544         (record_conditions): Declare.
36546 2017-03-15  Vladimir Makarov  <vmakarov@redhat.com>
36548         PR target/80017
36549         * lra-constraints.c (process_alt_operands): Increase reject for
36550         reloading an input/output operand.
36552 2017-03-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
36554         PR target/79038
36555         * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36556         insns to convert from signed/unsigned char/short to IEEE 128-bit
36557         floating point.
36558         (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36560 2017-03-15  Uros Bizjak  <ubizjak@gmail.com>
36562         PR target/80019
36563         * config/i386/i386.c (ix86_vector_duplicate_value): Create
36564         subreg of inner mode for values already in registers.
36566 2017-03-15  Bernd Schmidt  <bschmidt@redhat.com>
36568         * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36569         iteration reg is used after the loop.
36571 2017-03-14  Martin Sebor  <msebor@redhat.com>
36573         PR tree-optimization/79800
36574         * gimple-ssa-sprintf.c (format_floating: Add argument.  Handle
36575         precision in negative-positive range.
36576         (format_floating): Call non-const overload with adjusted precision.
36578 2017-03-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
36580         PR target/79947
36581         * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36582         -mpowerpc-gfxopt.
36584 2017-03-14  Martin Sebor  <msebor@redhat.com>
36586         PR middle-end/80020
36587         * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36588         * builtins.def (aligned_alloc): Use it.
36590         PR c/79936
36591         * Makefile.in (GTFILES): Add calls.c.
36592         * calls.c: Include "gt-calls.h".
36594 2017-03-14  Bernd Schmidt  <bschmidt@redhat.com>
36596         PR rtl-optimization/79728
36597         * regs.h (struct target_regs): New field
36598         x_contains_allocatable_regs_of_mode.
36599         (contains_allocatable_regs_of_mode): New macro.
36600         * reginfo.c (init_reg_sets_1): Initialize it, and change
36601         contains_reg_of_mode so it includes global regs as well.
36602         * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36603         rather than contains_regs_of_mode.
36605 2017-03-14  Martin Liska  <mliska@suse.cz>
36607         * doc/invoke.texi: Document options that can't be combined with
36608         -fcheck-pointer-bounds.
36610 2017-03-14  Martin Liska  <mliska@suse.cz>
36612         PR middle-end/79831
36613         * doc/invoke.texi (-Wchkp): Document the option.
36615 2017-03-14  Martin Liska  <mliska@suse.cz>
36617         * Makefile.in: Install gcov-dump.
36619 2017-03-14  Martin Liska  <mliska@suse.cz>
36621         * multiple_target.c (expand_target_clones): Bail out for
36622         an invalid attribute.
36624 2017-03-14  Richard Biener  <rguenther@suse.de>
36626         * alias.c (struct alias_set_entry): Pack properly.
36627         * cfgloop.h (struct loop): Likewise.
36628         * cse.c (struct set): Likewise.
36629         * ipa-utils.c (struct searchc_env): Likewise.
36630         * loop-invariant.c (struct invariant): Likewise.
36631         * lra-remat.c (struct cand): Likewise.
36632         * recog.c (struct change_t): Likewise.
36633         * rtl.h (struct address_info): Likewise.
36634         * symbol-summary.h (function_summary): Likewise.
36635         * tree-loop-distribution.c (struct partition): Likewise.
36636         * tree-object-size.c (struct object_size_info): Likewise.
36637         * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36638         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36639         * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36640         * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36641         * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36642         * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36643         (struct _stmt_vec_info): Likewise.
36645 2017-03-14  Martin Liska  <mliska@suse.cz>
36647         PR target/79892
36648         * multiple_target.c (create_dispatcher_calls): Check that
36649         a target can create a function dispatcher.
36651 2017-03-14  Martin Liska  <mliska@suse.cz>
36653         PR lto/66295
36654         * multiple_target.c (expand_target_clones): Drop local.local
36655         flag for default implementation.
36657 2017-03-14  Richard Biener  <rguenther@suse.de>
36659         PR tree-optimization/80030
36660         * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36662 2017-03-13  Kito Cheng  <kito.cheng@gmail.com>
36664         * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36665         gcc_fallthrough() instead of __attribute__((fallthrough));
36667 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36669         * doc/gcc.texi: Remove "up" link to (DIR).
36670         * doc/gccint.texi: Ditto.
36672 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36674         * doc/install.texi (Specific) <avr>: Remove reference to
36675         binutils 2.13.
36677 2017-03-13  Jeff Law  <law@redhat.com>
36679         * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36680         attribute rather than comments.
36682         * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36683         match_scratch operand is highest.
36685 2017-03-13  Martin Liska  <mliska@suse.cz>
36687         PR middle-end/78339
36688         * ipa-pure-const.c (warn_function_noreturn): If the declarations
36689         is a CHKP clone, use original declaration.
36691 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36693         * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36694         (arc_conditional_register_usage): Use a different allocation order
36695         when optimizing for size.
36696         * common/config/arc/arc-common.c (arc_option_optimization_table):
36697         Section anchors default on when optimizing for size.
36699 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36701         * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36703 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36705         * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36706         * config/arc/arc.md (cpu_facility): Add cd variant.
36707         (*movqi_insn): Add code density variant.
36708         (*movhi_insn): Likewise.
36709         (*movqi_insn): Likewise.
36710         (*addsi3_mixed): Likewise.
36711         (subsi3_insn): Likewise.
36713 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36715         * config/arc/arc.md (movsi_cond_exec): Update constraint.
36717 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36719         * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36720         expressions with MINUS and UNARY ops.
36722 2017-03-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36724         PR target/79911
36725         * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36726         Rename to...
36727         (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36728         between vec_select and vector argument.
36729         (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36730         (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36731         (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36732         (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36733         (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36734         (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36736 2017-03-13  Richard Biener  <rguenther@suse.de>
36738         PR other/79991
36739         * params.def (vect-max-peeling-for-alignment): Fix typo.
36741 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36743         * doc/install.texi (Specific) <mips-*-*>: Remove description of
36744         issue that only occurred with binutils below 2.18.
36746 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36748         * doc/install.texi (Specific) <cris-axis-elf>: No longer
36749         refer to binutils 2.11/2.12 minimum.
36751 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36753         * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36754         ftp.kernel.org and simplify binutils requirement.
36756 2017-03-11  Gerald Pfeifer  <gerald@pfeifer.com>
36758         * doc/invoke.texi (Warning Options): Fix spelling of link-time
36759         optimization.
36760         (Optimize Options): Ditto.  Also remove redundancy.
36762 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36764         PR translation/79848
36765         * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36766         "%qs".
36767         * ipa-pure-const.c (suggest_attribute): Likewise.  Convert _
36768         to G_ to avoid double translation.
36770 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36772         PR translation/79923
36773         * auto-profile.c (get_combined_location): Convert leading
36774         character of diagnostics to lower case and remove trailing period.
36775         (read_profile): Likewise for various diagnostics.
36776         * config/arm/arm.c (arm_option_override): Remove trailing period
36777         from various diagnostics.
36778         * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36779         (msp430_expand_delay_cycles): Likewise.
36781 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36783         PR target/79925
36784         * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36785         full command-line argument, rather than just "str".
36786         (aarch64_validate_march): Likewise.
36787         (aarch64_validate_mtune): Likewise.
36789 2017-03-10  Bernd Schmidt  <bschmidt@redhat.com>
36791         PR rtl-optimization/78911
36792         * lra-assigns.c (must_not_spill_p): New function.
36793         (spill_for): Use it.
36795 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36797         PR tree-optimization/79981
36798         * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36799         ATOMIC_COMPARE_EXCHANGE ifn result.
36800         (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36801         IFN_ATOMIC_COMPARE_EXCHANGE.
36803 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36805         PR driver/79875
36806         * opts.c (parse_sanitizer_options): Add missing question mark to
36807         "did you mean" message.
36809 2017-03-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36811         * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36812         built-in.
36813         (VMULEUH_UNS): Likewise.
36814         (VMULOUB_UNS): Likewise.
36815         (VMULOUH_UNS): Likewise.
36816         * config/rs6000/rs6000.c (builtin_function_type): Remove
36817         references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36819 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36821         PR bootstrap/79952
36822         * read-rtl-function.c (function_reader::read_rtx_operand): Update
36823         x with result of extra_parsing_for_operand_code_0.
36824         (function_reader::extra_parsing_for_operand_code_0): Convert
36825         return type from void to rtx, returning x.  When reading
36826         SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36827         larger size containing struct block_symbol.
36829 2017-03-10  Segher Boessenkool  <segher@kernel.crashing.org>
36831         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36832         -mfloat128-hardware without -m64.
36834 2017-03-10  Will Schmidt <will_schmidt@vnet.ibm.com>
36836         PR target/79941
36837         * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36838         entries to the case statement that marks unsigned arguments to
36839         overloaded functions.
36841 2017-03-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36843         * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36844         two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36846 2017-03-10  Pat Haugen  <pthaugen@us.ibm.com>
36848         PR target/79907
36849         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36850         TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36852 2017-03-10  Martin Liska  <mliska@suse.cz>
36854         PR target/65705
36855         PR target/69804
36856         * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36857         * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36858         FIELD != NULL.
36860 2017-03-10  Olivier Hainque  <hainque@adacore.com>
36862         * tree-switch-conversion (array_value_type): Start by resetting
36863         candidate type to it's main variant.
36865 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36867         PR rtl-optimization/79909
36868         * combine.c (try_combine): Use simplify_replace_rtx on individual
36869         CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36870         of the whole CALL_INSN_FUNCTION_USAGE.
36872         PR tree-optimization/79972
36873         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36874         get_range_info on SSA_NAMEs.  Formatting fixes.
36876 2017-03-10  Richard Biener  <rguenther@suse.de>
36877             Jakub Jelinek  <jakub@redhat.com>
36879         PR tree-optimization/77975
36880         * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36881         edge to be constant.
36882         (get_val_for): For constant x return it.  Formatting fix.
36883         (loop_niter_by_eval): Avoid pointless looping if the next iteration
36884         would use the same bases as the current one.
36886 2017-03-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36888         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36889         instead of vec_select for V1TImode.
36890         * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36891         longer needed.
36892         (VSX_LE_128): Add V1TI to this mode iterator.
36893         (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36894         (*vsx_le_perm_store_<mode>): Likewise.
36895         (pre-reload splitter for VSX stores): Likewise.
36896         (post-reload splitter for VSX stores): Likewise.
36897         (*vsx_xxpermdi2_le_<mode>): Likewise.
36898         (*vsx_lxvd2x2_le_<mode>): Likewise.
36899         (*vsx_stxvd2x2_le_<mode>): Likewise.
36901 2017-03-09  Michael Eager  <eager@eagercon.com>
36903         Correct failures with --enable-checking=yes,rtl.
36905         * config/microblaze/microblaze.c (microblaze_expand_shift):
36906         Replace GET_CODE test with CONST_INT_P and INTVAL test with
36907         test for const0_rtx.
36908         * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36909         lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36911 2017-03-09  Richard Biener  <rguenther@suse.de>
36913         PR tree-optimization/79977
36914         * graphite-scop-detection.c (scop_detection::merge_sese):
36915         Handle the case of extra exits to blocks dominating the entry.
36917 2017-03-09  Toma Tabacu  <toma.tabacu@imgtec.com>
36919         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
36920         Document rdynamic.
36922 2017-03-09  Vladimir Makarov  <vmakarov@redhat.com>
36924         PR rtl-optimization/79949
36925         * lra-constraints.c (process_alt_operands): Check memory when
36926         trying to predict a cycle.  Print about the overall increase.
36928 2017-03-09  Richard Biener  <rguenther@suse.de>
36930         PR middle-end/79971
36931         * gimple-expr.c (useless_type_conversion_p): Preserve
36932         TYPE_SATURATING for fixed-point types.
36934 2017-03-09  Richard Biener  <rguenther@suse.de>
36936         PR ipa/79970
36937         * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
36938         alignment of BLKmode params.
36940 2017-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36942         PR target/79913
36943         * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
36944         (VALL_NO_V2Q): Likewise.
36945         (VDQF_DF): Delete.
36946         * config/aarch64/aarch64-simd.md
36947         (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
36948         iterator.
36949         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
36950         VALL_NO_V2Q mode iterator.
36951         (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
36953 2017-03-09  Martin Liska  <mliska@suse.cz>
36955         PR tree-optimization/79631
36956         * tree-chkp-opt.c (chkp_is_constant_addr): Call
36957         tree_int_cst_sign_bit just for INTEGER constants.
36959 2017-03-09  Martin Liska  <mliska@suse.cz>
36961         PR target/65705
36962         PR target/69804
36963         * toplev.c (process_options): Disable -fcheck-pointer-bounds with
36964         sanitizers.
36966 2017-03-09  Marek Polacek  <polacek@redhat.com>
36968         PR c++/79672
36969         * tree.c (inchash::add_expr): Handle TREE_VEC.
36971 2017-03-09  Martin Liska  <mliska@suse.cz>
36973         PR ipa/79764
36974         (chkp_narrow_size_and_offset): New function.
36975         (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
36976         (void chkp_parse_bit_field_ref): New function.
36977         (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
36978         (chkp_process_stmt): Use chkp_parse_bit_field_ref.
36980 2017-03-09  Martin Liska  <mliska@suse.cz>
36982         PR ipa/79761
36983         * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
36984         (chkp_find_bounds_1): Remove gcc_unreachable.
36986 2017-03-09  Jakub Jelinek  <jakub@redhat.com>
36988         PR sanitizer/79944
36989         * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
36990         BUILT_IN_SYNC*, determine the access type from the size suffix and
36991         always build a MEM_REF with that type.  Handle forgotten
36992         BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
36994         PR target/79932
36995         * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
36996         _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
36997         _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
36998         _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
36999         _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
37000         _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
37001         _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
37002         _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
37003         _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
37004         _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
37005         _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
37006         _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
37007         _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
37008         _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
37009         _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
37010         _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
37011         _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
37012         _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
37013         _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
37014         _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
37015         _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
37016         _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
37017         _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
37018         _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
37019         _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
37020         _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
37021         _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
37022         _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
37023         _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
37024         _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
37025         definitions outside of __OPTIMIZE__ guarded section.
37027         PR target/79932
37028         * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
37029         _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
37030         _mm512_packus_epi32, _mm512_maskz_packus_epi32,
37031         _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
37032         guarded section.
37034 2017-03-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
37036         * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
37037         ("vfenez<mode>"): Add missing constraints.
37039 2017-03-08  Martin Sebor  <msebor@redhat.com>
37041         PR target/79928
37042         * config/nds32/nds32.c (nds32_option_override):
37043         Fix misspelled diagnostic.
37045 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37047         PR c/79940
37048         * gimplify.c (gimplify_omp_for): Replace index var in outer
37049         taskloop statement with an artificial variable and add
37050         OMP_CLAUSE_PRIVATE clause for it.
37052 2017-03-08  Richard Biener  <rguenther@suse.de>
37054         PR tree-optimization/79955
37055         * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
37056         for accesses that are completely outside of the variable.
37058 2017-03-08  Andrew Haley  <aph@redhat.com>
37060         PR tree-optimization/79943
37061         * tree-ssa-loop-split.c (compute_new_first_bound): When
37062         calculating the new upper bound, (END-BEG) should be added, not
37063         subtracted.
37065 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37067         * config/avr/avr.md (setmemhi): Make sure match_dup
37068         operand number comes before match_scratch.
37070 2017-03-08  Richard Biener  <rguenther@suse.de>
37072         PR tree-optimization/79920
37073         * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
37074         with ncopies == 1 to ...
37075         (vect_transform_slp_perm_load): ... here.  Properly compute
37076         all element loads by iterating VF times over the group.  Do
37077         not handle ncopies (computed in a broken way) in
37078         vect_create_mask_and_perm.
37080 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37082         PR sanitizer/79904
37083         * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
37084         is a uniform vector, use uniform_vector_p return value instead of
37085         building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
37087 2017-03-07  Marek Polacek  <polacek@redhat.com>
37089         PR middle-end/79809
37090         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
37091         (alloca_call_type): Likewise.
37093 2017-03-07  Martin Liska  <mliska@suse.cz>
37095         * gcov.c (process_args): Put comment to correct location.
37097 2017-03-07  Martin Liska  <mliska@suse.cz>
37099         PR middle-end/68270
37100         * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
37101         Use array_at_struct_end_p instead of DECL_CHAIN (field).
37102         (chkp_narrow_bounds_for_field): Likewise.
37103         (chkp_parse_array_and_component_ref): Pass one more argument to
37104         call.
37106 2017-03-07  Richard Biener  <rguenther@suse.de>
37108         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
37109         preheaders.
37111 2017-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
37113         * config/i386/i386.c (ix86_local_alignment): Align most aggregates
37114         of 16 bytes and more to 16 bytes, not those of 16 bits and more.
37116 2017-03-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37118         PR c/79855
37119         * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
37120         to end of description.
37121         (PARAM_MAX_STORES_TO_MERGE): Likewise.
37123 2017-03-07  Jakub Jelinek  <jakub@redhat.com>
37125         PR rtl-optimization/79901
37126         * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
37127         ...
37128         (*avx512f_<code><mode>3<mask_name>): ... this.
37129         (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
37130         iterator instead of VI8_AVX2_AVX512BW.
37132         PR rtl-optimization/79901
37133         * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
37134         min/max expander, expand it using expand_vec_cond_expr.
37136         PR sanitizer/79897
37137         * ubsan.c (ubsan_encode_value): Call mark_addressable on the
37138         temporary.
37140 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37142         PR c++/79821
37143         * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
37144         to void * for PCH reasons.
37145         * dwarf2out.c (output_loc_operands, output_die): Cast
37146         v.val_vec.array to unsigned char *.
37148 2017-03-06  John David Anglin  <danglin@gcc.gnu.org>
37150         PR target/77850
37151         * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
37152         vector types.
37154 2017-03-06  Vladimir Makarov  <vmakarov@redhat.com>
37156         PR rtl-optimization/79571
37157         * lra-constraints.c (process_alt_operands): Calculate static
37158         reject and subtract it from overall when only addresses will be
37159         reloaded.
37161 2017-03-06  Julia Koval  <julia.koval@intel.com>
37163         PR target/79793
37164         * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
37165         incoming stack boundary to 128 for 64-bit targets.
37167 2017-03-06  Richard Biener  <rguenther@suse.de>
37169         PR tree-optimization/79894
37170         * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
37171         to NULL after folding it.
37173 2017-03-06  Richard Biener  <rguenther@suse.de>
37175         PR tree-optimization/79824
37176         * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
37177         check disabling peeling for gaps.
37179 2017-03-06  Toma Tabacu  <toma.tabacu@imgtec.com>
37181         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
37182         attributes): Document gettimeofday.
37184 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37186         * config/s390/s390.c (s390_option_override_internal): Set
37187         PARAM_MIN_VECT_LOOP_BOUND
37189 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37191         * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
37192         * config/s390/s390.md: Likewise.
37194 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37196         PR target/79812
37197         * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
37198         (<avx2_avx512>_perm<mode>): Rename to ...
37199         (avx2_perm<mode>): ... this.  Use VI8F_256 iterator instead
37200         of VI8F_256_512.
37201         (<avx512>_perm<mode>_mask): Rename to ...
37202         (avx512vl_perm<mode>_mask): ... this.  Use VI8F_256 iterator instead
37203         of VI8F_256_512.
37204         (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
37205         (avx2_perm<mode>_1<mask_name): ... this.  Use VI8F_256 iterator
37206         instead of VI8F_256_512.
37207         (avx512f_perm<mode>): New define_expand.
37208         (avx512f_perm<mode>_mask): Likewise.
37209         (avx512f_perm<mode>_1<mask_name>): New define_insn.
37210         (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
37212 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37214         * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
37215         msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
37216         if_then_else.
37217         (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
37219 2017-03-06  Martin Liska  <mliska@suse.cz>
37221         PR sanitize/79783
37222         * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
37223         when having a SSA NAME w/o VAR_DECL assigned to it.
37225 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37227         * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
37228         msa_dpsub_<su>_d): Fix MODE for vec_select.
37230 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37232         * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
37233         argument.
37234         * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
37236 2017-03-06  Richard Biener  <rguenther@suse.de>
37238         * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
37239         * plugin.c (register_plugin_info): Likewise.
37240         * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
37242 2017-03-05  Jakub Jelinek  <jakub@redhat.com>
37244         * config/i386/sse.md (sse_storehps, sse_storelps,
37245         avx_<castmode><avxsizesuffix>_<castmode>,
37246         avx512f_<castmode><avxsizesuffix>_<castmode>,
37247         avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37248         in condition that at least one operand is not a MEM.
37250 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37252         PR middle-end/79805
37253         * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37254         ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37255         ECF_NOTHROW.
37256         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37257         gimple_call_nothrow_p flag based on whether original builtin can throw.
37258         If it can, emit following stmts on the fallthrough edge.
37259         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37260         don't create new bb if inserting just debug stmts on the edge, try to
37261         insert them on the fallthru bb or just reset debug stmts.
37263 2017-03-03  Segher Boesssenkool  <segher@kernel.crashing.org>
37265         PR target/43763
37266         * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37267         restore recog_data (including the operand rtxes inside it) around
37268         the call to get_insn_template.
37270 2017-03-03  Martin Sebor  <msebor@redhat.com>
37272         PR tree-optimization/79699
37273         * context.c (context::~context): Free MPFR caches to avoid
37274         a memory leak on program exit.
37276 2017-03-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37278         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37279         Use wide_int::ulow () instead of .elt (0).
37281 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37283         * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37284         (*pushxf): Limit oF constraint to 32bit targets and add oC
37285         constraint for 64bit targets.
37286         (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37287         (*pushdf): Change rmF constraint to rmC.
37289 2017-03-03  Martin Liska  <mliska@suse.cz>
37291         * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37292         Remove unused variable.
37294 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37296         PR target/79807
37297         * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37298         is a memory operand, increase num_memory.
37299         (ix86_expand_args_builtin): Likewise.
37301 2017-03-03  Jan Hubicka  <jh@suse.cz>
37303         PR lto/79760
37304         * ipa-devirt.c (maybe_record_node): Properly handle
37305         __cxa_pure_virtual visibility.
37307 2017-03-03  Martin Liska  <mliska@suse.cz>
37309         PR tree-optimization/79803
37310         * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37311         assert.
37312         (pass_loop_prefetch::execute): Disabled optimization if an
37313         assumption about L1 cache size is not met.
37315 2017-03-03  Martin Liska  <mliska@suse.cz>
37317         PR rtl-optimization/79574
37318         * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37319         (hash_scan_set): Likewise.
37320         (dump_hash_table): Likewise.
37321         (hoist_code): Likewise.
37323 2017-03-03  Richard Biener  <rguenther@suse.de>
37325         * fixed-value.c (fixed_from_string): Restore use of elt (1)
37326         in place of uhigh ().
37327         (fixed_convert_from_real): Likewise.
37329 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37331         PR target/79514
37332         * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37334 2017-03-03  Richard Biener  <rguenther@suse.de>
37336         PR middle-end/79818
37337         * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37338         TYPE_OVERFLOW_UNDEFINED check.
37340 2017-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37342         * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37343         numbers.
37344         (vector_ae_<mode>_p): Likewise.
37345         (vector_nez_<mode>_p): Likewise.
37346         (vector_ne_v2di_p): Likewise.
37347         (vector_ae_v2di_p): Likewise.
37348         (vector_ne_<mode>_p): Likewise.
37349         * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37350         numbers.
37351         (vsx_tsqrt<mode>2_fe): Likewise.
37353 2017-03-02  Uros Bizjak  <ubizjak@gmail.com>
37355         PR target/79514
37356         * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37358 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37360         PR rtl-optimization/79780
37361         * cprop.c (one_cprop_pass): When second and further conditional trap
37362         in a single basic block is turned into an unconditional trap, turn it
37363         into a deleted note to avoid RTL verification failures.
37365 2017-03-02  Richard Biener  <rguenther@suse.de>
37367         * fold-const.c (const_binop): Use ulow () instead of elt (0).
37369 2017-03-02  Richard Biener  <rguenther@suse.de>
37371         PR tree-optimization/79345
37372         PR c++/42000
37373         * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37374         param and abort the walk, returning -1 if it is hit.
37375         (walk_aliased_vdefs): Take a limit param and pass it on.
37376         * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37377         defaulting to 0 and return a signed int.
37378         * tree-ssa-uninit.c (struct check_defs_data): New struct.
37379         (check_defs): New helper.
37380         (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37381         about uninitialized memory.
37382         * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37383         bogus uninitialized warning.
37384         (fixed_convert_from_real): Likewise.
37386 2017-03-02  Bin Cheng  <bin.cheng@arm.com>
37388         PR tree-optimization/66768
37389         * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37390         iv_use if base object can't be determined.
37392 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37394         PR tree-optimization/79345
37395         * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37396         * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37397         (get_pattern_stats): Initialize it.
37398         * genemit.c (gen_expand): Verify match_scratch numbers come after
37399         match_operand/match_dup numbers.
37400         * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37401         match_scratch numbers.
37402         * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37403         Likewise.
37404         * config/s390/s390.md (trunctdsd2): Likewise.
37406 2017-03-02  Richard Biener  <rguenther@suse.de>
37408         * wide-int.h (wide_int_storage::operator=): Implement in terms
37409         of wi::copy.
37411 2017-03-02  Richard Biener  <rguenther@suse.de>
37413         PR tree-optimization/79777
37414         * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37415         the to insert expression to sth existing.
37417 2017-03-01  Martin Sebor  <msebor@redhat.com>
37419         PR middle-end/79692
37420         * gimple-ssa-sprintf.c
37421         (directive::known_width_and_precision): New function.
37422         (format_integer): Use it.
37423         (get_mpfr_format_length): Consider the full range of precision
37424         when computing %g output with the # flag.  Set the likely byte
37425         count to 3 rather than 1 when precision is indeterminate.
37426         (format_floating): Correct the lower bound of precision.
37428 2017-03-01  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37430         * doc/invoke.texi: Document default code model for 64-bit Linux.
37432 2017-03-01  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
37434         PR target/79752
37435         * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37436         udiv rather than div since input pattern is unsigned.
37438 2017-03-01  Uros Bizjak  <ubizjak@gmail.com>
37440         * config/i386/i386.c (print_reg): Warn for values of
37441         unsupported size in integer register.
37443 2017-03-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
37445         PR target/79439
37446         * config/rs6000/predicates.md (current_file_function_operand): Do
37447         not allow self calls to be local if the function is replaceable.
37449 2017-03-01  Kelvin Nilsen  <kelvin@gcc.gnu.org>
37451         PR target/79395
37452         * config/rs6000/altivec.h (vec_ctz and others): Change the
37453         preprocessor macro that controls conditional compilation from
37454         _ARCH_PWR9 to __POWER9_VECTOR__.
37455         (vec_all_ne): Change parameterization of __altivec_scalar_pred
37456         macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37457         control (instead of _ARCH_PWR9 control) so that template
37458         definition uses power9-specific function.
37459         (vec_any_eq): Likewise.
37460         (vec_all_ne): Change macro definition to use a power9-specific
37461         expansion under #ifdef __POWER9_VECTOR__ control (instead of
37462         _ARCH_PWR9 control).
37463         (vec_any_eq) Likewise.
37464         * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37465         expansion for CMPNEF to remove support for xvcmpnesp instruction.
37466         (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37467         support for xvcmpnedp instruction.
37468         (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37469         macro expansion so that Power9 implementation of vec_all_ne does
37470         not use the AltiVec predicate framework.
37471         (VCMPNEH_P): Likewise.
37472         (VCMPNEW_P): Likewise.
37473         (VCMPNED_P): Likewise.
37474         (VCMPNEFP_P): Likewise.
37475         (VCMPNEDP_P): Likewise.
37476         (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37477         implementation of vec_any_eq to not use AltiVec predicate
37478         framework.
37479         (VCMPAEH_P): Likewise.
37480         (VCMPAEW_P): Likewise.
37481         (VCMPAED_P): Likewise.
37482         (VCMPAEFP_P): Likewise.
37483         (VCMPAEDP_P): Likewise.
37484         (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37485         BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37486         not use the AltiVec predicate framework.
37487         (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37488         of vec_any_eq to not use AltiVec predicate framework.
37489         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37490         support for predefined __POWER9_VECTOR__ macro to indicate that
37491         Power9 instruction selection is enabled.
37492         (altivec_overloaded_builtins): Remove extraneous
37493         ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37494         function argument types RS6000_BTI_bool_V16QI and
37495         RS6000_BTI_bool_V16QI.  Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37496         entry for overloaded function argument types RS6000_BTI_bool_V4SI
37497         andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB.  Remove
37498         two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37499         P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37500         Power9 for implementations of vec_cmpne.  Change the signature for
37501         all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37502         (representing vec_all_ne) to remove the previously described first
37503         argument of type RS6000_BTI_INTSI, as this was an artifact of
37504         reliance on the AltiVec predicate framework, which is no longer
37505         used in the implementation of these functions.  Add
37506         P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37507         function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37508         since, unlike the AltiVec predicate framework implementation, we
37509         do not share function descriptors between vec_alle and vec_anyeq.
37510         (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37511         set of modes that receive special treatment even when
37512         TARGET_P9_VECTOR is true.  The special treatment emits code that
37513         does not depend on Power9 instructions.
37514         * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37515         define_expand to not rely on AltiVec predicate framework.
37516         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37517         function.
37518         (vector_ne_v2di_p): Change this define_expand to not rely on
37519         AltiVec predicate framework.
37520         (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37521         function.
37522         (vector_ne_<mode>_p): Change this define_expand to not rely on
37523         AltiVec predicate framework.
37524         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37525         function.
37526         * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37527         (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37528         define_insn pattern.
37529         (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37530         define_insn pattern because the xvcmpne<VSs>. instruction is not
37531         supported.
37532         (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37533         instruction is not supported.
37535 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37537         * config/nvptx/nvptx.c: Include intl.h.
37539 2017-03-01  Martin Jambor  <mjambor@suse.cz>
37541         PR lto/78140
37542         * ipa-prop.h (ipa_bits): Removed field known.
37543         (ipa_jump_func): Removed field vr_known.  Changed fields bits and m_vr
37544         to pointers.  Adjusted their comments to warn about their sharing.
37545         (ipcp_transformation_summary): Change bits to a vector of pointers.
37546         (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37547         (ipa_get_ipa_bits_for_value): Declare.
37548         * tree-vrp.h (value_range): Mark as GTY((for_user)).
37549         * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37550         (ipa_bits_hash_table): Likewise.
37551         (ipa_vr_ggc_hash_traits): Likewise.
37552         (ipa_vr_hash_table): Likewise.
37553         (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37554         being pointers and vr_known being removed.
37555         (ipa_set_jf_unknown): Likewise.
37556         (ipa_get_ipa_bits_for_value): New function.
37557         (ipa_set_jfunc_bits): Likewise.
37558         (ipa_get_value_range): New overloaded functions.
37559         (ipa_set_jfunc_vr): Likewise.
37560         (ipa_compute_jump_functions_for_edge): Use the above functions to
37561         construct bits and vr parts of jump functions.
37562         (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37563         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37564         exist.
37565         (ipcp_grow_transformations_if_necessary): Also allocate
37566         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37567         exist.
37568         (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37569         them.  Fix too long lines.
37570         (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37571         vr_known being removed.
37572         (ipa_read_jump_function): Use new setter functions to construct bits
37573         and vr parts of jump functions or set them to NULL.
37574         (write_ipcp_transformation_info): Adjust for bits being pointers.
37575         (read_ipcp_transformation_info): Likewise.
37576         (ipcp_update_bits): Likewise.  Fix excessively long lines a trailing
37577         space.
37578         Include gt-ipa-prop.h.
37579         * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37580         being pointers.
37581         (ipcp_store_bits_results): Likewise.
37582         (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37583         Do not write to existing jump functions but use a temporary instead.
37585 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37587         PR c++/79681
37588         * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37589         attempt to use its first operand as BIT_FIELD_REF base.
37591 2017-03-01  Richard Biener  <rguenther@suse.de>
37593         PR middle-end/79721
37594         * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37595         interpolating formula in wrapping arithmetic.
37596         (chrec_apply): Convert chrec_evaluate return value to wanted type.
37598 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37600         PR tree-optimization/79734
37601         * tree-vect-generic.c (expand_vector_condition): Optimize
37602         AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37603         Handle VEC_COND_EXPR where comparison has different inner width from
37604         type's inner width.
37606 2017-02-28  Sandra Loosemore  <sandra@codesourcery.com>
37608         * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37609         markup, and similar issues.  Remove @opindex entries for things
37610         that aren't options.  Add missing -mmpy-option entries.
37612 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37614         PR tree-optimization/79737
37615         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37616         a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37617         tmpbuf[byte_size - 1].  Call natice_encode_expr with byte_size - 1
37618         instead of byte_size.  Formatting fix.
37619         (shift_bytes_in_array_right): Formatting fix.
37621 2017-02-28  Eric Botcazou  <ebotcazou@adacore.com>
37623         PR target/79749
37624         * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37625         condition on optimize for the leaf function test.
37627 2017-02-28  Martin Liska  <mliska@suse.cz>
37629         PR lto/79625
37630         * read-rtl-function.c (function_reader::handle_unknown_directive):
37631         Bail out when one uses -flto.
37633 2017-02-28  Martin Liska  <mliska@suse.cz>
37635         * common.opt: Replace space with tabular for options of <number>
37636         type.
37637         * config/i386/i386.opt: Show <number> value for
37638         -mlarge-data-threshold.
37639         * opts.c (print_filtered_help): Do not display number in hexadecimal
37640         format.
37642 2017-02-28  Martin Liska  <mliska@suse.cz>
37644         * common.opt: Fix --help=option -Q for options which are of
37645         an enum type.
37647 2017-02-28  Uros Bizjak  <ubizjak@gmail.com>
37649         * config/i386/i386.c (print_reg): Error out for values
37650         of 8-bit size in invalid integer register.
37652 2017-02-28  Martin Sebor  <msebor@redhat.com>
37654         PR tree-optimization/79691
37655         * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37657 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37659         PR target/79729
37660         * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37661         gcc_unreachable with output_operand_lossage.
37663 2017-02-28  Richard Biener  <rguenther@suse.de>
37665         PR tree-optimization/79740
37666         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37667         inserts.
37668         (visit_nary_op): Insert the nary into the hashtable if we
37669         pattern-matched sth.
37670         * tree-ssa-pre.c (eliminate_insert): Robustify.
37672 2017-02-28  Richard Biener  <rguenther@suse.de>
37674         PR middle-end/79731
37675         * fold-const.c (decode_field_reference): Reject out-of-bound
37676         accesses.
37678 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37680         * config/i386/i386.c: Include intl.h.
37681         (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37682         instead of just cond ? "..." : "...".
37683         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37684         * coverage.c (read_counts_file): Likewise.
37685         * omp-offload.c: Include intl.h.
37686         (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37687         of just cond ? "..." : "...".
37688         * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37689         of just cond ? "..." : "...".
37691 2017-02-28  Richard Earnshaw  <rearnsha@arm.com>
37693         PR target/79742
37694         * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37695         entry, if present.
37696         * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37697         'tune for' CPU name.
37698         * config/arm/arm-cpu-data.h: Regenerated.
37700 2017-02-28  Richard Biener  <rguenther@suse.de>
37702         PR tree-optimization/79732
37703         * tree-inline.c (expand_call_inline): Do not shadow var.
37705 2017-02-28  Richard Biener  <rguenther@suse.de>
37707         PR tree-optimization/79723
37708         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37709         address-space properly.
37711 2017-02-28  Thomas Schwinge  <thomas@codesourcery.com>
37713         * doc/optinfo.texi (Optimization groups): Fix option used for
37714         OPTGROUP_ALL.
37715         * doc/invoke.texi (-fopt-info): Document "omp".
37716         * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37717         (OPTGROUP_ALL): Add OPTGROUP_OMP.
37718         * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37719         * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37720         * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37722         * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP.  Adjust
37723         all users.
37724         * dumpfile.c (optgroup_options): Instead of "openmp", associate
37725         OPTGROUP_OMP with "omp".
37727 2017-02-27  Pat Haugen  <pthaugen@us.ibm.com>
37729         PR target/79544
37730         * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37731         for arithmetic shift of unsigned V2DI.
37733 2017-02-27  Claudiu Zissulescu  <claziss@synopsys.com>
37735         * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37736         arc/linux.h headers.
37737         * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37738         (LINK_SPEC): Likewise.
37739         (ARC_TLS_EXTRA_START_SPEC): Likewise.
37740         (EXTRA_SPECS): Likewise.
37741         (STARTFILE_SPEC): Likewise.
37742         (ENDFILE_SPEC): Likewise.
37743         (LIB_SPEC): Likewise.
37744         (TARGET_SDATA_DEFAULT): Likewise.
37745         (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37746         (MULTILIB_DEFAULTS): Likewise.
37747         (DWARF2_UNWIND_INFO): Likewise.
37748         * config/arc/big.h: New file.
37749         * config/arc/elf.h: Likewise.
37750         * config/arc/linux.h: Likewise.
37751         * config/arc/t-uClibc: Remove.
37753 2017-02-27  Bin Cheng  <bin.cheng@arm.com>
37755         PR tree-optimization/77536
37756         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37757         (tree_transform_and_unroll_loop): Use above function to compute the
37758         estimated niter of unrolled loop and use it when scaling profile.
37759         Also use count info rather than frequency if it's non-zero.
37760         * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37761         * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37762         (vect_transform_loop): Call above function.
37764 2017-02-27  Richard Biener  <rguenther@suse.de>
37766         PR tree-optimization/45397
37767         * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37768         * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37769         (visit_nary_op): Add pattern matching for CSEing sign-changed
37770         or truncated operations with wider ones.
37772 2017-02-27  Richard Biener  <rguenther@suse.de>
37774         PR tree-optimization/79690
37775         * tree-vect-stmts.c (vectorizable_store): Use vector type
37776         built from the DR with address-space.
37778 2017-02-26  Gerald Pfeifer  <gerald@pfeifer.com>
37780         * doc/invoke.texi (Optimize Options): Refine the description
37781         of asan-use-after-return.
37783 2017-02-25  Alan Modra  <amodra@gmail.com>
37785         PR rtl-optimization/79584
37786         * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37787         base, not ad->base_term, the reg within base.  Remove assertion
37788         that ad->base == ad->base_term.  Replace gen_int_mode using
37789         bogus mode with const0_rtx.
37791 2017-02-25  Jakub Jelinek  <jakub@redhat.com>
37793         PR middle-end/79396
37794         * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37795         FMA_EXPR like tcc_binary or tcc_unary.
37797         * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37799         PR debug/77589
37800         * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37801         bitfield.
37802         (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37803         (output_loc_operands): Handle DW_OP_call_ref and
37804         DW_OP_GNU_variable_value.
37805         (struct variable_value_struct): New type.
37806         (struct variable_value_hasher): Likewise.
37807         (variable_value_hash): New variable.
37808         (string_types): Remove.
37809         (copy_loc_descr): New function.
37810         (add_loc_descr_to_each): Clarify comment.  Use copy_loc_descr.
37811         (prepend_loc_descr_to_each): New function.
37812         (add_loc_list): Fix comment typo.  Use prepend_loc_descr_to_each
37813         instead of add_loc_descr_to_each if the first argument is single
37814         location list and the second has multiple.
37815         (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37816         (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37817         when looking for variable value which doesn't have other location info.
37818         (loc_list_from_tree): Formatting fix.
37819         (gen_array_type_die): Simplify DW_AT_string_length handling.
37820         (adjust_string_types): Remove.
37821         (gen_subprogram_die): Don't call adjust_string_types nor test/set
37822         string_types.  Call resolve_variable_values.
37823         (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37824         (resolve_addr_in_expr): Likewise.  Add A argument.
37825         (copy_deref_exprloc): Remove deref argument.  Adjust for the
37826         original expression being DW_OP_GNU_variable_value with optionally
37827         DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37828         optionally after it.
37829         (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37830         (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37831         callers.  Set remove_AT_byte_size if removing DW_AT_string_length.
37832         (variable_value_hasher::hash, variable_value_hasher::equal): New
37833         methods.
37834         (resolve_variable_value_in_expr, resolve_variable_value,
37835         resolve_variable_values, note_variable_value_in_expr,
37836         note_variable_value): New functions.
37837         (dwarf2out_early_finish): Call note_variable_value on all toplevel
37838         DIEs.
37840 2017-02-24  Jakub Jelinek  <jakub@redhat.com>
37842         PR c/79677
37843         * opts.h (handle_generated_option): Add GENERATED_P argument.
37844         * opts-common.c (handle_option): Adjust function comment.
37845         (handle_generated_option): Add GENERATED_P argument, pass it to
37846         handle_option.
37847         (control_warning_option): Pass false to handle_generated_option
37848         GENERATED_P.
37849         * opts.c (maybe_default_option): Pass true to handle_generated_option
37850         GENERATED_P.
37851         * optc-gen.awk: Likewise.
37853 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37855         * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37856         a REG, look at the REG it is a SUBREG of.
37857         (splitter for cmpeqsi_t): Ditto.
37859 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37861         * config/pa/pa.c (pa_combine_instructions): Do not share RTL.  Make
37862         the special USEs with the pattern of the insn, not the insn itself.
37864 2017-02-24  Matthew Fortune  <matthew.fortune@imgtec.com>
37866         PR target/79473
37867         * doc/invoke.texi: Document -mload-store-pairs.
37869 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37870             Sandra Loosemore  <sandra@codesourcery.com>
37872         * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37873         argument isn't a CONST_INT.
37874         (nios2_alternate_compare_const): Assert op is a CONST_INT.
37875         (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37876         (nios2_validate_compare): Bypass alternate compare logic if *op2
37877         is not a CONST_INT.
37878         (ldstwm_operation_p): Return false if first_base is not a REG or
37879         if first_offset is not a CONST_INT.
37881 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37883         * config/cris/cris.md: Use correct operand in a define_peephole2.
37885 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37887         * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37889 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37891         * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37892         this_insn if it is an INSN or JUMP_INSN.
37893         (force_offsettable): Look at base, not at addr.
37894         * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37895         on things that aren't necessarily CONST_INTs.
37897 2017-02-24  Uros Bizjak  <ubizjak@gmail.com>
37899         * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37900         -mfpmath=sse is the default also for x86-32 targets with SSE2
37901         instruction set when @option{-ffast-math} is enabled
37903 2017-02-24  Jeff Law  <law@redhat.com>
37905         PR rtl-optimizatoin/79286
37906         * ira.c (update_equiv_regs): Drop may_trap_p exception to
37907         dominance test.
37909 2017-02-24  Richard Biener  <rguenther@suse.de>
37911         PR tree-optimization/79389
37912         * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37913         debug insns.
37915 2017-02-24  Aldy Hernandez  <aldyh@redhat.com>
37917         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
37918         function comment to reflect reality.
37919         (loop_exits_before_overflow): Fix typo in function description.
37921 2017-02-24  Richard Biener  <rguenther@suse.de>
37923         PR tree-optimization/79389
37924         * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
37925         properly that a threading opportunity exists.  Detect conditional
37926         copy/constant propagation opportunities.
37928 2017-02-23  Eric Botcazou  <ebotcazou@adacore.com>
37930         * config/visium/visium.md (type): Add trap.
37931         (b): New mode attribute.
37932         (*btst): Rename into...
37933         (*btst<mode>): ...this and adjust.
37934         (*cbranchsi4_btst_insn): Rename into...
37935         (*cbranch<mode>4_btst_insn): ...this and adjust.
37936         (trap): New define_insn.
37938 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37940         PR tree-optimization/79389
37941         * ifcvt.c (struct noce_if_info): Add rev_cond field.
37942         (noce_reversed_cond_code): New function.
37943         (noce_emit_store_flag): Use rev_cond if non-NULL instead of
37944         reversed_comparison_code.  Formatting fix.
37945         (noce_try_store_flag): Test rev_cond != NULL in addition to
37946         reversed_comparison_code.
37947         (noce_try_store_flag_constants): Likewise.
37948         (noce_try_store_flag_mask): Likewise.
37949         (noce_try_addcc): Use rev_cond if non-NULL instead of
37950         reversed_comparison_code.
37951         (noce_try_cmove_arith): Likewise.  Formatting fixes.
37952         (noce_try_minmax, noce_try_abs): Clear rev_cond.
37953         (noce_find_if_block): Initialize rev_cond.
37954         (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
37955         instead of false as last argument never attempt to reverse it
37956         afterwards.
37958 2017-02-23  Bin Cheng  <bin.cheng@arm.com>
37960         PR tree-optimization/79663
37961         * tree-predcom.c (combine_chains): Process refs in reverse order
37962         only for ZERO length chains, and add explaining comment.
37964 2017-02-23  Jeff Law  <law@redhat.com>
37966         PR tree-optimization/79578
37967         * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
37968         in call to operand_equal_p.
37970 2017-01-23  Dominique d'Humieres  <dominiq@lps.ens.fr>
37972         PR target/71017
37973         * config/i386/cpuid.h: Fix another undefined behavior.
37975 2017-02-23  Richard Biener  <rguenther@suse.de>
37977         PR tree-optimization/79683
37978         * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
37979         vector types for data-refs.
37981 2017-02-23  Martin Liska  <mliska@suse.cz>
37983         * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
37985 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37987         PR middle-end/79665
37988         * internal-fn.c (get_range_pos_neg): Moved to ...
37989         * tree.c (get_range_pos_neg): ... here.  No longer static.
37990         * tree.h (get_range_pos_neg): New prototype.
37991         * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
37992         are known to be in between 0 and signed maximum inclusive, try to
37993         expand both unsigned and signed divmod and use the cheaper one from
37994         those.
37996 2017-02-22  Jeff Law  <law@redhat.com>
37998         PR tree-optimization/79578
37999         * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
38000         to compare base operands.
38002 2017-02-22  Segher Boessenkool  <segher@kernel.crashing.org>
38004         PR target/79211
38005         * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
38006         gpc_reg_operand instead of fpr_reg_operand.
38008 2017-02-22  Sameera Deshpande  <sameera.deshpande@imgtec.com>
38010         * config/mips/mips.c (mips_return_in_memory): Force FP
38011         vector types to be returned in memory for o32 ABI.
38013 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
38015         * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
38016         instead of DW_TAG_member for static data member declarations and don't
38017         set no_linkage_name for static inline data members.
38018         (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
38019         to DW_TAG_member.
38021 2017-02-22  Martin Liska  <mliska@suse.cz>
38023         * doc/invoke.texi: Replace inequality signs with square brackets
38024         for -Wnormalized.
38026 2017-02-22  Matthew Fortune  <matthew.fortune@imgtec.com>
38028         PR target/78660
38029         * lra-constraints.c (simplify_operand_subreg): Handle
38030         WORD_REGISTER_OPERATIONS targets.
38032 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
38034         PR target/70465
38035         * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
38036         and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
38037         elimination by swapping fld*.
38039 2017-02-22  Richard Biener  <rguenther@suse.de>
38041         PR tree-optimization/79673
38042         * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
38043         convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
38044         irrelevant address-space qualifiers and avoiding a
38045         ADDR_SPACE_CONVERT_EXPR from fold_convert.
38047 2017-02-22  Richard Biener  <rguenther@suse.de>
38049         PR tree-optimization/79666
38050         * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
38051         to not symbolically negate if that may introduce undefined
38052         overflow.
38054 2017-02-22  Martin Liska  <mliska@suse.cz>
38056         PR lto/79587
38057         * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
38058         * data-streamer-out.c (streamer_write_gcov_count_stream):
38059         Likewise.
38060         * value-prof.c (stream_out_histogram_value): Make assert more
38061         precise based on type of counter.
38063 2017-02-21  Uros Bizjak  <ubizjak@gmail.com>
38065         PR target/79593
38066         * config/i386/i386.md (standard_x87sse_constant_load splitter):
38067         Use nonimmediate_operand instead of memory_operand for operand 1.
38068         (float-extend standard_x87sse_constant_load splitter): Ditto.
38070 2017-02-21  Jeff Law  <law@redhat.com>
38072         PR tree-optimization/79621
38073         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
38074         blocks with edges to themselves.
38076 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38078         PR target/79633
38079         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
38080         is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
38081         Use gimple_call_builtin_p.
38083         PR target/79570
38084         * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
38085         on temporarily removed DEBUG_INSNs.
38087         PR tree-optimization/79649
38088         * tree-loop-distribution.c (classify_partition): Give up on
38089         non-generic address space loads/stores.
38091 2017-02-21  Aldy Hernandez  <aldyh@redhat.com>
38093         * doc/loop.texi (Loop manipulation): Remove nonexistent
38094         tree_ssa_loop_version from the documentation.
38095         * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
38097 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38099         PR target/79494
38100         * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
38101         make_reg_eh_region_note_nothrow_nononlocal on call_insn.
38102         * config/rs6000/rs6000.c: Include except.h.
38103         (rs6000_expand_split_stack_prologue): Call
38104         make_reg_eh_region_note_nothrow_nononlocal on the call insn.
38106 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38108         PR lto/79579
38109         * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
38110         have been analyzed.
38112 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38114         * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
38115         for backward compatibility only.
38116         * doc/invoke.texi (Option Summary): Remove all references to
38117         -fipa-cp-alignment.
38119 2017-02-21  Matthew Fortune  <matthew.fortune@imgtec.com>
38121         PR target/78660
38122         Revert:
38123         2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38125         * lra-constraints.c (curr_insn_transform): Handle
38126         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38128 2017-02-21  Martin Liska  <mliska@suse.cz>
38130         * config/i386/i386.opt: Replace -masm-dialect with -masm.
38132 2017-02-21  Thomas Schwinge  <thomas@codesourcery.com>
38134         PR translation/79638
38135         * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
38137 2017-02-21  Eric Botcazou  <ebotcazou@adacore.com>
38139         PR ada/67205
38140         * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
38141         (arm_function_ok_for_sibcall): Return false for an indirect call by
38142         descriptor if all the argument registers are used.
38143         (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
38144         alignment of the function.
38146 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38148         PR tree-optimization/61441
38149         * simplify-rtx.c (simplify_const_unary_operation): For
38150         -fsignaling-nans and sNaN operand, return NULL_RTX rather than
38151         the sNaN unmodified.
38153 2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
38155         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
38156         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
38157         instead of SYSTEM_HEADER_DIR.
38159 2017-02-20  Gerald Pfeifer  <gerald@pfeifer.com>
38160             Martin LiÅ¡ka  <mliska@suse.cz>
38162         * doc/invoke.texi (use-after-scope-direct-emission-threshold):
38163         Fix typos and grammar, use active voice, and clarify.
38165 2017-02-20  Marek Polacek  <polacek@redhat.com>
38167         PR middle-end/79537
38168         * gimplify.c (gimplify_expr): Handle unused *&&L;.
38170         PR sanitizer/79558
38171         * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
38173 2017-02-20  Jakub Jelinek  <jakub@redhat.com>
38175         PR target/79568
38176         * config/i386/i386.c (ix86_expand_builtin): Handle
38177         OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
38178         ix86_builtins_isa[fcode].isa as a requirement of those
38179         flags and any other flag in the bitmask.
38180         (ix86_init_mmx_sse_builtins): Use 0 instead of
38181         ~OPTION_MASK_ISA_64BIT as mask.
38182         * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
38183         __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
38184         __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
38185         __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
38187 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38189         PR target/78012
38190         * lra-constraints.c (split_reg): Check requested split mode
38191         is supported by the register.
38193 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38195         * lra-constraints.c (simplify_operand_subreg): Remove early
38196         return false.
38198 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38200         PR target/78660
38201         * lra-constraints.c (curr_insn_transform): Tighten condition
38202         for converting SUBREG reloads from OP_OUT to OP_INOUT.
38204 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38206         PR target/78660
38207         * lra-constraints.c (curr_insn_transform): Handle
38208         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38210 2017-02-19  Uros Bizjak  <ubizjak@gmail.com>
38212         Revert:
38213         2016-05-30  Uros Bizjak  <ubizjak@gmail.com>
38215         * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
38217 2017-02-19  Jonathan Wakely  <jwakely@redhat.com>
38219         PR c++/69523
38220         * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
38221         description.
38223 2017-02-19  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38225         * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
38226         for FMA_EXPR.
38228 2017-02-18  Jakub Jelinek  <jakub@redhat.com>
38230         * final.c (last_columnnum, override_columnnum): New variables.
38231         (final_start_function): Set last_columnnum, pass it to begin_prologue
38232         hook and pass 0 to dwarf2out_begin_prologue.
38233         (final_scan_insn): Update override_columnnum.  Pass last_columnnum
38234         to source_line debug hook.
38235         (notice_source_line): Compute last_columnnum and for debug_column_info
38236         return true on column changes.
38237         * debug.h (struct gcc_debug_hooks): Add column argument to
38238         source_line and begin_prologue hooks.
38239         (debug_nothing_int_charstar_int_bool): Remove prototype.
38240         (debug_nothing_int_int_charstar,
38241         debug_nothing_int_int_charstar_int_bool): New prototypes.
38242         (dwarf2out_begin_prologue): Add column argument.
38243         * debug.c (do_nothing_debug_hooks): Adjust source_line and
38244         begin_prologue hooks.
38245         (debug_nothing_int_charstar_int_bool): Remove.
38246         (debug_nothing_int_int_charstar,
38247         debug_nothing_int_int_charstar_int_bool): New functions.
38248         * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38249         through to dwarf2out_source_line.
38250         (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38251         (dwarf2out_source_line): Add column argument, emit it if requested.
38252         * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38253         arguments.
38254         * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38255         * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38256         * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38257         through to dwarf2out_begin_prologue.
38258         (vmsdbgout_source_line): Add column argument, pass it through to
38259         dwarf2out_source_line.
38260         * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38261         dbxout_source_line caller.
38262         (dbxout_source_line): Add column argument.
38264         * common.opt (gno-column-info, gcolumn-info): New options.
38265         * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38266         (check_die): Also test for multiple DW_AT_decl_column attributes.
38267         (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38268         DW_AT_decl_column if requested.
38269         (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38270         if requested.
38271         (gen_variable_die): Likewise.
38272         (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38273         * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38275         PR target/79569
38276         * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38277         * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38278         (ix86_handle_option): Handle OPT_m3dnowa.
38279         * doc/invoke.texi (-m3dnowa): Document.
38280         * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38281         -m3dnowa instead of -m3dnow -march=athlon.
38283         PR target/79559
38284         * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38285         instead of gcc_assert for K, r and R code checks.  Formatting fixes.
38287 2017-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
38289         PR target/79261
38290         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38291         support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38292         * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38293         generator for vsx_xxpermdi_<mode>_be.
38294         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38295         force big-endian semantics.
38296         (vsx_xxpermdi_<mode>_be): New define_expand with same
38297         implementation as previous version of vsx_xxpermdi_<mode>.
38299 2017-02-17  Jakub Jelinek  <jakub@redhat.com>
38301         PR tree-optimization/79327
38302         * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38303         variable, its initialization and use.
38305 2017-02-17  Julia Koval  <julia.koval@intel.com>
38307         * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38308         (OPTION_MASK_ISA_PKU_UNSET): New.
38309         (ix86_handle_option): Handle -mrdpid.
38310         * config/i386/cpuid.h (bit_RDPID): New.
38311         * config/i386/driver-i386.c (host_detect_local_cpu):
38312         Detect RDPID feature.
38313         * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38314         * config/i386/i386-c.c (ix86_target_macros_internal):
38315         Handle RDPID flag.
38316         * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38317         (ix86_valid_target_attribute_inner_p): Add "rdpid".
38318         (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38319         * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38320         * config/i386/i386.md (define_insn "rdpid"): New.
38321         * config/i386/i386.opt Add -mrdpid.
38322         * config/i386/immintrin.h (_rdpid_u32): New.
38324 2017-02-17  Vladimir Makarov  <vmakarov@redhat.com>
38326         PR rtl-optimization/79541
38327         * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38328         instead of transforming it into USE.
38330 2017-02-17  Segher Boessenkool  <segher@kernel.crashing.org>
38332         * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38333         If HONOR_SNANS (SFmode) force the input to a register.
38334         (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38335         (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38336         an frsp or similar insn.
38338 2017-02-17  Martin Liska  <mliska@suse.cz>
38340         PR rtl-optimization/79577
38341         * params.def (selsched-max-sched-times): Increase minimum to 1.
38343 2017-02-17  Martin Liska  <mliska@suse.cz>
38345         PR rtl-optimization/79574
38346         * gcse.c (want_to_gcse_p): Prevent integer overflow.
38348 2017-02-17  Martin Liska  <mliska@suse.cz>
38350         PR tree-optimization/79529
38351         * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38352         ssa_defined_default_def_p to handle cases which are implicitly
38353         defined.
38354         * tree-ssa.c (ssa_defined_default_def_p): New function.
38355         (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38356         which are implicitly defined.
38357         * tree-ssa.h (ssa_defined_default_def_p): Declare.
38359 2017-02-17  Richard Biener  <rguenther@suse.de>
38361         PR middle-end/79576
38362         * params.def (max-ssa-name-query-depth): Limit to 10.
38364 2017-02-17  Richard Biener  <rguenther@suse.de>
38366         PR tree-optimization/79552
38367         * tree-ssa-structalias.c (visit_loadstore): Properly verify
38368         default defs.
38370 2017-02-17  Richard Biener  <rguenther@suse.de>
38372         PR bootstrap/79567
38373         * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38375 2017-02-17  Marek Polacek  <polacek@redhat.com>
38377         PR middle-end/79536
38378         * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38379         (fold_negate_expr): New wrapper.
38381 2017-02-16  Sandra Loosemore  <sandra@codesourcery.com>
38383         * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]: 
38384         Correct terminology and de-emphasize pre-standard behavior.
38386 2017-02-16  Alan Modra  <amodra@gmail.com>
38388         PR rtl-optimization/79286
38389         * ira.c (def_dominates_uses): New function.
38390         (update_equiv_regs): Don't create an equivalence for insns that
38391         may trap where the register def does not dominate the use.
38393 2017-02-16  Vladimir Makarov  <vmakarov@redhat.com>
38395         PR rtl-optimization/78127
38396         * lra.c (lra): Call lra_eliminate before finish the loop after
38397         lra_constraint.
38399 2017-02-16  Richard Biener  <rguenther@suse.de>
38401         * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38402         isl/isl_val.h.
38403         * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38404         (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38405         * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38406         (isl_val_int_from_wi): New function.
38407         (extract_affine_gmp): Rename to ...
38408         (extract_affine_wi): ... this, take a widest_int.
38409         (extract_affine_int): Just wrap extract_affine_wi.
38410         (add_param_constraints): Use isl_val_int_from_wi.
38411         (add_loop_constraints): Likewise, and extract_affine_wi.
38413 2017-02-15  Jeff Law  <law@redhat.com>
38415         PR middle-end/79521
38416         * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38417         ira_init_register_move_cost_if_necessary.
38419 2017-02-15  Martin Sebor  <msebor@redhat.com>
38421         PR middle-end/32003
38422         * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38423         removed in a prior commit.
38425 2017-02-15  Bin Cheng  <bin.cheng@arm.com>
38427         PR tree-optimization/79347
38428         * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38429         counters during peeling.
38431 2017-02-15  Thomas Schwinge  <thomas@codesourcery.com>
38433         * Makefile.in (site.exp): Remove "set ISLVER".
38435 2017-02-15  Jakub Jelinek  <jakub@redhat.com>
38437         PR target/79487
38438         * real.c (real_from_integer): Call real_convert even for decimal.
38440 2017-02-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38442         PR target/79421
38443         * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38445 2017-02-14  Andrew Pinski  <apinski@cavium.com>
38447         * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38448         cores and change the partno/implementer to be correct.
38449         (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38450         the 'B" as the implementer.
38451         * config/aarch64/aarch64-tune.md: Regenerate.
38453 2017-02-14  Carl Love  <cel@us.ibm.com>
38455         * config/rs6000/rs6000.c: Add case statement entry to make the
38456         xvcvuxdsp built-in argument unsigned.
38457         * config/rs6000/vsx.md: Fix the source and return operand types so they
38458         match the instruction definitions from the ISA document.  Fix typo
38459         in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38460         statement.
38462 2017-02-14  Vladimir Makarov  <vmakarov@redhat.com>
38464         PR target/79282
38465         * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38466         member early_clobber_alts.
38467         * lra-lives.c (reg_early_clobber_p): New.
38468         (process_bb_lives): Use it.
38469         * lra.c (new_insn_reg): New arg early_clobber_alts.  Use it.
38470         (debug_operand_data): Initialize early_clobber_alts.
38471         (setup_operand_alternative): Set up early_clobber_alts.
38472         (collect_non_operand_hard_regs): Ditto.  Pass early clobber
38473         alternatives to new_insn_reg.
38474         (add_regs_to_insn_regno_info): Add arg early_clobber_alts.  Use
38475         it.
38476         (lra_update_insn_regno_info): Pass the new arg.
38478 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38480         PR middle-end/79505
38481         * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38482         (new_oacc_loop_raw): Don't clear already cleared fields.
38484         PR target/79481
38485         * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38486         _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38487         _mm512_prefetch_i64gather_ps): New inline functions and macros.
38489 2017-02-14  Uros Bizjak  <ubizjak@gmail.com>
38491         PR target/79495
38492         * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38494 2017-02-14  H.J. Lu  <hongjiu.lu@intel.com>
38496         PR target/79498
38497         * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38498         the extra instruction to the right place to store 128-bit constant
38499         when needed.
38501 2017-02-14  Martin Sebor  <msebor@redhat.com>
38503         PR middle-end/79448
38504         * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38505           warning for strings of unknown length.
38507 2017-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
38509         * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38511 2017-02-14  Jeff Law  <law@redhat.com>
38513         PR target/79404
38514         * ira-costs.c (scan_one_insn): Initialize register move costs
38515         for pseudos seen in USE/CLOBBER insns.
38517         PR tree-optimization/79095
38518         * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38519         if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38520         (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38521         if the operands are known to be not equal, then the resulting range
38522         is ~[0,0].
38523         (intersect_ranges): If the new range is ~[0,0] and the old range is
38524         wide, then prefer ~[0,0].
38525         * tree-vrp.c (overflow_comparison_p_1): New function.
38526         (overflow_comparison_p): New function.
38527         * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38528         if NAME is used in an overflow test.
38529         (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38530         overflow check that can be expressed as an equality test, then adjust
38531         ops to be that equality test.
38533 2017-02-14  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38535         * config/s390/s390-builtin-types.def: Remove flags argument.
38536         * config/s390/s390.c (s390_init_builtins): Likewise.
38538 2017-02-14  Martin Liska  <mliska@suse.cz>
38540         * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38541         vector.  Fix trailing white spaces.
38543 2017-02-14  James Greenhalgh  <james.greenhalgh@arm.com>
38545         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38546         HFmode.
38548 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38550         PR rtl-optimization/68664
38551         * config/arm/arm.c (arm_sched_can_speculate_insn):
38552         New function.  Declare prototype.
38553         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38555 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38557         PR rtl-optimization/68664
38558         * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38559         New function.
38560         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38562 2017-02-14  Amit Pawar  <amit.pawar@amd.com>
38564         * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38565         max skip bytes for function, loop and jump.
38567 2017-02-14  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38569         * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38570         ABS_EXPR for gimple dump.
38572 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38574         PR target/79462
38575         * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38577         PR tree-optimization/79408
38578         * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38579         case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38580         (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38581         also if rhs1 is INTEGER_CST.
38583 2017-02-14  Richard Biener  <rguenther@suse.de>
38585         PR middle-end/79432
38586         * tree-into-ssa.c (insert_phi_nodes): When the function can
38587         have abnormal edges rewrite SSA names with broken use-def
38588         dominance out of SSA and register them for PHI insertion.
38590 2017-02-13  Martin Sebor  <msebor@redhat.com>
38592         PR middle-end/79496
38593         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38594         clearing info.nowrite flag when snprintf size argument is a range.
38596 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38598         * cprop.c (cprop_jump): Add missing space in string literal.
38599         * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38600         (get_constraint_for_component_ref): Likewise.
38601         * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38602         * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38603         * lra-constraints.c (process_alt_operands): Likewise.
38604         * ipa-inline.c (inline_small_functions): Likewise.
38605         * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38606         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38607         * trans-mem.c (diagnose_tm_1_op): Likewise.
38608         * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38609         (grid_parallel_clauses_gridifiable): Likewise.
38611         * config/nvptx/mkoffload.c (process): Add space in between
38612         , and %d.
38614         * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38615         "MOD4_SSE_REGS" and "ALL_REGS".
38617         * spellcheck.c (test_data): Add , in between "foo" and "food".
38619 2017-02-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38621         PR target/79449
38622         * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38623         boundary crossing check and subsequent code generation agree.
38625 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38627         * config/aarch64/aarch64.c (has_memory_op): Delete.
38628         (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38629         has_memory_op.
38631 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38633         PR rtl-optimization/79388
38634         PR rtl-optimization/79450
38635         * combine.c (distribute_notes): When removing TEM_INSN for which
38636         corresponding dest has last value recorded, invalidate that last
38637         value.
38639 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38641         * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38642         of explicit '@'.  Add missing assembly comment marker on branch costs
38643         printout.
38645 2017-02-13  Nathan Sidwell  <nathan@acm.org>
38647         * gengtype-lex.l (<in_struct>): Add '/'.
38649 2017-02-13  Martin Liska  <mliska@suse.cz>
38651         PR c/79471
38652         * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38654 2017-02-13  Richard Biener  <rguenther@suse.de>
38656         * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38657         Remove.
38658         * configure: Re-generate.
38659         * config.in: Likewise.
38660         * graphite-dependences.c: Simplify as if
38661         HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38662         * graphite-isl-ast-to-gimple.c: Likewise.
38663         * graphite-optimize-isl.c: Likewise.
38664         * graphite-poly.c: Likewise.
38665         * graphite-sese-to-poly.c: Likewise.
38666         * graphite.h: Likewise.
38667         * toplev.c: Include isl/version.h and use isl_version () for
38668         printing the ISL version.
38669         * doc/install.texi: Update ISL requirement.
38671 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38673         * doc/standards.texi (Standards): Update reference to
38674         Objective-C 2.0.
38676 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38678         * doc/extend.texi (Named Address Spaces): sourceware.org now
38679         defaults to https.
38680         * doc/install.texi (Binaries): Ditto.
38681         (Specific): Ditto.
38683 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38685         * doc/cpp.texi: Replace "stringify"/"stringification" with C
38686         standard terminology "stringize"/"stringizing" throughout.
38687         * doc/cppinternals.texi: Likewise.
38689 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38691         * doc/extend.texi: Fix some spelling mistakes and typos.
38692         * doc/invoke.texi: Likewise.
38694 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38696         PR ipa/79224
38697         * params.def (inline-min-speedup) Change from 10 to 8.
38699 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38701         * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38702         4.5.
38704 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38706         PR ipa/79224
38707         * ipa-inline-analysis.c (get_minimal_bb): New function.
38708         (record_modified): Use it.
38709         (remap_edge_change_prob): Handle also ancestor functions.
38711 2017-02-11  Gerald Pfeifer  <gerald@pfeifer.com>
38713         * doc/contrib.texi (Contributors): Remove broken link into
38714         the Mauve CVS repository.
38716 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38718         PR middle-end/79454
38719         * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38720         result computation whenever lhs doesn't have vector mode, not
38721         just when it has BLKmode.
38723 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38725         * doc/makefile.texi (profiledbootstrap): Refer to the
38726         installation instructions only in textual form.
38728 2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38730         PR target/79295
38731         * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38733 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38735         * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38736         (Specific): Update mingw-w64 reference.
38737         (Binaries): Ditto.
38738         (Specific): Remove broken link to Renesas RX processor.
38740 2017-02-10  Richard Biener  <rguenther@suse.de>
38742         * toplev.c (process_options): Do not mention obsolete graphite
38743         options when printing sorry message about missing graphite support.
38744         Mention -floop-nest-optimize.
38746 2017-02-10  Christophe Lyon  <christophe.lyon@linaro.org>
38748         * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38749         (vtst_p16): Likewise.
38750         (vtstq_p8): Likewise.
38751         (vtstq_p16): Likewise.
38752         (vtst_p64): New.
38753         (vtstq_p64): Likewise.
38754         * config/arm/arm_neon.h (vgetq_lane_p64): New.
38755         (vset_lane_p64): New.
38756         (vsetq_lane_p64): New.
38758 2017-02-10  Jakub Jelinek  <jakub@redhat.com>
38760         PR tree-optimization/79411
38761         * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38762         stmt operands are SSA_NAMEs used in abnormal phis.
38763         (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38764         phis.
38766 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38768         PR ipa/70795
38769         * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38770         flag if needed.
38772 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38774         * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38776 2017-02-09  Jakub Jelinek  <jakub@redhat.com>
38778         * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38779         to avoid warning.
38781         PR c/79413
38782         * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38783         not arbitrary TREE_CONSTANT.
38785         PR c/79431
38786         * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38787         "omp declare target link" attribute unless is_global_var.
38788         * omp-offload.c (find_link_var_op): Likewise.
38790 2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
38791             Chung-Lin Tang  <cltang@codesourcery.com>
38793         * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38794         OMP_CLAUSE_TILE.
38795         (gimplify_adjust_omp_clauses): Don't delete TILE.
38796         (gimplify_omp_for): Deal with TILE.
38797         * internal-fn.c (expand_GOACC_TILE): New function.
38798         * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38799         (GOACC_TILE): New.
38800         * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38801         (expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
38802         element fields.
38803         (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38804         avoid DIV for outermost collapse var.
38805         (expand_oacc_for): Insert tile element loop as needed.  Adjust.
38806         Remove out of date comments, fix whitespace.
38807         * omp-general.c (omp_extract_for_data): Deal with tiling.
38808         * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38809         adjust OLF_DIM_BASE value.
38810         (struct omp_for_data): Add tiling field.
38811         * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38812         (lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
38813         for auto loops.  Remove default auto determining, moved to
38814         oacc_loop_fixed_partitions.
38815         * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38816         stmts, add e_mask field.
38817         (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38818         (oacc_thread_numbers): Use oacc_dim_call.
38819         (oacc_xform_tile): New.
38820         (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38821         (finish_oacc_loop): Adjust for ifns vector.
38822         (oacc_loop_discover_walk): Append loop abstraction sites to list,
38823         add case for GOACC_TILE fns.
38824         (oacc_loop_xform_loop): Delete.
38825         (oacc_loop_process): Iterate over call list directly, and add
38826         handling for GOACC_TILE fns.
38827         (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38828         dump partitioning.
38829         (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38830         vector partitioning to outer loops.  Assign 2 partitions to loops
38831         when available. Add TILE handling.
38832         (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38833         (execite_oacc_device_lower): Process GOACC_TILE fns,
38834         ignore unknown specs.
38835         * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38836         * tree.c (omp_clause_num_ops): Adjust TILE ops.
38837         * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38839 2017-02-09  Gerald Pfeifer  <gerald@pfeifer.com>
38841         * configure.ac (ACX_BUGURL): Update.
38842         * configure: Regenerate.
38844 2017-02-09  Richard Biener  <rguenther@suse.de>
38846         PR tree-optimization/69823
38847         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38848         Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.
38850 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38852         * config/arc/arc-c.def: Add __NPS400__ definition.
38853         * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38854         (TARGET_NPS400): Define.
38856 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38858         * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38859         file.
38860         (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38861         pointer, arch_info.
38862         (arc_cpu_types): Fill the arch_info field with a pointer into the
38863         arc_arch_types table.
38864         (arc_selected_cpu): Declare.
38865         * config/arc/arc.c (arc_selected_cpu): Make global.
38866         (arc_selected_arch): Delete.
38867         (arc_base_cpu): Delete.
38868         (arc_override_options): Remove references to deleted variables,
38869         update access to arch information.
38870         (ARC_OPT): Update access to arch information.
38871         (ARC_OPTX): Likewise.
38872         * config/arc/arc.h (arc_base_cpu): Remove declaration.
38873         (TARGET_ARC600): Update access to arch information.
38874         (TARGET_ARC601): Likewise.
38875         (TARGET_ARC700): Likewise.
38876         (TARGET_EM): Likewise.
38877         (TARGET_HS): Likewise.
38878         * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38879         information.
38881 2017-02-08  Pat Haugen  <pthaugen@us.ibm.com>
38883         PR target/78604
38884         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38885         condition/operands for integer GE/LE/GEU/LEU operations.
38887 2017-02-08  Segher Boessenkool  <segher@kernel.crashing.org>
38889         PR translation/79397
38890         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38891         of AltiVec.
38893 2017-02-08  Martin Jambor  <mjambor@suse.cz>
38895         PR ipa/79375
38896         * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38897         whether allocation happened.
38898         (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38899         nothing was allocated.
38901 2017-02-08  Jakub Jelinek  <jakub@redhat.com>
38903         PR tree-optimization/79408
38904         * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38905         constant, but SSA_NAME with a known integer range, use the minimum
38906         of that range instead of op1 to determine if modulo can be replaced
38907         with its first operand.
38909 2016-02-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38911         * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38913 2017-02-08  Richard Biener  <rguenther@suse.de>
38915         PR tree-optimization/71824
38916         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38917         Check all loops contained in the merged region.
38919 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38921         * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
38923 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38925         * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
38926         (thunderxt88): Likewise.
38927         (thunderxt81): Disable LSE and change v8.1 to v8.
38928         (thunderxt83): Likewise.
38930 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38931             Richard Biener  <rguenther@suse.de>
38933         PR middle-end/79399
38934         * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
38935         type from int to size_t.
38936         * ira-costs.c (struct_costs_size): Change type from int to size_t.
38938 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38940         PR rtl-optimization/79386
38941         * cprop.c (bypass_conditional_jumps): Initialize
38942         bypass_last_basic_block already before splitting bbs after
38943         unconditional traps...
38944         (bypass_conditional_jumps): ... rather than here.
38946         PR target/79299
38947         * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
38948         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
38949         *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
38950         fix -masm=intel patterns.
38952 2017-02-07  Richard Biener  <rguenther@suse.de>
38954         PR tree-optimization/79256
38955         PR middle-end/79278
38956         * builtins.c (get_object_alignment_2): Use min_align_of_type
38957         to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
38958         and ADJUST_FIELD_ALIGN.
38960         * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
38961         type parameter.
38962         * doc/tm.texi: Regenerate.
38963         * stor-layout.c (layout_decl): Adjust.
38964         (update_alignment_for_field): Likewise.
38965         (place_field): Likewise.
38966         (min_align_of_type): Likewise.
38967         * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
38968         * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
38969         * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
38970         * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
38971         * config/frv/frv.c (frv_adjust_field_align): Likewise.
38972         * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
38973         * config/i386/i386.c (x86_field_alignment): Likewise.
38974         * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
38975         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
38976         * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
38977         * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
38978         * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
38979         * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
38980          Likewise.
38982         Revert
38983         2017-01-30  Richard Biener  <rguenther@suse.de>
38985         PR tree-optimization/79256
38986         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
38987         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
38988         alignment on TYPE.
38990 2017-02-07  Toma Tabacu  <toma.tabacu@imgtec.com>
38992         * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
38993         argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
38994         builtins to SImode and emit a zero-extend, if necessary.
38996 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38998         * docs/invoke.texi (RISC-V Options): Alphabetize.
39000 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
39002         * doc/invoke.texi (RISC-V Options): Use two spaces to separate
39003         options.
39005 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
39007         * config/riscv/riscv.c: New file.
39008         * common/config/riscv/riscv-common.c: Likewise.
39009         * config.gcc: Likewise.
39010         * config/riscv/constraints.md: Likewise.
39011         * config/riscv/elf.h: Likewise.
39012         * config/riscv/generic.md: Likewise.
39013         * config/riscv/linux.h: Likewise.
39014         * config/riscv/multilib-generator: Likewise.
39015         * config/riscv/peephole.md: Likewise.
39016         * config/riscv/pic.md: Likewise.
39017         * config/riscv/predicates.md: Likewise.
39018         * config/riscv/riscv-builtins.c: Likewise.
39019         * config/riscv/riscv-c.c: Likewise.
39020         * config/riscv/riscv-ftypes.def: Likewise.
39021         * config/riscv/riscv-modes.def: Likewise.
39022         * config/riscv/riscv-opts.h: Likewise.
39023         * config/riscv/riscv-protos.h: Likewise.
39024         * config/riscv/riscv.h: Likewise.
39025         * config/riscv/riscv.md: Likewise.
39026         * config/riscv/riscv.opt: Likewise.
39027         * config/riscv/sync.md: Likewise.
39028         * config/riscv/t-elf-multilib: Likewise.
39029         * config/riscv/t-linux: Likewise.
39030         * config/riscv/t-linux-multilib: Likewise.
39031         * config/riscv/t-riscv: Likewise.
39032         * configure.ac: Likewise.
39033         * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
39034         Waterman as RISC-V maintainers.
39035         * doc/install.texi: Add RISC-V entries.
39036         * doc/invoke.texi: Add RISC-V options section.
39037         * doc/md.texi: Add RISC-V constraints section.
39038         * configure: Regenerated.
39040 2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
39042         PR target/66144
39043         * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
39044         false values to be constant vectors with all 0 or all 1 bits set.
39045         (vcondu<mode><mode>): Likewise.
39046         * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
39047         predicate.
39048         (fpmask_comparison_operator): Update comment.
39049         (vecint_comparison_operator): New predicate.
39050         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
39051         vector conditionals when the true and false values are constant
39052         vectors with all 0 bits or all 1 bits set.
39054 2017-02-06  Martin Sebor  <msebor@redhat.com>
39056         PR  tree-optimization/79376
39057         * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
39059 2017-02-06  Uros Bizjak  <ubizjak@gmail.com>
39061         * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
39062         explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
39063         to simplify split condition.
39065 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39067         * omp-expand.c (oxpand_omp_atomic_fetch_op,
39068         expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
39069         false.
39071 2017-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
39073         PR rtl-optimization/68664
39074         * target.def (can_speculate_insn): New hook.
39075         * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
39076         * doc/tm.texi: Regenerate.
39077         * sched-rgn.c (can_schedule_ready_p): Use the new hook.
39078         * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
39079         (rs6000_sched_can_speculate_insn): New function.
39081 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39083         PR tree-optimization/79284
39084         * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
39085         * tree-vect-stmts.c (vect_get_vec_def_for_operand,
39086         vectorizable_mask_load_store, vectorizable_operation,
39087         vect_is_simple_cond, get_same_sized_vectype): Use it instead
39088         of comparing TREE_CODE of a type against BOOLEAN_TYPE.
39089         * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
39090         vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
39091         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
39092         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
39093         Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
39094         is_gimple_assign (stmt).  Replace another such test with
39095         is_gimple_assign (stmt).
39097 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39099         PR target/78883
39100         * config/avr/avr.c (rtl-iter.h): Include it.
39101         (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
39102         (avr_legitimate_combined_insn): ...and implementation.
39104 2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39106         * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
39107         * config/s390/s390.c (s390_const_operand_ok)
39108         (s390_canonicalize_comparison, s390_extract_part)
39109         (s390_single_part, s390_contiguous_bitmask_nowrap_p)
39110         (s390_contiguous_bitmask_p, s390_rtx_costs)
39111         (legitimize_pic_address): Likewise.
39112         * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
39113         * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
39114         ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
39115         ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
39116         * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
39118 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39120         * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
39121         REGNO($0) == REGNO($1).
39123 2017-02-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39125         * config/s390/linux.h(SIZE_TYPE): Add comment.
39127 2017-02-06  Julian Brown  <julian@codesourcery.com>
39128             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39129             Virendra Pathak  <virendra.pathak@broadcom.com>
39131         * config/aarch64/aarch64-cores.def: Change the scheduler
39132         to Thunderx2t99.
39133         * config/aarch64/aarch64.md: Include thunderx2t99.md.
39134         * config/aarch64/thunderx2t99.md: New file.
39136 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39138         * doc/standards.texi (Go Language): Update link to language
39139         standard.
39141 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39143         * tree-eh.c (lower_resx): Sanitize profile.
39144         (cleanup_empty_eh_move_lp): Likewise.
39146 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39148         PR tree-ssa/79347
39149         * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
39150         ELSE_PROB.
39151         * cfgloopmanip.h (loop_version): Update prototype.
39152         * modulo-sched.c (sms_schedule): Update call of loop_version.
39153         * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
39154         * tree-parloops.c (gen_parallel_loop): Likewise.
39155         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
39156         * tree-ssa-loop-split.c (split_loop): Likewise.
39157         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
39158         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
39160 2017-02-05  Martin Liska  <mliska@suse.cz>
39162         PR bootstrap/78985
39163         * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
39164         variable to NULL.
39165         (print_operand_address): Initialize a struct to zero.
39167 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39169         * doc/contrib.texi (Contributors): Refer to Hans Boehm's
39170         garbage collector only in textual form.
39172 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39174         * doc/extend.texi (x86 specific memory model extensions for
39175         transactional memory): Simplify a phrase.
39177 2017-02-05  Eric Botcazou  <ebotcazou@adacore.com>
39179         PR target/79353
39180         * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
39181         'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
39182         (atomic_storedi_1): Likewise.
39184 2017-02-04  Jakub Jelinek  <jakub@redhat.com>
39186         PR tree-optimization/79338
39187         * tree-parloops.c (gather_scalar_reductions): Don't call
39188         vect_analyze_loop_form for loop->inner before destroying loop's
39189         loop_vinfo.
39191 2017-02-03  Martin Sebor  <msebor@redhat.com>
39193         PR tree-optimization/79327
39194         * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
39195         when precision has resulted in leading zeros.
39196         (format_integer): Adjust the likely counter to assume an unknown
39197         argument that may be zero is non-zero.
39199 2017-02-03  Jason Merrill  <jason@redhat.com>
39201         PR c++/78689
39202         * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
39203         avoid copying non-taken branch.
39205 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39207         PR tree-optimization/79340
39208         * tree-vect-loop.c (vectorizable_reduction): Release
39209         vec_defs elements after safe_splicing them into other vectors.
39210         Formatting fixes.
39212         PR tree-optimization/79327
39213         * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
39214         true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
39215         dirtype.
39216         (format_integer): Use wide_int_to_tree instead of build_int_cst
39217         + to_?hwi.  If argmin is NULL, just set argmin and argmax to
39218         TYPE_{MIN,MAX}_VALUE of argtype.  Simplify and fix computation
39219         of shortest and longest sequence.
39221 2017-02-03  Uros Bizjak  <ubizjak@gmail.com>
39223         * config/i386/i386.c (dimode_scalar_chain::convert_reg):
39224         Use pextrd for TARGET_SSE4_1 when creating scalar copy.
39226 2017-02-03  Walter Lee  <walt@tilera.com>
39228         PR target/78862
39229         * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
39230         after initial stackframe link reg save.
39231         * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
39233 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39235         PR target/79354
39236         * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
39237         wu for stxssp alternative.
39239 2017-02-03  Martin Sebor  <msebor@redhat.com>
39241         PR tree-optimization/79352
39242         * gimple-fold.c (get_range_strlen): Add argument.
39243         (get_range_strlen): Change return type to bool.
39244         (get_maxval_strlen): Pass in a dummy argument.
39245         * gimple-fold.h (get_range_strlen): Change return type to bool.
39246         * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39247         * tree.h (array_at_struct_end_p): Add argument.
39248         * tree.c (array_at_struct_end_p): Handle it.
39250 2017-02-03  Martin Liska  <mliska@suse.cz>
39252         PR lto/66295
39253         * multiple_target.c (create_dispatcher_calls): Redirect edge
39254         from a caller of a dispatcher.
39255         (expand_target_clones): Make the clones local.
39256         (ipa_target_clone): Do both target clones and resolvers.
39257         (ipa_dispatcher_calls): Remove the pass.
39258         (pass_dispatcher_calls::gate): Likewise.
39259         (make_pass_dispatcher_calls): Likewise.
39260         * passes.def (pass_target_clone): Put as very first IPA early
39261         pass.
39263 2017-02-03  Martin Liska  <mliska@suse.cz>
39265         * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39266         in case of a function with ifunc attribute.
39268 2017-02-03  Martin Liska  <mliska@suse.cz>
39270         * cgraph.c (cgraph_node::dump): Dump function version info.
39271         * symtab.c (symtab_node::dump_base): Add missing new line.
39273 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39275         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39276         (ifcombine_ifandif): Use it.
39278 2017-02-03  Martin Liska  <mliska@suse.cz>
39280         * doc/invoke.texi: Document default value for
39281         use-after-scope-direct-emission-threshold.
39283 2017-02-03  Martin Liska  <mliska@suse.cz>
39285         PR tree-optimization/79339
39286         * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39287         (format_floating): Likewise.
39289 2017-02-03  Martin Liska  <mliska@suse.cz>
39291         PR ipa/79337
39292         * ipa-prop.c (ipa_node_params_t::insert): Remove current
39293         implementation.
39294         (ipa_node_params_t::remove): Likewise.
39295         * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39296         initialization from removed ipa_node_params_t::insert.
39297         (ipa_node_params::~ipa_node_params): Move from removed
39298         ipa_node_params_t::release.
39299         * symbol-summary.h (symbol_summary::m_released): New member.
39300         Do not release a summary twice.  Do not allow to call finalizer
39301         for types of a summary that live in GGC memory.
39303 2017-02-02  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39305         * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39306         cmp_branch fusion.
39308 2017-02-02  Martin Sebor  <msebor@redhat.com>
39310         PR middle-end/79275
39311         * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39312         (format_string): Tighten up the range of output for non-constant
39313         strings and correct the expected range for wide non-constant strings.
39315 2017-02-02  Martin Sebor  <msebor@redhat.com>
39317         * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39319         PR middle-end/32003
39320         * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39321         index.
39322         (-fdump-tree-@var): Add to index and document how to come up
39323         with pass-specific option and dump file names.
39324         (-fdump-passes): Clarify where to look for output.
39326 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39328         PR middle-end/77445
39329         * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39330         statistics of the analyzed path; allow threading for speed when
39331         any of BBs along the path are optimized for speed.
39333 2017-02-02  Eric Botcazou  <ebotcazou@adacore.com>
39335         PR middle-end/78468
39336         * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39337         settings of the virtual registers.
39339         Revert again
39340         2016-08-23  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39342         * explow.c (get_dynamic_stack_size): Take known alignment of stack
39343         pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39344         needed.
39346 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39348         * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39349         ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39351 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39353         * config/s390/s390.md: Add missing comments with the expanded
39354         mnemonics.
39355         * config/s390/vector.md: Likewise.
39356         * config/s390/vx-builtins.md: Likewise.
39358 2017-02-02  Jakub Jelinek  <jakub@redhat.com>
39360         PR target/79197
39361         * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39362         (fixuns_trunc<mode>di2): ... this, remove previous expander.  Put all
39363         conditions on a single line.
39365 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39367         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39368         __S390_VX__ to __VX__.
39370 2017-02-01  Andrew Pinski  <apinski@cavium.com>
39372         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39373         stmt_info to record_stmt_cost.
39374         (vect_get_known_peeling_cost): Pass stmt_info if known to
39375         record_stmt_cost.
39376         * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39377         cpu_vector_cost field into
39378         scalar_int_stmt_cost and scalar_fp_stmt_cost.  Split vec_stmt_cost
39379         field into vec_int_stmt_cost and vec_fp_stmt_cost.
39380         * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39381         splitting of scalar_stmt_cost and vec_stmt_cost.
39382         (thunderx_vector_cost): Likewise.
39383         (cortexa57_vector_cost): LIkewise.
39384         (exynosm1_vector_cost): Likewise.
39385         (xgene1_vector_cost): Likewise.
39386         (thunderx2t99_vector_cost): Improve after the splitting of the two
39387         fields.
39388         (aarch64_builtin_vectorization_cost): Update for the splitting of
39389         scalar_stmt_cost and vec_stmt_cost.
39391 2017-02-01  Torvald Riegel  <triegel@redhat.com>
39392             Richard Henderson  <rth@redhat.com>
39394         * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39395         conditional on existance of a fast atomic load.
39396         * optabs-query.c (can_atomic_load_p): New function.
39397         * optabs-query.h (can_atomic_load_p): Declare it.
39398         * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39399         no fast atomic load is available for the particular size of access.
39400         (expand_atomic_compare_and_swap): Likewise.
39401         (expand_atomic_load): Likewise.
39402         (expand_atomic_store): Likewise.
39403         (expand_atomic_fetch_op): Likewise.
39404         * testsuite/lib/target-supports.exp
39405         (check_effective_target_sync_int_128): Remove x86 because it provides
39406         no fast atomic load.
39407         (check_effective_target_sync_int_128_runtime): Likewise.
39409 2017-02-01  Richard Biener  <rguenther@suse.de>
39411         * graphite.c: Include tree-vectorizer.h for find_loop_location.
39412         (graphite_transform_loops): Provide opt-info for optimized nests.
39413         * tree-parloop.c (parallelize_loops): Provide opt-info for
39414         parallelized loops.
39416 2017-02-01  Richard Biener  <rguenther@suse.de>
39418         PR middle-end/79315
39419         * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39420         was not set before.
39422 2017-02-01  Richard Biener  <rguenther@suse.de>
39424         PR tree-optimization/71824
39425         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39426         Verify the loops are valid in the merged SESE region.
39427         (scop_detection::can_represent_loop_1): Check analyzing the
39428         evolution of the number of iterations in the region succeeds.
39430 2017-01-31  Ian Lance Taylor  <iant@golang.org>
39432         * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39433         REG_ARGS_SIZE note to 32-bit push insns and call insn.
39435 2017-01-31  David Malcolm  <dmalcolm@redhat.com>
39437         PR preprocessor/79210
39438         * input.c (get_substring_ranges_for_loc): Replace line_width
39439         assertion with error-handling.
39441 2017-01-31  Richard Biener  <rguenther@suse.de>
39443         PR tree-optimization/77318
39444         * graphite-sese-to-poly.c (extract_affine): Fix assert.
39445         (create_pw_aff_from_tree): Take loop parameter.
39446         (add_condition_to_pbb): Pass loop of the condition to
39447         create_pw_aff_from_tree.
39449 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39451         * config/s390/s390.c (s390_asan_shadow_offset): New function.
39452         (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39454 2017-01-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
39456         PR target/78597
39457         PR target/79038
39458         * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39459         no longer used.
39460         (convert_int_to_float128): Likewise.
39461         * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39462         (convert_int_to_float128): Likewise.
39463         * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39464         (UNSPEC_IEEE128_CONVERT): Likewise.
39465         (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39466         rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39467         Use local variables for IBM extended format.
39468         (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39469         (fix_trunc<mode>si2_fprs): Likewise.
39470         (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39471         (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39472         (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39473         to know that we can now have integers of all sizes in vector
39474         registers.
39475         (fix<uns>_<mode>di2_hw): Likewise.
39476         (float<uns>_<mode>si2_hw): Likewise.
39477         (fix_<mode>si2_hw): Likewise.
39478         (fixuns_<mode>si2_hw): Likewise.
39479         (float<uns>_<mode>di2_hw): Likewise.
39480         (float_<mode>di2_hw): Likewise.
39481         (float_<mode>si2_hw): Likewise.
39482         (floatuns_<mode>di2_hw): Likewise.
39483         (floatuns_<mode>si2_hw): Likewise.
39484         (xscvqp<su>wz_<mode>): Delete, no longer used.
39485         (xscvqp<su>dz_<mode>): Likewise.
39486         (xscv<su>dqp_<mode>): Likewise.
39487         (ieee128_mfvsrd_64bit): Likewise.
39488         (ieee128_mfvsrd_32bit): Likewise.
39489         (ieee128_mfvsrwz): Likewise.
39490         (ieee128_mtvsrw): Likewise.
39491         (ieee128_mtvsrd_64bit): Likewise.
39492         (ieee128_mtvsrd_32bit): Likewise.
39494 2017-01-31  Martin Liska  <mliska@suse.cz>
39496         PR ipa/79285
39497         * ipa-prop.c (ipa_free_all_node_params): Call release method
39498         instead of ~sumbol_summary to not to trigger double times
39499         dtor of hash_map.
39501 2017-01-31  Aldy Hernandez  <aldyh@redhat.com>
39503         PR tree-optimization/71691
39504         * bitmap.h (class auto_bitmap): New.
39505         * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39506         is_maybe_undefined instead of ssa_undefined_value_p.
39508 2017-01-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39510         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39511         __S390_ARCH_LEVEL__ to __ARCH__.
39513 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39515         PR tree-optimization/79267
39516         * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39517         if should_remove_lhs_p is true.
39519 2017-01-30  Alexandre Oliva  <aoliva@redhat.com>
39521         PR debug/63238
39522         * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39523         (add_alignment_attribute): New.
39524         (base_type_die): Add alignment attribute.
39525         (subrange_type_die): Likewise.
39526         (modified_type_die): Likewise.
39527         (gen_array_type_die): Likewise.
39528         (gen_descr_array_type_die: Likewise.
39529         (gen_enumeration_type_die): Likewise.
39530         (gen_subprogram_die): Likewise.
39531         (gen_variable_die): Likewise.
39532         (gen_field_die): Likewise.
39533         (gen_ptr_to_mbr_type_die): Likewise.
39534         (gen_struct_or_union_type_die): Likewise.
39535         (gen_subroutine_type_die): Likewise.
39536         (gen_typedef_die): Likewise.
39537         (base_type_cmp): Compare alignment attribute.
39539 2017-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
39541         PR target/79170
39542         * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39543         (setb_unsigned) New pattern for setb with CCUNS.
39544         * config/rs6000/rs6000.c (expand_block_compare): Use a different
39545         subfc./subfe sequence to avoid overflow problems.  Generate a
39546         shorter sequence with cmpld/setb for power9.
39547         * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39548         for generating subfc. instruction.
39549         (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39550         now uses this instruction.
39552 2017-01-30  Ian Lance Taylor  <iant@google.com>
39554         PR debug/79289
39555         * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39556         for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39558 2017-01-30  Martin Sebor  <msebor@redhat.com>
39560         * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39561         Move constant to the right of a relational operator.
39562         (get_mpfr_format_length, format_character, format_string): Ditto.
39563         (should_warn_p, maybe_warn): Same.
39565         * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39567 2017-01-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
39569         PR lto/79061
39570         * asan.c (get_translation_unit_decl): Remove function.
39571         (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39573 2017-01-30  Martin Liska  <mliska@suse.cz>
39575         PR gcov-profile/79259
39576         * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39577         -fprofile-generate.
39579 2017-01-30  Martin Liska  <mliska@suse.cz>
39581         PR bootstrap/78985
39582         * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39583         Initialize variables with NULL value.
39585 2017-01-30  Richard Earnshaw  <rearnsh@arm.com>
39587         PR target/79260
39588         * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39589         tm_p_file.
39590         * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39592 2017-01-30  Richard Biener  <rguenther@suse.de>
39594         PR tree-optimization/79276
39595         * tree-vrp.c (process_assert_insertions): Properly adjust common
39596         when removing a duplicate.
39598 2017-01-30  Richard Biener  <rguenther@suse.de>
39600         PR tree-optimization/79256
39601         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39602         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39603         alignment on TYPE.
39604         * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39606 2017-01-30  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39608         PR target/79240
39609         * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39610         ("*r<noxa>sbg_<mode>_sll_bitmask")
39611         ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39612         ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39613         Use contiguous_bitmask_nowrap_operand.
39615 2017-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
39617         PR target/79268
39618         * config/rs6000/altivec.h (vec_xl): Revise #define.
39619         (vec_xst): Likewise.
39621 2017-01-27  Uros Bizjak  <ubizjak@gmail.com>
39623         * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39625 2017-01-27  Bernd Schmidt  <bschmidt@redhat.com>
39627         PR rtl-optimization/79194
39628         * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39629         traps before call to bypass_conditional_jumps.
39631 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39633         PR tree-optimization/71374
39634         * lra-constraints.c (check_conflict_input_operands): New.
39635         (match_reload): Use it.
39637 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39639         PR target/79131
39640         * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39641         account to calculate conflict_set.
39643 2017-01-27  Bin Cheng  <bin.cheng@arm.com>
39645         PR rtl-optimization/78559
39646         * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39647         other_insn in combine.
39649 2017-01-27  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
39651         * builtin-types.def: Use unsigned_char_type_node for BT_UINT8.  Use
39652         uint16_type_node for BT_UINT16.
39654 2017-01-27  David Malcolm  <dmalcolm@redhat.com>
39656         * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39657         "RTL Tests" to menu.
39658         (GIMPLE Tests): New node.
39659         (RTL Tests): New node.
39661 2017-01-27  Richard Biener  <rguenther@suse.de>
39663         PR tree-optimization/79245
39664         * tree-loop-distribution.c (distribute_loop): Apply cost
39665         modeling also to detected patterns.
39667 2017-01-27  Richard Biener  <rguenther@suse.de>
39669         PR tree-optimization/71433
39670         * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39671         (compare_assert_loc): New function.
39672         (process_assert_insertions): Sort and optimize assert locations
39673         to remove duplicates and push down identical assertions on
39674         edges to their destination block.
39676 2017-01-27  Richard Biener  <rguenther@suse.de>
39678         PR tree-optimization/79244
39679         * tree-vrp.c (remove_range_assertions): Forcefully propagate
39680         out SSA names even if abnormal.
39682 2017-01-27  Jakub Jelinek  <jakub@redhat.com>
39684         * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39685         * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39686         instead of MPFR_RNDN.
39688 2017-01-27  Richard Earnshaw  <rearnsha@arm.com>
39690         PR target/79239
39691         * arm.c (arm_option_override): Don't call build_target_option_node
39692         until after doing all option overrides.
39693         (arm_valid_target_attribute_tree): Likewise.
39695 2017-01-27  Martin Liska  <mliska@suse.cz>
39697         * doc/invoke.texi (-fprofile-arcs): Document profiling support
39698         for {cd}tors and C++ {cd}tors.
39700 2017-01-27  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39702         * config/s390/s390.md ("*setmem_long_and")
39703         ("*setmem_long_and_31z"): Use zero_extend instead of and.
39705 2017-01-26  Martin Sebor  <msebor@redhat.com>
39707         * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39708         of precision.
39710 2017-01-26  Martin Sebor  <msebor@redhat.com>
39712         * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39713         HAVE_DFmode before using XFmode or DFmode.
39714         (parse_directive): Avoid using the z length modifier to avoid
39715         the ISO C++98 does not support the â€˜z’ gnu_printf length modifier.
39717         PR middle-end/78703
39718         * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39719         to accept adjustment as an array.
39720         (get_int_range): New function.
39721         (struct directive): Make width and prec arrays.
39722         (directive::set_width, directive::set_precision): Call get_int_range.
39723         (format_integer, format_floating): Handle width and precision ranges.
39724         (format_string, parse_directive): Same.
39726 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39728         PR debug/79129
39729         * dwarf2out.c (generate_skeleton_bottom_up): For children with
39730         comdat_type_p set, just clone them, but keep the children in the
39731         original DIE.
39733         PR debug/78835
39734         * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39735         which have direct callers with -fvar-tracking-assignments enabled
39736         in the current TU.
39737         (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39738         inside of type units.
39740 2017-01-26  Martin Sebor  <msebor@redhat.com>
39742         PR middle-end/78703
39743         * gimple-ssa-sprintf.c (struct result_range): Add likely and
39744         unlikely counters.
39745         (struct format_result): Replace number_chars, number_chars_min,
39746         and number_chars_max with a single member of struct result_range.
39747         Remove bounded.
39748         (format_result::operator+=): Adjust.
39749         (struct fmtresult): Remove bounded.  Handle likely and unlikely
39750         counters.
39751         (fmtresult::adjust_for_width_or_precision): New function.
39752         (fmtresult:type_max_digits): New function.
39753         (bytes_remaining): Handle likely and unlikely counters.
39754         (min_bytes_remaining): Remove.
39755         (format_percent): Simplify.
39756         (format_integer, format_floating): Set likely and unlikely counters.
39757         (get_string_length, format_character, format_string): Same.
39758         (format_plain, should_warn_p): New function.
39759         (maybe_warn): Call should_warn_p.  Update diagnostic messages
39760         and handle those for all directives, including plain strings.
39761         (format_directive): Handle likely and unlikely counters.
39762         Remove unnecessary quoting from diagnostics.  Add an informational
39763         note.
39764         (add_bytes): Remove.
39765         (pass_sprintf_length::compute_format_length): Simplify.
39766         (try_substitute_return_value): Handle likely and unlikely counters.
39768 2017-01-26  Carl Love  <cel@us.ibm.com>
39770         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39771         bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39773 2017-01-26  Vladimir Makarov  <vmakarov@redhat.com>
39775         PR target/79131
39776         * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39777         endianess for subregs into account.
39778         * lra-constraints.c (lra_constraints): Do risky transformations
39779         always on the first iteration.
39780         * lra-lives.c (check_pseudos_live_through_calls): Add arg
39781         last_call_used_reg_set.
39782         (process_bb_lives): Define and use last_call_used_reg_set.
39783         * lra.c (lra): Always continue after lra_constraints on the first
39784         iteration.
39786 2017-01-26  Kirill Yukhin  <kirill.yukhin@gmail.com>
39788         * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39789         constant.
39790         * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39792 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39794         * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39795         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39796         * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39797         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39798         * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39799         __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39800         OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39801         * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39802         (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39804 2017-01-26  Marek Polacek  <polacek@redhat.com>
39806         PR c/79199
39807         * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39808         for the third operand.
39810 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39812         PR middle-end/79236
39813         * omp-low.c (struct omp_context): Add simt_stmt field.
39814         (scan_omp_for): Return omp_context *.
39815         (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39816         context to the _simt_ SIMD stmt.
39817         (lower_omp_for): For combined SIMD with sibling _simt_
39818         SIMD, make sure to use the same decls in _looptemp_
39819         clauses as in the sibling.
39821 2017-01-26  David Sherwood  <david.sherwood@arm.com>
39823         PR middle-end/79212
39824         * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39825         all contexts.
39827 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39829         PR target/70465
39830         * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39831         emit fld b; fld a; if possible.
39833         * brig-builtins.def: Update copyright years.
39834         * config/arm/arm_acle_builtins.def: Update copyright years.
39836 2017-01-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
39838         PR target/79179
39839         * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39840         constraint instead of o for the stxsd instruction.
39842 2017-01-25  Carl Love  <cel@us.ibm.com>
39844         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39845         of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39847 2017-01-25  Jonathan Wakely  <jwakely@redhat.com>
39849         * doc/invoke.texi (C++ Dialect Options): Fix typo.
39851 2017-01-25  Richard Biener  <rguenther@suse.de>
39853         PR tree-optimization/69264
39854         * target.def (vector_alignment_reachable): Improve documentation.
39855         * doc/tm.texi: Regenerate.
39856         * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39857         and add a comment.
39858         * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39859         earlier changes with respect to TYPE_USER_ALIGN.
39860         (vector_alignment_reachable_p): Likewise.  Improve dumping.
39862 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39864         PR target/79145
39865         * config/arm/arm.md (xordi3): Force constant operand into a register
39866         for TARGET_IWMMXT.
39868 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39870         * doc/invoke.texi (-fstore-merging): Correct default optimization
39871         levels at which it is enabled.
39872         (-O): Move -fstore-merging from list to...
39873         (-O2): ... Here.
39875 2017-01-25  Richard Biener  <rguenther@suse.de>
39877         PR debug/78363
39878         * omp-expand.c: Include debug.h.
39879         (expand_omp_taskreg): Make sure to generate early debug before
39880         outlining anything from a function.
39881         (expand_omp_target): Likewise.
39882         (grid_expand_target_grid_body): Likewise.
39884 2017-01-25  Maxim Ostapenko  <m.ostapenko@samsung.com>
39886         PR lto/79061
39887         * asan.c (get_translation_unit_decl): New function.
39888         (asan_add_global): Extract modules file name from globals
39889         TRANSLATION_UNIT_DECL name.
39891 2017-01-24  Eric Botcazou  <ebotcazou@adacore.com>
39893         PR target/77439
39894         * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39895         for long calls with APCS frame and VFP.
39897 2017-01-24  David Malcolm  <dmalcolm@redhat.com>
39899         * cfg.c (original_copy_tables_initialized_p): New function.
39900         * cfg.h (original_copy_tables_initialized_p): New decl.
39901         * cfgrtl.c (relink_block_chain): Guard the call to
39902         free_original_copy_tables with a call to
39903         original_copy_tables_initialized_p.
39904         * cgraph.h (symtab_node::native_rtl_p): New decl.
39905         * cgraphunit.c (symtab_node::native_rtl_p): New function.
39906         (symtab_node::needed_p): Don't assert for early assembly output
39907         for __RTL functions.
39908         (cgraph_node::finalize_function): Set "force_output" for __RTL
39909         functions.
39910         (cgraph_node::analyze): Bail out early for __RTL functions.
39911         (analyze_functions): Update assertion to support __RTL functions.
39912         (cgraph_node::expand): Bail out early for __RTL functions.
39913         * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39914         __RTL functions.
39915         * function.h (struct function): Update comment for field
39916         "pass_startwith".
39917         * gimple-expr.c: Include "tree-pass.h".
39918         (gimple_has_body_p): Return false for __RTL functions.
39919         * Makefile.in (OBJS): Add run-rtl-passes.o.
39920         * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
39921         accessor.
39922         (gcc::pass_manager::get_clean_slate): New accessor.
39923         * passes.c: Include "insn-addr.h".
39924         (should_skip_pass_p): Add logging.  Update logic for running
39925         "expand" to be compatible with both __GIMPLE and __RTL.  Guard
39926         property-provider override so it is only done for gimple passes.
39927         Don't skip dfinit.
39928         (skip_pass): New function.
39929         (execute_one_pass): Call skip_pass when skipping passes.
39930         * read-md.c (md_reader::read_char): Support filtering
39931         the input to a subset of line numbers.
39932         (md_reader::md_reader): Initialize fields
39933         m_first_line and m_last_line.
39934         (md_reader::read_file_fragment): New function.
39935         * read-md.h (md_reader::read_file_fragment): New decl.
39936         (md_reader::m_first_line): New field.
39937         (md_reader::m_last_line): New field.
39938         * read-rtl-function.c (function_reader::create_function): Only
39939         create cfun if it doesn't already exist.  Set PROP_rtl on cfun's
39940         curr_properties.  Set DECL_INITIAL to a dummy block.
39941         (read_rtl_function_body_from_file_range): New function.
39942         * read-rtl-function.h (read_rtl_function_body_from_file_range):
39943         New decl.
39944         * run-rtl-passes.c: New file.
39945         * run-rtl-passes.h: New file.
39947 2017-01-24  Jeff Law  <law@redhat.com>
39949         * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
39950         buffer size.
39952 2017-01-24  Bin Cheng  <bin.cheng@arm.com>
39954         PR tree-optimization/79159
39955         * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
39956         (record_nonwrapping_iv): Improve boundary using above function if no
39957         value range information.
39959 2017-01-24  Pekka Jääskeläinen  <pekka@parmance.com>
39960             Martin Jambor  <mjambor@suse.cz>
39962         * brig-builtins.def: New file.
39963         * builtins.def (DEF_HSAIL_BUILTIN): New macro.
39964         (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
39965         (DEF_HSAIL_SAT_BUILTIN): Likewise.
39966         (DEF_HSAIL_INTR_BUILTIN): Likewise.
39967         (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
39968         * builtin-types.def (BT_INT8): New.
39969         (BT_INT16): Likewise.
39970         (BT_UINT8): Likewise.
39971         (BT_UINT16): Likewise.
39972         (BT_FN_ULONG): Likewise.
39973         (BT_FN_UINT_INT): Likewise.
39974         (BT_FN_UINT_ULONG): Likewise.
39975         (BT_FN_UINT_LONG): Likewise.
39976         (BT_FN_UINT_PTR): Likewise.
39977         (BT_FN_ULONG_PTR): Likewise.
39978         (BT_FN_INT8_FLOAT): Likewise.
39979         (BT_FN_INT16_FLOAT): Likewise.
39980         (BT_FN_UINT32_FLOAT): Likewise.
39981         (BT_FN_UINT16_FLOAT): Likewise.
39982         (BT_FN_UINT8_FLOAT): Likewise.
39983         (BT_FN_UINT64_FLOAT): Likewise.
39984         (BT_FN_UINT16_UINT32): Likewise.
39985         (BT_FN_UINT32_UINT16): Likewise.
39986         (BT_FN_UINT16_UINT16_UINT16): Likewise.
39987         (BT_FN_INT_PTR_INT): Likewise.
39988         (BT_FN_UINT_PTR_UINT): Likewise.
39989         (BT_FN_LONG_PTR_LONG): Likewise.
39990         (BT_FN_ULONG_PTR_ULONG): Likewise.
39991         (BT_FN_VOID_UINT64_UINT64): Likewise.
39992         (BT_FN_UINT8_UINT8_UINT8): Likewise.
39993         (BT_FN_INT8_INT8_INT8): Likewise.
39994         (BT_FN_INT16_INT16_INT16): Likewise.
39995         (BT_FN_INT_INT_INT): Likewise.
39996         (BT_FN_UINT_FLOAT_UINT): Likewise.
39997         (BT_FN_FLOAT_UINT_UINT): Likewise.
39998         (BT_FN_ULONG_UINT_UINT): Likewise.
39999         (BT_FN_ULONG_UINT_PTR): Likewise.
40000         (BT_FN_ULONG_ULONG_ULONG): Likewise.
40001         (BT_FN_UINT_UINT_UINT): Likewise.
40002         (BT_FN_VOID_UINT_PTR): Likewise.
40003         (BT_FN_UINT_UINT_PTR: Likewise.
40004         (BT_FN_UINT32_UINT64_PTR): Likewise.
40005         (BT_FN_INT_INT_UINT_UINT): Likewise.
40006         (BT_FN_UINT_UINT_UINT_UINT): Likewise.
40007         (BT_FN_UINT_UINT_UINT_PTR): Likewise.
40008         (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
40009         (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
40010         (BT_FN_LONG_LONG_UINT_UINT): Likewise.
40011         (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
40012         (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
40013         (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
40014         (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
40015         (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
40016         (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
40017         * doc/frontends.texi: List BRIG FE.
40018         * doc/install.texi (Testing): Add BRIG tesring requirements.
40019         * doc/invoke.texi (Overall Options): Mention BRIG.
40020         * doc/standards.texi (Standards): Doucment BRIG HSA version.
40022 2017-01-24  Richard Biener  <rguenther@suse.de>
40024         PR translation/79208
40025         * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
40027 2017-01-24  Martin Jambor  <mjambor@suse.cz>
40029         PR bootstrap/79198
40030         * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
40031         * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
40032         and known_contexts.
40034 2017-01-24  Aldy Hernandez  <aldyh@redhat.com>
40036         PR middle-end/79123
40037         * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
40038         casts from signed to unsigned really don't have a range.
40040 2017-01-24  Markus Trippelsdorf  <markus@trippelsdorf.de>
40042         * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
40043         GMP_RNDx for compatiblity.
40045 2017-01-24  Martin Liska  <mliska@suse.cz>
40047         PR bootstrap/79132
40048         * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
40049         that would prevent us to call alloca with -1 as argument.
40051 2017-01-24  Jakub Jelinek  <jakub@redhat.com>
40053         * dwarf2out.c (output_compilation_unit_header, output_file_names):
40054         Avoid -Wformat-security warning.
40056 2017-01-23  Andrew Pinski  <apinski@cavium.com>
40058         * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
40059         cost table.
40061 2017-01-23  Martin Sebor  <msebor@redhat.com>
40063         PR middle-end/78703
40064         * gimple-ssa-sprintf.c (warn_level): New global.
40065         (format_integer): Use it here and throughout the rest of the file.
40066         Use the same switch to compute sign as base.
40067         (maybe_warn): New function.
40068         (format_directive): Factor out warnings into maybe_warn.
40069         Add debugging output.  Use warn_level.
40070         (add_bytes): Use warn_level.
40071         (pass_sprintf_length::compute_format_length): Add debugging output.
40072         (try_substitute_return_value): Same.
40073         (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
40075         PR middle-end/78703
40076         * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
40077         (struct fmtresult, format_integer, format_floating): Adjust.
40078         (fmtresult::fmtresult): Set max correctly in two argument ctor.
40079         (get_string_length, format_string,format_directive): Same.
40080         (pass_sprintf_length::compute_format_length): Same.
40081         (try_substitute_return_value): Simplify slightly.
40083         PR middle-end/78703
40084         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
40085         (fmtresult::operator+=): Outlined.
40086         (struct fmtresult): Add ctors.
40087         (struct conversion_spec): Rename...
40088         (struct directive): ...to this.  Add and remove data members.
40089         (directive::set_width, directive::set_precision): New functions.
40090         (format_percent): Use fmtresult ctor.
40091         (get_width_and_precision): Remove.
40092         (format_integer): Make naming changes.  Avoid computing width and
40093         precision.
40094         (format_floating): Same.  Adjust indentation.
40095         (format_character, format_none): New functions.
40096         (format_string): Moved character handling to format_character.
40097         (format_directive): Remove arguments, change return type.
40098         (parse_directive): New function.
40099         (pass_sprintf_length::compute_format_length): Move directive
40100         parsing to parse_directive.
40102 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40104         * tree.h (assign_assembler_name_if_neeeded): Rename to ...
40105         (assign_assembler_name_if_needed): ... this.
40106         * tree.c (assign_assembler_name_if_neeeded): Rename to ...
40107         (assign_assembler_name_if_needed): ... this.
40108         (free_lang_data_in_cgraph): Adjust callers.
40109         * cgraphunit.c (cgraph_node::analyze): Likewise.
40110         * omp-expand.c (expand_omp_taskreg, expand_omp_target):
40111         Likewise.
40113 2017-01-23  Richard Biener  <rguenther@suse.de>
40115         PR tree-optimization/79088
40116         PR tree-optimization/79188
40117         * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
40118         resetting loop bounds after last path deletion.  Reset loop
40119         bounds of the target loop, make code match the comments.
40120         * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
40121         Make sure loops need no fixups.
40123 2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
40125         * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
40126         exponent support with double type for first argument.
40127         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
40128         type returned by __builtin_vec_extract_sig,
40129         __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
40130         functions from "vector int" to "vector unsigned int" or from
40131         "vector long long int" to "vector unsigned long long int".
40132         Changed type returned by __builtin_vec_extract_exp,
40133         __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
40134         functions from "vector int" to "vector unsigned int" or from
40135         "vector long long int" to "vector unsigned long long int".
40136         Changed return type of __builtin_vec_test_data_class,
40137         __builtin_vec_test_data_class_sp, and
40138         __builtin_vec_test_data_class_dp from "vector int" to
40139         "vector bool int" or from "vector long long int" to "vector bool
40140         long long int" and changed second argument type from "unsigned
40141         int" to "int".  Added new overloaded function forms "vector float
40142         __builtin_vec_insert_exp (vector float, vector unsigned int)" and
40143         "vector float __builtin_vec_insert_exp_sp (vector float, vector
40144         unsigned int)" and "vector double __builtin_vec_insert_exp (vector
40145         double, vector unsigned long long int)" and "vector double
40146         __builtin_vec_insert_exp_dp (vector double, vector unsigned long
40147         long int)".  Changed return type of
40148         __builtin_scalar_test_data_class and
40149         __builtin_scalar_test_data_class_sp and
40150         __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
40151         int" and changed second argument from "unsigned int" to "int".
40152         Changed type returned by __builtin_scalar_test_neg,
40153         __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
40154         from "int" to "bool int".  Added new overloaded function form
40155         "double __builtin_scalar_insert_exp (double, unsigned long long int)".
40156         * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
40157         exponent double-precision with floating point first argument.
40158         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
40159         documentation of scalar_test_data_class, scalar_test_neg,
40160         scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
40161         vector_extract_exp, vec_extract_sig, vec_insert_exp, and
40162         vec_test_data_class built-in functions to reflect refinements in
40163         their type signatures.
40165 2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>
40167         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
40168         size of buf.
40169         (aarch64_elf_asm_destructor): Likewise.
40171 2017-01-23  Bernd Schmidt  <bschmidt@redhat.com>
40173         PR rtl-optimization/78634
40174         * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
40175         (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
40176         * ifcvt.c (noce_try_cmove): Add missing cost check.
40178         PR rtl-optimization/71724
40179         * combine.c (if_then_else_cond): Look for situations where it is
40180         beneficial to undo the work of one of the recursive calls.
40182 2017-01-23  Bin Cheng  <bin.cheng@arm.com>
40184         PR tree-optimization/70754
40185         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
40186         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
40187         combined stmt before it if not NULL.
40188         (combine_chains): Process refs reversely and compute dominance point
40189         for root ref.
40191 2017-01-23  Martin Liska  <mliska@suse.cz>
40193         PR tree-optimization/79196
40194         * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
40195         (fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
40196         instead of memcmp.
40197         (strlen_optimize_stmt): Call the renamed function.
40199 2017-01-23  Michael Matz  <matz@suse.de>
40201         PR tree-optimization/78384
40202         * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
40204 2017-01-23  Richard Biener  <rguenther@suse.de>
40206         PR tree-optimization/79186
40207         * tree-vrp.c (register_new_assert_for): Make sure we've seen
40208         both incoming edges before moving an assert.
40210 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40212         * ipa-prop.c (load_from_param_1): Removed.
40213         (load_from_unmodified_param): Bits from load_from_param_1 put back
40214         here.
40215         (load_from_param): Removed.
40216         (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
40217         with stmt.  Reverted back to use of load_from_unmodified_param.
40219 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40221         PR ipa/79108
40222         * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
40223         (ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
40224         field a pointer to garbage collected vector, mark lattices and
40225         ipcp_orig_node with GTY((skip)).
40226         (ipa_get_param_count): Adjust to descriptors being a pointer.
40227         (ipa_get_param): Likewise.
40228         (ipa_get_type): Likewise.
40229         (ipa_get_param_move_cost): Likewise.
40230         (ipa_set_param_used): Likewise.
40231         (ipa_get_controlled_uses): Likewise.
40232         (ipa_set_controlled_uses): Likewise.
40233         (ipa_is_param_used): Likewise.
40234         (ipa_node_params_t): Move into garbage collector.  New methods insert
40235         and remove.
40236         (ipa_node_params_sum): Annotate wth GTY(()).
40237         (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
40238         garbage collected.
40239         (ipa_load_from_parm_agg): Adjust declaration.
40240         * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
40241         * ipa-profile.c (ipa_profile): Likewise.
40242         * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
40243         (ipa_populate_param_decls): Make descriptors parameter garbage
40244         collected.
40245         (ipa_dump_param): Adjust to descriptors being a pointer.
40246         (ipa_alloc_node_params): Likewise.
40247         (ipa_initialize_node_params): Likewise.
40248         (load_from_param_1): Make descriptors parameter garbage collected.
40249         (load_from_unmodified_param): Likewise.
40250         (load_from_param): Likewise.
40251         (ipa_load_from_parm_agg): Likewise.
40252         (ipa_node_params::~ipa_node_params): Removed.
40253         (ipa_free_all_node_params): Remove call to delete operator.
40254         (ipa_node_params_t::insert): New.
40255         (ipa_node_params_t::remove): Likewise.
40256         (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40257         copy known_csts and known_contexts vectors.
40258         (ipa_read_node_info): Adjust to descriptors being a pointer.
40259         (ipcp_modif_dom_walker): Make m_descriptors field garbage
40260         collected.
40261         (ipcp_transform_function): Make descriptors variable garbage
40262         collected.
40264 2017-01-23  Andrew Senkevich  <andrew.senkevich@intel.com>
40266         * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40267         * config/i386/avx512dqintrin.h: Ditto.
40268         * config/i386/avx512fintrin.h: Ditto.
40269         * config/i386/i386.c: Handle new builtins.
40270         * config/i386/i386-builtin.def: Add new builtins.
40271         * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40272         (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40274 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40275             Martin Liska  <mliska@suse.cz>
40277         * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40278         * asan.c (asan_expand_poison_ifn): Support stores and use
40279         appropriate ASAN report function.
40280         * internal-fn.c (expand_ASAN_POISON_USE): New function.
40281         * internal-fn.def (ASAN_POISON_USE): Declare.
40282         * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40283         (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40284         * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40285         ASAN_POISON calls w/o LHS.
40286         * tree-ssa.c (execute_update_addresses_taken): Create clobber
40287         for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40288         from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40289         * gimplify.c (asan_poison_variables): Add attribute
40290         use_after_scope_memory to variables that really needs to live
40291         in memory.
40292         * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40293         having the attribute.
40295 2017-01-23  Martin Liska  <mliska@suse.cz>
40297         * asan.c (create_asan_shadow_var): New function.
40298         (asan_expand_poison_ifn): Likewise.
40299         * asan.h (asan_expand_poison_ifn): New declaration.
40300         * internal-fn.c (expand_ASAN_POISON): Likewise.
40301         * internal-fn.def (ASAN_POISON): New builtin.
40302         * sanopt.c (pass_sanopt::execute): Expand
40303         asan_expand_poison_ifn.
40304         * tree-inline.c (copy_decl_for_dup_finish): Make function
40305         external.
40306         * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40307         * tree-ssa.c (is_asan_mark_p): New function.
40308         (execute_update_addresses_taken): Rewrite local variables
40309         (identified just by use-after-scope as addressable) into SSA.
40311 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40313         * doc/install.texi (Specific): opensource.apple.com uses https
40314         now. Remove trailing slash.
40316 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40318         * README.Portability: Remove note on an Irix compatibility issue.
40320 2017-01-22  Dimitry Andric <dim@FreeBSD.org>
40322         * gcov.c (INCLUDE_ALGORITHM): Define.
40323         (INCLUDE_VECTOR): Define.
40324         No longer include <vector> and <algorithm> directly.
40326 2017-01-21  Gerald Pfeifer  <gerald@pfeifer.com>
40328         * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40329         to https.
40330         * doc/invoke.texi (Code Gen Options): Ditto.
40332 2017-01-21  Jan Hubicka  <hubicka@ucw.cz>
40334         PR lto/78407
40335         * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40337 2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>
40339         rtl-optimization/79125
40340         * cprop.c (local_cprop_pass): Handle cases where we make an
40341         unconditional trap.
40343 2017-01-20  Segher Boessenkool  <segher@kernel.crashing.org>
40345         PR target/61729
40346         PR target/77850
40347         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40348         read from, for big endian.
40350 2017-01-20  Jiong Wang  <jiong.wang@arm.com>
40352         * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40353         register pauth builtins for LP64 only.
40355 2017-01-20  Marek Polacek  <polacek@redhat.com>
40357         PR c/79152
40358         * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40359         non-case labels.
40361 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40363         * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40364         of safelen status.
40365         * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40366         * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40367         * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40369 2017-01-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40371         PR target/71270
40372         * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40373         in big-endian mode when they are not a single duplicated value.
40375 2017-01-20  Richard Biener  <rguenther@suse.de>
40377         * BASE-VER: Bump to 7.0.1.
40379 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40381         * omp-low.c (omplow_simd_context): New struct.  Use it...
40382         (lower_rec_simd_input_clauses): ...here and...
40383         (lower_rec_input_clauses): ...here to hold common data.  Adjust all
40384         references to idx, lane, max_vf, is_simt.
40386 2017-01-20  Graham Markall  <graham.markall@embecosm.com>
40388         * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40389         mcpu=nps400.
40391 2017-01-20  Martin Jambor  <mjambor@suse.cz>
40393         * hsa.h: Renaed to hsa-common.h.  Adjusted a comment.
40394         * hsa.c: Renaed to hsa-common.c.  Change include of gt-hsa.h to
40395         gt-hsa-common.h.
40396         * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40397         (GTFILES): Rename hsa.c to hsa-common.c.
40398         * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40399         * hsa-dump.c: Likewise.
40400         * hsa-gen.c: Likewise.
40401         * hsa-regalloc.c: Likewise.
40402         * ipa-hsa.c: Likewise.
40403         * omp-expand.c: Likewise.
40404         * omp-low.c: Likewise.
40405         * toplev.c: Likewise.
40407 2017-01-20  Marek Polacek  <polacek@redhat.com>
40409         PR c/64279
40410         * doc/invoke.texi: Document -Wduplicated-branches.
40411         * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40412         COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40413         POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40414         STATEMENT_LIST, and RETURN_EXPR.  For non-pure non-const functions
40415         return 0 only when not OEP_LEXICOGRAPHIC.
40416         (fold_build_cleanup_point_expr): Use the expression
40417         location when building CLEANUP_POINT_EXPR.
40418         * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40419         * tree.c (add_expr): Handle error_mark_node.
40421 2017-01-20  Martin Liska  <mliska@suse.cz>
40423         PR lto/69188
40424         * tree-profile.c (init_ic_make_global_vars): Do not call
40425         finalize_decl.
40426         (gimple_init_gcov_profiler): Likewise.
40428 2017-01-20  Martin Liska  <mliska@suse.cz>
40430         PR ipa/71190
40431         * cgraph.h (maybe_create_reference): Remove argument and
40432         update comment.
40433         * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40434         argument.
40435         * ipa-cp.c (create_specialized_node): Likewise.
40436         * symtab.c (symtab_node::maybe_create_reference): Handle
40437         VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40439 2017-01-20  Martin Liska  <mliska@suse.cz>
40441         * read-rtl-function.c (function_reader::create_function): Use
40442         build_decl instread of build_decl_stat.
40444 2017-01-20  Andrew Senkevich  <andrew.senkevich@intel.com>
40446         * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40447         * config/i386/avx512dqintrin.h: Ditto.
40448         * config/i386/avx512fintrin.h: Ditto.
40449         * config/i386/i386-builtin-types.def: Add new types.
40450         * config/i386/i386.c: Handle new types.
40451         * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40452         (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40453         (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40454         (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40455         (__builtin_ia32_kshiftridi): New.
40456         * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40458 2017-01-19  Segher Boessenkool  <segher@kernel.crashing.org>
40460         PR target/78875
40461         PR target/79140
40462         * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40463         define to rs6000_init_stack_protect_guard.
40464         (rs6000_init_stack_protect_guard): New function.
40466 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40467             Yunqiang Su  <yunqiang.su@imgtec.com>
40469         * config.gcc (supported_defaults): Add madd4.
40470         (with_madd4): Add validation.
40471         (all_defaults): Add madd4.
40472         * config/mips/mips.opt (mmadd4): New option.
40473         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40474         mmadd4.
40475         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40476         __mips_no_madd4.
40477         (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40478         (ISA_HAS_FUSED_MADD4): Likewise.
40479         * doc/invoke.texi (-mmadd4): Document the new option.
40480         * doc/install.texi (--with-madd4): Document the new option.
40482 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40484         * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40485         entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40486         AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40487         (aarch64_init_pauth_hint_builtins): New.
40488         (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40489         (aarch64_expand_builtin): Expand new builtins.
40491 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40493         * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40494         * combine-stack-adj.c (no_unhandled_cfa): Handle
40495         REG_CFA_TOGGLE_RA_MANGLE.
40496         * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40497         * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40498         info for return address signing.
40499         (aarch64_expand_epilogue): Likewise.
40501 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40503         * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40504         * config/aarch64/aarch64-protos.h
40505         (aarch64_return_address_signing_enabled): New declaration.
40506         * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40507         New function.
40508         (aarch64_expand_prologue): Sign return address before it's pushed onto
40509         stack.
40510         (aarch64_expand_epilogue): Authenticate return address fetched from
40511         stack.
40512         (aarch64_override_options): Sanity check for ILP32 and ISA level.
40513         (aarch64_attributes): New function attributes for "sign-return-address".
40514         * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40515         UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40516         ("*do_return"): Generate combined instructions according to key index.
40517         ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40518         * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40519         iterators.
40520         (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40521         * config/aarch64/aarch64.opt (msign-return-address=): New.
40522         * doc/extend.texi (AArch64 Function Attributes): Documents
40523         "sign-return-address=".
40524         * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40526 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40528         * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40529         overall option summary.
40531 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40533         * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40534         * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40535         AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40536         * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40538 2017-01-19  Michael Meissner  <meissner@linux.vnet.ibm.com>
40540         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40541         -mpower9-minmax by default for -mcpu=power9.
40542         (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40543         128-bit floating point.
40545 2017-01-20  Alan Modra  <amodra@gmail.com>
40547         * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40548         optimizing for size.
40550 2017-01-20  Alan Modra  <amodra@gmail.com>
40552         PR target/79144
40553         * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40554         for strcmp and strncmp from corresponding builtin decl.
40556 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40558         * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40559         instead of i386/rtems-64.h.
40560         * config/i386/rtems-64.h: Remove.
40562 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40564         PR target/78478
40565         Revert:
40566         2013-11-05  Uros Bizjak  <ubizjak@gmail.com>
40568         * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40570 2017-01-19  Tamar Christina  <tamar.christina@arm.com>
40572         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40573         Change int to HOST_WIDE_INT.
40574         * config/aarch64/aarch64-protos.h
40575         (aarch64_simd_gen_const_vector_dup): Likewise.
40576         * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40578 2017-01-19  David Malcolm  <dmalcolm@redhat.com>
40580         * langhooks-def.h (lhd_type_for_size): New decl.
40581         (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40582         * langhooks.c (lhd_type_for_size): New function, taken from
40583         lto_type_for_size.
40585 2017-01-19  Pat Haugen  <pthaugen@us.ibm.com>
40587         * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40588         define_bypass for CR latency.
40589         (power9-cracked-alu): Update bypass latency and remove power9-branch.
40590         (power9-alu2): Add define_bypass for CR latency.
40591         (power9-cmp): New.
40592         (power9-mul): Update insn latency.
40593         (power9-mul-compare): Update insn latency, bypass latency and remove
40594         power9-branch.
40596 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40598         * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40599         Delete.
40600         * config/aarch64/aarch64.md
40601         (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40602         aarch64_nopcrelative_literal_loads.
40603         (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40605 2017-01-19  Chenghua Xu  <paul.hua.gm@gmail.com>
40607         * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40608         TARGET_LOONGSON_3A.
40609         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40611 2017-01-19  Doug Gilmore  <doug.gilmore@imgtec.com>
40613         PR target/78176
40614         * config.gcc (supported_defaults): Add lxc1-sxc1.
40615         (with_lxc1_sxc1): Add validation.
40616         (all_defaults): Add lxc1-sxc1.
40617         * config/mips/mips.opt (mlxc1-sxc1): New option.
40618         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40619         mlxc1-sxc1.
40620         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40621         __mips_no_lxc1_sxc1.
40622         (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40623         * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40624         * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40626 2017-01-19  Richard Biener  <rguenther@suse.de>
40628         PR tree-optimization/72488
40629         * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40630         sure to restore SSA info.
40631         * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40633 2017-01-19  Richard Earnshaw  <rearnsha@arm.com>
40635         PR rtl-optimization/79121
40636         * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40637         of the inner type when shifting an extended value.
40639 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40641         PR lto/78407
40642         * symtab.c (symtab_node::equal_address_to): Fix comparing of
40643         interposable aliases.
40645 2017-01-18  Peter Bergner  <bergner@vnet.ibm.com>
40647         PR target/78516
40648         * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40649         Use the evmergelohi instruction.
40650         (mov_si<mode>_e500_subreg4_2_le): Likewise.
40651         (mov_sitf_e500_subreg8_2_be): Likewise.
40652         (mov_sitf_e500_subreg12_2_le): Likewise.
40653         (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40654         (mov_si<mode>_e500_subreg4_2_be): Likewise.
40655         (mov_sitf_e500_subreg8_2_le): Likewise.
40656         (mov_sitf_e500_subreg12_2_be): Likewise.
40658 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40660         * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40661         attribute from vecsimple to vecperm.
40662         (altivec_vbpermq2): Likewise.
40664 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40666         PR target/79040
40667         * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40669 2017-01-18  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
40670         * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40671         * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40672         strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40673         case where N arg is SIZE_MAX.
40674         * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40675         (cmpstrsi): Add pattern.
40677 2017-01-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
40679         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40680         __builtin_vec_revb builtins.
40681         * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40682         built-in functions to support generation of the ISA 3.0 XXBR<x>
40683         vector byte reverse instructions.
40684         (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40685         (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40686         (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40687         (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40688         (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40689         (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40690         (P9V_BUILTIN_VEC_REVB): Likewise.
40691         * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40692         generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40693         (p9_xxbrq_v16qi): Likewise.
40694         (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40695         (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40696         (p9_xxbrh_v8hi): Likewise.
40697         * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40698         * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40699         vec_revb built-in functions.
40701 2017-01-18  Uros Bizjak  <ubizjak@gmail.com>
40703         PR rtl-optimization/78952
40704         * config/i386/i386.md (any_extract): New code iterator.
40705         (*insvqi_2): Use any_extract for source operand.
40706         (*insvqi_3): Use any_shiftrt for source operand.
40708 2017-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
40710         * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40711         New function.
40712         (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40714 2017-01-18  Matthias Klose  <doko@ubuntu.com>
40716         * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40718 2016-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40720         * config/rs6000/altivec.h (vec_bperm): Change #define.
40721         * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40722         (altivec_vbpermq2): New define_insn.
40723         (altivec_vbpermd): Likewise.
40724         * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40725         function interface.
40726         (VBPERMD): Likewise.
40727         (VBPERM): New polymorphic function interface.
40728         * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40729         Add entries for P9V_BUILTIN_VEC_VBPERM.
40730         * doc/extend.texi: Add interfaces for vec_bperm.
40732 2017-01-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
40734         * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40735         first letter of error messages.
40736         (s390_resolve_overloaded_builtin): Likewise.
40737         * config/s390/s390.c (s390_expand_builtin): Likewise.
40738         (s390_invalid_arg_for_unprototyped_fn): Likewise.
40739         (s390_valid_target_attribute_inner_p): Likewise.
40740         * config/s390/s390.md ("tabort"): Likewise.
40742 2017-01-18  Toma Tabacu  <toma.tabacu@imgtec.com>
40744         * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40745         (ISA_AVOID_DIV_HILO): New macro.
40746         (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40747         (ISA_HAS_DDIV): Likewise.
40749 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40751         * doc/invoke.texi (fabi-version): Correct number of occurrences.
40753 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40755         * doc/invoke.texi (fabi-version): Spelling fix.
40757 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40759         PR c++/70182
40760         * doc/invoke.texi (fabi-version): Mention mangling fix for
40761         operator names.
40763 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40765         PR c++/77489
40766         * doc/invoke.texi (fabi-version): Document discriminator mangling.
40768 2017-01-17  Segher Boessenkool  <segher@kernel.crashing.org>
40770         PR target/78875
40771         * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40772         * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40773         the new options.
40774         * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40775         flexible settings.
40776         (stack_protect_test): Ditto.
40777         * config/rs6000/rs6000.opt (mstack-protector-guard=,
40778         mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40779         options.
40780         * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40781         Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40782         -mstack-protector-guard-offset=.
40783         (RS/6000 and PowerPC Options): Ditto.
40785 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40787         * config/i386/i386.h (MASK_CLASS_P): New define.
40788         * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40789         there are no registers from different register sets also when
40790         mask registers are used.  Update function comment.
40791         * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40792         to (*k/*r) and (*k/*km) alternatives.
40794 2017-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
40796         * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40797         * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40798         (EH_RETURN_HANDLER_RTX): New define.
40799         * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40800         Force frame pointer in EH return functions.
40801         (aarch64_expand_epilogue): Add barrier for eh_return.
40802         (aarch64_final_eh_return_addr): Remove.
40803         (aarch64_eh_return_handler_rtx): New function.
40804         * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40805         Remove.
40806         (aarch64_eh_return_handler_rtx): New prototype.
40808 2017-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40810         * config/rs6000/altivec.h (vec_rlmi): New #define.
40811         (vec_vrlnm): Likewise.
40812         (vec_rlnm): Likewise.
40813         * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40814         (UNSPEC_VRLNM): Likewise.
40815         (VIlong): New mode iterator.
40816         (altivec_vrl<VI_char>mi): New define_insn.
40817         (altivec_vrl<VI_char>nm): Likewise.
40818         * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40819         function entry.
40820         (VRLDNM): Likewise.
40821         (RLNM): New polymorphic function entry.
40822         (VRLWMI): New monomorphic function entry.
40823         (VRLDMI): Likewise.
40824         (RLMI): New polymorphic function entry.
40825         * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40826         new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40827         * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40828         vec_vrlnm.
40830 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40832         PR debug/78839
40833         * dwarf2out.c (field_byte_offset): Restore the
40834         PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40835         and DECL_FIELD_BIT_OFFSET.  Use fold_build2 instead of build2 + fold.
40836         (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40837         of build2 + fold.
40839 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40841         PR ada/67205
40842         * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40844 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40846         PR debug/71669
40847         * dwarf2out.c (add_data_member_location_attribute): For constant
40848         offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40849         instead of DW_AT_data_member_location, DW_AT_bit_offset and
40850         DW_AT_byte_size attributes.
40852 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40854         * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40855         after forcing to constant memory when the code model is medium.
40857 2017-01-17  Julia Koval  <julia.koval@intel.com>
40859         PR target/76731
40860         * config/i386/avx512fintrin.h
40861         (_mm512_i32gather_ps): Change __addr type to void const*.
40862         (_mm512_mask_i32gather_ps): Ditto.
40863         (_mm512_i32gather_pd): Ditto.
40864         (_mm512_mask_i32gather_pd): Ditto.
40865         (_mm512_i64gather_ps): Ditto.
40866         (_mm512_mask_i64gather_ps): Ditto.
40867         (_mm512_i64gather_pd): Ditto.
40868         (_mm512_mask_i64gather_pd): Ditto.
40869         (_mm512_i32gather_epi32): Ditto.
40870         (_mm512_mask_i32gather_epi32): Ditto.
40871         (_mm512_i32gather_epi64): Ditto.
40872         (_mm512_mask_i32gather_epi64): Ditto.
40873         (_mm512_i64gather_epi32): Ditto.
40874         (_mm512_mask_i64gather_epi32): Ditto.
40875         (_mm512_i64gather_epi64): Ditto.
40876         (_mm512_mask_i64gather_epi64): Ditto.
40877         (_mm512_i32scatter_ps): Change __addr type to void*.
40878         (_mm512_mask_i32scatter_ps): Ditto.
40879         (_mm512_i32scatter_pd): Ditto.
40880         (_mm512_mask_i32scatter_pd): Ditto.
40881         (_mm512_i64scatter_ps): Ditto.
40882         (_mm512_mask_i64scatter_ps): Ditto.
40883         (_mm512_i64scatter_pd): Ditto.
40884         (_mm512_mask_i64scatter_pd): Ditto.
40885         (_mm512_i32scatter_epi32): Ditto.
40886         (_mm512_mask_i32scatter_epi32): Ditto.
40887         (_mm512_i32scatter_epi64): Ditto.
40888         (_mm512_mask_i32scatter_epi64): Ditto.
40889         (_mm512_i64scatter_epi32): Ditto.
40890         (_mm512_mask_i64scatter_epi32): Ditto.
40891         (_mm512_i64scatter_epi64): Ditto.
40892         (_mm512_mask_i64scatter_epi64): Ditto.
40893         * config/i386/avx512pfintrin.h
40894         (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40895         (_mm512_mask_prefetch_i32gather_ps): Ditto.
40896         (_mm512_mask_prefetch_i64gather_pd): Ditto.
40897         (_mm512_mask_prefetch_i64gather_ps): Ditto.
40898         (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40899         (_mm512_prefetch_i32scatter_ps): Ditto.
40900         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40901         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40902         (_mm512_prefetch_i64scatter_pd): Ditto.
40903         (_mm512_prefetch_i64scatter_ps): Ditto.
40904         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40905         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40906         * config/i386/avx512vlintrin.h
40907         (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40908         (_mm_mmask_i32gather_ps): Ditto.
40909         (_mm256_mmask_i32gather_pd): Ditto.
40910         (_mm_mmask_i32gather_pd): Ditto.
40911         (_mm256_mmask_i64gather_ps): Ditto.
40912         (_mm_mmask_i64gather_ps): Ditto.
40913         (_mm256_mmask_i64gather_pd): Ditto.
40914         (_mm_mmask_i64gather_pd): Ditto.
40915         (_mm256_mmask_i32gather_epi32): Ditto.
40916         (_mm_mmask_i32gather_epi32): Ditto.
40917         (_mm256_mmask_i32gather_epi64): Ditto.
40918         (_mm_mmask_i32gather_epi64): Ditto.
40919         (_mm256_mmask_i64gather_epi32): Ditto.
40920         (_mm_mmask_i64gather_epi32): Ditto.
40921         (_mm256_mmask_i64gather_epi64): Ditto.
40922         (_mm_mmask_i64gather_epi64): Ditto.
40923         (_mm256_i32scatter_ps): Change __addr type to void*.
40924         (_mm256_mask_i32scatter_ps): Ditto.
40925         (_mm_i32scatter_ps): Ditto.
40926         (_mm_mask_i32scatter_ps): Ditto.
40927         (_mm256_i32scatter_pd): Ditto.
40928         (_mm256_mask_i32scatter_pd): Ditto.
40929         (_mm_i32scatter_pd): Ditto.
40930         (_mm_mask_i32scatter_pd): Ditto.
40931         (_mm256_i64scatter_ps): Ditto.
40932         (_mm256_mask_i64scatter_ps): Ditto.
40933         (_mm_i64scatter_ps): Ditto.
40934         (_mm_mask_i64scatter_ps): Ditto.
40935         (_mm256_i64scatter_pd): Ditto.
40936         (_mm256_mask_i64scatter_pd): Ditto.
40937         (_mm_i64scatter_pd): Ditto.
40938         (_mm_mask_i64scatter_pd): Ditto.
40939         (_mm256_i32scatter_epi32): Ditto.
40940         (_mm256_mask_i32scatter_epi32): Ditto.
40941         (_mm_i32scatter_epi32): Ditto.
40942         (_mm_mask_i32scatter_epi32): Ditto.
40943         (_mm256_i32scatter_epi64): Ditto.
40944         (_mm256_mask_i32scatter_epi64): Ditto.
40945         (_mm_i32scatter_epi64): Ditto.
40946         (_mm_mask_i32scatter_epi64): Ditto.
40947         (_mm256_i64scatter_epi32): Ditto.
40948         (_mm256_mask_i64scatter_epi32): Ditto.
40949         (_mm_i64scatter_epi32): Ditto.
40950         (_mm_mask_i64scatter_epi32): Ditto.
40951         (_mm256_i64scatter_epi64): Ditto.
40952         (_mm256_mask_i64scatter_epi64): Ditto.
40953         (_mm_i64scatter_epi64): Ditto.
40954         (_mm_mask_i64scatter_epi64): Ditto.
40955         * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
40956         (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
40957         (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
40958         (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
40959         (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
40960         (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
40961         (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
40962         (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
40963         (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
40964         (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
40965         (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
40966         (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
40967         (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
40968         (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
40969         (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
40970         (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
40971         (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
40972         (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
40973         (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
40974         (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
40975         (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
40976         (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
40977         (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
40978         (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
40979         (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
40980         (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
40981         (VOID_QI_V8DI_PCINT_INT_INT): Remove.
40982         (V16SF_V16SF_PCVOID_V16SI_HI_INT,  V8DF_V8DF_PCVOID_V8SI_QI_INT)
40983         (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
40984         (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
40985         (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
40986         (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
40987         (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
40988         (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
40989         (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
40990         (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
40991         (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
40992         (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
40993         (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
40994         (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
40995         (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
40996         (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
40997         (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
40998         (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
40999         (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
41000         (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
41001         (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
41002         (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
41003         (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
41004         (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
41005         (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
41006         (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
41007         (VOID_QI_V8DI_PCVOID_INT_INT): Add.
41008         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
41009         definitions accordingly.
41011 2017-01-17  Kito Cheng  <kito.cheng@gmail.com>
41012             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
41014         PR target/79079
41015         * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
41016         gen_lowpart.
41018 2017-01-17  Vladimir Makarov  <vmakarov@redhat.com>
41020         PR target/79058
41021         * ira-conflicts.c (ira_build_conflicts): Update total conflict
41022         hard regs for inner regno.
41024 2017-01-17  Martin Liska  <mliska@suse.cz>
41026         PR ipa/71207
41027         * ipa-polymorphic-call.c (contains_type_p): Fix wrong
41028         assumption and add comment.
41030 2017-01-17  Nathan Sidwell  <nathan@acm.org>
41032         * ipa-visibility.c (localize_node): New function, broken out of ...
41033         (function_and_variable_visibility): ... here. Call it.
41035 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
41037         PR middle-end/77445
41038         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
41039         correctly set frequency of oudgoing edge.
41040         (duplicate_thread_path): Fix profile updating.
41042 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
41044         PR other/79046
41045         * configure.ac: Add GCC_BASE_VER.
41046         * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
41047         version from BASE-VER file.
41048         (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
41049         (gcc.o): Depend on $(BASEVER).
41050         * common.opt (dumpfullversion): New option.
41051         * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
41052         * doc/invoke.texi: Document -dumpfullversion.
41053         * doc/install.texi: Document --with-gcc-major-version-only.
41054         * configure: Regenerated.
41056 2017-01-17  Richard Biener  <rguenther@suse.de>
41058         PR tree-optimization/71433
41059         * tree-vrp.c (register_new_assert_for): Merge same asserts
41060         on all incoming edges.
41061         (process_assert_insertions_for): Handle insertions at the
41062         beginning of BBs.
41064 2017-01-17  Gerald Pfeifer  <gerald@pfeifer.com>
41066         * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
41067         * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
41069 2017-01-17  Kaz Kojima  <kkojima@gcc.gnu.org>
41071         PR target/78633
41072         * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
41073         RTL sharing.
41075 2017-01-17  Alan Modra  <amodra@gmail.com>
41077         PR target/79066
41078         * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
41079         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
41080         symbolic stack limit when pic.
41082 2017-01-16  Martin Sebor  <msebor@redhat.com>
41084         PR tree-optimization/78608
41085         * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
41087 2017-01-16  Jeff Law  <law@redhat.com>
41089         Revert:
41090         2016-12-02  Tadek Kijkowski  <tkijkowski@gmail.com>
41091         * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
41092         for several include directories that may be relative to sysroot.
41093         * config/i386/x-mingw32 (gplus_includedir): Define.
41094         (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
41095         (native_system_includedir): Likewise.
41096         * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
41097         override if TARGET_SYSTEM_ROOT is defined.
41098         (NATIVE_SYSTEM_HEADER_DIR): Likewise.
41100         PR tree-optimization/79090
41101         PR tree-optimization/33562
41102         PR tree-optimization/61912
41103         PR tree-optimization/77485
41104         * tree-ssa-dse.c (compute_trims): Accept STMT argument.  Dump STMT
41105         and computed trims into the dump file.
41107 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
41109         * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
41111 2017-01-16  Jakub Jelinek  <jakub@redhat.com>
41113         PR c/79089
41114         * gimplify.c (gimplify_init_constructor): If want_value and
41115         object == lhs, unshare lhs to avoid invalid tree sharing.  Formatting
41116         fix.
41118         PR target/79080
41119         * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
41120         sequence.  Formatting fixes.
41121         (doloop_optimize): Formatting fixes.
41123         PR driver/49726
41124         * gcc.c (debug_level_greater_than_spec_func): New function.
41125         (static_spec_functions): Add debug-level-gt spec function.
41126         (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
41127         !g0.
41128         * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41129         * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41130         * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
41131         gpubnames, ggnu-pubnames, gno-record-gcc-switches,
41132         grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
41133         gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
41135 2017-01-16  Uros Bizjak  <ubizjak@gmail.com>
41137         * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
41138         QImode fixups to general and mask registers only.
41140 2017-01-16  Carl Love  <cel@us.ibm.com>
41142         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
41143         for built-in functions
41144         vector signed char vec_nabs (vector signed char)
41145         vector signed short vec_nabs (vector signed short)
41146         vector signed int vec_nabs (vector signed int)
41147         vector signed long long vec_nabs (vector signed long long)
41148         vector float vec_nabs (vector float)
41149         vector double vec_nabs (vector double)
41150         * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
41151         and NABS overload.
41152         * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
41153         * config/rs6000/altivec.h: New define for vec_nabs built-in function.
41154         * doc/extend.texi: Update the documentation file for the new built-in
41155         functions.
41157 2017-01-16  Martin Sebor  <msebor@redhat.com>
41159         * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
41160         message.
41162 2017-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41164         * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
41165         UNSPEC_VSX__XXSPLTD to require special splat handling.
41167 2017-01-16  David Malcolm  <dmalcolm@redhat.com>
41169         PR bootstrap/78616
41170         * system.h: Poison strndup.
41172 2017-01-16  Alan Modra  <amodra@gmail.com>
41174         PR target/79098
41175         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
41176         use a switch.
41178 2017-01-16  Georg-Johann Lay  <avr@gjlay.de>
41180         * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
41182 2017-01-15  Uros Bizjak  <ubizjak@gmail.com>
41184         * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
41185         call recog here.  Assert that INSN_CODE (insn) is non-negative.
41187 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41189         PR target/72749
41190         * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
41191         fallthrough.
41192         * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
41193         in the currently scheduled RTL fragment.
41195 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41197         PR rtl-optimization/78751
41198         * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
41199         give up.
41201 2017-01-14  Jeff Law  <law@redhat.com>
41203         PR tree-optimization/79090
41204         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
41205         variable length stores.
41206         (compute_trims): Delete dead assignment to *trim_tail.
41207         (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
41208         zero length.
41210 2017-01-14  Bernd Schmidt  <bschmidt@redhat.com>
41212         PR rtl-optimization/78626
41213         PR rtl-optimization/78727
41214         * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
41215         of a block, and split such blocks after everything else is finished.
41217 2017-01-14  Alan Modra  <amodra@gmail.com>
41219         PR target/72749
41220         * combine.c (recog_for_combine_1): Set INSN_CODE before calling
41221         target legitimate_combined_insn.
41222         * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
41223         (rs6000_legitimate_combined_insn): New function.
41224         * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
41225         all uses.
41226         (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
41227         (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
41228         (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
41230 2017-01-14  Gerald Pfeifer  <gerald@pfeifer.com>
41232         * doc/frontends.texi (G++ and GCC): Remove references to Java.
41234 2017-01-13  Jeff Law  <law@redhat.com>
41236         PR tree-optimization/33562
41237         PR tree-optimization/61912
41238         PR tree-optimization/77485
41239         * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
41240         a statement.
41241         (delete_dead_assignment): Likewise.
41242         (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
41243         statement to delete_dead_call and delete_dead_assignment.
41245 2017-01-13  David Malcolm  <dmalcolm@redhat.com>
41247         PR c/78304
41248         * substring-locations.c (format_warning_va): Strengthen case 1 so
41249         that both endpoints of the substring must be within the format
41250         range for just the substring to be printed.
41252 2017-01-13  Uros Bizjak  <ubizjak@gmail.com>
41254         * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41255         * config/i386/i386.c (ix86_target_string): Add missing options
41256         to isa_opts and reorder options by implied ISAs.  Rename isa_opts2 to
41257         isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41258         flags_other and ix86_target_other to flags2_other.  Display unknown
41259         isa2 options.
41260         (ix86_valid_target_attribute_inner_p): Add missing options and
41261         reorder options by implied ISAs, as in ix86_target_string.
41263 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41265         * hash-table.h (hash_table::too_empty_p): New function.
41266         (hash_table::expand): Use it.
41267         (hash_table::traverse): Likewise.
41268         (hash_table::empty_slot): Use sizeof (value_type) instead of
41269         sizeof (PTR) to convert bytes to elements.  Shrink the table
41270         if the current size is excessive for the current number of
41271         elements.
41273 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41275         * ira-costs.c (record_reg_classes): Break from the inner loop
41276         early once alt_fail is known to be true.  Update outer loop
41277         handling accordingly.
41279 2017-01-13  Jeff Law  <law@redhat.com>
41281         * tree-ssa-dse.c (decrement_count): New function.
41282         (increment_start_addr, maybe_trim_memstar_call): Likewise.
41283         (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41284         when we know the partially dead statement is a mem* function.
41286         PR tree-optimization/61912
41287         PR tree-optimization/77485
41288         * tree-ssa-dse.c: Include expr.h.
41289         (maybe_trim_constructor_store): New function.
41290         (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41292         PR tree-optimization/33562
41293         PR tree-optimization/61912
41294         PR tree-optimization/77485
41295         * doc/invoke.texi: Document new dse-max-object-size param.
41296         * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41297         * tree-ssa-dse.c: Include params.h.
41298         (dse_store_status): New enum.
41299         (initialize_ao_ref_for_dse): New, partially extracted from
41300         dse_optimize_stmt.
41301         (valid_ao_ref_for_dse, normalize_ref): New.
41302         (setup_live_bytes_from_ref, compute_trims): Likewise.
41303         (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41304         (maybe_trim_partially_dead_store): Likewise.
41305         (maybe_trim_complex_store): Likewise.
41306         (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41307         Track what bytes live from the original store.  Return tri-state
41308         for dead, partially dead or live.
41309         (dse_dom_walker): Add constructor, destructor and new private members.
41310         (delete_dead_call, delete_dead_assignment): New extracted from
41311         dse_optimize_stmt.
41312         (dse_optimize_stmt): Make a member of dse_dom_walker.
41313         Use initialize_ao_ref_for_dse.
41315         PR tree-optimization/33562
41316         PR tree-optimization/61912
41317         PR tree-optimization/77485
41318         * sbitmap.h (bitmap_count_bits): Prototype.
41319         (bitmap_clear_range, bitmap_set_range): Likewise.
41320         * sbitmap.c (bitmap_clear_range): New function.
41321         (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41323 2017-01-13  Martin Liska  <mliska@suse.cz>
41325         PR ipa/79043
41326         * function.c (set_cfun): Add new argument force.
41327         * function.h (set_cfun): Likewise.
41328         * ipa-inline-transform.c (inline_call): Use the function when
41329         strict alising from is dropped for function we inline to.
41331 2017-01-13  Richard Biener  <rguenther@suse.de>
41333         * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41334         for dumping GIMPLE INTEGER_CSTs.
41336 2017-01-13  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
41338         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41339         to 201112L since C++17.
41341 2017-01-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
41343         PR sanitizer/78887
41344         * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41345         if -fsanitize=kernel-address is present.
41347 2017-01-13  Richard Biener  <rguenther@suse.de>
41349         * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41350         as _Literal ( type ) number in case usual suffixes do not
41351         preserve all information.
41353 2017-01-13  Richard Biener  <rguenther@suse.de>
41355         PR tree-optimization/77283
41356         * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41357         and ssa-iterators.h.
41358         (is_feasible_trace): Implement a cost model based on joiner
41359         PHI node uses.
41361 2017-01-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
41363         PR target/79004
41364         * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41365         char or short to __float128/_Float128 directly.
41367 2017-01-12  Martin Sebor  <msebor@redhat.com>
41369         to -Wformat-overflow.
41370         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41371         (min_bytes_remaining): Same.
41372         (get_string_length): Same.
41373         (format_string): Same.
41374         (format_directive): Same.
41375         (add_bytes): Same.
41376         (pass_sprintf_length::handle_gimple_call): Same.
41378 2017-01-12  Jakub Jelinek  <jakub@redhat.com>
41380         * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41381         info.nowrite calls with no lhs that can't throw.  Return bool
41382         whether gsi_remove has been called or not.
41383         (pass_sprintf_length::handle_gimple_call): Return bool whether
41384         try_substitute_return_value called gsi_remove.  Formatting fix.
41385         (pass_sprintf_length::execute): Don't use gsi_remove if
41386         handle_gimple_call returned true.
41388         PR bootstrap/79069
41389         * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41390         be removed due to side-effects, don't remove following barrier nor
41391         turn the successor edge into fallthru edge.
41393 2017-01-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41395         PR target/79044
41396         * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41397         element-reversing loads and stores as not swappable.
41399 2017-01-12  Nathan Sidwell  <nathan@acm.org>
41400             Nicolai Stange  <nicstange@gmail.com>
41402         * combine.c (try_combine): Don't ignore result of overlap checking
41403         loop.  Combine overlap & asm check into single loop.
41405 2017-01-12  Richard Biener  <rguenther@suse.de>
41407         * tree-pretty-print.c (dump_generic_node): Provide -gimple
41408         variant for MEM_REF.  Sanitize INTEGER_CST for -gimple.
41410 2017-01-12  Richard Biener  <rguenther@suse.de>
41412         * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41413         and TS_TARGET_OPTION directly derive from TS_BASE.
41414         * tree-core.h (tree_optimization_option): Derive from tree_base.
41415         (tree_target_option): Likewise.
41417 2017-01-11  Uros Bizjak  <ubizjak@gmail.com>
41419         * config/i386/i386.c (memory_address_length): Increase len
41420         only when rip_relative_addr_p returns false.
41422 2017-01-11  Julia Koval  <julia.koval@intel.com>
41424         * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41425         (OPTION_MASK_ISA_SGX_SET): New.
41426         (ix86_handle_option): Handle OPT_msgx.
41427         * config.gcc: Added sgxintrin.h.
41428         * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41429         * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41430         * config/i386/i386.c (ix86_target_string): Add -msgx.
41431         (PTA_SGX): New.
41432         (ix86_option_override_internal): Handle new options.
41433         (ix86_valid_target_attribute_inner_p): Add sgx.
41434         * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41435         * config/i386/i386.opt: Add msgx.
41436         * config/i386/sgxintrin.h: New file.
41437         * config/i386/x86intrin.h: Add sgxintrin.h.
41439 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41441         PR c++/71537
41442         * fold-const.c (maybe_nonzero_address): Return 1 for function
41443         local objects.
41444         (tree_single_nonzero_warnv_p): Don't handle function local objects
41445         here.
41447         PR c++/72813
41448         * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41449         of c-header.
41451 2017-01-11  David Malcolm  <dmalcolm@redhat.com>
41453         PR driver/78877
41454         * opts.c: Include "spellcheck.h"
41455         (struct string_fragment): New struct.
41456         (struct edit_distance_traits<const string_fragment &>): New
41457         struct.
41458         (get_closest_sanitizer_option): New function.
41459         (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41461 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41463         * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41464         by 12.
41465         (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41466         DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41467         (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41468         (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41469         for initial die_offset if dwarf_split_debug_info.
41470         (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41471         initial next_die_offset if dwo_id is non-NULL.  Don't emit padding
41472         fields.
41473         (output_skeleton_debug_sections): Formatting fix.  Use
41474         DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41475         DWARF_COMPILE_UNIT_HEADER_SIZE.  Don't emit padding.
41477 2017-01-11  Wilco Dijkstra  <wdijkstr@arm.com>
41479         * config/arm/cortex-a53.md: Add bypasses for
41480         cortex_a53_r2f_cvt.
41481         (cortex_a53_r2f): Only use for transfers.
41482         (cortex_a53_f2r): Likewise.
41483         (cortex_a53_r2f_cvt): Add reservation for conversions.
41484         (cortex_a53_f2r_cvt): Likewise.
41486 2017-01-11  Tamar Christina  <tamar.christina@arm.com>
41488         * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41489         to all inlined functions, change static to extern.
41491 2017-01-11  Christophe Lyon  <christophe.lyon@linaro.org>
41493         PR target/78253
41494         * config/arm/arm.c (legitimize_pic_address): Handle reference to
41495         weak symbol.
41496         (arm_assemble_integer): Likewise.
41498 2017-01-11  Richard Earnshaw  <rearnsha@arm.com>
41500         * config.gcc: Use new awk script to check CPU, FPU and architecture
41501         parameters for --with-... options.
41502         * config/arm/parsecpu.awk: New file
41503         * config/arm/arm-cpus.in: New file.
41504         * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41505         files.
41506         * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41507         files.
41508         * config/arm/t-arm: Update dependency rules.
41509         * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41510         of processing .def files.
41511         * config/arm/genopt.sh: Deleted.
41512         * config/arm/gentune.sh: Deleted.
41513         * config/arm/arm-cores.def: Deleted.
41514         * config/arm/arm-arches.def: Deleted.
41515         * config/arm/arm-fpus.def: Deleted.
41516         * config/arm/arm-tune.md: Regenerated.
41517         * config/arm/arm-tables.opt: Regenerated.
41518         * config/arm/arm-cpu.h: New generated file.
41519         * config/arm/arm-cpu-data.h: New generated file.
41520         * config/arm/arm-cpu-cdata.h: New generated file.
41522 2017-01-11  Maxim Ostapenko  <m.ostapenko@samsung.com>
41524         PR lto/79042
41525         * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41526         bit.
41527         (input_varpool_node): Unpack dynamically_initialized bit.
41529 2017-01-11  Eric Botcazou  <ebotcazou@adacore.com>
41531         PR rtl-optimization/79032
41532         * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41533         the alignment of the adjusted memory reference against that of MODE,
41534         instead of the alignment of the original memory reference.
41536 2017-01-11  Martin Jambor  <mjambor@suse.cz>
41538         * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41539         test.
41540         * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41541         decorated functions.
41543 2017-01-11  Richard Biener  <rguenther@suse.de>
41545         * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41546         set range/nonnull info for PHI results.  Do not set it on
41547         stmts marked for removal.
41549 2017-01-10  Eric Botcazou  <ebotcazou@adacore.com>
41551         * expr.c (store_field): In the bitfield case, fetch the return value
41552         from the registers before applying a single big-endian adjustment.
41553         Always do a final load for a BLKmode value not larger than a word.
41555 2017-01-10  David Malcolm  <dmalcolm@redhat.com>
41557         PR c++/77949
41558         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41559         that we correctly handle column numbers greater than
41560         LINE_MAP_MAX_COLUMN_NUMBER.
41562 2017-01-10  Martin Sebor  <msebor@redhat.com>
41564         PR middle-end/78245
41565         * gimple-ssa-sprintf.c (get_destination_size): Call
41566         {init,fini}object_sizes.
41567         * tree-object-size.c (addr_object_size): Adjust.
41568         (pass_through_call): Adjust.
41569         (pass_object_sizes::execute): Adjust.
41570         * tree-object-size.h (fini_object_sizes): Declare.
41572 2017-01-10  Martin Sebor  <msebor@redhat.com>
41574         PR tree-optimization/78775
41575         * builtins.c (get_size_range): Move...
41576         * calls.c: ...to here.
41577         (alloc_max_size): Accept zero argument.
41578         (operand_signed_p): Remove.
41579         (maybe_warn_alloc_args_overflow): Call get_size_range.
41580         * calls.h (get_size_range): Declare.
41582 2017-01-10  Joe Seymour  <joe.s@somniumtech.com>
41584         * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41585         from TI's devices.csv file as of September 2016.
41586         * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41588 2017-01-10  Sandra Loosemore  <sandra@codesourcery.com>
41590         * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41591         * doc/invoke.texi: Likewise.
41592         * doc/md.texi: Likewise.
41593         * doc/objc.texi: Likewise.
41595 2017-01-10  Joshua Conner  <joshconner@google.com>
41597         * config/arm/fuchsia-elf.h: New file.
41598         * config/fuchsia.h: New file.
41599         * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41600         (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41601         targets.
41602         * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41604 2016-01-10  Richard Biener  <rguenther@suse.de>
41606         PR tree-optimization/79034
41607         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41608         Propagate out degenerate PHIs in the joiner.
41610 2017-01-10  Martin Liska  <mliska@suse.cz>
41612         * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41613         (sort_congruence_classes_by_decl_uid): Likewise.
41614         (sort_congruence_class_groups_by_decl_uid): Likewise.
41615         (sem_item_optimizer::merge_classes): Sort class, groups in these
41616         classes and members in the groups by DECL_UID of declarations.
41617         This would make merge operations stable.
41619 2017-01-10  Martin Liska  <mliska@suse.cz>
41621         * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41622         usage of m_classes_vec.
41623         (sem_item_optimizer::~sem_item_optimizer):  Likewise.
41624         (sem_item_optimizer::get_group_by_hash): Likewise.
41625         (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41626         (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41627         (sem_item_optimizer::verify_classes): Likewise.
41628         (sem_item_optimizer::process_cong_reduction): Likewise.
41629         (sem_item_optimizer::dump_cong_classes): Likewise.
41630         (sem_item_optimizer::merge_classes): Likewise.
41631         * ipa-icf.h (congruence_class_hash): Rename from
41632         congruence_class_group_hash.  Remove declaration of m_classes_vec.
41634 2017-01-10  Andrew Senkevich  <andrew.senkevich@intel.com>
41636         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41637         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41638         * config.gcc: Add avx512vpopcntdqintrin.h.
41639         * config/i386/avx512vpopcntdqintrin.h: New.
41640         * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41641         * config/i386/i386-builtin-types.def: Add new types.
41642         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41643         __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41644         __builtin_ia32_vpopcountq_v8di_mask): New.
41645         * config/i386/i386-c.c (ix86_target_macros_internal): Define
41646         __AVX512VPOPCNTDQ__.
41647         * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41648         (PTA_AVX512VPOPCNTDQ): Define.
41649         * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41650         TARGET_AVX512VPOPCNTDQ_P): Define.
41651         * config/i386/i386.opt: Add mavx512vpopcntdq.
41652         * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41653         * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41655 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41657         PR middle-end/77484
41658         * predict.def (PRED_CALL): Set to 67.
41660 2017-01-09  Eric Botcazou  <ebotcazou@adacore.com>
41662         * expr.c (store_field): In the bitfield case, if the value comes from
41663         a function call and is of an aggregate type returned in registers, do
41664         not modify the field mode; extract the value in all cases if the mode
41665         is BLKmode and the size is not larger than a word.
41667 2017-01-09  Dominique d'Humieres  <dominiq@lps.ens.fr>
41669         PR target/71017
41670         * config/i386/cpuid.h: Fix undefined behavior.
41672 2017-01-04  Jeff Law  <law@redhat.com>
41674         PR tree-optimization/79007
41675         PR tree-optimization/67955
41676         * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41677         conservative for pt.null when flag_non_call_exceptions is on.
41679 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41681         PR translation/79019
41682         PR translation/79020
41683         * params.def (PARAM_INLINE_MIN_SPEEDUP,
41684         PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41685         PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41686         in descriptions.
41687         * config/avr/avr.opt (maccumulate-args): Likewise.
41688         * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41689         * common.opt (freport-bug): Likewise.
41690         * cif-code.def (CIF_FINAL_ERROR): Likewise.
41691         * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41692         * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41693         translatable string.
41694         * config/i386/i386.c (function_value_32): Likewise.
41695         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41696         * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41697         Likewise.
41698         * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41699         * common/config/msp430/msp430-common.c (msp430_handle_option):
41700         Likewise.
41701         * symtab.c (symtab_node::verify_base): Likewise.
41702         * opts.c (set_debug_level): Likewise.
41703         * tree.c (verify_type_variant): Likewise.  Fix typo in comment.
41704         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41705         missing whitespace to translatable strings.
41706         * config/avr/avr.md (bswapsi2): Fix typo in comment.
41707         * config/sh/superh.h: Likewise.
41708         * config/i386/xopintrin.h: Likewise.
41709         * config/i386/znver1.md: Likewise.
41710         * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41711         * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41712         * double-int.h (struct double_int): Likewise.
41713         * double-int.c (div_and_round_double): Likewise.
41714         * wide-int.cc: Likewise.
41715         * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41716         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41717         * cfgcleanup.c (crossjumps_occured): Renamed to ...
41718         (crossjumps_occurred): ... this.
41719         (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41720         Adjust all uses.
41722         PR tree-optimization/78899
41723         * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41724         returning bool return struct loop *, NULL for failure and the new
41725         loop on success.
41726         (versionable_outer_loop_p): Don't version outer loop if it has
41727         dont_vectorized bit set.
41728         (tree_if_conversion): When versioning outer loop, ensure
41729         tree_if_conversion is performed also on the inner loop of the
41730         non-vectorizable outer loop copy.
41731         * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix.  Fold
41732         LOOP_VECTORIZED in inner loop of the scalar outer loop and
41733         prevent vectorization of it.
41734         (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41735         the outer loop vectorization of the non-scalar version is attempted
41736         before vectorization of the inner loop in scalar version.  If
41737         outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41738         vectorization of its inner loop.
41739         * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41740         has 2 inner loops, rename also on edges from bb whose single pred
41741         is outer_loop->header.  Fix typo in function comment.
41743 2017-01-09  Martin Sebor  <msebor@redhat.com>
41745         PR bootstrap/79033
41746         * asan.c (asan_emit_stack_protection): Increase local buffer size
41747         to avoid snprintf truncation warning.
41749 2017-01-09  Andrew Pinski  <apinski@cavium.com>
41751         * config/aarch64/aarch64-cores.def: Add thunderx2t99.  Change vulcan
41752         to reference thunderx2t99 for the tuning structure
41753         * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41754         Rename to ...
41755         (thunderx2t99_extra_costs): This.
41756         * config/aarch64/aarch64-tune.md: Regenerate.
41757         * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41758         (vulcan_addrcost_table): This.
41759         (vulcan_regmove_cost): Rename to ...
41760         (thunderx2t99_regmove_cost): This.
41761         (vulcan_vector_cost): Rename to ...
41762         (thunderx2t99_vector_cost): this.
41763         (vulcan_branch_cost): Rename to ...
41764         (thunderx2t99_branch_cost): This.
41765         (vulcan_tunings): Rename to ...
41766         (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41767         * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41769 2017-01-09  Martin Jambor  <mjambor@suse.cz>
41771         PR ipa/78365
41772         PR ipa/78599
41773         * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41774         * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41775         (propagate_vr_accross_jump_function): Use the above function for all
41776         value range computations for pass-through jump functions and type
41777         converasion from explicit value range values.
41778         (ipcp_propagate_stage): Do not attempt to deduce types of formal
41779         parameters from TYPE_ARG_TYPES.
41780         * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41781         (ipa_write_node_info): Stream type of the actual argument.
41782         (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41784 2017-01-09  Martin Liska  <mliska@suse.cz>
41786         PR pch/78970
41787         * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41788         (lookup_compiler): Do not show error message with have_E.
41790 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41792         PR tree-optimization/78938
41793         * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41794         where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41795         BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41796         {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR.  Formatting
41797         fixes.
41799 2017-01-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
41801         * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41802         is const0_rtx.
41804 2017-01-09  Richard Biener  <rguenther@suse.de>
41806         PR tree-optimization/78997
41807         * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41808         name condition properly.
41810 2017-01-09  Richard Biener  <rguenther@suse.de>
41812         PR debug/79000
41813         * dwarf2out.c (is_cxx): New overload with context.
41814         (is_naming_typedef_decl): Use it.
41816 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41818         * invoke.texi (Option Summary): Correct spacing in option lists
41819         and add line breaks to fix over-long lines.
41821 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41823         PR middle-end/17660
41825         * extend.texi (Common Variable Attributes): Add xref to GCC
41826         Internals manual to explain mode attribute keywords.
41828 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41830         PR other/16519
41831         * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41832         and Preprocessor Options.
41833         (Options for Linking): Document -pthread here....
41834         (RS/6000 and PowerPC Options): ...not here.
41835         (Solaris 2 Options): ...or here.
41836         * doc/cppopts.texi: Document -pthread.
41838 2017-01-08  Martin Sebor  <msebor@redhat.com>
41840         PR middle-end/77708
41841         * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41842         * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41843         New member functions.
41844         (format_directive): Used them.
41845         (add_bytes): Same.
41846         (pass_sprintf_length::handle_gimple_call): Same.
41847         * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41848         to avoid truncation for any argument.
41849         (extract_affine_mul): Same.
41850         * tree.c (get_file_function_name): Same.
41852 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41854         PR middle-end/77484
41855         * predict.def (PRED_INDIR_CALL): Set to 86.
41857 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41859         PR preprocessor/54124
41860         * doc/cppopts.texi: Reformat -d subtable to list the full name
41861         of the options.  Add cross-reference to the docs for the general
41862         compiler -d options.
41863         * doc/invoke.texi (Developer Options): Add cross-reference to the
41864         preprocessor-specific -d option documentation.
41866 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41868         PR preprocessor/13498
41869         * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41870         redudant material, and reflect new command-line options.
41871         (System Headers): Likewise.
41873 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41875         * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41876         -isystem, and -idirafter.  Copy-edit.
41877         * doc/cppopts.texi: Copy-edit.  Remove contradiction about
41878         default for -ftrack-macro-expansion.  Delete obsolete and
41879         badly-formatted implementation details about -fdebug-cpp output.
41880         * doc/cppwarnopts.texi: Copy-edit.
41882 2017-01-07  David Malcolm  <dmalcolm@redhat.com>
41884         PR c++/72803
41885         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41886         that the transition from a max line width >= 1<<10 to narrower
41887         lines works correctly.
41889 2017-01-07  Alexandre Oliva  <aoliva@redhat.com>
41891         * doc/options.texi (PerFunction): New.
41892         * opt-functions.awk (switch_flags): Map both Optimization and
41893         PerFunction to CL_OPTIMIZATION.
41894         * opth-gen.awk: Test for PerFunction flag along with
41895         Optimization.
41896         * optc-save-gen.awk: Likewise.  Introduce var_opt_hash and set
41897         it only when the latter is present.  Skip those that don't in
41898         the hash function generator.
41899         * common.opt (fvar-tracking): Mark as PerFunction instead of
41900         Optimization.
41901         (fvar-tracking-assignments): Likewise.
41902         (fvar-tracking-assignments-toggle): Likewise.
41903         (fvar-tracking-uninit): Likewise.
41905 2017-01-07  Jakub Jelinek  <jakub@redhat.com>
41907         PR translation/79018
41908         * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41909         the and store.
41911 2017-01-06  Mikael Pettersson  <mikpelinux@gmail.com>
41913         PR target/57583
41914         * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41915         * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41916         TARGET_LONG_JUMP_TABLE_OFFSETS.
41917         * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
41918         * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
41919         * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
41920         (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
41921         * config/m68k/m68k.md (tablejump expander): Likewise.
41922         (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
41923         TARGET_LONG_JUMP_TABLE_OFFSETS.
41924         (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
41925         * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
41927 2017-01-06  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
41928             David Holsgrove <david.holsgrove@xilinx.com>
41930         * common/config/microblaze/microblaze-common.c
41931         (TARGET_EXCEPT_UNWIND_INFO): Remove.
41932         * config/microblaze/microblaze-protos.h (microblaze_eh_return):
41933         New prototype.
41934         * config/microblaze/microblaze.c (microblaze_must_save_register)
41935         (microblaze_expand_epilogue, microblaze_return_addr): Handle
41936         calls_eh_return.
41937         (microblaze_eh_return): New function.
41938         * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
41939         (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
41940         (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
41941         * config/microblaze/microblaze.md (eh_return): New pattern.
41943 2017-01-06  Jakub Jelinek  <jakub@redhat.com>
41945         * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
41946         GCC_DIAGNOSTIC_STRINGIFY): Define.
41948         * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
41950 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41952         * config/arm/arm.md (<mcrr>): New.
41953         (<mrrc>): New.
41954         * config/arm/arm.c (arm_arch5te): New.
41955         (arm_option_override): Set arm_arch5te.
41956         (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
41957         and mrrc2.
41958         * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
41959         (arm_mcrr_qualifiers): ... this. New.
41960         (MRRC_QUALIFIERS): Define to...
41961         (arm_mrrc_qualifiers): ... this. New.
41962         * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
41963         __arm_mrrc2): New.
41964         * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
41965         * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
41966         (MRRCI, mrrc, MRRC): New.
41967         * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
41968         VUNSPEC_MRRC2): New.
41970 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41972         * config/arm/arm.md (<mcr>): New.
41973         (<mrc>): New.
41974         * config/arm/arm.c (arm_coproc_builtin_available): Add
41975         support for mcr, mrc, mcr2 and mrc2.
41976         * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
41977         (arm_mcr_qualifiers): ... this. New.
41978         (MRC_QUALIFIERS): Define to ...
41979         (arm_mrc_qualifiers): ... this. New.
41980         (MCR_QUALIFIERS): Define to ...
41981         (arm_mcr_qualifiers): ... this. New.
41982         * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
41983         __arm_mrc2): New.
41984         * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
41985         * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
41986         * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
41987         VUNSPEC_MRC2): New.
41989 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41991         * config/arm/arm.md (*ldc): New.
41992         (*stc): New.
41993         (<ldc>): New.
41994         (<stc>): New.
41995         * config/arm/arm.c (arm_coproc_builtin_available): Add
41996         support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
41997         (arm_coproc_ldc_stc_legitimate_address): New.
41998         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
41999         'qualifier_const_pointer'.
42000         (LDC_QUALIFIERS): Define to...
42001         (arm_ldc_qualifiers): ... this. New.
42002         (STC_QUALIFIERS): Define to...
42003         (arm_stc_qualifiers): ... this. New.
42004         * config/arm/arm-protos.h
42005         (arm_coproc_ldc_stc_legitimate_address): New.
42006         * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
42007         __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
42008         * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
42009         stc2, stcl, stc2l): New.
42010         * config/arm/constraints.md (Uz): New.
42011         * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
42012         * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
42013         VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
42014         VUNSPEC_STC2L): New.
42016 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42018         * config/arm/arm.md (<cdp>): New.
42019         * config/arm/arm.c (neon_const_bounds): Rename this ...
42020         (arm_const_bounds): ... this.
42021         (arm_coproc_builtin_available): New.
42022         * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
42023         (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
42024         (CDP_QUALIFIERS): Define to...
42025         (arm_cdp_qualifiers): ... this. New.
42026         (void_UP): Define.
42027         (arm_expand_builtin_args): Add case for 6 arguments.
42028         * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
42029         (arm_const_bounds): ... this.
42030         (arm_coproc_builtin_available): New.
42031         * config/arm/arm_acle.h (__arm_cdp): New.
42032         (__arm_cdp2): New.
42033         * config/arm/arm_acle_builtins.def (cdp): New.
42034         (cdp2): New.
42035         * config/arm/iterators.md (CDPI,CDP,cdp): New.
42036         * config/arm/neon.md: Rename all 'neon_const_bounds' to
42037         'arm_const_bounds'.
42038         * config/arm/types.md (coproc): New.
42039         * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
42040         * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
42041         * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
42042         arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
42044 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42046         * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
42047         (UBINOP_QUALIFIERS): New.
42048         (si_UP): Define.
42049         (acle_builtin_data): New. Change comment.
42050         (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
42051         ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
42052         ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
42053         arm_acle_builtins.def.
42054         (ARM_BUILTIN_ACLE_PATTERN_START): Define.
42055         (arm_init_acle_builtins): New.
42056         (CRC32_BUILTIN): Remove.
42057         (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
42058         crc32cb, crc32ch and crc32cw.
42059         (arm_init_crc32_builtins): Remove.
42060         (arm_init_builtins): Use arm_init_acle_builtins rather
42061         than arm_init_crc32_builtins.
42062         (arm_expand_acle_builtin): New.
42063         (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
42064         * config/arm/arm_acle_builtins.def: New.
42066 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42068         * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
42069         (arm_builtin_datum): ... this.
42070         (arm_init_neon_builtin): Rename to ...
42071         (arm_init_builtin): ... this. Add a new parameters PREFIX
42072         and USE_SIG_IN_NAME.
42073         (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
42074         'arm_init_builtin'. Replace type 'neon_builtin_datum' with
42075         'arm_builtin_datum'.
42076         (arm_init_vfp_builtins): Likewise.
42077         (builtin_arg): Rename enum's replacing 'NEON_ARG' with
42078         'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
42079         (arm_expand_neon_args): Rename to ...
42080         (arm_expand_builtin_args): ... this. Rename builtin_arg
42081         enum values and differentiate between ARG_BUILTIN_MEMORY
42082         and ARG_BUILTIN_NEON_MEMORY.
42083         (arm_expand_neon_builtin_1): Rename to ...
42084         (arm_expand_builtin_1): ... this. Rename builtin_arg enum
42085         values, arm_expand_builtin_args and add bool parameter NEON.
42086         (arm_expand_neon_builtin): Use arm_expand_builtin_1.
42087         (arm_expand_vfp_builtin): Likewise.
42088         (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
42090 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42092         PR middle-end/77484
42093         * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
42094         * predict.c (tree_estimate_probability_bb): Reverse direction of
42095         polymorphic call predictor.
42097 2017-01-06  David Malcolm  <dmalcolm@redhat.com>
42099         * passes.c (execute_one_pass): Split out pass-skipping logic into...
42100         (determine_pass_name_match): ...this new function and...
42101         (should_skip_pass_p): ...this new function.
42103 2017-01-06  Nathan Sidwell  <nathan@acm.org>
42105         * ipa-visibility.c (function_and_variable_visibility): Reformat
42106         comments and long lines.  Remove extrneous if.
42107         * symtab.c (symtab_node::make_decl_local): Fix code format.
42108         (symtab_node::set_section_for_node): Fix comment typo.
42110 2017-01-06  Martin Liska  <mliska@suse.cz>
42112         PR bootstrap/79003
42113         * lra-constraints.c: Rename invariant to lra_invariant.
42114         * predict.c (set_even_probabilities): Initialize e to NULL.
42116 2017-01-05  Martin Sebor  <msebor@redhat.com>
42118         PR tree-optimization/78910
42119         * gimple-ssa-sprintf.c (tree_digits): Add an argument.
42120         (format_integer): Correct off-by-one error in the handling
42121         of precision with negative numbers in signed conversions..
42123 2017-01-05  Eric Botcazou  <ebotcazou@adacore.com>
42125         * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
42127 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42129         PR tree-optimization/71016
42130         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
42131         factor_out_conditional_conversion.  Formatting fix.
42132         (factor_out_conditional_conversion): Add cond_stmt argument.
42133         If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
42134         cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
42135         Formatting fix.
42137 2017-01-05  David Malcolm  <dmalcolm@redhat.com>
42139         * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
42140         read-rtl-function.o, and selftest-rtl.o.
42141         * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
42142         (selftest::aarch64_test_loading_full_dump): New function.
42143         (selftest::aarch64_run_selftests): New function.
42144         (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
42145         selftest::aarch64_run_selftests.
42146         * config/i386/i386.c
42147         (selftest::ix86_test_loading_dump_fragment_1): New function.
42148         (selftest::ix86_test_loading_call_insn): New function.
42149         (selftest::ix86_test_loading_full_dump): New function.
42150         (selftest::ix86_test_loading_unspec): New function.
42151         (selftest::ix86_run_selftests): Call the new functions.
42152         * emit-rtl.c (maybe_set_max_label_num): New function.
42153         * emit-rtl.h (maybe_set_max_label_num): New decl.
42154         * function.c (instantiate_decls): Guard call to
42155         instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
42156         * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
42157         "static".
42158         * gensupport.c (gen_reader::gen_reader): Pass "false"
42159         for new "compact" param of rtx_reader.
42160         * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
42161         rather than an empty string for NULL strings.
42162         * read-md.c: Potentially include config.h rather than bconfig.h.
42163         Wrap include of errors.h with #ifdef GENERATOR_FILE.
42164         (have_error): New global, copied from errors.c.
42165         (md_reader::read_name): Rename to...
42166         (md_reader::read_name_1): ...this, adding "out_loc" param,
42167         and converting "missing name or number" to returning false, rather
42168         than failing.
42169         (md_reader::read_name): Reimplement in terms of read_name_1.
42170         (md_reader::read_name_or_nil): New function.
42171         (md_reader::read_string): Handle "(nil)" by returning NULL.
42172         (md_reader::md_reader): Add new param "compact".
42173         (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
42174         (md_reader::read_file): New method.
42175         * read-md.h (md_reader::md_reader): Add new param "compact".
42176         (md_reader::read_file): New method.
42177         (md_reader::is_compact): New accessor.
42178         (md_reader::read_name): Convert return type from void to file_location.
42179         (md_reader::read_name_or_nil): New decl.
42180         (md_reader::read_name_1): New decl.
42181         (md_reader::m_compact): New field.
42182         (noop_reader::noop_reader): Pass "false" for new "compact" param
42183         of rtx_reader.
42184         (rtx_reader::rtx_reader): Add new "compact" param.
42185         (rtx_reader::read_rtx_operand): Make virtual and convert return
42186         type from void to rtx.
42187         (rtx_reader::read_until): New decl.
42188         (rtx_reader::handle_any_trailing_information): New virtual function.
42189         (rtx_reader::postprocess): New virtual function.
42190         (rtx_reader::finalize_string): New virtual function.
42191         (rtx_reader::m_in_call_function_usage): New field.
42192         (rtx_reader::m_reuse_rtx_by_id): New field.
42193         * read-rtl-function.c: New file.
42194         * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
42195         * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
42196         (selftest::verify_three_block_rtl_cfg): New decl.
42197         * read-rtl-function.h: New file.
42198         * read-rtl.c: Potentially include config.h rather than bconfig.h.
42199         For host, include function.h, memmodel.h, and emit-rtl.h.
42200         (one_time_initialization): New function.
42201         (struct compact_insn_name): New struct.
42202         (compact_insn_names): New array.
42203         (find_code): Handle insn codes in compact dumps.
42204         (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
42205         (bind_subst_iter_and_attr): Likewise.
42206         (add_condition_to_string): Likewise.
42207         (add_condition_to_rtx): Likewise.
42208         (apply_attribute_uses): Likewise.
42209         (add_current_iterators): Likewise.
42210         (apply_iterators): Likewise.
42211         (initialize_iterators): Guard usage of apply_subst_iterator with
42212         #ifdef GENERATOR_FILE.
42213         (read_conditions): Wrap with #ifdef GENERATOR_FILE.
42214         (md_reader::read_mapping): Likewise.
42215         (add_define_attr_for_define_subst): Likewise.
42216         (add_define_subst_attr): Likewise.
42217         (read_subst_mapping): Likewise.
42218         (check_code_iterator): Likewise.
42219         (rtx_reader::read_rtx): Likewise.  Move one-time initialization
42220         logic to...
42221         (one_time_initialization): New function.
42222         (rtx_reader::read_until): New method.
42223         (read_flags): New function.
42224         (parse_reg_note_name): New function.
42225         (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
42226         Handle reuse_rtx ids.
42227         Wrap iterator lookup within #ifdef GENERATOR_FILE.
42228         Add parsing support for RTL dumps, mirroring the special-cases in
42229         print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
42230         values, and calling handle_any_trailing_information.
42231         (rtx_reader::read_rtx_operand): Convert return type from void
42232         to rtx, returning return_rtx.  Handle case 'e'.  Call
42233         finalize_string on XSTR and XTMPL fields.
42234         (rtx_reader::read_nested_rtx):  Handle dumps in which trailing
42235          "(nil)" values were omitted.  Call the postprocess vfunc on the
42236         return_rtx.
42237         (rtx_reader::rtx_reader): Add new "compact" param and pass to base
42238         class ctor.  Initialize m_in_call_function_usage.  Call
42239         one_time_initialization.
42240         * rtl-tests.c (selftest::test_uncond_jump): Call
42241         set_new_first_and_last_insn.
42242         * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
42243         * selftest-rtl.c: New file.
42244         * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42245         (selftest::get_insn_by_uid): New decl.
42246         * selftest-run-tests.c (selftest::run_tests): Call
42247         read_rtl_function_c_tests.
42248         * selftest.h  (selftest::read_rtl_function_c_tests): New decl.
42249         * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42250         dumps.
42252 2017-01-05  Uros Bizjak  <ubizjak@gmail.com>
42254         * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42255         operands in a special way.  Assert that pos+len <= mode precision.
42257 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42259         * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42260         3 argument Alias with unlimited for the negative form.
42261         (fno-vect-cost-model): Removed.
42263 2017-01-05  Martin Liska  <mliska@suse.cz>
42265         * hsa-gen.c (gen_hsa_divmod): New function.
42266         (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42268 2017-01-05  Martin Liska  <mliska@suse.cz>
42270         PR pch/78970
42271         * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42272         header.
42274 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42276         * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42277         small constant length operands.
42279 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42281         * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42282         between loop iterations.
42284 2017-01-05  Martin Liska  <mliska@suse.cz>
42286         PR sanitizer/78815
42287         * gimplify.c (gimplify_decl_expr): Compare to
42288         asan_poisoned_variables instread of checking flags.
42289         (gimplify_target_expr): Likewise.
42290         (gimplify_expr): Likewise.
42291         (gimplify_function_tree): Conditionally initialize
42292         asan_poisoned_variables.
42294 2017-01-04  Jeff Law  <law@redhat.com>
42296         PR tree-optimizatin/78812
42297         * rtl.h (contains_mem_rtx_p): Prototype.
42298         * ifcvt.c (containts_mem_rtx_p): Move from here to...
42299         * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42300         * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42301         and prune MEMs that are not at the toplevel of a SET_SRC rtx.  Look
42302         through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42304 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42306         * input.c (assert_char_at_range): Default-initialize actual_range.
42308 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42310         * df-scan.c (df_ref_create_structure): Make regno unsigned,
42311         to match the caller.
42313 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42315         * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42316         insns after final jump in test to emit dummy move.
42318 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42320         * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42321         * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42323 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42325         * multiple_target.c (create_dispatcher_calls): Init e_next.
42326         * tree-ssa-loop-split.c (split_loop): Init border.
42327         * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42328         scalar_type.
42330 2017-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
42332         PR target/71977
42333         PR target/70568
42334         PR target/78823
42335         * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42336         (altivec_register_operand): Do not return true if the operand
42337         contains a SUBREG mixing SImode and SFmode.
42338         (vsx_register_operand): Likewise.
42339         (vsx_reg_sfsubreg_ok): New predicate.
42340         (vfloat_operand): Do not return true if the operand contains a
42341         SUBREG mixing SImode and SFmode.
42342         (vint_operand): Likewise.
42343         (vlogical_operand): Likewise.
42344         (gpc_reg_operand): Likewise.
42345         (int_reg_operand): Likewise.
42346         * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42347         * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42348         determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42349         SImode and SFmode.
42350         (rs6000_emit_move_si_sf_subreg): New helper function.
42351         (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42352         fixup SUBREGs involving SImode and SFmode.
42353         * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42354         numbers for the new peephole2 optimization.
42355         (peephole2 for SFmode unions): New peephole2 to optimize cases in
42356         the GLIBC math library that do AND/IOR/XOR operations on single
42357         precision floating point.
42358         * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42359         target macros to say whether we need to avoid SUBREGs mixing
42360         SImode and SFmode.
42361         (TARGET_ALLOW_SF_SUBREG): Likewise.
42362         * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42363         (UNSPEC_SI_FROM_SF): Likewise.
42364         (iorxor): Change spacing.
42365         (and_ior_xor): New iterator for AND, IOR, and XOR.
42366         (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42367         (movdi_from_sf_zero_ext): Likewise.
42368         (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42369         instead of gpc_reg_operand.  Add SImode/SFmode SUBREG support.
42370         (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42371         (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42372         (fms<mode>4): Likewise.
42373         (fnma<mode>4): Likewise.
42374         (fnms<mode>4): Likewise.
42375         (nfma<mode>4): Likewise.
42376         (nfms<mode>4): Likewise.
42378 2017-01-04  Marek Polacek  <polacek@redhat.com>
42380         PR c++/64767
42381         * doc/invoke.texi: Document -Wpointer-compare.
42383 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42385         * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42386         RejectNegative.
42388         * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42389         descriptions for -gdwarf-5 and emit them as uleb128 instead of
42390         2-byte data.
42392 2017-01-04  Kelvin Nilsen  <kelvin@gcc.gnu.org>
42394         PR target/78056
42395         * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42396         documentation of the powerpc_popcntb_ok attribute.
42397         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42398         code to issue warning messages if a requested CPU configuration is
42399         not supported by the binary (assembler and loader) toolchain.
42400         (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42401         made to define a built-in function that has been disabled.
42402         (paired_init_builtins): Add assertion to prevent ICE if attempt is
42403         made to define a built-in function that has been disabled.
42404         (altivec_init_builtins): Add comment explaining why definition
42405         of the DST built-in functions is not preceded by an assertion
42406         check.  Add assertions to prevent ICE if attempts are made to
42407         define an altivec predicate or an abs* built-in function that has
42408         been disabled.
42409         (htm_init_builtins): Add comment explaining why definition of the
42410         htm built-in functions is not preceded by an assertion check.
42412 2017-01-04  Jeff Law  <law@redhat.com>
42414         PR tree-optimizatin/67955
42415         * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42416         Allow any SSA_VAR_P as the base objects.  Use integer_zerop.  Verify
42417         the points-to solution does not include pt_null.  Use DECL_PT_UID
42418         unconditionally.
42420 2017-01-04  Uros Bizjak  <ubizjak@gmail.com>
42422         * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42423         Use gen_int_mode instead of gen_lopwart for const_int operands.
42425 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42427         PR tree-optimization/71563
42428         * match.pd: Simplify X << Y into X if Y is known to be 0 or
42429         out of range value - has low bits known to be zero.
42431 2017-01-04  Alan Modra  <amodra@gmail.com>
42433         * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42434         * configure: Regenerate.
42435         * config.in: Regenerate.
42437 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42439         PR bootstrap/77569
42440         * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42441         a substring of the message, but strcmp with the whole message.  Ifdef
42442         ENABLE_NLS, translate the message first using dgettext.
42444 2017-01-03  Jeff Law  <law@redhat.com>
42446         PR tree-optimizatin/78856
42447         * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42448         (mark_threaded_blocks): Remove code to truncate thread paths that
42449         cross multiple loop headers.  Instead invalidate the cached loop
42450         iteration information and handle case of a thread path walking
42451         into an irreducible region.
42453 2017-01-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
42455         PR target/78900
42456         * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42457         assertions.  Add support for doing the signbit if the IEEE 128-bit
42458         floating point value is in a GPR.
42459         * config/rs6000/rs6000.md (Fsignbit): Delete.
42460         (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42461         Update the length attribute if the value is in a GPR.
42462         (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42463         the sign or zero extension instruction, since the value is always 0/1.
42464         (signbit<mode>2_dm2): Delete using <Fsignbit>.
42466         PR target/78953
42467         * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42468         extracting SImode to a GPR register so that we can generate a
42469         store, limit the vector to be in a traditional Altivec register
42470         for the vextuwrx instruction.
42472 2017-01-03  Ian Lance Taylor  <iant@google.com>
42474         * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42476 2017-01-03  Martin Sebor  <msebor@redhat.com>
42478         PR tree-optimization/78696
42479         * gimple-ssa-sprintf.c (format_floating): Correct handling of
42480         precision.  Use MPFR for %f for greater fidelity.  Correct handling
42481         of %g.
42482         (pass_sprintf_length::compute_format_length): Set width and precision
42483         specified by asrerisk to void_node for vararg functions.
42484         (try_substitute_return_value): Adjust dump output.
42486 2017-01-03  David Edelsohn  <dje.gcc@gmail.com>
42488         * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42490 2017-01-03  Eric Botcazou  <ebotcazou@adacore.com>
42492         * doc/invoke.texi (SPARC options): Document -mlra as the default.
42493         * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42494         -mlra/-mno-lra was passed to the compiler.
42496 2017-01-03  James Cowgill  <James.Cowgill@imgtec.com>
42498         PR rtl-optimization/65618
42499         * emit-rtl.c (try_split): Move initialization of "before" and
42500         "after" to just before the call to emit_insn_after_setloc.
42502 2017-01-03  Gerald Pfeifer  <gerald@pfeifer.com>
42504         * doc/md.texi (Standard Names): Remove reference to Java frontend.
42506 2017-01-03  Pierre-Marie de Rodat  <derodat@adacore.com>
42508         * dwarf2out.c (gen_enumeration_type_die): When
42509         -gno-strict-dwarf, add a DW_AT_encoding attribute.
42511 2017-01-03  Jakub Jelinek  <jakub@redhat.com>
42513         PR tree-optimization/78965
42514         * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42515         Change first argument from const call_info & to call_info &.  For %n
42516         set info.nowrite to false.
42518         PR middle-end/78901
42519         * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42520         possibly throwing calls.
42522         * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42523         and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42524         and fns handling, rather than in a separate case SSA_NAME.
42526 2017-01-02  Jeff Law  <law@redhat.com>
42528         * config/darwin-driver.c (darwin_driver_init): Const-correctness
42529         fixes for first_period and second_period variables.
42531 2017-01-02  Uros Bizjak  <ubizjak@gmail.com>
42533         PR target/78967
42534         * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42535         (*insvqi_1): New insn pattern.
42536         (*insvqi_1_mem_rex64): Ditto.
42537         (*insvqi_2): Ditto.
42538         (*insvqi_3): Rename from *insvqi.
42540         (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42542 2017-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
42544         * doc/cfg.texi (Edges): Remove reference to Java.
42545         (Maintaining the CFG): Ditto.
42547 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42549         PR middle-end/77674
42550         * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42551         transparent aliases.
42553 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42555         PR middle-end/77484
42556         * predict.def (PRED_CALL): Update hitrate.
42557         (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42558         * predict.c (tree_estimate_probability_bb): Split CALL predictor
42559         into direct/indirect/polymorphic variants.
42561 2017-01-01  Jakub Jelinek  <jakub@redhat.com>
42563         Update copyright years.
42565         * gcc.c (process_command): Update copyright notice dates.
42566         * gcov-dump.c (print_version): Ditto.
42567         * gcov.c (print_version): Ditto.
42568         * gcov-tool.c (print_version): Ditto.
42569         * gengtype.c (create_file): Ditto.
42570         * doc/cpp.texi: Bump @copying's copyright year.
42571         * doc/cppinternals.texi: Ditto.
42572         * doc/gcc.texi: Ditto.
42573         * doc/gccint.texi: Ditto.
42574         * doc/gcov.texi: Ditto.
42575         * doc/install.texi: Ditto.
42576         * doc/invoke.texi: Ditto.
42578 Copyright (C) 2017 Free Software Foundation, Inc.
42580 Copying and distribution of this file, with or without modification,
42581 are permitted in any medium without royalty provided the copyright
42582 notice and this notice are preserved.