poly_int: REGMODE_NATURAL_SIZE
[official-gcc.git] / gcc / ChangeLog
blob5cb1c23a070f0009a5318064d426bfa9e07285c7
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         * combine.c (can_change_dest_mode): Handle polynomial
6         REGMODE_NATURAL_SIZE.
7         * expmed.c (store_bit_field_1): Likewise.
8         * expr.c (store_constructor): Likewise.
9         * emit-rtl.c (validate_subreg): Operate on polynomial mode sizes
10         and polynomial REGMODE_NATURAL_SIZE.
11         (gen_lowpart_common): Likewise.
12         * reginfo.c (record_subregs_of_mode): Likewise.
13         * rtlanal.c (read_modify_subreg_p): Likewise.
15 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
16             Alan Hayward  <alan.hayward@arm.com>
17             David Sherwood  <david.sherwood@arm.com>
19         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
20         numbers of elements.
22 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
23             Alan Hayward  <alan.hayward@arm.com>
24             David Sherwood  <david.sherwood@arm.com>
26         * match.pd: Cope with polynomial numbers of vector 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         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
33         in a POINTER_PLUS_EXPR.
35 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
36             Alan Hayward  <alan.hayward@arm.com>
37             David Sherwood  <david.sherwood@arm.com>
39         * omp-simd-clone.c (simd_clone_subparts): New function.
40         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
41         (ipa_simd_modify_function_body): Likewise.
43 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
44             Alan Hayward  <alan.hayward@arm.com>
45             David Sherwood  <david.sherwood@arm.com>
47         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
48         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
49         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
50         (expand_vector_condition, vector_element): Likewise.
51         (subparts_gt): New function.
52         (get_compute_type): Use subparts_gt.
53         (count_type_subparts): Delete.
54         (expand_vector_operations_1): Use subparts_gt instead of
55         count_type_subparts.
57 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
58             Alan Hayward  <alan.hayward@arm.com>
59             David Sherwood  <david.sherwood@arm.com>
61         * tree-vect-data-refs.c (vect_no_alias_p): Replace with...
62         (vect_compile_time_alias): ...this new function.  Do the calculation
63         on poly_ints rather than trees.
64         (vect_prune_runtime_alias_test_list): Update call accordingly.
66 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
67             Alan Hayward  <alan.hayward@arm.com>
68             David Sherwood  <david.sherwood@arm.com>
70         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
71         numbers of units.
72         (vect_schedule_slp_instance): Likewise.
74 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
75             Alan Hayward  <alan.hayward@arm.com>
76             David Sherwood  <david.sherwood@arm.com>
78         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
79         constant and extern definitions for variable-length vectors.
80         (vect_get_constant_vectors): Note that the number of units
81         is known to be constant.
83 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
84             Alan Hayward  <alan.hayward@arm.com>
85             David Sherwood  <david.sherwood@arm.com>
87         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
88         of units as polynomial.  Choose between WIDE and NARROW based
89         on multiple_p.
91 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
92             Alan Hayward  <alan.hayward@arm.com>
93             David Sherwood  <david.sherwood@arm.com>
95         * tree-vect-stmts.c (simd_clone_subparts): New function.
96         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
98 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
99             Alan Hayward  <alan.hayward@arm.com>
100             David Sherwood  <david.sherwood@arm.com>
102         * tree-vect-stmts.c (vectorizable_call): Treat the number of
103         vectors as polynomial.  Use build_index_vector for
104         IFN_GOMP_SIMD_LANE.
106 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
107             Alan Hayward  <alan.hayward@arm.com>
108             David Sherwood  <david.sherwood@arm.com>
110         * tree-vect-stmts.c (get_load_store_type): Treat the number of
111         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
112         for variable-length vectors.
113         (vectorizable_mask_load_store): Treat the number of units as
114         polynomial, asserting that it is constant if the condition has
115         already been enforced.
116         (vectorizable_store, vectorizable_load): Likewise.
118 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
119             Alan Hayward  <alan.hayward@arm.com>
120             David Sherwood  <david.sherwood@arm.com>
122         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
123         of units as polynomial.  Punt if we can't tell at compile time
124         which vector contains the final result.
126 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
127             Alan Hayward  <alan.hayward@arm.com>
128             David Sherwood  <david.sherwood@arm.com>
130         * tree-vect-loop.c (vectorizable_induction): Treat the number
131         of units as polynomial.  Punt on SLP inductions.  Use an integer
132         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
133         cast of such a series for variable-length floating-point
134         reductions.
136 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
137             Alan Hayward  <alan.hayward@arm.com>
138             David Sherwood  <david.sherwood@arm.com>
140         * tree.h (build_index_vector): Declare.
141         * tree.c (build_index_vector): New function.
142         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
143         of units as polynomial, forcibly converting it to a constant if
144         vectorizable_reduction has already enforced the condition.
145         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
146         to create a {1,2,3,...} vector.
147         (vectorizable_reduction): Treat the number of units as polynomial.
148         Choose vectype_in based on the largest scalar element size rather
149         than the smallest number of units.  Enforce the restrictions
150         relied on above.
152 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
153             Alan Hayward  <alan.hayward@arm.com>
154             David Sherwood  <david.sherwood@arm.com>
156         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
157         number of units as polynomial.
159 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
160             Alan Hayward  <alan.hayward@arm.com>
161             David Sherwood  <david.sherwood@arm.com>
163         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
164         * target.def (autovectorize_vector_sizes): Return the vector sizes
165         by pointer, using vector_sizes rather than a bitmask.
166         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
167         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
168         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
169         Likewise.
170         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
171         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
172         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
173         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
174         * omp-general.c (omp_max_vf): Likewise.
175         * omp-low.c (omp_clause_aligned_alignment): Likewise.
176         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
177         * tree-vect-loop.c (vect_analyze_loop): Likewise.
178         * tree-vect-slp.c (vect_slp_bb): Likewise.
179         * doc/tm.texi: Regenerate.
180         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
181         to a poly_uint64.
182         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
183         the vector size as a poly_uint64 rather than an unsigned int.
184         (current_vector_size): Change from an unsigned int to a poly_uint64.
185         (get_vectype_for_scalar_type): Update accordingly.
186         * tree.h (build_truth_vector_type): Take the size and number of
187         units as a poly_uint64 rather than an unsigned int.
188         (build_vector_type): Add a temporary overload that takes
189         the number of units as a poly_uint64 rather than an unsigned int.
190         * tree.c (make_vector_type): Likewise.
191         (build_truth_vector_type): Take the number of units as a poly_uint64
192         rather than an unsigned int.
194 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
195             Alan Hayward  <alan.hayward@arm.com>
196             David Sherwood  <david.sherwood@arm.com>
198         * target.def (get_mask_mode): Take the number of units and length
199         as poly_uint64s rather than unsigned ints.
200         * targhooks.h (default_get_mask_mode): Update accordingly.
201         * targhooks.c (default_get_mask_mode): Likewise.
202         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
203         * doc/tm.texi: Regenerate.
205 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
206             Alan Hayward  <alan.hayward@arm.com>
207             David Sherwood  <david.sherwood@arm.com>
209         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
210         * omp-general.c (omp_max_vf): Likewise.
211         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
212         (expand_omp_simd): Handle polynomial safelen.
213         * omp-low.c (omplow_simd_context): Add a default constructor.
214         (omplow_simd_context::max_vf): Change from int to poly_uint64.
215         (lower_rec_simd_input_clauses): Update accordingly.
216         (lower_rec_input_clauses): Likewise.
218 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
219             Alan Hayward  <alan.hayward@arm.com>
220             David Sherwood  <david.sherwood@arm.com>
222         * tree-vectorizer.h (vect_nunits_for_cost): New function.
223         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
224         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
225         (vect_analyze_slp_cost): Likewise.
226         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
227         (vect_model_load_cost): Likewise.
229 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
230             Alan Hayward  <alan.hayward@arm.com>
231             David Sherwood  <david.sherwood@arm.com>
233         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
234         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
235         from an unsigned int * to a poly_uint64_pod *.
236         (calculate_unrolling_factor): New function.
237         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
239 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
240             Alan Hayward  <alan.hayward@arm.com>
241             David Sherwood  <david.sherwood@arm.com>
243         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
244         from an unsigned int to a poly_uint64.
245         (_loop_vec_info::slp_unrolling_factor): Likewise.
246         (_loop_vec_info::vectorization_factor): Change from an int
247         to a poly_uint64.
248         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
249         (vect_get_num_vectors): New function.
250         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
251         (vect_get_num_copies): Use vect_get_num_vectors.
252         (vect_analyze_data_ref_dependences): Change max_vf from an int *
253         to an unsigned int *.
254         (vect_analyze_data_refs): Change min_vf from an int * to a
255         poly_uint64 *.
256         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
257         than an unsigned HOST_WIDE_INT.
258         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
259         (vect_analyze_data_ref_dependence): Change max_vf from an int *
260         to an unsigned int *.
261         (vect_analyze_data_ref_dependences): Likewise.
262         (vect_compute_data_ref_alignment): Handle polynomial vf.
263         (vect_enhance_data_refs_alignment): Likewise.
264         (vect_prune_runtime_alias_test_list): Likewise.
265         (vect_shift_permute_load_chain): Likewise.
266         (vect_supportable_dr_alignment): Likewise.
267         (dependence_distance_ge_vf): Take the vectorization factor as a
268         poly_uint64 rather than an unsigned HOST_WIDE_INT.
269         (vect_analyze_data_refs): Change min_vf from an int * to a
270         poly_uint64 *.
271         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
272         vfm1 as a poly_uint64 rather than an int.  Make the same change
273         for the returned bound_scalar.
274         (vect_gen_vector_loop_niters): Handle polynomial vf.
275         (vect_do_peeling): Likewise.  Update call to
276         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
277         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
278         be constant.
279         * tree-vect-loop.c (vect_determine_vectorization_factor)
280         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
281         (vect_get_known_peeling_cost): Likewise.
282         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
283         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
284         (vect_transform_loop): Likewise.  Use the lowest possible VF when
285         updating the upper bounds of the loop.
286         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
287         rather than an int.
288         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
289         polynomial unroll factors.
290         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
291         (vect_make_slp_decision): Likewise.
292         (vect_supported_load_permutation_p): Likewise, and polynomial
293         vf too.
294         (vect_analyze_slp_cost): Handle polynomial vf.
295         (vect_slp_analyze_node_operations): Likewise.
296         (vect_slp_analyze_bb_1): Likewise.
297         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
298         than an unsigned HOST_WIDE_INT.
299         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
300         (vectorizable_load): Handle polynomial vf.
301         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
302         a poly_uint64.
303         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
305 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
306             Alan Hayward  <alan.hayward@arm.com>
307             David Sherwood  <david.sherwood@arm.com>
309         * match.pd: Handle bit operations involving three constants
310         and try to fold one pair.
312 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
314         * tree-vect-loop-manip.c: Include gimple-fold.h.
315         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
316         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
317         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
318         Add a path that uses a step of VF instead of 1, but disable it
319         for now.
320         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
321         and niters_no_overflow parameters.  Update calls to
322         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
323         Create a new SSA name if the latter choses to use a ste other
324         than zero, and return it via niters_vector_mult_vf_var.
325         * tree-vect-loop.c (vect_transform_loop): Update calls to
326         vect_do_peeling, vect_gen_vector_loop_niters and
327         slpeel_make_loop_iterate_ntimes.
328         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
329         (vect_gen_vector_loop_niters): Update declarations after above changes.
331 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
333         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
334         128-bit round to integer instructions.
335         (ceil<mode>2): Likewise.
336         (btrunc<mode>2): Likewise.
337         (round<mode>2): Likewise.
339 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
341         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
342         unaligned VSX load/store on P8/P9.
343         (expand_block_clear): Allow the use of unaligned VSX
344         load/store on P8/P9.
346 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
348         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
349         New function.
350         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
351         swap associated with both a load and a store.
353 2018-01-02  Andrew Waterman  <andrew@sifive.com>
355         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
356         * config/riscv/riscv.md (clear_cache): Use it.
358 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
360         * web.c: Remove out-of-date comment.
362 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
364         * expr.c (fixup_args_size_notes): Check that any existing
365         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
366         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
367         (emit_single_push_insn): ...here.
369 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
371         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
372         (const_vector_encoded_nelts): New function.
373         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
374         (const_vector_int_elt, const_vector_elt): Declare.
375         * emit-rtl.c (const_vector_int_elt_1): New function.
376         (const_vector_elt): Likewise.
377         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
378         of CONST_VECTOR_ELT.
380 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
382         * expr.c: Include rtx-vector-builder.h.
383         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
384         directly on the tree encoding.
385         (const_vector_from_tree): Likewise.
386         * optabs.c: Include rtx-vector-builder.h.
387         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
388         sequence of "u" values.
389         * vec-perm-indices.c: Include rtx-vector-builder.h.
390         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
391         directly on the vec_perm_indices encoding.
393 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
395         * doc/rtl.texi (const_vector): Describe new encoding scheme.
396         * Makefile.in (OBJS): Add rtx-vector-builder.o.
397         * rtx-vector-builder.h: New file.
398         * rtx-vector-builder.c: Likewise.
399         * rtl.h (rtx_def::u2): Add a const_vector field.
400         (CONST_VECTOR_NPATTERNS): New macro.
401         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
402         (CONST_VECTOR_DUPLICATE_P): Likewise.
403         (CONST_VECTOR_STEPPED_P): Likewise.
404         (CONST_VECTOR_ENCODED_ELT): Likewise.
405         (const_vec_duplicate_p): Check for a duplicated vector encoding.
406         (unwrap_const_vec_duplicate): Likewise.
407         (const_vec_series_p): Check for a non-duplicated vector encoding.
408         Say that the function only returns true for integer vectors.
409         * emit-rtl.c: Include rtx-vector-builder.h.
410         (gen_const_vec_duplicate_1): Delete.
411         (gen_const_vector): Call gen_const_vec_duplicate instead of
412         gen_const_vec_duplicate_1.
413         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
414         (gen_const_vec_duplicate): Use rtx_vector_builder.
415         (gen_const_vec_series): Likewise.
416         (gen_rtx_CONST_VECTOR): Likewise.
417         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
418         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
419         Build a new vector rather than modifying a CONST_VECTOR in-place.
420         (handle_special_swappables): Update call accordingly.
421         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
422         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
423         Build a new vector rather than modifying a CONST_VECTOR in-place.
424         (handle_special_swappables): Update call accordingly.
426 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
428         * simplify-rtx.c (simplify_const_binary_operation): Use
429         CONST_VECTOR_ELT instead of XVECEXP.
431 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
433         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
434         the selector elements to be different from the data elements
435         if the selector is a VECTOR_CST.
436         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
437         ssizetype for the selector.
439 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
441         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
442         before testing each element individually.
443         * tree-vect-generic.c (lower_vec_perm): Likewise.
445 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
447         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
448         * selftest-run-tests.c (selftest::run_tests): Call it.
449         * vector-builder.h (vector_builder::operator ==): New function.
450         (vector_builder::operator !=): Likewise.
451         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
452         (vec_perm_indices::all_from_input_p): New function.
453         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
454         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
455         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
456         instead of reading the VECTOR_CST directly.  Detect whether both
457         vector inputs are the same before constructing the vec_perm_indices,
458         and update the number of inputs argument accordingly.  Use the
459         utility functions added above.  Only construct sel2 if we need to.
461 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
463         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
464         the broadcast of the low byte.
465         (expand_mult_highpart): Use an explicit encoding for the permutes.
466         * optabs-query.c (can_mult_highpart_p): Likewise.
467         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
468         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
469         (vectorizable_bswap): Likewise.
470         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
471         explicit encoding for the power-of-2 permutes.
472         (vect_permute_store_chain): Likewise.
473         (vect_grouped_load_supported): Likewise.
474         (vect_permute_load_chain): Likewise.
476 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
478         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
479         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
480         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
481         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
482         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
483         (vect_gen_perm_mask_any): Likewise.
485 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
487         * int-vector-builder.h: New file.
488         * vec-perm-indices.h: Include int-vector-builder.h.
489         (vec_perm_indices): Redefine as an int_vector_builder.
490         (auto_vec_perm_indices): Delete.
491         (vec_perm_builder): Redefine as a stand-alone class.
492         (vec_perm_indices::vec_perm_indices): New function.
493         (vec_perm_indices::clamp): Likewise.
494         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
495         (vec_perm_indices::new_vector): New function.
496         (vec_perm_indices::new_expanded_vector): Update for new
497         vec_perm_indices class.
498         (vec_perm_indices::rotate_inputs): New function.
499         (vec_perm_indices::all_in_range_p): Operate directly on the
500         encoded form, without computing elided elements.
501         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
502         encoding.  Update for new vec_perm_indices class.
503         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
504         the given vec_perm_builder.
505         (expand_vec_perm_var): Update vec_perm_builder constructor.
506         (expand_mult_highpart): Use vec_perm_builder instead of
507         auto_vec_perm_indices.
508         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
509         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
510         or double series encoding as appropriate.
511         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
512         vec_perm_indices instead of auto_vec_perm_indices.
513         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
514         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
515         (vect_permute_store_chain): Likewise.
516         (vect_grouped_load_supported): Likewise.
517         (vect_permute_load_chain): Likewise.
518         (vect_shift_permute_load_chain): Likewise.
519         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
520         (vect_transform_slp_perm_load): Likewise.
521         (vect_schedule_slp_instance): Likewise.
522         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
523         (vectorizable_mask_load_store): Likewise.
524         (vectorizable_bswap): Likewise.
525         (vectorizable_store): Likewise.
526         (vectorizable_load): Likewise.
527         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
528         vec_perm_indices instead of auto_vec_perm_indices.  Use
529         tree_to_vec_perm_builder to read the vector from a tree.
530         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
531         vec_perm_builder instead of a vec_perm_indices.
532         (have_whole_vector_shift): Use vec_perm_builder and
533         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
534         truncation to calc_vec_perm_mask_for_shift.
535         (vect_create_epilog_for_reduction): Likewise.
536         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
537         from auto_vec_perm_indices to vec_perm_indices.
538         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
539         instead of changing individual elements.
540         (aarch64_vectorize_vec_perm_const): Use new_vector to install
541         the vector in d.perm.
542         * config/arm/arm.c (expand_vec_perm_d::perm): Change
543         from auto_vec_perm_indices to vec_perm_indices.
544         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
545         instead of changing individual elements.
546         (arm_vectorize_vec_perm_const): Use new_vector to install
547         the vector in d.perm.
548         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
549         Update vec_perm_builder constructor.
550         (rs6000_expand_interleave): Likewise.
551         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
552         (rs6000_expand_interleave): Likewise.
554 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
556         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
557         to qimode could truncate the indices.
558         * optabs.c (expand_vec_perm_var): Likewise.
560 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
562         * Makefile.in (OBJS): Add vec-perm-indices.o.
563         * vec-perm-indices.h: New file.
564         * vec-perm-indices.c: Likewise.
565         * target.h (vec_perm_indices): Replace with a forward class
566         declaration.
567         (auto_vec_perm_indices): Move to vec-perm-indices.h.
568         * optabs.h: Include vec-perm-indices.h.
569         (expand_vec_perm): Delete.
570         (selector_fits_mode_p, expand_vec_perm_var): Declare.
571         (expand_vec_perm_const): Declare.
572         * target.def (vec_perm_const_ok): Replace with...
573         (vec_perm_const): ...this new hook.
574         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
575         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
576         * doc/tm.texi: Regenerate.
577         * optabs.def (vec_perm_const): Delete.
578         * doc/md.texi (vec_perm_const): Likewise.
579         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
580         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
581         expand_vec_perm for constant permutation vectors.  Assert that
582         the mode of variable permutation vectors is the integer equivalent
583         of the mode that is being permuted.
584         * optabs-query.h (selector_fits_mode_p): Declare.
585         * optabs-query.c: Include vec-perm-indices.h.
586         (selector_fits_mode_p): New function.
587         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
588         is defined, instead of checking whether the vec_perm_const_optab
589         exists.  Use targetm.vectorize.vec_perm_const instead of
590         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
591         fit in the vector mode before using a variable permute.
592         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
593         vec_perm_indices instead of an rtx.
594         (expand_vec_perm): Replace with...
595         (expand_vec_perm_const): ...this new function.  Take the selector
596         as a vec_perm_indices rather than an rtx.  Also take the mode of
597         the selector.  Update call to shift_amt_for_vec_perm_mask.
598         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
599         Use vec_perm_indices::new_expanded_vector to expand the original
600         selector into bytes.  Check whether the indices fit in the vector
601         mode before using a variable permute.
602         (expand_vec_perm_var): Make global.
603         (expand_mult_highpart): Use expand_vec_perm_const.
604         * fold-const.c: Includes vec-perm-indices.h.
605         * tree-ssa-forwprop.c: Likewise.
606         * tree-vect-data-refs.c: Likewise.
607         * tree-vect-generic.c: Likewise.
608         * tree-vect-loop.c: Likewise.
609         * tree-vect-slp.c: Likewise.
610         * tree-vect-stmts.c: Likewise.
611         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
612         Delete.
613         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
614         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
615         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
616         (aarch64_vectorize_vec_perm_const): ...this new function.
617         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
618         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
619         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
620         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
621         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
622         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
623         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
624         into...
625         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
626         check for NEON modes.
627         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
628         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
629         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
630         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
631         into...
632         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
633         the old VEC_PERM_CONST conditions.
634         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
635         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
636         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
637         (ia64_vectorize_vec_perm_const_ok): Merge into...
638         (ia64_vectorize_vec_perm_const): ...this new function.
639         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
640         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
641         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
642         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
643         * config/mips/mips.c (mips_expand_vec_perm_const)
644         (mips_vectorize_vec_perm_const_ok): Merge into...
645         (mips_vectorize_vec_perm_const): ...this new function.
646         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
647         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
648         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
649         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
650         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
651         (rs6000_expand_vec_perm_const): Delete.
652         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
653         Delete.
654         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
655         (altivec_expand_vec_perm_const_le): Take each operand individually.
656         Operate on constant selectors rather than rtxes.
657         (altivec_expand_vec_perm_const): Likewise.  Update call to
658         altivec_expand_vec_perm_const_le.
659         (rs6000_expand_vec_perm_const): Delete.
660         (rs6000_vectorize_vec_perm_const_ok): Delete.
661         (rs6000_vectorize_vec_perm_const): New function.
662         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
663         an element count and rtx array.
664         (rs6000_expand_extract_even): Update call accordingly.
665         (rs6000_expand_interleave): Likewise.
666         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
667         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
668         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
669         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
670         (rs6000_expand_vec_perm_const): Delete.
671         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
672         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
673         (altivec_expand_vec_perm_const_le): Take each operand individually.
674         Operate on constant selectors rather than rtxes.
675         (altivec_expand_vec_perm_const): Likewise.  Update call to
676         altivec_expand_vec_perm_const_le.
677         (rs6000_expand_vec_perm_const): Delete.
678         (rs6000_vectorize_vec_perm_const_ok): Delete.
679         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
680         reference to the SPE evmerge intructions.
681         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
682         an element count and rtx array.
683         (rs6000_expand_extract_even): Update call accordingly.
684         (rs6000_expand_interleave): Likewise.
685         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
686         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
687         new function.
688         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
690 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
692         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
693         vector mode and that that mode matches the mode of the data
694         being permuted.
695         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
696         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
697         directly using expand_vec_perm_1 when forcing selectors into
698         registers.
699         (expand_vec_perm_var): New function, split out from expand_vec_perm.
701 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
703         * optabs-query.h (can_vec_perm_p): Delete.
704         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
705         * optabs-query.c (can_vec_perm_p): Split into...
706         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
707         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
708         particular selector is valid.
709         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
710         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
711         (vect_grouped_load_supported): Likewise.
712         (vect_shift_permute_load_chain): Likewise.
713         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
714         (vect_transform_slp_perm_load): Likewise.
715         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
716         (vectorizable_bswap): Likewise.
717         (vect_gen_perm_mask_checked): Likewise.
718         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
719         implementations of variable permutation vectors into account
720         when deciding which selector to use.
721         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
722         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
723         with a false third argument.
724         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
725         to test whether the constant selector is valid and can_vec_perm_var_p
726         to test whether a variable selector is valid.
728 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
730         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
731         * optabs-query.c (can_vec_perm_p): Likewise.
732         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
733         instead of vec_perm_indices.
734         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
735         (vect_gen_perm_mask_checked): Likewise,
736         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
737         (vect_gen_perm_mask_checked): Likewise,
739 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
741         * optabs-query.h (qimode_for_vec_perm): Declare.
742         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
743         (qimode_for_vec_perm): ...this new function.
744         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
746 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
748         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
749         does not have a conditional at the top.
751 2018-01-02  Richard Biener  <rguenther@suse.de>
753         * ipa-inline.c (big_speedup_p): Fix expression.
755 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
757         PR target/81616
758         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
759         for generic 4->6.
761 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
763         PR target/81616
764         Generic tuning.
765         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
766         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
767         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
768         cond_taken_branch_cost 3->4.
770 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
772         PR tree-optimization/83581
773         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
774         TODO_cleanup_cfg if any changes have been made.
776         PR middle-end/83608
777         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
778         convert_modes if target mode has the right side, but different mode
779         class.
781         PR middle-end/83609
782         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
783         last argument when extracting from CONCAT.  If either from_real or
784         from_imag is NULL, use expansion through memory.  If result is not
785         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
786         the parts directly to inner mode, if even that fails, use expansion
787         through memory.
789         PR middle-end/83623
790         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
791         check for bswap in mode rather than HImode and use that in expand_unop
792         too.
794 2017-12-31  Jakub Jelinek  <jakub@redhat.com>
796         PR target/83536
797         * config/i386/i386.c (ix86_attribute_table): Remove excess
798         initializer for "shared" attribute.
800 2017-12-30  Tom de Vries  <tom@codesourcery.com>
802         PR libgomp/83046
803         * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
804         DECL_PRESERVE_P.
805         * lto-streamer-out.c (prune_offload_funcs): New function.  Remove
806         offload_funcs entries that no longer have a corresponding cgraph_node.
807         Mark the remaining ones as DECL_PRESERVE_P.
808         (output_lto): Call prune_offload_funcs.
810 2017-12-30  Jakub Jelinek  <jakub@redhat.com>
812         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
813         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
814         vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
815         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
816         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
817         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
818         vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
819         vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
820         vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
821         vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
822         vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
823         vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
824         vaesenc_<mode>, vpclmulqdq_<mode>,
825         avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
827 2017-12-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
829         * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
830         _Float<N> and _Float<N>X variants for rounding built-in
831         functions.
832         (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
833         (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
834         (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
835         (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
836         (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
837         * builtins.c (mathfn_built_in_2): Likewise.
838         * internal-fn.def (CEIL): Likewise.
839         (FLOOR): Likewise.
840         (NEARBYINT): Likewise.
841         (RINT): Likewise.
842         (ROUND): Likewise.
843         (TRUNC): Likewise.
844         * convert.c (convert_to_integer_1): Likewise.
845         * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
846         (integer_valued_real_call_p): Likewise.
847         * fold-const-call.c (fold_const_call_ss): Likewise.
848         * gencfn-macros.c (print_case_cfn): Change CFN and operator
849         printers to take a const char * suffix instead of a bool.
850         (print_define_operator_list): Likewise.
851         (fltall_suffixes): New list of suffixes, that include the
852         traditional suffixes as well as all of the _Float<N> and
853         _Float<N>X suffixes.
854         (main): For _Float<N> and _Float<N>X functions, emit both
855         <name>_FN and <name>_ALL variants.  The <macro>_FN variant only
856         has the _Float<N> and _Float<N>X case names or operators.  The
857         <name>_ALL variant has both the traditional and the
858         _Float<N>/_Float<N>X case names or operators.
859         * match.pd (COPYSIGN optimizations): Provide optimizations for
860         _Float<N> and _Float<N>X types where possible.
861         (MIN/MAX optimizations): Likewise.
862         (sqrt optimizations): Likewise.
863         (rounding optimizations): Likewise.
865 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
867         * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
868         gen_int_mode rather than GEN_INT.
869         * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
871 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
873         * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
874         instead of CONSTANT_P.
875         (gen_vec_series): Likewise.
876         * simplify-rtx.c (simplify_binary_operation_1): Likewise.
878 2017-12-28  Andreas Schwab  <schwab@linux-m68k.org>
880         * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
881         CC_STATUS_INIT.
883 2017-12-27  Kugan Vivekanandarajah  <kuganv@linaro.org>
885         * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
886         (aarch64_ld1x2<VDC:mode>): Likewise.
887         (aarch64_simd_ld1<mode>_x2): Likewise.
888         (aarch64_simd_ld1<mode>_x2): Likewise.
889         * config/aarch64/arm_neon.h (vld1_u8_x2): New.
890         (vld1_s8_x2): Likewise.
891         (vld1_u16_x2): Likewise.
892         (vld1_s16_x2): Likewise.
893         (vld1_u32_x2): Likewise.
894         (vld1_s32_x2): Likewise.
895         (vld1_u64_x2): Likewise.
896         (vld1_s64_x2): Likewise.
897         (vld1_f16_x2): Likewise.
898         (vld1_f32_x2): Likewise.
899         (vld1_f64_x2): Likewise.
900         (vld1_p8_x2): Likewise.
901         (vld1_p16_x2): Likewise.
902         (vld1_p64_x2): Likewise.
903         (vld1q_u8_x2): Likewise.
904         (vld1q_s8_x2): Likewise.
905         (vld1q_u16_x2): Likewise.
906         (vld1q_s16_x2): Likewise.
907         (vld1q_u32_x2): Likewise.
908         (vld1q_s32_x2): Likewise.
909         (vld1q_u64_x2): Likewise.
910         (vld1q_s64_x2): Likewise.
911         (vld1q_f16_x2): Likewise.
912         (vld1q_f32_x2): Likewise.
913         (vld1q_f64_x2): Likewise.
914         (vld1q_p8_x2): Likewise.
915         (vld1q_p16_x2): Likewise.
916         (vld1q_p64_x2): Likewise.
918 2017-12-27  Martin Liska  <mliska@suse.cz>
920         PR tree-optimization/83552
921         * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
922         of get_string_lenth to a SSA_NAME if not a GIMPLE value.
924 2017-12-27  Tom de Vries  <tom@codesourcery.com>
926         * config/nvptx/nvptx.c (nvptx_option_override): Disable
927         -gstatement-frontiers.
929 2017-12-26  Alexander Monakov  <amonakov@ispras.ru>
931         PR rtl-optimization/83513
932         * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
933         before priority comparison.
935 2017-12-25  Jakub Jelinek  <jakub@redhat.com>
937         PR target/83488
938         * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
939         ix86_isa_flags2 to ix86_isa_flags.
940         * config/i386/i386-c.c (ix86_target_macros_internal): Test
941         OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
942         isa_flags rather than isa_flags2.
943         * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
944         and -mavx512bitalg from isa2_opts to isa_opts.
945         (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
946         in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
947         and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
948         Formatting fixes.
949         (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
950         OPTION_MASK_ISA_AVX512F ored with another option similarly to
951         OPTION_MASK_ISA_AVX512VL.  Even for OPTION_MASK_ISA_AVX512VL don't
952         clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
953         (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
954         OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
955         OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
956         * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
957         builtins from bdesc_args2 to bdesc_args section.
958         (__builtin_ia32_compressstoreuqi512_mask,
959         __builtin_ia32_compressstoreuhi512_mask,
960         __builtin_ia32_compressstoreuqi256_mask,
961         __builtin_ia32_expandloadqi512_mask,
962         __builtin_ia32_expandloadqi512_maskz,
963         __builtin_ia32_expandloadhi512_mask,
964         __builtin_ia32_expandloadhi512_maskz,
965         __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
966         __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
967         __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
968         __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
969         __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
970         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
971         __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
972         __builtin_ia32_vpshldv_v32hi_maskz,
973         __builtin_ia32_vpopcountb_v64qi_mask,
974         __builtin_ia32_vpopcountw_v32hi_mask,
975         __builtin_ia32_vpshufbitqmb512_mask,
976         __builtin_ia32_vpshufbitqmb256_mask): Add
977         " | OPTION_MASK_ISA_AVX512BW".
978         (__builtin_ia32_expandloadqi256_mask,
979         __builtin_ia32_expandloadqi256_maskz,
980         __builtin_ia32_vpopcountb_v32qi_mask): Add
981         " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
982         (__builtin_ia32_expandloadhi256_mask,
983         __builtin_ia32_expandloadhi256_maskz,
984         __builtin_ia32_expandloadqi128_mask,
985         __builtin_ia32_expandloadqi128_maskz,
986         __builtin_ia32_expandloadhi128_mask,
987         __builtin_ia32_expandloadhi128_maskz,
988         __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
989         __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
990         __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
991         __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
992         __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
993         __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
994         __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
995         __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
996         __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
997         __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
998         __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
999         __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
1000         __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
1001         __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
1002         __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
1003         __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
1004         __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
1005         __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
1006         __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
1007         __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
1008         __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
1009         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
1010         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
1011         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
1012         __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
1013         __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
1014         __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
1015         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
1016         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
1017         __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
1018         __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
1019         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
1020         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1021         __builtin_ia32_vpopcountw_v8hi_mask): Add
1022         " | OPTION_MASK_ISA_AVX512VL".
1023         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
1024         _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1025         _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
1026         _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1027         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
1028         _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
1029         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
1030         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
1031         _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
1032         _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
1033         _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
1034         * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
1035         _mm_mask_bitshuffle_epi64_mask): Likewise.
1036         * common/config/i386/i386-common.c
1037         (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
1038         OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
1039         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1040         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
1041         OPTION_MASK_ISA_AVX512BITALG_UNSET.
1042         (OPTION_MASK_ISA2_AVX512F_UNSET,
1043         OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
1044         (ix86_handle_option): For -mno-general-regs-only, clear from
1045         ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
1046         just OPTION_MASK_ISA_MPX.  For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
1047         -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
1048         ix86_isa_flags2.  For -mno-avx512f likewise, instead of masking
1049         individually listed ISAs.  For -m{,no-}avx512{vpopcntdq,bitalg} adjust
1050         for moving from ix86_isa_flags2 to ix86_isa_flags.
1052 2017-12-23  Jakub Jelinek  <jakub@redhat.com>
1054         PR c++/83553
1055         * fold-const.c (struct contains_label_data): New type.
1056         (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
1057         inside of a SWITCH_BODY seen during the walk.
1058         (contains_label_p): Use walk_tree instead of
1059         walk_tree_without_duplicates, prepare data for contains_label_1 and
1060         provide own pset.
1062 2017-12-22  Jakub Jelinek  <jakub@redhat.com>
1064         PR debug/83547
1065         * tree-iterator.c (alloc_stmt_list): Start with cleared
1066         TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
1067         or old one reused.
1069         PR target/83488
1070         * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
1071         enable avx512f explicitly in #pragma GCC target.
1072         * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
1073         __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
1074         __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
1075         __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
1076         __builtin_ia32_vpdpbusds_v8si_mask,
1077         __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
1078         __builtin_ia32_vpdpbusds_v4si_mask,
1079         __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
1080         __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
1081         __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
1082         __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
1083         __builtin_ia32_vpdpwssds_v8si_mask,
1084         __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
1085         __builtin_ia32_vpdpwssds_v4si_mask,
1086         __builtin_ia32_vpdpwssds_v4si_maskz): Use
1087         OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
1088         just OPTION_MASK_ISA_AVX512VNNI.
1090 2017-12-22  Martin Jambor  <mjambor@suse.cz>
1092         PR lto/82027
1093         * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
1094         clones.
1096 2017-12-22  Julia Koval  <julia.koval@intel.com>
1097             Sebastian Peryt  <sebastian.peryt@intel.com>
1099         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
1100         OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
1101         (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
1102         * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
1103         * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
1104         _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
1105         _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
1106         _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
1107         _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
1108         _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
1109         _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
1110         _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
1111         _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
1112         * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
1113         _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
1114         _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
1115         _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
1116         _mm256_maskz_popcnt_epi64): New intrinsics.
1117         * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
1118         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
1119         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
1120         V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
1121         V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
1122         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
1123         __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
1124         __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
1125         __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
1126         __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
1127         __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
1128         __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
1129         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
1130         __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
1131         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1132         __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
1133         __builtin_ia32_vpshufbitqmb256_mask,
1134         __builtin_ia32_vpshufbitqmb512_mask): New builtins.
1135         * config/i386/i386-c.c (__AVX512BITALG__): New.
1136         * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
1137         (ix86_valid_target_attribute_inner_p): Ditto.
1138         (ix86_expand_args_builtin): Handle new types.
1139         * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
1140         * config/i386/i386.opt: Add -mavx512bitalg.
1141         * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
1142         avx512bitalgintrin.h.
1143         * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
1144         (vpopcount<mode><mask_name>): Add more types.
1145         (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
1146         * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
1148 2017-12-22  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
1150         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
1151         Or in OPTION_MASK_ISA_AVX512F_SET.
1152         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1153         OPTION_MASK_ISA_AVX512VNNI_UNSET.
1154         (ix86_handle_option): Adjust for
1155         OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
1156         * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
1157         section to ARGS.
1158         * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
1159         isa_flag instead of isa_flag2.
1160         * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
1161         isa_opts2 to isa_opts.
1162         * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
1163         to ix86_isa_flags.
1165 2017-12-22  Mike Stump  <mikestump@comcast.net>
1166             Eric Botcazou  <ebotcazou@adacore.com>
1168         * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
1170 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1172         PR middle-end/83487
1173         * config/i386/i386.c (ix86_function_arg_boundary): Return
1174         PARM_BOUNDARY for TYPE_EMPTY_P types.
1176         PR c/83448
1177         * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
1178         if navail is >= dir.len.
1180 2017-12-21  Steve Ellcey  <sellcey@cavium.com>
1182         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
1183         triplet for ilp32.
1185 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1187         PR rtl-optimization/80747
1188         PR rtl-optimization/83512
1189         * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
1190         succ edge from ENTRY, copy partition from e->dest to the newly
1191         created bb.
1192         * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
1193         ENTRY, use BB_PARTITION of its successor block as current_partition.
1194         Don't copy partition when splitting succ edge from ENTRY.
1196         PR tree-optimization/83523
1197         * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
1198         for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
1199         (convert_mult_to_fma): Likewise.
1201         PR tree-optimization/83521
1202         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
1203         gimple_build_assign without code on result of
1204         fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
1205         a VIEW_CONVERT_EXPR.
1207 2017-12-21  Andrew Pinski  <apinski@cavium.com>
1208             Steve Ellcey  <sellcey@cavium.com>
1210         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
1211         multi-arch for ilp32.
1213 2017-12-21  Uros Bizjak  <ubizjak@gmail.com>
1215         PR target/83467
1216         * config/i386/i386.md (*ashl<mode>3_mask): Add operand
1217         constraints to operand 2.
1218         (*ashl<mode>3_mask_1): Ditto.
1219         (*<shift_insn><mode>3_mask): Ditto.
1220         (*<shift_insn><mode>3_mask_1): Ditto.
1221         (*<rotate_insn><mode>3_mask): Ditto.
1222         (*<rotate_insn><mode>3_mask_1): Ditto.
1224 2017-12-21  Alexandre Oliva  <aoliva@redhat.com>
1226         * reorg.c (make_return_insns): Reemit each insn with its own location.
1228 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1230         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
1231         generation for cases where splatting a value is not useful.
1232         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
1233         across a vec_duplicate and a paradoxical subreg forming a vector
1234         mode to a vec_concat.
1236 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1238         * combine.c (simplify_set): Do not transform subregs to zero_extends
1239         if the destination is not a scalar int mode.
1241 2017-12-21  Eric Botcazou  <ebotcazou@adacore.com>
1243         PR c++/82872
1244         * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1245         shared zero if the input has overflowed.
1247 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1249         * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1250         support for Cortex-A55 and Cortex-A75.
1252 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1254         * common/config/arm/arm-common.c (compare_opt_names): Add function
1255         comment.  Use strcmp instead of manual loop.
1257 2017-12-21  Martin Liska  <mliska@suse.cz>
1259         PR gcov-profile/83509
1260         * gcov-dump.c (dump_gcov_file): Do not read info about
1261         support_unexecuted_blocks for gcda files.
1263 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1265         PR rtl-optimization/82973
1266         * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1267         (valid_for_const_vector_p): ... this.
1268         * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1269         (valid_for_const_vector_p): ... this.  Adjust function comment.
1270         (gen_vec_duplicate): Adjust caller.
1271         * optabs.c (expand_vector_broadcast): Likewise.
1272         * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1273         CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1274         constant.
1275         (simplify_const_binary_operation): Likewise.  Use CONST_FIXED_P macro
1276         instead of GET_CODE == CONST_FIXED.
1277         (simplify_subreg): Use CONST_FIXED_P macro instead of
1278         GET_CODE == CONST_FIXED.
1280         PR target/83488
1281         * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1282         -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1283         -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1284         (ix86_option_override_internal): Adjust for
1285         OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1286         and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1287         (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1288         (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1289         Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1290         and OPTION_MASK_ISA_CLZERO builtins.  Use def_builtin instead of
1291         def_builtin2 for CET builtins.
1292         (ix86_expand_builtin): Remove bdesc_special_args2 handling.  Fix
1293         up formatting in IX86_BUILTIN_RDPID code.
1294         * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1295         section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1296         * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1297         ix86_isa_flags2 to ix86_isa_flags.
1298         (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1299         ix86_isa_flags2.
1300         * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1301         OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1302         Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1303         of isa_flag2.
1304         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1305         Or in OPTION_MASK_ISA_AVX512F_SET.
1306         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1307         OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1308         (ix86_handle_option): Adjust for
1309         OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1310         and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1312 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1313             Alan Hayward  <alan.hayward@arm.com>
1314             David Sherwood  <david.sherwood@arm.com>
1316         * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1317         factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1318         * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1319         Track polynomial offsets.
1321 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1322             Alan Hayward  <alan.hayward@arm.com>
1323             David Sherwood  <david.sherwood@arm.com>
1325         * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
1326         Treat drb->init as a poly_int.  Fail if its misalignment wrt
1327         vector_alignment isn't known.
1329 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1330             Alan Hayward  <alan.hayward@arm.com>
1331             David Sherwood  <david.sherwood@arm.com>
1333         * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1334         field.
1335         (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1336         (vect_loop_versioning): Take the loop versioning threshold as a
1337         separate parameter.
1338         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1339         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1340         versioning_threshold.
1341         (vect_analyze_loop_2): Compute the loop versioning threshold
1342         whenever loop versioning is needed, and store it in the new
1343         field rather than combining it with the cost model threshold.
1344         (vect_transform_loop): Update call to vect_loop_versioning.
1345         Try to combine the loop versioning and cost thresholds here.
1347 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1348             Alan Hayward  <alan.hayward@arm.com>
1349             David Sherwood  <david.sherwood@arm.com>
1351         * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1352         poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1353         * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1354         * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1355         an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1356         (group_compare_offset): Update accordingly.
1357         (split_small_address_groups_p): Likewise.
1358         (record_use): Take addr_offset as a poly_uint64 rather than
1359         an unsigned HOST_WIDE_INT.
1360         (strip_offset): Return the offset as a poly_uint64 rather than
1361         an unsigned HOST_WIDE_INT.
1362         (record_group_use, split_address_groups): Track polynomial offsets.
1363         (add_iv_candidate_for_use): Likewise.
1364         (addr_offset_valid_p): Take the offset as a poly_int64 rather
1365         than a HOST_WIDE_INT.
1366         (strip_offset_1): Return the offset as a poly_int64 rather than
1367         a HOST_WIDE_INT.
1369 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1370             Alan Hayward  <alan.hayward@arm.com>
1371             David Sherwood  <david.sherwood@arm.com>
1373         * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1374         rather than a HOST_WIDE_INT.
1375         * tree.c (get_binfo_at_offset): Likewise.
1377 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1378             Alan Hayward  <alan.hayward@arm.com>
1379             David Sherwood  <david.sherwood@arm.com>
1381         * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1382         rather than a HOST_WIDE_INT.
1383         * tree-sra.c (build_ref_for_offset): Likewise.
1385 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1386             Alan Hayward  <alan.hayward@arm.com>
1387             David Sherwood  <david.sherwood@arm.com>
1389         * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1390         than an offset_int.
1391         * tree.c (mem_ref_offset): Likewise.
1392         (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1393         * builtins.c (get_object_alignment_2): Likewise.
1394         * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1395         * gimple-fold.c (get_base_constructor): Likewise.
1396         * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1397         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1398         Likewise.
1399         * ipa-polymorphic-call.c
1400         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1401         * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1402         (get_ancestor_addr_info): Likewise.
1403         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1404         * match.pd: Likewise.
1405         * tree-data-ref.c (dr_analyze_innermost): Likewise.
1406         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1407         * tree-eh.c (tree_could_trap_p): Likewise.
1408         * tree-object-size.c (addr_object_size): Likewise.
1409         * tree-ssa-address.c (copy_ref_info): Likewise.
1410         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1411         (indirect_refs_may_alias_p): Likewise.
1412         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1413         * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1414         (non_rewritable_mem_ref_base): Likewise.
1415         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1416         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1417         * varasm.c (decode_addr_const): Likewise.
1419 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1420             Alan Hayward  <alan.hayward@arm.com>
1421             David Sherwood  <david.sherwood@arm.com>
1423         * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1424         offsets for MEM_REFs.
1426 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1427             Alan Hayward  <alan.hayward@arm.com>
1428             David Sherwood  <david.sherwood@arm.com>
1430         * tree-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1431         as a poly_uint64 rather than an unsigned int.
1432         * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1434 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1435             Alan Hayward  <alan.hayward@arm.com>
1436             David Sherwood  <david.sherwood@arm.com>
1438         * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1439         to poly_int64.
1440         (decode_addr_const): Update accordingly.
1442 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1443             Alan Hayward  <alan.hayward@arm.com>
1444             David Sherwood  <david.sherwood@arm.com>
1446         * tree.h (bit_field_size, bit_field_offset): New functions.
1447         * hsa-gen.c (gen_hsa_addr): Use them.
1448         * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1449         (simplify_vector_constructor): Likewise.
1450         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1451         * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1452         BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1453         * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1454         tree_fits_uhwi_p.
1456 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1457             Alan Hayward  <alan.hayward@arm.com>
1458             David Sherwood  <david.sherwood@arm.com>
1460         * expr.h (emit_group_load, emit_group_load_into_temps)
1461         (emit_group_store): Take the size as a poly_int64 rather than an int.
1462         * expr.c (emit_group_load_1, emit_group_load): Likewise.
1463         (emit_group_load_into_temp, emit_group_store): Likewise.
1465 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1466             Alan Hayward  <alan.hayward@arm.com>
1467             David Sherwood  <david.sherwood@arm.com>
1469         * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1470         an unsigned int to a poly_uint64.
1471         * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1472         sizes as poly_uint64s rather than unsigned ints.
1473         * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1474         Likewise.
1476 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1477             Alan Hayward  <alan.hayward@arm.com>
1478             David Sherwood  <david.sherwood@arm.com>
1480         * lra-constraints.c (emit_inc): Change inc_amount from an int
1481         to a poly_int64.
1483 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1484             Alan Hayward  <alan.hayward@arm.com>
1485             David Sherwood  <david.sherwood@arm.com>
1487         * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1488         to a poly_uint64.
1489         (add_stack_var, stack_var_cmp, partition_stack_vars)
1490         (dump_stack_var_partition): Update accordingly.
1491         (alloc_stack_frame_space): Take the size as a poly_int64 rather
1492         than a HOST_WIDE_INT.
1493         (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1494         (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1495         (account_stack_vars, expand_one_var): Likewise.  Return a poly_uint64
1496         rather than a HOST_WIDE_INT.
1498 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1499             Alan Hayward  <alan.hayward@arm.com>
1500             David Sherwood  <david.sherwood@arm.com>
1502         * target.def (return_pops_args): Treat both the input and output
1503         sizes as poly_int64s rather than HOST_WIDE_INTS.
1504         * targhooks.h (default_return_pops_args): Update accordingly.
1505         * targhooks.c (default_return_pops_args): Likewise.
1506         * doc/tm.texi: Regenerate.
1507         * emit-rtl.h (incoming_args): Change pops_args, size and
1508         outgoing_args_size from int to poly_int64_pod.
1509         * function.h (expr_status): Change x_pending_stack_adjust and
1510         x_stack_pointer_delta from int to poly_int64.
1511         (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1512         (ARGS_SIZE_RTX): Update accordingly.
1513         * calls.c (highest_outgoing_arg_in_use): Change from int to
1514         unsigned int.
1515         (stack_usage_watermark, stored_args_watermark): New variables.
1516         (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1517         (emit_call_1): Change the stack_size and rounded_stack_size
1518         parameters from HOST_WIDE_INT to poly_int64.  Track n_popped
1519         as a poly_int64.
1520         (save_fixed_argument_area): Check stack_usage_watermark.
1521         (initialize_argument_information): Change old_pending_adj from
1522         a HOST_WIDE_INT * to a poly_int64_pod *.
1523         (compute_argument_block_size): Return the size as a poly_int64
1524         rather than an int.
1525         (finalize_must_preallocate): Track polynomial argument sizes.
1526         (compute_argument_addresses): Likewise.
1527         (internal_arg_pointer_based_exp): Track polynomial offsets.
1528         (mem_overlaps_already_clobbered_arg_p): Rename to...
1529         (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1530         size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1531         Check stored_args_used_watermark.
1532         (load_register_parameters): Update accordingly.
1533         (check_sibcall_argument_overlap_1): Likewise.
1534         (combine_pending_stack_adjustment_and_call): Take the unadjusted
1535         args size as a poly_int64 rather than an int.  Return a bool
1536         indicating whether the optimization was possible and return
1537         the new adjustment by reference.
1538         (check_sibcall_argument_overlap): Track polynomail argument sizes.
1539         Update stored_args_watermark.
1540         (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1541         (expand_call): Likewise.  Maintain stack_usage_watermark and
1542         stored_args_watermark.  Update calls to
1543         combine_pending_stack_adjustment_and_call.
1544         (emit_library_call_value_1): Handle polynomial argument sizes.
1545         Call stack_region_maybe_used_p and mark_stack_region_used.
1546         Maintain stack_usage_watermark.
1547         (store_one_arg): Likewise.  Update call to
1548         mem_overlaps_already_clobbered_arg_p.
1549         * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1550         HOST_WIDE_INT.
1551         * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1552         * config/microblaze/microblaze.c (microblaze_function_prologue):
1553         Likewise.
1554         * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1555         TARGET_RETURN_POPS_ARGS interface.
1556         (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1557         to HOST_WIDE_INT.
1558         * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1559         * config/i386/i386.c (ix86_return_pops_args): Update for new
1560         TARGET_RETURN_POPS_ARGS interface.
1561         (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1562         * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1563         * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1564         TARGET_RETURN_POPS_ARGS interface.
1565         * config/vax/vax.c (vax_return_pops_args): Likewise.
1566         * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1567         (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1568         * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1569         * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1570         * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1571         * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1572         * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1573         * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1574         * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1575         * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1576         and x_stack_pointer_delta from int to poly_int64.
1577         * dojump.c (do_pending_stack_adjust): Update accordingly.
1578         * explow.c (allocate_dynamic_stack_space): Handle polynomial
1579         stack_pointer_deltas.
1580         * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1581         (pad_to_arg_alignment): Track polynomial offsets.
1582         (assign_parm_find_stack_rtl): Likewise.
1583         (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1584         * toplev.c (output_stack_usage): Update reference to
1585         current_function_pushed_stack_size.
1587 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1588             Alan Hayward  <alan.hayward@arm.com>
1589             David Sherwood  <david.sherwood@arm.com>
1591         * function.c (in_arg_offset, var_offset, dynamic_offset)
1592         (out_arg_offset, cfa_offset): Change from int to poly_int64.
1593         (instantiate_new_reg): Return the new offset as a poly_int64_pod
1594         rather than a HOST_WIDE_INT.
1595         (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1596         (instantiate_virtual_regs_in_insn): Likewise.
1598 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1599             Alan Hayward  <alan.hayward@arm.com>
1600             David Sherwood  <david.sherwood@arm.com>
1602         * rtl.h (get_args_size, add_args_size_note): New functions.
1603         (find_args_size_adjust): Return a poly_int64 rather than a
1604         HOST_WIDE_INT.
1605         (fixup_args_size_notes): Likewise.  Make the same change to the
1606         end_args_size parameter.
1607         * rtlanal.c (get_args_size, add_args_size_note): New functions.
1608         * builtins.c (expand_builtin_trap): Use add_args_size_note.
1609         * calls.c (emit_call_1): Likewise.
1610         * explow.c (adjust_stack_1): Likewise.
1611         * cfgcleanup.c (old_insns_match_p): Update use of
1612         find_args_size_adjust.
1613         * combine.c (distribute_notes): Track polynomial arg sizes.
1614         * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1615         end_true_args_size, beg_delay_args_size and end_delay_args_size
1616         from HOST_WIDE_INT to poly_int64.
1617         (add_cfi_args_size): Take the args_size as a poly_int64 rather
1618         than a HOST_WIDE_INT.
1619         (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1620         (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1621         polynomial arg sizes.
1622         * emit-rtl.c (try_split): Use get_args_size.
1623         * recog.c (peep2_attempt): Likewise.
1624         * reload1.c (reload_as_needed): Likewise.
1625         * expr.c (find_args_size_adjust): Return the adjustment as a
1626         poly_int64 rather than a HOST_WIDE_INT.
1627         (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1628         to a poly_int64 and change the return type in the same way.
1629         (emit_single_push_insn): Track polynomial arg sizes.
1631 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1632             Alan Hayward  <alan.hayward@arm.com>
1633             David Sherwood  <david.sherwood@arm.com>
1635         * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1636         from HOST_WIDE_INT to poly_int64.
1637         * expr.c (push_block, emit_push_insn): Likewise.
1639 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1640             Alan Hayward  <alan.hayward@arm.com>
1641             David Sherwood  <david.sherwood@arm.com>
1643         * function.h (frame_space): Change start and length from HOST_WIDE_INT
1644         to poly_int64.
1645         (get_frame_size): Return the size as a poly_int64 rather than a
1646         HOST_WIDE_INT.
1647         (frame_offset_overflow): Take the offset as a poly_int64 rather
1648         than a HOST_WIDE_INT.
1649         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1650         (assign_stack_temp): Likewise for the size.
1651         * function.c (get_frame_size): Return a poly_int64 rather than
1652         a HOST_WIDE_INT.
1653         (frame_offset_overflow): Take the offset as a poly_int64 rather
1654         than a HOST_WIDE_INT.
1655         (try_fit_stack_local): Take the start, length and size as poly_int64s
1656         rather than HOST_WIDE_INTs.  Return the offset as a poly_int64_pod
1657         rather than a HOST_WIDE_INT.
1658         (add_frame_space): Take the start and end as poly_int64s rather than
1659         HOST_WIDE_INTs.
1660         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1661         (assign_stack_temp): Likewise for the size.
1662         (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1663         to poly_int64.
1664         (find_temp_slot_from_address): Handle polynomial offsets.
1665         (combine_temp_slots): Likewise.
1666         * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1667         to poly_int64.
1668         * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1669         poly_int64 rather than a HOST_WIDE_INT.
1670         (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1671         than a HOST_WIDE_INT.
1672         (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1673         polynomial frame offsets.
1674         * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1675         as a poly_int64 rather than an int.
1676         * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1677         * config/v850/v850-protos.h (compute_frame_size): Likewise.
1678         * config/v850/v850.c (compute_frame_size): Likewise.
1679         * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1680         * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1681         * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1682         * config/pa/pa.c (pa_compute_frame_size): Likewise.
1683         * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1684         rather than a HOST_WIDE_INT.
1685         * explow.c (get_dynamic_stack_base): Likewise.
1686         * final.c (final_start_function): Use the constant lower bound
1687         of the frame size for -Wframe-larger-than.
1688         * ira.c (do_reload): Adjust for new get_frame_size return type.
1689         * lra.c (lra): Likewise.
1690         * reload1.c (reload): Likewise.
1691         * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1692         * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1693         * rtlanal.c (get_initial_register_offset): Return the offset as
1694         a poly_int64 rather than a HOST_WIDE_INT.
1696 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1697             Alan Hayward  <alan.hayward@arm.com>
1698             David Sherwood  <david.sherwood@arm.com>
1700         * reload1.c (elim_table): Change initial_offset, offset and
1701         previous_offset from HOST_WIDE_INT to poly_int64_pod.
1702         (offsets_at): Change the target array's element type from
1703         HOST_WIDE_INT to poly_int64_pod.
1704         (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1705         (elimination_costs_in_insn, update_eliminable_offsets)
1706         (verify_initial_elim_offsets, set_offsets_for_label)
1707         (init_eliminable_invariants): Update after above changes.
1709 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1710             Alan Hayward  <alan.hayward@arm.com>
1711             David Sherwood  <david.sherwood@arm.com>
1713         * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1714         * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1715         (decomposition): Change start and end from HOST_WIDE_INT
1716         to poly_int64_pod.
1717         (decompose, immune_p): Update accordingly.
1718         (find_inc_amount): Return a poly_int64 rather than an int.
1719         * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1720         rather than an int.
1722 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1723             Alan Hayward  <alan.hayward@arm.com>
1724             David Sherwood  <david.sherwood@arm.com>
1726         * tree.h (get_inner_reference): Return the bitsize and bitpos
1727         as poly_int64_pods rather than HOST_WIDE_INT.
1728         * fold-const.h (ptr_difference_const): Return the pointer difference
1729         as a poly_int64_pod rather than a HOST_WIDE_INT.
1730         * expr.c (get_inner_reference): Return the bitsize and bitpos
1731         as poly_int64_pods rather than HOST_WIDE_INT.
1732         (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1733         offsets and sizes.
1734         * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1735         rather than a HOST_WIDE_INT.  Update call to get_inner_reference.
1736         (optimize_bit_field_compare): Update call to get_inner_reference.
1737         (decode_field_reference): Likewise.
1738         (fold_unary_loc): Track polynomial offsets and sizes.
1739         (split_address_to_core_and_offset): Return the bitpos as a
1740         poly_int64_pod rather than a HOST_WIDE_INT.
1741         (ptr_difference_const): Likewise for the pointer difference.
1742         * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1743         * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1744         * dbxout.c (dbxout_expand_expr): Likewise.
1745         * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1746         (loc_list_from_tree_1, fortran_common): Likewise.
1747         * gimple-laddress.c (pass_laddress::execute): Likewise.
1748         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1749         * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1750         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1751         * tree-affine.c (tree_to_aff_combination): Likewise.
1752         (get_inner_reference_aff): Likewise.
1753         * tree-data-ref.c (split_constant_offset_1): Likewise.
1754         (dr_analyze_innermost): Likewise.
1755         * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1756         * tree-sra.c (ipa_sra_check_caller): Likewise.
1757         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1758         * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1759         (instrument_bool_enum_load, instrument_object_size): Likewise.
1760         * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1761         to get_inner_reference.
1762         * hsa-gen.c (gen_hsa_addr): Likewise.
1763         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1764         * tsan.c (instrument_expr): Likewise.
1765         * match.pd: Update call to ptr_difference_const.
1767 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1768             Alan Hayward  <alan.hayward@arm.com>
1769             David Sherwood  <david.sherwood@arm.com>
1771         * fold-const.c (fold_comparison): Track sizes and offsets as
1772         poly_int64s rather than HOST_WIDE_INTs when folding address
1773         comparisons.
1775 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1776             Alan Hayward  <alan.hayward@arm.com>
1777             David Sherwood  <david.sherwood@arm.com>
1779         * expr.h (get_bit_range): Return the bitstart and bitend as
1780         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Return the bitpos
1781         as a poly_int64 rather than a HOST_WIDE_INT.
1782         * expr.c (get_bit_range): Likewise.
1783         (expand_assignment): Update call accordingly.
1784         * fold-const.c (optimize_bit_field_compare): Likewise.
1786 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1787             Alan Hayward  <alan.hayward@arm.com>
1788             David Sherwood  <david.sherwood@arm.com>
1790         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1791         (aarch64_legitimate_address_p): Use it instead of an rtx code,
1792         as an optional final parameter.
1793         * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1794         (aarch64_legitimate_address_p): Likewise.
1795         (aarch64_print_address_internal): Take an aarch64_addr_query_type
1796         instead of an rtx code.
1797         (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1798         (aarch64_legitimate_address_hook_p): Likewise.
1799         (aarch64_print_ldpstp_address): Likewise.
1800         (aarch64_print_operand_address): Likewise.
1801         (aarch64_address_cost): Likewise.
1802         * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1803         * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1804         (aarch64_mem_pair_lanes_operand): Likewise.
1806 2017-12-20  Richard Biener  <rguenther@suse.de>
1808         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1809         update_stmt_if_modified.
1811 2017-12-20  Wilco Dijkstra  <wdijkstr@arm.com>
1813         PR tree-optimization/83491
1814         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1815         before walking uses.  Improve coding style and comments.
1817 2017-12-20  Tom de Vries  <tom@codesourcery.com>
1819         * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1821 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
1823         PR ipa/83506
1824         * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1825         todo_flags_finish.
1826         (pass_ipa_free_fn_summary): Add small_p private data member,
1827         initialize to false in the ctor.
1828         (pass_ipa_free_fn_summary::clone,
1829         pass_ipa_free_fn_summary::set_pass_param,
1830         pass_ipa_free_fn_summary::gate): New methods.
1831         (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1832         | TODO_dump_symtab if small_p.
1833         * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1834         entry and add another instance of the pass with false parm after
1835         ipa-pure-const.
1836         * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1837         ipa_free_fn_summary here.
1839 2017-12-20  Paolo Carlini  <paolo.carlini@oracle.com>
1841         * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1843 2017-12-20  Martin Sebor  <msebor@redhat.com>
1845         PR testsuite/83131
1846         * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1847         (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1848         (expand_builtin_strncmp): Same.
1850 2017-12-20  Alexandre Oliva  <aoliva@redhat.com>
1852         PR bootstrap/83396
1853         * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1854         allowed debug stmts before labels.
1855         (expand_gimple_basic_block): Likewise.
1856         * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1857         * gimple-iterator.h (gsi_after_labels): Likewise.
1858         * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1859         rename reused variable, and simplify using gsi_move_before.
1860         * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1861         * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1862         (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1863         (gimple_verify_flow_info, gimple_block_label): Likewise.
1864         (make_blocks): Move debug markers after adjacent labels.
1865         * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1866         allowed debug insns outside blocks.
1867         * df-scan.c (df_insn_delete): Likewise.
1868         * lra-constraints.c (update_ebb_live_info): Likewise.
1869         * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1870         (vt_initialize, delete_vta_debug_insns): Likewise.
1871         (reemit_marker_as_note): Drop BB parm.  Adjust callers.
1873 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1874             Alan Hayward  <alan.hayward@arm.com>
1875             David Sherwood  <david.sherwood@arm.com>
1877         * poly-int-types.h (round_down_to_byte_boundary): New macro.
1878         (round_up_to_byte_boundary): Likewise.
1879         * expr.h (get_bit_range): Add temporary shim.
1880         * gimple-ssa-store-merging.c (store_operand_info): Change the
1881         bitsize, bitpos, bitregion_start and bitregion_end fields from
1882         unsigned HOST_WIDE_INT to poly_uint64.
1883         (merged_store_group): Likewise load_align_base.
1884         (compatible_load_p, compatible_load_p): Update accordingly.
1885         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1886         (split_group, imm_store_chain_info::output_merged_store): Likewise.
1887         (mem_valid_for_store_merging): Return the bitsize, bitpos,
1888         bitregion_start and bitregion_end as poly_uint64s rather than
1889         unsigned HOST_WIDE_INTs.  Track polynomial offsets internally.
1890         (handled_load): Take the bitsize, bitpos,
1891         bitregion_start and bitregion_end as poly_uint64s rather than
1892         unsigned HOST_WIDE_INTs.
1893         (pass_store_merging::process_store): Update call to
1894         mem_valid_for_store_merging.
1896 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1897             Alan Hayward  <alan.hayward@arm.com>
1898             David Sherwood  <david.sherwood@arm.com>
1900         * builtins.c (get_object_alignment_2): Track polynomial offsets
1901         and sizes.  Update the alignment handling.
1903 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1904             Alan Hayward  <alan.hayward@arm.com>
1905             David Sherwood  <david.sherwood@arm.com>
1907         * tree.h (get_inner_reference): Add a version that returns the
1908         offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1909         * cfgexpand.c (expand_debug_expr): Track polynomial offsets.  Simply
1910         the case in which bitpos is not associated with the first byte.
1912 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1913             Alan Hayward  <alan.hayward@arm.com>
1914             David Sherwood  <david.sherwood@arm.com>
1916         * tree-affine.h (get_inner_reference_aff): Return the size as a
1917         poly_widest_int.
1918         * tree-affine.c (get_inner_reference_aff): Likewise.
1919         * tree-data-ref.c (dr_may_alias_p): Update accordingly.
1920         * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
1922 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1923             Alan Hayward  <alan.hayward@arm.com>
1924             David Sherwood  <david.sherwood@arm.com>
1926         * fold-const.c (pointer_may_wrap_p): Take the offset as a
1927         HOST_WIDE_INT rather than a poly_int64.
1929 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1930             Alan Hayward  <alan.hayward@arm.com>
1931             David Sherwood  <david.sherwood@arm.com>
1933         * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
1934         HOST_WIDE_INT to poly_int64_pod.
1935         (perform_symbolic_merge): Update accordingly.
1936         (bswap_replace): Likewise.
1938 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1939             Alan Hayward  <alan.hayward@arm.com>
1940             David Sherwood  <david.sherwood@arm.com>
1942         * tree-affine.h (aff_tree::offset): Change from widest_int
1943         to poly_widest_int.
1944         (wide_int_ext_for_comb): Delete.
1945         (aff_combination_const, aff_comb_cannot_overlap_p): Take the
1946         constants as poly_widest_int rather than widest_int.
1947         (aff_combination_constant_multiple_p): Return the multiplier
1948         as a poly_widest_int.
1949         (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
1950         polynomial offsets.
1951         * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
1952         version static and add an overload for poly_widest_int.
1953         (aff_combination_const, aff_combination_add_cst)
1954         (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
1955         the constants as poly_widest_int rather than widest_int.
1956         (tree_to_aff_combination): Generalize INTEGER_CST case to
1957         poly_int_tree_p.
1958         (aff_combination_to_tree): Track offsets as poly_widest_ints.
1959         (aff_combination_add_product, aff_combination_mult): Handle
1960         polynomial offsets.
1961         (aff_combination_constant_multiple_p): Return the multiplier
1962         as a poly_widest_int.
1963         * tree-predcom.c (determine_offset): Return the offset as a
1964         poly_widest_int.
1965         (split_data_refs_to_components, suitable_component_p): Update
1966         accordingly.
1967         (valid_initializer_p): Update call to
1968         aff_combination_constant_multiple_p.
1969         * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
1970         * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
1971         as a poly_int64 rather than a HOST_WIDE_INT.
1972         (get_address_cost): Handle polynomial offsets.
1973         (iv_elimination_compare_lt): Likewise.
1974         (rewrite_use_nonlinear_expr): Likewise.
1976 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1977             Alan Hayward  <alan.hayward@arm.com>
1978             David Sherwood  <david.sherwood@arm.com>
1980         * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
1981         as a poly_int64_pod rather than a HOST_WIDE_INT.
1982         (get_addr_base_and_unit_offset): Likewise.
1983         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1984         (get_addr_base_and_unit_offset): Likewise.
1985         * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
1986         to poly_int64 in example.
1987         * fold-const.c (fold_binary_loc): Update call to
1988         get_addr_base_and_unit_offset.
1989         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1990         (maybe_canonicalize_mem_ref_addr): Likewise.
1991         (gimple_fold_stmt_to_constant_1): Likewise.
1992         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1993         Likewise.
1994         * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
1995         * match.pd: Likewise.
1996         * omp-low.c (lower_omp_target): Likewise.
1997         * tree-sra.c (build_ref_for_offset): Likewise.
1998         (build_debug_ref_for_model): Likewise.
1999         * tree-ssa-address.c (maybe_fold_tmr): Likewise.
2000         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
2001         * tree-ssa-ccp.c (optimize_memcpy): Likewise.
2002         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
2003         (constant_pointer_difference): Likewise.
2004         * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
2005         * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
2006         * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
2007         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
2008         (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
2009         (set_ssa_val_to): Likewise.
2010         * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
2011         (maybe_diag_stxncpy_trunc): Likewise.
2012         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
2013         * tree.c (build_simple_mem_ref_loc): Likewise.
2014         (array_at_struct_end_p): Likewise.
2016 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2017             Alan Hayward  <alan.hayward@arm.com>
2018             David Sherwood  <david.sherwood@arm.com>
2020         * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
2021         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
2022         (get_ref_base_and_extent_hwi): Declare.
2023         * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
2024         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
2025         (get_ref_base_and_extent_hwi): New function.
2026         * cfgexpand.c (expand_debug_expr): Update call to
2027         get_ref_base_and_extent.
2028         * dwarf2out.c (add_var_loc_to_decl): Likewise.
2029         * gimple-fold.c (get_base_constructor): Return the offset as a
2030         poly_int64_pod rather than a HOST_WIDE_INT.
2031         (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
2032         * ipa-polymorphic-call.c
2033         (ipa_polymorphic_call_context::set_by_invariant)
2034         (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
2035         (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
2036         (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
2037         rather than get_ref_base_and_extent.
2038         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
2039         * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
2040         (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
2041         Likewise.
2042         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
2043         call to get_ref_base_and_extent.
2044         * tree-sra.c (create_access, get_access_for_expr): Likewise.
2045         * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
2046         (stmt_kills_ref_p): Likewise.
2047         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2048         * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
2049         Likewise.
2050         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
2051         Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
2052         when calling native_encode_expr.
2053         * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
2054         call to get_ref_base_and_extent.
2055         (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
2056         get_ref_base_and_extent.
2057         * var-tracking.c (track_expr_p): Likewise.
2059 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2060             Alan Hayward  <alan.hayward@arm.com>
2061             David Sherwood  <david.sherwood@arm.com>
2063         * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
2064         HOST_WIDE_INT to poly_int64_pod.
2065         * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
2066         polynomail parameter offsets.
2068 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2069             Alan Hayward  <alan.hayward@arm.com>
2070             David Sherwood  <david.sherwood@arm.com>
2072         * gengtype.c (main): Handle poly_int64_pod.
2073         * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
2074         (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
2075         (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
2076         from HOST_WIDE_INT to poly_int64_pod.
2077         * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
2078         (copy_cfa): New function.
2079         (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
2080         (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
2081         (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
2082         (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
2083         (initial_return_save): Treat offsets as poly_ints.
2084         (def_cfa_0): Likewise.  Cache the CFA in dw_cfi_cfa_loc if either
2085         offset is nonconstant.
2086         (reg_save): Take the offset as a poly_int64.  Fall back to
2087         DW_CFA_expression for nonconstant offsets.
2088         (queue_reg_save): Take the offset as a poly_int64.
2089         * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
2091 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2092             Alan Hayward  <alan.hayward@arm.com>
2093             David Sherwood  <david.sherwood@arm.com>
2095         * rtl.h (operand_subword, operand_subword_force): Take the offset
2096         as a poly_uint64 an unsigned int.
2097         * emit-rtl.c (operand_subword, operand_subword_force): Likewise.
2099 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2100             Alan Hayward  <alan.hayward@arm.com>
2101             David Sherwood  <david.sherwood@arm.com>
2103         * doc/rtl.texi: Update documentation of SUBREG_BYTE.  Document the
2104         'p' format code.  Use INT_LIST rather than SUBREG as the example of
2105         a code with an XINT and an XEXP.  Remove the implication that
2106         accessing an rtx field using XINT is expected to work.
2107         * rtl.def (SUBREG): Change format from "ei" to "ep".
2108         * rtl.h (rtunion::rt_subreg): New field.
2109         (XCSUBREG): New macro.
2110         (SUBREG_BYTE): Use it.
2111         (subreg_shape): Change offset from an unsigned int to a poly_uint16.
2112         Update constructor accordingly.
2113         (subreg_shape::operator ==): Update accordingly.
2114         (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
2115         than an unsigned int.
2116         (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
2117         a poly_uint64 rather than an unsigned int.
2118         (subreg_lsb_1): Likewise.  Take the offset as a poly_uint64 rather
2119         than an unsigned int.
2120         (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
2121         (subreg_size_highpart_offset): Return a poly_uint64 rather than
2122         an unsigned int.  Take the sizes as poly_uint64s.
2123         (subreg_offset_from_lsb): Return a poly_uint64 rather than
2124         an unsigned int.  Take the shift as a poly_uint64 rather than
2125         an unsigned int.
2126         (subreg_regno_offset, subreg_offset_representable_p): Take the offset
2127         as a poly_uint64 rather than an unsigned int.
2128         (simplify_subreg_regno): Likewise.
2129         (byte_lowpart_offset): Return the memory offset as a poly_int64
2130         rather than an int.
2131         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2132         poly_uint64 rather than an unsigned int.
2133         (simplify_subreg, simplify_gen_subreg, subreg_get_info)
2134         (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
2135         poly_uint64 rather than an unsigned int.
2136         * rtl.c (rtx_format): Describe 'p' in comment.
2137         (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
2138         * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
2139         offset as a poly_uint64 rather than an unsigned int.
2140         (byte_lowpart_offset): Return the memory offset as a poly_int64
2141         rather than an int.
2142         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2143         poly_uint64 rather than an unsigned int.
2144         (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
2145         mode sizes as poly_uint64s rather than unsigned ints.  Return a
2146         poly_uint64 rather than an unsigned int.
2147         (subreg_lowpart_p): Treat subreg offsets as poly_ints.
2148         (copy_insn_1): Handle 'p'.
2149         * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
2150         (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
2151         an unsigned int.  Return the shift in the same way.
2152         (subreg_lsb): Return the shift as a poly_uint64 rather than an
2153         unsigned int.
2154         (subreg_size_offset_from_lsb): Take the sizes and shift as
2155         poly_uint64s rather than unsigned ints.  Return the offset as
2156         a poly_uint64.
2157         (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
2158         (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
2159         an unsigned int.
2160         * rtlhash.c (add_rtx): Handle 'p'.
2161         * genemit.c (gen_exp): Likewise.
2162         * gengenrtl.c (type_from_format, gendef): Likewise.
2163         * gensupport.c (subst_pattern_match, get_alternatives_number)
2164         (collect_insn_data, alter_predicate_for_insn, alter_constraints)
2165         (subst_dup): Likewise.
2166         * gengtype.c (adjust_field_rtx_def): Likewise.
2167         * genrecog.c (find_operand, find_matching_operand, validate_pattern)
2168         (match_pattern_2): Likewise.
2169         (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
2170         (rtx_test::subreg_field): New function.
2171         (operator ==, safe_to_hoist_p, transition_parameter_type)
2172         (print_nonbool_test, print_test): Handle SUBREG_FIELD.
2173         * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
2174         * genpeep.c (match_rtx): Likewise.
2175         * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
2176         (rtx_writer::print_rtx_operand): Handle 'p'.
2177         (print_value): Handle SUBREG.
2178         * read-rtl.c (apply_int_iterator): Likewise.
2179         (rtx_reader::read_rtx_operand): Handle 'p'.
2180         * alias.c (rtx_equal_for_memref_p): Likewise.
2181         * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
2182         * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
2183         as poly_ints.
2184         * calls.c (expand_call): Likewise.
2185         * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
2186         (make_extraction, gen_lowpart_for_combine): Likewise.
2187         * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
2188         Likewise.
2189         * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
2190         rather than an unsigned int.  Treat subreg offsets as poly_ints.
2191         (exp_equiv_p): Handle 'p'.
2192         (hash_rtx_cb): Likewise.  Treat subreg offsets as poly_ints.
2193         (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
2194         * dse.c (find_shift_sequence): Likewise.
2195         * dwarf2out.c (rtl_for_decl_location): Likewise.
2196         * expmed.c (extract_low_bits): Likewise.
2197         * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
2198         (expand_expr_real_2): Likewise.
2199         * final.c (alter_subreg): Likewise.
2200         (leaf_renumber_regs_insn): Handle 'p'.
2201         * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
2202         Treat subreg offsets as poly_ints.
2203         * fwprop.c (forward_propagate_and_simplify): Likewise.
2204         * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
2205         * ira.c (get_subreg_tracking_sizes): Likewise.
2206         * ira-conflicts.c (go_through_subreg): Likewise.
2207         * ira-lives.c (process_single_reg_class_operands): Likewise.
2208         * jump.c (rtx_renumbered_equal_p): Likewise.  Handle 'p'.
2209         * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
2210         as a poly_uint64 rather than an unsigned int.
2211         (simplify_gen_subreg_concatn, resolve_simple_move): Treat
2212         subreg offsets as poly_ints.
2213         * lra-constraints.c (operands_match_p): Handle 'p'.
2214         (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
2215         * lra-spills.c (assign_mem_slot): Likewise.
2216         * postreload.c (move2add_valid_value_p): Likewise.
2217         * recog.c (general_operand, indirect_operand): Likewise.
2218         * regcprop.c (copy_value, maybe_mode_change): Likewise.
2219         (copyprop_hardreg_forward_1): Likewise.
2220         * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
2221         (record_subregs_of_mode): Likewise.
2222         * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
2223         * reload.c (operands_match_p): Handle 'p'.
2224         (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
2225         * reload1.c (alter_reg, choose_reload_regs): Likewise.
2226         (compute_reload_subreg_offset): Likewise, and return an poly_int64.
2227         * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
2228         (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
2229         (simplify_const_poly_int_tests<N>::run): Likewise.
2230         (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
2231         a poly_uint64 rather than an unsigned int.
2232         * valtrack.c (debug_lowpart_subreg): Likewise.
2233         * var-tracking.c (var_lowpart): Likewise.
2234         (loc_cmp): Handle 'p'.
2236 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2237             Alan Hayward  <alan.hayward@arm.com>
2238             David Sherwood  <david.sherwood@arm.com>
2240         * ira.c (get_subreg_tracking_sizes): New function.
2241         (init_live_subregs): Take an integer size rather than a register.
2242         (build_insn_chain): Use get_subreg_tracking_sizes.  Update calls
2243         to init_live_subregs.
2245 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2246             Alan Hayward  <alan.hayward@arm.com>
2247             David Sherwood  <david.sherwood@arm.com>
2249         * expr.c (store_constructor_field): Change bitsize from a
2250         unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2251         HOST_WIDE_INT to a poly_int64.
2252         (store_constructor): Change size from a HOST_WIDE_INT to
2253         a poly_int64.
2254         (store_field): Likewise bitsize and bitpos.
2256 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2257             Alan Hayward  <alan.hayward@arm.com>
2258             David Sherwood  <david.sherwood@arm.com>
2260         * expmed.h (store_bit_field): Change bitregion_start and
2261         bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2262         * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2263         (store_bit_field_1, store_integral_bit_field, store_bit_field)
2264         (store_fixed_bit_field, store_split_bit_field): Likewise.
2265         * expr.c (store_constructor_field, store_field): Likewise.
2266         (optimize_bitfield_assignment_op): Likewise.  Make the same change
2267         to bitsize and bitpos.
2268         * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2269         and m_bitregion_end from HOST_WIDE_INT to poly_int64.  Make the
2270         same change in the constructor arguments.
2271         (get_best_mode): Change bitregion_start and bitregion_end from
2272         unsigned HOST_WIDE_INT to poly_uint64.
2273         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2274         Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2275         poly_int64.
2276         (bit_field_mode_iterator::next_mode): Update for new types
2277         of m_bitregion_start and m_bitregion_end.
2278         (get_best_mode): Change bitregion_start and bitregion_end from
2279         unsigned HOST_WIDE_INT to poly_uint64.
2281 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2282             Alan Hayward  <alan.hayward@arm.com>
2283             David Sherwood  <david.sherwood@arm.com>
2285         * rtl.h (simplify_gen_subreg): Add a temporary overload that
2286         accepts poly_uint64 offsets.
2287         * expmed.h (extract_bit_field): Take bitsize and bitnum as
2288         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2289         * expmed.c (lowpart_bit_field_p): Likewise.
2290         (extract_bit_field_as_subreg): New function, split out from...
2291         (extract_bit_field_1): ...here.  Take bitsize and bitnum as
2292         poly_uint64s rather than unsigned HOST_WIDE_INTs.  For vector
2293         extractions, check that BITSIZE matches the size of the extracted
2294         value and that BITNUM is an exact multiple of that size.
2295         If all else fails, try forcing the value into memory if
2296         BITNUM is variable, and adjusting the address so that the
2297         offset is constant.  Split the part that can only handle constant
2298         bitsize and bitnum out into...
2299         (extract_integral_bit_field): ...this new function.
2300         (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2301         rather than unsigned HOST_WIDE_INTs.
2303 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2304             Alan Hayward  <alan.hayward@arm.com>
2305             David Sherwood  <david.sherwood@arm.com>
2307         * expmed.h (store_bit_field): Take bitsize and bitnum as
2308         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2309         * expmed.c (simple_mem_bitfield_p): Likewise.  Add a parameter
2310         that returns the byte size.
2311         (store_bit_field_1): Take bitsize and bitnum as
2312         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Update call
2313         to simple_mem_bitfield_p.  Split the part that can only handle
2314         constant bitsize and bitnum out into...
2315         (store_integral_bit_field): ...this new function.
2316         (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2317         than unsigned HOST_WIDE_INTs.
2318         (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2320 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2321             Alan Hayward  <alan.hayward@arm.com>
2322             David Sherwood  <david.sherwood@arm.com>
2324         * lra-int.h (lra_reg): Change offset from int to poly_int64.
2325         (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2326         to poly_int64.
2327         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2328         update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2329         (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2330         offset as a poly_int64 rather than an int.
2331         * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2332         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2333         * lra-constraints.c (equiv_address_substitution): Track offsets
2334         as poly_int64s.
2335         (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2336         (curr_insn_transform): Handle the new form of sp_offset.
2337         * lra-eliminations.c (lra_elim_table): Change previous_offset
2338         and offset from HOST_WIDE_INT to poly_int64.
2339         (print_elim_table, update_reg_eliminate): Update accordingly.
2340         (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2341         (get_elimination): Update accordingly.
2342         (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2343         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2344         update_sp_offset from a HOST_WIDE_INT to a poly_int64.  Handle
2345         poly_int64 offsets generally.
2346         (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2347         (mark_not_eliminable, init_elimination): Update accordingly.
2348         (remove_reg_equal_offset_note): Return a bool and pass the new
2349         offset back by pointer as a poly_int64.
2350         * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2351         rather than a HOST_WIDE_INT.
2352         (do_remat): Track offsets poly_int64s.
2353         * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2355 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2356             Alan Hayward  <alan.hayward@arm.com>
2357             David Sherwood  <david.sherwood@arm.com>
2359         * rtl.h (mem_attrs): Add a default constructor.  Change size and
2360         offset from HOST_WIDE_INT to poly_int64.
2361         * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2362         (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2363         (widen_memory_access): Take the sizes and offsets as poly_int64s
2364         rather than HOST_WIDE_INTs.
2365         * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2366         (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2367         and ints.
2368         (adjust_offset_for_component_ref): Change the offset from a
2369         HOST_WIDE_INT to a poly_int64.
2370         (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2371         * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2372         * dce.c (find_call_stack_args): Likewise.
2373         * dse.c (record_store): Likewise.
2374         * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2375         * print-rtl.c (rtx_writer::print_rtx): Likewise.
2376         * read-rtl-function.c (test_loading_mem): Likewise.
2377         * rtlanal.c (may_trap_p_1): Likewise.
2378         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2379         * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2380         * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2381         (mem_attrs::mem_attrs): New function.
2382         (set_mem_attributes_minus_bitpos): Change bitpos from a
2383         HOST_WIDE_INT to poly_int64.
2384         (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2385         (clear_mem_offset, clear_mem_size, change_address)
2386         (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2387         initialize mem_attrs.
2388         (set_mem_offset, set_mem_size, adjust_address_1)
2389         (adjust_automodify_address_1, offset_address, widen_memory_access):
2390         Likewise.  Take poly_int64s rather than HOST_WIDE_INT.
2392 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2393             Alan Hayward  <alan.hayward@arm.com>
2394             David Sherwood  <david.sherwood@arm.com>
2396         * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2397         as poly_int64s rather than HOST_WIDE_INTs.  Use a size of -1
2398         rather than 0 to represent an unknown size.  Assert that the size
2399         is known when the mode isn't BLKmode.
2400         (may_trap_p_1): Use -1 for unknown sizes.
2401         (rtx_addr_can_trap_p): Likewise.  Pass BLKmode rather than VOIDmode.
2403 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2404             Alan Hayward  <alan.hayward@arm.com>
2405             David Sherwood  <david.sherwood@arm.com>
2407         * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2408         to poly_int64.  Update commentary for positions_needed.large.
2409         (read_info_type): Change offset and width from HOST_WIDE_INT
2410         to poly_int64.
2411         (set_usage_bits): Likewise.
2412         (canon_address): Return the offset as a poly_int64 rather than
2413         a HOST_WIDE_INT.  Use strip_offset_and_add.
2414         (set_all_positions_unneeded, any_positions_needed_p): Use
2415         positions_needed.large to track stores with non-constant widths.
2416         (all_positions_needed_p): Likewise.  Take the offset and width
2417         as poly_int64s rather than ints.  Assert that rhs is nonnull.
2418         (record_store): Cope with non-constant offsets and widths.
2419         Nullify the rhs of an earlier store if we can't tell which bytes
2420         of it are needed.
2421         (find_shift_sequence): Take the access_size and shift as poly_int64s
2422         rather than ints.
2423         (get_stored_val): Take the read_offset and read_width as poly_int64s
2424         rather than HOST_WIDE_INTs.
2425         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2426         non-constant offsets and widths.
2428 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2429             Alan Hayward  <alan.hayward@arm.com>
2430             David Sherwood  <david.sherwood@arm.com>
2432         * inchash.h (inchash::hash::add_poly_int): New function.
2433         * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2434         Use poly_int64 rather than HOST_WIDE_INT.
2435         (ao_ref::max_size_known_p): New function.
2436         * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2437         rather than HOST_WIDE_INT.
2438         * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2439         to temporaries until its interface is adjusted to match.
2440         (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2441         (aliasing_component_refs_p, decl_refs_may_alias_p)
2442         (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2443         the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2444         (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2445         ao_ref fields.
2446         * alias.c (ao_ref_from_mem): Likewise.
2447         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2448         * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2449         (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2450         (maybe_trim_complex_store, maybe_trim_constructor_store)
2451         (live_bytes_read, dse_classify_store): Likewise.
2452         * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2453         (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2454         (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2455         (vn_reference_lookup_3): Likewise.
2456         * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
2458 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2459             Alan Hayward  <alan.hayward@arm.com>
2460             David Sherwood  <david.sherwood@arm.com>
2462         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2463         (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2464         instead of ranges_overlap_p.
2466 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2467             Alan Hayward  <alan.hayward@arm.com>
2468             David Sherwood  <david.sherwood@arm.com>
2470         * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2471         sizes as poly_int64s rather than HOST_WIDE_INTs.
2473 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2474             Alan Hayward  <alan.hayward@arm.com>
2475             David Sherwood  <david.sherwood@arm.com>
2477         * gimple-fold.h (fold_ctor_reference): Take the offset and size
2478         as poly_uint64 rather than unsigned HOST_WIDE_INT.
2479         * gimple-fold.c (fold_ctor_reference): Likewise.
2481 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2482             Alan Hayward  <alan.hayward@arm.com>
2483             David Sherwood  <david.sherwood@arm.com>
2485         * target.def (dwarf_poly_indeterminate_value): New hook.
2486         * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2487         * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2488         * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2489         * doc/tm.texi: Regenerate.
2490         * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2491         offset as a poly_int64.
2492         * dwarf2out.c (new_reg_loc_descr): Move later in file.  Take the
2493         offset as a poly_int64.
2494         (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2495         Take the offset as a poly_int64.
2496         (build_cfa_loc): Likewise.  Use loc_descr_plus_const.
2497         (frame_pointer_fb_offset): Change to a poly_int64.
2498         (int_loc_descriptor): Take the offset as a poly_int64.  Use
2499         targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2500         (based_loc_descr): Take the offset as a poly_int64.
2501         Use strip_offset_and_add to handle (plus X (const)).
2502         Use new_reg_loc_descr instead of an open-coded version of the
2503         previous implementation.
2504         (mem_loc_descriptor): Handle CONST_POLY_INT.
2505         (compute_frame_pointer_to_fb_displacement): Take the offset as a
2506         poly_int64.  Use strip_offset_and_add to handle (plus X (const)).
2508 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2509             Alan Hayward  <alan.hayward@arm.com>
2510             David Sherwood  <david.sherwood@arm.com>
2512         * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2513         (gen_rtx_REG_offset): Take the offset as a poly_int64.
2514         * inchash.h (inchash::hash::add_poly_hwi): New function.
2515         * gengtype.c (main): Register poly_int64.
2516         * emit-rtl.c (reg_attr_hasher::hash): Use inchash.  Treat the
2517         offset as a poly_int.
2518         (reg_attr_hasher::equal): Use must_eq to compare offsets.
2519         (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2520         offset as a poly_int64.
2521         (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2522         * print-rtl.c (print_poly_int): New function.
2523         (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2524         a poly_int.
2525         * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2526         functions.
2527         (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2528         (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2529         (vt_get_decl_and_offset): Return the offset as a poly_int64.
2530         Enforce track_offset_p for parts of a PARALLEL.
2531         (vt_add_function_parameter): Use const_offset for the final
2532         offset to track.  Use get_tracked_reg_offset for the parts
2533         of a PARALLEL.
2535 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2536             Alan Hayward  <alan.hayward@arm.com>
2537             David Sherwood  <david.sherwood@arm.com>
2539         * target.def (truly_noop_truncation): Take poly_uint64s instead of
2540         unsigned ints.  Change default to hook_bool_puint64_puint64_true.
2541         * doc/tm.texi: Regenerate.
2542         * hooks.h (hook_bool_uint_uint_true): Delete.
2543         (hook_bool_puint64_puint64_true): Declare.
2544         * hooks.c (hook_bool_uint_uint_true): Delete.
2545         (hook_bool_puint64_puint64_true): New function.
2546         * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2547         instead of unsigned ints.
2548         * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2549         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
2551 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2552             Alan Hayward  <alan.hayward@arm.com>
2553             David Sherwood  <david.sherwood@arm.com>
2555         * optabs.h (expand_operand): Add an int_value field.
2556         (create_expand_operand): Add an int_value parameter and use it
2557         to initialize the new expand_operand field.
2558         (create_integer_operand): Replace with a declaration of a function
2559         that accepts poly_int64s.  Move the implementation to...
2560         * optabs.c (create_integer_operand): ...here.
2561         (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2562         the mode preserves the value of int_value, instead of calling
2563         const_int_operand on the rtx.  Use gen_int_mode to generate
2564         the new rtx.
2566 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2567             Alan Hayward  <alan.hayward@arm.com>
2568             David Sherwood  <david.sherwood@arm.com>
2570         * dumpfile.h (dump_dec): Declare.
2571         * dumpfile.c (dump_dec): New function.
2572         * pretty-print.h (pp_wide_integer): Turn into a function and
2573         declare a poly_int version.
2574         * pretty-print.c (pp_wide_integer): New function for poly_ints.
2576 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2577             Alan Hayward  <alan.hayward@arm.com>
2578             David Sherwood  <david.sherwood@arm.com>
2580         * doc/generic.texi (POLY_INT_CST): Document.
2581         * tree.def (POLY_INT_CST): New tree code.
2582         * treestruct.def (TS_POLY_INT_CST): New tree layout.
2583         * tree-core.h (tree_poly_int_cst): New struct.
2584         (tree_node): Add a poly_int_cst field.
2585         * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2586         (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2587         instead of a wide_int_ref.
2588         (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2589         of a HOST_WIDE_INT.
2590         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2591         instead of an unsigned HOST_WIDE_INT.
2592         (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2593         (ptrdiff_tree_p): Declare.
2594         (tree_to_poly_int64, tree_to_poly_uint64): Likewise.  Provide
2595         extern inline implementations if the target doesn't use POLY_INT_CST.
2596         (poly_int_tree_p): New function.
2597         (wi::unextended_tree): New class.
2598         (wi::int_traits <unextended_tree>): New override.
2599         (wi::extended_tree): Add a default constructor.
2600         (wi::extended_tree::get_tree): New function.
2601         (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2602         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2603         (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2604         (wi::tree_to_poly_wide_ref): New typedefs.
2605         (wi::ints_for): Provide overloads for extended_tree and
2606         unextended_tree.
2607         (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2608         (wi::to_wide): New functions.
2609         (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2610         * tree.c (poly_int_cst_hasher): New struct.
2611         (poly_int_cst_hash_table): New variable.
2612         (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2613         (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2614         POLY_INT_CST.
2615         (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2616         (init_ttree): Initialize poly_int_cst_hash_table.
2617         (build_int_cst, build_int_cst_type, build_invariant_address): Take
2618         a poly_int64 instead of a HOST_WIDE_INT.
2619         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2620         instead of an unsigned HOST_WIDE_INT.
2621         (wide_int_to_tree): Rename to...
2622         (wide_int_to_tree_1): ...this.
2623         (build_new_poly_int_cst, build_poly_int_cst): New functions.
2624         (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2625         (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2626         (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2627         (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2628         * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2629         TS_POLY_INT_CST.
2630         * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2631         (streamer_read_tree_body): Likewise.
2632         * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2633         (streamer_write_tree_body): Likewise.
2634         * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2635         * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2636         * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2637         * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2638         * gimple-expr.h (is_gimple_constant): Likewise.
2639         * gimplify.c (maybe_with_size_expr): Likewise.
2640         * print-tree.c (print_node): Likewise.
2641         * tree-data-ref.c (data_ref_compare_tree): Likewise.
2642         * tree-pretty-print.c (dump_generic_node): Likewise.
2643         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2644         * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2645         * tree-vrp.c (compare_values_warnv): Likewise.
2646         * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2647         (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2648         (force_expr_to_var_cost): Likewise.
2649         * tree-ssa-loop.c (for_each_index): Likewise.
2650         * fold-const.h (build_invariant_address, size_int_kind): Take a
2651         poly_int64 instead of a HOST_WIDE_INT.
2652         * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2653         (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2654         POLY_INT_CST.
2655         (size_binop_loc): Likewise.  Allow int_const_binop_1 to fail.
2656         (int_const_binop_2): New function, split out from...
2657         (int_const_binop_1): ...here.  Handle POLY_INT_CST.
2658         (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2659         * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2660         * gimple-ssa-strength-reduction.c (slsr_process_add)
2661         (slsr_process_mul): Check for INTEGER_CSTs before using them
2662         as candidates.
2663         * stor-layout.c (bits_from_bytes): New function.
2664         (bit_from_pos): Use it.
2665         (layout_type): Likewise.  For vectors, multiply the TYPE_SIZE_UNIT
2666         by BITS_PER_UNIT to get the TYPE_SIZE.
2667         * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2668         MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2670 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2671             Alan Hayward  <alan.hayward@arm.com>
2672             David Sherwood  <david.sherwood@arm.com>
2674         * doc/rtl.texi (const_poly_int): Document.  Also document the
2675         rtl sharing behavior.
2676         * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2677         * rtl.h (const_poly_int_def): New struct.
2678         (rtx_def::u): Add a cpi field.
2679         (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2680         (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2681         (wi::rtx_to_poly_wide_ref): New typedef
2682         (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2683         (poly_int_rtx_p): New functions.
2684         (trunc_int_for_mode): Declare a poly_int64 version.
2685         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2686         (immed_wide_int_const): Take a poly_wide_int_ref rather than
2687         a wide_int_ref.
2688         (strip_offset): Declare.
2689         (strip_offset_and_add): New function.
2690         * rtl.def (CONST_POLY_INT): New rtx code.
2691         * rtl.c (rtx_size): Handle CONST_POLY_INT.
2692         (shared_const_p): Use poly_int_rtx_p.
2693         * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2694         HOST_WIDE_INT.
2695         (gen_int_shift_amount): Likewise.
2696         * emit-rtl.c (const_poly_int_hasher): New class.
2697         (const_poly_int_htab): New variable.
2698         (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2699         (const_poly_int_hasher::hash): New function.
2700         (const_poly_int_hasher::equal): Likewise.
2701         (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2702         (immed_wide_int_const): Rename to...
2703         (immed_wide_int_const_1): ...this and make static.
2704         (immed_wide_int_const): New function, taking a poly_wide_int_ref
2705         instead of a wide_int_ref.
2706         (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2707         (gen_lowpart_common): Handle CONST_POLY_INT.
2708         * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2709         * cselib.c (cselib_hash_rtx): Likewise.
2710         * dwarf2out.c (const_ok_for_output_1): Likewise.
2711         * expr.c (convert_modes): Likewise.
2712         * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2713         * rtlhash.c (add_rtx): Likewise.
2714         * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2715         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2716         Handle existing CONST_POLY_INT rtxes.
2717         * expmed.h (expand_shift): Take a poly_int64 instead of a
2718         HOST_WIDE_INT.
2719         * expmed.c (expand_shift): Likewise.
2720         * rtlanal.c (strip_offset): New function.
2721         (commutative_operand_precedence): Give CONST_POLY_INT the same
2722         precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2723         and CONST_INT.
2724         * rtl-tests.c (const_poly_int_tests): New struct.
2725         (rtl_tests_c_tests): Use it.
2726         * simplify-rtx.c (simplify_const_unary_operation): Handle
2727         CONST_POLY_INT.
2728         (simplify_const_binary_operation): Likewise.
2729         (simplify_binary_operation_1): Fold additions of symbolic constants
2730         and CONST_POLY_INTs.
2731         (simplify_subreg): Handle extensions and truncations of
2732         CONST_POLY_INTs.
2733         (simplify_const_poly_int_tests): New struct.
2734         (simplify_rtx_c_tests): Use it.
2735         * wide-int.h (storage_ref): Add default constructor.
2736         (wide_int_ref_storage): Likewise.
2737         (trailing_wide_ints): Use GTY((user)).
2738         (trailing_wide_ints::operator[]): Add a const version.
2739         (trailing_wide_ints::get_precision): New function.
2740         (trailing_wide_ints::extra_size): Likewise.
2742 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2743             Alan Hayward  <alan.hayward@arm.com>
2744             David Sherwood  <david.sherwood@arm.com>
2746         * emit-rtl.h (gen_int_shift_amount): Declare.
2747         * emit-rtl.c (gen_int_shift_amount): New function.
2748         * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2749         instead of GEN_INT.
2750         * calls.c (shift_return_value): Likewise.
2751         * cse.c (fold_rtx): Likewise.
2752         * dse.c (find_shift_sequence): Likewise.
2753         * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2754         (expand_shift, expand_smod_pow2): Likewise.
2755         * lower-subreg.c (shift_cost): Likewise.
2756         * optabs.c (expand_superword_shift, expand_doubleword_mult)
2757         (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2758         (expand_vec_perm_var): Likewise.
2759         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2760         (simplify_binary_operation_1): Likewise.
2761         * combine.c (try_combine, find_split_point, force_int_to_mode)
2762         (simplify_shift_const_1, simplify_shift_const): Likewise.
2763         (change_zero_ext): Likewise.  Use simplify_gen_binary.
2765 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2767         * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2769 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2771         * doc/invoke.texi (ARM Options): Document accepted extension options
2772         for -march=armv8.3-a.
2774 2017-12-20  Richard Earnshaw  <rearnsha@arm.com>
2776         PR target/83105
2777         * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2778         or --with-float=softfp, set the default CPU to arm10e.
2780 2017-12-20  Eric Botcazou  <ebotcazou@adacore.com>
2782         * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2783         * config/visium/predicates.md (const_shift_operand): Likewise.
2784         * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2785         (visium_legitimize_reload_address): Likewise.
2787 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2789         PR target/82975
2790         * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2792 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
2794         PR c++/83490
2795         * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2797 2017-12-20  Julia Koval  <julia.koval@intel.com>
2799         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2800         OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2801         (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2802         * config.gcc: Include vpclmulqdqintrin.h.
2803         * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2804         * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2805         -mvpclmulqdq.
2806         * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2807         __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2808         * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2809         * config/i386/i386.c (isa2_opts): Add -mcx16.
2810         (isa_opts): Add -mpclmulqdq, remove -mcx16.
2811         (ix86_option_override_internal): Move mcx16 to flags2.
2812         (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2813         (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2814         * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2815         * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2816         * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2817         * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2818         * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2819         _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2820         * doc/invoke.texi: Add -mvpclmulqdq.
2822 2017-12-20  Tom de Vries  <tom@codesourcery.com>
2824         PR middle-end/83423
2825         * config/i386/i386.c (ix86_static_chain): Move
2826         DECL_STATIC_CHAIN test ...
2827         * calls.c (rtx_for_static_chain): ... here.  New function.
2828         * calls.h (rtx_for_static_chain): Declare.
2829         * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2830         instead of targetm.calls.static_chain.
2831         * df-scan.c (df_get_entry_block_def_set): Same.
2833 2017-12-19  Tom de Vries  <tom@codesourcery.com>
2835         PR tree-optimization/83493
2836         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2837         and lb.
2839 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2841         * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2842         inform with hardcoded english plural handling.
2844 2017-12-18  Jeff Law  <law@redhat.com>
2846         PR tree-optimization/83477
2847         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2848         a non-virtual PHI, always push a new range.
2850 2017-12-19  Martin Sebor  <msebor@redhat.com>
2852         PR middle-end/77608
2853         * builtins.c (compute_objsize): Handle non-constant offsets.
2855 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2857         PR tree-optimization/83444
2858         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2859         character load case, if get_stridx on MEM_REF's operand doesn't
2860         look usable, retry with get_addr_stridx.
2862 2017-12-19  Alexandre Oliva  <aoliva@redhat.com>
2864         PR debug/83422
2865         * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2866         (variable_tracking_main_1): Keep markers even when VTA fails.
2868         PR bootstrap/83396
2869         * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2870         even if there are markers before it.
2871         * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2873 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2875         * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2876         typical order conditions.
2877         * sel-sched.c (extract_new_fences_from): Likewise.
2878         * config/visium/constraints.md (J, K, L): Likewise.
2879         * config/visium/predicates.md (const_shift_operand): Likewise.
2880         * config/visium/visium.c (visium_legitimize_address,
2881         visium_legitimize_reload_address): Likewise.
2882         * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2883         * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2884         * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2885         * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2886         SET_DUMP_DETAIL): Likewise.
2887         * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2888         * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2889         avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2890         avr_set_core_architecture, avr_set_current_function,
2891         avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2892         output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2893         avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2894         avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2895         avr_map_decompose, avr_fold_builtin): Likewise.
2896         * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2897         * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2898         * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2899         * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2900         * config/m32c/m32c.c (m32c_conditional_register_usage,
2901         m32c_address_cost): Likewise.
2902         * config/m32c/predicates.md (shiftcount_operand,
2903         longshiftcount_operand): Likewise.
2904         * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2905         * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2906         can_use_cdx_ldstw): Likewise.
2907         * config/nios2/nios2.h (CDX_REG_P): Likewise.
2908         * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2909         Likewise.
2910         * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2911         * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
2912         * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
2913         * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
2914         * config/vax/vax.c (adjacent_operands_p): Likewise.
2915         * config/ft32/constraints.md (L, b, KA): Likewise.
2916         * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
2917         Likewise.
2918         * cfgexpand.c (expand_stack_alignment): Likewise.
2919         * gcse.c (insert_expr_in_table): Likewise.
2920         * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
2921         * cgraphunit.c (cgraph_node::expand): Likewise.
2922         * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
2923         * emit-rtl.c (add_insn): Likewise.
2924         * input.c (dump_location_info): Likewise.
2925         * passes.c (NEXT_PASS): Likewise.
2926         * read-rtl-function.c (parse_note_insn_name,
2927         function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
2928         Likewise.
2929         * sched-rgn.c (sched_rgn_init): Likewise.
2930         * diagnostic-show-locus.c (layout::show_ruler): Likewise.
2931         * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
2932         if_then_else_cond, simplify_shift_const_1, simplify_comparison):
2933         Likewise.
2934         * explow.c (eliminate_constant_term): Likewise.
2935         * final.c (leaf_renumber_regs_insn): Likewise.
2936         * cfgrtl.c (print_rtl_with_bb): Likewise.
2937         * genhooks.c (emit_init_macros): Likewise.
2938         * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
2939         * tree-data-ref.c (conflict_fn): Likewise.
2940         * selftest.c (assert_streq): Likewise.
2941         * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
2942         * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
2943         fold_binary_loc, multiple_of_p): Likewise.
2944         * reload.c (push_reload, find_equiv_reg): Likewise.
2945         * et-forest.c (et_nca, et_below): Likewise.
2946         * dbxout.c (dbxout_symbol_location): Likewise.
2947         * reorg.c (relax_delay_slots): Likewise.
2948         * dojump.c (do_compare_rtx_and_jump): Likewise.
2949         * gengtype-parse.c (type): Likewise.
2950         * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
2951         simplify_const_relational_operation): Likewise.
2952         * reload1.c (do_output_reload): Likewise.
2953         * dumpfile.c (get_dump_file_info_by_switch): Likewise.
2954         * gengtype.c (type_for_name): Likewise.
2955         * gimple-ssa-sprintf.c (format_directive): Likewise.
2957 2017-12-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2959         PR target/82975
2960         * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
2961         accessing it.  Adjust comment.
2963 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2965         PR middle-end/81914
2966         * predict.c (zero_one_minusone): New function.
2967         (apply_return_prediction): Avoid return prediction for functions
2968         returning only -1, 0 and 1 values, unless they only return -1 and 0
2969         or 0 and 1.
2971 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2973         * config/arc/arc.c (legitimate_scaled_address_p): Clean
2974         fall-through warning.
2975         (arc_compute_frame_size): Remove unused variables.
2976         (arc_print_operand): Fix fprintif format.
2977         (arc_can_follow_jump): Clean fall-through warning.
2979 2017-12-19  Marek Polacek  <polacek@redhat.com>
2981         PR c++/83489
2982         * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
2983         on an error node.
2985 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2987         * config/arc/arc.c (overriderregs): New variable.
2988         (arc_override_options): Track fixed/call saved/call options.
2989         (arc_conditional_register_usage): Check against overrideregs
2990         variable whenever we change register properties.
2992 2017-12-19  Nathan Sidwell  <nathan@acm.org>
2994         * opts.c (finish_options): Don't prefix dump_base_name if it
2995         already contains directories.
2997 2017-12-19  Martin Liska  <mliska@suse.cz>
2999         PR rtl-optimization/82675
3000         * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
3001         more element in sbitmap.
3003 2017-12-19  Paolo Carlini  <paolo.carlini@oracle.com>
3005         * gimplify.c (gimplify_expr): Use error_operand_p.
3007 2017-12-19  Sebastian Huber  <sebastian.huber@embedded-brains.de>
3009         PR target/83387
3010         * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
3011         multilib.
3013 2017-12-19  Daniel Cederman  <cederman@gaisler.com>
3015         * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
3016         the jump is to a label.
3018 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
3020         PR tree-optimization/83444
3021         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
3022         character loads.
3024         PR ipa/82801
3025         PR ipa/83346
3026         * ipa-inline.c (flatten_remove_node_hook): New function.
3027         (ipa_inline): Keep only nodes with flatten attribute at the end of
3028         the array in the order from ipa_reverse_postorder, only walk that
3029         portion of array for flattening, if there is more than one such
3030         node, temporarily register a removal hook and ignore removed nodes.
3032         PR tree-optimization/80631
3033         * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
3034         induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
3035         IFN_REDUC_MAX or IFN_REDUC_MIN.
3037 2017-12-18  Jeff Law  <law@redhat.com>
3039         * tree-ssa-dom.c (record_equivalences_from_phis): Do not
3040         record symbolic equivalences from backedges in the CFG.
3042         Revert
3043         2017-11-19  Jeff Law  <law@redhat.com>
3045         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
3046         of degenerates resulting from ignoring an edge.
3048 2017-12-18  Martin Sebor  <msebor@redhat.com>
3050         PR middle-end/83373
3051         PR tree-optimization/78450
3052         * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
3053         (handle_builtin_strlen): Call it.
3055 2017-12-18  Segher Boessenkool  <segher@kernel.crashing.org>
3057         PR rtl-optimization/83424
3058         * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
3060 2017-12-18  Marek Polacek  <polacek@redhat.com>
3062         PR middle-end/83463
3063         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
3064         Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
3065         values.
3067 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3069         * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
3070         (umaddsidi4, umaddsidi_split): Likewise.
3072 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3074         * config/arc/arc.c (arc_legitimate_constant_p): Always check all
3075         constants.
3077 2017-12-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
3079         PR target/83420
3080         * config/s390/s390.c (s390_option_override): Avoid strncpy.
3082 2017-12-18  Richard Biener  <rguenther@suse.de>
3084         PR tree-optimization/81877
3085         * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
3086         (outermost_indep_loop): Adjust.
3087         (ref_indep_loop_p_1): Likewise.  Remove safelen handling again.
3088         (can_sm_ref_p): Adjust.
3090 2017-12-18  Richard Biener  <rguenther@suse.de>
3092         PR middle-end/77291
3093         * tree.c (array_at_struct_end_p): Return true if the underlying
3094         object has space for at least one element in excess of what
3095         the array domain specifies.
3097 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3099         * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
3100         example to avoid overfull hbox.
3101         * doc/invoke.texi (Option Summary): Add missing @gol.
3102         (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
3103         overfull hbox.
3105 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3106             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3108         * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
3109         to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
3110         (AArch64 Options): Document -mverbose-cost-dump.
3111         (ARM Options): Likewise, plus -mflip-thumb.
3113 2017-12-17  Martin Sebor  <msebor@redhat.com>
3115         PR bootstrap/83446
3116         * gimple-ssa-warn-restrict.c
3117         (builtin_memref::offset_out_of_bounds): Correct the handling of
3118         anti-ranges.
3120 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3122         * doc/invoke.texi (ARC Options): Add missing -mlra entry.
3124 2017-12-17  John David Anglin  <danglin@gcc.gnu.org>
3126         * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
3128 2017-12-17  Markus Trippelsdorf  <markus@trippelsdorf.de>
3130         * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
3131         latencies.
3133 2017-12-16  Sandra Loosemore  <sandra@codesourcery.com>
3135         * doc/invoke.texi: Fix some typos.
3137 2017-12-16  Martin Sebor  <msebor@redhat.com>
3139         PR tree-optimization/78918
3140         * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
3141         * builtins.c (check_sizes): Rename...
3142         (check_access): ...to this.  Rename function arguments for clarity.
3143         (check_memop_sizes): Adjust names.
3144         (expand_builtin_memchr, expand_builtin_memcpy): Same.
3145         (expand_builtin_memmove, expand_builtin_mempcpy): Same.
3146         (expand_builtin_strcat, expand_builtin_stpncpy): Same.
3147         (check_strncat_sizes, expand_builtin_strncat): Same.
3148         (expand_builtin_strncpy, expand_builtin_memset): Same.
3149         (expand_builtin_bzero, expand_builtin_memcmp): Same.
3150         (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
3151         (maybe_emit_sprintf_chk_warning): Same.
3152         (expand_builtin_strcpy): Adjust.
3153         (expand_builtin_stpcpy): Same.
3154         (expand_builtin_with_bounds): Detect out-of-bounds accesses
3155         in pointer-checking forms of memcpy, memmove, and mempcpy.
3156         (gcall_to_tree_minimal, max_object_size): Define new functions.
3157         * builtins.h (max_object_size): Declare.
3158         * calls.c (alloc_max_size): Call max_object_size instead of
3159         hardcoding ssizetype limit.
3160         (get_size_range): Handle new argument.
3161         * calls.h (get_size_range): Add a new argument.
3162         * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
3163         * doc/invoke.texi (-Wrestrict): Adjust, add example.
3164         * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
3165         operations.
3166         (gimple_fold_builtin_memory_chk): Same.
3167         (gimple_fold_builtin_stxcpy_chk): New function.
3168         * gimple-ssa-warn-restrict.c: New source.
3169         * gimple-ssa-warn-restrict.h: New header.
3170         * gimple.c (gimple_build_call_from_tree): Propagate location.
3171         * passes.def (pass_warn_restrict): Add new pass.
3172         * tree-pass.h (make_pass_warn_restrict): Declare.
3173         * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
3174         operations.
3175         (handle_builtin_strcat): Same.
3176         (strlen_optimize_stmt): Rename...
3177         (strlen_check_and_optimize_stmt): ...to this.  Handle strncat,
3178         stpncpy, strncpy, and their checking forms.
3180 2017-12-16  Jan Hubicka  <hubicka@ucw.cz>
3182         PR rtl-optimization/82849
3183         * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
3184         and get_max_loop_iterations_int.
3186 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3187             Alan Hayward  <alan.hayward@arm.com>
3188             David Sherwood  <david.sherwood@arm.com>
3190         * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
3191         (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
3192         size as a poly_uint64.
3193         (mode_for_vector, mode_for_int_vector): Take the number of vector
3194         elements as a poly_uint64.
3195         * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
3196         size as a poly_uint64.
3197         (mode_for_vector, mode_for_int_vector): Take the number of vector
3198         elements as a poly_uint64.
3200 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3201             Alan Hayward  <alan.hayward@arm.com>
3202             David Sherwood  <david.sherwood@arm.com>
3204         * machmode.h (MACRO_MODE): New macro.
3205         * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
3206         * targhooks.c (default_libcall_value, default_secondary_reload)
3207         (default_memory_move_cost, default_register_move_cost)
3208         (default_class_max_nregs): Likewise.
3210 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3211             Alan Hayward  <alan.hayward@arm.com>
3212             David Sherwood  <david.sherwood@arm.com>
3214         * doc/sourcebuild.texi: Document IN_TARGET_CODE.
3215         * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
3216         target C file.
3217         * genautomata.c (main): Likewise.
3218         * genconditions.c (write_header): Likewise.
3219         * genemit.c (main): Likewise.
3220         * genextract.c (print_header): Likewise.
3221         * genopinit.c (main): Likewise.
3222         * genoutput.c (output_prologue): Likewise.
3223         * genpeep.c (main): Likewise.
3224         * genpreds.c (write_insn_preds_c): Likewise.
3225         * genrecog.c (writer_header): Likewise.
3226         * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
3227         * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
3228         * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
3229         * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
3230         * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
3231         * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
3232         * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
3233         * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
3234         * config/arc/arc.c (IN_TARGET_CODE): Likewise.
3235         * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
3236         * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
3237         * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3238         * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3239         * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3240         * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3241         * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3242         * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3243         * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3244         * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3245         * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3246         * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3247         * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3248         * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3249         * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3250         * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3251         * config/darwin.c (IN_TARGET_CODE): Likewise.
3252         * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3253         * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3254         * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3255         * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3256         * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3257         * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3258         * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3259         * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3260         * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3261         * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3262         * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3263         * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3264         * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3265         * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3266         * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3267         * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3268         * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3269         * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3270         * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3271         * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3272         * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3273         * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3274         * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3275         * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3276         * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3277         * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3278         * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3279         * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3280         * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3281         * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3282         * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3283         * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3284         * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3285         * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3286         * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3287         * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3288         * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3289         * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3290         * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3291         * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3292         * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3293         * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3294         * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3295         * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3296         * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3297         * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3298         * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3299         * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3300         * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3301         * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3302         * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3303         * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3304         * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3305         * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3306         * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3307         * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3308         * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3309         * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3310         * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3311         * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3312         * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3313         * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3314         * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3315         * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3316         * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3317         * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3318         * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3319         * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3320         * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3321         * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3322         * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3323         * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3324         * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3325         * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3326         * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3327         * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3328         * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3329         * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3330         * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3331         * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3332         * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3333         * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3334         * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3335         * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3336         * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3337         * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3338         * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3339         * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3340         * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3341         * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3342         * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3343         * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3344         * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3345         * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3346         * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3347         * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3348         * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3349         * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3350         * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3351         * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3352         * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3353         * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3354         * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3355         * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3356         * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3357         * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3359 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3360             Alan Hayward  <alan.hayward@arm.com>
3361             David Sherwood  <david.sherwood@arm.com>
3363         * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3364         checks for MEM_REF.
3366 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3367             Alan Hayward  <alan.hayward@arm.com>
3368             David Sherwood  <david.sherwood@arm.com>
3370         * doc/generic.texi (VEC_SERIES_EXPR): Document.
3371         * doc/md.texi (vec_series@var{m}): Document.
3372         * tree.def (VEC_SERIES_EXPR): New tree code.
3373         * tree.h (build_vec_series): Declare.
3374         * tree.c (build_vec_series): New function.
3375         * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3376         * tree-pretty-print.c (dump_generic_node): Likewise.
3377         * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3378         * tree-inline.c (estimate_operator_cost): Likewise.
3379         * expr.c (expand_expr_real_2): Likewise.
3380         * optabs-tree.c (optab_for_tree_code): Likewise.
3381         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3382         * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3383         * expmed.c (make_tree): Handle VEC_SERIES.
3384         * optabs.def (vec_series_optab): New optab.
3385         * optabs.h (expand_vec_series_expr): Declare.
3386         * optabs.c (expand_vec_series_expr): New function.
3387         * tree-vect-generic.c (expand_vector_operations_1): Check that
3388         the operands also have vector type.
3390 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3391             Alan Hayward  <alan.hawyard@arm.com>
3392             David Sherwood  <david.sherwood@arm.com>
3394         * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3395         (VEC_COND_EXPR): Add missing @tindex.
3396         * doc/md.texi (vec_duplicate@var{m}): Document.
3397         * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3398         * tree.c (build_vector_from_val): Add stubbed-out handling of
3399         variable-length vectors, using VEC_DUPLICATE_EXPR.
3400         (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3401         * cfgexpand.c (expand_debug_expr): Likewise.
3402         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3403         * tree-inline.c (estimate_operator_cost): Likewise.
3404         * tree-pretty-print.c (dump_generic_node): Likewise.
3405         * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3406         * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3407         (test_vec_duplicate_folding): New function.
3408         (fold_const_c_tests): Call it.
3409         * optabs.def (vec_duplicate_optab): New optab.
3410         * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3411         * optabs.h (expand_vector_broadcast): Declare.
3412         * optabs.c (expand_vector_broadcast): Make non-static.  Try using
3413         vec_duplicate_optab.
3414         * expr.c (store_constructor): Try using vec_duplicate_optab for
3415         uniform vectors.
3416         (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3418 2017-12-15  Markus Trippelsdorf  <markus@trippelsdorf.de>
3420         PR target/83358
3421         * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3422         div/mod latencies a bit.
3424 2017-12-15  Jeff Law  <law@redhat.com>
3426         PR tree-optimization/36550
3427         * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3428         (mark_threaded_blocks): Rewrite code to avoid block copying when
3429         optimizing for size.  Don't pessimize blocks which will be
3430         copied, but all the statements will be dead.
3432 2017-12-15  Alexandre Oliva  <aoliva@redhat.com>
3434         PR tree-optimization/81165
3435         * tree-ssa-threadupdate.c (uses_in_bb): New.
3436         (estimate_threading_killed_stmts): New.
3437         * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3438         * tree-ssa-threadedge.c
3439         (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3440         when its hit.
3442 2017-12-15  Jeff Law  <law@redhat.com>
3444         PR tree-optimization/83410
3445         * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3446         threads when parallelizing loops.
3448 2017-12-15  Jakub Jelinek  <jakub@redhat.com>
3450         * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3451         handler fields.
3452         * config/alpha/alpha.c (vms_attribute_table): Swap
3453         affects_type_identity and handler fields, adjust comments.
3454         * config/mips/mips.c (mips_attribute_table): Likewise.
3455         * config/visium/visium.c (visium_attribute_table): Likewise.
3456         * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3457         * config/microblaze/microblaze.c (microblaze_attribute_table):
3458         Likewise.
3459         * config/spu/spu.c (spu_attribute_table): Likewise.
3460         * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3461         * config/arc/arc.c (arc_attribute_table): Likewise.
3462         * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3463         * config/v850/v850.c (v850_handle_interrupt_attribute,
3464         v850_handle_data_area_attribute): Formatting fixes.
3465         (v850_attribute_table): Swap affects_type_identity and handler
3466         fields, adjust comments.
3467         * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3468         * config/arm/arm.c (arm_attribute_table): Likewise.
3469         * config/avr/avr.c (avr_attribute_table): Likewise.
3470         * config/s390/s390.c (s390_attribute_table): Likewise.
3471         * config/sh/sh.c (sh_attribute_table): Likewise.
3472         * config/i386/i386.c (ix86_handle_cconv_attribute,
3473         ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3474         (ix86_attribute_table): Swap affects_type_identity and handler
3475         fields, adjust comments.
3476         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3477         * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3478         * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3479         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3480         * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3481         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3482         * config/rx/rx.c (rx_attribute_table): Likewise.
3483         * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3484         * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3485         * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3486         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3487         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3488         * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3489         * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3490         * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3491         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3492         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3493         * doc/plugins.texi (user_attr): Likewise.  Add NULL for
3494         exclude.
3495         * attribs.c (empty_attribute_table): Swap affects_type_identity and
3496         handler fields.
3497         (register_scoped_attributes, decl_attributes): Formatting fixes.
3499         PR tree-optimization/83269
3500         * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3501         subtraction in arg0's type if type is signed and arg0 is unsigned.
3502         Formatting fix.
3504         PR sanitizer/81281
3505         * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3506         convert? on @0 instead of convert.  Check type of @1, not @0.
3507         ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3508         convert? on @0 instead of convert.  Check type of @1, not @0.
3509         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3510         only optimize if either both @1 and @2 types are narrower
3511         precision, or both are wider or equal precision, and in the former
3512         case only if both have undefined overflow.
3514 2017-12-15  Richard Biener  <rguenther@suse.de>
3516         PR lto/83388
3517         * internal-fn.def (IFN_NOP): Add.
3518         * internal-fn.c (expand_NOP): Do nothing.
3519         * lto-streamer-in.c (input_function): Instead of removing
3520         sanitizer calls replace them with IFN_NOP calls.
3522 2017-12-15  Richard Sandiford  <richard.sandiford@linaro.org>
3523             Alan Hayward  <alan.hayward@arm.com>
3524             David Sherwood  <david.sherwood@arm.com>
3526         * dse.c (store_info, read_info_type): Replace begin and end with
3527         offset and width.
3528         (print_range): New function.
3529         (set_all_positions_unneeded, any_positions_needed_p)
3530         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3531         accordingly.
3532         (record_store): Likewise.  Optimize the case in which all positions
3533         are unneeded.
3534         (get_stored_val): Replace read_begin and read_end with read_offset
3535         and read_width.
3536         (replace_read): Update call accordingly.
3538 2017-12-15  Bin Cheng  <bin.cheng@arm.com>
3540         * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3541         (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3542         (loop_cand::loop_cand): Initialize above members.
3543         (loop_cand::supported_operations): Delete.
3544         (loop_cand::can_interchange_p): Inline above function.
3545         (loop_cand::classify_simple_reduction): Record number of constant
3546         initialized simple reductions.
3547         (should_interchange_loops): New parameters.  Check stmt cost of loops
3548         to be interchange.
3549         (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3550         Update call to should_interchange_loops.
3551         (should_interchange_loop_nest): Update call to
3552         should_interchange_loops.
3554 2017-12-15  Eric Botcazou  <ebotcazou@adacore.com>
3556         PR target/66488
3557         * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3558         * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3559         * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3561 2017-12-15  Julia Koval  <julia.koval@intel.com>
3563         * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3564         __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3565         New.
3566         * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3567         * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3568         _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3570 2017-12-15  Julia Koval  <julia.koval@intel.com>
3572         * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3573         __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3574         * config/i386/sse.md (vaesenc_<mode>): New pattern.
3575         * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3576         _mm_aesenc_epi128): New intrinsics.
3578 2017-12-15  Julia Koval  <julia.koval@intel.com>
3580         * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3581         __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3582         New.
3583         * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3584         * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3585         _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3587 2017-12-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3589         * gimple-ssa-strength-reduction.c (analyze_increments):
3590         Distinguish replacement costs for constant strides from those for
3591         unknown strides.
3593 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3595         * var-tracking.c (variable_tracking_main_1): Formatting fix.
3597 2017-12-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3599         * doc/invoke.texi: Document -Wcast-function-type.
3600         * recog.h (stored_funcptr): Change signature.
3601         * tree-dump.c (dump_node): Avoid warning.
3602         * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3604 2017-12-14  Qing Zhao  <qing.zhao@oracle.com>
3606         PR middle_end/79538
3607         * gimple-fold.c (get_range_strlen): Add the handling of non-member 
3608         array.
3610 2017-12-14  David Malcolm  <dmalcolm@redhat.com>
3612         PR tree-optimization/83312
3613         * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3614         * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3615         "val" param, and to cope with arbitrary basic blocks.
3616         (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3617         handle NULL_TREE for "val", dropping "bb" param.
3618         (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3619         drop "bb" param.  Handle NULL_TREE for "val".
3620         (find_case_label_for_value): Make "switch_stmt" param const.
3621         * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3622         of dom_walker.
3623         (vrp_prop::check_all_array_refs): Reimplement as...
3624         (check_array_bounds_dom_walker::before_dom_children): ...this new
3625         vfunc.  Replace linear search through BB block list, excluding
3626         those with non-executable in-edges via dominator walk.
3628 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3630         * config/arm/arm.opt (mverbose-cost-dump): New option.
3631         * config/arm/arm.c (arm_rtx_costs): Use it.
3633 2017-12-14  Andreas Schwab  <schwab@linux-m68k.org>
3635         PR bootstrap/83396
3636         * reload1.c (emit_input_reload_insns): Skip debug markers.
3638 2017-12-14  Alexandre Oliva  <aoliva@redhat.com>
3640         * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3641         for its nop_endbr.
3643         PR bootstrap/83396
3644         * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3645         * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3646         * config/sh/sh.c: Skip debug insns besides notes.
3647         * config/sh/sh.md: Likewise.
3648         * config/sh/sh_treg_combine.cc: Likewise.
3649         * config/sh/sync.md: Likewise.
3651 2017-12-14  Tom de Vries  <tom@codesourcery.com>
3653         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3654         Add item for weak_undefined.
3656 2017-12-14  Richard Biener  <rguenther@suse.de>
3658         PR tree-optimization/67842
3659         * sese.h (bb_in_region): Remove #if 0'ed code.
3661 2017-12-14  Richard Biener  <rguenther@suse.de>
3663         PR tree-optimization/83326
3664         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3665         may_be_zero parameter and handle it by not marking the first
3666         peeled copy as not exiting the loop.
3667         (try_peel_loop): Likewise.
3668         (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3669         to handle the case of constant or zero iterations and perform
3670         loop header copying on-the-fly.
3672 2017-12-14  Richard Biener  <rguenther@suse.de>
3674         PR tree-optimization/83418
3675         * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3676         Instead of asserting we don't get unfolded comparisons deal with them.
3678 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3680         PR bootstrap/83396
3681         * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3682         basic blocks.  Assert debug bind insns don't appear outside of bbs,
3683         don't reset them.  Assert insns without BLOCK_FOR_INSN are outside of
3684         bb.  Simplify.
3686         PR tree-optimization/83198
3687         * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3688         dir.modifier, regardless of TREE_TYPE (arg).  Assume non-REAL_CST
3689         value if arg is a REAL_CST with incompatible type.
3691 2017-12-14  Sudakshina Das  <sudi.das@arm.com>
3692             Bin Cheng  <bin.cheng@arm.com>
3694         PR target/81228
3695         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3696         to CCFPEmode.
3697         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3698         LTGT.
3700 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3702         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3703         cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3705 2017-12-14  Richard Sandiford  <richard.sandiford@linaro.org>
3706             Alan Hayward  <alan.hayward@arm.com>
3707             David Sherwood  <david.sherwood@arm.com>
3709         * poly-int.h: New file.
3710         * poly-int-types.h: Likewise.
3711         * coretypes.h: Include them.
3712         (POLY_INT_CONVERSION): Define.
3713         * target.def (estimated_poly_value): New hook.
3714         * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3715         * doc/tm.texi: Regenerate.
3716         * doc/poly-int.texi: New file.
3717         * doc/gccint.texi: Include it.
3718         * doc/rtl.texi: Describe restrictions on subreg modes.
3719         * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3720         * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3721         (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3722         * targhooks.h (default_estimated_poly_value): Declare.
3723         * targhooks.c (default_estimated_poly_value): New function.
3724         * target.h (estimated_poly_value): Likewise.
3725         * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3726         (wi::unary_traits): Delete.
3727         (wi::binary_traits::signed_shift_result_type): Define for
3728         offset_int << HOST_WIDE_INT, etc.
3729         (generic_wide_int::operator <<=): Define for all types and use
3730         wi::lshift instead of <<.
3731         (wi::hwi_with_prec): Add a default constructor.
3732         (wi::ints_for): New class.
3733         (operator <<): Define for all wide-int types.
3734         (operator /): New function.
3735         (operator %): Likewise.
3736         * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3737         (ASSERT_MAYBE_NE_AT): New macros.
3739 2017-12-13  Eric Botcazou  <ebotcazou@adacore.com>
3740             Dominik Vogt  <vogt@linux.vnet.ibm.com>
3742         PR middle-end/78468
3743         * emit-rtl.c (init_emit): Remove ??? comment.
3744         * explow.c (get_dynamic_stack_size): Take known alignment of stack
3745         pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3746         * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3747         alignment of 3 virtual registers to BITS_PER_WORD.
3749         * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3751 2017-12-13  Peter Bergner  <bergner@vnet.ibm.com>
3753         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3754         * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3756 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3758         PR bootstrap/83396
3759         * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3761 2017-12-13  Jakub Jelinek  <jakub@redhat.com>
3763         * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3764         with the exception of debug begin stmt markers appear before
3765         labels.
3767         PR bootstrap/83396
3768         * final.c (rest_of_handle_final): Call variable_tracking_main only
3769         if !flag_var_tracking.
3771 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3772             Jakub Jelinek  <jakub@redhat.com>
3774         PR bootstrap/83396
3775         PR debug/83391
3776         * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3777         labels debug stmts that can only appear after labels.
3779 2017-12-13  Alexander Monakov  <amonakov@ispras.ru>
3781         PR rtl-optimization/82398
3782         * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3783         EXPR_USEFULNESS in priority comparison.
3785 2017-12-13  Segher Boessenkool  <segher@kernel.crashing.org>
3787         PR rtl-optimization/83393
3788         * combine.c (move_deaths): If reg_stat points to a too new insn in
3789         last_death, do not use it: find the proper insn instead.
3791 2017-12-12  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3793         PR target/82190
3794         * config/rs6000/rs6000-string.c (expand_block_compare,
3795         expand_strn_compare): Fix set_mem_size() calls.
3797 2017-12-12  Jeff Law  <law@redhat.com>
3799         PR tree-optimization/83298
3800         PR tree-optimization/83362
3801         PR tree-optimization/83383
3802         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3803         push_value_range a public interface.  Add new argument to
3804         record_ranges_from_stmt.
3805         * gimple-ssa-evrp-analyze.c
3806         (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3807         Update comments.  Handle recording temporary equivalences.
3808         * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3809         new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3810         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3811         * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3812         gimple-ssa-evrp-analyze.h.
3813         (record_temporary_equivalences_from_phis): Add new argument.  When
3814         the PHI arg is an SSA_NAME, set the result's range to the range
3815         of the PHI arg.
3816         (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3817         from statements too.
3818         (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3819         Pass it down to children as needed.
3820         (thread_outgoing_edges): Likewise.
3821         (thread_across_edge): Likewise.   Push/pop range state as needed.
3822         * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3824 2017-12-12  Julia Koval  <julia.koval@intel.com>
3826         * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3827         (PTA_CANNONLAKE): Remove PTA_CLWB.
3829 2017-12-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3831         PR target/83332
3832         * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3833         (vcondv2div2df): Likewise.
3834         (vconduv2dfv2di): Likewise.
3836 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3838         * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3839         BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3840         BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3841         ATTR_MATHFN_FPROUNDING_ERRNO.
3843 2017-12-12  Richard Biener  <rguenther@suse.de>
3845         PR tree-optimization/83385
3846         * graphite-scop-detection.c (get_order, order): Remove.
3847         (bb_to_rpo): New global.
3848         (cmp_pbbs): Adjust.
3849         (build_scops): Sort pbbs in RPO order.
3851 2017-12-12  James Greenhalgh  <james.greenhalgh@arm.com>
3853         * combine.c (simplify_set): Do not transform subregs to zero_extends
3854         if the destination mode is a vector mode.
3856 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3858         PR tree-optimization/83359
3859         * tree-cfg.h (fold_loop_internal_call): Declare.
3860         * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3861         * tree-cfg.c (fold_loop_internal_call): ... here.  No longer static.
3862         (find_loop_dist_alias): New function.
3863         (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3864         updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3865         calls.
3867         PR tree-optimization/80631
3868         * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3869         (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3870         arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3871         hardcoding zero as the value if COND_EXPR is never true.  For
3872         INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3873         INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3874         hardcoding MAX_EXPR as the reduction operation.
3875         (is_nonwrapping_integer_induction): Allow negative step.
3876         (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3877         vect_create_epilog_for_reduction, if no value is suitable, don't
3878         use INTEGER_INDUC_COND_REDUCTION for now.  Formatting fixes.
3880 2017-12-12  Richard Biener  <rguenther@suse.de>
3882         PR tree-optimization/81889
3883         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3884         range info from the non-wrapping IV instead of just the range
3885         of the type.
3887 2017-12-12  Julia Koval  <julia.koval@intel.com>
3889         * config.gcc: Add vaesintrin.h.
3890         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3891         New type.
3892         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3893         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3894         New builtins.
3895         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3896         * config/i386/immintrin.h: Include vaesintrin.h.
3897         * config/i386/sse.md (vaesdec_<mode>): New pattern.
3898         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3899         _mm_aesdec_epi128): New intrinsics.
3901 2017-12-12  Julia Koval  <julia.koval@intel.com>
3903         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3904         OPTION_MASK_ISA_VAES_UNSET): New.
3905         (ix86_handle_option): Handle -mvaes.
3906         * config/i386/cpuid.h: Define bit_VAES.
3907         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3908         * config/i386/i386-c.c (__VAES__): New.
3909         * config/i386/i386.c (ix86_target_string): Add -mvaes.
3910         (ix86_valid_target_attribute_inner_p): Ditto.
3911         * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
3912         * config/i386/i386.opt: Add -mvaes.
3913         * doc/invoke.texi: Ditto.
3915 2017-12-12  Alexandre Oliva  <aoliva@redhat.com>
3917         * debug.h (gcc_debug_hooks): Add inline_entry.
3918         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
3919         * debug.c (do_nothing_debug_hooks): Likewise.
3920         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
3921         * dwarf2out.c (dwarf2_debug_hooks): Likewise.
3922         (dwarf2_lineno_debug_hooks): Likewise.
3924         * common.opt (gstatement-frontiers): New, setting
3925         debug_nonbind_markers_p.
3926         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
3927         * toplev.c (process_options): Autodetect value for debug statement
3928         frontiers option.
3929         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
3930         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
3932         * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
3933         markers.  Integrate source bind into debug stmt expand loop.
3934         (pass_expand::execute): Check debug marker limit.  Avoid deep
3935         TER and expand debug locations for debug bind insns only.
3936         * cse.c (insn_live_p): Keep nonbind markers and debug bindings
3937         followed by them.
3938         * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
3939         * final.c (reemit_insn_block_notes): Take current block from
3940         nonbind markers.  Declare note where it's first set.
3941         (final_scan_insn): Handle begin stmt notes.  Emit is_stmt according to
3942         begin stmt markers if enabled.
3943         (notice_source_line): Handle nonbind markers.  Fail if their
3944         location is unknown or that of builtins.
3945         (rest_of_handle_final): Convert begin stmt markers to notes if
3946         var-tracking didn't run.
3947         (rest_of_clean_state): Skip begin stmt markers.
3948         * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
3949         markers.
3950         * function.c (allocate_struct_function): Set begin_stmt_markers.
3951         * function.h (struct function): Add debug_marker_count counter
3952         and debug_nonbind_markers flag.
3953         * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
3954         * gimple-low.c (lower_function_body): Adjust
3955         debug_nonbind_markers.
3956         (lower_stmt): Drop or skip gimple debug stmts.
3957         (lower_try_catch): Skip debug stmts.
3958         * gimple.c (gimple_build_debug_begin_stmt): New.
3959         (gimple_copy): Increment debug_marker_count if copying one.
3960         * gimple.h (gimple_build_debug_begin_stmt): Declare.
3961         * gimplify.c (rexpr_location): New.
3962         (rexpr_has_location): New.
3963         (warn_switch_unreachable_r): Handle gimple debug stmts.
3964         (shortcut_cond_r): Call expr_location.
3965         (find_goto): New.
3966         (find_goto_label): New.
3967         (shortcut_cond_expr): Call expr_has_location, expr_location, and
3968         find_goto_label.
3969         (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
3970         expr_location.
3971         (gimplify_expr): Handle begin stmt markers.  Reject debug expr decls.
3972         * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New.  Add to...
3973         (LANG_HOOKS_INITIALIZER): ... this.
3974         * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
3975         * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
3976         debug insns.
3977         (update_ebb_live_info): Skip debug insn markers.
3978         * lra.c (debug_insn_static_data): Rename to...
3979         (debug_bind_static_data): ... this.
3980         (debug_marker_static_data): New.
3981         (lra_set_insn_recog_data): Select one of the above depending
3982         on debug insn kind.
3983         (lra_update_isn_regno_info): Don't assume debug insns have
3984         freqs.
3985         (push_insns): Skip debug insns.
3986         * lto-streamer-in.c (input_function): Drop debug stmts
3987         depending on active options.  Adjust debug_nonbind_markers.
3988         * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
3989         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3990         begin stmt marker notes.
3991         (print_insn): Likewise.
3992         * recog.c (extract_insn): Recognize rtl for debug markers.
3993         * rtl.def (DEBUG_MARKER): New.
3994         * tree-inline.c: Include params.h.
3995         (remap_gimple_stmt): Handle nonbind markers.
3996         (maybe_move_debug_stmts_to_successors): Likewise.
3997         (copy_debug_stmt): Likewise.
3998         * tree-iterator.c (append_to_statement_list_1): Append begin stmt
3999         markers regardless of no side effects.
4000         (tsi_link_before): Don't update container's side effects when adding
4001         a begin stmt marker.
4002         (tsi_link_after): Likewise.
4003         (expr_first): Skip begin stmt markers.
4004         (expr_last): Likewise.
4005         * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
4006         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4007         Disregard nonbind markers.
4008         * tree.c (make_node_stat): Don't set side effects for begin stmt
4009         markers.
4010         (build1_stat): Likewise.
4011         * tree.def (DEBUG_BEGIN_STMT): New.
4012         * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
4013         * var-tracking.c (delete_debug_insns): Renamed to...
4014         (delete_vta_debug_insns): ... this.
4015         (reemit_marker_as_note): New.
4016         (vt_initialize): Reemit markers.
4017         (delete_vta_debug_insns): Likewise.
4018         (vt_debug_insns_local): Reemit or delete markers.
4019         (variable_tracking_main_1): Likewise.
4020         * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
4021         * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
4022         (gimple_debug_nonbind_marker_p): New.
4023         (gimple_build_debug_bind): Adjust.
4024         (gimple_build_debug_begin_stmt): New.
4025         * doc/invoke.texi (max-debug-marker-count): New param.
4026         * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
4027         (debug_parameter_ref, debug_marker): New.
4028         (NOTE_INSN_BEGIN_STMT): New.
4029         (DEBUG_INSN): Describe begin stmt markers.
4031         * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
4032         without debug insns, we wouldn't, but clean up debug insns
4033         after a control flow insn nevertheless.
4035         * cfgcleanup.c (delete_unreachable_blocks): Use alternate
4036         block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
4037         * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
4038         * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
4039         (rtl_tidy_fallthru_edge): Likewise.
4040         (rtl_verify_fallthru): Likewise.
4041         (rtl_verify_bb_layout): Likewise.
4042         (skip_insns_after_block): Likewise.
4043         (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
4044         * dwarf2out.c: Include print-rtl.h.
4045         (dwarf2out_next_real_insn): New.
4046         (dwarf2out_var_location): Call it.  Disregard begin stmt markers.
4047         Dump debug binds in asm comments.
4048         * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
4049         * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
4050         callers to use gsi_start_nondebug_bb instead.
4051         (gsi_after_labels): Skip gimple debug stmts.
4052         (gsi_start_nondebug): New.
4053         * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
4054         (proper_loop_form_for_interchange): Adjust.
4055         * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
4056         * gimple.h (gimple_seq_last_nondebug_stmt): New.
4057         * gimplify.c (last_stmt_in_scope): Skip debug stmts.
4058         (collect_fallthrough_labels): Likewise.
4059         (should_warn_for_implicit_fallthrough): Likewise.
4060         (warn_implicit_fallthrough_r): Likewise.
4061         (expand_FALLTHROUGH_r): Likewise.
4062         * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
4063         (graphite_copy_stmts_from_block): Skip nonbind markers.
4064         * haifa-sched.c (sched_extend_bb): Skip debug insns.
4065         * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
4066         * jump.c (clean_barriers): Skip debug insns.
4067         * omp-expand.c (expand_parallel_call): Skip debug insns.
4068         (expand_task_call): Likewise.
4069         (remove_exit_barrier): Likewise.
4070         (expand_omp_taskreg): Likewise.
4071         (expand_omp_for_init_counts): Likewise.
4072         (expand_omp_for_generic): Likewise.
4073         (expand_omp_for_static_nochunk): Likewise.
4074         (expand_omp_for_static_chunk): Likewise.
4075         (expand_omp_simd): Likewise.
4076         (expand_omp_taskloop_for_outer): Likewise.
4077         (expand_omp_taskloop_for_inner): Likewise.
4078         (expand_oacc_for): Likewise.
4079         (expand_omp_sections): Likewise.
4080         (expand_omp_single): Likewise.
4081         (expand_omp_synch): Likewise.
4082         (expand_omp_atomic_load): Likewise.
4083         (expand_omp_atomic_store): Likewise.
4084         (expand_omp_atomic_fetch_op): Likewise.
4085         (expand_omp_atomic_pipeline): Likewise.
4086         (expand_omp_atomic_mutex): Likewise.
4087         (expand_omp_target): Likewise.
4088         (grid_expand_omp_for_loop): Likewise.
4089         (grid_expand_target_grid_body): Likewise.
4090         (build_omp_regions_1): Likewise.
4091         * omp-low.c (check_combined_parallel): Skip debug stmts.
4092         * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
4093         * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
4094         * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
4095         test.
4096         * tree-cfg.c (make_blobs_1): Skip debug stmts.
4097         (make_edges): Likewise.
4098         (cleanup_dead_labels): Likewise.
4099         (gimple_can_merge_blocks_p): Likewise.
4100         (stmt_starts_bb_p): Likewise.
4101         (gimple_block_label): Likewise.
4102         (gimple_redirect_edge_and_branch): Likewise.
4103         * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
4104         of debug stmts.
4105         (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
4106         TDF_SLIM.
4107         * tree-pretty-print (print_declaration): Omit initializer in slim
4108         dumps.
4109         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
4110         markers.
4111         (eliminate_unnecessary_stmts): Stabilize block removal order.
4112         * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
4113         * var-tracking.c (get_first_insn): New.
4114         (vt_emit_notes): Call it.
4115         (vt_initialize): Walk any insns before the first BB.
4116         (delete_debug_insns): Likewise.
4118         * gimple.h (enum gimple_debug_subcode): Add
4119         GIMPLE_DEBUG_BEGIN_STMT.
4120         (gimple_debug_begin_stmt_p): New.
4121         (gimple_debug_nonbind_marker_p): New.
4122         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
4123         (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
4124         (MAY_HAVE_DEBUG_STMTS): ... this.  Check both.
4125         * insn-notes.def (BEGIN_STMT): New.
4126         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
4127         (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
4128         (MAY_HAVE_DEBUG_INSNS): ... this.  Check both.
4129         (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
4130         (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
4131         (INSN_DEBUG_MARKER_KIND): New.
4132         (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
4133         (INSN_VAR_LOCATION): Check for VAR_LOCATION.
4134         (INSN_VAR_LOCATION_PTR): New.
4135         * cfgexpand.c (expand_debug_locations): Handle debug bind insns
4136         only.
4137         (expand_gimple_basic_block): Likewise.  Emit debug temps for TER
4138         deps only if debug bind insns are enabled.
4139         (pass_expand::execute): Avoid deep TER and expand
4140         debug locations for debug bind insns only.
4141         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
4142         debug stmts special handling down to debug bind stmts.
4143         * combine.c (try_combine): Narrow debug insns special handling
4144         down to debug bind insns.
4145         * cse.c (delete_trivially_dead_insns): Handle debug bindings.
4146         Narrow debug insns preexisting special handling down to debug
4147         bind insns.
4148         * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
4149         handling down to debug bind insns.
4150         * function.c (instantiate_virtual_regs): Skip debug markers,
4151         adjust handling of debug binds.
4152         * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
4153         temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
4154         * haifa-sched.c (schedule_insn): Narrow special handling of debug
4155         insns to debug bind insns.
4156         * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
4157         special handling of debug stmts to debug bind stmts.
4158         * ipa-split.c (split_function): Likewise.
4159         * ira.c (combine_and_move_insns): Adjust debug bind insns only.
4160         * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
4161         debug insns.
4162         * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
4163         * regrename.c (build_def_use): Likewise.
4164         * regcprop.c (copyprop_hardreg_forward_1): Likewise.
4165         (pass_cprop_hardreg): Narrow special casing of debug insns to
4166         debug bind insns.
4167         * regstat.c (regstat_init_n_sets_and_refs): Likewise.
4168         * reload1.c (reload): Likewise.
4169         * sese.c (sese_insert_phis_for_liveouts): Narrow special
4170         casing of debug stmts to debug bind stmts.
4171         * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
4172         * ssa-iterators.h (num_imm_uses): Likewise.
4173         * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
4174         debug stmts to debug bind stmts.
4175         * tree-inline.c (tree_function_versioning): Narrow special casing
4176         of debug stmts to debug bind stmts.
4177         * tree-loop-distribution.c (generate_loops_for_partition):
4178         Narrow special casing of debug stmts to debug bind stmts.
4179         * tree-sra.c (analyze_access_subtree): Narrow special casing
4180         of debug stmts to debug bind stmts.
4181         * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
4182         stmts to debug bind stmts.
4183         * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
4184         casing of debug stmts to debug bind stmts.
4185         * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
4186         * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
4187         casing of debug stmts to debug bind stmts.
4188         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4189         Likewise.
4190         * tree-ssa.c (flush_pending_stmts): Narrow special casing of
4191         debug stmts to debug bind stmts.
4192         (gimple_replace_ssa_lhs): Likewise.
4193         (insert_debug_temp_for_var_def): Likewise.
4194         (insert_debug_temps_for_defs): Likewise.
4195         (reset_debug_uses): Likewise.
4196         * tree-ssanames.c (release_ssa_name_fn): Likewise.
4197         * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
4198         (adjust_debug_stmts): Likewise.
4199         (adjust_phi_and_debug_stmts): Likewise.
4200         (vect_do_peeling): Likewise.
4201         * tree-vect-loop.c (vect_transform_loop): Likewise.
4202         * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
4203         * var-tracking.c (adjust_mems): Narrow special casing of debug
4204         insns to debug bind insns.
4205         (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
4206         (compute_bb_dataflow, vt_find_locations): Likewise.
4207         (vt_expand_loc, emit_notes_for_changes): Likewise.
4208         (vt_init_cfa_base): Likewise.
4209         (vt_emit_notes): Likewise.
4210         (vt_initialize): Likewise.
4211         (vt_finalize): Likewise.
4213         * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
4214         (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
4215         (next_nonnote_nondebug_insn_bb): New.
4216         (prev_nonnote_nondebug_insn_bb): New.
4217         (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
4218         * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
4219         (prev_nonnote_nondebug_insn_bb): Declare.
4220         (next_nonnote_nondebug_insn_bb): Declare.
4221         * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
4222         * cfgrtl.c (get_last_bb_insn): Likewise.
4223         * lra.c (push_insns): Likewise.
4225 2017-12-11  David Malcolm  <dmalcolm@redhat.com>
4227         PR c/82050
4228         * selftest-run-tests.c (selftest::run_tests): Move start/finish code
4229         to...
4230         * selftest.c (selftest::test_runner::test_runner): New ctor.
4231         (selftest::test_runner::~test_runner): New dtor.
4232         * selftest.h (class selftest::test_runner): New class.
4234 2017-12-11  Carl Love  <cel@us.ibm.com>
4236         * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
4237         vec_extract_fp32_from_shortl]): Add #defines.
4238         * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4239         * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4240         ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4241         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4242         ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4243         ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4244         * doc/extend.texi: Add documentation for the added builtins.
4246 2017-12-11  Alexandre Oliva  <aoliva@redhat.com>
4248         PR rtl-optimization/80693
4249         PR rtl-optimization/81019
4250         PR rtl-optimization/81020
4251         * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4252         are not mentioned in i3.  Place the REG_UNUSED note on i2,
4253         possibly modified to REG_DEAD, if it did not originate in i3.
4255 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4257         * recog.c (store_data_bypass_p_1): New function.
4258         (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER.  Use
4259         store_data_bypass_p_1 to avoid code duplication.  Formatting fixes.
4261 2017-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
4263         PR rtl-optimization/83361
4264         * ifcvt.c (if_convert): Call fixup_partitions.
4266 2017-12-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
4268         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4269         early folding of splat_u{8,16,32}.
4271 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4273         * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4274         output_operand_lossage first argument with capital letter.
4275         (aarch64_override_options): Don't start error and sorry first argument
4276         with capital letter.
4278 2017-12-11  Andi Kleen  <ak@linux.intel.com>
4280         PR gcov-profile/83355
4281         * auto-profile.c (string_table::get_index_by_decl): Don't
4282         recurse when abstract origin points to itself.
4284 2017-12-11  Bin Cheng  <bin.cheng@arm.com>
4286         PR tree-optimization/83320
4287         * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4288         (prune_datarefs_not_in_loop): Ditto.
4290 2017-12-10  Gerald Pfeifer  <gerald@pfeifer.com>
4292         * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4294 2017-12-10  Jakub Jelinek  <jakub@redhat.com>
4296         PR tree-optimization/83337
4297         * gimple-loop-interchange.cc (compute_access_stride): Handle
4298         bitfield DRs properly.
4300 2017-12-09  Jakub Jelinek  <jakub@redhat.com>
4302         PR tree-optimization/83338
4303         * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4304         vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4305         vector of unsigned integers to vector of signed integers.
4307 2017-12-08  Vladimir Makarov  <vmakarov@redhat.com>
4309         PR rtl-optimization/83317
4310         * lra-constraints.c (process_address_1): Add insn code check.
4312 2017-12-08  Michael Matz  <matz@suse.de>
4314         Fix PR tree-optimization/83323
4315         * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4316         head-controlled loops and loop BBs.
4317         * common.opt (funroll-and-jam): Remove, instead ...
4318         (floop-unroll-and-jam): ... reuse this option.
4319         * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4320         * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4321         (-floop-unroll-and-jam): ... this option.
4323 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4325         * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4326         a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4327         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4329 2017-12-08  Richard Biener  <rguenther@suse.de>
4331         PR middle-end/81782
4332         * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4333         handle accesses outside of zero-sized vars.
4335 2017-12-08  Martin Jambor  <mjambor@suse.cz>
4337         PR tree-optimization/83141
4338         * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4339         test for MEM_REFs implicitely changing types with padding.  Remove
4340         inline keyword.
4341         (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4343 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4345         * config/arc/arc.c (arc_attribute_table): Add exclusions to
4346         the comment.
4347         * config/avr/avr.c (avr_attribute_table): Likewise.
4348         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4349         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4350         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4351         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4352         of struct attribute_spec.
4353         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4355 2017-12-08  Ulrich Weigand  <uweigand@de.ibm.com>
4357         PR target/82960
4358         * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4360 2017-12-08  Jan Hubicka  <hubicka@ucw.cz>
4362         PR middle-end/83609
4363         * profile-count.c (profile_count::from_gcov_type): Move from
4364         profile-count.h; handle overflow.
4365         * profile-count.h (profile_count::from_gcov_type): Move offline.
4367 2017-12-08  Segher Boessenkool  <segher@kernel.crashing.org>
4369         PR rtl-optimization/83304
4370         * combine.c (move_deaths): If we do not know where a register died,
4371         search for it.
4373 2017-12-08  Richard Biener  <rguenther@suse.de>
4375         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4376         Provide -fopt-info-loop feedback when we interchange in a nest.
4378 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4380         * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4381         for armv6 ARM CPU IDs.
4383 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4385         * common/config/arm/arm-common.c: Include <algorithm>.
4386         (INCLUDE_VECTOR): Define.
4387         (compare_opt_names): New function.
4388         (arm_rewrite_selected_arch): Only strip out extensions that can be
4389         expressed through -mfpu.  Sort the remaining extensions
4390         alphabetically.
4392 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4394         * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4395         (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4396         (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4397         * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4399 2017-06-08  Tristan Gingold  <gindold@adacore.com>
4401         PR ada/81470
4402         * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4403         .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4405 2017-12-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4407         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4408         of struct attribute_spec.
4410 2017-12-08  Julia Koval  <julia.koval@intel.com>
4412         * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4413         _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4414         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4415         _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4416         _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4417         _mm_maskz_dpwssds_epi32): Ditto.
4419 2017-12-08  Richard Biener  <rguenther@suse.de>
4421         PR tree-optimization/81303
4422         * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4423         conditions try to create a comparison vector type matching
4424         the data vector type.
4425         (vectorizable_condition): Adjust.
4426         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4427         Leave invariant conditions alone in case we can vectorize those.
4429 2017-12-08  Julia Koval  <julia.koval@intel.com>
4431         * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4432         _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4433         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4434         _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4435         _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4437 2017-12-08  Julia Koval  <julia.koval@intel.com>
4439         * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4440         _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4441         * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4442         _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4443         _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4444         _mm_maskz_dpbusds_epi32): New intrinsics.
4446 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4448         * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4449         operand.
4450         (ld<bh>io_signed): Likewise.
4451         (st<bhw>io): Likewise.
4452         * config/nios2/predicates.md (ldstio_memory_operand): Allow
4453         SMALL_INT12 constant integer operand.
4455 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4457         * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4458         Rename to...
4459         (nios2_large_constant_allowed): ...this.  Adjust uses.
4460         (nios2_plus_symbolic_constant_p): Rename to...
4461         (nios2_plus_large_constant_p): ...this.  Adjust uses.
4462         (nios2_legitimate_address_p): Correct CONST_INT handling.
4463         (nios2_symbolic_memory_operand_p): Rename to...
4464         (nios2_large_constant_memory_operand_p): ...this.  Adjust uses.
4465         (nios2_large_constant_p): Check for large constant integers too.
4466         (nios2_split_large_constant): Handle constant integers.
4467         (nios2_split_symbolic_memory_operand): Rename to...
4468         (nios2_split_large_constant_memory_operand): ...this.  Adjust uses.
4469         (nios2_legitimize_constant_address): Handle constant integers.
4470         (r0rel_constant_p): Handle small constant integers.
4471         (nios2_print_operand_address): Handle r0-relative integer addresses.
4472         * config/nios2/nios2-protos.h: Adjust for renamed functions.
4473         * config/nios2/nios2.md: Adjust for renamed functions.
4475 2017-12-07  Andrew Waterman  <andrew@sifive.com>
4477         * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4478         (TARGET_HAVE_SRODATA_SECTION): New define.
4479         (riscv_select_section): New function.
4481 2017-12-08  Joseph Myers  <joseph@codesourcery.com>
4482             Alexander Monakov  <amonakov@ispras.ru>
4483             Jakub Jelinek  <jakub@redhat.com>
4485         PR target/81906
4486         * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4488 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4490         * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4491         a bool success value.  Don't call output_operand_lossage here.
4492         (aarch64_print_ldpstp_address): Return a bool success value.
4493         (aarch64_print_operand_address): Call output_addr_const if
4494         aarch64_print_address_internal fails.
4495         (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4496         'y'; call output_operand_lossage instead.  Call output_operand_lossage
4497         if aarch64_print_ldpstp_address fails.
4499 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4501         * tree-vector-builder.h
4502         (tree_vector_builder::binary_encoded_nelts): Declare.
4503         * tree-vector-builder.c
4504         (tree_vector_builder::binary_encoded_nelts): New function.
4505         * fold-const.c (negate_expr_p): Likewise.
4506         (operand_equal_p, fold_checksum_tree): Likewise.
4507         * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4508         * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4509         (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4510         (uniform_vector_p): Likewise.
4511         * varasm.c (const_hash_1, compare_constant): Likewise.
4512         * tree-ssa-ccp.c: Include tree-vector-builder.h.
4513         (valid_lattice_transition): Operate directly on the VECTOR_CST
4514         encoding.
4515         * ipa-icf.c: Include tree-vector-builder.h.
4516         (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4517         * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4519 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4521         * tree.c (build_vector): Delete.
4522         * tree.h (build_vector): Make static and move into the self-testing
4523         block.
4525 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4527         * vector-builder.h (vector_builder::derived): New const overload.
4528         (vector_builder::elt): New function.
4529         * tree-vector-builder.h (tree_vector_builder::type): New function.
4530         (tree_vector_builder::apply_step): Declare.
4531         * tree-vector-builder.c (tree_vector_builder::apply_step): New
4532         function.
4533         * gimple-fold.h (tree_vector_builder): Declare.
4534         (gimple_build_vector): Take a tree_vector_builder instead of a
4535         type and vector of elements.
4536         * gimple-fold.c (gimple_build_vector): Likewise.
4537         * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4538         accordingly.
4539         (get_initial_defs_for_reduction): Likewise.
4540         (vectorizable_induction): Likewise.
4542 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4544         * tree-vector-builder.h
4545         (tree_vector_builder::new_binary_operation): Declare.
4546         * tree-vector-builder.c
4547         (tree_vector_builder::new_binary_operation): New function.
4548         * fold-const.c (fold_relational_const): Use it.
4549         (const_binop): Likewise.  Check that both input vectors have
4550         the same number of elements, thus excluding things like WIDEN_SUM.
4551         Check whether it is possible to operate directly on the encodings
4552         of stepped inputs.
4554 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4556         * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4557         new_unary_operation, operating only on the encoded elements.
4558         (const_unop): Likewise.
4559         (exact_inverse): Likewise.
4560         (distributes_over_addition_p): New function.
4561         (const_binop): Use tree_vector_builder and new_unary_operation
4562         for combinations of VECTOR_CST and INTEGER_CST.  Operate only
4563         on the encoded elements unless the encoding is strided and the
4564         operation does not distribute over addition.
4565         (fold_convert_const):  Use tree_vector_builder and
4566         new_unary_operation.  Operate only on the encoded elements
4567         for truncating integer conversions, or for non-stepped encodings.
4569 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4571         * config/sparc/sparc.c: Include tree-vector-builder.h.
4572         (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4573         * expmed.c: Include tree-vector-builder.h.
4574         (make_tree): Use tree_vector_builder instead of build_vector.
4575         * fold-const.c: Include tree-vector-builder.h.
4576         (const_binop): Use tree_vector_builder instead of build_vector.
4577         (const_unop): Likewise.
4578         (native_interpret_vector): Likewise.
4579         (fold_vec_perm): Likewise.
4580         (fold_ternary_loc): Likewise.
4581         * gimple-fold.c: Include tree-vector-builder.h.
4582         (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4583         of build_vector.
4584         * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4585         (simplify_vector_constructor): Use tree_vector_builder instead
4586         of build_vector.
4587         * tree-vect-generic.c: Include tree-vector-builder.h.
4588         (add_rshift): Use tree_vector_builder instead of build_vector.
4589         (expand_vector_divmod): Likewise.
4590         (optimize_vector_constructor): Likewise.
4591         * tree-vect-loop.c: Include tree-vector-builder.h.
4592         (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4593         of build_vector.  Explicitly use a stepped encoding for
4594         { 1, 2, 3, ... }.
4595         * tree-vect-slp.c: Include tree-vector-builder.h.
4596         (vect_get_constant_vectors): Use tree_vector_builder instead
4597         of build_vector.
4598         (vect_transform_slp_perm_load): Likewise.
4599         (vect_schedule_slp_instance): Likewise.
4600         * tree-vect-stmts.c: Include tree-vector-builder.h.
4601         (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4602         (vect_gen_perm_mask_any): Likewise.
4603         (vectorizable_call): Likewise.  Explicitly use a stepped encoding.
4604         * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4605         of build_vector.
4606         (build_vector_from_val): Likewise.  Explicitly use a duplicate
4607         encoding.
4609 2017-12-07  Richard Sandiford  <richard.sandiford@arm.com>
4611         * doc/generic.texi (VECTOR_CST): Describe new representation of
4612         vector constants.
4613         * vector-builder.h: New file.
4614         * tree-vector-builder.h: Likewise.
4615         * tree-vector-builder.c: Likewise.
4616         * Makefile.in (OBJS): Add tree-vector-builder.o.
4617         * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4618         * tree-core.h (tree_base): Add a vector_cst field to the u union.
4619         (tree_vector): Change the number of elements to
4620         vector_cst_encoded_nelts.
4621         * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4622         (VECTOR_CST_ELTS): Delete.
4623         (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4624         (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4625         (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4626         (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4627         (VECTOR_CST_ENCODED_ELT): Likewise.
4628         (vector_cst_encoded_nelts): New function.
4629         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4630         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4631         (vector_cst_int_elt, vector_cst_elt): Declare.
4632         * tree.c: Include tree-vector-builder.h.
4633         (tree_code_size): Abort if passed VECTOR_CST.
4634         (tree_size): Update for new VECTOR_CST layout.
4635         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4636         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4637         (build_vector): Use tree_vector_builder.
4638         (vector_cst_int_elt, vector_cst_elt): New functions.
4639         (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4640         encoded elements and then create the vector in the canonical form.
4641         (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4642         (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4643         (tree_c_tests): Call test_vector_cst_patterns.
4644         * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4645         VECTOR_CST fields.
4646         (hash_tree): Likewise.
4647         * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4648         (streamer_write_tree_header): Likewise.
4649         * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4650         (streamer_alloc_tree): Likewise.  Update call to make_vector.
4651         * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4653 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4655         * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4656         (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4657         * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4659 2017-12-07  Bin Cheng  <bin.cheng@arm.com>
4660             Richard Biener  <rguenther@suse.de>
4662         PR tree-optimization/81303
4663         * Makefile.in (gimple-loop-interchange.o): New object file.
4664         * common.opt (floop-interchange): Reuse the option from graphite.
4665         * doc/invoke.texi (-floop-interchange): Ditto.  New document for
4666         -floop-interchange and mention it for -O3.
4667         * opts.c (default_options_table): Enable -floop-interchange at -O3.
4668         * gimple-loop-interchange.cc: New file.
4669         * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4670         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4671         * passes.def (pass_linterchange): New pass.
4672         * timevar.def (TV_LINTERCHANGE): New time var.
4673         * tree-pass.h (make_pass_linterchange): New declaration.
4674         * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4675         interchange.  Record IV before/after increment in new parameters.
4676         * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4677         * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4678         path check into...
4679         (check_reduction_path): ...New function here.
4680         * tree-vectorizer.h (check_reduction_path): New declaration.
4682 2017-12-07  Vladimir Makarov  <vmakarov@redhat.com>
4684         PR target/83252
4685         PR rtl-optimization/80818
4686         * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4687         always early clobbered.
4688         * lra-lives.c (process_bb_lives): Check input hard regs for early
4689         clobbered non-operand hard reg.
4691 2017-12-07  Jakub Jelinek  <jakub@redhat.com>
4693         PR middle-end/83164
4694         * tree-cfg.c (verify_gimple_assign_binary): Don't require
4695         types_compatible_p, just that TYPE_MODE is the same.
4697 2017-12-07  Martin Sebor  <msebor@redhat.com>
4699         PR c/81544
4700         * attribs.c (empty_attribute_table): Initialize new member of
4701         struct attribute_spec.
4702         (decl_attributes): Add argument.  Handle mutually exclusive
4703         combinations of attributes.
4704         (selftests::test_attribute_exclusions): New function.
4705         (selftests::attribute_c_tests): Ditto.
4706         * attribs.h (decl_attributes): Add default argument.
4707         * selftest.h (attribute_c_tests): Declare.
4708         * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4709         * tree-core.h (attribute_spec::exclusions, exclude): New type and
4710         member.
4711         * doc/extend.texi (Common Function Attributes): Update const and pure.
4712         * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4713         of struct attribute_spec.
4714         * config/arc/arc.c (arc_attribute_table): Same.
4715         * config/arm/arm.c (arm_attribute_table): Same.
4716         * config/avr/avr.c ( avr_attribute_table): Same.
4717         * config/bfin/bfin.c (bfin_attribute_table): Same.
4718         * config/cr16/cr16.c (cr16_attribute_table): Same.
4719         * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4720         * config/h8300/h8300.c (h8300_attribute_table): Same.
4721         * config/i386/i386.c (ix86_attribute_table): Same.
4722         * config/ia64/ia64.c (ia64_attribute_table): Same.
4723         * config/m32c/m32c.c (m32c_attribute_table): Same.
4724         * config/m32r/m32r.c (m32r_attribute_table): Same.
4725         * config/m68k/m68k.c (m68k_attribute_table): Same.
4726         * config/mcore/mcore.c (mcore_attribute_table): Same.
4727         * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4728         * config/mips/mips.c (mips_attribute_table): Same.
4729         * config/msp430/msp430.c (msp430_attribute_table): Same.
4730         * config/nds32/nds32.c (nds32_attribute_table): Same.
4731         * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4732         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4733         * config/rl78/rl78.c (rl78__attribute_table): Same.
4734         * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4735         * onfig/rx/rx.c (rx_attribute_table): Same.
4736         * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4737         * config/sh/sh.c (sh_attribute_table): Same.
4738         * config/sparc/sparc.c (sparc_attribute_table): Same.
4739         * config/spu/spu.c (spu_attribute_table): Same.
4740         * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4741         * config/v850/v850.c (v850_attribute_table): Same.
4742         * config/visium/visium.c (visium_attribute_table): Same.
4744 2017-12-07  Tamar Christina  <tamar.christina@arm.com>
4746         PR target/82641
4747         * config/arm/arm.c (INCLUDE_STRING): Define.
4748         (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4749         (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4750         and .fpu.
4752 2017-12-07  Michael Matz  <matz@suse.de>
4754         Add unroll and jam pass
4756         * gimple-loop-jam.c: New file.
4757         * Makefile.in (OBJS): Add gimple-loop-jam.o.
4758         * common.opt (funroll-and-jam): New option.
4759         * opts.c (default_options_table): Add unroll-and-jam at -O3.
4760         * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4761         (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4762         * passes.def: Add pass_loop_jam.
4763         * timevar.def (TV_LOOP_JAM): Add.
4764         * tree-pass.h (make_pass_loop_jam): Declare.
4765         * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4766         * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4767         * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4768         to flow_loop_tree_node_add.
4769         (duplicate_subloops, copy_loops_to): Append to sibling list.
4770         * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4771         * doc/invoke.texi (-funroll-and-jam): Document new option.
4772         (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4774 2017-12-07  Richard Biener  <rguenther@suse.de>
4776         PR tree-optimization/83296
4777         PR tree-optimization/67769
4778         * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4779         flow sensitive info in an unrelated BB.
4780         (value_replacement): Use reset_flow_sensitive_info.
4781         (minmax_replacement): Reset flow sensitive info on the def
4782         we move.  Do not reset flow sensitive info in the whole BB
4783         we move the stmt to.
4784         (abs_replacement): Likewise.
4786 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4788         PR target/43871
4789         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4790         rs6000_cpu to the given -mcpu=, or to the default processor.
4792 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4794         * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4795         * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4796         instead of rs6000_cpu_attr.
4797         (rs6000_adjust_cost): Ditto.
4798         (is_microcoded_insn): Ditto.
4799         (rs6000_adjust_priority): Ditto.
4800         (rs6000_issue_rate): Ditto.
4801         (rs6000_use_sched_lookahead): Ditto.
4802         (rs6000_use_sched_lookahead_guard): Ditto.
4803         (rs6000_sched_reorder): Ditto.
4804         (force_new_group): Ditto.
4805         * config/rs6000/rs6000.md (cpu attribute): Ditto.
4806         (group_ending_nop): Ditto.
4808 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4810         * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4811         * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4812         rs6000_tune.  Use rs6000_tune instead of rs6000_cpu where appropriate.
4813         (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4814         appropriate.
4815         (rs6000_reassociation_width): Ditto.
4816         (rs6000_emit_epilogue): Ditto.
4817         (rs6000_adjust_cost): Ditto.
4818         (is_microcoded_insn): Ditto.
4819         (is_cracked_insn): Ditto.
4820         (rs6000_adjust_priority): Ditto.
4821         (rs6000_sched_reorder): Ditto.
4822         (rs6000_sched_reorder2): Ditto.
4823         (insn_must_be_first_in_group): Ditto.
4824         (insn_must_be_last_in_group): Ditto.
4825         (rs6000_register_move_cost): Ditto.
4826         * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4827         rs6000_cpu.
4829 2017-12-07  Julia Koval  <julia.koval@intel.com>
4831         * config.gcc: Add vaesintrin.h.
4832         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4833         New type.
4834         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4835         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4836         New builtins.
4837         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4838         * config/i386/immintrin.h: Include vaesintrin.h.
4839         * config/i386/sse.md (vaesdec_<mode>): New pattern.
4840         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4841         _mm_aesdec_epi128): New intrinsics.
4843 2017-12-06  David Malcolm  <dmalcolm@redhat.com>
4845         * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4846         * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4847         c-family/c-spellcheck.cc.
4848         (best_macro_match::best_macro_match): Likewise.
4849         * spellcheck-tree.h
4850         (struct edit_distance_traits<cpp_hashnode *>): Move to
4851         c-family/c-spellcheck.h.
4852         (class best_macro_match): Likewise.
4854 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4856         PR tree-optimization/83293
4857         * gimple-ssa-strength-reduction.c (insert_initializers): Use
4858         GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4859         might insert into empty bb.
4861         PR sanitizer/81281
4862         * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4863         simplify for plus with :c added, and pointer_plus without that.
4864         ((T)P - (T)(P + A) -> -(T) A): Likewise.  If type is integral
4865         with undefined overflow and the conversion is not widening,
4866         perform negation in utype and only convert to type afterwards.
4867         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4868         simplify for plus with :c added, and pointer_plus without that.
4869         If type is integral with undefined overflow and the conversion is
4870         not widening, perform minus in utype and only convert to type
4871         afterwards.  Move the last pointer_diff_expr simplify into the
4872         two outermost ifs.
4874 2017-12-06  Martin Sebor  <msebor@redhat.com>
4876         PR tree-optimization/82646
4877         * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4878         strncpy, not maxlen.
4880 2017-12-06  Martin Sebor  <msebor@redhat.com>
4882         * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4883         nonstring.
4885         PR tree-optimization/83075
4886         * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4887         strncat/strncpy don't change length of source string.
4889 2017-12-06  Eric Botcazou  <ebotcazou@adacore.com>
4891         Revert
4892         2017-11-29  Martin Aberg  <maberg@gaisler.com>
4894         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4895         to prevent b2bst errata sequence.
4896         (sqrtdf2_fix): Likewise.
4898 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4900         PR tree-optimization/81945
4901         * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4902         * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4903         to dest_cfun has orig_loop_num set, either remap it to the new
4904         loop number if the loop got moved too, or clear it.
4906 2017-12-05  Steve Ellcey  <sellcey@cavium.com>
4908         * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4909         to reservation.
4910         (thunderx2t99_nothing): New insn reservation.
4911         (thunderx2t99_mrs): New insn reservation.
4912         (thunderx2t99_multiple): New insn reservation.
4913         (thunderx2t99_alu_basi): Add bfx to reservation.
4914         (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
4916 2017-12-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4918         PR target/82248
4919         * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
4921 2017-12-05  Bin Cheng  <bin.cheng@arm.com>
4923         * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
4924         tree-ssa-pre.c::remove_dead_inserted_code.
4925         * tree-ssa-dce.h: New file.
4926         * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
4927         (remove_dead_inserted_code): Move and rename to function
4928         tree-ssa-dce.c::simple_dce_from_worklist.
4929         (pass_pre::execute): Update use.
4931 2017-12-05  Richard Biener  <rguenther@suse.de>
4933         PR tree-optimization/83277
4934         * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
4935         to code-gen liveout vars.
4937 2017-12-05  Richard Sandiford  <richard.sandiford@linaro.org>
4939         * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
4940         (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
4941         (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
4942         split condition.
4944 2017-12-05  Max Filippov  <jcmvbkbc@gmail.com>
4946         * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
4947         function.
4948         (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
4949         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
4950         ASAN is enabled.
4952 2017-12-05  Richard Biener  <rguenther@suse.de>
4954         * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
4955         * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
4956         (pass_data_cse_sincos): Use TV_TREE_SINCOS.
4957         (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
4959 2017-12-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4961         * dbxout.c (dbxout_block): Grow buf to 30 bytes.
4963 2017-12-05  Martin Liska  <mliska@suse.cz>
4964             Jakub Jelinek  <jakub@redhat.com>
4966         * doc/invoke.texi: Document the options.
4967         * flag-types.h (enum sanitize_code): Add
4968         SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4969         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
4970         of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4971         * opts.c: Define new sanitizer options.
4972         * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
4973         (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
4975 2017-12-05  Julia Koval  <julia.koval@intel.com>
4977         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
4978         OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
4979         (ix86_handle_option): Handle -mavx512vnni.
4980         * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
4981         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
4982         * config/i386/i386-c (__AVX512VNNI__): New.
4983         * config/i386/i386.c (ix86_target_string): Handle new option.
4984         (ix86_valid_target_attribute_inner_p): Handle new option.
4985         * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
4986         * config/i386/i386.opt (mavx512vnni): New option.
4988 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4990         PR target/81616
4991         * athlon.md: Disable for generic.
4992         * haswell.md: Enable for generic.
4993         * i386.c (ix86_sched_init_global): Add core hooks for generic.
4994         * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
4995         to 4.
4996         (ix86_adjust_cost): Move generic to haswell path.
4998 2017-12-04  Eric Botcazou  <ebotcazou@adacore.com>
5000         * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
5001         instead of MEM_P in a couple more places.  Fix formatting issues.
5003 2017-12-04  Jim Wilson  <jimw@sifive.com>
5005         * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
5006         instead of GP_REG_LAST-1.
5007         (riscv_adjust_libcall_cfi_prologue): Likewise.
5008         (riscv_adjust_libcall_cri_epilogue): Likewise.
5009         * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
5010         comment.
5012 2017-12-04  Luis Machado  <luis.machado@linaro.org>
5014         * ipa-pure-const.c (check_decl): Add missing newline.
5015         (state_from_flags): Likewise.
5017 2017-12-04  Jeff Law  <law@redhat.com>
5019         PR tree-optimizatin/78496
5020         * gimple-ssa-evrp-analyze.h
5021         (evrp_range_analyzer::get_vr_values): Simplify.
5022         * gimple-ssa-evrp-analyze.c: Corresponding changes.
5023         * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
5024         and gimple-ssa-evrp-analyze.h.
5025         (dom_opt_dom_walker class): Add evrp_range_analyzer member.
5026         (simplify_stmt_for_jump_threading): Copy a blob of code from
5027         tree-vrp.c to use ranges to simplify statements.
5028         (dom_opt_dom_walker::before_dom_children): Call
5029         evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
5030         (dom_opt_dom_walker::after_dom_children): Similarly for
5031         evrp_range_analyzer::leave.
5032         (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
5033         conditionals.
5035         * gimple-ssa-evrp-analyze.c
5036         (evrp_range_analyzer::extract_range_from_stmt):  Always use
5037         vr_values::update_value_range so preexisting range info is
5038         medged with new range info, even if the new range is VR_VARYING.
5040 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5042         * combine.c: Adjust comment.
5043         (use_crosses_set_p): Delete.
5044         (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
5045         (try_combine): Ditto.
5047 2017-12-04  Richard Biener  <rguenther@suse.de>
5049         PR tree-optimization/83255
5050         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
5051         Re-add zero-iteration check.
5053 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5055         PR rtl-optimization/83245
5056         * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
5057         hard registers as earlyclobber, also if not in an asm.
5059 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5061         PR bootstrap/83265
5062         Revert
5063         2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5065         PR target/43871
5066         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5067         rs6000_cpu based on cpu_index, not tune_index.
5069 2017-12-04  Richard Biener  <rguenther@suse.de>
5071         PR tree-optimization/83238
5072         * graphite-scop-detection.c (scop_detection::merge_sese): Make
5073         code match comment, rejecting invalid SESE regions.
5075 2017-12-03  John David Anglin  <danglin@gcc.gnu.org>
5077         * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
5078         require base operand is a REG_POINTER prior to reload on targets
5079         with non-equivalent space registers.
5081 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5083         * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
5084         (update_specialized_profile): Fix updating of counts.
5085         (perhaps_add_new_callers): Likewise.
5087 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5089         PR target/81616
5090         * x86-tune.def: Remove obsolette FIXMEs.
5091         (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
5092         (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
5093         X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
5094         Enable for generic.
5095         (X86_TUNE_PAD_RETURNS): Disable for generic.
5097 2017-12-02  Jakub Jelinek  <jakub@redhat.com>
5099         PR tree-optimization/83170
5100         PR tree-optimization/83241
5101         * gimple-ssa-store-merging.c
5102         (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
5103         gimple_vuse (ins_stmt) in case it has changed.
5104         (imm_store_chain_info::output_merged_store): Likewise.
5106         * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
5107         POINTER_DIFF_EXPR.
5109         PR c++/81212
5110         * tree-cfg.c (pass_warn_function_return::execute): Handle
5111         __builtin_ubsan_handle_missing_return like __builtin_unreachable
5112         with BUILTINS_LOCATION.
5114         PR target/78643
5115         PR target/80583
5116         * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
5117         is BLKmode for vector field with vector raw mode, use TYPE_MODE
5118         instead of DECL_MODE.
5120         * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
5121         last argument to rtx pointer.
5122         * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
5123         with OPERANDS.  For AVX+ 128-bit VEX encoded instructions over 256-bit
5124         or 512-bit.  If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
5125         depending on the chosen ISAs.
5126         * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
5127         *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
5128         *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
5129         callers.
5130         * config/i386/sse.md (mov<mode>_internal): Likewise.
5131         * config/i386/mmx.md (*mov<mode>_internal): Likewise.
5133 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5135         * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
5137 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5139         * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
5140         parameters from prototype.
5141         * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
5142         parameters.  Don't print a comment.
5143         (emit_fusion_gpr_load): Adjust.
5144         (emit_fusion_load_store): Adjust.
5145         * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
5146         * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
5147         comment on the second line.
5149 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5151         PR target/43871
5152         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5153         rs6000_cpu based on cpu_index, not tune_index.
5155 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5157         * final.c (output_asm_name): Print insn_cost.  Shorten output.  Print
5158         which_alternative instead of which_alternative + 1.
5159         (output_asm_insn): Print an extra tab if the template is short.
5161 2017-12-01  Jim Wilson  <jimw@sifive.com>
5163         * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
5164         comment.
5165         * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
5166         * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
5167         reference.
5168         * doc/tm.texi: Regenerate.
5170 2017-12-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
5172         PR target/81959
5173         * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
5174         whether we can allocate pseudos before trying to fix an address.
5175         * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
5176         memory address is indexed or indirect.
5177         (floatuns_<mode>si2_hw2): Likewise.
5179 2017-12-01  Jason Merrill  <jason@redhat.com>
5181         * Makefile.in (TAGS): Add c-family/*.cc.
5183 2017-12-01  Wilco Dijkstra  <wdijkstr@arm.com>
5185         * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
5186         (call_value_insn): Likewise.
5187         (sibcall_insn): Likewise.
5188         (sibcall_value_insn): Likewise.
5189         (movsi_aarch64): Likewise.
5190         (movdi_aarch64): Likewise.
5191         (add_losym_): Likewise.
5192         (ldr_got_small_): Likewise.
5193         (ldr_got_small_sidi): Likewise.
5194         (ldr_got_small_28k_): Likewise.
5195         (ldr_got_small_28k_sidi): Likewise.
5196         * config/aarch64/aarch64.c (aarch64_print_address_internal):
5197         Move output_addr_const to symbolic case. Add error check.
5199 2017-12-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5201         * config/s390/predicates.md (plus16_Q_operand): New predicate.
5202         * config/s390/s390.md: Disable MVC merging peephole if it would
5203         disable operand forwarding.
5204         (new peephole2): Split MVCs if it would turn them into up to 2
5205         forwardable MVCs.
5207 2017-12-01  Richard Biener  <rguenther@suse.de>
5209         PR tree-optimization/83232
5210         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
5211         detection of same access. Instead of breaking the group here
5212         do not consider the duplicate.  Add comment explaining real fix.
5214 2017-12-01  Jonathan Wakely  <jwakely@redhat.com>
5216         * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
5218 2017-12-01  Sudakshina Das  <sudi.das@arm.com>
5220         * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
5222 2017-12-01  Jakub Jelinek  <jakub@redhat.com>
5224         * function.h (struct function): Remove cilk_frame_decl,
5225         is_cilk_function and calls_cilk_spawn fields.
5226         * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
5227         field.
5228         * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
5229         cilk_elemental field.
5230         * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
5231         * target.def: Adjust comment.
5232         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5233         Don't test cilk_elemental.
5235         PR tree-optimization/83233
5236         * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
5237         bswap_stat name for the struct.
5239         PR c/79153
5240         * tree.h (SWITCH_BREAK_LABEL_P): Define.
5241         * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5242         starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5243         SWITCH_BREAK_LABEL_P set on the label.
5244         (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5245         added for default case if it was missing and not all cases covered.
5246         Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5247         switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5248         set on the label.
5249         * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5250         to avoid -Wimplicit-fallthrough warning.
5251         * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5252         FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5254         PR tree-optimization/83221
5255         * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5256         down by 16.
5257         (init_reassoc): Formatting fix.
5259         PR sanitizer/81275
5260         * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5261         C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5262         is diagnosed.
5264         PR sanitizer/83219
5265         * tree-cfg.c: Include asan.h.
5266         (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5268 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5270         * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5272 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5274         * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5275         vpcmpeqd instruction.
5277 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5279         * config/i386/i386.c (standard_sse_constant_opcode): Fix
5280         registers type for 128bit mode.
5282 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5284         * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5285         of ;;.
5286         * gengtype-state.c (read_state_pair): Likewise.
5287         * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5288         * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5289         * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5290         * ifcvt.c (noce_try_store_flag_constants): Likewise.
5291         * tree-ssa-ccp.c (ccp_finalize): Likewise.
5292         * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5293         * builtins.c (fold_builtin_3): Likewise.
5294         * graphite-scop-detection.c
5295         (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5296         * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5297         Likewise.
5299 2017-12-01  Maxim Ostapenko  <m.ostapenko@samsung.com>
5301         PR sanitizer/81697
5302         * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5303         parameter. Return true if ignore_decl_rtl_set_p is true and other
5304         conditions are satisfied.
5305         * asan.h (asan_protect_global): Add new parameter.
5306         * varasm.c (categorize_decl_for_section): Pass true as second parameter
5307         to asan_protect_global calls.
5309 2017-11-30  Jim Wilson  <jimw@sifive.com>
5311         * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5312         -mno-memcpy options.  For -mplt, -mfdiv, -mdiv, -msave-restore, and
5313         -mstrict-align, add info on default value.  Delete redundant lines for
5314         -mabi.  Add missing -mexplicit-relocs docs.
5316 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5318         * config/arc/arc.md (trap): New pattern.
5320 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5322         * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5323         instruction to end into a delay slot.
5324         * config/arc/arc.md (cond_delay_insn): Check if the instruction
5325         can be placed into a delay slot against reg_note.
5327 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5329         * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5330         labels number of usages.
5332 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5334         * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5335         function.
5336         (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5338 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5340         PR target/83210
5341         * internal-fn.c (expand_mul_overflow): Optimize unsigned
5342         multiplication by power of 2 constant into two shifts + comparison.
5344 2017-11-30  Jan Hubicka  <hubicka@ucw.cz>
5346         PR target/81616
5347         * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5349 2017-11-30  Richard Biener  <rguenther@suse.de>
5351         PR tree-optimization/83202
5352         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5353         allow_peel argument and guard peeling.
5354         (canonicalize_loop_induction_variables): Likewise.
5355         (canonicalize_induction_variables): Pass false.
5356         (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5357         peeling from cunrolli.
5359 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5361         * combine.c (try_combine): Print a message to dump file whenever
5362         I0, I1, or I2 cannot be combined into I3.
5364 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5366         PR rtl-optimization/83156
5367         PR rtl-optimization/82621
5368         * combine.c (try_combine): Don't split an I2 if one of the dests is
5369         set again before I3.  Allow unused dests.
5371 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5373         * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5375 2017-11-29  Vladimir Makarov  <vmakarov@redhat.com>
5377         PR rtl-optimization/80818
5378         * lra.c (collect_non_operand_hard_regs): New arg insn.  Pass it
5379         recursively.  Use insn code for clobber.
5380         (lra_set_insn_recog_data): Pass the new arg to
5381         collect_non_operand_hard_regs.
5382         (add_regs_to_insn_regno_info): Pass insn instead of uid.  Use insn
5383         code for clobber.
5384         (lra_update_insn_regno_info): Pass insn to
5385         add_regs_to_insn_regno_info.
5387 2017-11-29  Jim Wilson  <jimw@sifive.com>
5388             Andrew Waterman  <andrew@sifive.com>
5390         * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5391         (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5392         SINGLE_SHIFT_COST.  Case LT and ZERO_EXTEND, likewise.  Case ASHIFT,
5393         use SINGLE_SHIFT_COST.
5394         * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5395         (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5397 2017-11-29  Julia Koval  <julia.koval@intel.com>
5399         * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5400         _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5401         _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5402         _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5403         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5404         _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5405         _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5406         _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5407         _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5408         _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5409         _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5410         * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5411         __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5412         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5413         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5414         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5415         __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5416         __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5417         __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5418         __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5419         __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5420         __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5421         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5422         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5423         __builtin_ia32_vpshldv_v2di_mask,
5424         __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5425         * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5426         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5428 2017-11-29  Julia Koval  <julia.koval@intel.com>
5430         * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5431         _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5432         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5433         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5434         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5435         _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5436         _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5437         _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5438         _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5439         _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5440         _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5441         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5442         V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5443         V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5444         V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5445         V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5446         V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5447         V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5448         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5449         * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5450         vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5452 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5454         * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5455         movsi_pic_gotdata_op instruction as a load for the UT699 errata
5456         workaround.
5458 2017-11-29  Martin Aberg  <maberg@gaisler.com>
5460         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5461         to prevent b2bst errata sequence.
5462         (sqrtdf2_fix): Likewise.
5464 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5466         * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5467         (div_sqrt_insn_p): New function.
5468         (sparc_do_work_around_errata): Insert NOP instructions to
5469         prevent sequences that could trigger the TN-0013 errata for
5470         certain LEON3 processors.
5471         (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5472         (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5473         * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5474         (in_branch_delay): Prevent div and sqrt in delay slot if
5475         fix_lost_divsqrt.
5476         * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5478 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5480         * config/sparc/sparc.c (atomic_insn_p): New function.
5481         (sparc_do_work_around_errata): Insert NOP instructions to
5482         prevent sequences that could trigger the TN-0010 errata for
5483         UT700.
5484         * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5485         instruction referable in atomic_insns_p.
5487 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5489         * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5490         (atomic_compare_and_swap_leon3_1): Likewise.
5491         (ldstub): Likewise.
5493 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5495         * config/sparc/sparc.c (fpop_insn_p): New function.
5496         (sparc_do_work_around_errata): Insert NOP instructions to
5497         prevent sequences that could trigger the TN-0012 errata for
5498         GR712RC.
5499         (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5500         * config/sparc/sparc.md (fix_gr712rc): New attribute.
5501         (in_branch_annul_delay): Prevent floating-point instructions
5502         in delay slot of annulled integer branch.
5504 2017-11-29  Richard Biener  <rguenther@suse.de>
5506         PR tree-optimization/83202
5507         * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5508         (bst_fail): Use it.
5509         (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5510         nodes vectorized to the same stmts multiple times.
5511         (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5512         (vect_analyze_slp_instance): Adjust.
5513         (scalar_stmts_to_slp_tree_map_t): New typedef.
5514         (vect_schedule_slp_instance): Add a map recording the SLP node
5515         representing the vectorized stmts for a set of scalar stmts.
5516         Avoid code-generating redundancies.
5517         (vect_schedule_slp): Allocate map and pass it down.
5519 2017-11-29  Nathan Sidwell  <nathan@acm.org>
5521         PR c++/83187
5522         * tree.c (build_complex_type): Fix canonicalization.  Only fill in
5523         type if it is new.
5525 2017-11-29  Wilco Dijkstra  <wdijkstr@arm.com>
5527         * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5528         cases for printing LDP/STP memory addresses.
5529         (aarch64_print_address_internal): Renamed from
5530         aarch64_print_operand_address, added parameter, add Pmode check.
5531         (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5532         (aarch64_print_operand_address): Indirect to
5533         aarch64_print_address_internal.
5534         * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5535         'y' operand output specifier.
5537 2017-11-29  Jakub Jelinek  <jakub@redhat.com>
5539         PR middle-end/83185
5540         * tree.c (build_simple_mem_ref_loc): Handle
5541         get_addr_base_and_unit_offset returning a MEM_REF.
5543         PR middle-end/80929
5544         * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5546         PR target/80819
5547         * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5548         alternative.
5550 2017-11-28  David Malcolm  <dmalcolm@redhat.com>
5552         * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5553         m_x_offset.
5554         (layout::move_to_column): Likewise.
5556 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5558         PR sanitizer/81275
5559         * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5560         is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5562 2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5563             Martin Jambor  <mjambor@suse.cz>
5565         PR ipa/82808
5566         * tree.h (expr_type_first_operand_type_p): Declare
5567         * tree.c (expr_type_first_operand_type_p): New function.
5568         * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5569         (ipa_value_from_jfunc): Adjust declaration.
5570         * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5571         Use it as result type for arithmetics, unless it is NULL in which case
5572         be more conservative.
5573         (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5574         ipa_get_jf_pass_through_result.
5575         (propagate_vals_across_pass_through): Likewise.
5576         (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5577         is to propagate_vals_across_pass_through.
5578         (propagate_constants_across_call): Pass PARM_TYPE to
5579         propagate_scalar_across_jump_function.
5580         (find_more_scalar_values_for_callers_subset): Pass parameter type to
5581         ipa_value_from_jfunc.
5582         (cgraph_edge_brings_all_scalars_for_node): Likewise.
5583         * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5584         to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5585         * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5586         target_type, pass it to ipa_value_from_jfunc.
5587         (update_indirect_edges_after_inlining): Pass parameter type to
5588         try_make_edge_direct_simple_call.
5590 2017-11-28  Jeff Law  <law@redhat.com>
5592         * gimple-ssa-evrp-analyze.c
5593         (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5594         refine ranges if scev_initialized_p returns true.
5595         * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5597 2017-11-28  Julia Koval  <julia.koval@intel.com>
5599         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5600         _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5601         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5602         _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5603         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5604         _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5605         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5606         _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5607         _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5608         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5609         _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5610         * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5611         __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5612         __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5613         __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5614         __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5615         __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5616         __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5617         __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5618         __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5619         __builtin_ia32_vpshrd_v2di_mask): New builtins.
5620         * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5622 2017-11-28  Julia Koval  <julia.koval@intel.com>
5624         * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5625         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5626         _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5627         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5628         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5629         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5630         _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5631         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5632         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5633         _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5634         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5635         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5636         V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5637         V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5638         V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5639         V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5640         V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5641         V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5642         V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5643         V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5644         * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5645         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5646         __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5647         __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5648         __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5649         __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5650         __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5651         __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5652         __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5653         __builtin_ia32_vpshld_v2di_mask): New builtins.
5654         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5655         * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5657 2017-11-28  Richard Biener  <rguenther@suse.de>
5659         PR tree-optimization/80776
5660         * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5661         Declare.
5662         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5663         New function.
5664         (evrp_range_analyzer::record_ranges_from_incoming_edges):
5665         If the incoming edge is an effective fallthru because the other
5666         edge only reaches a __builtin_unreachable () then record ranges
5667         derived from the controlling condition in SSA info.
5668         (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5669         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5671 2017-11-28  Olivier Hainque  <hainque@adacore.com>
5673         * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5674         on mingw build hosts.
5676 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5678         PR debug/81307
5679         * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5681 2017-11-28  Richard Biener  <rguenther@suse.de>
5683         PR middle-end/83141
5684         * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5685         copies generated from memcpy use a character array as reference
5686         type.
5688 2017-11-28  Julia Koval  <julia.koval@intel.com>
5689             Sebastian Peryt  <sebastian.peryt@intel.com>
5691         * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5692         c-family/c-cilkplus.o, c-family/array-notation-common.o,
5693         cilk-common.o, cilk.h, cilk-common.c): Remove.
5694         * builtin-types.def
5695         (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5696         * builtins.c (is_builtin_name): Remove cilkplus condition.
5697         (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5698         * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5699         cilk-builtins.def, cilkplus.def): Remove.
5700         * cif-code.def (CILK_SPAWN): Remove.
5701         * cilk-builtins.def: Delete.
5702         * cilk-common.c: Ditto.
5703         * cilk.h: Ditto.
5704         * cilkplus.def: Ditto.
5705         * config/darwin.h (fcilkplus): Delete.
5706         * cppbuiltin.c: Ditto.
5707         * doc/extend.texi: Remove cilkplus doc.
5708         * doc/generic.texi: Ditto.
5709         * doc/invoke.texi: Ditto.
5710         * doc/passes.texi: Ditto.
5711         * gcc.c (fcilkplus): Remove.
5712         * gengtype.c (cilk.h): Remove.
5713         * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5714         support.
5715         * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5716         Remove.
5717         * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5718         gimplify_call_expr,
5719         is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5720         gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5721         cilkplus conditions.
5722         * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5723         inline_read_section): Ditto.
5724         * ipa-inline-analysis.c (cilk.h): Remove.
5725         * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5726         * lto-wrapper.c (merge_and_complain, append_compiler_options,
5727         append_linker_options): Remove condition for fcilkplus.
5728         * lto/lto-lang.c (cilk.h): Remove.
5729         (lto_init): Remove condition for fcilkplus.
5730         * omp-expand.c (expand_cilk_for_call): Delete.
5731         (expand_omp_taskreg, expand_omp_for_static_chunk,
5732         expand_omp_for): Remove cilkplus
5733         conditions.
5734         (expand_cilk_for): Delete.
5735         * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5736         * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5737         execute_lower_omp, diagnose_sb_0): Ditto.
5738         * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5739         * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5740         * tree-nested.c: Ditto.
5741         * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5742         (dump_generic_node): Ditto.
5743         * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5744         * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5745         Delete.
5746         * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5748 2017-11-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5750         * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5751         * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5752         for vector mode and !TARGET_SIMD.
5754 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5756         * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5757         Adjust comment.
5758         * tree.h (SWITCH_LABELS): Remove.
5759         * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5760         assert SWITCH_BODY is non-NULL.
5761         * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5762         handling.
5763         * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5765         PR tree-optimization/80788
5766         * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5767         has TREE_OVERFLOW set, call drop_tree_overflow.
5769 2017-11-28  Richard Biener  <rguenther@suse.de>
5771         PR tree-optimization/83158
5772         * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5774 2017-11-28  Segher Boessenkool  <segher@kernel.crashing.org>
5776         PR 81288/target
5777         * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5778         TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.
5780 2017-11-27  Segher Boessenkool  <segher@kernel.crashing.org>
5782         * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5783         LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5785 2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5787         PR middle_end/82333
5788         * varasm.c (compare_constant): Take the mode of the constants into
5789         account when comparing floating point constants.
5791 2017-11-27  Gerald Pfeifer  <gerald@pfeifer.com>
5793         * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5794         from explicit instantiation of debug_helper.
5795         * vec.h (DEFINE_DEBUG_VEC): Ditto.
5797 2017-11-27  Richard Biener  <rguenther@suse.de>
5799         * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5800         refactor a bit.
5802 2017-11-27  Richard Biener  <rguenther@suse.de>
5804         * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5805         (type_hash_canon): Also clear int_cst_hash_table entry for
5806         TYPE_MIN/MAX_VALUE.
5807         (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5809 2017-11-27  Tamar Christina  <tamar.christina@arm.com>
5811         * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5812         * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5813         * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5815 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5817         * hash-map.h (gt_cleare_cache): Avoid UB.
5819 2017-11-27  Eric Botcazou  <ebotcazou@adacore.com>
5821         * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5822         * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5823         (decide_unroll_runtime_iterations): Likewise.
5824         (decide_unroll_stupid): Likewise.
5826 2017-11-27  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5828         PR target/83109
5829         * config/i386/i386.md: Add a loop with incssp.
5831 2017-11-27  Martin Jambor  <mjambor@suse.cz>
5833         PR tree-optimization/81248
5834         * tree-sra.c (splice_param_accesses): Remove size check.
5835         (decide_one_param_reduction): Fix size check.
5836         * gimple-pretty-print.c (dump_profile): Silence warning.
5837         * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5839 2017-11-27  Jakub Jelinek  <jakub@redhat.com>
5841         PR debug/81307
5842         * dbxout.c (lastlineno): New variable.
5843         (dbx_debug_hooks): Use dbxout_switch_text_section as
5844         switch_text_section debug hook.
5845         (dbxout_function_end): Switch to current_function_section
5846         rather than function_section.  If crtl->has_bb_partition,
5847         output just one N_FUN, depending on in_cold_section_p.
5848         (dbxout_source_line): Remember last lineno in lastlineno.
5849         (dbxout_switch_text_section): New function.
5850         (dbxout_function_decl): Adjust dbxout_block caller.
5851         (dbx_block_with_cold_children): New function.
5852         (dbxout_block): Return true if any LBRAC/RBRAC have been
5853         emitted.  Use dbx_block_with_cold_children at depth == 0
5854         in second partition.  Add PARENT_BLOCKNUM argument, pass
5855         it optionally adjusted to children.  Output LBRAC/RBRAC
5856         around recursive call only if the block is in the current
5857         partition, if not and anything was output, emit empty
5858         range LBRAC/RBRAC.
5859         * final.c (final_scan_insn): Compute cold_function_name
5860         before calling switch_text_section debug hook.  Call
5861         that hook even if dwarf2out_do_frame if not emitting
5862         dwarf debug info.
5864         PR target/83100
5865         * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5866         TREE_READONLY decls.
5868 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5870         PR rtl-optimization/82488
5871         * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5873 2017-11-26  Julia Koval  <julia.koval@intel.com>
5875         * config/i386/i386.c (processor_target_table): Add skylake_cost for
5876         skylake-avx512.
5877         * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5878         skylake_cost): New.
5880 2017-11-26  Julia Koval  <julia.koval@intel.com>
5882         * config/i386/driver-i386.c (host_detect_local_cpu):
5883         Detect skylake-avx512.
5885 2017-11-26  Julia Koval  <julia.koval@intel.com>
5887         * config.gcc: Add -march=cannonlake.
5888         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5889         * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5890         * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5891         (PTA_CANNONLAKE): New.
5892         (processor_target_table): Add cannonlake.
5893         (ix86_option_override_internal): Ditto.
5894         (fold_builtin_cpu): Ditto.
5895         (get_builtin_code_for_version): Handle cannonlake.
5896         (M_INTEL_COREI7_CANNONLAKE): New.
5897         * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5898         * doc/invoke.texi: Add -march=cannonlake.
5900 2017-11-14  Boris Kolpackov  <boris@codesynthesis.com>
5902         * plugin.c (add_new_plugin): Use platform-specific library extensions.
5903         (try_init_one_plugin): Alternative implementation for MinGW.
5904         * Makefile.in (plugin_implib): New.
5905         (gengtype-lex.c): Fix broken AIX workaround.
5906         * configure: Regenerate.
5907         * doc/plugins.texi: Document support for MinGW.
5909 2017-11-25  Jakub Jelinek  <jakub@redhat.com>
5911         PR rtl-optimization/81553
5912         * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
5913         to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
5914         is a shift where C1 has different mode than the whole shift, use C1's
5915         mode for MULT rather than the shift's mode.
5917         PR target/82848
5918         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
5919         builtins not enabled in the currently selected ISA.
5921 2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>
5923         PR tree-optimization/71026
5924         * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
5925         (insert_reciprocals): Change to insert reciprocals before a division
5926         by a square and to insert the square of a reciprocal.
5927         (execute_cse_reciprocals_1): Change to consider division by a square.
5928         (register_division_in): Add importance parameter.
5930 2017-11-24  Richard Biener  <rguenther@suse.de>
5932         PR tree-optimization/82402
5933         * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
5934         set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
5936 2017-11-24  Marc Glisse  <marc.glisse@inria.fr>
5938         * match.pd (0-ptr): New transformation.
5940 2017-11-24  Jan Hubicka  <hubicka@ucw.cz>
5942         PR bootstrap/83015
5943         * ipa-inline.c (inline_small_functions): Set current badnes correctly
5944         when skipping checking.
5946 2017-11-24  Richard Biener  <rguenther@suse.de>
5948         PR tree-optimization/83128
5949         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
5950         (vn_reference_lookup_3): Likewise.
5952 2017-11-24  Jakub Jelinek  <jakub@redhat.com>
5954         PR sanitizer/83014
5955         * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
5956         instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
5957         tree_to_uhwi twice.
5959         * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
5960         ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
5961         check for the latter with gimple_call_builtin_p.  Do not handle
5962         BUILT_IN_STPNCPY_CHK which is not a pass through call.
5964 2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>
5966         * config/arm/arm_neon.h: Fix pragma GCC push_options before
5967         vdot_u32.
5969 2017-11-23  Julia Koval  <julia.koval@intel.com>
5971         * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
5972         _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
5973         _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
5974         _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
5975         _mm512_maskz_expandloadu_epi16): New intrinsics.
5976         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
5977         _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
5978         _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
5979         _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
5980         _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
5981         _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
5982         _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
5983         _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
5984         _mm256_maskz_expandloadu_epi8): New intrinsics.
5985         * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
5986         V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
5987         V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
5988         V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
5989         * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
5990         * config/i386/sse.md (VI248_VLBW): New iterator.
5991         (expand<mode>_mask, expand<mode>_maskz): New patterns.
5993 2017-11-23  Julia Koval  <julia.koval@intel.com>
5995         * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
5996         * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
5997         _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
5998         _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
5999         _mm512_mask_compressstoreu_epi16): New.
6000         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
6001         _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
6002         _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
6003         _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
6004         _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
6005         _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
6006         _mm256_mask_compressstoreu_epi8): New.
6007         * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
6008         VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
6009         VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
6010         VOID_FTYPE_PV8HI_V8HI_UQI): New types.
6011         * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
6012         __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
6013         __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
6014         __builtin_ia32_compresshi128_mask,
6015         __builtin_ia32_compressstoreuqi512_mask,
6016         __builtin_ia32_compressstoreuhi512_mask,
6017         __builtin_ia32_compressstoreuqi256_mask,
6018         __builtin_ia32_compressstoreuqi128_mask,
6019         __builtin_ia32_compressstoreuhi256_mask,
6020         __builtin_ia32_compressstoreuhi128_mask): New builtins.
6021         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
6022         array for flags2.
6023         (ix86_expand_special_args_builtin): Handle new types.
6024         (s4fma_expand): Handle new builtin array.
6025         * config/i386/immintrin.h: Include new headers.
6026         * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
6027         (compress<mode>_mask, compressstore<mode>_mask): New patterns.
6029 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6031         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
6032         fixes.  Declare temp and g variables at the top in order to avoid
6033         {} in most of the cases.
6035 2017-11-23  Marc Glisse  <marc.glisse@inria.fr>
6037         * match.pd (ptr-0): New transformation.
6039 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
6041         * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
6042         (struct addr_mode_cost_table): New.
6043         (struct tune_params): Add field addr_mode_costs.
6044         * config/arm/arm.c (generic_addr_mode_costs): New.
6045         (arm_slowmul_tune): Initialise addr_mode_costs field.
6046         (arm_fastmul_tune): Likewise.
6047         (arm_strongarm_tune): Likewise.
6048         (arm_xscale_tune): Likewise.
6049         (arm_9e_tune): Likewise.
6050         (arm_marvell_pj4_tune): Likewise.
6051         (arm_v6t2_tune): Likewise.
6052         (arm_cortex_tune): Likewise.
6053         (arm_cortex_a8_tune): Likewise.
6054         (arm_cortex_a7_tune): Likewise.
6055         (arm_cortex_a15_tune): Likewise.
6056         (arm_cortex_a35_tune): Likewise.
6057         (arm_cortex_a53_tune): Likewise.
6058         (arm_cortex_a57_tune): Likewise.
6059         (arm_exynosm1_tune): Likewise.
6060         (arm_xgene1_tune): Likewise.
6061         (arm_cortex_a5_tune): Likewise.
6062         (arm_cortex_a9_tune): Likewise.
6063         (arm_cortex_a12_tune): Likewise.
6064         (arm_cortex_a73_tune): Likewise.
6065         (arm_v7m_tune): Likewise.
6066         (arm_cortex_m7_tune): Likewise.
6067         (arm_v6m_tune): Likewise.
6068         (arm_fa726te_tune): Likewise.
6069         (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
6071 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
6073         * config/arm/arm.c (arm_mem_costs): New function.
6074         (arm_rtx_costs_internal): Use arm_mem_costs.
6076 2017-11-23  Mark Wielaard  <mark@klomp.org>
6078         * dwarf2out.c (init_sections_and_labels): Use generation to create
6079         unique ranges_section_label and ranges_base_label. Return generation.
6080         (output_rnglists): Add generation argument. Use generation to create
6081         unique ranges labels.
6082         (dwarf2out_finish): Get generation from init_sections_and_labels
6083         and pass generation to output_rnglists.
6085 2017-11-23  Mike Stump  <mikestump@comcast.net>
6086             Eric Botcazou  <ebotcazou@adacore.com>
6088         * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
6089         * cfgloop.h (struct loop): Add unroll field.
6090         * function.h (struct function): Add has_unroll bitfield.
6091         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
6092         (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
6093         * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
6094         (pass_rtl_unroll_loops::gate): Likewise.
6095         * loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
6096         for which loop->unroll==1.
6097         (decide_unroll_constant_iterations): Use note for consistency and
6098         take loop->unroll into account.  Return early if loop->unroll is set.
6099         Fix thinko in existing test.
6100         (decide_unroll_runtime_iterations): Use note for consistency and
6101         take loop->unroll into account.
6102         (decide_unroll_stupid): Likewise.
6103         * lto-streamer-in.c (input_cfg): Read loop->unroll.
6104         * lto-streamer-out.c (output_cfg): Write loop->unroll.
6105         * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
6106         New case.
6107         (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
6108         (print_loop): Print loop->unroll if set.
6109         * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
6110         * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
6111         * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
6112         New case.
6113         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
6114         loop->unroll is set and smaller than the trip count.  Otherwise bypass
6115         entirely the heuristics if loop->unroll is set.  Remove dead note.
6116         Fix off-by-one bug in other note.
6117         (try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
6118         (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
6119         is greater than 1.
6120         (tree_unroll_loops_completely): Make static.
6121         (pass_complete_unroll::execute): Use correct type for variable.
6122         (pass_complete_unrolli::execute): Fix formatting.
6123         * tree.def (ANNOTATE_EXPR): Add 3rd operand.
6125 2017-11-23  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6127         * config/i386/i386.h (TARGET_PREFER_AVX256): Also
6128         enable when TARGET_PREFER_AVX128 is set.
6130 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6132         * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
6133         frequency.
6134         * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
6135         (cgraph_edge::maybe_hot_p): Use sreal frequencies.
6137 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6139         * ipa-fnsummary.c (record_modified_bb_info): Use sreal
6140         frequencies. Fix estimation of aggregate parameters.
6142 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6144         * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
6145         when inlining.
6147 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6149         * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
6151 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6153         * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
6154         macro.
6156 2017-11-23  Oleg Endo  <olegendo@gcc.gnu.org>
6158         PR target/83111
6159         * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
6160         sibcall_value_pcrel_fdpic): Use local variable instead of
6161         operands[3].
6162         (calli_tbr_rel): Add missing operand 2.
6163         (call_valuei_tbr_rel): Add missing operand 3.
6165 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6167         PR middle-end/82253
6168         * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
6169         bitpos/bitsize covering the whole destination, use store_expr only if
6170         the complex mode is the same.  Otherwise, use expand_normal and if
6171         it returns CONCAT, subreg each part separately instead of trying to
6172         subreg the whole result.
6174 2017-11-23  Richard Biener  <rguenther@suse.de>
6176         PR tree-optimization/23094
6177         * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
6178         come here from walking over backedges in the first iteration.
6179         (vn_reference_lookup_3): Skip clobbers that store the same value.
6181 2017-11-23  Richard Biener  <rguenther@suse.de>
6183         PR tree-optimization/81403
6184         * tree-ssa-pre.c (get_representative_for): Add parameter specifying
6185         a block we need a leader relative to.
6186         (phi_translate_1): For nary processing require a leader from
6187         get_representative_for given we run expression simplification
6188         using match-and-simplify.  Remove previous fix.
6190 2017-11-22  Jeff Law  <law@redhat.com>
6192         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
6193         Use new method allocate_value_range rather than accessing the
6194         vrp_value_range_pool data member directly.
6195         * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
6196         to use extract_range_from_stmt method to avoid need for
6197         extract_range_from_assignment method.
6198         (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
6199         method rather than setting values_propgated data member directly.
6200         * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
6201         and values propagated data members and extract_range_from_assignment
6202         method.  Reorder private data members to conform to standards.
6203         Add new methods set_lattice_propagation_complete and
6204         allocate_value_range.
6206 2017-11-22  Eric Botcazou  <ebotcazou@adacore.com>
6208         PR rtl-optimization/83030
6209         * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
6210         for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
6211         (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
6212         * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
6213         CROSSING_JUMP_P flag is already set before setting it.
6214         * cfgrtl.c (fixup_partition_crossing): Likewise.
6215         * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
6216         insn as useless.
6218 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6220         * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
6221         Handle the case where both arguments are using gen_const_vec_series.
6223 2017-11-22  David Malcolm  <dmalcolm@redhat.com>
6225         PR c++/62170
6226         * pretty-print.c (pp_format): Move quoting implementation to
6227         pp_begin_quote and pp_end_quote.  Update pp_format_decoder call
6228         to pass address of "quote" local.
6229         (pp_begin_quote): New function.
6230         (pp_end_quote): New function.
6231         * pretty-print.h (printer_fn): Convert penultimate param from bool
6232         to bool *.
6233         (pp_begin_quote): New decl.
6234         (pp_end_quote): New decl.
6235         * tree-diagnostic.c (default_tree_printer): Convert penultimate
6236         param from bool to bool *.
6237         * tree-diagnostic.h (default_tree_printer): Likewise.
6239 2017-11-22  Jeff Law  <law@redhat.com>
6241         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6242         Initialize vr_values.
6243         (evrp_range_analyzer::try_find_new_range): Call methods attached to
6244         vr_values via vr_values class instance rather than delegators.
6245         (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6246         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6247         (evrp_range_analyzer::push_value_range): Likewise.
6248         (evrp_range_analyzer::pop_value_range): Likewise.
6249         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6250         most delegators.  Those remaining are exposed as public interfaces.
6251         Make vr_values a pointer and private.
6252         (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6253         vr_values.
6254         (evrp_range_analyzer::get_vr_value): New method.
6255         * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6256         (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6257         (class evrp_dom_walker): Attach evrp_folder class, initialize
6258         it in the ctor.  Remove temporary delegators.
6259         (evrp_dom_walker::before_dom_children): Call methods in attached
6260         evrp_range_analyzer class via class instance pointer.  Use
6261         free value_range_constant_singleton to remove need for
6262         op_with_constant_singleton_value delegator method.  Do not
6263         create a vrp_prop class instance for every call!  Narrow
6264         scope of a couple variables.
6265         (evrp_dom_walker::cleanup): Call methods in attached
6266         evrp_range_analyzer class via class instance pointer.
6267         * vr-values.h (class vr_values): Privatize many methods and
6268         data members.
6270 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6272         * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6273         * cfgexpand.c (expand_debug_expr): Remove handling for them.
6274         * expr.c (expand_expr_real_2): Likewise.
6275         * fold-const.c (const_unop): Likewise.
6276         * optabs-tree.c (optab_for_tree_code): Likewise.
6277         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6278         * tree-inline.c (estimate_operator_cost): Likewise.
6279         * tree-pretty-print.c (dump_generic_node): Likewise.
6280         (op_code_prio): Likewise.
6281         (op_symbol_code): Likewise.
6282         * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6283         (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6284         * internal-fn.c (direct_internal_fn_optab): New function.
6285         (direct_internal_fn_array, direct_internal_fn_supported_p
6286         (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6287         * fold-const-call.c (fold_const_reduction): New function.
6288         (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6289         CFN_REDUC_MIN.
6290         * tree-vect-loop.c: Include internal-fn.h.
6291         (reduction_code_for_scalar_code): Rename to...
6292         (reduction_fn_for_scalar_code): ...this and return an internal
6293         function.
6294         (vect_model_reduction_cost): Take an internal_fn rather than
6295         a tree_code.
6296         (vect_create_epilog_for_reduction): Likewise.  Build calls rather
6297         than assignments.
6298         (vectorizable_reduction): Use internal functions rather than tree
6299         codes for the reduction operation.  Update calls to the functions
6300         above.
6301         * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6302         Use calls to internal functions rather than REDUC tree codes.
6303         * config/aarch64/aarch64-simd.md: Update comment accordingly.
6305 2017-11-22  Olivier Hainque  <hainque@adacore.com>
6307         * config/vxworks.c (vxworks_override_options): Pick default
6308         dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6309         * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6310         DWARF_GNAT_ENCODINGS_DEFAULT.
6311         * config/vxworksae.h: Likewise.
6313 2017-11-22  Marc Glisse  <marc.glisse@inria.fr>
6315         PR tree-optimization/83104
6316         * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6317         not result.
6319 2017-11-22  Marek Polacek  <polacek@redhat.com>
6320             H.J. Lu  <hongjiu.lu@intel.com>
6321             Jason Merrill  <jason@redhat.com>
6323         PR c++/60336
6324         PR middle-end/67239
6325         PR target/68355
6326         * calls.c (initialize_argument_information): Call
6327         warn_parameter_passing_abi target hook.
6328         (store_one_arg): Use 0 for empty record size.  Don't push 0 size
6329         argument onto stack.
6330         (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6331         * common.opt: Update -fabi-version description.
6332         * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6333         (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6334         int_size_in_bytes.
6335         (ix86_is_empty_record): New function.
6336         (ix86_warn_parameter_passing_abi): New function.
6337         (TARGET_EMPTY_RECORD_P): Redefine.
6338         (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6339         * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6340         * doc/tm.texi: Regenerated.
6341         * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6342         TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6343         * dwarf2out.c (get_ultimate_context): Move to tree.c.
6344         * explow.c (hard_function_value): Call arg_int_size_in_bytes
6345         instead of int_size_in_bytes.
6346         * expr.c (copy_blkmode_to_reg): Likewise.
6347         * function.c (aggregate_value_p): Return 0 for empty types.
6348         (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6349         target hook.
6350         (locate_and_pad_parm): Call arg size_in_bytes instead
6351         size_in_bytes.
6352         * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6353         * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6354         * target.def (empty_record_p, warn_parameter_passing_abi): New target
6355         hooks.
6356         * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6357         (std_gimplify_va_arg_expr): Skip empty records.  Call
6358         arg_size_in_bytes instead size_in_bytes.
6359         * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6360         * tree-core.h (tree_type_common): Add empty_flag.
6361         (tree_decl_common): Update comments.
6362         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6363         DECL_PADDING_P.
6364         (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6365         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6366         DECL_PADDING_P.
6367         (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6368         * tree.c (default_is_empty_type): New function.
6369         (default_is_empty_record): New function.
6370         (arg_int_size_in_bytes): New function.
6371         (arg_size_in_bytes): New function.
6372         (get_ultimate_context): New function.
6373         * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6374         TRANSLATION_UNIT_WARN_EMPTY_P.
6375         (default_is_empty_record, arg_int_size_in_bytes,
6376         arg_size_in_bytes, get_ultimate_context): Declare.
6378 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6380         * config/arm/arm.c (cmse_clear_registers): New function.
6381         (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6382         code by call to cmse_clear_registers.
6383         (cmse_nonsecure_entry_clear_before_return): Likewise.
6385 2017-11-22  Tamar Christina  <tamar.christina@arm.com>
6387         * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6388         (vdot_s32, vdotq_s32): New.
6389         (vdot_lane_u32, vdotq_lane_u32): New.
6390         (vdot_lane_s32, vdotq_lane_s32): New.
6393 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6395         PR middle-end/82547
6396         * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6397         for unsigned values with fewer HWIs than the precision.
6398         (test_overflow): New function.
6399         (wide_int_cc_tests): Call it.
6401 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6403         * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6404         has the same mode class as Pmode.
6406 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6408         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6409         padding_bits_to_clear_ptr.
6410         (cmse_nonsecure_entry_clear_before_return): Likewise.
6412 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6414         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6415         auto_sbitap instead of integer bitfield to control register needing
6416         clearing.
6418 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6420         PR tree-optimization/83044
6421         * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6422         INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6423         that need the upper bound.  Subtract offset from
6424         get_addr_base_and_unit_offset only if positive and subtract it
6425         before division by eltsize rather than after it.
6427         PR debug/83084
6428         * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6429         debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6430         (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6431         asm.
6433         PR middle-end/82875
6434         * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6435         expand_binop with *mul_widen_optab, make sure at least one of the
6436         operands doesn't have VOIDmode.
6438         PR debug/83034
6439         * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6441         PR rtl-optimization/82044
6442         PR tree-optimization/82042
6443         * dse.c (record_store): Check for overflow.
6444         (check_mem_read_rtx): Properly check for overflow if width == -1, call
6445         add_wild_read instead of clear_rhs_from_active_local_stores on
6446         overflow and log it into dump_file.
6448 2017-11-22  Richard Biener  <rguenther@suse.de>
6450         * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6451         fake edges to exit when looking for a place to insert.
6452         * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6453         and remove.
6454         (insert_into_preds_of_block): Commit edge insertion immediately,
6455         assert that doesn't require new BBs.
6456         (fini_pre): Release expressions.
6457         (pass_pre::execute): Shuffle things around a bit, if the fn
6458         is too large do not compute AVAIL either as this is really the
6459         quadratic bit.
6461 2017-11-22  Richard Biener  <rguenther@suse.de>
6463         PR tree-optimization/83089
6464         * tree-if-conv.c (pass_if_conversion::execute): If anything
6465         changed reset SCEV and free the number of iteration estimates.
6467 2017-11-21  Martin Sebor  <msebor@redhat.com>
6469         PR tree-optimization/82945
6470         * calls.h (warn_nonstring_bound): Remove unused function.
6472 2017-11-21  Martin Sebor  <msebor@redhat.com>
6474         PR tree-optimization/82945
6475         * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6476         * calls.h (maybe_warn_nonstring_arg): Declare new function.
6477         * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6478         functions.
6479         (initialize_argument_information): Call maybe_warn_nonstring_arg.
6480         * calls.h (get_attr_nonstring_decl): Declare new function.
6481         * doc/extend.texi (attribute nonstring): Update.
6482         * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6483         get_attr_nonstring_decl and handle it.
6484         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same.  Improve
6485         detection of nul-termination.
6486         (strlen_to_stridx): Change to a pointer.
6487         (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6488         (pass_strlen::execute): Same.
6490 2017-11-21  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6492         * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6493         for the new option -mprefer-vector-width=[none|128|256|512].
6494         * config/i386/i386.c (ix86_target_string): remove old style options
6495         -mprefer-avx256 and make -mprefer-avx128 as alias.
6496         (ix86_option_override_internal):  Apply defaults for the
6497         -mprefer-vector-width=[128|256] option.
6498         * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6499         Implement macros to work with -mprefer-vector-width=.
6500         * config/i386/i386.opt: Implemented option
6501         -mprefer-vector-width=[none|128|256|512].
6502         * doc/invoke.texi: Documentation for
6503         -mprefer-vector-width=[none|128|256|512].
6505 2017-11-21  Pat Haugen  <pthaugen@us.ibm.com>
6507         * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6509 2017-11-21  Jonathan Wakely  <jwakely@redhat.com>
6511         * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6513 2017-11-21  Marc Glisse  <marc.glisse@inria.fr>
6515         * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6516         POINTER_PLUS_EXPR.
6517         * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6518         * expr.c (expand_expr_real_2): Likewise.
6519         * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6520         fold_binary_loc): Likewise.
6521         * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6522         P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6523         (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6524         MINUS_EXPR transformations.
6525         * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6526         * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6527         * tree-inline.c (estimate_operator_cost): Likewise.
6528         * tree-pretty-print.c (dump_generic_node, op_code_prio,
6529         op_symbol_code): Likewise.
6530         * tree-vect-stmts.c (vectorizable_operation): Likewise.
6531         * vr-values.c (extract_range_from_binary_expr): Likewise.
6532         * varasm.c (initializer_constant_valid_p_1): Likewise.
6533         * tree.def: New tree code POINTER_DIFF_EXPR.
6535 2017-11-21  Uros Bizjak  <ubizjak@gmail.com>
6537         * config/i386/i386.md (*bswap<mode>2_movbe): Add
6538         integer suffix to movbe mnemonic.
6539         (*bswaphi2_movbe): Ditto.
6540         (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6542 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6544         PR c++/83045
6545         * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6546         Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6547         Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6549 2017-11-21  Martin Liska  <mliska@suse.cz>
6551         * tree-inline.c (expand_call_inline): Remove not needed
6552         xstrdup_for_dump.
6554 2017-11-21  James Cowgill  <James.Cowgill@imgtec.com>
6555             Jakub Jelinek  <jakub@redhat.com>
6557         PR target/82880
6558         * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6559         Remove static keyword from f variable.
6561 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6563         PR tree-optimization/83086
6564         * gimple-ssa-store-merging.c
6565         (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6566         rather than n.base_addr.
6568 2017-11-21  Martin Liska  <mliska@suse.cz>
6570         PR rtl-optimization/82044
6571         PR tree-optimization/82042
6572         * dse.c (check_mem_read_rtx): Check for overflow.
6574 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6576         * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6577         typo in comment.
6579 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6581         * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6582         and its corresponding call arg location note.
6584 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6585             Andrew Burgess  <andrew.burgess@embecosm.com>
6587         * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6588         declaration.
6589         (arc_return_slot_offset): Likewise.
6590         (arc_eh_return_address_location): New declaration.
6591         * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6592         (MUST_SAVE_REGISTER): Add exception handler case.
6593         (MUST_SAVE_RETURN_ADDR): Likewise.
6594         (arc_frame_pointer_required): Likewise.
6595         (arc_frame_pointer_needed): New function.
6596         (arc_compute_frame_size): Changed.
6597         (arc_expand_prologue): Likewise.
6598         (arc_expand_epilogue): Likewise.
6599         (arc_initial_elimination_offset): Likewise.
6600         (arc_return_slot_offset): Delete.
6601         (arc_eh_return_address_location): New function.
6602         (arc_builtin_setjmp_frame_value): Likewise.
6603         * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6604         (EH_RETURN_STACKADJ_RTX): Define.
6605         (EH_RETURN_HANDLER_RTX): Likewise.
6606         * config/arc/arc.md (eh_return): Delete.
6608 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6610         * print-tree.h (debug_vec_tree): Remove prototype.
6611         * gdbinit.in (pvt): Remove macro.
6613 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6615         PR tree-optimization/83047
6616         * gimple-ssa-store-merging.c
6617         (imm_store_chain_info::output_merged_store): If the loads with the
6618         same vuse are in different basic blocks, for load_gsi pick a load
6619         location that is dominated by the other loads.
6621         PR c++/83059
6622         * config/i386/i386.c (ix86_memmodel_check): Start
6623         -Winvalid-memory-model diagnostics with lowercase letter.
6625         PR debug/82718
6626         * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6627         set in_cold_section_p to the partition containing loc_list->first.
6628         When seeing loc_list->last_before_switch node, update secname and
6629         perform range_across_switch second partition handling only after that.
6631         PR debug/82933
6632         * run-rtl-passes.c: Include debug.h.
6633         (run_rtl_passes): Call debug_hooks->assembly_start.
6634         * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6635         multiple times.
6637         PR target/82981
6638         * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6639         OPTAB_DIRECT in calls to expand_simple_binop.
6641 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6643         PR c/81404
6644         * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6646 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6648         PR c++/72786
6649         * spellcheck.h (best_match::blithely_get_best_candidate): New
6650         accessor.
6652 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6654         * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6655         with lower case letter.
6657 2017-11-20  Uros Bizjak  <ubizjak@gmail.com>
6659         * config/i386/i386.md (bswaphi2): New expander.
6660         (*bswaphi2_movbe): New insn pattern.
6661         (bswaphi -> rorhi pepehole2): New peephole pattern.
6663 2017-11-20  Jeff Law  <law@redhat.com>
6665         * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6666         * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6667         * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6668         * gimple-ssa-evrp.c: Remove bits moved into new files.  Include
6669         gimple-ssa-evrp-analyze.h.
6671         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6672         set BB_VISITED here.
6673         (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6675 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6677         * tree-predcom.c: Add general comment on Store-Store chains.
6678         (split_data_refs_to_components): Postpone clearing eliminate_store_p
6679         flag in component.
6680         (get_chain_last_ref_at): Rename into...
6681         (get_chain_last_write_at): ...this.
6682         (get_chain_last_write_before_load): New function.
6683         (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6684         CT_STORE_STORE when write reference is added.
6685         (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6686         (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6687         (initialize_root_vars_store_elim_1): Ditto.
6688         (initialize_root_vars_store_elim_2): Ditto.  Replace rhs once default
6689         definition is created.
6690         (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6691         chain by replacing it with dominant stored value.
6693 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6695         * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6697 2017-11-20  Marc Glisse  <marc.glisse@inria.fr>
6699         * vr-values.c (extract_range_from_binary_expr): Use a full range
6700         for VR_VARYING.
6702 2017-11-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6704         * config/arm/arm.md (R4_REGNUM): Define constant.
6705         (nonsecure_call_internal): Remove r4 clobber.
6706         (nonsecure_call_value_internal): Likewise.
6707         * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6708         clobber and resequence match_operands.
6709         (nonsecure_call_value_reg_thumb1_v5): Likewise.
6710         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6711         (nonsecure_call_value_reg_thumb2): Likewise.
6713 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6715         PR tree-optimization/78821
6716         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6717         if base is TARGET_MEM_REF.  If base is not MEM_REF, set base_addr
6718         to the address of the base rather than the base itself.
6719         (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6720         (find_bswap_or_nop_finalize): New function.
6721         (find_bswap_or_nop): Use it.
6722         (bswap_replace): Return a tree rather than bool, change first
6723         argument from gimple * to gimple_stmt_iterator, allow inserting
6724         into an empty sequence, allow ins_stmt to be NULL - then emit
6725         all stmts into gsi.  Fix up MEM_REF address gimplification.
6726         (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6727         (struct store_immediate_info): Add N and INS_STMT non-static
6728         data members.
6729         (store_immediate_info::store_immediate_info): Initialize them
6730         from newly added ctor args.
6731         (merged_store_group::apply_stores): Formatting fixes.  Sort by
6732         bitpos at the end.
6733         (stmts_may_clobber_ref_p): For stores call also
6734         refs_anti_dependent_p.
6735         (gather_bswap_load_refs): New function.
6736         (imm_store_chain_info::try_coalesce_bswap): New method.
6737         (imm_store_chain_info::coalesce_immediate_stores): Use it.
6738         (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6739         (imm_store_chain_info::output_merged_store): Fail if number of
6740         new estimated stmts is bigger or equal than old.  Handle LROTATE_EXPR
6741         and NOP_EXPR rhs_code.
6742         (pass_store_merging::process_store): Compute n and ins_stmt, if
6743         ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6744         LROTATE_EXPR rhs_code.  Pass n and ins_stmt to store_immediate_info
6745         ctor.
6746         (pass_store_merging::execute): Calculate dominators.
6748         * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6749         BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6750         CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6751         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6752         find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6753         class pass_optimize_bswap, bswap_replace,
6754         pass_optimize_bswap::execute): Moved to ...
6755         * gimple-ssa-store-merging.c: ... this file.
6756         Include optabs-tree.h.
6757         (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6758         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6759         find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6760         anonymous namespace, remove static keywords.
6761         (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6762         (pass_optimize_bswap::execute): ... rather than here.  Formatting fix.
6764 2017-11-20  Jan Hubicka  <hubicka@ucw.cz>
6766         PR bootstrap/83062
6767         * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6769 2017-11-20  Aldy Hernandez  <aldyh@redhat.com>
6771         * vec.h (debug_helper): New function.
6772         (DEFINE_DEBUG_VEC): New macro.
6773         * hash-set.h (debug_helper): New function.
6774         (DEFINE_DEBUG_HASH_SET): New macro.
6775         * cfg.c (debug_slim (edge)): New function.
6776         Call DEFINE_DEBUG_VEC for edges.
6777         Call DEFINE_DEBUG_HASH_SET for edges.
6778         * cfghooks.c (debug_slim (basic_block)): New function.
6779         Call DEFINE_DEBUG_VEC for basic blocks.
6780         Call DEFINE_DEBUG_HASH_SET for basic blocks.
6781         * print-tree.c (debug_slim): New function to handle trees.
6782         Call DEFINE_DEBUG_VEC for trees.
6783         Call DEFINE_DEBUG_HASH_SET for trees.
6784         (debug (vec<tree, va_gc>) &): Remove.
6785         (debug (<vec<tree, va_gc>) *): Remove.
6786         * print-rtl.c (debug_slim): New function to handle const_rtx.
6787         Call DEFINE_DEBUG_VEC for rtx_def.
6788         Call DEFINE_DEBUG_VEC for rtx_insn.
6789         Call DEFINE_DEBUG_HASH_SET for rtx_def.
6790         Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6791         * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6792         (debug (vec<rtx_insn *> *ptr): Remove.
6793         (debug_insn_vector): Remove.
6794         * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6796 2017-11-20  Tom de Vries  <tom@codesourcery.com>
6798         PR rtl-optimization/82020
6799         * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6800         IF_THEN_ELSE condition.
6802 2017-11-19  Jeff Law  <law@redhat.com>
6804         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6805         of degenerates resulting from ignoring an edge.
6807 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6809         PR ipa/81360
6810         * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6812 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6814         PR ipa/83001
6815         * profile-count.c (profile_count::to_sreal_scale): Fix return value
6816         for uninitialied counts.
6818 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6820         PR ipa/60243
6821         * tree-inline.c (estimate_num_insns): Set to 1 at least.
6823 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6825         PR target/82713
6826         * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6828 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6830         * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6831         "do while (0)".
6833 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6835         * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6836         "do {} while (0)".
6838 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6840         * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6841         macro body.
6843 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6845         * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6846         "do {} while (0)".
6847         * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6849 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6851         * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6852         "do {} while (0)".
6853         * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6854         semicolon after MCORE_EXPORT_NAME call.
6856 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6858         PR target/82961
6859         * vmsdbgout.c (vmsdbgout_early_finish): New function.
6860         (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6862 2017-11-18  Jan Hubicka  <hubicka@ucw.cz>
6864         * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6865         (cgraph_edge::clone): Cleanup updating of profile.
6866         * ipa-cp.c (update_profiling_info): Likewise.
6867         * ipa-inline-transform.c (inline_transform): Likewise.
6868         * ipa-inline.c (inline_small_functions): Add missing space to dump.
6869         * ipa-split.c (execute_split_functions): Do not split when function
6870         is cold.
6871         * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6872         * profile-count.c (profile_count::dump): Add global0.
6873         (profile_count::to_cgraph_frequency): Do not ICE when entry is
6874         undefined.
6875         (profile_count::to_sreal_scale): Likewise.
6876         (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6877         (profile_count::combine_with_ipa_count): New function.
6878         * profile-count.h (profile_guessed_global0adjusted): New.
6879         (profile_count::adjusted_zero): New.
6880         (profile_count::global0adjusted): New.
6881         (profile_count::combine_with_ipa_count): New.
6882         * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6883         correct profile of return block of split functions.
6884         (copy_cfg_body): Remove unused profile_count.
6885         (copy_body): Likewise.
6886         (expand_call_inline): Update.
6887         (tree_function_versioning): Update.
6889 2017-11-18  Aldy Hernandez  <aldyh@redhat.com>
6891         * hash-set.h (hash_set::empty): New.
6892         * tree-ssa-threadbackward.h: Delete.
6893         * tree-ssa-threadbackward.c (class thread_jumps): New.
6894         Move max_threaded_paths into class.
6895         (fsm_find_thread_path): Remove arguments that are now in class.
6896         (profitable_jump_thread_path): Rename to...
6897         (thread_jumps::profitable_jump_thread_path): ...this.
6898         (convert_and_register_jump_thread_path): Rename to...
6899         (thread_jumps::convert_and_register_current_path): ...this.
6900         (check_subpath_and_update_thread_path): Rename to...
6901         (thread_jumps::check_subpath_and_update_thread_path): ...this.
6902         (register_jump_thread_path_if_profitable): Rename to...
6903         (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6904         (handle_phi): Rename to...
6905         (thread_jumps::handle_phi): ...this.
6906         (handle_assignment): Rename to...
6907         (thread_jumps::handle_assignment): ...this.
6908         (fsm_find_control_statement_thread_paths): Rename to...
6909         (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6910         (find_jump_threads_backwards): Rename to...
6911         (thread_jumps::find_jump_threads_backwards): ...this.
6912         Initialize path local data.
6913         (pass_thread_jumps::execute): Call find_jump_threads_backwards
6914         from within thread_jumps class.
6915         (pass_early_thread_jumps::execute): Same.
6917 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6919         * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
6921 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6923         * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
6924         to canonical location.
6926 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6928         PR target/81356
6929         * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
6930         Remove.
6931         (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
6933 2017-11-17  H.J. Lu  <hongjiu.lu@intel.com>
6935         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
6936         rcrt1.o%s/grcrt1.o%s for -static-pie.
6938 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6940         * i386.c (ix86_multiplication_cost, ix86_division_cost,
6941         ix86_shift_rotate_cost): Break out from ...
6942         (ix86_rtx_costs): ... here.
6943         (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
6944         vector operations.
6946 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6948         * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
6949         when entry block was promoted unlikely.
6950         (estimate_bb_frequencies): Increase frequency scale.
6951         * profile-count.h (profile_count): Export precision info.
6953 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6955         * tree-tailcall.c (eliminate_tail_call): Be more careful about not
6956         disturbin profile of entry block.
6958 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6960         * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
6961         roundoff errors.
6963 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6965         * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
6966         * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
6968 2017-11-17  Jeff Law  <law@redhat.com>
6970         * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
6971         from evrp_dom_walker class.  Various methods moved into new class.
6972         (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
6973         (evrp_range_analyzer::enter): New method.
6974         (evrp_range_analyzer::leave): New method.
6975         (evrp_dom_walker): Remove delegators no longer needed by this class.
6976         Replace vr_values data member with evrp_range_analyzer
6978         * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
6979         method extracted from evrp_dom_walker::before_dom_children.
6980         (evrp_dom_walker::record_ranges_from_stmt): Likewise.
6981         (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
6983         * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
6984         Add private copy constructor and move assignment operators.
6985         Privatize methods and class data where trivially possible.
6986         (evrp_dom_walker::cleanup): New function, extracted from
6987         execute_early_vrp.  Simplify access to class data.
6989         * vr-values.h (get_output_for_vrp): Prototype.
6990         * vr-values.c (get_output_for_vrp): New function extracted from
6991         vrp_visit_assignment_or_call and extract_range_from_stmt.
6992         (vrp_visit_assignment_or_call): Use get_output_for_vrp.  Simplify.
6994 2017-11-17  Luis Machado  <luis.machado@linaro.org>
6996         * config/aarch64/aarch64.c
6997         (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
6998         (qdf24xx_tunings) <autoprefetcher_model>: Set to
6999         tune_params::AUTOPREFETCHER_WEAK.
7001 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
7003         PR target/82641
7004         * config/arm/arm.c (arm_valid_target_attribute_rec):
7005         Parse "arch=" and "+<ext>".
7006         (arm_valid_target_attribute_tree): Re-init global options.
7007         (arm_option_override): Make non-static.
7008         (arm_options_perform_arch_sanity_checks): Make errors fatal.
7009         * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
7010         (__ARM_FEATURE_CRC32): Support undef.
7011         * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
7012         * doc/extend.texi (ARM Function Attributes): Add pragma and target.
7014 2017-11-17  David Malcolm  <dmalcolm@redhat.com>
7016         * gdbinit.in (break-on-diagnostic): New command.
7018 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7020         * config/i386/i386.c (ix86_expand_epilogue): Change simple
7021         return to indirect jump for EH return if control-flow
7022         protection is enabled. Change explicit 'false' argument in
7023         pro_epilogue_adjust_stack with a value of
7024         flag_cf_protection.
7025         * config/i386/i386.md (simple_return_indirect_internal):
7026         Remove SImode restriction to support 64-bit.
7028 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
7030         * combine.c (added_notes_insn): New.
7031         (try_combine): Handle added_notes_insn like added_links_insn.
7032         Rewrite return value code.
7033         (distribute_notes): Set added_notes_insn to the earliest insn we added
7034         a note to.
7036 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
7038         PR rtl-optimization/82621
7039         * combine.c (try_combine): Do not split PARALLELs of two SETs if the
7040         dest of one of those SETs is unused.
7042 2017-11-17  Richard Biener  <rguenther@suse.de>
7044         PR fortran/83017
7045         * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
7046         * tree-pretty-print.c (dump_generic_node): Handle
7047         annot_expr_parallel_kind.
7048         * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
7049         * gimplify.c (gimple_boolify): Likewise.
7051 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7053         * config.gcc (extra_headers): Add cet.h for x86 targets.
7054         * config/i386/cet.h: New file.
7055         * doc/install.texi: Add --enable-cet/--disable-cet.
7057 2017-11-17  Richard Biener  <rguenther@suse.de>
7059         PR tree-optimization/83017
7060         * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
7061         (gen_parallel_loop): Properly count iterations.
7062         (parallelize_loops): Handle loop->can_be_parallel independent
7063         of flag_loop_parallelize_all.  Make static profitability test match
7064         the runtime one.
7065         * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
7066         * invoke.texi (parloops-min-per-thread): Document.
7068 2017-11-17  Vineet Gupta  <vgupta@synopsys.com>
7070         * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
7071         upstreaming review comments.
7073 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
7075         * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
7076         with fast unaligned access.
7077         * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
7079 2017-11-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
7081         * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
7082         to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
7084 2017-11-17  Richard Biener  <rguenther@suse.de>
7086         * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
7087         folding of references.
7089 2017-11-17  Qing Zhao  <qing.zhao@oracle.com>
7091         PR middle-end/78809
7092         * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
7093         of replacing call to strncmp with corresponding call to strcmp when
7094         meeting conditions.
7096 2017-11-17  Sergey Shalnov  <Sergey.Shalnov@intel.com>
7098         * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
7099         option prefer-avx256 for skylake-avx512 configuration.
7100         * config/i386/i386.c (ix86_option_override_internal): Ditto.
7101         (get_builtin_code_for_version): Ditto.
7103 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7104             Monk Chiang  <sh.chiang04@gmail.com>
7106         * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
7107         (FIXED_REGISTERS): Reserve more register numbers.
7108         (CALL_USED_REGISTERS): Likewise.
7109         (REG_ALLOC_ORDER): Likewise.
7110         (REG_CLASS_CONTENTS): Likewise.
7111         (REGISTER_NAMES): Likewise.
7113 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7114             Kito Cheng  <kito.cheng@gmail.com>
7116         * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
7117         V2SI.
7118         * config/nds32/iterators.md: Add vector mode iterators and attributes.
7120 2017-11-16  Steven Munroe  <munroesj@gcc.gnu.org>
7122         * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
7123         parameter list for vec_splats.
7125 2017-11-16  Joseph Myers  <joseph@codesourcery.com>
7127         * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
7128         date of C17.
7129         (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
7131 2017-11-16  Thomas Koenig  <tkoenig@gcc.gnu.org>
7133         PR bootstrap/82856
7134         * doc/install.texi: Document incompatibility of Perl >=5.6.26
7135         with the required version of automake 1.11.6.
7137 2017-11-16  Pat Haugen  <pthaugen@us.ibm.com>
7139         * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
7140         for it.
7141         (DU_C2_3_power9): Correct reservation combinations.
7142         (FP_DIV_power9, VEC_DIV_power9): New.
7143         (power9-alu): Split out rotate/shift...
7144         (power9-rot): ...to here, correct dispatch resource.
7145         (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
7146         resource.
7147         (power9-fp): Correct latency.
7148         (power9-sdiv): Add div/sqrt resource.
7149         (power9-ddiv): Correct latency, add div/sqrt resource.
7150         (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
7151         (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
7152         resource.
7153         (power9-qpdiv, power9-qpmul): Adjust resource usage.
7155 2017-11-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
7157         * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
7158         switch statement mapping KF built-ins to TF built-ins if we don't
7159         have the proper ISA 3.0 assembler support.
7161 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7163         * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
7164         (lower_emutls_function_body): Do not compute it.
7166 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7168         * ipa-split.c (split_bb_info): Turn time to sreal.
7169         (split_point): Likewise.
7170         (dump_split_point): Likewise.
7171         (fine_split_points): Likewise.
7172         (execute_split_functions): Only zero split_bbs; turn time to sreals.
7174 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7176         * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
7177         in sreal.
7179 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7181         * predict.c (combine_predictions_for_bb): Preserve zero predicted
7182         edges.
7183         (expensive_function_p): Remove useless assert.
7184         (determine_unlikely_bbs): Propagate also forward; determine cold blocks
7186 2017-11-16  Martin Sebor  <msebor@redhat.com>
7188         PR tree-optimization/82588
7189         PR tree-optimization/82583
7190         * tree-vrp.c (check_array_ref): Handle flexible array members,
7191         string literals, and inner indices.
7192         (search_for_addr_array): Add detail to diagnostics.
7194 2017-11-16  Nathan Sidwell  <nathan@acm.org>
7196         PR c++/82836
7197         PR c++/82737
7198         * tree.h (COPY_DECL_RTL): Rename parms for clarity.
7199         (SET_DECL_ASSEMBLER_NAME): Forward to
7200         overwrite_decl_assembler_name.
7201         (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
7202         (overwrite_decl_assembler_name): Declare.
7203         * tree.c (overwrite_decl_assembler_name): New.
7204         * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
7205         (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
7206         (LANG_HOOKS_INITIALIZER): Add it.
7207         * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
7208         * langhooks.c (lhd_set_decl_assembler_name): Use
7209         SET_DECL_ASSEMBLER_NAME.
7210         (lhd_overwrite_decl_assembler_name): Default implementation.
7212 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7213             Jackson Woodruff  <jackson.woodruff@arm.com>
7215         PR tree-optimization/71026
7216         * match.pd: Canonicalize constant multiplies in division.
7218 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7220         * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
7221         * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
7222         Remove OPT_fomit_frame_pointer entry.
7223         * common/config/arc/arc-common.c: Likewise.
7224         * common/config/arm/arm-common.c: Likewise.
7225         * common/config/avr/avr-common.c: Likewise.
7226         * common/config/c6x/c6x-common.c: Likewise.
7227         * common/config/cr16/cr16-common.c: Likewise.
7228         * common/config/cris/cris-common.c: Likewise.
7229         * common/config/epiphany/epiphany-common.c: Likewise.
7230         * common/config/fr30/fr30-common.c: Likewise.
7231         * common/config/frv/frv-common.c: Likewise.
7232         * common/config/ia64/ia64-common.c: Likewise.
7233         * common/config/iq2000/iq2000-common.c: Likewise.
7234         * common/config/lm32/lm32-common.c: Likewise.
7235         * common/config/m32r/m32r-common.c: Likewise.
7236         * common/config/mcore/mcore-common.c: Likewise.
7237         * common/config/microblaze/microblaze-common.c: Likewise.
7238         * common/config/mips/mips-common.c: Likewise.
7239         * common/config/mmix/mmix-common.c: Likewise.
7240         * common/config/mn10300/mn10300-common.c: Likewise.
7241         * common/config/nios2/nios2-common.c: Likewise.
7242         * common/config/pa/pa-common.c: Likewise.
7243         * common/config/pdp11/pdp11-common.c: Likewise.
7244         * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7245         * common/config/riscv/riscv-common.c: Likewise.
7246         * common/config/rs6000/rs6000-common.c: Likewise.
7247         * common/config/rx/rx-common.c: Likewise.
7248         * common/config/s390/s390-common.c: Likewise.
7249         * common/config/sh/sh-common.c: Likewise.
7250         * common/config/sparc/sparc-common.c: Likewise.
7251         * common/config/tilegx/tilegx-common.c: Likewise.
7252         * common/config/tilepro/tilepro-common.c: Likewise.
7253         * common/config/v850/v850-common.c: Likewise.
7254         * common/config/visium/visium-common.c: Likewise.
7255         * common/config/xstormy16/xstormy16-common.c: Likewise.
7256         * common/config/xtensa/xtensa-common.c: Likewise.
7257         * invoke.texi (-fomit-frame-pointer): Update documentation.
7259 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7261         * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7263 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7265         * cfg.c (scale_bbs_frequencies_int,
7266         cale_bbs_frequencies_gcov_type): Remove.
7267         * cfg.h (scale_bbs_frequencies_int,
7268         cale_bbs_frequencies_gcov_type): Remove.
7270 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7272         * tree-ssa-loop-manip.c
7273         (scale_dominated_blocks_in_loop): Update to profile counts.
7274         (tree_transform_and_unroll_loop): Likewise.
7276 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7278         * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7279         scale_bbs_frequencies_int.
7281 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7283         * final.c (compute_alignments): Use counts rather than frequencies.
7285 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7287         * cfgloopanal.c: Include sreal.h
7288         (average_num_loop_insns): Use counts and sreal for accounting.
7290 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7292         * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7293         manipulation.
7295 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7296             Kito Cheng  <kito.cheng@gmail.com>
7298         * config/nds32/constraints.md: Provide more constraints.
7299         * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7300         REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7301         support constraints usage.
7303 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7305         * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7307 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7308             Kito Cheng  <kito.cheng@gmail.com>
7310         * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7311         * config/nds32/nds32.opt: Refine the layout.
7312         * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7313         TARGET_EXT_STRING): Support new options.
7314         * config/nds32/nds32.h: Likewise.
7315         * config/nds32/nds32.md: Likewise.
7316         * config/nds32/nds32-predicates.c: Likewise.
7317         * config/nds32/constraints.md: Likewise.
7318         * common/config/nds32/nds32-common.c: Likewise.
7320 2017-11-16  Julia Koval  <julia.koval@intel.com>
7322         PR target/82983
7323         * config/i386/gfniintrin.h: Add sse check.
7324         * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7326 2017-11-16  Julia Koval  <julia.koval@intel.com>
7328         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7329         OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7330         (ix86_handle_option): Handle -mavx512vbmi2.
7331         * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7332         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7333         * config/i386/i386-c.c (__AVX512VBMI2__): New.
7334         * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7335         (ix86_valid_target_attribute_inner_p): Ditto.
7336         * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7337         * config/i386/i386.opt (mavx512vbmi2): New option.
7338         * doc/invoke.texi: Add new option.
7340 2017-11-16  Julia Koval  <julia.koval@intel.com>
7342         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7343         _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7344         _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7345         _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7346         _mm512_gf2p8mul_epi8): New intrinsics.
7347         * config/i386/i386-builtin-types.def
7348         (V64QI_FTYPE_V64QI_V64QI): New type.
7349         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7350         __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7351         __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7352         __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7353         * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7354         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7356 2017-11-15  Uros Bizjak  <ubizjak@gmail.com>
7358         * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7359         explicitly as a stream of bytes.
7361 2017-11-15  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7363         * config/rs6000/altivec.h (vec_xst_be): New #define.
7364         * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7365         and externalize from *altivec_vperm_<mode>_internal.
7366         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7367         instantiation.
7368         (XL_BE_V8HI): Likewise.
7369         (XL_BE_V4SI): Likewise.
7370         (XL_BE_V4SI): Likewise.
7371         (XL_BE_V2DI): Likewise.
7372         (XL_BE_V4SF): Likewise.
7373         (XL_BE_V2DF): Likewise.
7374         (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7375         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7376         all array entries with these keys: VSX_BUILTIN_VEC_XL,
7377         VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST.  Add entries for key
7378         VSX_BUILTIN_VEC_XST_BE.
7379         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7380         (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7381         built-ins.
7382         (altivec_init_builtins): Replace conditional calls to def_builtin
7383         for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7384         __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7385         with unconditional calls.  Remove calls to def_builtin for
7386         __builtin_vsx_le_be_<mode>.  Add a call to def_builtin for
7387         __builtin_vec_xst_be.
7388         * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7389         to define_expand, and add alternate RTL generation for P8.
7390         (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7391         vsx_ld_elemrev_v8hi.
7392         (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7393         add alternate RTL generation for P8.
7394         (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7395         vsx_ld_elemrev_v16qi.
7396         (vsx_st_elemrev_v8hi): Convert define_insn
7397         to define_expand, and add alternate RTL generation for P8.
7398         (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7399         vsx_st_elemrev_v8hi.
7400         (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7401         add alternate RTL generation for P8.
7402         (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7403         vsx_st_elemrev_v16qi.
7405 2017-11-15  H.J. Lu  <hongjiu.lu@intel.com>
7407         PR target/82990
7408         * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7409         TARGET_AVX512ER check.
7410         (ix86_option_override_internal): Set MASK_VZEROUPPER if
7411         neither -mzeroupper nor -mno-zeroupper is used and
7412         TARGET_EMIT_VZEROUPPER is set.
7413         * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7414         * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7416 2017-11-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7418         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7419         folding of vector compares.
7420         (fold_build_vec_cmp): New helper function.
7421         (fold_compare_helper): New helper function.
7422         (builtin_function_type): Add compare builtins to the list of functions
7423         having unsigned arguments.  Cosmetic updates to comment indentation.
7424         * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7425         the not+eq combination.
7427 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7429         PR tree-optimization/82726
7430         PR tree-optimization/70754
7431         * tree-predcom.c (order_drefs_by_pos): New function.
7432         (combine_chains): Move code setting has_max_use_after to...
7433         (try_combine_chains): ...here.  New parameter.  Sort combined chains
7434         according to position information.
7435         (tree_predictive_commoning_loop): Update call to above function.
7436         (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7438 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7440         PR tree-optimization/82726
7441         Revert
7442         2017-01-23  Bin Cheng  <bin.cheng@arm.com>
7444         PR tree-optimization/70754
7445         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7446         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
7447         combined stmt before it if not NULL.
7448         (combine_chains): Process refs reversely and compute dominance point
7449         for root ref.
7451         Revert
7452         2017-02-23  Bin Cheng  <bin.cheng@arm.com>
7454         PR tree-optimization/79663
7455         * tree-predcom.c (combine_chains): Process refs in reverse order
7456         only for ZERO length chains, and add explaining comment.
7458 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7460         * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7461         * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7462         * doc/invoke.texi (ARM Options): Add armv8.3-a.
7464 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7466         * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7468 2017-11-15  Martin Liska  <mliska@suse.cz>
7470         * tree-cfg.c (pass_warn_function_return::execute):
7471         Compare warn_return_type for greater than zero.
7473 2017-11-15  Sebastian Peryt  <sebastian.peryt@intel.com>
7475         PR target/82941
7476         PR target/82942
7477         * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7478         to return true on Xeon and not on Xeon Phi.
7479         (ix86_check_avx256_register): Changed to ...
7480         (ix86_check_avx_upper_register): ... this. Add extra check for
7481         VALID_AVX512F_REG_OR_XI_MODE.
7482         (ix86_avx_u128_mode_needed): Changed
7483         ix86_check_avx256_register to ix86_check_avx_upper_register.
7484         (ix86_check_avx256_stores): Changed to ...
7485         (ix86_check_avx_upper_stores): ... this. Changed
7486         ix86_check_avx256_register to ix86_check_avx_upper_register.
7487         (ix86_avx_u128_mode_after): Changed
7488         avx_reg256_found to avx_upper_reg_found. Changed
7489         ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7490         (ix86_avx_u128_mode_entry): Changed
7491         ix86_check_avx256_register to ix86_check_avx_upper_register.
7492         (ix86_avx_u128_mode_exit): Ditto.
7493         * config/i386/i386.h: (host_detect_local_cpu): New define.
7495 2017-11-15  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
7497         * config/arm/xgene1.md (xgene1): Split into automatons
7498         xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7499         (xgene1_f_load): Adjust reservations and/or types.
7500         (xgene1_f_store): Likewise.
7501         (xgene1_load_pair): Likewise.
7502         (xgene1_store_pair): Likewise.
7503         (xgene1_fp_load1): Likewise.
7504         (xgene1_load1): Likewise.
7505         (xgene1_store1): Likewise.
7506         (xgene1_move): Likewise.
7507         (xgene1_alu): Likewise.
7508         (xgene1_simd): Likewise.
7509         (xgene1_bfm): Likewise.
7510         (xgene1_neon_load1): Likewise.
7511         (xgene1_neon_store1): Likewise.
7512         (xgene1_neon_logic): Likewise.
7513         (xgene1_neon_st1): Likewise.
7514         (xgene1_neon_ld1r): Likewise.
7515         (xgene1_alu_cond): Added.
7516         (xgene1_shift_reg): Likwise.
7517         (xgene1_bfx): Likewise.
7518         (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7520 2017-11-15  Jakub Jelinek  <jakub@redhat.com>
7522         PR target/82981
7523         * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7524         ssa-iterators.h.
7525         (can_widen_mult_without_libcall): New function.
7526         (expand_mul_overflow): If only checking unsigned mul overflow,
7527         not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7528         Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7529         no other way works.  Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7530         (expand_DIVMOD): Formatting fix.
7531         * expmed.h (expand_mult): Add NO_LIBCALL argument.
7532         * expmed.c (expand_mult): Likewise.  Use OPTAB_WIDEN rather
7533         than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7535         PR tree-optimization/82977
7536         * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7537         constructed temporary to strlen_to_stridx.put.
7539 2017-11-15  Martin Liska  <mliska@suse.cz>
7541         * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7542         * configure: Regenerate.
7544 2017-11-15  Martin Liska  <mliska@suse.cz>
7546         PR target/82927
7547         * config/sh/sh-mem.cc: Use proper probability for
7548         REG_BR_PROB_NOTE.
7550 2017-11-14  Jeff Law  <law@redhat.com>
7552         * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7553         the red zone for stack_clash_protection_final_dynamic_probe targets
7554         when the total dynamic stack size is zero bytes.
7556         * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7557         blocks is post order.
7559 2017-11-15  Alexandre Oliva  <aoliva@redhat.com>
7561         * dumpfile.h (TDF_COMPARE_DEBUG): New.
7562         * final.c (rest_of_clean_state): Set it for the
7563         -fcompare-debug dump.
7564         * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7565         class when TDF_COMPARE_DEBUG is set.
7567         * dwarf2out.c (gen_producer_string): Discard
7568         OPT_fcompare_debug.
7570 2017-11-15  Joseph Myers  <joseph@codesourcery.com>
7572         PR c/81156
7573         * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7574         * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7575         (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7576         (__TGMATH_REAL_2_3): Remove macros.
7577         (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7578         (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7579         __builtin_tgmath.
7580         (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7581         __TGMATH_REAL_2.
7582         (remquo): Define using __TGMATH_REAL_3.
7584 2017-11-14  Jeff Law  <law@redhat.com>
7586         * vr-values.c: New file with contents extracted from tree-vrp.c.
7587         * Makefile.in (OBJS): Add vr-values.o
7588         * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7589         (set_value_range, set_and_canonicalize_value_range): Likewise.
7590         (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7591         (value_range_constant_singleton, symbolic_range_p): Likewise.
7592         (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7593         (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7594         (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7595         (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7596         (find_case_label_range, find_case_label_index): Likewise.
7597         (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7598         (range_int_cst_singleton_p, value_inside_range): Likewise.
7599         (get_single_symbol): Likewise.
7600         (switch_update): Move structure definition here.
7601         (to_remove_edges, to_update_switch_stmts): Provide externs.
7602         * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7603         (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7604         (vrp_val_is_min, set_value_range): Likewise.
7605         (set_and_canonicalize_value_range, copy_value_range): Likewise.
7606         (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7607         (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7608         (range_is_nonnull, range_int_cst_p): Likewwise.
7609         (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7610         (get_single_symbol, operand_less_p): Likewise
7611         (compare_values_warnv, compare_values): Likewise.
7612         (value_inside_range, value_range_constant_singleton): Likewise.
7613         (zero_nonzero_bitgs_from_vr): Likewise.
7614         (extract_range_from_binary_expr_1): Likewise.
7615         (overflow_comparison_p): Likewise.
7616         (to_remove_edges, to_update_switch_stmts): Likewise.
7617         (find_case_label-index, find_case_label_range): Likewise.
7618         (switch_update, set_value_range_to_nonnegative): Remove.
7619         (set_value_range_to_truthvalue): Likewise.
7620         (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7621         (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7622         (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7623         (find_case_label_ranges, test_for_singularity): Likewise.
7624         (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7625         (x_vr_values): Move to its remaining use site.
7627 2017-11-10  Jeff Law  <law@redhat.com>
7629         * vr-values.h (VR_INITIALIZER): Move #define here.
7630         * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7631         * Makefile.in (OBJS): Add tree-evrp.o
7632         * tree-vrp.h (assert_info): Move structure definition here.
7633         (set_value_range_to_varying): Prototype.
7634         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7635         (infer_value_range, register_edge_assert_for): Likewise.
7636         (stmt_interesting_for_vrp): Likewise.
7637         * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7638         (set_value_range_to_varying): No longer static.
7639         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7640         (infer_value_range, register_edge_assert_for): Likewise.
7642 2017-11-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
7644         * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7645         generate the XXBRD instruction.
7647         * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7648         (rs6000_builtin_type_compatible): Treat _Float128 and long double
7649         as being compatible if -mabi=ieeelongdouble.
7650         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7651         to setup float128 built-ins with hardware support.
7652         (BU_FLOAT128_HW_2): Likewise.
7653         (BU_FLOAT128_HW_3): Likewise.
7654         (BU_FLOAT128_HW_VSX_1): Likewise.
7655         (BU_FLOAT128_HW_VSX_2): Likewise.
7656         (scalar_extract_expq): Change float128 built-in functions to
7657         accommodate having both KFmode and TFmode functions.  Use the
7658         KFmode variant as the default.
7659         (scalar_extract_sigq): Likewise.
7660         (scalar_test_neg_qp): Likewise.
7661         (scalar_insert_exp_q): Likewise.
7662         (scalar_insert_exp_qp): Likewise.
7663         (scalar_test_data_class_qp): Likewise.
7664         (sqrtf128_round_to_odd): Delete processing the round to odd
7665         built-in functions as special built-in functions, and define them
7666         as float128 built-ins.  Use the KFmode variant as the default.
7667         (truncf128_round_to_odd): Likewise.
7668         (addf128_round_to_odd): Likewise.
7669         (subf128_round_to_odd): Likewise.
7670         (mulf128_round_to_odd): Likewise.
7671         (divf128_round_to_odd): Likewise.
7672         (fmaf128_round_to_odd): Likewise.
7673         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7674         support for KFmode and TFmode xststdcqp calls.
7675         (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7676         point, switch the built-in handlers for the get/set float128
7677         exponent, get float128 mantissa, float128 test built-ins, and the
7678         float128 round to odd built-in functions.  Eliminate creating the
7679         float128 round to odd built-in functions as special built-ins.
7680         (rs6000_init_builtins): Eliminate special creation of the float128
7681         round to odd built-in functions.
7682         * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7683         function insns to support both TFmode and KFmode variants.
7684         (xsxsigqp_<mode>): Likewise.
7685         (xsiexpqpf_<mode>): Likewise.
7686         (xsiexpqp_<mode>): Likewise.
7687         (xststdcqp_<mode>): Likewise.
7688         (xststdcnegqp_<mode>): Likewise.
7689         (xststdcqp_<mode>): Likewise.
7691 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7693         * tree-ssa-threadupdate.c (compute_path_counts): Remove
7694         unused path_in_freq_ptr parameter.
7695         (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7697 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7699         * ipa-inline.c (edge_badness): Dump sreal frequency.
7700         (compute_inlined_call_time): Match natural implementaiton ...
7701         * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7702         forgotten division by CGRAPH_FREQ_BASE.
7704 2017-11-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7706         * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7707         Solaris 11.  Update comment.
7708         * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7709         renaming.
7710         * config/sol2.h (STARTFILE_SPEC): Likewise.
7711         * configure: Regenerate.
7713 2017-11-14  Carl Love  <cel@us.ibm.com>
7715         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7716         le_ and be_ prefixes to swap* variables.  Remove
7717         if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7718         statements.
7720 2017-11-14  Jason Merrill  <jason@redhat.com>
7722         Support GTY((cache)) on hash_map.
7723         * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7724         (ggc_cache_remove): Override it instead of ggc_mx.
7725         * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7726         (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7727         * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7728         (simple_cache_map_traits): Override maybe_mx.
7729         * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7730         (hash_map): Friend gt_cleare_cache.
7731         (gt_cleare_cache): New.
7732         * tree.h (tree_cache_traits): New hash_map traits class.
7733         (tree_cache_map): New typedef.
7735 2017-11-14  Richard Biener  <rguenther@suse.de>
7737         * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7738         paramter and handling.
7739         (cleanup_control_flow_bb): Likewise.
7740         (cleanup_control_flow_pre): New helper performing a DFS walk
7741         to call cleanup_control_flow_bb in PRE order.
7742         (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7743         via cleanup_control_flow_pre.
7745 2017-11-14  James Greenhalgh  <james.greenhalgh@arm.com>
7747         * config/aarch64/aarch64-simd.md
7748         (aarch64_simd_bsl<mode>_internal): Remove DImode.
7749         (*aarch64_simd_bsl<mode>_alt): Likewise.
7750         (aarch64_simd_bsldi_internal): New.
7751         (aarch64_simd_bsldi_alt): Likewise.
7753 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7755         * tracer.c (better_p): Do not compare frequencies.
7756         * reg-stack.c (better_edge): Likewise.
7757         * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7758         and back.
7760 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7762         * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7763         * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7764         * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7765         * lto-streamer-in.c (input_function): Use update_max_bb_count.
7766         * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7767         * predict.c (maybe_hot_frequency_p): Inline to ...
7768         (maybe_hot_count_p): ... here; rewrite to counts.
7769         (counts_to_freqs): Rename to ...
7770         (update_max_bb_count): ... this one.
7771         (expensive_function_p): Use counts.
7772         (estimate_bb_frequencies): Update.
7773         (rebuild_frequencies): Update.
7774         * predict.h (counts_to_freqs): Rename to ...
7775         (update_max_bb_count): ... this one.
7776         * profile.c (compute_branch_probabilities): Add debug info
7777         * tree-inline.c (expand_call_inline): Update debug info.
7778         (optimize_inline_calls): Use update_max_bb_count..
7779         (tree_function_versioning): Use update_max_bb_count..
7780         * value-prof.c (gimple_value_profile_transformations):
7781         Do not use update_max_bb_count.
7783 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7785         * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7786         always use frequencies.
7788 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7790         * bb-reorder.c: Remove frequencies from comments.
7791         (better_edge_p): Use profile counts.
7792         (find_traces): Dump profile counts.
7793         (rotate_loop): Use profile counts.
7794         (find_traces_1_round): Likewise.
7795         (connect_better_edge_p): Use counts instead of probabilities for
7796         reverse walk.
7797         (copy_bb_p): Drop early check for non-0 frequency.
7798         (sanitize_hot_paths): Update comments.
7800 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7802         * ipa-split.c (struct split_point): Add count.
7803         (consider_split): Do not compute incoming frequency; compute incoming
7804         count and store it to split_point.
7805         (split_function): Set count of the call to split part correctly.
7807 2017-11-13  Carl Love  <cel@us.ibm.com>
7809         * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7811 2017-11-13  Tom Tromey  <tom@tromey.com>
7813         * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7815 2017-11-13  Carl Love  <cel@us.ibm.com>
7817         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7818         Add support for builtins:
7819         unsigned int vec_first_{,miss}_match_{,or_eos}index,
7820         vector {un,}signed {char,int,short},
7821         vector {un,}signed {char,int,short}) arguments.
7822         * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7823         VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7824         Add BU_P9V_AV_2 expansions for the builtins.
7825         * config/rs6000/altivec.h (vec_first_match_index,
7826         vec_first_mismatch_index, vec_first_match_or_eos_index,
7827         vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7828         * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7829         new extern declaration.
7830         * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7831         * config/rs6000/vsx.md (first_match_index_<mode>,
7832         first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7833         first_mismatch_or_eos_index_<mode>): Add define expand.
7834         (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7835         * doc/extend.texi: Update the built-in documenation file for the new
7836         built-in functions.
7838 2017-11-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
7840         * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7841         into the min/max operations for _Float<N> and _Float<N>X types.
7843 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7845         PR lto/81351
7846         * dwarf2out.c (do_eh_frame): New static variable.
7847         (dwarf2out_begin_prologue): Set it.
7848         (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7850 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7852         * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7854         * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7856 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7858         PR lto/81351
7859         * debug.h (dwarf2out_do_eh_frame): Declare.
7860         * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7861         (dwarf2out_do_frame): Use it.
7862         (dwarf2out_do_cfi_asm): Likewise.
7863         * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7864         (dwarf2out_assembly_start): Likewise.
7865         (dwarf2out_begin_prologue): Fix comment.
7866         * toplev.c (compile_file): Always call dwarf2out_frame_finish
7867         if the target needs either debug or unwind DWARF2 info.
7868         * lto-opts.c (lto_write_options): Do not save -fexceptions,
7869         -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7870         -ftrapping-math, -ftrapv and -fwrapv.
7872 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7874         * cgraph.c (cgraph_edge::sreal_frequency): New function.
7875         * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7876         * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7877         (estimate_edge_size_and_time): Likewise.
7878         (ipa_merge_fn_summary_after_inlining): Likewise.
7879         * ipa-inline.c (cgraph_freq_base_rec): Remove.
7880         (compute_uninlined_call_time): Use sreal_frequency.
7881         (compute_inlined_call_time): Likewise.
7882         (ipa_inline): Do not initialize cgraph_freq_base_rec.
7883         * profile-count.c: Include sreal.h.
7884         (profile_count::to_sreal_scale): New.
7885         * profile-count.h: Forward declare sreal.
7886         (profile_count::to_sreal_scale): Declare.
7888 2017-11-13  Nathan Sidwell  <nathan@acm.org>
7890         * diagnostic.c (maybe_line_and_column): New.
7891         (diagnostic_get_location_text): Use it.
7892         (diagnostic_report_current_module): Likewise.
7893         (test_diagnostic_get_location_text): Add tests.
7895 2017-11-13  Luis Machado  <luis.machado@linaro.org>
7897         * doc/md.texi (Specifying processor pipeline description): Fix
7898         incorrect latency for the div instruction example.
7900 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7902         PR tree-optimization/78821
7903         * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7904         that bit_not_p is the same.
7905         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7906         (split_group): Count precisely bit_not_p bits in each statement.
7907         (invert_op): New function.
7908         (imm_store_chain_info::output_merged_store): Use invert_op to
7909         emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7910         but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
7912 2017-11-13  Martin Liska  <mliska@suse.cz>
7914         * gcov.c (struct coverage_info): Remove typedef of coverage_t.
7915         (struct source_info): Likewise.
7916         (add_branch_counts): Likewise.
7917         (add_line_counts): Likewise.
7918         (function_summary): Likewise.
7919         (output_intermediate_line): Likewise.
7920         (generate_results): Likewise.
7922 2017-11-13  Martin Liska  <mliska@suse.cz>
7924         * gcov.c (struct block_info): Remove typedef for block_t.
7925         (struct line_info): Likewise.
7926         (line_info::has_block): Likewise.
7927         (EXIT_BLOCK): Likewise.
7928         (unblock): Likewise.
7929         (circuit): Likewise.
7930         (get_cycles_count): Likewise.
7931         (process_file): Likewise.
7932         (read_graph_file): Likewise.
7933         (solve_flow_graph): Likewise.
7934         (find_exception_blocks): Likewise.
7935         (add_line_counts): Likewise.
7936         (accumulate_line_info): Likewise.
7937         (output_line_details): Likewise.
7939 2017-11-13  Martin Liska  <mliska@suse.cz>
7941         * gcov.c (struct arc_info): Remove typedef for arc_t.
7942         (struct line_info): Likewise.
7943         (add_branch_counts): Likewise.
7944         (output_branch_count): Likewise.
7945         (function_info::~function_info): Likewise.
7946         (circuit): Likewise.
7947         (output_intermediate_line): Likewise.
7948         (read_graph_file): Likewise.
7949         (solve_flow_graph): Likewise.
7950         (find_exception_blocks): Likewise.
7951         (add_line_counts): Likewise.
7952         (accumulate_line_info): Likewise.
7953         (output_line_details): Likewise.
7954         (output_function_details): Likewise.
7956 2017-11-13  Martin Liska  <mliska@suse.cz>
7958         * gcov.c (struct function_info): Remove typedef for function_t.
7959         (struct source_info): Likewise.
7960         (source_info::get_functions_at_location): Likewise.
7961         (solve_flow_graph): Likewise.
7962         (find_exception_blocks): Likewise.
7963         (add_line_counts): Likewise.
7964         (output_intermediate_file): Likewise.
7965         (process_file): Likewise.
7966         (generate_results): Likewise.
7967         (release_structures): Likewise.
7968         (read_graph_file): Likewise.
7969         (read_count_file): Likewise.
7970         (accumulate_line_counts): Likewise.
7971         (output_lines): Likewise.
7973 2017-11-13  Martin Liska  <mliska@suse.cz>
7975         * gcov.c (function_info::function_info): Remove num_counts
7976         and add vector<gcov_type>.
7977         (function_info::~function_info): Use the vector.
7978         (process_file): Likewise.
7979         (read_graph_file): Likewise.
7980         (read_count_file): Likewise.
7981         (solve_flow_graph): Likewise.
7983 2017-11-13  Martin Liska  <mliska@suse.cz>
7985         * gcov.c (function_info::is_artificial): New function.
7986         (process_file): Erase all artificial early.
7987         (generate_results): Skip as all artificial are already
7988         removed.
7990 2017-11-13  Martin Liska  <mliska@suse.cz>
7992         * gcov.c (read_graph_file): Store to global vector of functions.
7993         (read_count_file): Iterate the vector.
7994         (process_file): Likewise.
7995         (generate_results): Likewise.
7996         (release_structures): Likewise.
7998 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
8000         PR tree-optimization/82954
8001         * gimple-ssa-store-merging.c
8002         (imm_store_chain_info::coalesce_immediate_stores): If
8003         !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
8005 2017-11-13  Richard Sandiford  <richard.sandiford@linaro.org>
8007         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
8008         Upddate call to ENDIAN_LANE_N.
8009         (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
8010         (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
8011         (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
8012         and use aarch64_endian_lane_rtx.
8013         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
8015 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8017         * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
8019 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8021         * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
8023 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8025         * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
8026         body.
8027         * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
8028         * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
8029         * defaults.h (ASM_OUTPUT_LABELREF): Same.
8031 2017-11-11  Martin Sebor  <msebor@redhat.com>
8033         PR c/81117
8034         * doc/extend.texi (attribute nonstring): Remove spurious argument.
8036         PR bootstrap/82948
8037         * prefic.c (translate_name): Replace strncpy with memcpy to
8038         avoid -Wstringop-truncation.
8040 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8042         * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
8044 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8046         * predict.c (maybe_hot_frequency_p): Do not use cfun.
8048 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8050         * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
8051         merging.
8053 2017-11-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
8055         * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
8056         enable generating XXBRH if the value is in a vector register.
8057         (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
8058         value is in a vector register.
8059         (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
8060         register to register bswap64's instead of doing the GPR sequence
8061         used on previous machines.
8062         (bswapdi2_xxbrd): New insn.
8063         (bswapdi2_reg): Disallow on ISA 3.0.
8064         (register to register bswap64 splitter): Do not split the insn on
8065         ISA 3.0 systems that use XXBRD.
8067 2017-11-10  Martin Sebor  <msebor@redhat.com>
8069         PR c/81117
8070         * config/darwin-c.c (framework_construct_pathname): Replace strncpy
8071         with memcpy.
8072         (find_subframework_file): Same.
8074 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8076         * auto-profile.c (afdo_indirect_call): Drop frequency.
8077         * cgraph.c (symbol_table::create_edge): Drop frequency argument.
8078         (cgraph_node::create_edge): Drop frequency argument.
8079         (cgraph_node::create_indirect_edge): Drop frequency argument.
8080         (cgraph_edge::make_speculative): Drop frequency arguments.
8081         (cgraph_edge::resolve_speculation): Do not update frequencies
8082         (cgraph_edge::dump_edge_flags): Do not dump frequency.
8083         (cgraph_node::dump): Check consistency in IPA mode.
8084         (cgraph_edge::maybe_hot_p): Use IPA counter.
8085         (cgraph_edge::verify_count_and_frequency): Rename to ...
8086         (cgraph_edge::verify_count): ... this one; drop frequency checking.
8087         (cgraph_node::verify_node): Update.
8088         * cgraph.h (struct cgraph_edge): Drop frequency.
8089         (cgraph_edge::frequency): New function.
8090         * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
8091         frequencies.
8092         (cgraph_edge::rebuild_edges): Likewise.
8093         * cgraphclones.c (cgraph_edge::clone): Scale only counts.
8094         (duplicate_thunk_for_node): Do not pass frequency.
8095         (cgraph_node::create_clone): Scale only counts.
8096         (cgraph_node::create_virtual_clone): Do not pass frequency.
8097         (cgraph_node::create_edge_including_clones): Do not pass frequency.
8098         (cgraph_node::create_version_clone): Do not pass frequency.
8099         * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
8100         (cgraph_node::expand_thunk): Do not pass frequency.
8101         (cgraph_node::create_wrapper): Do not pass frequency.
8102         * gimple-iterator.c (update_call_edge_frequencies): Do not pass
8103         frequency.
8104         * gimple-streamer-in.c (input_bb): Scale only IPA counts.
8105         * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
8106         * ipa-cp.c (ipcp_lattice::print): Use frequency function.
8107         (gather_caller_stats): Use frequency function.
8108         (ipcp_cloning_candidate_p): Use frequency function.
8109         (ipcp_propagate_stage): Use frequency function.
8110         (get_info_about_necessary_edges): Use frequency function.
8111         (update_profiling_info): Update only IPA profile.
8112         (update_specialized_profile): Use frequency functoin.
8113         (perhaps_add_new_callers): Update only IPA profile.
8114         * ipa-devirt.c (ipa_devirt): Use IPA profile.
8115         * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
8116         (dump_ipa_call_summary): Use frequency function.
8117         (estimate_edge_size_and_time): Use frequency function.
8118         (ipa_merge_fn_summary_after_inlining): Use frequency function.
8119         * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
8120         * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
8121         (update_noncloned_counts): ... ths one; scale counts only.
8122         (clone_inlined_nodes): Do not scale frequency.
8123         (inline_call): Do not pass frequency.
8124         * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
8125         (compute_inlined_call_time): Use IPA profile.
8126         (want_inline_small_function_p): Use IPA profile.
8127         (want_inline_self_recursive_call_p): Use IPA profile.
8128         (edge_badness): Use IPA profile.
8129         (lookup_recursive_calls): Use IPA profile.
8130         (recursive_inlining): Do not pass frequency.
8131         (resolve_noninline_speculation): Do not update frequency.
8132         (inline_small_functions): Collect max of IPA profile.
8133         (dump_overall_stats): Dump IPA porfile.
8134         (dump_inline_stats): Dump IPA porfile.
8135         (ipa_inline): Collect IPA stats.
8136         * ipa-inline.h (clone_inlined_nodes): Update prototype.
8137         * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
8138         (ipa_propagate_frequency): Use frequency function.
8139         (ipa_profile): Cleanup.
8140         * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
8141         * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
8142         * lto-cgraph.c (lto_output_edge): Do not stream frequency.
8143         (input_node): Do not stream frequency.
8144         (input_edge): Do not stream frequency.
8145         (merge_profile_summaries): Scale only IPA profiles.
8146         * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
8147         * predict.c (drop_profile): Do not recompute frequency.
8148         * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
8149         (ipa_tm_insert_gettmclone_call): Do not pass frequency.
8150         * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
8151         * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
8152         * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
8153         * tree-inline.c (copy_bb): Do not scale frequency.
8154         (expand_call_inline): Do not scale frequency.
8155         (tree_function_versioning): Do not scale frequency.
8156         * ubsan.c (ubsan_create_edge): Do not pass frequency.
8158 2017-11-10  Julia Koval  <julia.koval@intel.com>
8160         * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
8161         (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
8162         (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
8163         (_mm256_maskz_gf2p8affine_epi64_epi8)
8164         (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
8165         (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
8166         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
8167         (__builtin_ia32_vgf2p8affineqb_v32qi)
8168         (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
8169         * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
8171 2017-11-10  Tamar Christina  <tamar.christina@arm.com>
8173         PR target/82641
8174         * config/arm/arm.c
8175         (arm_option_override): Refactor.
8176         (arm_option_reconfigure_globals): New.
8177         (arm_options_perform_arch_sanity_checks): New.
8178         * config/arm/arm-protos.h (arm_option_reconfigure_globals):
8179         New prototype.
8180         (arm_options_perform_arch_sanity_checks): Likewise
8182 2017-11-10  Pat Haugen  <pthaugen@us.ibm.com>
8184         * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
8185         (power9-qpmul): New.
8186         * rs6000/rs6000.md ("type" attr): Add qmul.
8187         (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
8188         *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
8189         *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
8191 2017-11-10  Martin Sebor  <msebor@redhat.com>
8193         PR c/81117
8194         * builtins.c (compute_objsize): Handle arrays that
8195         compute_builtin_object_size likes to fail for.  Make extern.
8196         * builtins.h (compute_objsize): Declare.
8197         (check_strncpy_sizes): New function.
8198         (expand_builtin_strncpy): Call check_strncpy_sizes.
8199         * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
8200         -Wstringop-truncation.
8201         (gimple_fold_builtin_strncat): Same.
8202         * gimple.c (gimple_build_call_from_tree): Set call location.
8203         * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
8204         (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
8205         (handle_builtin_stxncpy, handle_builtin_strncat): Same.
8206         (handle_builtin_strlen): Use strlen_to_stridx.
8207         (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
8208         stpncpy.
8209         Use strlen_to_stridx.
8210         (pass_strlen::execute): Release strlen_to_stridx.
8211         * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
8212         (-Wstringop-truncation): Document new option.
8214 2017-11-10  Martin Liska  <mliska@suse.cz>
8216         PR gcov-profile/82702
8217         * gcov.c (main): Handle intermediate files in a different
8218         way.
8219         (get_gcov_intermediate_filename): New function.
8220         (output_gcov_file): Remove support of intermediate files.
8221         (generate_results): Allocate intermediate file.
8222         (release_structures): Clean-up properly fn_end.
8223         (output_intermediate_file): Start iterating with line 1.
8225 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8227         PR tree-optimization/82929
8228         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8229         ops_swapped_p non-static data member.
8230         (store_immediate_info::store_immediate_info): Clear it.
8231         (imm_store_chain_info::coalesce_immediate_stores): If swapping
8232         ops set ops_swapped_p.
8233         (count_multiple_uses): Handle ops_swapped_p.
8235 2017-11-10  Martin Liska  <mliska@suse.cz>
8237         * coverage.c (coverage_init): Stream information about
8238         support of has_unexecuted_blocks.
8239         * doc/gcov.texi: Document that.
8240         * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8241         * gcov.c (read_graph_file): Likewise.
8242         (output_line_beginning): Fix a small issue with
8243         color output.
8245 2017-11-10  Bin Cheng  <bin.cheng@arm.com>
8247         * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8248         reference of trivial component.
8250 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8252         PR bootstrap/82916
8253         * gimple-ssa-store-merging.c
8254         (pass_store_merging::terminate_all_aliasing_chains): For
8255         gimple_store_p stmts also call refs_output_dependent_p.
8257         PR rtl-optimization/82913
8258         * compare-elim.c (try_merge_compare): Punt if def_insn is not
8259         single set.
8261 2017-11-09  Jeff Law  <law@redhat.com>
8263         * vr-values.h: New file with vr_values class.
8264         * tree-vrp.c: Include vr-values.h
8265         (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8266         data objects into the vr_values class.
8267         (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8268         (get_value_range): Make it a member function within vr_values class.
8269         (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8270         (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8271         (op_with_constant_singleton_value_range): Likewise.
8272         (extract_range_for_var_from_comparison_expr): Likewise.
8273         (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8274         (extract_range_from_binary_expr): Likewise.
8275         (extract_range_from_unary_expr): Likewise.
8276         (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8277         (check_for_binary_op_overflow, extract_range_basic): Likewise.
8278         (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8279         (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8280         (compare_name_with_value, compare_names): Likewise.
8281         (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8282         (vrp_evaluate_conditional_warnv_with_ops): Likewise.  Remove prototype.
8283         (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8284         (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8285         (extract_range_from_phi_node): Likewise.
8286         (simplify_truth_ops_using_ranges): Likewise.
8287         (simplify_div_or_mod_using_ranges): Likewise.
8288         (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8289         (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8290         (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8291         (simplify_float_conversion_using_ranges): Likewise.
8292         (simplify_internal_call_using_ranges): Likewise.
8293         (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8294         (vrp_visit_assignment_or_call): Likewise.  Smuggle class instance
8295         poitner via x_vr_values for calls into gimple folder.
8296         (vrp_initialize_lattice): Make this the vr_values ctor.
8297         (vrp_free_lattice): Make this the vr_values dtor.
8298         (set_vr_value): New function.
8299         (class vrp_prop): Add vr_values data member.  Add various member
8300         functions as well as member functions that delegate to vr_values.
8301         (check_array_ref): Make a member function within vrp_prop class.
8302         (search_for_addr_array, vrp_initialize): Likewise.
8303         (vrp_finalize): Likewise.  Revamp to avoid direct access to
8304         vr_value, values_propagated, etc.
8305         (check_array_bounds): Extract vrp_prop class instance pointer from
8306         walk info structure.  Use it to call member functions.
8307         (check_all_array_refs): Make a member function within vrp_prop class.
8308         Smuggle class instance pointer via walk info structure.
8309         (x_vr_values): New local static.
8310         (vrp_valueize): Use x_vr_values to get class instance.
8311         (vr_valueize_1): Likewise.
8312         (class vrp_folder): Add vr_values data member.  Add various member
8313         functions as well as member functions that delegate to vr_values.
8314         (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8315         (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8316         class instance from vr_values.  Use it to call member functions.
8317         (vrp_dom_walker): Add vr_values data member.
8318         (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8319         instance via x_vr_values.
8320         (identify_jump_threads): Accept vr_values as argument.  Store
8321         it into the walker structure.
8322         (evrp_dom_walker): Add vr_values class data member.  Add various
8323         delegators.
8324         (evrp_dom_walker::try_find_new_range): Use vr_values data
8325         member to access the memory allocator.
8326         (evrp_dom_walker::before_dom_children): Store vr_values class
8327         instance into the vrp_folder class.
8328         (evrp_dom_walker::push_value_range): Rework to avoid direct
8329         access to num_vr_values and vr_value.
8330         (evrp_dom_walker::pop_value_range): Likewise.
8331         (execute_early_vrp): Remove call to vrp_initialize_lattice.
8332         Use vr_values to get to dump_all_value_ranges member function.
8333         Remove call to vrp_free_lattice.  Call vrp_initialize, vrp_finalize,
8334         and simplify_cond_using_ranges_2 via vrp_prop class instance.
8335         Pass vr_values class instance down to identify_jump_threads.
8336         Remove call to vrp_free_lattice.
8337         (debug_all_value_ranges): Remove.
8339         * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8340         (vrp_folder): Likewise.
8342         * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8343         Get it from the existing bitmap instead.
8344         (vrp_intersect_ranges_1): Likewise.
8346 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8348         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8349         bit_not_p field.
8350         (store_immediate_info::store_immediate_info): Add bitnotp argument,
8351         set bit_not_p to it.
8352         (imm_store_chain_info::coalesce_immediate_stores): Break group
8353         if bit_not_p is different.
8354         (count_multiple_uses, split_group,
8355         imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8356         (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8357         (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8358         result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8359         to store_immediate_info ctor.
8361 2017-11-09  Jim Wilson  <jimw@sifive.com>
8363         * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8364         (scan_prog_file): Likewise.
8366 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8368         * bb-reorder.c (max_entry_frequency): Remove.
8369         (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8370         connect_traces, push_to_next_round_p): Remove prototypes.
8371         (find_traces_1_round): Use counts only.
8372         (push_to_next_round_p): Likewise.
8373         (find_traces): Likewise.
8374         (rotate_loop): Likewise.
8375         (find_traces_1_round): Likewise.
8376         (connect_traces): Likewise.
8377         (edge_order): Likewise.
8379 2017-11-09  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8381         * config/arm/arm.c (output_return_instruction): Add comments to
8382         indicate requirement for cmse_nonsecure_entry return to account
8383         for the size of clearing instruction output here.
8384         (thumb_exit): Likewise.
8385         * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8386         return in hardfloat mode.
8388 2017-11-09  Segher Boessenkool  <segher@kernel.crashing.org>
8390         * config/rs6000/rs6000.c (machine_function): Add a bool,
8391         "toc_is_wrapped_separately".
8392         (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8393         if it wasn't explicitly set or unset, we are optimizing for speed, and
8394         doing separate shrink-wrapping.
8395         (rs6000_get_separate_components): Enable the TOC component if
8396         saving the TOC register in the prologue.
8397         (rs6000_components_for_bb): Handle the TOC component.
8398         (rs6000_emit_prologue_components): Store the TOC register where needed.
8399         (rs6000_set_handled_components): Mark TOC as handled, if handled.
8400         (rs6000_emit_prologue): Don't save the TOC if that is already done.
8402 2017-11-09  Martin Jambor  <mjambor@suse.cz>
8404         * ipa-param-manipulation.c: New file.
8405         * ipa-param-manipulation.h: Likewise.
8406         * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8407         (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8408         * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8409         (ipa_parm_adjustment): Likewise.
8410         (ipa_parm_adjustment_vec): Likewise.
8411         (ipa_get_vector_of_formal_parms): Moved declaration to
8412         ipa-param-manipulation.h.
8413         (ipa_get_vector_of_formal_parm_types): Likewise.
8414         (ipa_modify_formal_parameters): Likewise.
8415         (ipa_modify_call_arguments): Likewise.
8416         (ipa_combine_adjustments): Likewise.
8417         (ipa_dump_param_adjustments): Likewise.
8418         (ipa_modify_expr): Likewise.
8419         (ipa_get_adjustment_candidate): Likewise.
8420         * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8421         ipa-param-manipulation.c.
8422         (ipa_get_vector_of_formal_parm_types): Likewise.
8423         (ipa_modify_formal_parameters): Likewise.
8424         (ipa_modify_call_arguments): Likewise.
8425         (ipa_modify_expr): Likewise.
8426         (get_ssa_base_param): Likewise.
8427         (ipa_get_adjustment_candidate): Likewise.
8428         (index_in_adjustments_multiple_times_p): Likewise.
8429         (ipa_combine_adjustments): Likewise.
8430         (ipa_dump_param_adjustments): Likewise.
8431         * tree-sra.c: Also include ipa-param-manipulation.h
8432         * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8433         ipa-param.h.
8435 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8436             Alan Hayward  <alan.hayward@arm.com>
8437             David Sherwood  <david.sherwood@arm.com>
8439         * doc/sourcebuild.texi (vect_masked_store): Document.
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_align_stack_vars): 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_variable_length): 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_unaligned_possible): 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_element_align_preferred): 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_perm_short, vect_perm_byte): Document
8470         previously undocumented selectors.
8471         (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8473 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8475         * doc/rtl.texi (const_vector): Say that elements can be
8476         const_wide_ints too.
8477         * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8478         * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8479         (gen_vec_duplicate): Use it instead of CONSTANT_P.
8480         * optabs.c (expand_vector_broadcast): Likewise.
8482 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8483             Alan Hayward  <alan.hayward@arm.com>
8484             David Sherwood  <david.sherwood@arm.com>
8486         * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8487         scaled index even if the unscaled address was invalid.
8488         Don't increase the complexity of using a scale in that case.
8490 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8491             Alan Hayward  <alan.hayward@arm.com>
8492             David Sherwood  <david.sherwood@arm.com>
8494         * doc/rtl.texi: Rewrite the subreg rules so that they partition
8495         the inner register into REGMODE_NATURAL_SIZE bytes rather than
8496         UNITS_PER_WORD bytes.
8497         * emit-rtl.c (validate_subreg): Divide subregs into blocks
8498         based on REGMODE_NATURAL_SIZE of the inner mode.
8499         (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8500         !SCALAR_FLOAT_MODE_P cases.  Use REGMODE_NATURAL_SIZE for the latter.
8501         * expmed.c (lowpart_bit_field_p): Divide the value up into
8502         chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8503         * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8504         whether something is likely to occupy more than one register.
8506 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8508         PR ipa/82879
8509         * ipa-inline-transform.c (update_noncloned_frequencies): Use
8510         profile_count::adjust_for_ipa_scaling.
8511         * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8512         * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8513         function.
8514         * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8516 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8518         * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8519         (split_group): Add total_orig and total_new arguments, estimate the
8520         number of statements related to the store group without store merging
8521         and with store merging.
8522         (imm_store_chain_info::output_merged_store): Adjust split_group
8523         callers, punt if estimated number of statements with store merging
8524         is not smaller than estimated number of statements without it.
8525         Formatting fix.
8526         (handled_load): Remove has_single_use checks.
8527         (pass_store_merging::process_store): Likewise.
8529 2017-11-09  Richard Biener  <rguenther@suse.de>
8531         PR tree-optimization/82902
8532         * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8534 2017-11-09  Martin Liska  <mliska@suse.cz>
8536         PR target/82863
8537         * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8538         uninitialized.
8540 2017-11-09  Martin Liska  <mliska@suse.cz>
8542         PR tree-optimization/82669
8543         * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8545 2017-11-09  Martin Liska  <mliska@suse.cz>
8547         PR gcov-profile/48463
8548         * coverage.c (coverage_begin_function): Output also end locus
8549         of a function and information whether the function is
8550         artificial.
8551         * gcov-dump.c (tag_function): Parse and print the information.
8552         * gcov.c (INCLUDE_MAP): Add include.
8553         (INCLUDE_SET): Likewise.
8554         (struct line_info): Move earlier in the source file because
8555         of vector<line_info> in function_info structure.
8556         (line_info::line_info): Likewise.
8557         (line_info::has_block): Likewise.
8558         (struct source_info): Add new member index.
8559         (source_info::get_functions_at_location): New function.
8560         (function_info::group_line_p): New function.
8561         (output_intermediate_line): New function.
8562         (output_intermediate_file): Use the mentioned function.
8563         (struct function_start): New.
8564         (struct function_start_pair_hash): Likewise.
8565         (process_file): Add code that identifies group functions.
8566         Assign lines either to global or function scope.
8567         (generate_results): Skip artificial functions.
8568         (find_source): Assign index for each source file.
8569         (read_graph_file): Read new flag artificial and end_line.
8570         (add_line_counts): Assign it either to global of function scope.
8571         (accumulate_line_counts): Isolate core of the function to
8572         accumulate_line_info and call it for both function and global
8573         scope lines.
8574         (accumulate_line_info): New function.
8575         (output_line_beginning): Fix GNU coding style.
8576         (print_source_line): New function.
8577         (output_line_details): Likewise.
8578         (output_function_details): Likewise.
8579         (output_lines): Iterate both source (global) scope and function
8580         scope.
8581         (struct function_line_start_cmp): New class.
8582         * doc/gcov.texi: Reflect changes in documentation.
8584 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8586         PR debug/82837
8587         * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8588         (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8589         and similarly for not instead of neg.
8591 2017-11-08  Andi Kleen  <ak@linux.intel.com>
8593         * config/i386/i386.opt: Add -mforce-indirect-call.
8594         * config/i386/predicates.md: Check for flag_force_indirect_call.
8595         * doc/invoke.texi: Document -mforce-indirect-call
8597 2017-11-08  Kito Cheng  <kito.cheng@gmail.com>
8599         * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8600         New extern.
8601         (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8602         config/riscv/riscv.c (predict.h): New include.
8603         (riscv_slow_unaligned_access_p): No longer static.
8604         (riscv_block_move_straight): Add require.
8605         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8607 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8609         PR target/82855
8610         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8611         <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8612         nonimmediate_operand predicate for operand 1 instead of
8613         register_operand.
8615 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8617         * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8618         New pattern.
8619         * config/aarch64/constraints.md (Uml): New constraint.
8620         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8621         predicate.
8623 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8625         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8626         of two vec_duplicates into a vec_concat.
8628 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8630         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8631         Simplify vec_merge of vec_duplicate and vec_concat.
8632         * config/aarch64/constraints.md (Utq): New constraint.
8633         * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8634         define_insn.
8636 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8638         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8639         Simplify vec_merge of vec_duplicate and const_vector.
8640         * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8641         New predicate.
8642         * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8643         mode iterator.  Update predicate on operand 1 to
8644         handle non-const_vec constants.  Delete constraints.
8645         (*aarch64_combinez_be<mode>): Likewise for operand 2.
8647 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8649         PR tree-optimization/78821
8650         * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8651         data member.
8652         (store_operand_info::store_operand_info): Initialize it to false.
8653         (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8654         ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8655         store in the group, and if chain_info is non-NULL, to ignore altogether
8656         that chain.
8657         (compatible_load_p): Fail if bit_not_p does not match.
8658         (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8659         (handled_load): Fill in bit_not_p.  Handle BIT_NOT_EXPR.
8660         (pass_store_merging::process_store): Adjust
8661         terminate_all_aliasing_chains calls to pass NULL in all current spots,
8662         call terminate_all_aliasing_chains newly when adding a store into
8663         a chain with non-NULL chain_info.
8665 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8667         * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8669 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8671         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8672         Remove.
8673         (aarch64_layout_frame): Initialise emit_frame_chain.
8674         (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8675         (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8677 2017-11-08  Martin Liska  <mliska@suse.cz>
8679         * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8680         of gimple_call_internal_p.
8682 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8684         * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8686 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8688         * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8689         "do {} while (0)".
8691 2017-11-08  Martin Liska  <mliska@suse.cz>
8693         PR sanitizer/82792
8694         * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8696 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8698         * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8700 2017-11-07  Jakub Jelinek  <jakub@redhat.com>
8702         PR target/82855
8703         * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8704         (*cmp<mode>_ccz_1): New insn with $k alternative.
8706         PR target/82855
8707         * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8708         RTX_COMM_COMPARE as commutative as well.
8709         (ix86_binary_operator_ok): Formatting fix.
8710         * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8711         *<code><mode>3<mask_name><round_saeonly_name>,
8712         *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8713         *<s>mul<mode>3_highpart<mask_name>,
8714         *vec_widen_umult_even_v16si<mask_name>,
8715         *vec_widen_umult_even_v8si<mask_name>,
8716         *vec_widen_umult_even_v4si<mask_name>,
8717         *vec_widen_smult_even_v16si<mask_name>,
8718         *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8719         *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8720         *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8721         *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8722         *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8723         <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8724         *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8725         *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8726         *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8727         !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8728         ix86_binary_operator_ok.  Formatting fixes.
8729         (*<plusminus_insn><mode>3<mask_name><round_name>,
8730         *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8731         fixes.
8733 2017-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
8735         * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8736         ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8737         eq and ne if TARGET_ISEL.
8738         (cmp): New code_iterator.
8739         (UNS, UNSU_, UNSIK): New code_attrs.
8740         (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8741         ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8742         to...
8743         ("eq<mode>3"): ... this.
8744         ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8745         to...
8746         ("ne<mode>3"): ... this.
8748 2017-11-07  Julia Koval  <julia.koval@intel.com>
8750         PR target/82812
8751         * common/config/i386/i386-common.c
8752         (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8753         (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8754         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8755         * config/i386/i386.opt: Ditto.
8756         * config/i386/i386.c (ix86_target_string): Ditto.
8757         (ix86_option_override_internal): Ditto.
8758         (ix86_init_mpx_builtins): Move MPX to args2.
8759         (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8760         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8761         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8762         __builtin_ia32_vgf2p8affineinvqb_v32qi,
8763         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8764         __builtin_ia32_vgf2p8affineinvqb_v16qi,
8765         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8767 2017-11-07  Uros Bizjak  <ubizjak@gmail.com>
8769         PR target/80425
8770         * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8771         and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8772         (zero-extendsidi peephole2): Remove peephole.
8774 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8776         PR c/53037
8777         * stor-layout.c: Include attribs.h.
8778         (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8779         explicit lookup of "aligned" attribute.
8781 2017-11-07  Andrew Waterman  <andrew@sifive.com>
8783         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8784         (riscv_expand_block_move): Likewise.
8785         * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8786         implementation.
8787         (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8788         (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8789         * config/riscv/riscv.c (riscv_block_move_straight): New function.
8790         (riscv_adjust_block_mem): Likewise.
8791         (riscv_block_move_loop): Likewise.
8792         (riscv_expand_block_move): Likewise.
8793         * config/riscv/riscv.md (movmemsi): New pattern.
8795 2017-11-07  Michael Clark  <michaeljclark@mac.com>
8797         * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8798         (MUSL_DYNAMIC_LINKER): Likewise.
8800 2017-11-07  Richard Sandiford  <richard.sandiford@linaro.org>
8802         * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8803         (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8804         nonmmory_operand.
8806 2017-11-07  Richard Biener  <rguenther@suse.de>
8808         * match.pd: Fix build.
8810 2017-11-07  Wilco Dijkstra  <wdijkstr@arm.com>
8811             Jackson Woodruff  <jackson.woodruff@arm.com>
8813         PR tree-optimization/71026
8814         * match.pd: Canonicalize negate in division.
8816 2017-11-07  Sudakshina Das  <sudi.das@arm.com>
8818         PR middle-end/80131
8819         * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8821 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8823         * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8824         (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8826 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8828         * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8829         non-scalar integral types.
8830         * match.pd (negate_expr_p): Handle MINUS_EXPR.
8831         (-(A-B), -(~A)): New transformations.
8833 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8835         * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8836         semicolon after "do {} while (0)".
8837         * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8838         * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8839         * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8840         * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8841         * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8843 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8845         * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8846         after "do {} while (0)".
8847         * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8848         * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8849         * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8850         * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8851         * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8853 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8855         PR other/82784
8856         * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8857         "while {} do (0)".
8858         (arm_rtx_costs_internal): Add missing semicolon after
8859         HANDLE_NARROW_SHIFT_ARITH call.
8861 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8863         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8864         disable isel if it was not set explicitly.
8866 2017-11-06  James Bowman  <james.bowman@ftdichip.com>
8868         * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8869         (add_type_attribute) likewise.
8871 2017-11-06  H.J. Lu  <hongjiu.lu@intel.com>
8873         * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8874         of struct ix86_frame.
8875         (ix86_initial_elimination_offset): Likewise.
8876         (ix86_expand_split_stack_prologue): Likewise.
8878 2017-11-06  Marc Glisse  <marc.glisse@inria.fr>
8880         * tree-vrp.h (enum value_range_type): Update stale comment.
8882 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8883             Alan Hayward  <alan.hayward@arm.com>
8884             David Sherwood  <david.sherwood@arm.com>
8886         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8887         (aarch64_expand_vec_perm_const): Take the number of units too.
8888         * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8889         (aarch64_expand_vec_perm_const): Likewise.
8890         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8891         (vec_perm<mode>): Update accordingly.
8893 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8894             Alan Hayward  <alan.hayward@arm.com>
8895             David Sherwood  <david.sherwood@arm.com>
8897         * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8898         Take the number of units too.
8899         * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8900         (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8901         but check for a vector mode before rather than after the call.
8902         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8903         (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8904         (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8905         (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8906         (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8907         (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8908         (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8909         (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8910         (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8911         (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
8912         (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
8913         (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
8914         (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
8915         (aarch64_sqdmull2_n<mode>): Update accordingly.
8917 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8918             Alan Hayward  <alan.hayward@arm.com>
8919             David Sherwood  <david.sherwood@arm.com>
8921         * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
8922         the number of units too.
8923         * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
8924         * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
8925         (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
8926         (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
8927         (vec_store_lanesxi<mode>): Update accordingly.
8929 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8930             Alan Hayward  <alan.hayward@arm.com>
8931             David Sherwood  <david.sherwood@arm.com>
8933         * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
8934         * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
8935         * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
8936         of units rather than the mode.
8937         * config/aarch64/iterators.md (nunits): New mode attribute.
8938         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
8939         Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
8940         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
8941         (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
8942         (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
8943         (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
8944         (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
8945         (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
8946         (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
8947         (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
8948         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
8949         (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
8950         (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
8951         (reduc_<maxmin_uns>_scal_<mode>): Likewise.
8952         (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
8953         (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
8954         (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
8955         (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
8956         (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
8957         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
8958         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
8959         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
8960         (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
8961         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
8962         (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
8963         (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
8964         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
8965         (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
8966         (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
8967         (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
8968         (aarch64_vec_load_lanesci_lane<mode>): Likewise.
8969         (aarch64_vec_store_lanesci_lane<mode>): Likewise.
8970         (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
8971         (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
8972         (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
8973         (aarch64_simd_vec_setv2di): Likewise.
8975 2017-11-06  Carl Love  <cel@us.ibm.com>
8977         * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
8978         definitions.
8979         (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
8980         * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
8981         to power 8.
8982         * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
8983         extern declaration.
8984         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
8985         * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
8986         Add power 8 macro expansions.
8987         (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
8988         * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
8989         power 8 instructions.  (VSX_XXBR): Add iterator.
8991 2017-11-06  Wilco Dijkstra  <wdijkstr@arm.com>
8993         * config/arm/arm.md (predicable_short_it): Change default to "no",
8994         improve documentation, remove uses that are identical to the default.
8995         (enabled_for_depr_it): Rename to enabled_for_short_it.
8996         * gcc/config/arm/arm-fixed.md (predicable_short_it):
8997         Remove default uses.
8998         * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
8999         * gcc/config/arm/sync.md (predicable_short_it): Likewise.
9000         * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
9001         * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
9003 2017-11-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
9005         PR target/82748
9006         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
9007         float128 helper macros, which are no longer used after deleting
9008         the old 'q' built-in functions, and moving the round to odd
9009         built-in functions to being special built-in functions.
9010         (BU_FLOAT128_2): Likewise.
9011         (BU_FLOAT128_1_HW): Likewise.
9012         (BU_FLOAT128_2_HW): Likewise.
9013         (BU_FLOAT128_3_HW): Likewise.
9014         (FABSQ): Delete old 'q' built-in functions.
9015         (COPYSIGNQ): Likewise.
9016         (SQRTF128_ODD): Move round to odd built-in functions to be
9017         special built-in functions, so that we can handle
9018         -mabi=ieeelongdouble.
9019         (TRUNCF128_ODD): Likewise.
9020         (ADDF128_ODD): Likewise.
9021         (SUBF128_ODD): Likewise.
9022         (MULF128_ODD): Likewise.
9023         (DIVF128_ODD): Likewise.
9024         (FMAF128_ODD): Likewise.
9025         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
9026         built-in names to 'f128'.
9027         * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
9028         old 'q' built-in functions, as the machine independent code for
9029         'f128' built-in functions handles this.
9030         (rs6000_expand_builtin): Add expansion for float128 round to odd
9031         functions, keying off on -mabi=ieeelongdouble of whether to use
9032         the KFmode or TFmode variant.
9033         (rs6000_init_builtins): Initialize the _Float128 round to odd
9034         built-in functions.
9035         * doc/extend.texi (PowerPC Built-in Functions): Document the old
9036         _Float128 'q' built-in functions are now mapped into the new
9037         'f128' built-in functions.
9039 2017-11-06  David Edelsohn  <dje.gcc@gmail.com>
9041         * collect2.c (add_lto_object): Compile for OBJECT_COFF.
9042         (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
9044 2017-11-06  David Malcolm  <dmalcolm@redhat.com>
9046         PR jit/82826
9047         * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
9048         * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
9049         * toplev.c: Include "ipa-fnsummary.h".
9050         (toplev::finalize): Call ipa_fnsummary_c_finalize.
9052 2017-11-06  Jakub Jelinek  <jakub@redhat.com>
9054         PR tree-optimization/82838
9055         * gimple-ssa-store-merging.c
9056         (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
9057         on a separate gimple_seq which is then appended to seq.
9059 2017-11-06  Jeff Law  <law@redhat.com>
9061         PR target/82788
9062         * config/i386/i386.c (PROBE_INTERVAL): Remove.
9063         (get_probe_interval): New functions.
9064         (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
9065         (ix86_adjust_stack_and_probe): Likewise.
9066         (output_adjust_stack_and_probe): Likewise.
9067         (ix86_emit_probe_stack_range): Likewise.
9068         (ix86_expand_prologue): Likewise.
9070 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9072         PR tree-optimization/82816
9073         * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
9074         if the modes of the two types are the same.
9075         (convert_plusminus_to_widen): Likewise.
9077 2017-11-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9079         * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
9080         p9_vadu<mode>3.
9081         (usadv16qi): New define_expand.
9082         (usadv8hi): New define_expand.
9084 2017-11-06  Jan Hubicka  <hubicka@ucw.cz>
9086         PR bootstrap/82832
9087         * ipa-inline-transform.c (update_noncloned_frequencies): Always
9088         scale.
9089         (inline_transform): Likewise.
9090         * predict.c (counts_to_freqs): Remove useless conditional.
9091         * profile-count.h (profile_count::apply_scale): Move sanity check.
9092         * tree-inline.c (copy_bb): Always scale.
9093         (copy_cfg_body): Likewise.
9095 2017-11-06  Christophe Lyon  <christophe.lyon@linaro.org>
9097         PR target/67591
9098         * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
9099         attribute.
9100         (*cmp_ite0): Add enabled_for_depr_it attribute.
9101         (*cmp_ite1): Likewise.
9103 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
9105         * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
9106         TYPE_MFCRF.
9108 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9110         * tree-vrp.c (vrp_int_const_binop): Return true on success and
9111         return the value by pointer.
9112         (extract_range_from_multiplicative_op_1): Update accordingly.
9113         Return as soon as an operation fails.
9115 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9117         PR other/82784
9118         * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
9119         (DEF_SANITIZER_BUILTIN): ... here.
9120         (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
9121         DEF_SANITIZER_BUILTIN in if stmt.  Add missing semicolon.
9123 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9125         PR other/82784
9126         * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
9127         macro body.
9128         (ASM_OUTPUT_CASE_LABEL): Add semicolon after
9129         ASM_OUTPUT_BEFORE_CASE_LABEL call.
9130         * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
9131         after macro body.
9132         * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9133         * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9134         * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9136 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9138         PR other/82784
9139         * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
9140         "do {} while (0)".
9142 2017-11-04  Michael Clark  <michaeljclark@mac.com>
9144         * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
9145         config/riscv/riscv.md (addsi3): Use 'i' for immediates.
9146         (adddi3): Likewise.
9147         (*addsi3_extended): Likewise.
9148         (*addsi3_extended2): Likewise.
9149         (<optab>si3): Likewise.
9150         (<optab>di3): Likewise.
9151         (<optab><mode>3): Likewise.
9152         (<*optabe>si3_internal): Likewise.
9153         (zero_extendqi<SUPERQI:mode>2): Likewise.
9154         (*add<mode>hi3): Likewise.
9155         (*xor<mode>hi3): Likewise.
9156         (<optab>di3): Likewise.
9157         (*<optab>si3_extend): Likewise.
9158         (*sge<u>_<X:mode><GPR:mode>): Likewise.
9159         (*slt<u>_<X:mode><GPR:mode>): Likewise.
9160         (*sle<u>_<X:mode><GPR:mode>): Likewise.
9162 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9164         * config/riscv/riscv.c (riscv_option_override): Conditionally set
9165         TARGET_STRICT_ALIGN based upon -mtune argument.
9167 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9169         * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
9171 2017-11-04  Daniel Santos  <daniel.santos@pobox.com>
9173         * config/i386/i386.c (choose_basereg): Use optional scratch
9174         register and add assertion.
9175         (x86_emit_outlined_ms2sysv_save): Use scratch register when
9176         needed, and don't allocate stack.
9177         (ix86_expand_prologue): Rearrange where SSE saves/stub call is
9178         emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
9179         (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
9181 2017-11-03  Jeff Law  <law@redhat.com>
9183         * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
9184         (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
9185         to probe at the start of a noreturn function.
9187 2017-11-03  Jakub Jelinek  <jakub@redhat.com>
9189         PR tree-optimization/78821
9190         * gimple-ssa-store-merging.c: Update the file comment.
9191         (MAX_STORE_ALIAS_CHECKS): Define.
9192         (struct store_operand_info): New type.
9193         (store_operand_info::store_operand_info): New constructor.
9194         (struct store_immediate_info): Add rhs_code and ops data members.
9195         (store_immediate_info::store_immediate_info): Add rhscode, op0r
9196         and op1r arguments to the ctor, initialize corresponding data members.
9197         (struct merged_store_group): Add load_align_base and load_align
9198         data members.
9199         (merged_store_group::merged_store_group): Initialize them.
9200         (merged_store_group::do_merge): Update them.
9201         (merged_store_group::apply_stores): Pick the constant for
9202         encode_tree_to_bitpos from one of the two operands, or skip
9203         encode_tree_to_bitpos if neither operand is a constant.
9204         (class pass_store_merging): Add process_store method decl.  Remove
9205         bool argument from terminate_all_aliasing_chains method decl.
9206         (pass_store_merging::terminate_all_aliasing_chains): Remove
9207         var_offset_p argument and corresponding handling.
9208         (stmts_may_clobber_ref_p): New function.
9209         (compatible_load_p): New function.
9210         (imm_store_chain_info::coalesce_immediate_stores): Terminate group
9211         if there is overlap and rhs_code is not INTEGER_CST.  For
9212         non-overlapping stores terminate group if rhs is not mergeable.
9213         (get_alias_type_for_stmts): Change first argument from
9214         auto_vec<gimple *> & to vec<gimple *> &.  Add IS_LOAD, CLIQUEP and
9215         BASEP arguments.  If IS_LOAD is true, look at rhs1 of the stmts
9216         instead of lhs.  Compute *CLIQUEP and *BASEP in addition to the
9217         alias type.
9218         (get_location_for_stmts): Change first argument from
9219         auto_vec<gimple *> & to vec<gimple *> &.
9220         (struct split_store): Remove orig_stmts data member, add orig_stores.
9221         (split_store::split_store): Create orig_stores rather than orig_stmts.
9222         (find_constituent_stmts): Renamed to ...
9223         (find_constituent_stores): ... this.  Change second argument from
9224         vec<gimple *> * to vec<store_immediate_info *> *, push pointers
9225         to info structures rather than the statements.
9226         (split_group): Rename ALLOW_UNALIGNED argument to
9227         ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
9228         it.  Adjust find_constituent_stores caller.
9229         (imm_store_chain_info::output_merged_store): Handle rhs_code other
9230         than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
9231         get_location_for_stmts callers.  Set MR_DEPENDENCE_CLIQUE and
9232         MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
9233         (mem_valid_for_store_merging): New function.
9234         (handled_load): New function.
9235         (pass_store_merging::process_store): New method.
9236         (pass_store_merging::execute): Use process_store method.  Adjust
9237         terminate_all_aliasing_chains caller.
9239 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9241         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9242         Return true for more constants, symbols and label references.
9243         (aarch64_valid_floating_const): Remove unused function.
9245 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9247         PR target/82786
9248         * config/aarch64/aarch64.c (aarch64_layout_frame):
9249         Undo forcing of LR at bottom of frame.
9251 2017-11-03  Jeff Law  <law@redhat.com>
9253         PR target/82823
9254         * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9255         for int_registers_saved.
9257         * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9258         extracted from tree-ssa-dom.c.
9259         * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9260         * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9261         (record_equivalences_from_incoming_edge): Add additional argument
9262         to single_pred_edge_ignoring_loop_edges call.
9263         * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9264         (uncprop_dom_walker::before_dom_children): Add additional argument
9265         to single_pred_edge_ignoring_loop_edges call.
9266         * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9267         single_pred_edge_ignoring_loop_edges rather than open coding.
9268         * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9270 2017-11-03  Marc Glisse  <marc.glisse@inria.fr>
9272         * match.pd (-(-A)): Rewrite.
9274 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9276         * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9277         (rs6000_emit_int_cmove): New declaration.
9278         * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9279         (rs6000_emit_sISEL): Delete.
9280         (rs6000_emit_int_cmove): Make non-static.
9281         * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9282         instead of rs6000_emit_sISEL.
9284 2017-11-03  Jan Hubicka  <hubicka@ucw.cz>
9286         * asan.c (create_cond_insert_point): Maintain profile.
9287         * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9288         merged.
9289         * basic-block.h (struct basic_block_def): Remove frequency.
9290         (EDGE_FREQUENCY): Use to_frequency
9291         * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9292         heuristics.
9293         (find_traces): Update to use to_frequency.
9294         (find_traces_1_round): Likewise; use only IPA counts.
9295         (bb_to_key): Likewise.
9296         (connect_traces): Use IPA counts only.
9297         (copy_bb_p): Update to use to_frequency.
9298         (fix_up_crossing_landing_pad): Likewise.
9299         (sanitize_hot_paths): Likewise.
9300         * bt-load.c (basic_block_freq): Likewise.
9301         * cfg.c (init_flow): Set count_max to uninitialized.
9302         (check_bb_profile): Remove frequencies; check counts.
9303         (dump_bb_info): Do not dump frequencies.
9304         (update_bb_profile_for_threading): Update counts only.
9305         (scale_bbs_frequencies_int): Likewise.
9306         (MAX_SAFE_MULTIPLIER): Remove.
9307         (scale_bbs_frequencies_gcov_type): Update counts only.
9308         (scale_bbs_frequencies_profile_count): Update counts only.
9309         (scale_bbs_frequencies): Update counts only.
9310         * cfg.h (struct control_flow_graph): Add count-max.
9311         (update_bb_profile_for_threading): Update prototype.
9312         * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9313         (find_many_sub_basic_blocks): Likewise.
9314         * cfgcleanup.c (try_forward_edges): Likewise.
9315         (try_crossjump_to_edge): Likewise.
9316         * cfgexpand.c (expand_gimple_cond): Likewise.
9317         (expand_gimple_tailcall): Likewise.
9318         (construct_init_block): Likewise.
9319         (construct_exit_block): Likewise.
9320         * cfghooks.c (verify_flow_info): Check consistency of counts.
9321         (dump_bb_for_graph): Do not dump frequencies.
9322         (split_block_1): Do not update frequencies.
9323         (split_edge): Do not update frequencies.
9324         (make_forwarder_block): Do not update frequencies.
9325         (duplicate_block): Do not update frequencies.
9326         (account_profile_record): Do not update frequencies.
9327         * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9328         for global heuristics.
9329         * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9330         (expected_loop_iterations_unbounded): Use counts only.
9331         * cfgloopmanip.c (scale_loop_profile): Simplify.
9332         (create_empty_loop_on_edge): Simplify
9333         (loopify): Simplify
9334         (duplicate_loop_to_header_edge): Simplify
9335         * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9336         (update_br_prob_note): Take care of removing note when profile
9337         becomes undefined.
9338         (relink_block_chain): Do not dump frequency.
9339         (rtl_account_profile_record): Use to_frequency.
9340         * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9341         (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9342         (cgraph_update_edges_for_call_stmt_node): Likewise.
9343         (cgraph_edge::verify_count_and_frequency): Update.
9344         (cgraph_node::verify_node): Temporarily disable frequency verification.
9345         * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9346         to_cgraph_frequency.
9347         (cgraph_edge::rebuild_edges): Convert to ipa counts.
9348         * cgraphunit.c (init_lowered_empty_function): Do not initialize
9349         frequencies.
9350         (cgraph_node::expand_thunk): Update profile.
9351         * except.c (dw2_build_landing_pads): Do not update frequency.
9352         * final.c (compute_alignments): Use to_frequency.
9353         (dump_basic_block_info): Do not dump frequency.
9354         * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9355         (dump_gimple_bb_header): Do not dump frequency.
9356         * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9357         do update count.
9358         * gimple-streamer-in.c (input_bb): Do not stream frequency.
9359         * gimple-streamer-out.c (output_bb): Do not stream frequency.
9360         * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9361         (init_before_recovery): Do not update frequency.
9362         (sched_create_recovery_edges): Do not update frequency.
9363         * hsa-gen.c (convert_switch_statements): Do not update frequency.
9364         * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9365         (ipa_cp_c_finalize): Set max_count to uninitialized.
9366         * ipa-fnsummary.c (get_minimal_bb): Use counts.
9367         (param_change_prob): Use counts.
9368         * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9369         local profiles.
9370         * ipa-split.c (consider_split): Use to_frequency.
9371         (split_function): Use to_frequency.
9372         * ira-build.c (loop_compare_func): Likewise.
9373         (mark_loops_for_removal): Likewise.
9374         (mark_all_loops_for_removal): Likewise.
9375         * loop-doloop.c (doloop_modify): Do not update frequency.
9376         * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9377         frequency.
9378         * lto-streamer-in.c (input_function): Update count_max.
9379         * omp-expand.c (expand_omp_taskreg): Update count_max.
9380         * omp-simd-clone.c (simd_clone_adjust): Update profile.
9381         * predict.c (maybe_hot_frequency_p): Use to_frequency.
9382         (maybe_hot_count_p): Use ipa counts only.
9383         (maybe_hot_bb_p): Simplify.
9384         (maybe_hot_edge_p): Simplify.
9385         (probably_never_executed): Do not take frequency argument.
9386         (probably_never_executed_bb_p): Do not pass frequency.
9387         (probably_never_executed_edge_p): Likewise.
9388         (combine_predictions_for_bb): Check that profile is nonzero.
9389         (propagate_freq): Do not set frequency.
9390         (drop_profile): Simplify.
9391         (counts_to_freqs): Simplify.
9392         (expensive_function_p): Use to_frequency.
9393         (propagate_unlikely_bbs_forward): Simplify.
9394         (determine_unlikely_bbs): Simplify.
9395         (estimate_bb_frequencies): Add hack to silence graphite issues.
9396         (compute_function_frequency): Use ipa counts.
9397         (pass_profile::execute): Update.
9398         (rebuild_frequencies): Use counts only.
9399         (force_edge_cold): Use counts only.
9400         * profile-count.c (profile_count::dump): Dump new count types.
9401         (profile_count::differs_from_p): Check compatiblity.
9402         (profile_count::to_frequency): New function.
9403         (profile_count::to_cgraph_frequency): New function.
9404         * profile-count.h (struct function): Declare.
9405         (enum profile_quality): Add profile_guessed_local and
9406         profile_guessed_global0.
9407         (class profile_proability): Decrease number of bits to 29;
9408         update from_reg_br_prob_note and to_reg_br_prob_note.
9409         (class profile_count: Update comment; decrease number of bits
9410         to 61. Check compatibility.
9411         (profile_count::compatible_p): New private member function.
9412         (profile_count::ipa_p): New member function.
9413         (profile_count::operator<): Handle global zero correctly.
9414         (profile_count::operator>): Handle global zero correctly.
9415         (profile_count::operator<=): Handle global zero correctly.
9416         (profile_count::operator>=): Handle global zero correctly.
9417         (profile_count::nonzero_p): New member function.
9418         (profile_count::force_nonzero): New member function.
9419         (profile_count::max): New member function.
9420         (profile_count::apply_scale): Handle IPA scalling.
9421         (profile_count::guessed_local): New member function.
9422         (profile_count::global0): New member function.
9423         (profile_count::ipa): New member function.
9424         (profile_count::to_frequency): Declare.
9425         (profile_count::to_cgraph_frequency): Declare.
9426         * profile.c (OVERLAP_BASE): Delete.
9427         (compute_frequency_overlap): Delete.
9428         (compute_branch_probabilities): Do not use compute_frequency_overlap.
9429         * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9430         * sched-ebb.c (rank): Use counts only.
9431         * shrink-wrap.c (handle_simple_exit): Use counts only.
9432         (try_shrink_wrapping): Use counts only.
9433         (place_prologue_for_one_component): Use counts only.
9434         * tracer.c (find_best_predecessor): Use to_frequency.
9435         (find_trace): Use to_frequency.
9436         (tail_duplicate): Use to_frequency.
9437         * trans-mem.c (expand_transaction): Do not update frequency.
9438         * tree-call-cdce.c: Do not update frequency.
9439         * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9440         (gimple_merge_blocks): Likewise.
9441         (gimple_split_edge): Likewise.
9442         (gimple_duplicate_sese_region): Likewise.
9443         (gimple_duplicate_sese_tail): Likewise.
9444         (move_sese_region_to_fn): Likewise.
9445         (gimple_account_profile_record): Likewise.
9446         (insert_cond_bb): Likewise.
9447         * tree-complex.c (expand_complex_div_wide): Likewise.
9448         * tree-eh.c (lower_resx): Update profile.
9449         * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9450         frequencies.
9451         (initialize_cfun): Do not initialize frequencies
9452         (freqs_to_counts): Delete.
9453         (copy_cfg_body): Ignore count parameter.
9454         (copy_body): Update.
9455         (expand_call_inline): Update count_max.
9456         (optimize_inline_calls): Update count_max.
9457         (tree_function_versioning): Update count_max.
9458         * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9459         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9460         frequency.
9461         * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9462         * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9463         (try_peel_loop): Likewise.
9464         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9465         to_frequency.
9466         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9467         (tree_transform_and_unroll_loop): Do not use frequencies
9468         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9469         Use reliable prediction only.
9470         * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9471         * tree-ssa-sink.c (select_best_block): Use to_frequency.
9472         * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9473         probability scaling.
9474         * tree-ssa-threadupdate.c (create_block_for_threading): Do
9475         not update frequency
9476         (any_remaining_duplicated_blocks): Likewise.
9477         (update_profile): Likewise.
9478         (estimated_freqs_path): Delete.
9479         (freqs_to_counts_path): Delete.
9480         (clear_counts_path): Delete.
9481         (ssa_fix_duplicate_block_edges): Likewise.
9482         (duplicate_thread_path): Likewise.
9483         * tree-switch-conversion.c (gen_inbound_check): Use counts.
9484         * tree-tailcall.c (decrease_profile): Do not update frequency.
9485         (eliminate_tail_call): Likewise.
9486         * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9487         * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9488         (optimize_mask_stores): Likewise.
9489         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9490         * ubsan.c (ubsan_expand_null_ifn): Update profile.
9491         (ubsan_expand_ptr_ifn): Update profile.
9492         * value-prof.c (gimple_ic): Simplify.
9493         * value-prof.h (gimple_ic): Update prototype.
9494         * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9495         * ipa-inline.c (compute_uninlined_call_time): Be sure that
9496         counts are nonzero.
9497         (want_inline_self_recursive_call_p): Likewise.
9498         (resolve_noninline_speculation): Only cummulate defined counts.
9499         (inline_small_functions): Use nonzero_p.
9500         (ipa_inline): Do not access freed node.
9502 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9504         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9505         Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9507 2017-11-03  Kito Cheng  <kito.cheng@gmail.com>
9509         * config/riscv/riscv.c (riscv_legitimize_move): Handle
9510         non-legitimate address.
9512 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9514         * config/rs6000/rs6000.md (*lt0_disi): Delete.
9515         (*lt0_<mode>di, *lt0_<mode>si): New.
9517 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9519         * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9520         TARGET_PAIRED_FLOAT.
9522 2017-11-03  Siddhesh Poyarekar  <siddhesh.poyarekar@linaro.org>
9523             Jim Wilson  <jim.wilson@linaro.org>
9525         * config/aarch64/aarch64-cores.def (saphira): New CPU.
9526         * config/aarch64/aarch64-tune.md: Regenerated.
9527         * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9528         * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9530 2017-11-03  Cupertino Miranda  <cmiranda@synopsys.com>
9532         * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9533         (arc_expand_prologue): Restore blink for millicode.
9534         * config/arc/linux.h (LINK_EH_SPEC): Defined.
9536 2017-11-03  Richard Sandiford  <richard.sandiford@linaro.org>
9538         PR target/82809
9539         * config/i386/i386.c (ix86_vector_duplicate_value): Use
9540         gen_vec_duplicate after forcing the scalar into a register.
9542 2017-11-02  Segher Boessenkool  <segher@kernel.crashing.org>
9544         * combine (try_combine): Print the insns input to try_combine to the
9545         dump file.
9547 2017-11-02  Steve Ellcey  <sellcey@cavium.com>
9549         PR target/79868
9550         * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9551         Remove second argument from aarch64_process_target_attr call.
9552         * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9553         Ditto.
9554         * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9555         field type.
9556         (aarch64_handle_attr_arch): Remove second argument.
9557         (aarch64_handle_attr_cpu): Ditto.
9558         (aarch64_handle_attr_tune): Ditto.
9559         (aarch64_handle_attr_isa_flags): Ditto.
9560         (aarch64_process_one_target_attr): Ditto.
9561         (aarch64_process_target_attr): Ditto.
9562         (aarch64_option_valid_attribute_p): Remove second argument.
9563         on aarch64_process_target_attr call.
9565 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9567         * diagnostic.c: Include "selftest-diagnostic.h".
9568         (selftest::assert_location_text): New function.
9569         (selftest::test_diagnostic_get_location_text): New function.
9570         (selftest::diagnostic_c_tests): Call it.
9572 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9574         * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9575         * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9576         (class selftest::test_diagnostic_context): Move to...
9577         * selftest-diagnostic.c: New file.
9578         * selftest-diagnostic.h: New file.
9580 2017-11-02  James Bowman  <james.bowman@ftdichip.com>
9582         * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9583         offset range for FT32B.
9584         * config/ft32/ft32.h: option "mcompress" enables relaxation.
9585         * config/ft32/ft32.md: Add TARGET_NOPM.
9586         * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9587         * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9589 2017-11-02  Wilco Dijkstra  <wdijkstr@arm.com>
9591         * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9593 2017-11-02  Jeff Law  <law@redhat.com>
9595         * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9596         virtual keyword on FINAL OVERRIDE members.
9598         * tree-ssa-propagate.h (ssa_propagation_engine): Group
9599         virtuals together.  Add virtual destructor.
9600         (substitute_and_fold_engine): Similarly.
9602 2017-11-02  Jan Hubicka  <hubicka@ucw.cz>
9604         * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9606 2017-11-02  Richard Biener  <rguenther@suse.de>
9608         PR tree-optimization/82795
9609         * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9611 2017-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9613         * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9614         gcc_SUN_LD_VERSION.
9615         (gcc_GAS_CHECK_FEATURE): Remove.
9616         * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9617         gcc_AC_INITFINI_ARRAY here.  Update for Solaris 11.4 changes.
9618         * configure: Regenerate.
9620 2017-11-02  Claudiu Zissulescu <claziss@synopsys.com>
9622         * config/arc/arc.c (hwloop_optimize): Account for empty
9623         body loops.
9625 2017-11-02  Richard Biener  <rguenther@suse.de>
9627         PR middle-end/82765
9628         * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9629         Truncate ARRAY_REF index and element size.
9631 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9633         * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9635 2017-11-01  Jeff Law  <law@redhat.com>
9637         * tree-ssa-ccp.c (ccp_folder): New class derived from
9638         substitute_and_fold_engine.
9639         (ccp_folder::get_value): New member function.
9640         (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9641         (ccp_fold_stmt): Remove prototype.
9642         (ccp_finalize): Call substitute_and_fold from the ccp_class.
9643         * tree-ssa-copy.c (copy_folder): New class derived from
9644         substitute_and_fold_engine.
9645         (copy_folder::get_value): Renamed from get_value.
9646         (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9647         * tree-vrp.c (vrp_folder): New class derived from
9648         substitute_and_fold_engine.
9649         (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9650         (vrp_folder::get_value): New member function.
9651         (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9652         (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9653         * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9654         provide a class interface to folder/substitute routines.
9655         (ssa_prop_fold_stmt_fn): Remove typedef.
9656         (ssa_prop_get_value_fn): Likewise.
9657         (subsitute_and_fold): Remove prototype.
9658         (replace_uses_in): Likewise.
9659         * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9660         Renamed from replace_uses_in.  Call the virtual member function
9661         (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9662         (substitute_and_fold_dom_walker): Remove initialization of
9663         data member entries for calbacks.  Add substitute_and_fold_engine
9664         member and initialize it.
9665         (substitute_and_fold_dom_walker::before_dom_children0: Use the
9666         member functions for get_value, replace_phi_args_in c
9667         replace_uses_in, and fold_stmt calls.
9668         (substitute_and_fold_engine::substitute_and_fold): Renamed from
9669         substitute_and_fold.  Remove assert.   Update ctor call.
9671         * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9672         (ssa_prop_visit_phi_fn): Likewise.
9673         (class ssa_propagation_engine): New class to provide an interface
9674         into ssa_propagate.
9675         * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9676         variable.
9677         (ssa_prop_visit_phi): Likewise.
9678         (ssa_propagation_engine::simulate_stmt): Moved into class.
9679         Call visit_phi/visit_stmt from the class rather than via
9680         file scoped static variables.
9681         (ssa_propagation_engine::simulate_block): Moved into class.
9682         (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9683         (ssa_propagation_engine::ssa_propagate): Similarly.  No longer
9684         set file scoped statics for the visit_stmt/visit_phi callbacks.
9685         * tree-complex.c (complex_propagate): New class derived from
9686         ssa_propagation_engine.
9687         (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9688         (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9689         (tree_lower_complex): Call ssa_propagate via the complex_propagate
9690         class.
9691         * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9692         ssa_propagation_engine.
9693         (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9694         (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9695         (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9696         * tree-ssa-copy.c (copy_prop): New class derived from
9697         ssa_propagation_engine.
9698         (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9699         (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9700         (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9701         * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9702         (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9703         (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9704         (execute_vrp): Call ssa_propagate from the vrp_prop class.
9706 2017-11-01  Jakub Jelinek  <jakub@redhat.com>
9708         PR rtl-optimization/82778
9709         PR rtl-optimization/82597
9710         * compare-elim.c (struct comparison): Add in_a_setter field.
9711         (find_comparison_dom_walker::before_dom_children): Remove killed
9712         bitmap and df_simulate_find_defs call, instead walk the defs.
9713         Compute last_setter and initialize in_a_setter.  Merge definitions
9714         with first initialization for a few variables.
9715         (try_validate_parallel): Use insn_invalid_p instead of
9716         recog_memoized.  Return insn rather than just the pattern.
9717         (try_merge_compare): Fix up comment.  Don't uselessly test if
9718         in_a is a REG_P.  Use cmp->in_a_setter instead of walking UD
9719         chains.
9720         (execute_compare_elim_after_reload): Remove df_chain_add_problem
9721         call.
9723 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9724             Alan Hayward  <alan.hayward@arm.com>
9725             David Sherwood  <david.sherwood@arm.com>
9727         * config/aarch64/aarch64.c (aarch64_rtx_costs): Use
9728         aarch64_hard_regno_nregs to get the number of registers
9729         in a mode.
9731 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9732             Alan Hayward  <alan.hayward@arm.com>
9733             David Sherwood  <david.sherwood@arm.com>
9735         * config/aarch64/constraints.md (Upl): Rename to...
9736         (Uaa): ...this.
9737         * config/aarch64/aarch64.md
9738         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9739         Update accordingly.
9741 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9742             Alan Hayward  <alan.hayward@arm.com>
9743             David Sherwood  <david.sherwood@arm.com>
9745         * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9746         (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9747         earlier in file.
9749 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9750             Alan Hayward  <alan.hayward@arm.com>
9751             David Sherwood  <david.sherwood@arm.com>
9753         * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9754         (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9755         (aarch64_evpc_dup): Generate rtl direcly, rather than using
9756         named expanders.
9757         (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9758         of a single element.
9759         * config/aarch64/iterators.md: Add a comment above the permute
9760         unspecs to say that they are generated directly by
9761         aarch64_expand_vec_perm_const.
9762         * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9764 2017-11-01  Nathan Sidwell  <nathan@acm.org>
9766         * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9768 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9770         * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9771         and medany code models, and describe what they do.
9773 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9775         Revert accidental duplicate:
9777         * combine.c (can_change_dest_mode): Reject changes in
9778         REGMODE_NATURAL_SIZE.
9780 2017-11-01  Segher Boessenkool  <segher@kernel.crashing.org>
9782         PR rtl-optimization/64682
9783         PR rtl-optimization/69567
9784         PR rtl-optimization/69737
9785         PR rtl-optimization/82683
9786         * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9787         register mentioned in the note, drop the note, unless it came from I3,
9788         in which case it should go to I3 again.
9790 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9792         * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9793         and return false if not.
9794         (clear_bytes_written_by, live_bytes_read): Update accordingly.
9796 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9798         * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9799         range is known to be empty.
9801 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9802             Alan Hayward  <alan.hayward@arm.com>
9803             David Sherwood  <david.sherwood@arm.com>
9805         * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9806         and CONST_VECTOR_NUNITS instead of computing the number of units from
9807         the byte sizes of the vector and element.
9808         (simplify_binary_operation_1): Likewise.
9809         (simplify_const_binary_operation): Likewise.
9810         (simplify_ternary_operation): Likewise.
9812 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9813             Alan Hayward  <alan.hayward@arm.com>
9814             David Sherwood  <david.sherwood@arm.com>
9816         * var-tracking.c (INT_MEM_OFFSET): Replace with...
9817         (int_mem_offset): ...this new function.
9818         (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9819         (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9820         (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9821         Update accordingly.
9823 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9824             Alan Hayward  <alan.hayward@arm.com>
9825             David Sherwood  <david.sherwood@arm.com>
9827         * lower-subreg.c (interesting_mode_p): New function.
9828         (compute_costs, find_decomposable_subregs, decompose_register)
9829         (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9830         (resolve_clobber, dump_choices): Use it.
9832 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9833             Alan Hayward  <alan.hayward@arm.com>
9834             David Sherwood  <david.sherwood@arm.com>
9836         * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
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         * alias.c (find_base_value, find_base_term): Only process integer
9843         truncations.  Check the precision rather than the size.
9845 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9846             Alan Hayward  <alan.hayward@arm.com>
9847             David Sherwood  <david.sherwood@arm.com>
9849         * machmode.h (is_narrower_int_mode): New function
9850         * optabs.c (expand_float, expand_fix): Use it.
9851         * dwarf2out.c (rotate_loc_descriptor): Likewise.
9853 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9854             Alan Hayward  <alan.hayward@arm.com>
9855             David Sherwood  <david.sherwood@arm.com>
9857         * rtl.h (narrower_subreg_mode): New function.
9858         * ira-color.c (update_costs_from_allocno): Use it.
9860 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9861             Alan Hayward  <alan.hayward@arm.com>
9862             David Sherwood  <david.sherwood@arm.com>
9864         * optabs-query.h (convert_optab_p): New function, split out from...
9865         (convert_optab_handler): ...here.
9866         (widening_optab_handler): Delete.
9867         (find_widening_optab_handler): Remove permit_non_widening parameter.
9868         (find_widening_optab_handler_and_mode): Likewise.  Provide an
9869         override that operates on mode class wrappers.
9870         * optabs-query.c (widening_optab_handler): Delete.
9871         (find_widening_optab_handler_and_mode): Remove permit_non_widening
9872         parameter.  Assert that the two modes are the same class and that
9873         the "from" mode is narrower than the "to" mode.  Use
9874         convert_optab_handler instead of widening_optab_handler.
9875         * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9876         instead of widening_optab_handler.
9877         * expr.c (expand_expr_real_2): Update calls to
9878         find_widening_optab_handler.
9879         * optabs.c (expand_widen_pattern_expr): Likewise.
9880         (expand_binop_directly): Take the insn_code as a parameter.
9881         (expand_binop): Only call find_widening_optab_handler for
9882         conversion optabs; use optab_handler otherwise.  Update calls
9883         to find_widening_optab_handler and expand_binop_directly.
9884         Use convert_optab_handler instead of widening_optab_handler.
9885         * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9886         find_widening_optab_handler and use scalar_mode rather than
9887         machine_mode.
9888         (convert_plusminus_to_widen): Likewise.
9890 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9891             Alan Hayward  <alan.hayward@arm.com>
9892             David Sherwood  <david.sherwood@arm.com>
9894         * machmode.h (fixed_size_mode): New class.
9895         * rtl.h (get_pool_mode): Return fixed_size_mode.
9896         * gengtype.c (main): Add fixed_size_mode.
9897         * target.def (get_raw_result_mode): Return a fixed_size_mode.
9898         (get_raw_arg_mode): Likewise.
9899         * doc/tm.texi: Regenerate.
9900         * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9901         * targhooks.c (default_get_reg_raw_mode): Likewise.
9902         * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9903         * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9904         * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9905         (msp430_get_raw_result_mode): Likewise.
9906         * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9907         * dbxout.c (dbxout_parms): Require fixed-size modes.
9908         * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9909         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9910         * omp-low.c (lower_oacc_reductions): Likewise.
9911         * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
9912         (simplify_subreg): Update accordingly.
9913         * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
9914         (force_const_mem): Update accordingly.  Return NULL_RTX for modes
9915         that aren't fixed-size.
9916         (get_pool_mode): Return a fixed_size_mode.
9917         (output_constant_pool_2): Take a fixed_size_mode.
9919 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9920             Alan Hayward  <alan.hayward@arm.com>
9921             David Sherwood  <david.sherwood@arm.com>
9923         * doc/rtl.texi (vec_series): Document.
9924         (const): Say that the operand can be a vec_series.
9925         * rtl.def (VEC_SERIES): New rtx code.
9926         * rtl.h (const_vec_series_p_1): Declare.
9927         (const_vec_series_p): New function.
9928         * emit-rtl.h (gen_const_vec_series): Declare.
9929         (gen_vec_series): Likewise.
9930         * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
9931         (gen_vec_series): Likewise.
9932         * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
9933         * simplify-rtx.c (simplify_unary_operation): Handle negations
9934         of vector series.
9935         (simplify_binary_operation_series): New function.
9936         (simplify_binary_operation_1): Use it.  Handle VEC_SERIES.
9937         (test_vector_ops_series): New function.
9938         (test_vector_ops): Call it.
9939         * config/powerpcspe/altivec.md (altivec_lvsl): Use
9940         gen_const_vec_series.
9941         (altivec_lvsr): Likewise.
9942         * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
9944 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9945             Alan Hayward  <alan.hayward@arm.com>
9946             David Sherwood  <david.sherwood@arm.com>
9948         * doc/rtl.texi (const): Update description of address constants.
9949         Say that vector constants are allowed too.
9950         * common.md (E, F): Use CONSTANT_P instead of checking for
9951         CONST_VECTOR.
9952         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9953         checking for CONST_VECTOR.
9954         * expmed.c (make_tree): Use build_vector_from_val for a CONST
9955         VEC_DUPLICATE.
9956         * expr.c (expand_expr_real_2): Check for vector modes instead
9957         of checking for CONST_VECTOR.
9958         * rtl.h (const_vec_p): New function.
9959         (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
9960         (unwrap_const_vec_duplicate): Handle them here too.
9962 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9963             David Malcolm  <dmalcolm@redhat.com>
9964             Alan Hayward  <alan.hayward@arm.com>
9965             David Sherwood  <david.sherwood@arm.com>
9967         * rtl.h (vec_duplicate_p): New function.
9968         * selftest-rtl.c (assert_rtx_eq_at): New function.
9969         * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
9970         (assert_rtx_eq_at): Declare.
9971         * selftest.h (selftest::simplify_rtx_c_tests): Declare.
9972         * selftest-run-tests.c (selftest::run_tests): Call it.
9973         * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
9974         (simplify_unary_operation_1): Recursively handle vector duplicates.
9975         (simplify_binary_operation_1): Likewise.  Handle VEC_SELECTs of
9976         vector duplicates.
9977         (simplify_subreg): Handle subregs of vector duplicates.
9978         (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
9979         (selftest::simplify_rtx_c_tests): New functions.
9981 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9982             Alan Hayward  <alan.hayward@arm.com>
9983             David Sherwood  <david.sherwood@arm.com>
9985         * emit-rtl.h (gen_const_vec_duplicate): Declare.
9986         (gen_vec_duplicate): Likewise.
9987         * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
9988         out from...
9989         (gen_const_vector): ...here.
9990         (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
9991         (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
9992         whose elements are all equal.
9993         * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
9994         * simplify-rtx.c (simplify_const_unary_operation): Likewise.
9995         (simplify_relational_operation): Likewise.
9996         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
9997         Likewise.
9998         (aarch64_simd_dup_constant): Use gen_vec_duplicate.
9999         (aarch64_expand_vector_init): Likewise.
10000         * config/arm/arm.c (neon_vdup_constant): Likewise.
10001         (neon_expand_vector_init): Likewise.
10002         (arm_expand_vec_perm): Use gen_const_vec_duplicate.
10003         (arm_block_set_unaligned_vect): Likewise.
10004         (arm_block_set_aligned_vect): Likewise.
10005         * config/arm/neon.md (neon_copysignf<mode>): Likewise.
10006         * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
10007         (expand_vec_perm_even_odd_pack): Likewise.
10008         (ix86_vector_duplicate_value): Use gen_vec_duplicate.
10009         * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
10010         * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
10011         gen_const_vec_duplicate.
10012         * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
10013         * config/mips/mips.c (mips_gen_const_int_vector): Use
10014         gen_const_vec_duplicate.
10015         (mips_expand_vector_init): Use CONST0_RTX.
10016         * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
10017         (define_split): Use gen_const_vec_duplicate.
10018         * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
10019         (define_split): Use gen_const_vec_duplicate.
10020         * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
10021         (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
10022         * config/spu/spu.c (spu_const): Likewise.
10024 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
10025             Alan Hayward  <alan.hayward@arm.com>
10026             David Sherwood  <david.sherwood@arm.com>
10028         * combine.c (can_change_dest_mode): Reject changes in
10029         REGMODE_NATURAL_SIZE.
10031 2017-10-31  Sandra Loosemore  <sandra@codesourcery.com>
10033         * configure.ac (--enable-libssp): New.
10034         (gcc_cv_libc_provides_ssp): Check for explicit setting before
10035         trying to determine target-specific default.  Adjust indentation.
10036         * configure: Regenerated.
10037         * doc/install.texi (Configuration): Expand --disable-libssp
10038         documentation.
10040 2017-10-31  Daniel Santos  <daniel.santos@pobox.com>
10042         config/i386/i386.c (ix86_expand_epilogue): Correct stack
10043         calculation.
10045 2017-10-31  Martin Jambor  <mjambor@suse.cz>
10047         PR c++/81702
10048         * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
10050 2017-10-31  David Malcolm  <dmalcolm@redhat.com>
10052         * auto-profile.c (autofdo_source_profile::read): Use
10053         UNKNOWN_LOCATION rather than 0.
10054         * diagnostic-core.h (warning_at_rich_loc): Rename to...
10055         (warning_at): ...this overload.
10056         (warning_at_rich_loc_n): Rename to...
10057         (warning_n): ...this overload.
10058         (error_at_rich_loc): Rename to...
10059         (error_at): ...this overload.
10060         (pedwarn_at_rich_loc): Rename to...
10061         (pedwarn): ...this overload.
10062         (permerror_at_rich_loc): Rename to...
10063         (permerror): ...this overload.
10064         (inform_at_rich_loc): Rename to...
10065         (inform): ...this overload.
10066         * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
10067         (diagnostic_n_impl_richloc): Rename to...
10068         (diagnostic_n_impl): ...this rich_location *-based decl.
10069         (inform_at_rich_loc): Rename to...
10070         (inform): ...this, and add an assertion.
10071         (inform_n): Update for removal of location_t-based diagnostic_n_impl.
10072         (warning_at_rich_loc): Rename to...
10073         (warning_at): ...this, and add an assertion.
10074         (warning_at_rich_loc_n): Rename to...
10075         (warning_n): ...this, and add an assertion.
10076         (warning_n): Update location_t-based implementation for removal of
10077         location_t-based diagnostic_n_impl.
10078         (pedwarn_at_rich_loc): Rename to...
10079         (pedwarn): ...this, and add an assertion.
10080         (permerror_at_rich_loc): Rename to...
10081         (permerror): ...this, and add an assertion.
10082         (error_n): Update for removal of location_t-based diagnostic_n_impl.
10083         (error_at_rich_loc): Rename to...
10084         (error_at): ...this, and add an assertion.
10085         * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
10086         (driver::do_spec_on_infiles): Likewise.
10087         * substring-locations.c (format_warning_va): Update for renaming
10088         of inform_at_rich_loc.
10090 2017-10-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
10092         * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
10093         _Float<N>X built-in functions so that the variant without the
10094         "__builtin_" prefix is only enabled for the GNU C and Objective C
10095         languages when they are in non-strict ANSI/ISO mode.
10096         (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
10097         * target.def (floatn_builtin_p): Add a target hook to control
10098         whether _Float<N> and _Float<N>X built-in functions without the
10099         "__builtin_" prefix are enabled, and return true for C and
10100         Objective C in the default hook.  Include langhooks.h in
10101         targhooks.c.
10102         * targhooks.h (default_floatn_builtin_p): Likewise.
10103         * targhooks.c (default_floatn_builtin_p): Likewise.
10104         * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
10105         floatn_builtin_p target hook.
10106         * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
10108 2017-10-31  Matthew Fortune  <matthew.fortune@imgtec.com>
10109             Eric Botcazou  <ebotcazou@adacore.com>
10111         PR rtl-optimization/81803
10112         * lra-constraints.c (curr_insn_transform): Also reload the whole
10113         register for a strict subreg no wider than a word if this is for
10114         a WORD_REGISTER_OPERATIONS target.
10116 2017-10-31  Jason Merrill  <jason@redhat.com>
10118         * gdbinit.in: Skip over inlines from timevar.h.
10120 2017-10-31  Martin Liska  <mliska@suse.cz>
10122         * doc/gcov.texi: Document new option.
10123         * gcov.c (print_usage): Likewise print it.
10124         (process_args): Support the argument.
10125         (format_count): New function.
10126         (format_gcov): Use the function.
10128 2017-10-31  Martin Liska  <mliska@suse.cz>
10130         * gcov.c (struct name_map): do not use typedef.
10131         Define operator== and operator<.
10132         (name_search): Remove.
10133         (name_sort): Remove.
10134         (main): Do not allocate names.
10135         (process_file): Add vertical space.
10136         (generate_results): Use std::find.
10137         (release_structures): Do not release memory.
10138         (find_source): Use std::find.
10140 2017-10-31  Martin Liska  <mliska@suse.cz>
10142         * gcov.c (struct line_info): Remove it's typedef.
10143         (line_info::line_info): Add proper ctor.
10144         (line_info::has_block): Do not use a typedef.
10145         (struct source_info): Do not use typedef.
10146         (circuit): Likewise.
10147         (get_cycles_count): Likewise.
10148         (output_intermediate_file): Iterate via vector iterator.
10149         (add_line_counts): Use std::vector methods.
10150         (accumulate_line_counts): Likewise.
10151         (output_lines): Likewise.
10153 2017-10-31  Martin Liska  <mliska@suse.cz>
10155         * gcov.c (struct source_info): Remove typedef.
10156         (source_info::source_info): Add proper ctor.
10157         (accumulate_line_counts): Use struct, not it's typedef.
10158         (output_gcov_file): Likewise.
10159         (output_lines): Likewise.
10160         (main): Do not allocate an array.
10161         (output_intermediate_file): Use size of vector container.
10162         (process_file): Resize the vector.
10163         (generate_results): Do not preallocate, use newly added vector
10164         lines.
10165         (release_structures): Do not release sources.
10166         (find_source): Use vector methods.
10167         (add_line_counts): Do not use typedef.
10169 2017-10-31  Martin Liska  <mliska@suse.cz>
10171         * doc/gcov.texi: Document that.
10172         * gcov.c (add_line_counts): Mark lines with a non-executed
10173         statement.
10174         (output_line_beginning): Handle such lines.
10175         (output_lines): Pass new argument.
10176         (output_intermediate_file): Print it in intermediate format.
10178 2017-10-31  Martin Liska  <mliska@suse.cz>
10180         * color-macros.h: New file.
10181         * diagnostic-color.c: Factor out color related to macros to
10182         color-macros.h.
10183         * doc/gcov.texi: Document -k option.
10184         * gcov.c (INCLUDE_STRING): Include string.h.
10185         (print_usage): Add -k option.
10186         (process_args): Parse it.
10187         (pad_count_string): New function.
10188         (output_line_beginning): Likewise.
10189         (DEFAULT_LINE_START): New macro.
10190         (output_lines): Support color output.
10192 2017-10-31  Martin Liska  <mliska@suse.cz>
10194         PR gcov-profile/82633
10195         * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
10196         their interaction with GCOV infrastructure.
10197         * configure.ac: Add -fkeep-{inline,static}-functions to
10198         coverage_flags.
10199         * configure: Regenerate.
10201 2017-10-31  Uros Bizjak  <ubizjak@gmail.com>
10203         PR target/82772
10204         * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
10206 2017-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
10208         PR target/82674
10209         * config/rs6000/rs6000.md (allocate_stack): Force update interval
10210         into a register if it does not fit into an immediate offset field.
10212 2017-10-31  Olivier Hainque  <hainque@adacore.com>
10214         * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
10216 2017-10-31  Julia Koval  <julia.koval@intel.com>
10218         * config.gcc: Add gfniintrin.h.
10219         * config/i386/gfniintrin.h: New.
10220         * config/i386/i386-builtin-types.def
10221         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10222         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
10223         __builtin_ia32_vgf2p8affineinvqb_v32qi,
10224         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
10225         __builtin_ia32_vgf2p8affineinvqb_v16qi,
10226         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
10227         * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
10228         V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
10229         V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
10230         V64QI_FTYPE_V64QI_V64QI_INT): New types.
10231         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
10232         * config/i386/immintrin.h: Include gfniintrin.h.
10233         * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
10235 2017-10-30  Eric Botcazou  <ebotcazou@adacore.com>
10237         * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10239 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10241         * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10242         (arm_ashldi3_1bit): Remove pattern.
10243         (ashrdi3): Remove shift by 1 expansion.
10244         (arm_ashrdi3_1bit): Remove pattern.
10245         (lshrdi3): Remove shift by 1 expansion.
10246         (arm_lshrdi3_1bit): Remove pattern.
10247         * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10248         cost of ashldi3 by 1.
10249         * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10250         (<shift>di3_neon): Likewise.
10252 2017-10-30  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
10254         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10255         both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10256         and (*aarch64_simd_mov<VQ:mode>).
10257         (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10258         pattern alternative.
10259         (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10260         attributes to match pattern alternative.
10262 2017-10-30  Steven Munroe  <munroesj@gcc.gnu.org>
10264         * config.gcc (powerpc*-*-*): Add emmintrin.h.
10265         * config/rs6000/emmintrin.h: New file.
10266         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10268 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10270         * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10271         * (movdi_vfp_cortexa8): Remove pattern.
10273 2017-10-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10275         * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10276         etc. reference.
10277         (Specific, alpha*-dec-osf5.1): Remove.
10278         (Specific, mips-sgi-irix5): Remove.
10279         (Specific, mips-sgi-irix6): Remove.
10281 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10283         PR middle-end/22141
10284         * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10285         arguments to clear_bit_region_be.
10287 2017-10-30  Jim Wilson  <wilson@tuliptree.org>
10289         * gimplify.c: Include memmodel.h.
10291 2017-10-30  Martin Jambor  <mjambor@suse.cz>
10293         * omp-grid.c (grid_attempt_target_gridification): Also insert a
10294         condition whether loop should be executed at all.
10296 2017-10-30  Will Schmidt  <will_schmidt@vnet.ibm.com>
10298         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10299         gimple folding of vec_madd() intrinsics.
10300         * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10301         fmav8hi4.  (altivec_vmladduhm): Rename to fmav8hi4.
10302         * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10304 2017-10-30  Richard Biener  <rguenther@suse.de>
10306         PR tree-optimization/82762
10307         Revert
10308         2017-10-23  Richard Biener  <rguenther@suse.de>
10310         PR tree-optimization/82129
10311         Revert
10312         2017-08-01  Richard Biener  <rguenther@suse.de>
10314         PR tree-optimization/81181
10315         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10316         (compute_antic): ... end of iteration here.
10318 2017-10-30  Joseph Myers  <joseph@codesourcery.com>
10320         * doc/invoke.texi (C Dialect Options): Document -std=c17,
10321         -std=iso9899:2017 and -std=gnu17.
10322         * doc/standards.texi (C Language): Document C17 support.
10323         * doc/cpp.texi (Overview): Mention -std=c17.
10324         (Standard Predefined Macros): Document C11 and C17 values of
10325         __STDC_VERSION__.  Do not refer to C99 support as incomplete.
10326         * doc/extend.texi (Inline): Do not list individual options for
10327         standards newer than C99.
10328         * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10329         "GNU C17".
10330         * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10331         language name.
10333 2017-10-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
10335         * asan.c (asan_finish_file): Align asan globals array by shadow
10336         granularity.
10338 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10340         PR middle-end/22141
10341         * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10342         (struct store_immediate_info): Add bitregion_start and bitregion_end
10343         fields.
10344         (store_immediate_info::store_immediate_info): Add brs and bre
10345         arguments and initialize bitregion_{start,end} from those.
10346         (struct merged_store_group): Add bitregion_start, bitregion_end,
10347         align_base and mask fields.  Drop unnecessary struct keyword from
10348         struct store_immediate_info.  Add do_merge method.
10349         (clear_bit_region_be): Use memset instead of loop storing zeros.
10350         (merged_store_group::do_merge): New method.
10351         (merged_store_group::merge_into): Use do_merge.  Allow gaps in between
10352         stores as long as the surrounding bitregions have no gaps.
10353         (merged_store_group::merge_overlapping): Use do_merge.
10354         (merged_store_group::apply_stores): Test that bitregion_{start,end}
10355         is byte aligned, rather than requiring that start and width are
10356         byte aligned.  Drop unnecessary struct keyword from
10357         struct store_immediate_info.  Allocate and populate also mask array.
10358         Make start of the arrays relative to bitregion_start rather than
10359         start and size them according to bitregion_{end,start} difference.
10360         (struct imm_store_chain_info): Drop unnecessary struct keyword from
10361         struct store_immediate_info.
10362         (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10363         (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10364         struct keyword from struct store_immediate_info.
10365         (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10366         between stores as long as the surrounding bitregions have no gaps.
10367         Formatting fixes.
10368         (struct split_store): Add orig non-static data member.
10369         (split_store::split_store): Initialize orig to false.
10370         (find_constituent_stmts): Return store_immediate_info *, non-NULL
10371         if there is exactly a single original stmt.  Change stmts argument
10372         to pointer from reference, if NULL, don't push anything to it.  Add
10373         first argument, use it to optimize skipping over orig stmts that
10374         are known to be before bitpos already.  Simplify.
10375         (split_group): Return unsigned int count how many stores are or
10376         would be needed rather than a bool.  Add allow_unaligned argument.
10377         Change split_stores argument from reference to pointer, if NULL,
10378         only do a dry run computing how many stores would be produced.
10379         Rewritten algorithm to use both alignment and misalign if
10380         !allow_unaligned and handle bitfield stores with gaps.
10381         (imm_store_chain_info::output_merged_store): Set start_byte_pos
10382         from bitregion_start instead of start.  Compute allow_unaligned
10383         here, if true, do 2 split_group dry runs to compute which one
10384         produces fewer stores and prefer aligned if equal.  Punt if
10385         new count is bigger or equal than original before emitting any
10386         statements, rather than during that.  Remove no longer needed
10387         new_ssa_names tracking.  Replace num_stmts with
10388         split_stores.length ().  Use 32-bit stack allocated entries
10389         in split_stores auto_vec.  Try to reuse original store lhs/rhs1
10390         if possible.  Handle bitfields with gaps.
10391         (pass_store_merging::execute): Ignore bitsize == 0 stores.
10392         Compute bitregion_{start,end} for the stores and construct
10393         store_immediate_info with that.  Formatting fixes.
10395 2017-10-30  Uros Bizjak  <ubizjak@gmail.com>
10397         PR target/82725
10398         * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10399         UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10401 2017-10-29  Jim Wilson  <wilson@tuliptree.org>
10403         * gimplify.c: Include tm_p.h.
10405         * common.opt (gcoff): Re-add as ignored option.
10406         (gcoff1, gcoff2, gcoff3): Likewise.
10408         * Makefile.in (OBJS): Delete sdbout.o.
10409         (GTFILES): Delete $(srcdir)/sdbout.c.
10410         * debug.h: Delete sdb_debug_hooks.
10411         * final.c: Delete sdbout.h include.
10412         (final_scan_insn): Delete SDB_DEBUG check.
10413         (rest_of_clean_state): Likewise.
10414         * output.h: Delete sdb_begin_function_line.
10415         * sdbout.c: Delete.
10416         * sdbout.h: Delete.
10417         * toplev.c: Delete sdbout.h include.
10418         (process_options): Delete SDB_DEBUG check.
10419         * tree-core.h (tree_type_common): Delete pointer field of
10420         tree_type_symtab.
10421         * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10422         TYPE_SYMTAB_POINTER.
10423         * tree.h (TYPE_SYMTAB_POINTER): Delete.
10424         (TYPE_SYMTAB_IS_POINTER): Delete.
10425         (TYPE_SYMTAB_IS_DIE): Renumber.
10426         * xcoffout.c: Refer to former sdbout.c file.
10427         (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10429         * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10430         * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10431         (Debugging Options): Delete -gcoff.
10432         (-gstabs): Delete SDB reference.
10433         (-gcoff): Delete.
10434         (-gcoff@var{level}): Delete.
10435         * doc/passes.texi (Debugging information output): Delete SDB and
10436         sdbout.c references.
10437         * doc/tm.texi: Regenerate.
10438         * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10439         (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10440         references.
10441         (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10442         (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10443         (SDB_DEBUGGING_INFO): Delete.
10444         (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10445         SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10446         * target.def (output_source_filename): Delete COFF reference.
10448         * common.opt (gcoff): Delete.
10449         (gxcoff+): Update Negative chain.
10450         * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10451         SDB_DEBUG.
10452         * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10453         * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10454         * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10455         and SDB references.
10456         (expand_function_start): Change sdb reference to past tense.
10457         (expand_function_end): Change sdb reference to past tense.
10458         * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10459         * opts.c (debug_type_names): Delete coff entry.
10460         (common_handle_option): Delete OPT_gcoff case.
10461         * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10463         * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10464         * config/cris/cris.h: Delete SDB reference in comment.
10465         * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10466         (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10467         * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10468         * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10469         to past tense.
10470         (ix86_expand_prologue): Likewise.
10471         * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10472         * config/ia64/ia64.h: Likewise.
10473         * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10474         * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10475         support.
10476         * config/mmix/mmix.h: Likewise.
10477         * config/nds32/nds32.c: Likewise.
10478         * config/stormy/storym16.h: Likewise.
10479         * config/visium/visium.h: Likewise.
10480         * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10482 2017-10-28  Sandra Loosemore  <sandra@codesourcery.com>
10484         * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10485         * config/nios2/nios2.c (nios2_initial_elimination_offset):  Make
10486         FRAME_POINTER_REGNUM point at high end of local var area.
10488 2017-10-27  Eric Botcazou  <ebotcazou@adacore.com>
10490         * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10491         Move comment around.  Do not reset best_edge for a copiable
10492         destination if the copy would cause a partition change.
10493         (better_edge_p): Remove redundant check.
10495 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10497         * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10499 2017-10-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
10501         * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10502         for math functions that have _Float<N> and _Float<N>X variants.
10503         (mathfn_built_in_2): Add support for math functions that have
10504         _Float<N> and _Float<N>X variants.
10505         (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10506         (expand_builtin_mathfn_ternary): Add support for fma with
10507         _Float<N> and _Float<N>X variants.
10508         (expand_builtin): Likewise.
10509         (fold_builtin_3): Likewise.
10510         * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10511         create math function _Float<N> and _Float<N>X variants as external
10512         library builtins.
10513         (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10514         DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10515         the __builtin_ prefix and if not strict ansi, without the prefix.
10516         (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10517         (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10518         (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10519         (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10520         (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10521         (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10522         * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10523         function signatures for fma _Float<N> and _Float<N>X variants.
10524         (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10525         (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10526         (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10527         (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10528         (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10529         (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10530         * gencfn-macros.c (print_case_cfn): Add support for math functions
10531         that have _Float<N> and _Float<N>X variants.
10532         (print_define_operator_list): Likewise.
10533         (fltfn_suffixes): Likewise.
10534         (main): Likewise.
10535         * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10536         for math functions that have _Float<N> and _Float<N>X variants.
10537         (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10538         and _Float<N>X variants.
10539         (COPYSIGN): Likewise.
10540         (FMIN): Likewise.
10541         (FMAX): Likewise.
10542         * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10543         copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10544         variants.
10545         (integer_valued_read_call_p): Likewise.
10546         * fold-const-call.c (fold_const_call_ss): Likewise.
10547         (fold_const_call_sss): Add support for copysign, fmin, and fmax
10548         _Float<N> and _Float<N>X variants.
10549         (fold_const_call_ssss): Add support for fma _Float<N> and
10550         _Float<N>X variants.
10551         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10552         support for copysign and fma _Float<N> and _Float<N>X variants.
10553         (backprop::process_builtin_call_use): Likewise.
10554         * tree-call-cdce.c (can_test_argument_range); Add support for
10555         sqrt _Float<N> and _Float<N>X variants.
10556         (edom_only_function): Likewise.
10557         (get_no_error_domain): Likewise.
10558         * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10559         * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10560         copysign _Float<N> and _Float<N>X variants.
10561         * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10562         handled by machine independent code.
10563         (FMAF128): Likewise.
10564         * doc/cpp.texi (Common Predefined Macros): Document defining
10565         __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10566         fma _Float<N> and _Float<N>X variants.
10568 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10570         PR target/82692
10571         * config/i386/i386-modes.def (CCFPU): Remove definition.
10572         * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10573         (ix86_cc_modes_compatible): Ditto.
10574         (ix86_expand_carry_flag_compare): Ditto.
10575         (ix86_expand_int_movcc): Ditto.
10576         (ix86_expand_int_addcc): Ditto.
10577         (ix86_reverse_condition): Ditto.
10578         (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10579         Return true/false for unordered/ordered fp comparisons.
10580         (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10581         (ix86_prepare_fp_compare_args): Update for rename.
10582         (ix86_expand_fp_compare): Update for rename.  Generate unordered
10583         compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10584         (ix86_expand_sse_compare_and_jump): Ditto.
10585         * config/i386/predicates.md (fcmov_comparison_operator):
10586         Remove CCFPU mode handling.
10587         (ix86_comparison_operator): Ditto.
10588         (ix86_carry_flag_operator): Ditto.
10589         * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10590         (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10591         (*cmpu<mode>_cc_i387): Ditto.
10592         (FPCMP): Remove mode iterator.
10593         (unord): Remove mode attribute.
10594         (unord_subst): New define_subst transformation
10595         (unord): New define_subst attribute.
10596         (unordered): Ditto.
10597         (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10598         (*cmpi<unord>xf_i387): Ditto.
10599         * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10600         from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10601         using unord_subst transformation.
10602         * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10603         (round_saeonly): Also handle CCFP mode.
10604         * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10605         Remove UNSPEC_SAHF unspec handling.
10607 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10609         * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10611 2017-10-27  Jeff Law  <law@redhat.com>
10613         * gimple-ssa-sprintf.c: Include domwalk.h.
10614         (class sprintf_dom_walker): New class, derived from dom_walker.
10615         (sprintf_dom_walker::before_dom_children): New function.
10616         (struct call_info): Moved into sprintf_dom_walker class
10617         (compute_formath_length, handle_gimple_call): Likewise.
10618         (sprintf_length::execute): Call the dominator walker rather
10619         than walking the statements.
10621         * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10622         gimple statement locations.
10623         (check_array_bounds): Corresponding changes.  Get the statement's
10624         location directly from wi->stmt.
10626 2017-10-27  Palmer Dabbelt  <palmer@dabbelt.com>
10628         PR target/82717
10629         * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10631 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10633         * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10634         X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10636 2017-10-27  Jakub Jelinek  <jakub@redhat.com>
10638         PR target/82703
10639         * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10640         * config/i386/i386.c (maybe_get_pool_constant): Removed.
10641         (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10642         maybe_get_pool_constant.
10643         * config/i386/predicates.md (zero_extended_scalar_load_operand):
10644         Likewise.
10646 2017-10-27  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10648         * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10649         2.26 caveat.  Update gas and gld versions.
10650         (Specific, *-*-solaris2*): Update binutils version.  Remove caveat
10651         reference.
10653 2017-10-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10655         * cgraph.h (set_malloc_flag): Declare.
10656         * cgraph.c (set_malloc_flag_1): New function.
10657         (set_malloc_flag): Likewise.
10658         * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10659         * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10660         false.
10661         (read_ipa_call_summary): Add support for reading is_return_callee.
10662         (write_ipa_call_summary): Stream is_return_callee.
10663         * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10664         * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10665         ipa-prop.h, ipa-fnsummary.h.
10666         (pure_const_names): Change to static.
10667         (malloc_state_e): Define.
10668         (malloc_state_names): Define.
10669         (funct_state_d): Add field malloc_state.
10670         (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10671         (check_retval_uses): New function.
10672         (malloc_candidate_p): Likewise.
10673         (analyze_function): Add support for malloc attribute.
10674         (pure_const_write_summary): Stream malloc_state.
10675         (pure_const_read_summary): Add support for reading malloc_state.
10676         (dump_malloc_lattice): New function.
10677         (propagate_malloc): New function.
10678         (warn_function_malloc): New function.
10679         (ipa_pure_const::execute): Call propagate_malloc and
10680         ipa_free_fn_summary.
10681         (pass_local_pure_const::execute): Add support for malloc attribute.
10682         * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10683         * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10685 2017-10-27  Martin Liska  <mliska@suse.cz>
10687         PR gcov-profile/82457
10688         * doc/invoke.texi: Document that one needs a non-strict ISO mode
10689         for fork-like functions to be properly instrumented.
10691 2017-10-27  Richard Biener  <rguenther@suse.de>
10693         PR middle-end/81659
10694         * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10695         info when we redirected EH.
10697 2017-10-26  Michael Collison  <michael.collison@arm.com>
10699         * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10700         New pattern.
10701         (<optab>_trunchf<GPI:mode>2: New pattern.
10702         (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10703         * config/aarch64/iterators.md (wv): New mode attribute.
10704         (vf, VF): New mode attributes.
10705         (vgp, VGP): New mode attributes.
10706         (s): Update attribute with SImode and DImode prefixes.
10708 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10710         * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10711         * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10712         * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10713         (nios2_option_overide): Initialize it.  Don't allow R0-relative
10714         addressing with PIC.
10715         (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10716         (nios2_symbolic_constant_p): Likewise.
10717         (nios2_legitimate_address_p): Likewise.
10718         (nios2_r0rel_section_name_p): New.
10719         (nios2_symbol_ref_in_r0rel_data_p): New.
10720         (nios2_emit_move_sequence): Handle r0rel_constant_p.
10721         (r0rel_constant_p): New.
10722         (nios2_print_operand_address): Handle r0rel_constant_p.
10723         (nios2_cdx_narrow_form_p): Likewise.
10724         * config/nios2/nios2.opt (mr0rel-sec=): New option.
10725         * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10726         (Nios II Options): Document -mr0rel-sec.
10728 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10730         * config/nios2/nios2.c: Include xregex.h.
10731         (nios2_gprel_sec_regex): New.
10732         (nios2_option_overide): Initialize it.  Don't allow GP-relative
10733         addressing with PIC.
10734         (nios2_small_section_name_p): Check for regex match.
10735         * config/nios2/nios2.opt (mgprel-sec=): New option.
10736         * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10737         (Nios II Options): Document -mgprel-sec.
10739 2017-10-26  Jim Wilson  <wilson@tuliptree.org>
10741         * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10743 2017-10-26  Tom de Vries  <tom@codesourcery.com>
10745         PR tree-optimization/82707
10746         * gimple.c (gimple_copy): Fix unsharing of
10747         GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10749 2017-10-26  Olga Makhotina  <olga.makhotina@intel.com>
10751         * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10752         _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10753         _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10754         _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10755         _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10756         _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10757         _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10758         _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10759         _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10760         _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10761         _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10762         _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10763         _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10764         _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10765         _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10766         _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10767         _mm512_mask_cmpunord_ps_mask): New intrinsics.
10769 2017-10-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
10771         * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10772         default to IBM.
10773         * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10774         * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10775         warning to rs6000.c.  Remove the Undocumented flag, since it has
10776         been documented.
10777         (-mabi=ibmlongdouble): Likewise.
10778         * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10779         already set, set the default format for long double.
10780         (rs6000_debug_reg_global): Print whether long double is IBM or
10781         IEEE.
10782         (rs6000_option_override_internal): Rework setting long double
10783         format.  Only warn if the user is changing the long double default
10784         and they did not use -Wno-psabi.
10785         * doc/invoke.texi (PowerPC options): Update the documentation for
10786         -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10788 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10789             Alan Hayward  <alan.hayward@arm.com>
10790             David Sherwood  <david.sherwood@arm.com>
10792         * rtl.h (wider_subreg_mode): New function.
10793         * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10794         rather than an unsigned int *.
10795         * ira-color.c (regno_max_ref_width): Replace with...
10796         (regno_max_ref_mode): ...this new variable.
10797         (coalesced_pseudo_reg_slot_compare): Update accordingly.
10798         Use wider_subreg_mode.
10799         (ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
10800         rather than an unsigned int *.
10801         * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10802         (process_alt_operands): Likewise.
10803         (invariant_p): Likewise.
10804         * lra-spills.c (assign_mem_slot): Likewise.
10805         (add_pseudo_to_slot): Likewise.
10806         * lra.c (collect_non_operand_hard_regs): Likewise.
10807         (add_regs_to_insn_regno_info): Likewise.
10808         * reload1.c (regno_max_ref_width): Replace with...
10809         (regno_max_ref_mode): ...this new variable.
10810         (reload): Update accordingly.  Update call to
10811         ira_sort_regnos_for_alter_reg.
10812         (alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
10813         (init_eliminable_invariants): Update to use regno_max_ref_mode.
10814         (scan_paradoxical_subregs): Likewise.
10816 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10818         * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10819         (aarch64_frame): Add emit_frame_chain boolean.
10820         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10821         Move eh_return case to aarch64_layout_frame.
10822         (aarch64_layout_frame): Initialize emit_frame_chain.
10823         (aarch64_expand_prologue): Use emit_frame_chain.
10825 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10827         * config/aarch64/aarch64.c (aarch64_layout_frame):
10828         Ensure LR is always stored at the bottom of the callee-saves.
10829         Remove rarely used frame layout which saves callee-saves at top of
10830         frame, so the store of LR can be used as a valid probe in all cases.
10832 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10834         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10835         Improve unaligned TImode/TFmode base/offset split.
10837 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10838             Alan Hayward  <alan.hayward@arm.com>
10839             David Sherwood  <david.sherwood@arm.com>
10841         * caller-save.c (mark_referenced_regs):  Use read_modify_subreg_p.
10842         * combine.c (find_single_use_1): Likewise.
10843         (expand_field_assignment): Likewise.
10844         (move_deaths): Likewise.
10845         * lra-constraints.c (simplify_operand_subreg): Likewise.
10846         (curr_insn_transform): Likewise.
10847         * lra.c (collect_non_operand_hard_regs): Likewise.
10848         (add_regs_to_insn_regno_info): Likewise.
10849         * rtlanal.c (reg_referenced_p): Likewise.
10850         (covers_regno_no_parallel_p): Likewise.
10852 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10854         * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10855         Don't print any bits outside the precision of the value.
10856         * wide-int.cc (test_printing): Add some new tests.
10858 2017-10-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10860         * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10861         supports -xbrace_comment option.
10862         * configure: Regenerate.
10863         * config.in: Regenerate.
10864         * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10865         (ASM_CPU_SPEC): Use it.
10867 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10869         * target.def (static_rtx_alignment): New hook.
10870         * targhooks.h (default_static_rtx_alignment): Declare.
10871         * targhooks.c (default_static_rtx_alignment): New function.
10872         * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10873         * doc/tm.texi: Regenerate.
10874         * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10875         instead of targetm.constant_alignment.  Remove call to
10876         set_mem_attributes.
10877         * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10878         (cris_preferred_mininum_alignment): New function, split out from...
10879         (cris_constant_alignment): ...here.
10880         (cris_static_rtx_alignment): New function.
10881         * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10882         split out from...
10883         (ix86_constant_alignment): ...here.
10884         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10885         * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10886         (mmix_static_rtx_alignment): New function.
10887         * config/spu/spu.c (spu_static_rtx_alignment): New function.
10888         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10890 2017-10-26  Tamar Christina  <tamar.christina@arm.com>
10892         PR target/81800
10893         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10894         Add flag_trapping_math and flag_fp_int_builtin_inexact.
10896 2017-10-25  Palmer Dabbelt  <palmer@dabbelt.com>
10898         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10899         * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10900         mark as a sign-extending load.
10901         (local_pic_load_u): Define.
10903 2017-10-25  Eric Botcazou  <ebotcazou@adacore.com>
10905         PR middle-end/82062
10906         * fold-const.c (operand_equal_for_comparison_p): Also return true
10907         if ARG0 is a simple variant of ARG1 with narrower precision.
10908         (fold_ternary_loc): Always pass unstripped operands to the predicate.
10910 2017-10-25  Jan Hubicka  <hubicka@ucw.cz>
10912         * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
10913         cost correctly.
10914         * i386.h (processor_costs): Add gather_static, gather_per_elt,
10915         scatter_static, scatter_per_elt.
10916         * x86-tune-costs.h: Add new cost entries.
10918 2017-10-25  Richard Biener  <rguenther@suse.de>
10920         * tree-ssa-sccvn.h (vn_eliminate): Declare.
10921         * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
10922         class pass_fre): Move to ...
10923         * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
10924         class pass_fre): ... here and adjust for statistics.
10926 2017-10-25  Jakub Jelinek  <jakub@redhat.com>
10928         PR libstdc++/81706
10929         * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
10930         for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
10931         (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
10932         * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
10933         declarations.
10935 2017-10-25  Richard Biener  <rguenther@suse.de>
10937         * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
10938         el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
10939         eliminate_push_avail, eliminate_insert): Move inside...
10940         (class eliminate_dom_walker): ... this class in preparation
10941         of move.
10942         (fini_eliminate): Remove by merging with ...
10943         (eliminate): ... this function.  Adjust for class changes.
10944         (pass_pre::execute): Remove fini_eliminate call.
10945         (pass_fre::execute): Likewise.
10947 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10949         PR target/82460
10950         * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
10951         (VPERMI2, VPERMI2I): New mode iterators.
10952         (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
10953         (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
10954         patterns.
10955         (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
10956         mode iterator.  Remove 3 old define_insn patterns.
10957         (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
10958         (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
10959         VPERMI2 mode iterator, remove the other two expanders.
10960         (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
10961         to use VPERMI2 mode iterator, add another alternative for vpermi2*
10962         instructions, remove the other two patterns.
10963         (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
10964         mode iterator, remove the other two patterns.
10965         * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
10966         (ix86_expand_vec_perm_vpermt2): ... this.  Swap mask and op0
10967         arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
10968         and adjust argument order accordingly.
10969         (ix86_expand_vec_perm): Adjust caller.
10970         (expand_vec_perm_1): Likewise.
10971         (expand_vec_perm_vpermi2_vpshub2): Rename to ...
10972         (expand_vec_perm_vpermt2_vpshub2): ... this.
10973         (ix86_expand_vec_perm_const_1): Adjust caller.
10974         (ix86_vectorize_vec_perm_const_ok): Adjust comments.
10976         PR target/82370
10977         * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
10978         (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
10979         (vec_shl_<mode>): Remove unused expander.
10980         (avx512bw_<shift_insn><mode>3): New define_insn.
10981         (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
10982         (<sse2_avx2>_<shift_insn><mode>3): ... this.  New define_insn.
10984 2017-10-24  Paolo Carlini  <paolo.carlini@oracle.com>
10986         PR c++/82466
10987         * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
10988         description.
10990 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10992         PR rtl-optimization/82396
10993         * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
10994         (autopref_multipass_init): Simplify initialization.
10995         (autopref_rank_data): Simplify sort order.
10996         * gcc/sched-int.h (autopref_multipass_data_): Remove
10997         multi_mem_insn_p, min_offset and max_offset.
10999 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
11001         PR middle-end/60580
11002         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
11003         Check special value of flag_omit_frame_pointer.
11004         (aarch64_can_eliminate): Likewise.
11005         (aarch64_override_options_after_change_1): Simplify handling of
11006         -fomit-frame-pointer and -fomit-leaf-frame-pointer.
11008 2017-10-24  Richard Biener  <rguenther@suse.de>
11010         PR tree-optimization/82697
11011         * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
11012         zero for conditional load and unconditional store.
11014 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
11016         * doc/install.texi: Document bootstrap-cet.
11018 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
11020         PR target/82659
11021         * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
11022         ENDBR instruction at function entrance if function is only
11023         called directly.
11025 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
11027         PR target/82628
11028         * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
11029         patterns to better describe from which operation the CF is computed.
11030         (addcarry<mode>_0, subborrow<mode>_0): New patterns.
11031         * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
11032         one LTU with [DT]Imode and another one with [SD]Imode.  If arg0
11033         is 0, use _0 suffixed expanders instead of emitting a comparison
11034         before it.
11036 2017-10-06  Sergey Shalnov  <Sergey.Shalnov@intel.com>
11038         * config/i386/i386.md(*movsf_internal, *movdf_internal):
11039         Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
11041 2017-10-24  Eric Botcazou  <ebotcazou@adacore.com>
11043         PR middle-end/82569
11044         * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
11045         * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
11046         * loop-iv.c (iv_get_reaching_def): Likewise.
11047         * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
11048         variable is promoted and the partition contains undefined values.
11050 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11052         * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
11053         reflect reality.
11054         (nios2_address_cost): Define.
11055         (nios2_legitimize_address): Recognize (exp + constant) directly.
11056         (TARGET_ADDRESS_COST): Define.
11058 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11060         * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
11061         (nios2_symbolic_memory_operand_p): Declare.
11062         (nios2_split_large_constant): Declare.
11063         (nios2_split_symbolic_memory_operand): Declare.
11064         * config/nios2/nios2.c: Adjust includes.
11065         (nios2_symbolic_constant_allowed): New.
11066         (nios2_symbolic_constant_p): New.
11067         (nios2_plus_symbolic_constant_p): New.
11068         (nios2_valid_addr_expr_p): Recognize addresses involving
11069         symbolic constants.
11070         (nios2_legitimate_address_p): Likewise, also LO_SUM.
11071         (nios2_symbolic_memory_operand_p): New.
11072         (nios2_large_constant_p): New.
11073         (nios2_split_large_constant): New.
11074         (nios2_split_plus_large_constant): New.
11075         (nios2_split_symbolic_memory_operand): New.
11076         (nios2_legitimize_address): Code refactoring.  Handle addresses
11077         involving symbolic constants.
11078         (nios2_emit_move_sequence): Likewise.
11079         (nios2_print_operand): Improve error output.
11080         (nios2_print_operand_address): Handle LO_SUM.
11081         (nios2_cdx_narrow_form_p): Likewise.
11082         * config/nios2/nios2.md (movqi_internal): Add splitter for memory
11083         operands involving symbolic constants.
11084         (movhi_internal, movsi_internal): Likewise.
11085         (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
11086         (extendhisi2, extendqi<mode>2): Likewise.
11088 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11090         * tree-pass.h (PROP_rtl_split_insns): Define.
11091         * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
11093 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11095         * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
11097 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11099         PR debug/82630
11100         * target.def (const_not_ok_for_debug_p): Default to
11101         default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
11102         * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
11103         * targhooks.c (default_const_not_ok_for_debug_p): New function.
11104         * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
11105         which targetm.const_not_ok_for_debug_p returned true.
11106         * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
11107         for UNSPECs.
11108         * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
11109         Likewise.
11110         * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
11111         * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
11112         UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
11113         if !base_term_p.
11114         (ix86_const_not_ok_for_debug_p): New function.
11115         (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
11116         (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
11118 2017-10-23  David Malcolm  <dmalcolm@redhat.com>
11120         PR bootstrap/82610
11121         * system.h: Conditionally include "unique-ptr.h" if
11122         INCLUDE_UNIQUE_PTR is defined.
11123         * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
11124         of defining INCLUDE_UNIQUE_PTR before including "system.h".
11126 2017-10-23  Sebastian Perta  <sebastian.perta@renesas.com>
11128         * config/rl78/rl78.md: New define_expand "subdi3".
11130 2017-10-23  H.J. Lu  <hongjiu.lu@intel.com>
11132         PR target/82673
11133         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
11134         DF_REF_INSN if DF_REF_INSN_INFO is false.
11136 2017-10-23  Jan Hubicka  <hubicka@ucw.cz>
11138         * i386.c (dimode_scalar_chain::compute_convert_gain): Use
11139         xmm_move instead of sse_move.
11140         (sse_store_index): New function.
11141         (ix86_register_move_cost): Be more sensible about mismatch stall;
11142         model AVX moves correctly; make difference between sse->integer and
11143         integer->sse.
11144         (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
11145         moves; make difference between SSE and AVX.
11146         * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
11147         and zmm_move. Increase size of sse load and store tables;
11148         add unaligned load and store tables; add ssemmx_to_integer.
11149         * x86-tune-costs.h: Update all entries according to real
11150         move latencies from Agner Fog's manual and chip documentation.
11152 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11154         PR target/82628
11155         * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
11156         * config/i386/constraints.md (Wf): New constraint.
11157         * config/i386/i386.md (UNSPEC_SBB): New unspec.
11158         (cmp<dwi>_doubleword): Removed.
11159         (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
11160         (sub<mode>3_carry_ccgz): Use unspec instead of compare.
11161         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
11162         expand with cmp<dwi>_doubleword.  For LTU and GEU use
11163         sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
11165         * common.opt (gcolumn-info): Enable by default.
11166         * doc/invoke.texi (gcolumn-info): Document new default.
11168 2017-10-23  Richard Biener  <rguenther@suse.de>
11170         PR tree-optimization/82672
11171         * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
11172         Fold the stmt if we propagated into it.
11174 2017-10-23  Richard Biener  <rguenther@suse.de>
11176         * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
11177         (bitmap_remove_expr_from_set): ... this.  All callers call this
11178         for non-constant values.
11179         (bitmap_set_subtract): Rename to...
11180         (bitmap_set_subtract_expressions): ... this.  Adjust and
11181         optimize.
11182         (bitmap_set_contains_value): Remove superfluous check.
11183         (bitmap_set_replace_value): Inline into single caller ...
11184         (bitmap_value_replace_in_set): ... here and simplify.
11185         (dependent_clean): Merge into ...
11186         (clean): ... this using an overload.  Adjust.
11187         (prune_clobbered_mems): Adjust.
11188         (compute_antic_aux): Likewise.
11189         (compute_partial_antic_aux): Likewise.
11191 2017-10-23  Richard Biener  <rguenther@suse.de>
11193         PR tree-optimization/82129
11194         Revert
11195         2017-08-01  Richard Biener  <rguenther@suse.de>
11197         PR tree-optimization/81181
11198         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
11199         (compute_antic): ... end of iteration here.
11201 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11203         * target.def (starting_frame_offset): New hook.
11204         * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
11205         (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
11206         * doc/tm.texi.in: Regenerate.
11207         * hooks.h (hook_hwi_void_0): Declare.
11208         * hooks.c (hook_hwi_void_0): New function.
11209         * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
11210         STARTING_FRAME_OFFSET.
11211         * builtins.c (expand_builtin_setjmp_receiver): Likewise.
11212         * reload1.c (reload): Likewise.
11213         * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
11214         instead of STARTING_FRAME_OFFSET.
11215         * function.c (try_fit_stack_local): Likewise.
11216         (assign_stack_local_1): Likewise
11217         (instantiate_virtual_regs): Likewise.
11218         * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
11219         * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
11220         * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
11221         * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
11222         * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
11223         * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
11224         * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
11225         * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
11226         * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
11227         * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
11228         * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
11229         * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
11230         * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
11231         * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
11232         * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
11233         * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
11234         * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
11235         * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
11236         * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
11237         * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11238         * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11239         * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11240         * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11241         * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11242         * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11243         * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11244         * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11245         * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11246         * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11247         * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11248         * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11249         * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11250         * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11251         * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11252         * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11253         * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11254         * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11255         * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11256         * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11257         * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11258         * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11259         * config/avr/avr.c (avr_starting_frame_offset): Make static and
11260         return a HOST_WIDE_INT.
11261         (avr_builtin_setjmp_frame_value): Use it instead of
11262         STARTING_FRAME_OFFSET.
11263         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11264         * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11265         * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11266         New function.
11267         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11268         * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11269         * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11270         (TARGET_CONSTANT_ALIGNMENT): Redefine.
11271         * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11272         * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11273         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11274         * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11275         * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11276         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11277         * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11278         * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11279         New function.
11280         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11281         * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11282         * config/mips/mips.c (mips_compute_frame_info): Refer to
11283         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11284         (mips_starting_frame_offset): New function.
11285         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11286         * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11287         * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11288         * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11289         and return a HOST_WIDE_INT.
11290         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11291         (mmix_initial_elimination_offset): Refer to
11292         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11293         * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11294         * config/pa/pa.c (pa_starting_frame_offset): New function.
11295         (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11296         (pa_expand_prologue): Likewise.
11297         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11298         * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11299         !FRAME_GROWS_DOWNWARD handling to...
11300         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11301         * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11302         !FRAME_GROWS_DOWNWARD handling to...
11303         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11304         * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11305         !FRAME_GROWS_DOWNWARD handling to...
11306         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11307         * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11308         Redefine.
11309         (rs6000_starting_frame_offset): New function.
11310         * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11311         !FRAME_GROWS_DOWNWARD handling to...
11312         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11313         * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11314         !FRAME_GROWS_DOWNWARD handling to...
11315         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11316         * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11317         !FRAME_GROWS_DOWNWARD handling to...
11318         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11319         * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11320         (rs6000_starting_frame_offset): New function.
11321         * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11322         * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11323         * config/vax/vax.c (vax_starting_frame_offset): New function.
11324         (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11325         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11326         * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11327         * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11328         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11329         * system.h (STARTING_FRAME_OFFSET): Poison.
11331 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11333         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11334         SCALAR_TYPE_MODE instead of TYPE_MODE.
11336 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11337             Alan Hayward  <alan.hayward@arm.com>
11338             David Sherwood  <david.sherwood@arm.com>
11340         * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_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         * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11348 2017-10-23  Richard Biener  <rguenther@suse.de>
11350         PR tree-optimization/82129
11351         * tree-ssa-pre.c (bitmap_set_and): Remove.
11352         (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11353         canonicalizing expressions in the set to those with lowest
11354         ID rather than taking that from the first edge.
11356 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11358         * combine.c (rtx_equal_for_field_assignment_p): Use
11359         byte_lowpart_offset.
11361 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11362             Alan Hayward  <alan.hayward@arm.com>
11363             David Sherwood  <david.sherwood@arm.com>
11365         * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11366         to a SUBREG_PROMOTED_VAR.
11368 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11369             Alan Hayward  <alan.hayward@arm.com>
11370             David Sherwood  <david.sherwood@arm.com>
11372         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11373         (expand_debug_source_expr): Likewise.
11374         * combine.c (combine_simplify_rtx): Likewise.
11375         * cse.c (fold_rtx): Likewise.
11376         * optabs.c (expand_float): Likewise.
11377         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11378         (simplify_binary_operation_1): Likewise.
11380 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11381             Alan Hayward  <alan.hayward@arm.com>
11382             David Sherwood  <david.sherwood@arm.com>
11384         * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11385         (record_promoted_value): Likewise.
11386         * expr.c (expand_expr_real_2): Likewise.
11387         * ree.c (update_reg_equal_equiv_notes): Likewise.
11388         (combine_set_extension): Likewise.
11389         * rtlanal.c (low_bitmask_len): Likewise.
11390         * simplify-rtx.c (neg_const_int): Likewise.
11391         (simplify_binary_operation_1): Likewise.
11393 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11394             Alan Hayward  <alan.hayward@arm.com>
11395             David Sherwood  <david.sherwood@arm.com>
11397         * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11398         * regcprop.c (maybe_mode_change): Likewise.
11399         * reload1.c (alter_reg): Likewise.
11401 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11403         * inchash.h (inchash::hash::add_wide_int): New function.
11404         * lto-streamer-out.c (hash_tree): Use it.
11406 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11408         * inchash.h (inchash::hash::add_wide_int): Rename to...
11409         (inchash::hash::add_hwi): ...this.
11410         * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11411         (polymorphic_call_target_hasher::hash): Likewise.
11412         * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11413         (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11414         (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11415         * lto-streamer-out.c (hash_tree): Likewise.
11416         * optc-save-gen.awk: Likewise.
11417         * tree.c (add_expr): Likewise.
11419 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11421         PR target/52451
11422         * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11423         for ordered inequality comparisons even with TARGET_IEEE_FP.
11425 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11427         PR target/82628
11428         * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11429         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11430         Expand with cmp<dwi>_doubleword.
11432 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11434         * extend.texi: Add x86 specific to 'nocf_check' attribute.
11435         List CET intrinsics.
11436         * invoke.texi: Add -mcet, -mibt, -mshstk options.  Add x86
11437         specific to -fcf-protection option.
11439 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11441         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11442         (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11443         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11444         (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11445         (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11446         * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11447         (extra_objs): Add cet.o for Linux/x86 targets.
11448         (tmake_file): Add i386/t-cet for Linux/x86 targets.
11449         * config/i386/cet.c: New file.
11450         * config/i386/cetintrin.h: Likewise.
11451         * config/i386/t-cet: Likewise.
11452         * config/i386/cpuid.h (bit_SHSTK): New.
11453         (bit_IBT): Likewise.
11454         * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11455         pass IBT and SHSTK bits.
11456         * config/i386/i386-builtin-types.def
11457         (VOID_FTYPE_UNSIGNED_PVOID): New.
11458         (VOID_FTYPE_UINT64_PVOID): Likewise.
11459         * config/i386/i386-builtin.def: Add CET intrinsics.
11460         * config/i386/i386-c.c (ix86_target_macros_internal): Add
11461         OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11462         * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11463         * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11464         prototype.
11465         * config/i386/i386.c (rest_of_insert_endbranch): New.
11466         (pass_data_insert_endbranch): Likewise.
11467         (pass_insert_endbranch): Likewise.
11468         (make_pass_insert_endbranch): Likewise.
11469         (ix86_notrack_prefixed_insn_p): Likewise.
11470         (ix86_target_string): Add -mibt, -mshstk flags.
11471         (ix86_option_override_internal): Add flag_cf_protection
11472         processing.
11473         (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11474         (ix86_print_operand): Add 'notrack' prefix output.
11475         (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11476         (ix86_expand_builtin): Expand CET intrinsics.
11477         (x86_output_mi_thunk): Add 'endbranch' instruction.
11478         * config/i386/i386.h (TARGET_IBT): New.
11479         (TARGET_IBT_P): Likewise.
11480         (TARGET_SHSTK): Likewise.
11481         (TARGET_SHSTK_P): Likewise.
11482         * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11483         UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11484         UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11485         (builtin_setjmp_setup): New pattern.
11486         (builtin_longjmp): Likewise.
11487         (rdssp<mode>): Likewise.
11488         (incssp<mode>): Likewise.
11489         (saveprevssp): Likewise.
11490         (rstorssp): Likewise.
11491         (wrss<mode>): Likewise.
11492         (wruss<mode>): Likewise.
11493         (setssbsy): Likewise.
11494         (clrssbsy): Likewise.
11495         (nop_endbr): Likewise.
11496         * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11497         options.
11498         * config/i386/immintrin.h: Include <cetintrin.h>.
11499         * config/i386/linux-common.h
11500         (file_end_indicate_exec_stack_and_cet): New prototype.
11501         (TARGET_ASM_FILE_END): New.
11503 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11505         * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11506         latencies instead of having separate table; make difference between
11507         integer and float costs.
11508         * i386.h (processor_costs): Remove scalar_stmt_cost,
11509         scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11510         scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11511         vec_store_cost.
11512         * x86-tune-costs.h: Remove entries which has been removed in
11513         procesor_costs from all tables; make cond_taken_branch_cost
11514         and cond_not_taken_branch_cost COST_N_INSNS based.
11516 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11518         * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11520 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11522         * config/i386/i386.md (isa): Remove fma_avx512f.
11523         * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11524         <avx512>_fmadd_<mode>_mask3<round_name>,
11525         <avx512>_fmsub_<mode>_mask<round_name>,
11526         <avx512>_fmsub_<mode>_mask3<round_name>,
11527         <avx512>_fnmadd_<mode>_mask<round_name>,
11528         <avx512>_fnmadd_<mode>_mask3<round_name>,
11529         <avx512>_fnmsub_<mode>_mask<round_name>,
11530         <avx512>_fnmsub_<mode>_mask3<round_name>,
11531         <avx512>_fmaddsub_<mode>_mask<round_name>,
11532         <avx512>_fmaddsub_<mode>_mask3<round_name>,
11533         <avx512>_fmsubadd_<mode>_mask<round_name>,
11534         <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11535         (*vec_widen_umult_even_v16si<mask_name>,
11536         *vec_widen_smult_even_v16si<mask_name>): Likewise.
11537         (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11539 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11541         * extend.texi: Add 'nocf_check' documentation.
11542         * gimple.texi: Add second parameter to
11543         gimple_build_call_from_tree.
11544         * invoke.texi: Add -fcf-protection documentation.
11545         * rtl.texi: Add REG_CALL_NOTRACK documenation.
11547 2017-10-20  Richard Biener  <rguenther@suse.de>
11549         PR tree-optimization/82473
11550         * tree-vect-loop.c (vectorizable_reduction): Properly get at
11551         the largest input type.
11553 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11555         * c-attribs.c (handle_nocf_check_attribute): New function.
11556         (c_common_attribute_table): Add 'nocf_check' handling.
11557         * gimple-parser.c: Add second argument NULL to
11558         gimple_build_call_from_tree.
11559         * attrib.c (comp_type_attributes): Check nocf_check attribute.
11560         * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11561         call insn.
11562         * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11563         * common.opt: Add fcf-protection flag.
11564         * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11565         * flag-types.h: Add enum cf_protection_level.
11566         * gimple.c (gimple_build_call_from_tree): Add second parameter.
11567         Add 'nocf_check' attribute propagation to gimple call.
11568         * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11569         (gimple_build_call_from_tree): Update prototype.
11570         (gimple_call_nocf_check_p): New function.
11571         (gimple_call_set_nocf_check): Likewise.
11572         * gimplify.c: Add second argument to gimple_build_call_from_tree.
11573         * ipa-icf.c: Add nocf_check attribute in statement hash.
11574         * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11575         * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11576         * toplev.c (process_options): Add flag_cf_protection handling.
11578 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11580         * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11582 2017-10-20  Richard Biener  <rguenther@suse.de>
11584         PR tree-optimization/82603
11585         * tree-if-conv.c (predicate_mem_writes): Make sure to only
11586         remove false predicated stores.
11588 2017-10-20  Richard Biener  <rguenther@suse.de>
11590         * graphite-isl-ast-to-gimple.c
11591         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11592         Remove return value and simplify, dump copied stmt after lhs
11593         adjustment.
11594         (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11595         Reduce dump verbosity.
11596         (gsi_insert_earliest): Likewise.
11597         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11598         * graphite.c (print_global_statistics): Adjust dumping.
11599         (print_graphite_scop_statistics): Likewise.
11600         (print_graphite_statistics): Do not dump loops here.
11601         (graphite_transform_loops): But here.
11603 2017-10-20  Nicolas Roche  <roche@adacore.com>
11605         * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11606         * configure: Regenerate.
11608 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11610         PR target/82158
11611         * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11612         functions when optimizing replace GIMPLE_RETURN stmts with
11613         calls to __builtin_unreachable ().
11615         PR sanitizer/82595
11616         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11617         for -fsanitize=thread link of executables.
11618         (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11619         link of executables.
11621         PR target/82370
11622         * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11623         New mode iterators.
11624         (<shift_insn><mode>3<mask_name>): Change the last of the 3
11625         define_insns for logical vector shifts to use VI248_AVX512BW
11626         iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11627         condition, useless isa and prefix attributes.  Change the first
11628         2 of these define_insns to ...
11629         (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11630         define_insn for avx512vl.
11631         (<shift_insn><mode>3): ... and this, new define_insn without
11632         masking for non-avx512vl.
11634         PR target/82370
11635         * config/i386/sse.md (*andnot<mode>3,
11636         <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11637         (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11638         and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11639         not applied use empty suffix even for TARGET_AVX512VL.
11640         * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11641         is applied, supply evex,evex or evex,evex,evex instead of just
11642         evex.
11644 2017-10-20  Julia Koval  <julia.koval@intel.com>
11646         * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11647         (OPTION_MASK_ISA_GFNI_UNSET): New.
11648         (ix86_handle_option): Handle OPT_mgfni.
11649         * config/i386/cpuid.h (bit_GFNI): New.
11650         * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11651         * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11652         * config/i386/i386.c (ix86_target_string): Add -mgfni.
11653         (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11654         * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11655         * config/i386/i386.opt: Add mgfni.
11657 2017-10-20  Orlando Arias  <oarias@knights.ucf.edu>
11659         * config/msp430/msp430.c (msp430_option_override): Disable
11660         -fdelete-null-pointer-checks.
11661         * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11663 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11665         * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11666         of x87 and SSE instructions.
11668 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11670         * asan.c (create_cond_insert_point): Do not update edge count.
11671         * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11672         (afdo_propagate_circuit): Likewise.
11673         (afdo_calculate_branch_prob): Likewise.
11674         (afdo_annotate_cfg): Likewise.
11675         * basic-block.h (struct edge_def): Remove count.
11676         (edge_def::count): New accessor.
11677         * bb-reorder.c (rotate_loop): Update.
11678         (find_traces_1_round): Update.
11679         (connect_traces): Update.
11680         (sanitize_hot_paths): Update.
11681         * cfg.c (unchecked_make_edge): Update.
11682         (make_single_succ_edge): Update.
11683         (check_bb_profile): Update.
11684         (dump_edge_info): Update.
11685         (update_bb_profile_for_threading): Update.
11686         (scale_bbs_frequencies_int): Update.
11687         (scale_bbs_frequencies_gcov_type): Update.
11688         (scale_bbs_frequencies_profile_count): Update.
11689         (scale_bbs_frequencies): Update.
11690         * cfganal.c (connect_infinite_loops_to_exit): Update.
11691         * cfgbuild.c (compute_outgoing_frequencies): Update.
11692         (find_many_sub_basic_blocks): Update.
11693         * cfgcleanup.c (try_forward_edges): Update.
11694         (try_crossjump_to_edge): Update
11695         * cfgexpand.c (expand_gimple_cond): Update
11696         (expand_gimple_tailcall): Update
11697         (construct_exit_block): Update
11698         * cfghooks.c (verify_flow_info): Update
11699         (redirect_edge_succ_nodup): Update
11700         (split_edge): Update
11701         (make_forwarder_block): Update
11702         (duplicate_block): Update
11703         (account_profile_record): Update
11704         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11705         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11706         * cfgloopmanip.c (scale_loop_profile): Update.
11707         (loopify): Update.
11708         (lv_adjust_loop_entry_edge): Update.
11709         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11710         (force_nonfallthru_and_redirect): Update.
11711         (purge_dead_edges): Update.
11712         (rtl_flow_call_edges_add): Update.
11713         * cgraphunit.c (init_lowered_empty_function): Update.
11714         (cgraph_node::expand_thunk): Update.
11715         * gimple-pretty-print.c (dump_probability): Update.
11716         (dump_edge_probability): Update.
11717         * gimple-ssa-isolate-paths.c (isolate_path): Update.
11718         * haifa-sched.c (sched_create_recovery_edges): Update.
11719         * hsa-gen.c (convert_switch_statements): Update.
11720         * ifcvt.c (dead_or_predicable): Update.
11721         * ipa-inline-transform.c (inline_transform): Update.
11722         * ipa-split.c (split_function): Update.
11723         * ipa-utils.c (ipa_merge_profiles): Update.
11724         * loop-doloop.c (add_test): Update.
11725         * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11726         * lto-streamer-in.c (input_cfg): Update.
11727         (input_function): Update.
11728         * lto-streamer-out.c (output_cfg): Update.
11729         * modulo-sched.c (sms_schedule): Update.
11730         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11731         * predict.c (maybe_hot_edge_p): Update.
11732         (unlikely_executed_edge_p): Update.
11733         (probably_never_executed_edge_p): Update.
11734         (dump_prediction): Update.
11735         (drop_profile): Update.
11736         (propagate_unlikely_bbs_forward): Update.
11737         (determine_unlikely_bbs): Update.
11738         (force_edge_cold): Update.
11739         * profile.c (compute_branch_probabilities): Update.
11740         * reg-stack.c (better_edge): Update.
11741         * shrink-wrap.c (handle_simple_exit): Update.
11742         * tracer.c (better_p): Update.
11743         * trans-mem.c (expand_transaction): Update.
11744         (split_bb_make_tm_edge): Update.
11745         * tree-call-cdce.c: Update.
11746         * tree-cfg.c (gimple_find_sub_bbs): Update.
11747         (gimple_split_edge): Update.
11748         (gimple_duplicate_sese_region): Update.
11749         (gimple_duplicate_sese_tail): Update.
11750         (gimple_flow_call_edges_add): Update.
11751         (insert_cond_bb): Update.
11752         (execute_fixup_cfg): Update.
11753         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11754         * tree-complex.c (expand_complex_div_wide): Update.
11755         * tree-eh.c (lower_resx): Update.
11756         (unsplit_eh): Update.
11757         (cleanup_empty_eh_move_lp): Update.
11758         * tree-inline.c (copy_edges_for_bb): Update.
11759         (freqs_to_counts): Update.
11760         (copy_cfg_body): Update.
11761         * tree-ssa-dce.c (remove_dead_stmt): Update.
11762         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11763         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11764         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11765         (unloop_loops): Update.
11766         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11767         * tree-ssa-loop-split.c (connect_loops): Update.
11768         (split_loop): Update.
11769         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11770         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11771         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11772         * tree-ssa-reassoc.c (branch_fixup): Update.
11773         * tree-ssa-tail-merge.c (replace_block_by): Update.
11774         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11775         (compute_path_counts): Update.
11776         (update_profile): Update.
11777         (recompute_probabilities): Update.
11778         (update_joiner_offpath_counts): Update.
11779         (estimated_freqs_path): Update.
11780         (freqs_to_counts_path): Update.
11781         (clear_counts_path): Update.
11782         (ssa_fix_duplicate_block_edges): Update.
11783         (duplicate_thread_path): Update.
11784         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11785         (case_bit_test_cmp): Update.
11786         (collect_switch_conv_info): Update.
11787         (gen_inbound_check): Update.
11788         (do_jump_if_equal): Update.
11789         (emit_cmp_and_jump_insns): Update.
11790         * tree-tailcall.c (decrease_profile): Update.
11791         (eliminate_tail_call): Update.
11792         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11793         (vect_do_peeling): Update.
11794         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11795         * ubsan.c (ubsan_expand_null_ifn): Update.
11796         (ubsan_expand_ptr_ifn): Update.
11797         * value-prof.c (gimple_divmod_fixed_value): Update.
11798         (gimple_mod_pow2): Update.
11799         (gimple_mod_subtract): Update.
11800         (gimple_ic): Update.
11801         (gimple_stringop_fixed_value): Update.
11803 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11805         PR target/82618
11806         * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11808 2017-10-19  Alexander Monakov  <amonakov@ispras.ru>
11810         PR rtl-optimization/82395
11811         * ira-color.c (allocno_priority_compare_func): Fix comparison step
11812         based on non_spilled_static_chain_regno_p.
11814 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11816         * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11817         (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11818         (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11820 2017-10-19  Martin Sebor  <msebor@redhat.com>
11822         PR tree-optimization/82596
11823         * tree.c (array_at_struct_end_p): Handle STRING_CST.
11825 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11827         * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11828         (get_mem_refs_of_builtin_call): Likewise.
11829         * builtins.c (expand_builtin_apply): Adjust call to
11830         allocate_dynamic_stack_space.
11831         (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11832         the third argument to allocate_dynamic_stack_space, otherwise -1.
11833         (expand_builtin): Deal with all alloca variants.
11834         (is_inexpensive_builtin): Likewise.
11835         * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11836         * calls.c (special_function_p): Deal with all alloca variants.
11837         (initialize_argument_information): Adjust call to
11838         allocate_dynamic_stack_space.
11839         (expand_call): Likewise.
11840         * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11841         * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11842         * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11843         use it for the stack usage computation.
11844         * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11845         * function.c (gimplify_parameters): Call build_alloca_call_expr.
11846         * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11847         Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11848         (in_loop_p): Remove first argument and useless check.
11849         (pass_walloca::execute): Remove useless test and adjust call to above.
11850         * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11851         * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11852         (gimplify_call_expr): Deal with all alloca variants.
11853         * hsa-gen.c (gen_hsa_alloca): Likewise.
11854         (gen_hsa_insns_for_call): Likewise.
11855         * ipa-pure-const.c (special_builtin_state): Likewise.
11856         * tree-chkp.c (chkp_build_returned_bound): Likewise.
11857         * tree-object-size.c (alloc_object_size): Likewise.
11858         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11859         (call_may_clobber_ref_p_1): Likewise.
11860         * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11861         (ccp_fold_stmt): Likewise.
11862         (optimize_stack_restore): Likewise.
11863         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11864         (mark_all_reaching_defs_necessary_1): Likewise.
11865         (propagate_necessity): Likewise.
11866         (eliminate_unnecessary_stmts): Likewise.
11867         * tree.c (build_common_builtin_nodes): Build
11868         BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11869         (build_alloca_call_expr): New function.
11870         * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11871         (CASE_BUILT_IN_ALLOCA): Likewise.
11872         (build_alloca_call_expr): Declare.
11873         * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11875 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11877         PR debug/82509
11878         * dwarf2out.c (new_die_raw): New static inline function.
11879         (new_die): Use it to create the DIE.
11880         (add_AT_external_die_ref): Likewise.
11881         (clone_die): Likewise.
11882         (clone_as_declaration): Likewise.
11883         (dwarf2out_vms_debug_main_pointer): Likewise.
11884         (base_type_die): Likewise.  Remove early return for corner cases.
11885         Do not call add_pubtype on the DIE here.
11886         (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11887         (modified_type_die): Adjust the lookup for reverse order DIEs.  Skip
11888         typedefs for base types with DW_AT_endianity.  Make sure a DIE with
11889         native order exists for base types, attach the DIE manually and call
11890         add_pubtype on it.  Do not equate a reverse order DIE to the type.
11892 2017-10-19  Richard Earnshaw  <rearnsha@arm.com>
11894         * config/arm/arm.c (align_ok_ldrd_strd): New function.
11895         (mem_ok_for_ldrd_strd): New parameter align.  Extract the alignment of
11896         the mem into it.
11897         (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11899 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11901         * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN.  Or
11902         SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11903         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11904         BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11905         * opts.c (sanitizer_opts): Add builtin.
11906         * ubsan.c (instrument_builtin): New function.
11907         (pass_ubsan::execute): Call it.
11908         (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11909         * doc/invoke.texi: Document -fsanitize=builtin.
11911         * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
11912         builtins, store max (log2 (align), 0) into uchar field instead of
11913         align into uptr field.
11914         (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
11915         store uchar 0 field instead of uptr 0 field.
11916         (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
11917         instead of passing one address of struct with 2 locations pass
11918         two addresses of structs with 1 location each.
11919         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
11920         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
11921         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
11922         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
11923         (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
11924         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
11925         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
11926         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
11928 2017-10-19  Martin Liska  <mliska@suse.cz>
11930         PR driver/81829
11931         * file-find.c (remove_prefix): Remove.
11932         * file-find.h (remove_prefix): Likewise.
11933         * gcc-ar.c: Remove smartness of lookup.
11935 2017-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
11937         * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
11938         *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
11939         *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
11941 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11943         PR target/82580
11944         * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
11945         (setcc + and to xor + setcc): New peephole2.
11947 2017-10-19  Tom de Vries  <tom@codesourcery.com>
11949         * doc/sourcebuild.texi (Test Directives, Variants of
11950         dg-require-support): Add dg-require-stack-size.
11952 2017-10-19  Martin Liska  <mliska@suse.cz>
11954         PR sanitizer/82517
11955         * gimplify.c (gimplify_decl_expr): Do not instrument variables
11956         that have a large alignment.
11957         (gimplify_target_expr): Likewise.
11959 2017-10-18  Segher Boessenkool  <segher@kernel.crashing.org>
11961         PR rtl-optimization/82602
11962         * ira.c (rtx_moveable_p): Return false for volatile asm.
11964 2017-10-18  Uros Bizjak  <ubizjak@gmail.com>
11966         PR target/82580
11967         * config/i386/i386-modes.def (CCGZ): New CC mode.
11968         * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
11969         * config/i386/predicates.md (ix86_comparison_operator):
11970         Handle CCGZmode.
11971         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11972         Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
11973         with double-word subtraction.
11974         (put_condition_code): Handle CCGZmode.
11976 2017-10-18  Aldy Hernandez  <aldyh@redhat.com>
11978         * wide-int.cc (debug (const wide_int &)): New.
11979         (debug (const wide_int *)): New.
11980         (debug (const widest_int &)): New.
11981         (debug (const widest_int *)): New.
11983 2017-10-18  Vladimir Makarov  <vmakarov@redhat.com>
11985         PR middle-end/82556
11986         * lra-constraints.c (curr_insn_transform): Use non-input operand
11987         instead of output one for matched reload.
11989 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11991         * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
11992         (tree-ssa-loop-ivopts.h): New header file.
11993         (struct builtin_info): New fields.
11994         (classify_builtin_1): Compute and record base and offset parts for
11995         memset builtin partition by calling strip_offset.
11996         (offset_cmp, fuse_memset_builtins): New functions.
11997         (finalize_partitions): Fuse adjacent memset partitions by calling
11998         above function.
11999         * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
12000         Expose the interface.
12001         * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
12003 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
12005         PR tree-optimization/82574
12006         * tree-loop-distribution.c (find_single_drs): New parameter.  Check
12007         that data reference must be executed exactly once per iteration
12008         against the outermost loop in nest.
12009         (classify_partition): Update call to above function.
12011 2017-10-18  Richard Biener  <rguenther@suse.de>
12013         PR tree-optimization/82591
12014         * graphite.c (graphite_transform_loops): Move code gen message
12015         printing ...
12016         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12017         Here.  Handle scop_to_isl_ast failing.
12018         (scop_to_isl_ast): Limit the number of ISL operations.
12020 2017-10-18  Richard Biener  <rguenther@suse.de>
12022         * graphite-isl-ast-to-gimple.c
12023         (translate_isl_ast_to_gimple::set_rename): Simplify.
12024         (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
12025         (graphite_copy_stmts_from_block): ... here.
12026         (copy_bb_and_scalar_dependences): Simplify.
12027         (add_parameters_to_ivs_params): Canonicalize.
12028         (generate_entry_out_of_ssa_copies): Simplify.
12029         * graphite-sese-to-poly.c (extract_affine_name): Simplify
12030         by passing in ISL dimension.
12031         (parameter_index_in_region_1): Rename to ...
12032         (parameter_index_in_region): ... this.
12033         (extract_affine): Adjust assert, pass down parameter index.
12034         (add_param_constraints): Use range-info when available.
12035         (build_scop_context): Adjust.
12036         * sese.c (new_sese_info): Adjust.
12037         (free_sese_info): Likewise.
12038         * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
12039         Remove unused typedefs.
12040         (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
12042 2017-10-18  Martin Liska  <mliska@suse.cz>
12044         * combine.c (simplify_compare_const): Add gcc_fallthrough.
12046 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12048         * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
12049         (s390_sched_init): Do not reset s390_sched_state if we entered the
12050         current basic block via a fallthru edge and all others are unlikely.
12052 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12054         * config/s390/s390.c (NUM_SIDES): New variable.
12055         (LONGRUNNING_THRESHOLD): New variable.
12056         (LATENCY_FACTOR): New variable.
12057         (s390_sched_score): Decrease score for long-running instructions on
12058         wrong side.
12059         (s390_sched_variable_issue): Perform bookkeeping for long-running
12060         instructions.
12062 2017-10-18  Richard Biener  <rguenther@suse.de>
12064         * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
12065         Simplify with removal of the parameter rename map.
12066         (set_rename): Likewise.
12067         (should_copy_to_new_region): Likewise.
12068         (graphite_copy_stmts_from_block): Likewise.
12069         (copy_bb_and_scalar_dependences): Remove initialization of
12070         unused copied_bb_map.
12071         (copy_def): Remove.
12072         (copy_internal_parameters): Likewise.
12073         (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
12074         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12075         Use INTEGRAL_TYPE_P.
12076         (parameter_index_in_region_1): Rename to ...
12077         (assign_parameter_index_in_region): ... this.  Assert we have
12078         a parameter we handle.
12079         (scan_tree_for_params): Adjust.
12080         * sese.h (parameter_rename_map_t): Remove.
12081         (struct sese_info_t): Remove unused parameter_rename_map and
12082         copied_bb_map members.
12083         * sese.c (new_sese_info): Adjust.
12084         (free_sese_info): Likewise.
12086 2017-10-18  Martin Liska  <mliska@suse.cz>
12088         PR sanitizer/82545
12089         * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
12090         on an abnormal edge.
12092 2017-10-18  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12094         * doc/invoke.texi (ffunction-sections and fdata-sections):
12095         Update.
12097 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12099         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
12100         the use statement can throw internally.
12102 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12104         * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
12105         any RTX present on the RHS of a SET.
12106         * compare-elim.c (try_eliminate_compare): Restore comment.
12108 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12110         * langhooks.h (struct lang_hooks): Document that tree_size langhook
12111         may be also called on tcc_type nodes.
12112         * langhooks.c (lhd_tree_size): Likewise.
12114 2017-10-17  David Malcolm  <dmalcolm@redhat.com>
12116         * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
12117         format_warning_at_substring.
12118         (maybe_warn): Convert source_range * param to a location_t.  Pass
12119         UNKNOWN_LOCATION rather than NULL to fmtwarn.
12120         (format_directive): Remove code to extract source_ranges and
12121         source_range * in favor of just a location_t.
12122         (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
12123         fmtwarn.
12124         * substring-locations.c (format_warning_va): Convert
12125         source_range * param to a location_t.
12126         (format_warning_at_substring): Likewise.
12127         * substring-locations.h (format_warning_va): Likewise.
12128         (format_warning_at_substring): Likewise.
12130 2017-10-17  Jan Hubicka  <hubicka@ucw.cz>
12132         * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
12133         vec_scatter_store
12134         * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
12135         and scatter/gather ops.
12137         * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
12138         vec_gather_load and vec_scatter_store.
12139         * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
12140         * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
12141         * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
12142         * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
12143         * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
12144         * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
12146 2017-10-17  Uros Bizjak  <ubizjak@gmail.com>
12148         * reg-stack.c (compare_for_stack_reg): Add bool argument.
12149         Detect FTST instruction and handle its register pops.  Only pop
12150         second operand if can_pop_second_op is true.
12151         (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
12152         set can_pop_second_op to false in the compare_for_stack_reg call.
12154         * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
12155         output_fp_compare for stack register operands.
12156         * config/i386/i386.c (output_fp_compare): Do not output SSE compare
12157         instructions here.  Do not emit stack register pops here.  Assert
12158         that FCOMPP pops next to top stack register.  Rewrite function.
12160 2017-10-17  Nathan Sidwell  <nathan@acm.org>
12162         PR middle-end/82577
12163         * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
12164         use DECL_ASSEMBLER_NAME_RAW.
12166         PR middle-end/82546
12167         * tree.c (tree_code_size): Reformat.  Punt to lang hook for unknown
12168         TYPE nodes.
12170 2017-10-17  Qing Zhao <qing.zhao@oracle.com>
12171             Wilco Dijkstra <wilco.dijkstra@arm.com>
12173         * builtins.c (expand_builtin_update_setjmp_buf): Add a
12174         converstion to Pmode from the buf_addr.
12176 2017-10-17  Richard Biener  <rguenther@suse.de>
12178         * graphite-dependences.c (scop_get_reads_and_writes): Change
12179         output parameters to references.
12181 2017-10-17  Jackson Woodruff  <jackson.woodruff@arm.com>
12183         PR 71026/tree-optimization
12184         * fold-const.c (distribute_real_division): Removed.
12185         (fold_binary_loc): Remove calls to distribute_real_divison.
12187 2017-10-17  Richard Biener  <rguenther@suse.de>
12189         * graphite-scop-detection.c
12190         (scop_detection::stmt_has_simple_data_refs_p): Always use
12191         the full nest as region.
12192         (try_generate_gimple_bb): Likewise.
12193         * sese.c (scalar_evolution_in_region): Simplify now that
12194         SCEV can handle instantiation in regions.
12195         * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
12196         in the non-loop part of a function if requested.
12198 2017-10-17  Richard Biener  <rguenther@suse.de>
12200         PR tree-optimization/82563
12201         * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
12202         New function.
12203         (graphite_regenerate_ast_isl): Call it.
12204         * graphite-scop-detection.c (build_scops): Remove entry edge split.
12206 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12208         PR tree-optimization/82549
12209         * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
12210         Formatting fixes.  Instead of calling make_bit_field_ref with negative
12211         bitpos return 0.
12213 2017-10-17  Olga Makhotina  <olga.makhotina@intel.com>
12215         * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
12216         _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
12217         _mm_maskz_reduce_ss): New.
12218         * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
12219         __builtin_ia32_reducess_mask): Ditto..
12220         (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
12221         * config/i386/sse.md (reduces<mode>): Renamed to ...
12222         (reduces<mode><mask_scalar_name>): ... this.
12223         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
12224         Changed to ...
12225         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
12226         %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
12228 2017-10-16  David Malcolm  <dmalcolm@redhat.com>
12230         * Makefile.in (OBJS): Add unique-ptr-tests.o.
12231         * selftest-run-tests.c (selftest::run_tests): Call
12232         selftest::unique_ptr_tests_cc_tests.
12233         * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
12234         * unique-ptr-tests.cc: New file.
12236 2017-10-16  Vladimir Makarov  <vmakarov@redhat.com>
12238         PR sanitizer/82353
12239         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12240         locations.
12241         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12242         (make_hard_regno_born, make_hard_regno_dead): Update
12243         bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12245 2017-10-16  Jeff Law  <law@redhat.com>
12247         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12249 2017-10-16  Jan Hubicka  <hubicka@ucw.cz>
12251         * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12253 2017-10-16  Olivier Hainque  <hainque@adacore.com>
12255         * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12256         with_cpu if we were configured for an e500v2 target cpu name.
12258 2017-10-16  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12260         * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12261         * doc/invoke.texi: Document +nodsp as a valid extension for
12262         -mcpu=cortex-m33.
12264 2017-10-16  Martin Liska  <mliska@suse.cz>
12266         * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12267         (test_set_range): Likewise.
12268         (test_range_functions): Rename to ...
12269         (test_bit_in_range): ... this.
12270         (sbitmap_c_tests): Add new test.
12272 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12274         * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12275         New.
12276         (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12277         (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12279 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12281         * config/aarch64/aarch64-builtins.c
12282         (aarch64_types_quadopu_lane_qualifiers): New.
12283         (TYPES_QUADOPU_LANE): New.
12284         * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12285         (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12286         (aarch64_<sur>dot_laneq<vsi2qi>): New.
12287         * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12288         (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12289         * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12290         (Vdottype, DOTPROD): New.
12291         (sur): Add SDOT and UDOT.
12293 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12295         * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12296         (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12297         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12298         Add TARGET_DOTPROD.
12299         * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12300         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12301         Enable TARGET_DOTPROD.
12302         (cortex-a75.cortex-a55): Likewise.
12303         * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12305 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12307         * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12308         (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12309         New.
12310         * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12311         New.
12312         * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12313         (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12314         * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12315         (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12316         * config/arm/types.md (neon_dot, neon_dot_q): New.
12317         * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12319 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12321         * config/arm/arm.h (TARGET_DOTPROD): New.
12322         * config/arm/arm.c (arm_arch_dotprod): New.
12323         (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12324         * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12325         * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12326         (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12327         (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12328         * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12329         * doc/invoke.texi (armv8.2-a): Document dotprod
12331 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12333         * i386.c (ix86_vec_cost): New function.
12334         (ix86_rtx_costs): Handle vector operations better.
12335         * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12336         * x86-tune-costs.h: Add new costs to all tables.
12338 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12340         * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12341         operations.
12342         * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12343         divsd, sqrtss and sqrtsd
12344         * x86-tune-costs.h: Add new entries to all costs.
12345         (znver1_cost): Fix to match real instruction latencies.
12347 2017-10-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12348             Michael Collison <michael.collison@arm.com>
12350         * compare-elim.c: Include emit-rtl.h.
12351         (can_merge_compare_into_arith): New function.
12352         (try_validate_parallel): Likewise.
12353         (try_merge_compare): Likewise.
12354         (try_eliminate_compare): Call the above when no previous clobber
12355         is available.
12356         (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12357         dataflow problems.
12359 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12361         PR middle-end/62263
12362         PR middle-end/82498
12363         * tree-ssa-phiopt.c (value_replacement): Comment fix.  Handle
12364         up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12366         PR middle-end/62263
12367         PR middle-end/82498
12368         * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12369         to be any operand_equal_p operands.  For & (B - 1) require
12370         B to be power of 2.  Recognize
12371         (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12373 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12375         PR bootstrap/82553
12376         * optabs.c (expand_memory_blockage): Fix call of
12377         targetm.have_memory_blockage.
12379 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12381         PR bootstrap/82548
12382         * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12383         x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12384         objects to extra_objs instead of overwriting it.
12386 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12388         * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12389         Use any_fp_register_operand as operand[3] predicate.  Simplify
12390         equality test for operands[2] and operands[4] memory location.
12391         (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12392         (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12393         (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12394         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12395         any_fp_register_operand as operand[1] predicate.  Simplify
12396         equality test for operands[0] and operands[3] memory location.
12397         (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12398         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12399         (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12401 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12403         * target-insns.def: Add memory_blockage.
12404         * optabs.c (expand_memory_blockage): New function.
12405         (expand_asm_memory_barrier): Rename ...
12406         (expand_asm_memory_blockage): ... to this.
12407         (expand_mem_thread_fence): Call expand_memory_blockage
12408         instead of expand_asm_memory_barrier.
12409         (expand_mem_singnal_fence): Ditto.
12410         (expand_atomic_load): Ditto.
12411         (expand_atomic_store): Ditto.
12412         * doc/md.texi (Standard Pattern Names For Generation):
12413         Document memory_blockage instruction pattern.
12415 2017-10-13  Sebastian Perta  <sebastian.perta@renesas.com>
12417         * config/rl78/rl78.c (rl78_emit_libcall): New function.
12418         * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12419         * config/rl78/rl78.md: New define_expand "adddi3".
12421 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12423         * cfghooks.c (verify_flow_info): Disable check that all probabilities
12424         are set correctly.
12426 2017-10-13  Jeff Law  <law@redhat.com>
12428         * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12430 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12432         PR target/82274
12433         * internal-fn.c (expand_mul_overflow): If both operands have
12434         the same highpart of -1 or 0 and the topmost bit of lowpart
12435         is different, overflow is if res <= 0 rather than res < 0.
12437 2017-10-13  Pat Haugen  <pthaugen@us.ibm.com>
12439         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12440         TARGET_P9_VECTOR code for unaligned_load case.
12442 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12444         * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12446 2017-10-13  Nathan Sidwell  <nathan@acm.org>
12448         * tree-core.h (tree_contains_struct): Make bool.
12449         * tree.c (tree_contains_struct): Likewise.
12450         * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12451         (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12452         MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12453         MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12454         MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12456 2017-10-13  Richard Biener  <rguenther@suse.de>
12458         * graphite-isl-ast-to-gimple.c
12459         (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12460         parameters and dominance check.
12461         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12462         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12463         (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12464         Do not update SSA form here or do intermediate IL verification.
12465         * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12466         (graphite_initialize): Remove check on the number of loops in
12467         the function and inline into graphite_transform_loops.
12468         (graphite_finalize): Inline into graphite_transform_loops.
12469         (graphite_transform_loops): Perform SSA update and IL verification
12470         here.
12471         * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12473 2017-10-13  Richard Biener  <rguenther@suse.de>
12475         * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12476         graphite_expression_type_precision): Avoid global constructor
12477         by moving ...
12478         (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12479         (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12480         (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12481         (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12482         (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12483         * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12485 2017-10-13  H.J. Lu  <hongjiu.lu@intel.com>
12487         PR target/82499
12488         * config/i386/i386.h (ix86_red_zone_size): New.
12489         * config/i386/i386.md (push peephole2s): Replace
12490         "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12492 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12493             Alan Hayward  <alan.hayward@arm.com>
12494             David Sherwood  <david.sherwood@arm.com>
12496         * combine.c (can_change_dest_mode): Reject changes in
12497         REGMODE_NATURAL_SIZE.
12499 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12500             Alan Hayward  <alan.hayward@arm.com>
12501             David Sherwood  <david.sherwood@arm.com>
12503         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12504         (expand_debug_source_expr): Likewise.
12505         * combine.c (combine_simplify_rtx): Likewise.
12506         * cse.c (fold_rtx): Likewise.
12507         * fwprop.c (canonicalize_address): Likewise.
12508         * targhooks.c (default_shift_truncation_mask): Likewise.
12510 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12511             Alan Hayward  <alan.hayward@arm.com>
12512             David Sherwood  <david.sherwood@arm.com>
12514         * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12515         (widened_mode): Likewise.
12516         (expand_unop): Likewise.
12517         * ree.c (transform_ifelse): Likewise.
12518         (merge_def_and_ext): Likewise.
12519         (combine_reaching_defs): Likewise.
12520         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12522 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12523             Alan Hayward  <alan.hayward@arm.com>
12524             David Sherwood  <david.sherwood@arm.com>
12526         * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12527         * combine.c (gen_lowpart_for_combine): Likewise.
12528         * dwarf2out.c (rtl_for_decl_location): Likewise.
12529         * final.c (alter_subreg): Likewise.
12530         * rtlhooks.c (gen_lowpart_general): Likewise.
12531         (gen_lowpart_if_possible): Likewise.
12533 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12534             Alan Hayward  <alan.hayward@arm.com>
12535             David Sherwood  <david.sherwood@arm.com>
12537         * calls.c (expand_call): Use subreg_lowpart_offset.
12538         * cse.c (cse_insn): Likewise.
12539         * regcprop.c (copy_value): Likewise.
12540         (copyprop_hardreg_forward_1): Likewise.
12542 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12544         PR target/82524
12545         * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12546         *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12547         =Q constraints to +Q and into insn condition add check
12548         that operands[0] and operands[1] are equal.
12549         (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12550         =Q constraints to +Q and into insn condition add check
12551         that operands[0] is equal to either operands[1] or operands[2].
12553         PR target/82498
12554         * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12555         instead of handling MINUS_EXPR twice (once for each argument),
12556         canonicalize operand order and handle just once, use rtype where
12557         possible.  Handle (A << B) | (A >> (-B & (Z - 1))).
12559         PR target/82498
12560         * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12561         any values of __C while still being pattern recognizable as a simple
12562         rotate instruction.
12564 2017-10-13  Richard Biener  <rguenther@suse.de>
12566         PR tree-optimization/82451
12567         Revert
12568         2017-10-02  Richard Biener  <rguenther@suse.de>
12570         PR tree-optimization/82355
12571         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12572         a mapping for the enclosing loop but avoid generating one for
12573         the loop tree root.
12574         (copy_bb_and_scalar_dependences): Remove premature codegen
12575         error on PHIs in blocks duplicated into multiple places.
12576         * graphite-scop-detection.c
12577         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12578         in the region use it as loop and nest to analyze the DR in.
12579         (try_generate_gimple_bb): Likewise.
12580         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12581         (add_loop_constraints): For blocks in a loop not in the region
12582         create a dimension with a single iteration.
12583         * sese.h (gbb_loop_at_index): Remove assert.
12585         * cfgloop.c (loop_preheader_edge): For the loop tree root
12586         return the single successor of the entry block.
12587         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12588         Reset the SCEV hashtable and niters.
12589         * graphite-scop-detection.c
12590         (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12591         assert that we only have POLYNOMIAL_CHREC that vary in loops
12592         contained in the region.
12593         (scop_detection::graphite_can_represent_expr): Adjust.
12594         (scop_detection::stmt_has_simple_data_refs_p): For loops
12595         not in the region set loop to NULL.  The nest is now the
12596         entry edge to the region.
12597         (try_generate_gimple_bb): Likewise.
12598         * sese.c (scalar_evolution_in_region): Adjust for
12599         instantiate_scev change.
12600         * tree-data-ref.h (graphite_find_data_references_in_stmt):
12601         Make nest parameter the edge into the region.
12602         (create_data_ref): Likewise.
12603         * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12604         entry edge into a region and adjust instantiate_scev calls.
12605         (create_data_ref): Likewise.
12606         (graphite_find_data_references_in_stmt): Likewise.
12607         (find_data_references_in_stmt): Pass the loop preheader edge
12608         from the nest argument.
12609         * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12610         parameter the edge into the region.
12611         (instantiate_parameters): Use the loop preheader edge as entry.
12612         * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12613         NULL loop.
12614         (get_instantiated_value_entry): Make instantiate_below parameter
12615         the edge into the region.
12616         (instantiate_scev_name): Likewise.  Adjust dominance checks,
12617         when we cannot use loop-based instantiation instantiate by
12618         walking use-def chains.
12619         (instantiate_scev_poly): Adjust.
12620         (instantiate_scev_binary): Likewise.
12621         (instantiate_scev_convert): Likewise.
12622         (instantiate_scev_not): Likewise.
12623         (instantiate_array_ref): Remove.
12624         (instantiate_scev_3): Likewise.
12625         (instantiate_scev_2): Likewise.
12626         (instantiate_scev_1): Likewise.
12627         (instantiate_scev_r): Do not blindly handle N-operand trees.
12628         Do not instantiate array-refs.  Handle all constants and invariants.
12629         (instantiate_scev): Make instantiate_below parameter
12630         the edge into the region.
12631         (resolve_mixers): Use the loop preheader edge for the region
12632         parameter to instantiate_scev_r.
12633         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12635 2017-10-13  Richard Biener  <rguenther@suse.de>
12637         PR tree-optimization/82525
12638         * graphite-isl-ast-to-gimple.c
12639         (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12640         out from ...
12641         (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12642         Fail code generation when we cannot represent the isl integer.
12643         (binary_op_to_tree): Elide modulo operations that are no-ops
12644         in the type we code generate.  Remove now superfluous code
12645         generation errors.
12647 2017-10-13  Richard Biener  <rguenther@suse.de>
12649         * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12650         (scop_detection::harmful_loop_in_region): Remove premature
12651         IV type restriction.
12652         (scop_detection::graphite_can_represent_scev): We can handle
12653         pointer IVs just fine.
12655 2017-10-13  Alan Modra  <amodra@gmail.com>
12657         * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12658         "Clobbers and Scratch Registers".  Add paragraph on
12659         alternative to clobbers for scratch registers and OpenBLAS
12660         example.
12662 2017-10-13  Alan Modra  <amodra@gmail.com>
12664         * doc/extend.texi (Clobbers): Correct vax example.  Delete old
12665         example of a memory input for a string of known length.  Move
12666         commentary out of table.  Add a number of new examples
12667         covering array memory inputs.
12669 2017-10-12  Martin Liska  <mliska@suse.cz>
12671         PR tree-optimization/82493
12672         * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12673         (test_range_functions): New function.
12674         (sbitmap_c_tests): Likewise.
12675         * selftest-run-tests.c (selftest::run_tests): Run new tests.
12676         * selftest.h (sbitmap_c_tests): New function.
12678         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12680 2017-10-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
12682         * config/rs6000/amo.h: Fix spacing issue.
12684 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12686         PR target/82498
12687         * config/i386/i386.md (*ashl<mode>3_mask_1,
12688         *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12689         *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12690         patterns.
12692 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12694         * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12695         (profile_probability): Set max_probability
12696         to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12697         in temporaries.
12698         * profile-count.c (profile_probability::differs_from_p): Do not
12699         rely on max_probaiblity == 10000
12701 2017-10-12  Jeff Law  <law@redhat.com>
12703         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12704         negative offsets.
12706 2017-10-12  Martin Sebor  <msebor@redhat.com>
12708         PR other/82301
12709         PR c/82435
12710         * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12711         (handle_alias_pairs): Call it.
12712         * common.opt (-Wattribute-alias): New option.
12713         * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12714         * doc/invoke.texi (-Wattribute-alias): Document.
12716 2017-10-12  Vladimir Makarov  <vmakarov@redhat.com>
12718         Revert
12719         2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12720         PR sanitizer/82353
12721         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12722         locations.
12723         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12724         (make_hard_regno_born, make_hard_regno_dead): Update
12725         bb_killed_pseudos and bb_gen_pseudos.
12727 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12729         * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12731 2017-10-12  Uros Bizjak  <ubizjak@gmail.com>
12733         * config/alpha/alpha.c (alpha_split_conditional_move):
12734         Use std::swap instead of manually swapping.
12735         (alpha_stdarg_optimize_hook): Ditto.
12736         (alpha_canonicalize_comparison): Ditto.
12738 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12740         * tree-loop-distribution.c (struct builtin_info): New struct.
12741         (struct partition): Refactor fields into struct builtin_info.
12742         (partition_free): Free struct builtin_info.
12743         (build_size_arg_loc, build_addr_arg_loc): Delete.
12744         (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12745         information from struct builtin_info.
12746         (find_single_drs): New function refactored from classify_partition.
12747         Also moved builtin validity checks to this function.
12748         (compute_access_range, alloc_builtin): New functions.
12749         (classify_builtin_st, classify_builtin_ldst): New functions.
12750         (classify_partition): Refactor code into functions find_single_drs,
12751         classify_builtin_st and classify_builtin_ldst.
12752         (distribute_loop): Don't do runtime alias check when distributing
12753         loop nest.
12754         (find_seed_stmts_for_distribution): New function.
12755         (pass_loop_distribution::execute): Refactor code finding seed
12756         stmts into above function.  Support distribution for the innermost
12757         two-level loop nest.  Adjust dump information.
12759 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12761         * tree-loop-distribution.c: Adjust the general comment.
12762         (NUM_PARTITION_THRESHOLD): New macro.
12763         (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12764         (classify_partition): Skip builtin pattern of loop nest's inner loop.
12765         (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12766         in call to build_partition_graph.
12767         (finalize_partitions): New parameter.  Make loop distribution more
12768         conservative by fusing more partitions.
12769         (distribute_loop): Don't do runtime alias check in case of loop nest
12770         distribution.
12771         (find_seed_stmts_for_distribution): New function.
12772         (prepare_perfect_loop_nest): New function.
12773         (pass_loop_distribution::execute): Refactor code finding seed stmts
12774         and loop nest into above functions.  Support loop nest distribution.
12775         Adjust dump information accordingly.
12777 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12779         * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12780         and set PTYPE_SEQUENTIAL for merged partition.
12782 2017-10-12  Richard Biener  <rguenther@suse.de>
12784         PR tree-optimization/69728
12785         Revert
12786         2017-09-19  Richard Biener  <rguenther@suse.de>
12788         PR tree-optimization/69728
12789         * graphite-sese-to-poly.c (schedule_error): New global.
12790         (add_loop_schedule): Handle empty domain by failing the
12791         schedule.
12792         (build_original_schedule): Handle schedule_error.
12794         * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12795         domain by returning an unchanged schedule.
12797 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12799         * genrecog.c (validate_pattern): For VEC_SELECT verify that
12800         CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12802 2017-10-12  Aldy Hernandez  <aldyh@redhat.com>
12804         * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12805         Handle params.def.
12807 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12809         PR c++/82159
12810         * expr.c (store_field): Don't optimize away bitsize == 0 store
12811         from CALL_EXPR with addressable return type.
12813 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12815         * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12816         * config/rs6000/rs6000.md (sel): Delete mode attribute.
12817         (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12818         *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12819         TARGET_ISEL instead of TARGET_ISEL<sel>.
12821 2017-10-11  David Edelsohn  <dje.gcc@gmail.com>
12823         * config/rs6000/rs6000.c
12824         (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12826 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12828         * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12829         Move up in file.
12830         (reg_or_cint_operand): Fix comment.
12831         (reg_or_zero_operand): New predicate.
12832         * config/rs6000/rs6000-protos.h (output_isel): Delete.
12833         * config/rs6000/rs6000.c (output_isel): Delete.
12834         * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12835         instead of reg_or_cint_operand.  Output instruction directly (not via
12836         output_isel).
12837         (isel_unsigned_<mode>): Ditto.
12838         (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12839         gpc_reg_operand.  Add an instruction alternative for this.  Output
12840         instruction directly.
12841         (*isel_reversed_unsigned_<mode>): Ditto.
12843 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12845         * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12846         (TARGET_CANONICALIZE_COMPARISON): Define.
12848 2017-10-11  Qing Zhao  <qing.zhao@oracle.com>
12850         PR target/81422
12851         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12852         Check whether the dest is REG before adding REG_EQUIV note.
12854 2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12856         PR sanitizer/82353
12857         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12858         locations.
12859         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12860         (make_hard_regno_born, make_hard_regno_dead): Update
12861         bb_killed_pseudos and bb_gen_pseudos.
12863 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12865         * incpath.h (enum incpath_kind): Name enum, prefix values.
12866         (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12867         * incpath.c (heads, tails): Use INC_MAX.
12868         (add_env_var_paths, add_standard_paths): Use incpath_kind.
12869         (merge_include_chains, split_quote_chain,
12870         register_include_chains): Update incpath_kind names.
12871         (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12872         * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12873         names.
12874         (add_framework_path, darwin_register_objc_includes): Likewise.
12875         * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12877 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12879         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12880         Do not use float_operator operator predicate.
12881         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12882         * config/i386/predicates.md (float_operator): Remove predicate.
12884 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12886         * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12887         (*jccxf_i387): Ditto.
12888         (*jcc<mode>_i387): Ditto.
12889         (*jccu<mode>_i387): Ditto.
12890         (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12891         (*jcc_*_i387 splitters): Remove.
12892         * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12893         * config/i386/i386.c (ix86_split_fp_branch): Remove.
12894         * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12895         Remove predicate.
12897 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12899         * profile-count.h (slow_safe_scale_64bit): New function.
12900         (safe_scale_64bit): New inline.
12901         (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12902         * profile-count.c: Include wide-int.h
12903         (slow_safe_scale_64bit): New.
12905 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12907         * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12908         HAS_DECL_ASSEMBLER_NAME_P.
12909         * gimple-expr.c (gimple_decl_printable_name: Check
12910         HAS_DECL_ASSEMBLER_NAME_P too.
12911         * ipa-utils.h (type_in_anonymous_namespace_p): Check
12912         DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
12913         (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
12914         * passes.c (rest_of_decl_compilation): Check
12915         HAS_DECL_ASSEMBLER_NAME_P too.
12916         * recog.c (verify_changes): Likewise.
12917         * tree-pretty-print.c (dump_decl_name): Likewise.
12918         * tree-ssa-structalias.c (alias_get_name): Likewise.  Reimplement.
12920         * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
12921         (SET_DECL_ASSEMBLER_NAME): Use it.
12922         (DECL_ASSEMBLER_NAME_SET_P): Likewise.
12923         (COPY_DECL_ASSEMBLER_NAME): Likewise.
12924         * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
12926 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12928         * config.gcc (i386, x86_64): Add extra objects.
12929         * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
12930         (ix86_min_insn_size): Declare.
12931         (ix86_issue_rate): Declare.
12932         (ix86_adjust_cost): Declare.
12933         (ia32_multipass_dfa_lookahead): Declare.
12934         (ix86_macro_fusion_p): Declare.
12935         (ix86_macro_fusion_pair_p): Declare.
12936         (ix86_bd_has_dispatch): Declare.
12937         (ix86_bd_do_dispatch): Declare.
12938         (ix86_core2i7_init_hooks): Declare.
12939         (ix86_atom_sched_reorder): Declare.
12940         * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
12941         (COSTS_N_BYTES): Move to x86-tune-costs.h.
12942         (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
12943         (rip_relative_addr_p): Rename to ...
12944         (ix86_rip_relative_addr_p): ... this one; export.
12945         (memory_address_length): Update.
12946         (ix86_issue_rate): Move to x86-tune-sched.c.
12947         (ix86_flags_dependent): Move to x86-tune-sched.c.
12948         (ix86_agi_dependent): Move to x86-tune-sched.c.
12949         (exact_dependency_1): Move to x86-tune-sched.c.
12950         (exact_store_load_dependency): Move to x86-tune-sched.c.
12951         (ix86_adjust_cost): Move to x86-tune-sched.c.
12952         (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
12953         (ix86_macro_fusion_p): Move to x86-tune-sched.c.
12954         (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
12955         (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
12956         (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
12957         (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
12958         (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
12959         (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
12960         (min_insn_size): Rename to ...
12961         (ix86_min_insn_size): ... this one; export.
12962         (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
12963         (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
12964         (core2i7_first_cycle_multipass_backtrack): Move to
12965         x86-tune-sched-core.c.
12966         (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
12967         (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
12968         (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
12969         (ix86_avoid_jump_mispredicts): Update.
12970         (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
12971         (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
12972         (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
12973         (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
12974         (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
12975         (MAX_INSN): Move to ix86-tune-sched-bd.c.
12976         (MAX_IMM): Move to ix86-tune-sched-bd.c.
12977         (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
12978         (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
12979         (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
12980         (MAX_LOAD): Move to ix86-tune-sched-bd.c.
12981         (MAX_STORE): Move to ix86-tune-sched-bd.c.
12982         (BIG): Move to ix86-tune-sched-bd.c.
12983         (enum dispatch_group): Move to ix86-tune-sched-bd.c.
12984         (enum insn_path): Move to ix86-tune-sched-bd.c.
12985         (get_mem_group): Move to ix86-tune-sched-bd.c.
12986         (is_cmp): Move to ix86-tune-sched-bd.c.
12987         (dispatch_violation): Move to ix86-tune-sched-bd.c.
12988         (is_branch): Move to ix86-tune-sched-bd.c.
12989         (is_prefetch): Move to ix86-tune-sched-bd.c.
12990         (init_window): Move to ix86-tune-sched-bd.c.
12991         (allocate_window): Move to ix86-tune-sched-bd.c.
12992         (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
12993         (is_end_basic_block): Move to ix86-tune-sched-bd.c.
12994         (process_end_window): Move to ix86-tune-sched-bd.c.
12995         (allocate_next_window): Move to ix86-tune-sched-bd.c.
12996         (find_constant): Move to ix86-tune-sched-bd.c.
12997         (get_num_immediates): Move to ix86-tune-sched-bd.c.
12998         (has_immediate): Move to ix86-tune-sched-bd.c.
12999         (get_insn_path): Move to ix86-tune-sched-bd.c.
13000         (get_insn_group): Move to ix86-tune-sched-bd.c.
13001         (count_num_restricted): Move to ix86-tune-sched-bd.c.
13002         (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
13003         (add_insn_window): Move to ix86-tune-sched-bd.c.
13004         (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
13005         (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
13006         (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
13007         (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
13008         (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
13009         (do_dispatch): Move to ix86-tune-sched-bd.c.
13010         (has_dispatch): Move to ix86-tune-sched-bd.c.
13011         * config/i386/t-i386: Add new object files.
13012         * config/i386/x86-tune-costs.h: New file.
13013         * config/i386/x86-tune-sched-atom.c: New file.
13014         * config/i386/x86-tune-sched-bd.c: New file.
13015         * config/i386/x86-tune-sched-core.c: New file.
13016         * config/i386/x86-tune-sched.c: New file.
13018 2017-10-11  Liu Hao  <lh_mouse@126.com>
13020         * pretty-print.c [_WIN32] (colorize_init): Remove.  Use
13021         the generic version below instead.
13022         (should_colorize): Recognize Windows consoles as terminals
13023         for MinGW targets.
13024         * pretty-print.c [__MINGW32__] (write_all): New function.
13025         [__MINGW32__] (find_esc_head): Likewise.
13026         [__MINGW32__] (find_esc_terminator): Likewise.
13027         [__MINGW32__] (eat_esc_sequence): Likewise.
13028         [__MINGW32__] (mingw_ansi_fputs): New function that handles
13029         ANSI escape codes.
13030         (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
13031         for MinGW targets.
13033 2017-10-11  Richard Biener  <rguenther@suse.de>
13035         * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
13036         Properly call analyze_scalar_evolution with the loop of the stmt.
13038 2017-10-11  Richard Biener  <rguenther@suse.de>
13040         * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
13041         * tree-core.h (tree_base): Add chrec_var union member.
13042         * tree.h (CHREC_VAR): Remove.
13043         (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
13044         * tree-chrec.h (build_polynomial_chrec): Adjust.
13045         * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
13046         * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
13048 2017-10-11  Marc Glisse  <marc.glisse@inria.fr>
13050         * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
13051         * match.pd: ... here.
13052         ((T) X == (T) Y): Relax condition.
13054 2017-10-11  Bin Cheng  <bin.cheng@arm.com>
13056         PR tree-optimization/82472
13057         * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
13058         comment.
13059         (break_alias_scc_partitions): Update postorder number.
13061 2017-10-11  Martin Liska  <mliska@suse.cz>
13063         PR sanitizer/82490
13064         * opts.c (parse_no_sanitize_attribute): Do not use error_value
13065         variable.
13066         * opts.h (parse_no_sanitize_attribute): Remove last argument.
13068 2017-10-11  Martin Liska  <mliska@suse.cz>
13070         * print-rtl.c (print_insn): Move declaration of idbuf
13071         to same scope as name.
13073 2017-10-11  Martin Liska  <mliska@suse.cz>
13075         Revert r253637:
13077         PR sanitizer/82484
13078         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13079         volatile arguments.
13081 2017-10-11  Martin Liska  <mliska@suse.cz>
13083         PR sanitizer/82484
13084         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13085         volatile arguments.
13087 2017-10-11  Adhemerval Zanella  <adhemerval.zanella@linaro.org>
13089         * config.gcc (default_gnu_indirect_function): Default to yes for
13090         arm*-*-linux* with glibc.
13092 2017-10-11  Richard Biener  <rguenther@suse.de>
13094         * tree-scalar-evolution.c (get_scalar_evolution): Handle
13095         default-defs and types we do not want to analyze.
13096         (interpret_loop_phi): Replace unreachable code with an assert.
13097         (compute_scalar_evolution_in_loop): Remove and inline ...
13098         (analyze_scalar_evolution_1): ... here, replacing condition with
13099         what makes the intent clearer.  Remove handling of cases
13100         get_scalar_evolution now handles.
13102 2017-10-10  Jim Wilson  <wilson@tuliptree.org>
13104         PR rtl-optimization/81434
13105         * haifa-sched.c (prune_ready_list): Init min_cost_group to 0.  Update
13106         comment for main loop.  In sched_group_found if, also add checks for
13107         pass and min_cost_group.
13109 2017-10-10  Segher Boessenkool  <segher@kernel.crashing.org>
13111         * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
13112         (rs6000_insn_cost): New function.
13113         * config/rs6000/rs6000.md (cost): New attribute.
13115 2017-10-10  Jakub Jelinek  <jakub@redhat.com>
13116             H.J. Lu  <hongjiu.lu@intel.com>
13118         PR target/79565
13119         PR target/82483
13120         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
13121         OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
13122         __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
13123         (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
13124         to OPTION_MASK_ISA_AVX512VL - builtins that have both
13125         OPTION_MASK_ISA_MMX and some other bit set require both
13126         mmx and the ISAs without the mmx bit.
13127         * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
13128         __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
13129         __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
13130         __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
13131         __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
13132         __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
13133         __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
13134         __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
13135         __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
13136         __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
13137         __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
13138         __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
13139         __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
13140         __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
13141         Add OPTION_MASK_ISA_MMX.
13143 2017-10-10  Andreas Tobler  <andreast@gcc.gnu.org>
13145         * config.gcc (armv7*-*-freebsd*): New target.
13146         (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
13148 2017-10-10  Jan Hubicka  <hubicka@ucw.cz>
13150         * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
13151         X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
13152         spot in the file.
13154 2017-10-10  Richard Sandiford  <richard.sandiford@linaro.org>
13156         * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
13157         a template parameter.
13158         (WIDE_INT_REF_FOR): Update accordingly.
13159         * tree.h (wi::int_traits <const_tree>): Delete.
13160         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
13161         (wi::to_widest, wi::to_offset): Use them.  Expand commentary.
13162         (wi::tree_to_wide_ref): New typedef.
13163         (wi::to_wide): New function.
13164         * calls.c (get_size_range): Use wi::to_wide when operating on
13165         trees as wide_ints.
13166         * cgraph.c (cgraph_node::create_thunk): Likewise.
13167         * config/i386/i386.c (ix86_data_alignment): Likewise.
13168         (ix86_local_alignment): Likewise.
13169         * dbxout.c (stabstr_O): Likewise.
13170         * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
13171         * expr.c (const_vector_from_tree): Likewise.
13172         * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
13173         * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
13174         (fold_negate_expr_1, int_const_binop_1, const_binop)
13175         (fold_convert_const_int_from_real, optimize_bit_field_compare)
13176         (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
13177         (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
13178         (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
13179         (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
13180         (fold_not_const, round_up_loc): Likewise.
13181         * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
13182         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
13183         (alloca_call_type): Likewise.
13184         * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
13185         * godump.c (go_output_typedef): Likewise.
13186         * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
13187         * internal-fn.c (get_min_precision): Likewise.
13188         * ipa-cp.c (ipcp_store_vr_results): Likewise.
13189         * ipa-polymorphic-call.c
13190         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
13191         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
13192         (ipa_modify_call_arguments): Likewise.
13193         * match.pd: Likewise.
13194         * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
13195         * print-tree.c (print_node_brief, print_node): Likewise.
13196         * stmt.c (expand_case): Likewise.
13197         * stor-layout.c (layout_type): Likewise.
13198         * tree-affine.c (tree_to_aff_combination): Likewise.
13199         * tree-cfg.c (group_case_labels_stmt): Likewise.
13200         * tree-data-ref.c (dr_analyze_indices): Likewise.
13201         (prune_runtime_alias_test_list): Likewise.
13202         * tree-dump.c (dequeue_and_dump): Likewise.
13203         * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
13204         * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
13205         * tree-pretty-print.c (dump_generic_node): Likewise.
13206         * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
13207         (simple_iv_with_niters): Likewise.
13208         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
13209         * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
13210         * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
13211         * tree-ssa-loop-niter.c (split_to_var_and_offset)
13212         (refine_value_range_using_guard, number_of_iterations_ne_max)
13213         (number_of_iterations_lt_to_ne, number_of_iterations_lt)
13214         (get_cst_init_from_scev, record_nonwrapping_iv)
13215         (scev_var_range_cant_overflow): Likewise.
13216         * tree-ssa-phiopt.c (minmax_replacement): Likewise.
13217         * tree-ssa-pre.c (compute_avail): Likewise.
13218         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
13219         (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
13220         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
13221         * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
13222         * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
13223         * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
13224         (dump_case_nodes, try_switch_expansion): Likewise.
13225         * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
13226         (vect_do_peeling): Likewise.
13227         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13228         * tree-vect-stmts.c (vectorizable_load): Likewise.
13229         * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
13230         (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
13231         (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
13232         (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
13233         (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
13234         (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
13235         (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
13236         (evrp_dom_walker::before_dom_children): Likewise.
13237         * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13238         (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13239         (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13240         (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13241         (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13242         (get_range_pos_neg): Likewise.
13243         * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13244         * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13245         * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13246         * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13247         * config/avr/avr.c (avr_fold_builtin): Likewise.
13248         * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13249         * config/msp430/msp430.c (msp430_attr): Likewise.
13250         * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13251         * config/powerpcspe/powerpcspe-c.c
13252         (altivec_resolve_overloaded_builtin): Likewise.
13253         * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13254         (rs6000_expand_ternop_builtin): Likewise.
13255         * config/rs6000/rs6000-c.c
13256         (altivec_resolve_overloaded_builtin): Likewise.
13257         * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13258         (rs6000_expand_ternop_builtin): Likewise.
13259         * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13261 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13263         * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13264         when copying loop nest with only one inner loop.
13266 2017-10-10  Richard Biener  <rguenther@suse.de>
13268         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13269         blocks if SCEV is active.
13270         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13271         dead code.
13272         (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13273         (scev_initialize): Assert we are not yet initialized.
13275 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13277         * tree-loop-distribution.c (generate_loops_for_partition): Remove
13278         inner loop's exit stmt by making it always exit the loop, otherwise
13279         we would generate an infinite empty loop.
13281 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13283         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13284         renaming variables in new preheader if it's deleted.
13286 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13288         * tree-loop-distribution.c (struct partition): Remove unused field
13289         loops of the structure.
13290         (partition_alloc, partition_free): Ditto.
13291         (build_rdg_partition_for_vertex): Ditto.
13293 2017-10-09  Jeff Law  <law@redhat.com>
13295         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13296         return type to match prototype and documentation.
13298 2010-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13300         * config/rs6000/rs6000.c (processor_costs): Move to ...
13301         * config/rs6000/rs6000.h: ... here.
13302         (rs6000_cost): Declare.
13304 2017-10-09  Eric Botcazou  <ebotcazou@adacore.com>
13306         * except.c (setjmp_fn): New global variable.
13307         (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13308         (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13309         if DONT_USE_BUILTIN_SETJMP is defined.
13311 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13313         * target.def (insn_cost): New hook.
13314         * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13315         * doc/tm.texi: Regenerate.
13316         * rtlanal.c (insn_cost): Use the new hook.
13318 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13320         * combine.c (combine_validate_cost): Compute the new insn_cost,
13321         not just pattern_cost.
13322         (try_combine): Adjust comment.
13324 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13326         * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13327         insn_cost.
13328         * combine.c (uid_insn_cost): Adjust comment.
13329         (combine_validate_cost): Adjust comment.  Use pattern_cost instead
13330         of insn_rtx_cost
13331         (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13332         * dse.c (find_shift_sequence): Ditto.
13333         * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13334         (bb_valid_for_noce_process_p): Use pattern_cost.
13335         * rtl.h (insn_rtx_cost): Delete.
13336         (pattern_cost): New prototype.
13337         (insn_cost): New prototype.
13338         * rtlanal.c (insn_rtx_cost): Rename to...
13339         (pattern_cost): ... this.
13340         (insn_cost): New.
13342 2017-10-09  Uros Bizjak  <ubizjak@gmail.com>
13344         * config/i386/i386.md (*jcc_2): Remove insn pattern.
13345         (*jcc<mode>_0_r_i387): Ditto.
13346         (*jccxf_r_i387): Ditto.
13347         (*jcc<mode>_r_i387): Ditto.
13348         (*jccu<mode>_r_i387): Ditto.
13349         (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13350         (*jcc): Rename from *jcc_1.
13352 2017-10-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13354         * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13355         deferred rescans after the lvx/stvx recombination pre-pass.
13357 2017-10-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
13359         * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13360         memory operation instruction support.
13361         * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13362         (rs6000-ibm-aix[789]*): Likewise.
13363         * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13364         Document new functions.
13366 2017-10-09  Richard Biener  <rguenther@suse.de>
13368         PR tree-optimization/82397
13369         * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13370         equality only for semantically equal trees.
13372 2017-10-09  Richard Biener  <rguenther@suse.de>
13374         PR tree-optimization/82449
13375         * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13376         * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13377         allow constant addresses.
13378         * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13379         are linear.
13381 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13383         * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13384         flags.
13386 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13388         PR target/82463
13389         * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13390         definitions.
13392 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13394         PR target/82465
13395         * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13397 2017-10-09  Jakub Jelinek  <jakub@redhat.com>
13399         PR target/82464
13400         * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13401         s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13403 2017-10-09  Richard Sandiford  <richard.sandiford@linaro.org>
13405         * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13406         (WI_BINARY_PREDICATE_RESULT): Likewise.
13407         (wi::binary_traits::operator_result): New type.
13408         (wi::binary_traits::predicate_result): Likewise.
13409         (generic_wide_int::operator~, unary generic_wide_int::operator-)
13410         (generic_wide_int::operator==, generic_wide_int::operator!=)
13411         (generic_wide_int::operator&, generic_wide_int::and_not)
13412         (generic_wide_int::operator|, generic_wide_int::or_not)
13413         (generic_wide_int::operator^, generic_wide_int::operator+
13414         (binary generic_wide_int::operator-, generic_wide_int::operator*):
13415         Delete.
13416         (operator~, unary operator-, operator==, operator!=, operator&)
13417         (operator|, operator^, operator+, binary operator-, operator*): New
13418         functions.
13419         * expr.c (get_inner_reference): Use wi::bit_and_not.
13420         * fold-const.c (fold_binary_loc): Likewise.
13421         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13422         * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13423         (bit_value_binop): Likewise.
13424         * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13425         * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13426         (extract_range_from_binary_expr_1): Likewise.
13427         (masked_increment): Likewise.
13428         (simplify_bit_ops_using_ranges): Likewise.
13430 2017-10-09  Martin Jambor  <mjambor@suse.cz>
13432         PR hsa/82416
13433         * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13434         * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13435         (hsa_type_for_scalar_tree_type): Use it.  Always force min32int for
13436         COMPLEX types.
13437         (hsa_fixup_mov_insn_type): New function.
13438         (hsa_op_with_type::get_in_type): Use it.
13439         (hsa_build_append_simple_mov): Likewise.  Allow sub-32bit
13440         immediates in an assert.
13441         (hsa_op_with_type::extend_int_to_32bit): New method.
13442         (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13443         types.  Convert to dest type if necessary.
13444         (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13445         (reg_for_gimple_ssa): Pass false as min32int to
13446         hsa_type_for_scalar_tree_type.
13447         (gen_hsa_addr): Fixup type when creating addresable temporary.
13448         (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13449         (gen_hsa_unary_operation): Extend operands and convert to dest type if
13450         necessary.  Call hsa_fixup_mov_insn_type.
13451         (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13452         extend operands and convert to dest type if necessary.
13453         (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13454         to dest type if necessary.
13455         (set_output_in_type): Call hsa_fixup_mov_insn_type.  Just ude dest
13456         if conversion nt necessary and size matches.
13457         (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13458         to dest type if necessary.
13459         (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13460         (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13461         necessary.
13462         (gen_hsa_clrsb): Likewise.
13463         (gen_hsa_ffs): Likewise.
13464         (gen_hsa_divmod): Extend operands and convert to dest type if
13465         necessary.
13466         (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13468 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13470         * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13471         Remove empty default arguments.  Use a brace block as output
13472         statement.
13473         (conditional return): Ditto.
13474         (jump): Ditto.
13475         (indirect_jump): Ditto.  Use b%T0 instead of bctr/blr.
13476         (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13477         Ditto.
13478         (group_ending_nop): Ditto.
13479         (doloop_end): Ditto.
13480         (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13481         (splitters for those): Ditto.
13483 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13485         * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13486         a conditional jump (and the compare for it) so that pc_rtx is the
13487         last operand.
13488         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13489         for the deleted and renamed ctr<mode>_internal[234] patterns.
13490         * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13491         Delete second conditional return pattern.
13492         (ctr<mode>_internal2): Delete this second bdnz pattern.
13493         (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13494         (ctr<mode>_internal4): Delete this second bdz pattern.
13496 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13498         * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13499         (always_initialized_rtx_for_ssa_name_p): New predicate.
13500         * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13501         (finish_out_of_ssa): Free new field of SA.
13502         * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13503         * tree-ssa-coalesce.c: Include tree-ssa.h.
13504         (get_parm_default_def_partitions): Remove extern keyword.
13505         (get_undefined_value_partitions): New function.
13506         * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13507         not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13508         uninitialized bits.
13509         * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13511 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13513         * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13515 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13517         * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13518         for targets that preffer 128bit.
13520 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13522         * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13524 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13526         * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13527         on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13528         prevent DSE.
13529         (thumb_set_return_address): Likewise.
13531 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13533         * common/config/arm/arm-common.c (arm_except_unwind_info):
13534         Handle DWARF2_UNWIND_INFO.
13536 2017-10-07  Michael Collison <michael.collison@arm.com>
13538         * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13539         New pattern.
13541 2017-10-07  Eric Botcazou  <ebotcazou@adacore.com>
13543         * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13544         of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13545         * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13546         defined, force the creation of a new block for a dispatch label.
13548 2017-10-07  Jan Hubicka  <hubicka@ucw.cz>
13550         * invoke.texi (Wsuggest-attribute=cold): Document.
13551         * common.opt (Wsuggest-attribute=cold): New
13552         * ipa-pure-const.c (warn_function_cold): New function.
13553         * predict.c (compute_function_frequency): Use it.
13554         * predict.h (warn_function_cold): Declare.
13556 2017-10-06  Jan Hubicka  <hubicka@ucw.cz>
13558         * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13559         Update profile.
13561 2017-10-06  Martin Liska  <mliska@suse.cz>
13563         * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13564         keyword for member functions.
13565         (struct sanopt_tree_couple): New struct.
13566         (struct sanopt_tree_couple_hash): New function.
13567         (struct sanopt_ctx): Add new hash_map.
13568         (has_dominating_ubsan_ptr_check): New function.
13569         (record_ubsan_ptr_check_stmt): Likewise.
13570         (maybe_optimize_ubsan_ptr_ifn): Likewise.
13571         (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13572         (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13574 2017-10-06  Sudakshina Das  <sudi.das@arm.com>
13576         PR target/82440
13577         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13578         aarch64_simd_valid_immediate on CONST_VECTORs.
13579         (aarch64_reg_or_bic_imm): Likewise.
13581 2017-10-06  Wilco Dijkstra  <wdijkstr@arm.com>
13583         PR rtl-optimization/82396
13584         * haifa-sched.c (ready_sort_real): Disable qsort checking.
13586 2017-10-06  Sebastian Pop  <sebpop@gmail.com>
13588         * graphite-dependences.c (scop_get_reads): Move code to...
13589         (scop_get_must_writes): Move code to...
13590         (scop_get_may_writes): Move code to...
13591         (scop_get_reads_and_writes): ... here.
13592         (scop_get_dependences): Call scop_get_reads_and_writes.
13594 2017-10-06  Jakub Jelinek  <jakub@redhat.com>
13596         PR tree-optimization/82434
13597         * fold-const.h (can_native_encode_type_p,
13598         can_native_encode_string_p): Remove.
13599         * fold-const.c (native_encode_int): Formatting fixes.  If ptr is NULL,
13600         don't encode anything, just return what would be otherwise returned.
13601         (native_encode_fixed, native_encode_complex, native_encode_vector):
13602         Likewise.
13603         (native_encode_string): Likewise.  Inline by hand
13604         can_native_encode_string_p.
13605         (can_native_encode_type_p): Remove.
13606         (can_native_encode_string_p): Remove.
13607         * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13608         STRING_CSTs using can_native_encode_string_p, test all
13609         CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13610         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13611         argument from native_encode_expr.
13612         (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13613         (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13614         but just 2.
13616 2017-10-06  Richard Biener  <rguenther@suse.de>
13618         PR tree-optimization/82397
13619         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13620         operand_equal_p but rely on data_ref_compare_tree for detecting
13621         equalities.
13622         (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13623         to match up with dr_group_sort_cmp.
13625 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13627         PR target/82322
13628         * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13629         builtin.
13630         * config/s390/s390-builtin-types.def: Regenerate.
13632 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13634         PR target/82317
13635         * config/s390/s390-builtin-types.def: Regenerate.
13636         * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13637         Change flag from B_VXE to B_VX.
13638         (s390_vec_min_dbl): Remove B_VXE flag.
13640 2017-10-06  Richard Biener  <rguenther@suse.de>
13642         * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13643         (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13644         translate_isl_ast_to_gimple::is_valid_rename,
13645         translate_isl_ast_to_gimple::get_rename,
13646         translate_isl_ast_to_gimple::get_def_bb_for_const,
13647         translate_isl_ast_to_gimple::get_new_name,
13648         translate_isl_ast_to_gimple::collect_all_ssa_names,
13649         translate_isl_ast_to_gimple::copy_loop_phi_args,
13650         translate_isl_ast_to_gimple::collect_all_ssa_names,
13651         translate_isl_ast_to_gimple::copy_loop_phi_args,
13652         translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13653         translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13654         translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13655         translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13656         translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13657         translate_isl_ast_to_gimple::copy_cond_phi_args,
13658         translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13659         translate_isl_ast_to_gimple::edge_for_new_close_phis,
13660         translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13661         translate_isl_ast_to_gimple::rename_uses,
13662         translate_isl_ast_to_gimple::rename_all_uses): Remove.
13663         (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13664         (set_rename_for_each_def): Likewise.
13665         (graphite_copy_stmts_from_block): Handle debug stmt resetting
13666         here.  Handle rewriting SCEV analyzable uses here.
13667         (copy_bb_and_scalar_dependences): Generate code for PHI
13668         copy-in/outs.
13669         (graphite_regenerate_ast_isl): Adjust.
13670         * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13671         (add_write, add_read): New functions.
13672         (build_cross_bb_scalars_def): Use it and simplify.
13673         (build_cross_bb_scalars_use): Likewise.
13674         (graphite_find_cross_bb_scalar_vars): Inline into...
13675         (try_generate_gimple_bb): ...here.  Add dependences for PHIs,
13676         simulating out-of-SSA.  Compute liveout and add dependencies.
13677         (build_scops): Force an empty entry block.
13678         * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13679         members.
13680         (sese_build_liveouts): Declare.
13681         (sese_trivially_empty_bb_p): Likewise.
13682         * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13683         compute liveout and debug_liveout.
13684         (sese_bad_liveouts_use): Remove.
13685         (sese_reset_debug_liveouts_bb): Likewise.
13686         (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13687         (sese_build_liveouts): Build liveout and debug_liveout and store
13688         it in region.
13689         (new_sese_info): Adjust.
13690         (free_sese_info): Likewise.
13691         (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13692         do not build liveout here.
13693         (move_sese_in_condition): Adjust region entry.
13694         (scev_analyzable_p): Match up with chrec_apply requirements.
13695         (sese_trivially_empty_bb_p): New.
13696         * tree-into-ssa.c (get_reaching_def): Properly support generating
13697         default-defs for incremental rewrite of anonymous names.
13699 2017-10-06  Richard Biener  <rguenther@suse.de>
13701         * graphite-sese-to-poly.c (extract_affine): For casts increasing
13702         precision do not perform modulo reduction.
13704 2017-10-06  Richard Biener  <rguenther@suse.de>
13706         PR tree-optimization/82436
13707         * tree-vect-slp.c (vect_supported_load_permutation_p): More
13708         conservatively choose the vectorization factor when checking
13709         whether we can perform the required load permutation.
13710         (vect_transform_slp_perm_load): Assert when we may not fail.
13712 2017-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
13714         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13715         message for incompatible -msdata=* and -mcall-* options.
13717 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13719         * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13720         rate for post-reload scheduling.
13722 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13724         * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13726 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13728         * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13729         to improve monte carlo in scimark.
13731 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13733         * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13734         pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13735         athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13736         pentium4_cost, nocona_cost): Set reassociation width to 1.
13737         (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13738         width to 2 for fp operations and 1 otherwise.
13739         (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13740         for int and fp.
13741         (atom_cost): Set reassociation width to 2.
13742         (slm_cost, generic_cost): Set fp reassociation width
13743         to 2 and 1 otherwise.
13744         (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13745         (core_cost): Set fp reassociation width to 4 and vector to 2.
13746         (ix86_reassociation_width): Rewrite using cost table; special case
13747         plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13748         and TARGET_AVX128_OPTIMAL.
13749         * config/i386/i386.h (processor_costs): Add
13750         reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13751         (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13752         TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13753         * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13754         (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13755         (X86_TUNE_VECTOR_PARALLEL_EXECUTION):  Remove.
13757 2017-10-05  Nathan Sidwell  <nathan@acm.org>
13759         * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13761 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13763         * config/arm/arm.c (arm_test_fpu_data): New.
13764         (arm_run_selftests): Call arm_test_fpu_data.
13766 2017-10-04  Nathan Sidwell  <nathan@acm.org>
13768         * toplev.c (toplev::main): Remove excess parens on pretty_printer
13769         decl.
13770         * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13772 2017-10-04  Sudakshina Das  <sudi.das@arm.com>
13774         * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13775         check type for aarch64_simd_valid_immediate.
13776         (aarch64_output_simd_mov_immediate): Update prototype.
13777         (aarch64_simd_valid_immediate): Update prototype.
13778         * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13779         support for ORR-immediate.
13780         (and<mode>3): modified pattern to add support for BIC-immediate.
13781         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13782         now checks for valid immediate for BIC and ORR based on new enum
13783         argument.
13784         (aarch64_output_simd_mov_immediate): Function now used to output
13785         BIC/ORR imm as well based on new enum argument.
13786         * config/aarch64/constraints.md (Do): New vector immediate constraint.
13787         (Db) : Likewise.
13788         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13789         (aarch64_reg_or_bic_imm): Likewise.
13791 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13793         * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13794         ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13796 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13798         Revert r253399:
13800         PR rtl-optimization/82396
13801         * haifa-sched.c (autopref_multipass_init): Simplify
13802         initialization.
13803         (autopref_rank_data): Simplify sort order.
13804         * sched-int.h (autopref_multipass_data_): Remove
13805         multi_mem_insn_p, min_offset and max_offset.
13807 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13809         * doc/sourcebuild.texi: Document vect_peeling_profitable.
13811 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13813         * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13814         vect_doubleint_cvt.
13816 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13818         * doc/sourcebuild.texi: Document vect_long_mult.
13820 2017-10-04  Richard Sandiford  <richard.sandiford@linaro.org>
13822         PR tree-optimization/82413
13823         * fold-const.c (build_range_check): Use widest_int when comparing
13824         the maximum ETYPE value with HIGH.
13826 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13828         PR rtl-optimization/82396
13829         * haifa-sched.c (autopref_multipass_init): Simplify
13830         initialization.
13831         (autopref_rank_data): Simplify sort order.
13832         * sched-int.h (autopref_multipass_data_): Remove
13833         multi_mem_insn_p, min_offset and max_offset.
13835 2017-10-04  Jakub Jelinek  <jakub@redhat.com>
13837         PR tree-optimization/82381
13838         * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13839         oeN->rank first.  Return 1 or -1 if one op is SSA_NAME and the other
13840         is not.
13842         PR tree-optimization/82374
13843         * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13844         DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13845         DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13846         current_function_decl to the new decl.
13848 2017-10-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
13850         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13851         helper macro for IEEE float128 hardware built-in functions.
13852         (SQRTF128_ODD): Add built-in functions with the round-to-odd
13853         semantics.
13854         (TRUNCF128_ODD): Likewise.
13855         (ADDF128_ODD): Likewise.
13856         (SUBF128_ODD): Likewise.
13857         (MULF128_ODD): Likewise.
13858         (DIVF128_ODD): Likewise.
13859         (FMAF128_ODD): Likewise.
13860         * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13861         UNSPEC_TRUNC_ROUND_TO_ODD.
13862         (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13863         (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13864         floating point round to odd instructions.
13865         (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13866         (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13867         (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13868         (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13869         (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13870         (trunc<mode>sf2_hw): Change the truncate with round to odd
13871         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13872         (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13873         to odd hardware instructions.
13874         (sub<mode>3_odd): Likewise.
13875         (mul<mode>3_odd): Likewise.
13876         (div<mode>3_odd): Likewise.
13877         (sqrt<mode>2_odd): Likewise.
13878         (fma<mode>4_odd): Likewise.
13879         (fms<mode>4_odd): Likewise.
13880         (nfma<mode>4_odd): Likewise.
13881         (nfms<mode>4_odd): Likewise.
13882         (trunc<mode>df2_odd): Change the truncate with round to odd
13883         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.  Add a generator
13884         function.
13885         * doc/extend.texi (PowerPC built-in functions): Update documentation
13886         for existing IEEE float128-bit built-in functions.  Add built-in
13887         functions that generate the IEEE 128-bit floating point round to
13888         odd instructions.
13890 2017-10-03  Segher Boessenkool  <segher@kernel.crashing.org>
13892         PR rtl-optimization/77729
13893         * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13894         to (X&(C1&~C2))|C2 transformations.
13896 2017-10-03  Martin Jambor  <mjambor@suse.cz>
13898         PR tree-optimization/82363
13899         * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13900         mismatch, mark lacc written regardless of racc.
13902 2017-10-03  Jakub Jelinek  <jakub@redhat.com>
13904         PR tree-optimization/82381
13905         * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13906         stmt_to_insert nor wheather SSA_NAMEs are default defs.
13907         Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13908         fallthrough into reassoc_stmt_dominates_stmt_p.
13910         PR target/82386
13911         * combine.c (combine_instructions): Don't combine in unreachable
13912         basic blocks.
13914 2017-10-02  Peter Bergner  <bergner@vnet.ibm.com>
13916         PR target/80210
13917         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
13918         function to not use the have_cpu variable.  Do not set cpu_index,
13919         rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
13920         or the default cpu.
13921         (rs6000_valid_attribute_p): Remove duplicate initializations of
13922         old_optimize and func_optimize.
13923         (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
13924         (rs6000_activate_target_options): Make global.
13925         * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
13926         prototype.
13928 2017-10-02  Jakub Jelinek  <jakub@redhat.com>
13930         * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
13931         if *poffset + *pmax_size overflows in HOST_WIDE_INT.
13932         Set *poffset to 0 and *psize and *pmax_size to -1 if
13933         *poffset + *psize overflows in HOST_WIDE_INT.
13935         PR tree-optimization/82387
13936         PR tree-optimization/82388
13937         PR tree-optimization/82389
13938         * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
13939         instead of live_bytes non-NULL.
13941 2017-10-02  Georg-Johann Lay  <avr@gjlay.de>
13943         PR target/41076
13944         * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
13945         alternative.
13947 2017-10-02  Richard Biener  <rguenther@suse.de>
13949         * graphite-isl-ast-to-gimple.c (set_codegen_error): With
13950         -fchecking and --param graphite-allow-codegen-errors=0 ICE.
13951         * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
13953 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13955         * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
13956         requested precision matches the type's.
13957         * calls.c (alloc_max_size): Calculate the new candidate size as
13958         a widest_int and use wi::to_widest when comparing it with the
13959         current candidate size.
13960         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
13961         zero rather than integer_zero_node.
13962         * match.pd: Check for a no-op conversion before using wi::add
13963         rather than after.  Use tree_to_uhwi when summing small shift
13964         counts into an unsigned int.
13966 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13967             Alan Hayward  <alan.hayward@arm.com>
13968             David Sherwood  <david.sherwood@arm.com>
13970         PR target/71307
13971         * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
13972         (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
13973         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13974         POINTER_AND_FP_REGS.
13976 2017-10-02  Richard Biener  <rguenther@suse.de>
13978         PR tree-optimization/82355
13979         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
13980         a mapping for the enclosing loop but avoid generating one for
13981         the loop tree root.
13982         (copy_bb_and_scalar_dependences): Remove premature codegen
13983         error on PHIs in blocks duplicated into multiple places.
13984         * graphite-scop-detection.c
13985         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
13986         in the region use it as loop and nest to analyze the DR in.
13987         (try_generate_gimple_bb): Likewise.
13988         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
13989         (add_loop_constraints): For blocks in a loop not in the region
13990         create a dimension with a single iteration.
13991         * sese.h (gbb_loop_at_index): Remove assert.
13993 2017-10-01  Kevin Buettner  <kevinb@redhat.com>
13995         * omp-expand.c (adjust_context_scope): New function.
13996         (expand_parallel_call): Call adjust_context_scope.
13998 2017-10-01  Jeff Law  <law@redhat.com>
14000         * tree-ssa-dom.c (optimize_stmt): Make this a method within the
14001         dom_opt_dom_walker class with direct access to private members.
14002         Add comments.  Call test_for_singularity.
14003         (dom_opt_dom_walker::before_dom_children): Corresponding changes.
14004         (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
14005         m_dummy_cond anymore.
14006         (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
14007         class ctor.
14008         (pass_dominator:execute): Build the dummy_cond here and pass it
14009         to the dom_opt_dom_walker ctor.
14010         (test_for_singularity): New function.
14012 2017-09-30  Krister Walfridsson  <krister.walfridsson@gmail.com>
14013             Maya Rashish  <coypu@sdf.org>
14015         * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
14016         netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
14017         (alpha*-*-netbsd*) Use nbsd_tm_file.
14018         (arm*-*-netbsdelf*) Likewise.
14019         (i[34567]86-*-netbsdelf*) Likewise.
14020         (x86_64-*-netbsd*) Likewise.
14021         (mips*-*-netbsd*) Likewise.
14022         (powerpc-*-netbsd*) Likewise.
14023         (sh*-*-netbsd*) Likewise.
14024         (sparc-*-netbsdelf*) Likewise.
14025         (sparc64-*-netbsd*) Likewise.
14026         (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
14027         to tm_defines.
14028         (vax-*-netbsdelf*) Likewise.
14029         * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
14030         (UINT_FAST8_TYPE) Likewise.
14031         (INT_FAST16_TYPE) Check CHAR_FAST16.
14032         (UINT_FAST16_TYPE) Likewise.
14034 2017-09-30  Jakub Jelinek  <jakub@redhat.com>
14036         PR target/82361
14037         * config/i386/i386.md
14038         (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
14039         (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
14040         *divmodsi4_zext_2): New define_insn_and_split.
14041         (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
14042         (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
14043         (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
14044         *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
14045         New define_insn_and_split.
14046         (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
14047         * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
14048         operands[1] having DImode when mode is SImode.
14050         * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
14051         always SImode for DIV and MOD in REG_EQUAL notes.
14053 2017-09-29  Yury Gribov  <tetra2005@gmail.com>
14055         PR middle-end/82319
14056         * match.pd: Fix handling of NaNs in pattern.
14058 2017-09-29  Jeff Law  <law@redhat.com>
14060         * sbitmap.c (bitmap_bit_in_range_p): New function.
14061         * sbitmap.h (bitmap_bit_in_range_p): Prototype.
14062         * tree-ssa-dse.c (live_bytes_read): New function.
14063         (dse_classify_store): Ignore reads of dead bytes.
14065         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
14066         typos and whitespace errors.
14067         * config/i386/predicates.md (address_no_seg_operand): Likewise.
14068         * config/s390/s390.c (s390_emit_prologue): Likewise.
14070 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14072         PR target/81481
14073         * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
14074         with a symbol for LRA.
14076 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14078         PR rtl-optimization/82338
14079         * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
14081 2017-09-29  Alexander Monakov  <amonakov@ispras.ru>
14083         * genmodes.c (calc_wider_mode): Suppress qsort macro.
14084         * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
14085         (qsort_chk): Declare.
14086         * vec.c [CHECKING_P] (qsort_chk_error): New static function.
14087         (qsort_chk): New function.
14089 2017-09-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14091         PR tree-optimization/82337
14092         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
14093         phi definition if the PHI result appears in an abnormal PHI.
14094         (find_basis_for_base_expr): Don't record a basis if the LHS of the
14095         basis appears in an abnormal PHI.
14097 2017-09-29  Richard Biener  <rguenther@suse.de>
14099         * graphite-isl-ast-to-gimple.c
14100         (translate_isl_ast_to_gimple::set_codegen_error): New function.
14101         (binary_op_to_tree): Use it.
14102         (get_rename_from_scev): Likewise.
14103         (copy_loop_phi_nodes): Likewise.
14104         (copy_bb_and_scalar_dependences): Likewise.
14105         (translate_pending_phi_nodes): Likewise.
14107 2017-09-29  Jakub Jelinek  <jakub@redhat.com>
14109         PR target/82339
14110         * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
14111         for movabsq $(i32 << shift), r64.
14113 2017-09-28  Uros Bizjak  <ubizjak@gmail.com>
14115         * config/i386/i386.c (ix86_print_operand_address_as): Do not check
14116         index when encoding %esp as %rsp to avoid 0x67 prefix.
14118 2017-09-28  Sergey Shalnov  <Sergey.Shalnov@intel.com>
14120         * config/i386/i386.md (*movsf_internal, *movdf_internal):
14121         Return 256-bit AVX modes for TARGET_PREFER_AVX256.
14123 2017-09-28  Thomas Preud'homme  <thomas.preudhomme@arm.com>
14125         * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
14126         Extensions with more than 16 double VFP registers.
14127         (cmse_nonsecure_entry_clear_before_return): Remove second entry of
14128         to_clear_mask and all code related to it.  Replace the remaining
14129         entry by a sbitmap and adapt code accordingly.
14131 2017-09-28  Henry Linjamäki  <henry.linjamaki@parmance.com>
14133         * brig-builtins.def: Change pure attributes to const.
14135 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14137         * config.gcc (default_gnu_indirect_function): Default to yes for
14138         sparc*-*-linux* with glibc.
14140 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14142         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
14143         (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
14144         when creating .init_array and .fini_array sections with priority
14145         specified.
14147 2017-09-27  Christophe Lyon  <christophe.lyon@linaro.org>
14149         PR target/71727
14150         * config/aarch64/aarch64.c
14151         (aarch64_builtin_support_vector_misalignment): Always return false
14152         when misalignment is unknown.
14154 2017-09-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14156         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
14157         this function to return false if the definition used by the swap
14158         instruction is artificial, or if the memory address from which the
14159         constant value is loaded is not represented by a base address held
14160         in a register or if the base address register is a frame or stack
14161         pointer.  Additionally, return false if the base address of the
14162         loaded constant is a SYMBOL_REF but is not considered to be a
14163         constant.
14164         (replace_swapped_load_constant): New function.
14165         (rs6000_analyze_swaps): Add a new pass to replace a swap of a
14166         loaded constant vector with a load of a swapped constant vector.
14168 2017-09-27  Carl Love  <cel@us.ibm.com>
14170         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14171         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14172         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14173         fctiw instruction.
14175 2017-09-27  Alexander Monakov  <amonakov@ispras.ru>
14177         * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
14178         first, always call autopref_rank_data otherwise.
14180 2017-09-27  Richard Biener  <rguenther@suse.de>
14182         * graphite-scop-detection.c (find_scop_parameters): Move
14183         loop bound handling ...
14184         (gather_bbs::before_dom_children): ... here, avoiding the need
14185         to build scop_info->loop_nest.
14186         (record_loop_in_sese): Remove.
14187         * sese.h (sese_info_t::loop_nest): Remove.
14188         * sese.c (new_sese_info): Do not allocate loop_nest.
14189         (free_sese_info): Do not free loop_nest.
14191 2017-09-27  Jakub Jelinek  <jakub@redhat.com>
14193         PR c++/82159
14194         * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
14195         lhs from calls if the lhs has addressable type.
14197 2017-09-27  Richard Biener  <rguenther@suse.de>
14199         * graphite.h (scop::max_alias_set): New member.
14200         * graphite-scop-detection.c: Remove references to non-existing
14201         --param in comments.
14202         (build_alias_sets): Record the maximum alias set used for drs.
14203         (build_scops): Support zero as unlimited for
14204         --param graphite-max-arrays-per-scop.
14205         * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
14206         and inline into ...
14207         (build_poly_sr_1): ... here.  Compute alias set based on the
14208         maximum alias set used for drs rather than
14209         PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
14211 2017-09-27  Richard Biener  <rguenther@suse.de>
14213         * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
14214         --param loop-block-tile-size=0 to disable tiling.
14216 2017-09-27  Richard Biener  <rguenther@suse.de>
14218         * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
14219         (graphite-max-nb-scop-params): Document special value zero.
14220         * domwalk.h (dom_walker::STOP): New symbolical constant.
14221         (dom_walker::dom_walker): Add optional parameter for bb to
14222         RPO mapping.
14223         (dom_walker::~dom_walker): Declare.
14224         (dom_walker::before_dom_children): Document STOP return value.
14225         (dom_walker::m_user_bb_to_rpo): New member.
14226         (dom_walker::m_bb_to_rpo): Likewise.
14227         * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
14228         mapping here if not provided by the user.
14229         (dom_walker::~dom_walker): Free bb to RPO mapping if not
14230         provided by the user.
14231         (dom_walker::STOP): Define.
14232         (dom_walker::walk): Do not compute bb to RPO mapping here.
14233         Support STOP return value from before_dom_children to stop
14234         walking.
14235         * graphite-optimize-isl.c (optimize_isl): If the schedule
14236         is the same still generate code if -fgraphite-identity
14237         or -floop-parallelize-all are given.
14238         * graphite-scop-detection.c: Include cfganal.h.
14239         (gather_bbs::gather_bbs): Get and pass through bb to RPO
14240         mapping.
14241         (gather_bbs::before_dom_children): Return STOP for BBs
14242         not in the region.
14243         (build_scops): Compute bb to RPO mapping and pass it to
14244         the domwalk.  Treat --param graphite-max-nb-scop-params=0
14245         as not limiting the number of params.
14246         * graphite.c (graphite_initialize): Remove limit on the
14247         number of basic-blocks in a function.
14248         * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14249         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14250         default value of 10.
14252 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14254         * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14255         Adjust code to eliminate needing to do the shift right 32-bits
14256         operation after XSCVDPSPN.
14258 2017-09-26  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14260         * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14261         ((X / Y) != 0 -> X >= Y): Likewise.
14263 2017-09-26  Carl Love  <cel@us.ibm.com>
14265         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14266         P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14267         vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14268         void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14269         * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14270         * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14271         definitions and overloading.
14272         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14273         statement for P9V_BUILTIN_XST_LEN_R.
14274         (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14275         * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14276         define_expand and define_insn for the instructions and builtins.
14277         * doc/extend.texi: Update the built-in documentation file for the new
14278         built-in functions.
14279         * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14280         define_insn for the instructions
14282 2017-09-26  Krister Walfridsson  <krister.walfridsson@gmail.com>
14284         PR target/39570
14285         * gcc/config/netbsd-protos.h: New file.
14286         * gcc/config/netbsd.c: New file.
14287         * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14288         * gcc/config/t-netbsd: New file.
14289         * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14290         (tmake_file) Add t-netbsd.
14291         (extra_objs) Add netbsd.o.
14293 2017-09-26  Janus Weil  <janus@gcc.gnu.org>
14295         PR fortran/82143
14296         PR fortran/82324
14297         * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14299 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14301         * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14302         sign extension from a vector register to a GPR by doing a 32-bit
14303         direct move and then an EXTSW.
14304         (extendsi<mode>2 splitter): Likewise.
14305         (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14306         right or vector extract after doing XSCVDPSPN.  Use
14307         zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14308         the GPRs.
14309         (movdi_from_sf_zero_ext): Likewise.
14310         (reload_gpr_from_vsxsf): Likewise.
14311         (p8_mfvsrd_4_disf): Delete, no longer used.
14312         (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14313         then needing to move the SFmode to a GPR to use the XSCVDPSP
14314         instruction instead of FRSP and XSCVDPSPN.
14315         * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14316         it is adjacent to the other XSCVSPDP insns.
14317         (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14318         SFmode to be in traditional Altivec registers.
14319         (vsx_xscvdpspn): Eliminate useless alternative constraint.
14320         (vsx_xscvspdpn): Likewise.
14321         (vsx_xscvspdpn_scalar): Likewise.
14323 2017-09-26  Martin Jambor  <mjambor@suse.cz>
14325         * tree-sra.c (compare_access_positions): Put integral types first,
14326         stabilize sorting of integral types, remove conditions putting
14327         non-full-precision integers last.
14328         (sort_and_splice_var_accesses): Disable scalarization if a
14329         non-integert would be represented by a non-full-precision integer.
14331 2017-09-26  Joseph Myers  <joseph@codesourcery.com>
14333         * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14334         * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14335         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14336         * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14337         Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14338         conditionals inside the function instead of around it.  Call
14339         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14340         (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14342 2017-09-26  Richard Biener  <rguenther@suse.de>
14344         * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14345         fold in ...
14346         (scop_detection::build_scop_breadth): ... this.  Removed.
14347         (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14348         (scop_detection::harmful_stmt_in_bb): Likewise.
14349         (scop_detection::graphite_can_represent_stmt): Likewise.
14350         (scop_detection::loop_body_is_valid_scop): Likewise.  Remove recursion.
14351         (scop_detection::can_represent_loop): Remove recursion, fold in ...
14352         (scop_detection::can_represent_loop_1): ... this.  Removed.
14353         (scop_detection::harmful_loop_in_region): Simplify after inlining
14354         the above and remove more quadraticness.
14355         (build_scops): Adjust.
14356         * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14357         quadraticness.
14359 2017-09-26  Jakub Jelinek  <jakub@redhat.com>
14361         PR target/82267
14362         * config/i386/i386.c (ix86_print_operand_address_as): Only test
14363         REGNO (base) == SP_REG if base is a REG.
14365         PR middle-end/35691
14366         * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14367         if it is different SSA_NAME.
14368         (optimize_range_tests_cmp_bitwise): New function.
14369         (optimize_range_tests): Call it.
14371 2017-09-26  Richard Biener  <rguenther@suse.de>
14373         PR tree-optimization/82321
14374         * graphite.c (canonicalize_loop_closed_ssa): Properly check
14375         for the def being inside the loop.
14377 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14379         * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14380         assembler output.
14381         * config/s390/s390-builtins.def: Fix constraint on op4.
14383 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14385         * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14386         independent expanders.
14387         * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14388         ("vec_ordered", "vec_unordered"): New expanders.
14390 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14392         * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14393         for SFmode.
14395 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14397         * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14398         vec_unpacks_lo_v16qi.
14399         ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14401 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14403         * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14404         ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14405         ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14407 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14409         * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14410         predicate.
14411         * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14412         and V16QI.
14413         ("*vec_slb<mode>"): New insn pattern.
14414         ("vec_shr_<mode>"): New expander.
14415         * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14416         and force the shift count operand to V16QImode.
14417         ("vec_srb<mode>"): Set shift count mode to V16QI.
14419 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14421         * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14422         ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14423         ("vec_widen_smult_hi_<mode>"): New expander definitions.
14425 2017-09-26  Richard Earnshaw  <rearnsha@arm.com>
14427         PR target/82175
14428         * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14430 2017-09-26  Richard Biener  <rguenther@suse.de>
14432         PR tree-optimization/82320
14433         * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14434         isn't a change.
14436 2017-09-25  Jeff Law  <law@redhat.com>
14438         * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14439         prototype for new argument.
14440         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14441         mostly extracted from rs6000_emit_allocate_stack.
14442         (rs6000_emit_probe_stack_range_stack_clash): New function.
14443         (rs6000_emit_allocate_stack): Call
14444         rs6000_emit_probe_stack_range_stack_clash as needed.
14445         (rs6000_emit_probe_stack_range): Add additional argument
14446         to call to gen_probe_stack_range{si,di}.
14447         (output_probe_stack_range): New.
14448         (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14449         (output_probe_stack_range_stack_clash): New.
14450         (rs6000_emit_prologue): Emit notes into dump file as requested.
14451         * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14452         (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14453         Add additional operand and pass it to output_probe_stack_range.
14455 2017-09-25  Bin Cheng  <bin.cheng@arm.com>
14457         PR tree-optimization/82163
14458         * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14459         (checking_verify_loop_closed_ssa): New parameter.
14460         * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14461         (check_loop_closed_ssa_stmt): Delete.
14462         (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14463         (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14464         (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14465         changed loops.
14467 2017-09-25  Pekka Jaaskelainen <pekka@parmance.com>
14469         * brig-builtins.def: Treat HSAIL barrier builtins as
14470         setjmp/longjump style functions.
14472 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14474         * target.def (constant_alignment): New hook.
14475         * defaults.h (CONSTANT_ALIGNMENT): Delete.
14476         * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14477         (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14478         * doc/tm.texi: Regenerate.
14479         * targhooks.h (default_constant_alignment): Declare.
14480         (constant_alignment_word_strings): Likewise.
14481         * targhooks.c (default_constant_alignment): New function.
14482         (constant_alignment_word_strings): Likewise.
14483         * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14484         instead of CONSTANT_ALIGNMENT.
14485         * varasm.c (align_variable, get_variable_align, build_constant_desc)
14486         (force_const_mem): Likewise.
14487         * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14488         * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14489         (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14490         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14491         * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14492         definition.
14493         * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14494         * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14495         constant_alignment_word_strings.
14496         * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14497         (CONSTANT_ALIGNMENT): Likewise.
14498         * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14499         (arm_constant_alignment): New function.
14500         * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14501         * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14502         constant_alignment_word_strings.
14503         * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14504         * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14505         constant_alignment_word_strings.
14506         * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14507         * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14508         (cris_constant_alignment): New function.
14509         * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14510         * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14511         (epiphany_constant_alignment): New function.
14512         * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14513         * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14514         constant_alignment_word_strings.
14515         * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14516         * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14517         constant_alignment_word_strings.
14518         * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14519         * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14520         constant_alignment_word_strings.
14521         * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14522         * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14523         * config/i386/i386.c (ix86_constant_alignment): Make static.
14524         Use the same interface as the target hook.
14525         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14526         * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14527         * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14528         constant_alignment_word_strings.
14529         * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14530         * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14531         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14532         * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14533         * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14534         constant_alignment_word_strings.
14535         * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14536         * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14537         constant_alignment_word_strings.
14538         * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14539         * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14540         constant_alignment_word_strings.
14541         * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14542         * config/microblaze/microblaze.c (microblaze_constant_alignment):
14543         New function.
14544         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14545         * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14546         * config/mips/mips.c (mips_constant_alignment): New function.
14547         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14548         * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14549         * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14550         * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14551         (mmix_constant_alignment): Make static.  Use the same interface
14552         as the target hook.
14553         * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14554         * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14555         constant_alignment_word_strings.
14556         * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14557         * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14558         constant_alignment_word_strings.
14559         * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14560         * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14561         constant_alignment_word_strings.
14562         * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14563         * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14564         (rs6000_constant_alignment): New function.
14565         * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14566         * config/riscv/riscv.c (riscv_constant_alignment): New function.
14567         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14568         * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14569         * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14570         (rs6000_constant_alignment): New function.
14571         * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14572         * config/s390/s390.c (s390_constant_alignment): New function.
14573         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14574         * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14575         * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14576         constant_alignment_word_strings.
14577         * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14578         * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14579         (sparc_constant_alignment): New function.
14580         * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14581         * config/spu/spu.c (spu_constant_alignment): New function.
14582         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14583         * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14584         * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14585         constant_alignment_word_strings.
14586         * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14587         * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14588         constant_alignment_word_strings.
14589         * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14590         * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14591         constant_alignment_word_strings.
14592         * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14593         * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14594         (visium_constant_alignment): New function.
14595         * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14596         * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14597         (xtensa_constant_alignment): New function.
14598         * system.h (CONSTANT_ALIGNMENT): Poison.
14600 2017-09-25  Will Schmidt  <will_schmidt@vnet.ibm.com>
14602         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14603         for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14604         (rs6000_builtin_valid_without_lhs): New helper function.
14605         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14606         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14608 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14610         * target.h (vec_perm_indices): Use unsigned short rather than
14611         unsigned char.
14612         (auto_vec_perm_indices): Likewise.
14613         * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14614         Use unsigned int rather than unsigned char.
14615         * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14617 2017-09-25  Richard Biener  <rguenther@suse.de>
14619         * cfgloop.h (sort_sibling_loops): Declare.
14620         * cfgloop.c (sort_sibling_loops_cmp): New helper.
14621         (sort_sibling_loops): New function sorting the sibling loop list
14622         in RPO order.
14623         * graphite.c (graphite_transform_loops): Sort sibling loops.
14625 2017-09-25  Richard Sandiford  <richard.sandifird@linaro.org>
14627         * target.def (vec_perm_const_ok): Change sel parameter to
14628         vec_perm_indices.
14629         * optabs-query.c (can_vec_perm_p): Update accordingly.
14630         * doc/tm.texi: Regenerate.
14631         * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14632         auto_vec_perm_indices and remove separate nelt field.
14633         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14634         (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14635         (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14636         (aarch64_expand_vec_perm_const): Update accordingly.
14637         (aarch64_vectorize_vec_perm_const_ok): Likewise.  Change sel
14638         to vec_perm_indices.
14639         * config/arm/arm.c (expand_vec_perm_d): Change perm to
14640         auto_vec_perm_indices and remove separate nelt field.
14641         (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14642         (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14643         (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14644         accordingly.
14645         (arm_vectorize_vec_perm_const_ok): Likewise.  Change sel
14646         to vec_perm_indices.
14647         * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14648         sel to vec_perm_indices.
14649         * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14650         * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14651         * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14652         Likewise.
14653         * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14654         Likewise.
14656 2017-09-25  Pierre-Marie de Rodat  <derodat@adacore.com>
14658         PR debug/82155
14659         * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14660         on the FUNCTION_DECL function context if it has a DIE that is a
14661         declaration.
14663 2017-09-25  Richard Biener  <rguenther@suse.de>
14665         PR tree-optimization/82285
14666         * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14667         enumeral types.
14669 2017-09-25  Tom de Vries  <tom@codesourcery.com>
14671         PR target/80035
14672         PR target/81069
14673         * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14674         noreturn function.
14676 2017-09-25  Richard Biener  <rguenther@suse.de>
14678         * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14679         ISL errors other than isl_error_quota happen.  Dump if the
14680         schedule is the same.
14681         * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14682         errors instead of aborting inside ISL.
14684 2017-09-25  Iain Sandoe  <iain@codesourcery.com>
14686         PR target/80556
14687         * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14688         of libgcc_eh for m64.
14689         * config/i386/darwin64.h: Likewise.
14691 2017-09-25  Richard Biener  <rguenther@suse.de>
14693         PR middle-end/82144
14694         * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14695         attribute for incomplete types nor twice for complete ones.
14697 2017-09-24  Uros Bizjak  <ubizjak@gmail.com>
14699         PR target/82267
14700         * config/i386/i386.c (ix86_print_operand_address_as): Encode
14701         %esp as %rsp to avoid 0x67 prefix if there is no index or base
14702         register.
14704 2017-09-23  Uros Bizjak  <ubizjak@gmail.com>
14706         PR bootstrap/82306
14707         * config/i386/i386.opt (mprefer-avx256): Use
14708         ix86_target_flags variable.
14709         * config/i386/i386.c (ix86_target_string): Move
14710         -mprefer-avx256 to flag2_opts.
14712 2017-09-22  Jakub Jelinek  <jakub@redhat.com>
14714         PR middle-end/35691
14715         * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14716         and x != -1 | y != -1 into (x & y) != -1.
14718 2017-09-22  Steve Ellcey  <sellcey@cavium.com>
14720         * config.gcc: Add new case statement to set
14721         default_gnu_indirect_function.  Remove it from x86_64-*-linux*,
14722         i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14723         s390x-*-linux* case statements.   Added aarch64 to the list of
14724         supported architectures.
14726 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14728         PR tree-optimization/82289
14729         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14730         STMT_VINFO_RELEVANT_P.
14732 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14733             Alan Hayward  <alan.hayward@arm.com>
14734             David Sherwood  <david.sherwood@arm.com>
14736         * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14737         for VR_RANGE only; don't allow VR_ANTI_RANGE.
14738         (extract_range_from_binary_expr_1): Don't call
14739         extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14741 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14742             Alan Hayward  <alan.hayward@arm.com>
14743             David Sherwood  <david.sherwood@arm.com>
14745         * target.def (preferred_vector_alignment): New hook.
14746         * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14747         hook.
14748         * doc/tm.texi: Regenerate.
14749         * targhooks.h (default_preferred_vector_alignment): Declare.
14750         * targhooks.c (default_preferred_vector_alignment): New function.
14751         * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14752         Expand commentary.
14753         (DR_TARGET_ALIGNMENT): New macro.
14754         (aligned_access_p): Update commentary.
14755         (vect_known_alignment_in_bytes): New function.
14756         * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14757         function.
14758         (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14759         Calculate the misalignment based on the target alignment rather than
14760         the vector size.
14761         (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14762         rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14763         (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14764         the target alignment, rather than masking the element misalignment
14765         with the number of elements in a vector.  Also use the target
14766         alignment when calculating the maximum number of peels.
14767         (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14768         instead of TYPE_ALIGN_UNIT.
14769         (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14770         Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14771         (vect_create_addr_base_for_vector_ref): Update call accordingly.
14772         (vect_create_data_ref_ptr): Likewise.
14773         (vect_setup_realignment): Realign by ANDing with
14774         -DR_TARGET_MISALIGNMENT.
14775         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14776         the number of peels based on DR_TARGET_ALIGNMENT.
14777         * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14778         with the guaranteed alignment boundary when deciding whether
14779         overrun is OK.
14780         (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14781         relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14782         (ensure_base_align): Remove stmt_info parameter.  Get the
14783         target base alignment from DR_TARGET_ALIGNMENT.
14784         (vectorizable_store): Update call accordingly.   Interpret
14785         DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14786         TYPE_ALIGN_UNIT.
14787         (vectorizable_load): Likewise.
14789 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14790             Alan Hayward  <alan.hayward@arm.com>
14791             David Sherwood  <david.sherwood@arm.com>
14793         * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14794         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14795         (vect_enhance_data_refs_alignment): Likewise.
14797 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14799         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14800         error.  Only quit immediately if parsing is complete.
14801         (BEGIN): Initialize fatal_err and parse_done.
14802         (begin fpu, end fpu): Check number of arguments.
14803         (begin arch, end arch): Likewise.
14804         (begin cpu, end cpu): Likewise.
14805         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14806         (optalias): Likewise.
14808 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14810         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14811         * config/arm/arm-isa.h: Delete.  Move definitions to ...
14812         * arm-cpus.in: ... here.  Use new feature and fgroup values.
14813         * config/arm/arm.c (arm_option_override): Use lower case for feature
14814         bit names.
14815         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14816         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14817         * config/arm/parsecpu.awk (END): Add new command 'isa'.
14818         (isa_pfx): Delete.
14819         (print_isa_bits_for): New function.
14820         (gen_isa): New function.
14821         (gen_comm_data): Use print_isa_bits_for.
14822         (define feature): New keyword.
14823         (define fgroup): New keyword.
14824         * config/arm/t-arm (TM_H): Remove.
14825         (GTM_H): Add arm-isa.h.
14826         (arm-isa.h): Add rule to generate file.
14827         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14828         case for feature bit names.
14830 2017-09-22  Richard Biener  <rguenther@suse.de>
14832         * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14833         single caller.
14834         (graphite_regenerate_ast_isl): Do not reset SCEV.  Move debug
14835         print of no dependency loops ...
14836         * graphite.c (graphite_transform_loops): ... here.
14837         (canonicalize_loop_closed_ssa_form): Work from inner to outer
14838         loops.
14839         (same_close_phi_node, remove_duplicate_close_phi,
14840         make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14841         (canonicalize_loop_closed_ssa): ... here and simplify.
14842         * graphite-optimize-isl.c: Include tree-vectorizer.h.
14843         (optimize_isl): Use dump_printf_loc to tell when we stopped
14844         optimizing because of an ISL timeout.
14846 2017-09-22  Richard Biener  <rguenther@suse.de>
14848         PR tree-optimization/82291
14849         * tree-if-conv.c (predicate_mem_writes): Make sure to
14850         remove writes in blocks predicated with false.
14852 2017-09-22  Richard Biener  <rguenther@suse.de>
14854         * sese.c: Include cfganal.h.
14855         (if_region_set_false_region): Remove.
14856         (create_if_region_on_edge): Likewise.
14857         (move_sese_in_condition): Re-implement without destroying
14858         dominators.
14860 2017-09-22  Richard Biener  <rguenther@suse.de>
14862         * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14863         Verify both BBs contain loop PHI nodes before dispatching to
14864         copy_loop_phi_args.
14865         (graphite_regenerate_ast_isl): Do not recompute dominators,
14866         do not verify three times.  Restructure for clarity.
14867         * graphite-scop-detection.c (same_close_phi_node,
14868         remove_duplicate_close_phi, make_close_phi_nodes_unique,
14869         defined_in_loop_p, canonicalize_loop_closed_ssa,
14870         canonicalize_loop_closed_ssa_form): Simplify, remove excess
14871         checking and SSA rewrite, move to ...
14872         * graphite.c: ... here.  Include ssa.h and tree-ssa-loop-manip.h.
14873         (graphite_initialize): Do not pass in ctx, do not reset the
14874         SCEV cache, compute only dominators.
14875         (graphite_transform_loops): Allocate ISL ctx after
14876         graphite_initialize.  Call canonicalize_loop_closed_ssa_form.
14877         Maintain post-dominators only around build_scops.
14878         * sese.c (if_region_set_false_region): Make static.  Free
14879         and recompute dominators.
14880         (move_sese_in_condition): Assert we don't get called with
14881         post-dominators computed.
14882         * sese.h (if_region_set_false_region): Remove.
14884 2017-09-22  Sergey Shalnov  <sergey.shalnov@intel.com>
14886         * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14887         mode attribute for TARGET_AVX512VL.
14889 2017-09-21  Sergey Shalnov  <sergey.shalnov@intel.com>
14891         * config/i386/i386.opt (mprefer-avx256): New option.
14892         * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14893         to flag_opts.
14894         (ix86_preferred_simd_mode): Return 256-bit AVX modes
14895         for TARGET_PREFER_AVX256.
14896         * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14898 2017-09-21  Jeff Law  <law@redhat.com>
14900         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14901         Fix dump output if the only stack space is for pushed registers.
14903 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14905         * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14906         of insn_cost.
14908 2017-09-21  Martin Sebor  <msebor@redhat.com>
14910         PR c/81882
14911         * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
14912         code (in C++) or code that triggers warnings.
14914 2017-09-21  Eric Botcazou  <ebotcazou@adacore.com>
14916         * stor-layout.c (bit_from_pos): Do not distribute the conversion.
14918 2017-09-21  Segher Boessenkool  <segher@kernel.crashing.org>
14920         * haifa-sched.c: Rename insn_cost to insn_sched_cost.
14921         * sched-rgn.c: Ditto.
14922         * sel-sched-ir.c: Ditto.
14924 2017-09-21  Alexander Monakov  <amonakov@ispras.ru>
14926         * toplev.h (set_random_seed): Adjust return type.
14927         * toplev.c (init_local_tick): Move eager initialization of random_seed
14928         to get_random_seed.  Adjust comment.
14929         (init_random_seed): Inline to get_random_seed, delete.
14930         (get_random_seed): Initialize random_seed lazily.
14931         (set_random_seed): Do not return previous value.
14932         (print_switch_value): Do not call get_random_seed.
14934 2017-09-21  Evgeny Kudryashov  <kudryashov@ispras.ru>
14936         * cgraph.c (delete_function_version): New, broken out from...
14937         (cgraph_node::delete_function_version): ...here.  Rename to
14938         cgraph_node::delete_function_version_by_decl.  Update all uses.
14939         (cgraph_node::remove): Call delete_function_version.
14941 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14943         PR sanitizer/81715
14944         * tree-inline.c (expand_call_inline): Emit clobber stmts for
14945         VAR_DECLs to which addressable non-volatile parameters are mapped
14946         and for id->retvar after the return value assignment.  Clear
14947         id->retval and id->retbnd after inlining.
14949 2017-09-21  Richard Biener  <rguenther@suse.de>
14951         PR tree-optimization/82276
14952         PR tree-optimization/82244
14953         * tree-vrp.c (build_assert_expr_for): Set
14954         SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
14955         has it set.
14956         (remove_range_assertions): Revert earlier change.
14958 2017-09-21  Wilco Dijkstra  <wdijkstr@arm.com>
14960         PR target/71951
14961         * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
14963 2017-09-21  Richard Biener  <rguenther@suse.de>
14965         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
14966         Restore valid IL after code generation errors.
14967         * graphite.c (graphite_transform_loops): Diagnose code
14968         generation issues as MSG_MISSED_OPTIMIZATION and continue
14969         with processing SCOPs.
14971 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14972             Alan Hayward  <alan.hayward@arm.com>
14973             David Sherwood  <david.sherwood@arm.com>
14975         * calls.c (compute_argument_addresses): Use simplify_gen_binary
14976         rather than choosing between plus_constant and gen_rtx_<CODE>.
14977         * expr.c (emit_push_insn): Likewise.
14978         (expand_expr_real_2): Likewise.
14980 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14981             Alan Hayward  <alan.hayward@arm.com>
14982             David Sherwood  <david.sherwood@arm.com>
14984         * loop-unroll.c (split_iv): Call copy_rtx on the step.
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         * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
14991         calling tree_to_uhwi.
14993 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14994             Alan Hayward  <alan.hayward@arm.com>
14995             David Sherwood  <david.sherwood@arm.com>
14997         * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
14998         INTEGER_CST rather than a negative test for ADDR_EXPR.
15000 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
15001             Alan Hayward  <alan.hayward@arm.com>
15002             David Sherwood  <david.sherwood@arm.com>
15004         * tree-vrp.c (extract_range_from_binary_expr_1): Check
15005         int_cst_rangeN before calling value_range_constant_singleton (&vrN).
15007 2017-09-21  Richard Biener  <rguenther@suse.de>
15009         PR tree-optimization/71351
15010         * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
15011         graphite_create_new_loop_guard): Remove, fold remaining parts
15012         into caller ...
15013         (translate_isl_ast_node_for): ... here and simplify.
15015 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
15017         PR target/82260
15018         * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
15019         with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
15020         latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
15021         alternative always use QI mode, for -Os imov (=R,R) alternative
15022         always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
15023         ignore -Os.
15025 2017-09-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
15026             Jeff Law  <law@redhat.com>
15028         * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
15029         (allocate_stack_space): New function, partially extracted from
15030         s390_emit_prologue.
15031         (s390_emit_prologue): Track offset to most recent stack probe.
15032         Code to allocate space moved into allocate_stack_space.
15033         Dump actions when no stack is allocated.
15034         (s390_prologue_plus_offset): New function.
15035         (s390_emit_stack_probe): Likewise.
15037 2017-09-20  Alexandre Oliva  <aoliva@redhat.com>
15039         * common.opt (Wa, Wl, Wp, g, gz=): Add
15040         RejectNegative.
15041         (gno-column-info): Remove.
15042         (gcolumn-info): Drop RejectNegative.
15043         (gno-): New prefix.
15044         (gno-record-gcc-switches): Remove.
15045         (grecord-gcc-switches): Drop RejectNegative.
15046         (gno-split-dwarf): Remove.
15047         (gsplit-dwarf): Drop RejectNegative.
15048         (gno-strict-dwarf): Remove.
15049         (gstrict-dwarf): Drop RejectNegative.
15050         * config/darwin.opt (gfull, gused): Add RejectNegative.
15051         * dwarf2out.c (gen_producer_string): Drop
15052         gno-record-gcc-switches handler.
15053         * optc-gen.awk: Add g to prefixes with negative forms.
15054         * opts-common.c (remapping_prefix_p): New.
15055         (find_opt): Check it.
15056         (generate_canonical_option): Test g prefix.
15057         (option_map): Add -gno- mapping.
15058         (add_misspelling_candidates): Check remapping_prefix_p.
15060 2017-09-20  Jeff Law  <law@redhat.com>
15062         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
15063         thinko in stack clash protection support.
15065         * explow.c (compute_stack_clash_protection_loop_data): Use
15066         CONST_INT_P instead of explicit test.  Verify object is a
15067         CONST_INT_P before looking at INTVAL.
15068         (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
15069         instead of explicit test.
15071 2017-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
15073         PR target/77687
15074         * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
15075         address instead of to r1 and r11.
15077 2017-09-20  Sebastian Peryt  <sebastian.peryt@intel.com>
15079         * config.gcc: Support "knm".
15080         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
15081         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15082         PROCESSOR_KNM.
15083         * config/i386/i386.c (m_KNM): Define.
15084         (processor_target_table): Add "knm".
15085         (PTA_KNM): Define.
15086         (ix86_option_override_internal): Add "knm".
15087         (ix86_issue_rate): Add PROCESSOR_KNM.
15088         (ix86_adjust_cost): Ditto.
15089         (ia32_multipass_dfa_lookahead): Ditto.
15090         (get_builtin_code_for_version): Handle PROCESSOR_KNM.
15091         (fold_builtin_cpu): Add M_INTEL_KNM.
15092         * config/i386/i386.h (processor_costs): Define TARGET_KNM.
15093         (processor_type): Add PROCESSOR_KNM.
15094         * config/i386/x86-tune.def: Add m_KNM.
15095         * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
15097 2017-09-20  Richard Biener  <rguenther@suse.de>
15099         PR tree-optimization/80213
15100         * graphite-scop-detection.c (trivially_empty_bb_p): Labels
15101         are allowed in empty BBs as well.
15102         (canonicalize_loop_closed_ssa): Also look for other complex
15103         edges.
15104         (scop_detection::get_sese): Include the loop-closed PHI block
15105         in loop SESEs.
15106         (scop_detection::merge_sese): Remove code adding extra blocks.
15107         (scop_detection::region_has_one_loop): Adjust for get_sese changes.
15108         (build_scops): Assert the final returned scop is invalid.
15110 2017-09-20  Richard Biener  <rguenther@suse.de>
15112         PR tree-optimization/82264
15113         * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
15114         for GIMPLE_CONDs.
15115         (vn_phi_lookup): Likewise.
15116         (vn_phi_insert): Likewise.
15118 2017-09-20  Jakub Jelinek  <jakub@redhat.com>
15120         * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
15121         that fits into uhwi or shwi, add DW_AT_const_value regardless
15122         of early_dwarf without going through RTL, using add_AT_unsigned
15123         or add_AT_int.
15125         * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
15126         (DEBUG_LTO_ABBREV_SECTION): Likewise.
15127         (DEBUG_LTO_MACINFO_SECTION): Likewise.
15128         (DEBUG_MACRO_SECTION): Likewise.
15129         (DEBUG_LTO_MACRO_SECTION): Likewise.
15130         (DEBUG_STR_DWO_SECTION): Likewise.
15131         (DEBUG_LTO_STR_DWO_SECTION): Likewise.
15132         (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
15133         (DEBUG_LTO_DWO_LINE_SECTION): Define.
15134         (DEBUG_LTO_LINE_STR_SECTION): Define.
15135         (init_sections_and_labels): Initialize debug_line_str_section
15136         variable.  Initialize debug_loc_section for -gdwarf-5 to
15137         DEBUG_LOCLISTS_SECTION.  Formatting fixes.
15139 2017-09-20  Richard Biener  <rguenther@suse.de>
15141         * graphite-sese-to-poly.c (extract_affine): Properly handle
15142         POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
15144 2017-09-20  Richard Biener  <rguenther@suse.de>
15146         PR tree-optimization/81373
15147         * graphite-scop-detection.c (build_cross_bb_scalars_def):
15148         Force SESE live-out defs to be handled even if they are
15149         scev_analyzable_p.
15151 2017-09-19  Jeff Law  <law@redhat.com>
15153         * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
15154         nothing for stack adjustments with REG_STACK_CHECK.
15155         * sched-deps.c (parse_add_or_inc): Reject insns with
15156         REG_STACK_CHECK from dependency breaking.
15157         * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
15158         (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
15159         * reg-notes.def (STACK_CHECK): New note.
15161         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
15162         (ix86_expand_prologue): Dump stack clash info as needed.
15163         Call ix86_adjust_stack_and_probe_stack_clash as needed.
15165         * function.c (dump_stack_clash_frame_info): New function.
15166         * function.h (dump_stack_clash_frame_info): Prototype.
15167         (enum stack_clash_probes): New enum.
15169         * config/alpha/alpha.c (alpha_expand_prologue): Also check
15170         flag_stack_clash_protection.
15171         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
15172         (arm_expand_prologue, thumb1_expand_prologue): Likewise.
15173         (arm_frame_pointer_required): Likewise.
15174         * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
15175         (ia64_expand_prologue): Likewise.
15176         * config/mips/mips.c (mips_expand_prologue): Likewise.
15177         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
15178         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15179         (sparc_flat_expand_prologue): Likewise.
15180         * config/spu/spu.c (spu_expand_prologue): Likewise.
15182         * explow.c: Include "params.h".
15183         (anti_adjust_stack_and_probe_stack_clash): New function.
15184         (get_stack_check_protect): Likewise.
15185         (compute_stack_clash_protection_loop_data): Likewise.
15186         (emit_stack_clash_protection_loop_start): Likewise.
15187         (emit_stack_clash_protection_loop_end): Likewise.
15188         (allocate_dynamic_stack_space): Use get_stack_check_protect.
15189         Use anti_adjust_stack_and_probe_stack_clash.
15190         * explow.h (compute_stack_clash_protection_loop_data): Prototype.
15191         (emit_stack_clash_protection_loop_start): Likewise.
15192         (emit_stack_clash_protection_loop_end): Likewise.
15193         * rtl.h (get_stack_check_protect): Prototype.
15194         * target.def (stack_clash_protection_final_dynamic_probe): New hook.
15195         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
15196         * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
15197         Prototype.
15198         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
15199         Add @hook.
15200         * doc/tm.texi: Rebuilt.
15201         * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
15202         get_stack_check_protect.
15203         * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
15204         * config/arm/arm.c (arm_expand_prologue): Likewise.
15205         (arm_frame_pointer_required): Likewise.
15206         * config/i386/i386.c (ix86_expand_prologue): Likewise.
15207         * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
15208         * config/mips/mips.c (mips_expand_prologue): Likewise.
15209         * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
15210         * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
15211         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15212         (sparc_flat_expand_prologue): Likewise.
15214         * common.opt (-fstack-clash-protection): New option.
15215         * flag-types.h (enum stack_check_type): Note difference between
15216         -fstack-check= and -fstack-clash-protection.
15217         * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
15218         (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
15219         * toplev.c (process_options): Issue warnings/errors for cases
15220         not handled with -fstack-clash-protection.
15221         * doc/invoke.texi (-fstack-clash-protection): Document new option.
15222         (-fstack-check): Note additional problem with -fstack-check=generic.
15223         Note that -fstack-check is primarily for Ada and refer users
15224         to -fstack-clash-protection for stack-clash-protection.
15225         Document new params for stack clash protection.
15227 2017-09-19  Uros Bizjak  <ubizjak@gmail.com>
15229         * config/i386/i386.c (ix86_split_long_move): Do not handle
15230         address used for LEA in a special way.
15232 2017-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
15234         * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
15236 2017-09-19  Martin Sebor  <msebor@redhat.com>
15238         PR c/81854
15239         * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15240         of incompatible types.
15242 2017-09-19  Will Schmidt  <will_schmidt@vnet.ibm.com>
15244         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15245         for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15246         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15247         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15249 2017-09-19  Richard Biener  <rguenther@suse.de>
15251         PR tree-optimization/82244
15252         * tree-vrp.c (remove_range_assertions): Do not propagate
15253         a constant to abnormals but replace the assert with a copy.
15255 2017-09-19  Alexander Monakov  <amonakov@ispras.ru>
15257         PR rtl-optimization/57878
15258         PR rtl-optimization/68988
15259         * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15260         avoidance test involving non_reload_pseudos.  Move frequency test
15261         below the general fragmentation avoidance test.
15263 2017-09-19  Richard Biener  <rguenther@suse.de>
15265         PR tree-optimization/69728
15266         * graphite-sese-to-poly.c (schedule_error): New global.
15267         (add_loop_schedule): Handle empty domain by failing the
15268         schedule.
15269         (build_original_schedule): Handle schedule_error.
15271 2017-09-19  Richard Biener  <rguenther@suse.de>
15273         * graphite-scop-detection.c (scop_detection::can_represent_loop):
15274         Do not iterate to sibling loops but only to siblings of inner
15275         loops.
15277 2017-09-18  Andreas Schwab  <schwab@linux-m68k.org>
15279         PR target/81613
15280         * config/m68k/m68k.md (moveq feeding equality comparison): Check
15281         that the registers are different.
15283 2017-09-18  Uros Bizjak  <ubizjak@gmail.com>
15285         * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15286         to processor_model and "amdfam17h" to arch_names_table.
15287         * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15289 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15291         PR c/82234
15292         * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15294 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15295             Alan Hayward  <alan.hayward@arm.com>
15296             David Sherwood  <david.sherwood@arm.com>
15298         * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15299         with a vec_info *.
15300         * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15301         accordingly.
15302         * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15303         parameter.  Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15304         vect_schedule_slp_instance.
15305         (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15306         Update call to vect_slp_analyze_node_operations.  Simplify return
15307         value.
15308         (vect_slp_analyze_bb_1): Update call accordingly.
15309         (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15310         Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15311         (vect_schedule_slp): Update call accordingly.
15313 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15314             Alan Hayward  <alan.hayward@arm.com>
15315             David Sherwood  <david.sherwood@arm.com>
15317         * tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
15318         with types that aren't in fact scalar.
15320 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15322         * tree-vect-slp.c (vect_record_max_nunits): New function,
15323         split out from...
15324         (vect_build_slp_tree_1): ...here.
15325         (vect_build_slp_tree_2): Call it for phis too.
15327 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15329         * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15330         to vect_get_vec_def_for_operand when getting the mask operand.
15332 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15333             Alan Hayward  <alan.hayward@arm.com>
15334             David Sherwood  <david.sherwood@arm.com>
15336         * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15337         bitstart.
15339 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15340             Alan Hayward  <alan.hayward@arm.com>
15341             David Sherwood  <david.sherwood@arm.com>
15343         * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15344         calculation for vector booleans.
15346 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15347             Alan Hayward  <alan.hayward@arm.com>
15348             David Sherwood  <david.sherwood@arm.com>
15350         * tree-vect-stmts.c (can_vectorize_live_stmts): New function,
15351         split out from...
15352         (vect_transform_stmt): ...here.
15353         (vect_analyze_stmt): Use it instead of calling
15354         vectorizable_live_operation directly.
15356 2017-09-18  Cesar Philippidis  <cesar@codesourcery.com>
15358         * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15359         non-SIMT targets in acc vector loops.
15361 2017-09-18  Claudiu Zissulescu  <claziss@synopsys.com>
15363         * configure.ac: Add arc and check if assembler supports gdwarf2.
15364         * configure: Regenerate.
15366 2017-09-18  Richard Biener  <rguenther@suse.de>
15368         PR tree-optimization/82220
15369         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15370         epilogue niters from the min_profitable_iters compute.
15372 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15374         PR target/82145
15375         * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15376         changes.  Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15377         (ix86_init_pic_reg): Revert 2017-09-01 changes.
15379 2017-09-18  Eric Botcazou  <ebotcazou@adacore.com>
15381         PR target/81361
15382         * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15383         switching to a new text section.
15385 2017-09-18  Richard Biener  <rguenther@suse.de>
15387         * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15388         Simplify.
15389         (build_alias_set): Reject aliases with no access function.
15391 2017-09-18  Richard Biener  <rguenther@suse.de>
15393         PR tree-optimization/79622
15394         * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15395         handle PHIs.
15396         (build_cross_bb_scalars_use): Likewise.
15398 2017-09-18  Pierre-Marie de Rodat  <derodat@adacore.com>
15400         * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15402 2017-09-18  Alan Modra  <amodra@gmail.com>
15404         PR target/81996
15405         * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15406         stack_pointer_rtx for count 0.  Update comments.  Break up
15407         large rtl expression.
15409 2017-09-17  Daniel Santos  <daniel.santos@pobox.com>
15411         * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15412         Increase to 20 bytes.
15413         (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15414         (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15415         or avx version of the stub.
15417 2017-09-17  H.J. Lu  <hongjiu.lu@intel.com>
15419         PR target/82166
15420         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15421         compute the minimum stack alignment.  Also update preferred stack
15422         boundary for leaf functions.
15424 2017-09-16  Richard Sandiford  <richard.sandiford@linaro.org>
15426         PR tree-optimization/82228
15427         * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15428         of ncopies.
15430 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15432         * common/config/nds32/nds32-common.c
15433         (nds32_option_optimization_table): Refine formatting.
15434         (nds32_option_optimization_table): Use -fsched-pressure and
15435         -fomit-frame-pointer for specific optimization level.
15437 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15439         * config/nds32/nds32.c: Refine formatting and comments.
15440         * config/nds32/nds32.h: Likewise.
15441         * config/nds32/nds32.md: Likewise.
15442         * config/nds32/nds32-cost.c: Likewise.
15443         * config/nds32/nds32-isr.c: Likewise.
15444         * config/nds32/nds32-md-auxiliary.c: Likewise.
15445         * config/nds32/nds32-multiple.md: Likewise.
15446         * config/nds32/nds32-predicates.c: Likewise.
15448 2017-09-15  Andrew Sutton  <andrew.n.sutton@gmail.com>
15449             Jakub Jelinek  <jakub@redhat.com>
15451         Add support for -std=c++2a.
15452         * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15453         or -std=gnu+2a.
15454         * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15456 2017-09-15  Steve Ellcey  <sellcey@cavium.com>
15458         PR target/82066
15459         * doc/extend.texi (Common Function Attributes): Add 
15460         references to ARM, AArch64, and S/390 specific attributes.
15461         (Function Specific Option Pragmas): Add AArch64 and S/390
15462         to list of back ends that support the target pragma.
15464 2017-09-15  Nathan Sidwell  <nathan@acm.org>
15466         * doc/standards.texi: Fix C++17 description.  Update URLs for
15467         C++11 & 14.
15469 2017-09-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15471         * common.opt (Wcast-align=strict): New warning option.
15472         * doc/invoke.texi: Document -Wcast-align=strict. 
15474 2017-09-15  Pierre-Marie de Rodat  <derodat@adacore.com>
15476         * cgraph.h (cgraph_thunk_info): Add comments.
15477         * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15478         assert for VIRTUAL_* arguments stricter.
15480 2017-09-15  Jackson Woodruff  <jackson.woodruff@arm.com>
15482         PR tree-optimization/71026
15483         * match.pd: Move RDIV patterns from fold-const.c
15484         * fold-const.c (distribute_real_division): Removed.
15485         (fold_binary_loc): Remove calls to distribute_real_divison.
15487 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15489         * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15490         c++1z and gnu++1z as deprecated.  Change other references to
15491         -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15492         Change -Wc++1z-compat to -Wc++17-compat.
15493         * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15494         * dwarf2out.c (highest_c_language): Handle C++17.
15495         (gen_compile_unit_die): Likewise.
15497 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15499         PR rtl-optimization/82192
15500         * combine.c (make_extraction): Don't look through non-paradoxical
15501         SUBREGs or TRUNCATE if pos + len is or might be bigger than
15502         inner's mode.
15504 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15505             Alan Hayward  <alan.hayward@arm.com>
15506             David Sherwood  <david.sherwood@arm.com>
15508         * target.def (function_arg_offset): New hook.
15509         * targhooks.h (default_function_arg_offset): Declare.
15510         * targhooks.c (default_function_arg_offset): New function.
15511         * function.c (locate_and_pad_parm): Use
15512         targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15513         * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15514         (TARGET_FUNCTION_ARG_OFFSET): ...this.
15515         * doc/tm.texi: Regenerate.
15516         * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15517         * config/spu/spu.c (spu_function_arg_offset): New function.
15518         (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15519         * system.h (FUNCTION_ARG_OFFSET): Poison.
15521 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15522             Alan Hayard  <alan.hayward@arm.com>
15523             David Sherwood  <david.sherwood@arm.com>
15525         * target.def (truly_noop_truncation): New hook.
15526         (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15527         than TRULY_NOOP_TRUNCATION.
15528         * hooks.h (hook_bool_uint_uint_true): Declare.
15529         * hooks.c (hook_bool_uint_uint_true): New function.
15530         * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15531         (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15532         * doc/tm.texi: Regenerate.
15533         * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15534         rather than TRULY_NOOP_TRUNCATION in comments.
15535         (simplify_comparison): Likewise.
15536         (record_truncated_value): Likewise.
15537         * expmed.c (extract_bit_field_1): Likewise.
15538         (extract_split_bit_field): Likewise.
15539         * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15540         instead of TRULY_NOOP_TRUNCATION.
15541         * function.c (assign_parm_setup_block): Likewise.
15542         * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15543         * rtlhooks.c: Include target.h.
15544         * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15545         * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15546         * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15547         * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15548         * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15549         * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15550         * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15551         * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15552         * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15553         * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15554         * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15555         * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15556         * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15557         * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15558         * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15559         * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15560         * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15561         * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15562         * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15563         * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15564         * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15565         * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15566         * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15567         * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15568         * config/mips/mips.c (mips_truly_noop_truncation): New function.
15569         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15570         * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15571         rather than TRULY_NOOP_TRUNCATION in comments.
15572         * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15573         * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15574         * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15575         * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15576         * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15577         * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15578         * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15579         * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15580         * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15581         * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15582         * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15583         * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15584         rather than TRULY_NOOP_TRUNCATION in comments.
15585         * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15586         * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15587         * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15588         * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15589         * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15590         TRULY_NOOP_TRUNCATION condition.
15591         (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15592         (TRULY_NOOP_TRUNCATION): Delete.
15593         * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15594         * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15595         * config/spu/spu.c (spu_truly_noop_truncation): New function.
15596         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15597         * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15598         * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15599         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15600         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15601         * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15602         rather than TRULY_NOOP_TRUNCATION in comments.
15603         * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15604         * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15605         * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15606         * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15607         * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15608         * system.h (TRULY_NOOP_TRUNCATION): Poison.
15610 2017-09-15  Christophe Lyon  <christophe.lyon@linaro.org>
15612         PR target/67591
15613         * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15614         (*cmp_ior): Likewise.
15615         (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15616         (*ior_scc_scc_cmp): Likewise.
15617         (*and_scc_scc): Likewise.
15618         (*and_scc_scc_cmp): Likewise.
15620 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15621             Alan Hayard  <alan.hayward@arm.com>
15622             David Sherwood  <david.sherwood@arm.com>
15624         * target.def (can_change_mode_class): New hook.
15625         (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15626         (hard_regno_nregs): Likewise.
15627         * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15628         * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15629         * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15630         (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15631         (LOAD_EXTEND_OP): Update accordingly.
15632         * doc/tm.texi: Regenerate.
15633         * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15634         CANNOT_CHANGE_MODE_CLASS.
15635         * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15636         (REG_CAN_CHANGE_MODE_P): ...this new macro.
15637         * combine.c (simplify_set): Update accordingly.
15638         * emit-rtl.c (validate_subreg): Likewise.
15639         * recog.c (general_operand): Likewise.
15640         * regcprop.c (mode_change_ok): Likewise.
15641         * reload1.c (choose_reload_regs): Likewise.
15642         (inherit_piecemeal_p): Likewise.
15643         * rtlanal.c (simplify_subreg_regno): Likewise.
15644         * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15645         instead of CANNOT_CHANGE_MODE_CLASS.
15646         (reload_cse_simplify_operands): Likewise.
15647         * reload.c (push_reload): Use targetm.can_change_mode_class
15648         instead of CANNOT_CHANGE_MODE_CLASS.
15649         (push_reload): Likewise.  Also use REG_CAN_CHANGE_MODE_P instead of
15650         REG_CANNOT_CHANGE_MODE_P.
15651         * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15652         * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15653         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15654         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15655         * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15656         (arm_can_change_mode_class): New function.
15657         * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15658         than CANNOT_CHANGE_MODE_CLASS in comments.
15659         * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15660         * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15661         * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15662         (ix86_can_change_mode_class): ...this new function, inverting the
15663         sense of the return value.
15664         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15665         * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15666         * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15667         (ia64_can_change_mode_class): New function.
15668         * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15669         * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15670         * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15671         (m32c_can_change_mode_class): ...this new function, inverting the
15672         sense of the return value.
15673         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15674         * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15675         * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15676         * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15677         (mips_can_change_mode_class): ...this new function, inverting the
15678         sense of the return value.
15679         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15680         * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15681         * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15682         (msp430_can_change_mode_class): New function.
15683         * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15684         * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15685         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15686         * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15687         * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15688         * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15689         * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15690         (pa_cannot_change_mode_class): Replace with...
15691         (pa_can_change_mode_class): ...this new function, inverting the
15692         sense of the return value.
15693         (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15694         than CANNOT_CHANGE_MODE_CLASS in comments.
15695         * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15696         * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15697         * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15698         (pdp11_cannot_change_mode_class): Replace with...
15699         (pdp11_can_change_mode_class): ...this new function, inverting the
15700         sense of the return value.
15701         * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15702         * config/powerpcspe/powerpcspe-protos.h
15703         (rs6000_cannot_change_mode_class_ptr): Delete.
15704         * config/powerpcspe/powerpcspe.c
15705         (rs6000_cannot_change_mode_class_ptr): Delete.
15706         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15707         (rs6000_option_override_internal): Assign to
15708         targetm.can_change_mode_class instead of
15709         rs6000_cannot_change_mode_class_ptr.
15710         (rs6000_cannot_change_mode_class): Replace with...
15711         (rs6000_can_change_mode_class): ...this new function, inverting the
15712         sense of the return value.
15713         (rs6000_debug_cannot_change_mode_class): Replace with...
15714         (rs6000_debug_can_change_mode_class): ...this new function.
15715         * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15716         * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15717         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15718         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15719         * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15720         Delete.
15721         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15722         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15723         (rs6000_option_override_internal): Assign to
15724         targetm.can_change_mode_class instead of
15725         rs6000_cannot_change_mode_class_ptr.
15726         (rs6000_cannot_change_mode_class): Replace with...
15727         (rs6000_can_change_mode_class): ...this new function, inverting the
15728         sense of the return value.
15729         (rs6000_debug_cannot_change_mode_class): Replace with...
15730         (rs6000_debug_can_change_mode_class): ...this new function.
15731         * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15732         * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15733         * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15734         (s390_can_change_mode_class): ...this new function, inverting the
15735         sense of the return value.
15736         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15737         * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15738         * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15739         * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15740         (sh_cannot_change_mode_class): Replace with...
15741         (sh_can_change_mode_class): ...this new function, inverting the
15742         sense of the return value.
15743         * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15744         * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15745         (sparc_can_change_mode_class): New function.
15746         * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15747         * config/spu/spu.c (spu_can_change_mode_class): New function.
15748         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15749         * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15750         * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15751         (visium_can_change_mode_class): New function.
15752         * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15754 2017-09-15  Richard Biener  <rguenther@suse.de>
15756         PR tree-optimization/82217
15757         * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15758         but not undefined case.
15760 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15762         PR target/82145
15763         * postreload.c (reload_cse_simplify_operands): Skip
15764         NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15766 2017-09-15  Richard Biener  <rguenther@suse.de>
15768         PR tree-optimization/68823
15769         * graphite-scop-detection.c (build_alias_set): If we have a
15770         possible dependence check whether we can handle them by just
15771         looking at the DRs DR_ACCESS_FNs.
15772         (build_scops): If build_alias_set fails, fail the SCOP.
15774 2017-09-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
15776         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15777         to support float128 built-in functions that require the ISA 3.0
15778         hardware.
15779         (BU_FLOAT128_3_HW): Likewise.
15780         (SQRTF128): Add support for the IEEE 128-bit square root and fma
15781         built-in functions.
15782         (FMAF128): Likewise.
15783         (FMAQ): Likewise.
15784         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15785         support for built-in functions that need the ISA 3.0 IEEE 128-bit
15786         floating point instructions.
15787         (rs6000_invalid_builtin): Likewise.
15788         (rs6000_builtin_mask_names): Likewise.
15789         * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15790         (RS6000_BTM_FLOAT128_HW): Likewise.
15791         (RS6000_BTM_COMMON): Likewise.
15792         * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15793         function.
15794         * doc/extend.texi (RS/6000 built-in functions): Document the
15795         IEEE 128-bit floating point square root and fused multiply-add
15796         built-in functions.
15798 2017-09-14  Pat Haugen  <pthaugen@us.ibm.com>
15800         * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15801         reg (r2) isn't in the set of registers defined in the prologue.
15803 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15804             Alan Hayward  <alan.hayward@arm.com>
15805             David Sherwood  <david.sherwood@arm.com>
15807         * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15808         (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15809         (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15810         (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15811         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15812         accordingly.
15813         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15814         max_vectorization_factor.
15815         (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
15817 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15818             Alan Hayward  <alan.hayward@arm.com>
15819             David Sherwood  <david.sherwood@arm.com>
15821         * tree-vectorizer.h (vect_min_worthwhile_factor): Delete.
15822         (vect_worthwhile_without_simd_p): Declare.
15823         * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15824         (vectorizable_reduction): Use it.
15825         * tree-vect-stmts.c (vectorizable_shift): Likewise.
15826         (vectorizable_operation): Likewise.
15828 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15829             Alan Hayward  <alan.hayward@arm.com>
15830             David Sherwood  <david.sherwood@arm.com>
15832         * tree-vectorizer.h (vect_get_num_copies): New function.
15833         * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15834         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15835         (vectorizable_induction): Likewise.
15836         (vectorizable_live_operation): Likewise.
15837         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15838         (vectorizable_bswap): Likewise.
15839         (vectorizable_call): Likewise.
15840         (vectorizable_conversion): Likewise.
15841         (vectorizable_assignment): Likewise.
15842         (vectorizable_shift): Likewise.
15843         (vectorizable_operation): Likewise.
15844         (vectorizable_store): Likewise.
15845         (vectorizable_load): Likewise.
15846         (vectorizable_condition): Likewise.
15847         (vectorizable_comparison): Likewise.
15848         (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15850 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15851             Alan Hayward  <alan.hayward@arm.com>
15852             David Sherwood  <david.sherwood@arm.com>
15854         * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15855         of vect_init_vector.
15857 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15858             Alan Hayward  <alan.hayward@arm.com>
15859             David Sherwood  <david.sherwood@arm.com>
15861         * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15862         an inline wrapper that provides a location.
15863         (gimple_build_vector): Likewise.
15864         * gimple-fold.c (gimple_build_vector_from_val): New function.
15865         (gimple_build_vector): Likewise.
15866         * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15867         functions to build the initial value.  Always return a gimple value.
15868         (get_initial_defs_for_reduction): Likewise.  Only compute
15869         neutral_vec once.
15870         (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15871         vect_init_vector on the results from get_initial_def(s)_for_reduction.
15872         (vectorizable_induction): Use gimple_build_vector rather than
15873         vect_init_vector.
15875 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15876             Alan Hayward  <alan.hayward@arm.com>
15877             David Sherwood  <david.sherwood@arm.com>
15879         * target.h (vec_perm_indices): New typedef.
15880         (auto_vec_perm_indices): Likewise.
15881         * optabs-query.h: Include target.h
15882         (can_vec_perm_p): Take a vec_perm_indices *.
15883         * optabs-query.c (can_vec_perm_p): Likewise.
15884         (can_mult_highpart_p): Update accordingly.  Use auto_vec_perm_indices.
15885         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15886         * tree-vect-generic.c (lower_vec_perm): Likewise.
15887         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15888         (vect_grouped_load_supported): Likewise.
15889         (vect_shift_permute_load_chain): Likewise.
15890         (vect_permute_store_chain): Use auto_vec_perm_indices.
15891         (vect_permute_load_chain): Likewise.
15892         * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15893         (fold_ternary_loc): Update accordingly.  Use auto_vec_perm_indices.
15894         Update uses of can_vec_perm_p.
15895         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15896         mode with a number of elements.  Take a vec_perm_indices *.
15897         (vect_create_epilog_for_reduction): Update accordingly.
15898         Use auto_vec_perm_indices.
15899         (have_whole_vector_shift): Likewise.  Update call to can_vec_perm_p.
15900         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15901         (vect_transform_slp_perm_load): Likewise.
15902         (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15903         * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15904         (vect_gen_perm_mask_checked): Likewise.
15905         * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15906         (vect_gen_perm_mask_checked): Likewise.
15907         (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15908         (vectorizable_store): Likewise.
15909         (vectorizable_load): Likewise.
15910         (perm_mask_for_reverse): Likewise.  Update call to can_vec_perm_p.
15911         (vectorizable_bswap): Likewise.
15913 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15914             Alan Hayward  <alan.hayward@arm.com>
15915             David Sherwood  <david.sherwood@arm.com>
15917         * tree.h (build_vector): Take a vec<tree> instead of a tree *.
15918         * tree.c (build_vector): Likewise.
15919         (build_vector_from_ctor): Update accordingly.
15920         (build_vector_from_val): Likewise.
15921         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
15922         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15923         * tree-vect-generic.c (add_rshift): Likewise.
15924         (expand_vector_divmod): Likewise.
15925         (optimize_vector_constructor): Likewise.
15926         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
15927         (vect_transform_slp_perm_load): Likewise.
15928         (vect_schedule_slp_instance): Likewise.
15929         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15930         (vectorizable_call): Likewise.
15931         (vect_gen_perm_mask_any): Likewise.  Add elements in order.
15932         * expmed.c (make_tree): Likewise.
15933         * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
15934         a vector passed to build_vector.
15935         (fold_convert_const): Likewise.
15936         (exact_inverse): Likewise.
15937         (fold_ternary_loc): Likewise.
15938         (fold_relational_const): Likewise.
15939         (const_binop): Likewise.  Use VECTOR_CST_ELT directly when operating
15940         on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
15941         (const_unop): Likewise.  Store the reduction accumulator in a
15942         variable rather than an array.
15943         (vec_cst_ctor_to_array): Take the number of elements as a parameter.
15944         (fold_vec_perm): Update calls accordingly.  Use auto_vec<tree> for
15945         the new vector, rather than constructing it after the input arrays.
15946         (native_interpret_vector): Use auto_vec<tree> when building
15947         a vector passed to build_vector.  Add elements in order.
15948         * tree-vect-loop.c (get_initial_defs_for_reduction): Use
15949         auto_vec<tree> when building a vector passed to build_vector.
15950         (vect_create_epilog_for_reduction): Likewise.
15951         (vectorizable_induction): Likewise.
15952         (get_initial_def_for_reduction): Likewise.  Fix indentation of
15953         case statements.
15954         * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
15955         to a vec<tree> *.
15956         (sparc_fold_builtin): Use auto_vec<tree> when building a vector
15957         passed to build_vector.
15959 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15960             Alan Hayward  <alan.hayward@arm.com>
15961             David Sherwood  <david.sherwood@arm.com>
15963         * tree-core.h (tree_base::u): Add an "nelts" field.
15964         (tree_vector): Use VECTOR_CST_NELTS as the length.
15965         * tree.c (tree_size): Likewise.
15966         (make_vector): Initialize VECTOR_CST_NELTS.
15967         * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
15968         * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
15969         TYPE_VECTOR_SUBPARTS.
15970         * expr.c (const_vector_mask_from_tree): Consistently use "units"
15971         as the number of units, setting it from VECTOR_CST_NELTS.
15972         (const_vector_from_tree): Likewise.
15973         * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
15974         TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
15975         (fold_negate_expr_1): Likewise.
15976         (fold_convert_const): Likewise.
15977         (const_binop): Likewise.  Differentiate the number of output and
15978         input elements.
15979         (const_unop): Likewise.
15980         (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
15981         in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
15982         in cases that did the opposite.
15984 2017-09-14  Richard Biener  <rguenther@suse.de>
15986         * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
15987         to VN_TOP.
15989 2017-09-14  Eric Botcazou  <ebotcazou@adacore.com>
15991         * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
15993 2017-09-14  Jakub Jelinek  <jakub@redhat.com>
15995         PR target/81325
15996         * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
15997         if and where to split a bb, except for splitting before debug insn
15998         sequences followed by non-label real insn.  Delete debug insns
15999         in between basic blocks.
16001         * combine.c (make_compound_operation_int): Formatting fixes.
16003         * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
16004         * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16005         * config/netbsd.h (LINK_EH_SPEC): Likewise.
16006         * config/sol2.h (LINK_EH_SPEC): Likewise.
16007         * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16008         * config/s390/linux.h (LINK_SPEC): Likewise.
16009         * config/freebsd.h (LINK_EH_SPEC): Likewise.
16010         * config/openbsd.h (LINK_EH_SPEC): Likewise.
16011         * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16012         * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
16013         * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
16014         * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16015         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
16016         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
16018 2017-09-13  Jakub Jelinek  <jakub@redhat.com>
16020         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
16021         support.
16022         (ENDFILE_LINUX_SPEC): Likewise.
16023         (LINK_EH_SPEC): Likewise.
16024         * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
16025         (LINK_OS_LINUX_SPEC32): Likewise.
16026         (LINK_OS_LINUX_SPEC64): Likewise.
16027         * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
16028         (LINK_OS_LINUX_SPEC): Likewise.
16030 2017-09-13  Martin Liska  <mliska@suse.cz>
16032         PR middle-end/82154
16033         * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
16034         CASE_HIGH is NULL_TREE.
16036 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16037             Alan Hayward  <alan.hayward@arm.com>
16038             David Sherwood  <david.sherwood@arm.com>
16040         * target.def (secondary_memory_needed): New hook.
16041         (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
16042         instead of SECONDARY_MEMORY_NEEDED.
16043         (secondary_memory_needed_mode): Likewise.
16044         * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
16045         * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
16046         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
16047         (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
16048         (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
16049         * doc/tm.texi: Regenerate.
16050         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
16051         * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
16052         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16053         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
16054         * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
16055         * config/i386/i386.c (inline_secondary_memory_needed): Put the
16056         mode argument first and change the reg_class arguments to reg_class_t.
16057         (ix86_secondary_memory_needed): Likewise.  Remove the strict parameter.
16058         Make static.  Update the call to inline_secondary_memory_needed.
16059         (ix86_register_move_cost): Update the call to
16060         inline_secondary_memory_needed.
16061         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16062         * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
16063         definition.
16064         * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
16065         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16066         in comment.
16067         * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
16068         * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
16069         * config/mips/mips.c (mips_secondary_memory_needed): Make static
16070         and match hook interface.  Add comment from mips.h.
16071         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16072         * config/mmix/mmix.md (truncdfsf2): Refer to
16073         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16074         in comment.
16075         * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
16076         (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
16077         * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16078         (pa_secondary_memory_needed): New function.
16079         * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
16080         * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
16081         * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16082         (pdp11_secondary_memory_needed): Make static and match hook interface.
16083         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
16084         * config/powerpcspe/powerpcspe-protos.h
16085         (rs6000_secondary_memory_needed_ptr): Delete.
16086         * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
16087         Delete.
16088         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16089         (rs6000_option_override_internal): Assign to
16090         targetm.secondary_memory_needed rather than
16091         rs6000_secondary_memory_needed_ptr.
16092         (rs6000_secondary_memory_needed): Match hook interface.
16093         (rs6000_debug_secondary_memory_needed): Likewise.
16094         * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
16095         * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
16096         (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
16097         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16098         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
16099         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
16100         Delete.
16101         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): Delete.
16102         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16103         (rs6000_option_override_internal): Assign to
16104         targetm.secondary_memory_needed rather than
16105         rs6000_secondary_memory_needed_ptr.
16106         (rs6000_secondary_memory_needed): Match hook interface.
16107         (rs6000_debug_secondary_memory_needed): Likewise.
16108         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
16109         * config/s390/s390.c (s390_secondary_memory_needed): New function.
16110         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16111         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
16112         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16113         (sparc_secondary_memory_needed): New function.
16114         * lra-constraints.c (check_and_process_move): Refer to
16115         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16116         in comment.
16117         (curr_insn_transform): Likewise.
16118         (process_alt_operands): Use targetm.secondary_memory_needed
16119         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16120         (check_secondary_memory_needed_p): Likewise.
16121         (choose_split_class): Likewise.
16122         * reload.c: Unconditionally include code that was previously
16123         conditional on SECONDARY_MEMORY_NEEDED.
16124         (push_secondary_reload): Use targetm.secondary_memory_needed
16125         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16126         (push_reload): Likewise.
16127         * reload1.c: Unconditionally include code that was previously
16128         conditional on SECONDARY_MEMORY_NEEDED.
16129         (choose_reload_regs): Use targetm.secondary_memory_needed
16130         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16131         (gen_reload): Likewise.
16132         * system.h (SECONDARY_MEMORY_NEEDED): Poison.
16134 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16135             Alan Hayward  <alan.hayward@arm.com>
16136             David Sherwood  <david.sherwood@arm.com>
16138         * target.def (secondary_memory_needed_mode): New hook:
16139         * targhooks.c (default_secondary_memory_needed_mode): Declare.
16140         * targhooks.h (default_secondary_memory_needed_mode): New function.
16141         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
16142         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
16143         * doc/tm.texi: Regenerate.
16144         * lra-constraints.c (check_and_process_move): Use
16145         targetm.secondary_memory_needed_mode instead of
16146         TARGET_SECONDARY_MEMORY_NEEDED_MODE.
16147         (curr_insn_transform): Likewise.
16148         * reload.c (get_secondary_mem): Likewise.
16149         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16150         * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
16151         function.
16152         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16153         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16154         * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
16155         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16156         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
16157         Delete.
16158         * config/powerpcspe/powerpcspe-protos.h
16159         (rs6000_secondary_memory_needed_mode): Delete.
16160         * config/powerpcspe/powerpcspe.c
16161         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16162         (rs6000_secondary_memory_needed_mode): Make static.
16163         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16164         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
16165         Delete.
16166         * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16167         Redefine.
16168         (rs6000_secondary_memory_needed_mode): Make static.
16169         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16170         * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
16171         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16172         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16173         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16174         Redefine.
16175         (sparc_secondary_memory_needed_mode): New function.
16176         * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
16178 2017-09-13  Jackson Woodruff  <jackson.woodruff@arm.com>
16180         * config/aarch64/constraints.md (Umq): New constraint.
16181         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
16182         Change to use Umq.
16183         (mov<mode>): Update condition.
16185 2017-09-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16187         * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
16188         when bitposition is the same.
16190 2017-09-13  Richard Biener  <rguenther@suse.de>
16192         * dwarf2out.c (output_die_symbol): Remove.
16193         (output_die): Do not output a DIEs symbol.
16195 2017-09-13  Richard Biener  <rguenther@suse.de>
16197         PR middle-end/82128
16198         * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
16199         default-def to avoid breaking iterator update with the weird
16200         interaction with cgraph_update_edges_for_call_stmt_node.
16202 2017-09-13  Richard Biener  <rguenther@suse.de>
16204         * tree-cfg.c (verify_gimple_assign_binary): Add verification
16205         for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
16206         VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
16207         (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
16209 2017-09-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
16211         * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
16212         Disable pc relative literal load irrespective of
16213         TARGET_FIX_ERR_A53_84341 for default.
16215 2017-09-12  Eric Botcazou  <ebotcazou@adacore.com>
16217         * config/sparc/sparc.c (output_return): Output the source location of
16218         the insn in the delay slot, if any.
16219         (output_sibcall): Likewise.
16221 2017-09-12  H.J. Lu  <hongjiu.lu@intel.com>
16223         PR driver/81498
16224         * common.opt (-static-pie): New alias.
16225         (shared): Negate static-pie.
16226         (-no-pie): Update help text.
16227         (-pie): Likewise.
16228         (static-pie): New option.
16229         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
16230         -static-pie support.
16231         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
16232         (LINK_EH_SPEC): Likewise.
16233         (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16234         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16235         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16236         * gcc.c (LINK_COMMAND_SPEC): Likewise.
16237         (init_gcc_specs): Likewise.
16238         (init_spec): Likewise.
16239         (display_help): Update help message for -pie.
16240         * doc/invoke.texi: Update -pie, -no-pie and -static.  Document
16241         -static-pie.
16243 2017-09-12  Wilco Dijkstra  <wdijkstr@arm.com>
16245         * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16246         (movdi_aarch64): Likewise.
16247         (movti_aarch64): Likewise.
16249 2017-09-12 Simon Wright <simon@pushface.org>
16251         PR target/80204
16252         * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16253         calculation of the minor version, always output as 0.
16255 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16257         PR target/82112
16258         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16259         ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16260         on it early, rather than manual conversion late.  For
16261         ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16262         instead of performing manual conversion.
16264 2017-09-12  Carl Love  <cel@us.ibm.com>
16266         * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16267         vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16268         vmulouw, vmulosw.
16269         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16270         VMULOSW): Add definitions.
16271         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16272         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16273         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16274         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16275         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16277 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16279         * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16280         types correctly.
16281         (movti_aarch64): Likewise.
16282         (movdf_aarch64): Likewise.
16283         (movtf_aarch64): Likewise.
16284         (load_pairdi): Likewise.
16285         (store_pairdi): Likewise.
16286         (load_pairdf): Likewise.
16287         (store_pairdf): Likewise.
16288         (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16289         (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16290         (ldr_got_small_<mode>): Likewise.
16291         (ldr_got_small_28k_<mode>): Likewise.
16292         (ldr_got_tiny): Likewise.
16293         * config/aarch64/iterators.md (ldst_sz): New.
16294         (ldpstp_sz): Likewise.
16295         * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16296         to store_16.
16297         (thunderx_load): Split load_8 to load_16.
16298         * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16299         load_8 to load_16.
16300         (thunderx2t99_storepair_basic): Split store_8 to store_16.
16301         * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16302         (xgene1_store_pair): Split store_8 to store_16.
16303         * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16304         (falkor_st_0_st_sd): Split store_8 to store_16.
16306 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16308         * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16309         and store1/2/3/4 to store_4/8/12/16.
16310         * config/aarch64/aarch64.md: Update for rename.
16311         * config/arm/arm.md: Likewise.: Likewise.
16312         * config/arm/arm.c: Likewise.
16313         * config/arm/thumb1.md: Likewise.
16314         * config/arm/thumb2.md: Likewise.
16315         * config/arm/vfp.md: Likewise.
16316         * config/arm/arm-generic.md: Likewise.
16317         * config/arm/arm1020e.md: Likewise.
16318         * config/arm/arm1026ejs.md: Likewise.
16319         * config/arm/arm1136jfs.md: Likewise.
16320         * config/arm/arm926ejs.md: Likewise.
16321         * config/arm/cortex-a15.md: Likewise.
16322         * config/arm/cortex-a17.md: Likewise.
16323         * config/arm/cortex-a5.md: Likewise.
16324         * config/arm/cortex-a53.md: Likewise.
16325         * config/arm/cortex-a57.md: Likewise.
16326         * config/arm/cortex-a7.md: Likewise.
16327         * config/arm/cortex-a8.md: Likewise.
16328         * config/arm/cortex-a9.md: Likewise.
16329         * config/arm/cortex-m4.md: Likewise.
16330         * config/arm/cortex-m7.md: Likewise.
16331         * config/arm/cortex-r4.md: Likewise.
16332         * config/arm/exynos-m1.md: Likewise.
16333         * config/arm/fa526.md: Likewise.
16334         * config/arm/fa606te.md: Likewise.
16335         * config/arm/fa626te.md: Likewise.
16336         * config/arm/fa726te.md: Likewise.
16337         * config/arm/fmp626.md: Likewise.
16338         * config/arm/iwmmxt.md: Likewise.
16339         * config/arm/ldmstm.md: Likewise.
16340         * config/arm/marvell-pj4.md: Likewise.
16341         * config/arm/xgene1.md: Likewise.
16342         * config/aarch64/thunderx.md: Likewise.
16343         * config/aarch64/thunderx2t99.md: Likewise.
16344         * config/aarch64/falkor.md: Likewise.
16346 2017-09-12  Martin Liska  <mliska@suse.cz>
16348         * attribs.c (private_lookup_attribute): New function.
16349         * attribs.h (private_lookup_attribute): Declared here.
16350         (lookup_attribute): Called from this place.
16352 2017-09-12  Richard Biener  <rguenther@suse.de>
16354         PR tree-optimization/82157
16355         * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16356         stmts with side-effects.
16358 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16359             Alan Hayward  <alan.hayward@arm.com>
16360             David Sherwood <david.sherwood@arm.com>
16362         * target.def (hard_regno_nregs): New hook.
16363         (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16364         * targhooks.h (default_hard_regno_nregs): Declare.
16365         * targhooks.c (default_hard_regno_nregs): New function.
16366         * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16367         (TARGET_HARD_REGNO_NREGS): ...this hook.
16368         (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16369         (CLASS_MAX_NREGS): Likewise.
16370         * doc/tm.texi: Regenerate.
16371         * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16372         instead of HARD_REGNO_NREGS.
16373         * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16374         HARD_REGNO_NREGS in the comment.
16375         * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16376         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16377         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16378         Return an unsigned int.
16379         (TARGET_HARD_REGNO_NREGS): Redefine.
16380         * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16381         * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16382         * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16383         (arc_hard_regno_nregs): New function.
16384         * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16385         * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16386         (arm_hard_regno_nregs): New function.
16387         * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16388         * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16389         * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16390         (TARGET_HARD_REGNO_NREGS): Redefine.
16391         * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16392         * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16393         (HARD_REGNO_NREGS): Delete.
16394         * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16395         (cr16_hard_regno_nregs): New function.
16396         (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16397         * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16398         * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16399         (cris_hard_regno_nregs): New function.
16400         * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16401         * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16402         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16403         * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16404         (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16405         * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16406         * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16407         (frv_hard_regno_nregs): Make static.  Take and return an
16408         unsigned int.
16409         (frv_class_max_nregs): Remove outdated copy of documentation.
16410         * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16411         * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16412         * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16413         * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16414         * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16415         * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16416         (TARGET_HARD_REGNO_NREGS): Redefine.
16417         * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16418         (CLASS_MAX_NREGS): Update comment.
16419         * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16420         (ia64_hard_regno_nregs): New function.
16421         * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16422         * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16423         * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16424         * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16425         * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16426         an unsigned int.
16427         (m32c_hard_regno_nregs): Likewise.  Make static.
16428         (TARGET_HARD_REGNO_NREGS): Redefine.
16429         * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16430         * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16431         * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16432         (m68k_hard_regno_nregs): New function.
16433         * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16434         * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16435         * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16436         * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16437         * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16438         Take and return an unsigned int.
16439         (TARGET_HARD_REGNO_NREGS): Redefine.
16440         * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16441         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16442         * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16443         * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16444         * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16445         * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16446         * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16447         (msp430_hard_regno_nregs): Make static.  Take and return an
16448         unsigned int.
16449         * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16450         * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16451         * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16452         (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16453         * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16454         * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16455         * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16456         (TARGET_HARD_REGNO_NREGS): Redefine.
16457         * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16458         (PA_HARD_REGNO_NREGS): ...this.
16459         * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16460         (PA_HARD_REGNO_NREGS): ...this.
16461         * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16462         (pa_hard_regno_nregs): New function.
16463         * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16464         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16465         (pdp11_hard_regno_nregs): New function.
16466         * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16467         * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16468         (rs6000_hard_regno_nregs_hook): New function.
16469         * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16470         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16471         * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16472         Take and return an unsigned int.  Move earlier in file.
16473         (TARGET_HARD_REGNO_NREGS): Redefine.
16474         * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16475         * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16476         * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16477         (rl78_hard_regno_nregs): Make static.  Take and return an
16478         unsigned int.
16479         * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16480         * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16481         (rs6000_hard_regno_nregs_hook): New function.
16482         * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16483         * config/rx/rx.c (rx_hard_regno_nregs): New function.
16484         (TARGET_HARD_REGNO_NREGS): Redefine.
16485         * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16486         * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16487         instead of HARD_REGNO_NREGS.
16488         (s390_hard_regno_nregs): New function.
16489         (s390_hard_regno_mode_ok): Add comment from s390.h.
16490         (TARGET_HARD_REGNO_NREGS): Redefine.
16491         * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16492         * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16493         (sh_hard_regno_nregs): New function.
16494         (sh_pass_in_reg_p): Use it.
16495         * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16496         * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16497         (sparc_hard_regno_nregs): New function.
16498         * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16499         * config/spu/spu.c (spu_hard_regno_nregs): New function.
16500         (spu_function_arg_advance): Use it, supplying a valid register number.
16501         (TARGET_HARD_REGNO_NREGS): Redefine.
16502         * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16503         * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16504         * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16505         * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16506         * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16507         * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16508         (CLASS_MAX_NREGS): Remove copy of old documentation.
16509         * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16510         (visium_hard_regno_nregs): New function.
16511         (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16512         * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16513         * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16514         xtensa_hard_regno_nregs): New function.
16515         * system.h (HARD_REGNO_NREGS): Poison.
16517 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16519         * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16520         hard_regno_nregs instead of HARD_REGNO_NREGS.
16521         (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16522         * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16523         (c6x_expand_epilogue): Likewise.
16524         * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16525         (frv_read_iacc_argument): Likewise.
16526         * config/sh/sh.c: Include regs.h.
16527         (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16528         (regs_used): Likewise.
16529         (output_stack_adjust): Likewise.
16530         * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16531         * expmed.c: Include regs.h.
16532         (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16533         * ree.c: Include regs.h.
16534         (combine_reaching_defs): Use hard_regno_nregs instead of
16535         HARD_REGNO_NREGS.
16536         (add_removable_extension): Likewise.
16538 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16540         * regs.h (hard_regno_nregs): Turn into a function.
16541         (end_hard_regno): Update accordingly.
16542         * caller-save.c (setup_save_areas): Likewise.
16543         (save_call_clobbered_regs): Likewise.
16544         (replace_reg_with_saved_mem): Likewise.
16545         (insert_restore): Likewise.
16546         (insert_save): Likewise.
16547         * combine.c (can_change_dest_mode): Likewise.
16548         (move_deaths): Likewise.
16549         (distribute_notes): Likewise.
16550         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16551         * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16552         (rs6000_split_multireg_move): Likewise.
16553         (rs6000_register_move_cost): Likewise.
16554         (rs6000_memory_move_cost): Likewise.
16555         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16556         (rs6000_split_multireg_move): Likewise.
16557         (rs6000_register_move_cost): Likewise.
16558         (rs6000_memory_move_cost): Likewise.
16559         * cselib.c (cselib_reset_table): Likewise.
16560         (cselib_lookup_1): Likewise.
16561         * emit-rtl.c (set_mode_and_regno): Likewise.
16562         * function.c (aggregate_value_p): Likewise.
16563         * ira-color.c (setup_profitable_hard_regs): Likewise.
16564         (check_hard_reg_p): Likewise.
16565         (calculate_saved_nregs): Likewise.
16566         (assign_hard_reg): Likewise.
16567         (improve_allocation): Likewise.
16568         (calculate_spill_cost): Likewise.
16569         * ira-emit.c (modify_move_list): Likewise.
16570         * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16571         (ira_hard_reg_in_set_p): Likewise.
16572         * ira.c (setup_reg_mode_hard_regset): Likewise.
16573         (clarify_prohibited_class_mode_regs): Likewise.
16574         (check_allocation): Likewise.
16575         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16576         (lra_setup_reg_renumber): Likewise.
16577         (setup_try_hard_regno_pseudos): Likewise.
16578         (spill_for): Likewise.
16579         (assign_hard_regno): Likewise.
16580         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16581         * lra-constraints.c (in_class_p): Likewise.
16582         (lra_constraint_offset): Likewise.
16583         (simplify_operand_subreg): Likewise.
16584         (lra_constraints): Likewise.
16585         (split_reg): Likewise.
16586         (split_if_necessary): Likewise.
16587         (invariant_p): Likewise.
16588         (inherit_in_ebb): Likewise.
16589         * lra-lives.c (process_bb_lives): Likewise.
16590         * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16591         (get_hard_regs): Likewise.
16592         (do_remat): Likewise.
16593         * lra-spills.c (assign_spill_hard_regs): Likewise.
16594         * mode-switching.c (create_pre_exit): Likewise.
16595         * postreload.c (reload_combine_recognize_pattern): Likewise.
16596         * recog.c (peep2_find_free_register): Likewise.
16597         * regcprop.c (kill_value_regno): Likewise.
16598         (set_value_regno): Likewise.
16599         (copy_value): Likewise.
16600         (maybe_mode_change): Likewise.
16601         (find_oldest_value_reg): Likewise.
16602         (copyprop_hardreg_forward_1): Likewise.
16603         * regrename.c (check_new_reg_p): Likewise.
16604         (regrename_do_replace): Likewise.
16605         * reload.c (push_reload): Likewise.
16606         (combine_reloads): Likewise.
16607         (find_dummy_reload): Likewise.
16608         (operands_match_p): Likewise.
16609         (find_reloads): Likewise.
16610         (find_equiv_reg): Likewise.
16611         (reload_adjust_reg_for_mode): Likewise.
16612         * reload1.c (count_pseudo): Likewise.
16613         (count_spilled_pseudo): Likewise.
16614         (find_reg): Likewise.
16615         (clear_reload_reg_in_use): Likewise.
16616         (free_for_value_p): Likewise.
16617         (allocate_reload_reg): Likewise.
16618         (choose_reload_regs): Likewise.
16619         (reload_adjust_reg_for_temp): Likewise.
16620         (emit_reload_insns): Likewise.
16621         (delete_output_reload): Likewise.
16622         * rtlanal.c (subreg_get_info): Likewise.
16623         * sched-deps.c (sched_analyze_reg): Likewise.
16624         * sel-sched.c (init_regs_for_mode): Likewise.
16625         (mark_unavailable_hard_regs): Likewise.
16626         (choose_best_reg_1): Likewise.
16627         (verify_target_availability): Likewise.
16628         * valtrack.c (dead_debug_insert_temp): Likewise.
16629         * var-tracking.c (track_loc_p): Likewise.
16630         (emit_note_insn_var_location): Likewise.
16631         * varasm.c (make_decl_rtl): Likewise.
16632         * reginfo.c (choose_hard_reg_mode): Likewise.
16633         (init_reg_modes_target): Refer directly to
16634         this_target_regs->x_hard_regno_nregs.
16636 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16638         * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16639         instead of hard_regno_nregs.
16641 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16643         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16644         end_hard_regno instead of hard_regno_nregs.
16645         * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16646         * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16647         * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16648         * ira-color.c (improve_allocation): Likewise.
16649         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16650         * lra-lives.c (mark_regno_live): Likewise.
16651         (mark_regno_dead): Likewise.
16652         * lra-remat.c (operand_to_remat): Likewise.
16653         * lra.c (collect_non_operand_hard_regs): Likewise.
16654         * postreload.c (reload_combine_note_store): Likewise.
16655         (move2add_valid_value_p): Likewise.
16656         * reload.c (regno_clobbered_p): Likewise.
16658 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16660         * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16661         hard_regno_nregs.
16662         * config/v850/v850.c (v850_reorg): Likewise.
16663         * reload.c (refers_to_regno_for_reload_p): Likewise.
16664         (find_equiv_reg): Likewise.
16665         * reload1.c (reload_reg_reaches_end_p): Likewise.
16667 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16669         * caller-save.c (add_used_regs): Use REG_NREGS instead of
16670         hard_regno_nregs.
16671         * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16672         * config/arm/arm.c (output_move_neon): Likewise.
16673         (arm_attr_length_move_neon): Likewise.
16674         (neon_split_vcombine): Likewise.
16675         * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16676         (c6x_mark_reg_written): Likewise.
16677         (c6x_dwarf_register_span): Likewise.
16678         * config/i386/i386.c (ix86_save_reg): Likewise.
16679         * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16680         (rws_access_reg): Likewise.
16681         * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16682         * mode-switching.c (create_pre_exit): Likewise.
16683         * ree.c (combine_reaching_defs): Likewise.
16684         (add_removable_extension): Likewise.
16685         * regcprop.c (find_oldest_value_reg): Likewise.
16686         (copyprop_hardreg_forward_1): Likewise.
16687         * reload.c (reload_inner_reg_of_subreg): Likewise.
16688         (push_reload): Likewise.
16689         (combine_reloads): Likewise.
16690         (find_dummy_reload): Likewise.
16691         (reload_adjust_reg_for_mode): Likewise.
16692         * reload1.c (find_reload_regs): Likewise.
16693         (forget_old_reloads_1): Likewise.
16694         (reload_reg_free_for_value_p): Likewise.
16695         (reload_adjust_reg_for_temp): Likewise.
16696         (emit_reload_insns): Likewise.
16697         (delete_output_reload): Likewise.
16698         * sel-sched.c (choose_best_reg_1): Likewise.
16699         (choose_best_pseudo_reg): Likewise.
16701 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16702             Alan Hayward  <alan.hayward@arm.com>
16703             David Sherwood <david.sherwood@arm.com>
16705         * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16706         * target.def (slow_unaligned_access): New hook.
16707         * targhooks.h (default_slow_unaligned_access): Declare.
16708         * targhooks.c (default_slow_unaligned_access): New function.
16709         * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16710         (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16711         * doc/tm.texi: Regenerate.
16712         * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16713         * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16714         * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16715         definition.
16716         * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16717         * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16718         Redefine.
16719         (rs6000_slow_unaligned_access): New function.
16720         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16721         (expand_block_compare): Likewise.
16722         (expand_strn_compare): Likewise.
16723         (rs6000_rtx_costs): Likewise.
16724         * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16725         (riscv_slow_unaligned_access): Likewise.
16726         * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16727         (riscv_slow_unaligned_access_p): ...this and make static.
16728         (riscv_option_override): Update accordingly.
16729         (riscv_slow_unaligned_access): New function.
16730         (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16731         * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16732         * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16733         (rs6000_slow_unaligned_access): New function.
16734         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16735         (rs6000_rtx_costs): Likewise.
16736         * config/rs6000/rs6000-string.c (expand_block_compare)
16737         (expand_strn_compare): Use targetm.slow_unaligned_access instead
16738         of SLOW_UNALIGNED_ACCESS.
16739         * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16740         * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16741         * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16742         of SLOW_UNALIGNED_ACCESS.
16743         * expmed.c (simple_mem_bitfield_p): Likewise.
16744         * expr.c (alignment_for_piecewise_move): Likewise.
16745         (emit_group_load_1): Likewise.
16746         (emit_group_store): Likewise.
16747         (copy_blkmode_from_reg): Likewise.
16748         (emit_push_insn): Likewise.
16749         (expand_assignment): Likewise.
16750         (store_field): Likewise.
16751         (expand_expr_real_1): Likewise.
16752         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16753         * lra-constraints.c (simplify_operand_subreg): Likewise.
16754         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16755         * gimple-ssa-store-merging.c: Likewise in block comment at start
16756         of file.
16757         * tree-ssa-strlen.c: Include target.h.
16758         (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16759         of SLOW_UNALIGNED_ACCESS.
16760         * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16762 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16764         PR rtl-optimization/82185
16765         * expmed.c (emit_store_flag_int): Only test tem if it has been
16766         initialized.
16768 2017-09-12  Richard Biener  <rguenther@suse.de>
16770         PR middle-end/82149
16771         * match.pd ((FTYPE) N CMP CST): Fix typo.
16773 2017-09-12  Simon Atanasyan  <simon.atanasyan@imgtec.com>
16775         * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16776         attribute.
16777         (mips_near_type_p): Add 'short_call' attribute as a synonym
16778         for 'near'.
16779         * doc/extend.texi (short_call): Document new function attribute.
16781 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16783         PR target/82112
16784         * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16785         assertion check that in the condition.
16786         (get_atomic_generic_size): Likewise.  Before testing if parameter
16787         has pointer type, if it has array type, call for C++
16788         default_conversion to perform array-to-pointer conversion.
16790 2017-09-12  Richard Biener  <rguenther@suse.de>
16792         * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16793         for operations we cannot scalarize.
16795 2017-09-12  Aldy Hernandez  <aldyh@redhat.com>
16797         * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16798         vectors heap vectors.  Clean up comments.
16799         Make visited_bbs a reference.
16800         (profitable_jump_thread_path): Make GC
16801         vectors heap vectors.  Clean up comments.
16802         Misc cleanups.
16803         (convert_and_register_jump_thread_path): Make GC vectors heap
16804         vectors.
16805         (check_subpath_and_update_thread_path): Same.  Clean up comments.
16806         Make visited_bbs a reference.
16807         (handle_phi): Abstract common code to to
16808         register_jump_thread_path_if_profitable.
16809         Rename VAR_BB to DEF_BB.
16810         Update comments.
16811         Make GC vectors heap vectors.
16812         Make visited_bbs a reference.
16813         (handle_assignment): Same.
16814         (register_jump_thread_path_if_profitable): New.
16815         (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16816         DEF_BB.
16817         Make GC vectors heap vectors.  Clean up comments.
16818         Make visited_bbs a reference.
16819         (find_jump_threads_backwards): Make visited_bbs live in the stack.
16820         * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16821         comment.
16823 2017-09-11  Max Filippov  <jcmvbkbc@gmail.com>
16825         PR target/82181
16826         * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16827         words of E_DImode object are reachable by xtensa_uimm8x4 access.
16829 2017-09-11  Vidya Praveen  <vidyapraveen@arm.com>
16831         Revert r251800 and r251799.
16833 2017-09-11  Martin Jambor  <mjambor@suse.cz>
16835         PR hsa/82119
16836         * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16837         arguments in advance.
16838         * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16839         use it to find predecessor edges.
16840         (naive_outof_ssa): Collect vector of predecessors.
16842 2017-09-08  Jason Merrill  <jason@redhat.com>
16844         PR c++/70029 - ICE with ref-qualifier and -flto
16845         * langhooks.h (struct lang_hooks_for_types): Add
16846         copy_lang_qualifiers.
16847         * attribs.c (build_type_attribute_qual_variant): Use it.
16848         * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16849         NULL.
16850         (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16851         * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16853 2017-09-08  Eric Botcazou  <ebotcazou@adacore.com>
16855         PR target/81988
16856         * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16857         (*mulsi3_sp64): New instruction.
16858         (mulsi3): New expander.
16860 2017-09-08  Uros Bizjak  <ubizjak@gmail.com>
16862         * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16864 2017-09-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16866         * sancov.c: Include memmodel.h.
16868 2017-09-07  Eric Botcazou  <ebotcazou@adacore.com>
16870         PR target/80897
16871         * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16872         large offsets.
16874 2017-09-07  Carl Love  <cel@us.ibm.com>
16876         * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16877         the sldi instruction.
16879 2017-09-07  David Edelsohn  <dje.gcc@gmail.com>
16881         * sancov.c: Include tm_p.h.
16883 2017-09-07  Jakub Jelinek  <jakub@redhat.com>
16885         PR target/81979
16886         * output.h (switch_to_other_text_partition): New declaration.
16887         * varasm.c (switch_to_other_text_partition): New function.
16888         * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16889         NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16890         (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16891         to the other text partition before emitting LCL label and switch back
16892         after emitting the word after it.
16894 2017-09-07  Richard Biener  <rguenther@suse.de>
16896         * passes.def (pass_split_crit_edges): Remove instance before PRE.
16897         * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16898         critical edges here, after loop init.
16899         (pass_data_pre): Remove PROP_no_crit_edges flags.
16900         * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16901         for valueization of call args to avoid leaking VN_TOP.
16902         (visit_use): Assert we do not visit default defs.
16903         (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16904         Use error_mark_node to more easily detect leaking VN_TOP.
16905         All default-defs are varying, not VN_TOP.  Mark them visited.
16906         (run_scc_vn): Make code match comment.
16908 2017-09-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
16910         * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16911         OPTION_MASK_FLOAT128_KEYWORD.
16912         (POWERPC_MASKS): Likewise.
16913         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
16914         support for the -mfloat128-type option, and make -mfloat128
16915         default on PowerPC Linux systems.  Define or undefine
16916         __FLOAT128__ and  __FLOAT128_HARDWARE__ for the current options.
16917         Define __float128 to be __ieee128 if IEEE 128-bit support is
16918         enabled, or undefine it.
16919         (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
16920         Delete defining __FLOAT128_TYPE__.
16921         * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
16922         -mfloat128-type option and make -mfloat128 default on PowerPC
16923         Linux systems.
16924         (TARGET_FLOAT128_TYPE): Likewise.
16925         (-mfloat128-type): Likewise.
16926         * config/rs6000/rs6000.c (rs6000_option_override_internal):
16927         Delete the -mfloat128-type option and make -mfloat128 default on
16928         PowerPC Linux systems.  Always use __ieee128 to be the keyword for
16929         the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
16930         128-bit floating point is enabled.  Change tests from using
16931         -mfloat128-type to -mfloat128.
16932         (rs6000_mangle_type): Use the correct mangling for the __float128
16933         type even if normal long double is restricted to 64-bits.
16934         (floatn_mode): Enable the _Float128 type by default on VSX Linux
16935         systems.
16936         * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
16937         (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
16938         (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
16939         -mfloat128-type.
16940         * doc/invoke.texi (RS/6000 and PowerPC Options): Update
16941         documentation for -mfloat128.
16943 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16945         * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
16947 2017-09-06  Wish Wu  <wishwu007@gmail.com>
16948             Jakub Jelinek  <jakub@redhat.com>
16950         * asan.c (initialize_sanitizer_builtins): Add
16951         BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
16952         BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
16953         BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
16954         BT_FN_VOID_UINT64_PTR variables.
16955         * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
16956         (BT_FN_VOID_UINT16_UINT16): Likewise.
16957         (BT_FN_VOID_UINT32_UINT32): Likewise.
16958         (BT_FN_VOID_FLOAT_FLOAT): Likewise.
16959         (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
16960         (BT_FN_VOID_UINT64_PTR): Likewise.
16961         * common.opt (flag_sanitize_coverage): New variable.
16962         (fsanitize-coverage=trace-pc): Remove.
16963         (fsanitize-coverage=): Add.
16964         * flag-types.h (enum sanitize_coverage_code): New enum.
16965         * fold-const.c (fold_range_test): Disable non-short-circuit
16966         optimization if flag_sanitize_coverage.
16967         (fold_truth_andor): Likewise.
16968         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16969         * opts.c (COVERAGE_SANITIZER_OPT): Define.
16970         (coverage_sanitizer_opts): New array.
16971         (get_closest_sanitizer_option): Add OPTS argument, handle also
16972         OPT_fsanitize_coverage_.
16973         (parse_sanitizer_options): Adjusted to also handle
16974         OPT_fsanitize_coverage_.
16975         (common_handle_option): Add OPT_fsanitize_coverage_.
16976         * sancov.c (instrument_comparison, instrument_switch): New function.
16977         (sancov_pass): Add trace-cmp support.
16978         * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
16979         BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
16980         BUILT_IN_SANITIZER_COV_TRACE_CMP8,
16981         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
16982         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
16983         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
16984         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
16985         BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
16986         BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
16987         * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
16989 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16991         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
16992         error.  Only quit immediately if parsing is complete.
16993         (BEGIN): Initialize fatal_err and parse_done.
16994         (begin fpu, end fpu): Check number of arguments.
16995         (begin arch, end arch): Likewise.
16996         (begin cpu, end cpu): Likewise.
16997         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
16998         (optalias): Likewise.
17000 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
17002         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
17003         * config/arm/arm-isa.h: Delete.  Move definitions to ...
17004         * arm-cpus.in: ... here.  Use new feature and fgroup values.
17005         * config/arm/arm.c (arm_option_override): Use lower case for feature
17006         bit names.
17007         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
17008         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
17009         * config/arm/parsecpu.awk (END): Add new command 'isa'.
17010         (isa_pfx): Delete.
17011         (print_isa_bits_for): New function.
17012         (gen_isa): New function.
17013         (gen_comm_data): Use print_isa_bits_for.
17014         (define feature): New keyword.
17015         (define fgroup): New keyword.
17016         * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
17017         (arm-isa.h): Add rule to generate file.
17018         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
17019         case for feature bit names.
17021 2017-09-06  Richard Biener  <rguenther@suse.de>
17023         * tree-ssa-pre.c (NECESSARY): Remove.
17024         (create_expression_by_pieces): Do not touch pass-local flags.
17025         (insert_into_preds_of_block): Likewise.
17026         (do_pre_regular_insertion): Likewise.
17027         (eliminate_insert): Likewise.
17028         (eliminate_dom_walker::before_dom_children): Likewise.
17029         (fini_eliminate): Do not look at inserted_exprs.
17030         (mark_operand_necessary): Remove.
17031         (remove_dead_inserted_code): Replace with simple work-list
17032         algorithm based on inserted_exprs and SSA uses.
17033         (pass_pre::execute): Re-order fini_eliminate and
17034         remove_dead_inserted_code.
17036 2017-09-06  Olivier Hainque  <hainque@adacore.com>
17038         * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
17039         for VxWorks 7.  Adjust surrounding comments.
17041 2017-09-06  Richard Biener  <rguenther@suse.de>
17043         * gimple-ssa-strength-reduction.c
17044         (find_candidates_dom_walker::before_dom_children): Also allow
17045         pointer types.
17047 2017-09-06  Richard Biener  <rguenther@suse.de>
17049         PR tree-optimization/82108
17050         * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
17051         for gap in the non-permutation SLP case.
17053 2017-09-06  Martin Jambor  <mjambor@suse.cz>
17055         PR tree-optimization/82078
17056         * tree-sra.c (sort_and_splice_var_accesses): Move call to
17057         add_access_to_work_queue...
17058         (build_accesses_from_assign): ...here.
17059         (propagate_all_subaccesses): Make sure racc is the group
17060         representative, if there is one.
17062 2017-09-06  Jakub Jelinek  <jakub@redhat.com>
17064         PR middle-end/82095
17065         * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
17066         NULL DECL_INITIAL.
17068 2017-09-06  Richard Biener  <rguenther@suse.de>
17070         * gimple-ssa-strength-reduction.c
17071         (find_candidates_dom_walker::before_doom_children): Use a
17072         type and not a mode check.
17074 2017-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17076         PR target/77308
17077         * config/arm/predicates.md (arm_general_adddi_operand): Create new
17078         non-vfp predicate.
17079         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
17081 2017-09-05  Jeff Law  <law@redhat.com>
17083         PR tree-optimization/64910
17084         * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
17085         cases where we have 3 or more operands.
17087 2017-09-05  Jakub Jelinek  <jakub@redhat.com>
17089         PR middle-end/81768
17090         * omp-low.c (lower_omp_for): Recompute tree invariant if
17091         gimple_omp_for_initial/final is ADDR_EXPR.
17093         PR middle-end/81768
17094         * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
17095         into gimple val before gimplification fo the COND_EXPR.
17097 2017-09-05  Aldy Hernandez  <aldyh@redhat.com>
17099         * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
17100         REGION_COPY argument.
17101         (thread_through_all_blocks): Remove unused argument to
17102         duplicate_thread_path.
17104 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17105             Alan Hayward  <alan.hayward@arm.com>
17106             David Sherwood  <david.sherwood@arm.com>
17108         * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
17109         Take a scalar_mode rather than a machine_mode.
17110         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17111         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
17112         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17113         (aarch64_gen_adjusted_ldpstp): Likewise.
17114         (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
17116 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17117             Alan Hayward  <alan.hayward@arm.com>
17118             David Sherwood  <david.sherwood@arm.com>
17120         * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
17121         Take a scalar_int_mode instead of a machine_mode.
17122         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17123         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17124         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17125         (aarch64_simd_attr_length_rglist): Delete.
17126         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
17127         a scalar_int_mode instead of a machine_mode.
17128         (aarch64_add_offset): Likewise.
17129         (aarch64_internal_mov_immediate): Likewise
17130         (aarch64_add_constant_internal): Likewise.
17131         (aarch64_add_constant): Likewise.
17132         (aarch64_movw_imm): Likewise.
17133         (aarch64_rtx_arith_op_extract_p): Likewise.
17134         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17135         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17136         Remove assert that the mode isn't a vector.
17137         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17138         (aarch64_expand_mov_immediate): Update calls after above changes.
17139         (aarch64_output_casesi): Use as_a <scalar_int_mode>.
17140         (aarch64_and_bitmask_imm): Check for scalar integer modes.
17141         (aarch64_move_imm): Likewise.
17142         (aarch64_can_const_movi_rtx_p): Likewise.
17143         (aarch64_strip_extend): Likewise.
17144         (aarch64_extr_rtx_p): Likewise.
17145         (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
17146         a CONST_INT when the mode parameter is VOIDmode.
17147         (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
17149 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17151         * machmode.h (bitwise_mode_for_mode): Return opt_mode.
17152         * stor-layout.c (bitwise_mode_for_mode): Likewise.
17153         (bitwise_type_for_mode): Update accordingly.
17155 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17157         * stor-layout.h (mode_for_size_tree): Return an opt_mode.
17158         * stor-layout.c (mode_for_size_tree): Likewise.
17159         (mode_for_array): Update accordingly.
17160         (layout_decl): Likewise.
17161         (compute_record_mode): Likewise.  Only set the mode once.
17163 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17165         * target.def (get_mask_mode): Change return type to opt_mode.
17166         Expand commentary.
17167         * doc/tm.texi: Regenerate.
17168         * targhooks.h (default_get_mask_mode): Return an opt_mode.
17169         * targhooks.c (default_get_mask_mode): Likewise.
17170         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17171         * optabs-query.c (can_vec_mask_load_store_p): Update use of
17172         targetm.get_mask_mode.
17173         * tree.c (build_truth_vector_type): Likewise.
17175 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17177         * machmode.h (mode_for_vector): Return an opt_mode.
17178         * stor-layout.c (mode_for_vector): Likewise.
17179         (mode_for_int_vector): Update accordingly.
17180         (layout_type): Likewise.
17181         * config/i386/i386.c (emit_memmov): Likewise.
17182         (ix86_expand_set_or_movmem): Likewise.
17183         (ix86_expand_vector_init): Likewise.
17184         (ix86_get_mask_mode): Likewise.
17185         * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
17186         Likewise.
17187         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
17188         * expmed.c (extract_bit_field_1): Likewise.
17189         * expr.c (expand_expr_real_2): Likewise.
17190         * optabs-query.c (can_vec_perm_p): Likewise.
17191         (can_vec_mask_load_store_p): Likewise.
17192         * optabs.c (expand_vec_perm): Likewise.
17193         * targhooks.c (default_get_mask_mode): Likewise.
17194         * tree-vect-stmts.c (vectorizable_store): Likewise.
17195         (vectorizable_load): Likewise.
17196         (get_vectype_for_scalar_type_and_size): Likewise.
17198 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17200         * machmode.h (mode_for_int_vector): New function.
17201         * stor-layout.c (mode_for_int_vector): Likewise.
17202         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
17203         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
17204         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
17205         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
17206         (s390_expand_vcond): Likewise.
17208 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17210         * machmode.h (opt_machine_mode): New type.
17211         (opt_mode<T>): Allow construction from anything that can be
17212         converted to a T.
17213         (is_a, as_a, dyn_cast): Add overloads for opt_mode.
17214         (mode_for_size): Return an opt_machine_mode.
17215         * stor-layout.c (mode_for_size): Likewise.
17216         (mode_for_size_tree): Update call accordingly.
17217         (bitwise_mode_for_mode): Likewise.
17218         (make_fract_type): Likewise.
17219         (make_accum_type): Likewise.
17220         * caller-save.c (replace_reg_with_saved_mem): Update call
17221         accordingly.
17222         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17223         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17224         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17225         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17226         * expmed.c (extract_bit_field_1): Likewise.
17227         * reload.c (get_secondary_mem): Likewise.
17228         * varasm.c (assemble_integer): Likewise.
17229         * lower-subreg.c (simplify_subreg_concatn): Likewise.  Move
17230         early-out.
17232 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17234         * machmode.h (decimal_float_mode_for_size): New function.
17235         * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
17236         (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
17237         (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17238         (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17239         (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17240         (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17242 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17244         * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17245         (get_builtin_sync_mode): Likewise.
17246         (expand_ifn_atomic_compare_exchange): Likewise.
17247         (expand_builtin_atomic_clear): Likewise.
17248         (expand_builtin_atomic_test_and_set): Likewise.
17249         (fold_builtin_atomic_always_lock_free): Likewise.
17250         * calls.c (compute_argument_addresses): Likewise.
17251         (emit_library_call_value_1): Likewise.
17252         (store_one_arg): Likewise.
17253         * combine.c (combine_instructions): Likewise.
17254         * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17255         * config/arm/arm.c (arm_function_value): Likewise.
17256         (aapcs_allocate_return_reg): Likewise.
17257         * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17258         * config/i386/i386.c (construct_container): Likewise.
17259         (ix86_gimplify_va_arg): Likewise.
17260         (ix86_expand_sse_cmp): Likewise.
17261         (emit_memmov): Likewise.
17262         (emit_memset): Likewise.
17263         (expand_small_movmem_or_setmem): Likewise.
17264         (ix86_expand_pextr): Likewise.
17265         (ix86_expand_pinsr): Likewise.
17266         * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17267         * config/microblaze/microblaze.c (microblaze_block_move_straight):
17268         Likewise.
17269         * config/mips/mips.c (mips_function_value_1) Likewise.
17270         (mips_block_move_straight): Likewise.
17271         (mips_expand_ins_as_unaligned_store): Likewise.
17272         * config/powerpcspe/powerpcspe.c
17273         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17274         (rs6000_darwin64_record_arg_flush): Likewise.
17275         * config/rs6000/rs6000.c
17276         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17277         (rs6000_darwin64_record_arg_flush): Likewise.
17278         * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17279         (sparc_function_value_1): Likewise.
17280         * config/spu/spu.c (adjust_operand): Likewise.
17281         (spu_emit_branch_or_set): Likewise.
17282         (arith_immediate_p): Likewise.
17283         * emit-rtl.c (gen_lowpart_common): Likewise.
17284         * expr.c (expand_expr_real_1): Likewise.
17285         * function.c (assign_parm_setup_block): Likewise.
17286         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17287         * reload1.c (alter_reg): Likewise.
17288         * stor-layout.c (mode_for_vector): Likewise.
17289         (layout_type): Likewise.
17291 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17293         * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17294         (spu_convert_move): Likewise.
17295         * lower-subreg.c (resolve_simple_move): Likewise.
17297 2017-09-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17299         PR target/81833
17300         * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17301         define_insn to a define_expand.
17302         (altivec_vsum2sws_direct): New define_insn.
17303         (altivec_vsumsws): Convert from a define_insn to a define_expand.
17305 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17307         * config/arm/arm.c (arm_option_params_internal): Improve setting of
17308         max_insns_skipped.
17310 2017-09-05  H.J. Lu  <hongjiu.lu@intel.com>
17312         PR target/59501
17313         PR target/81624
17314         PR target/81769
17315         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17316         realign stack if stack alignment needed is less than incoming
17317         stack boundary.
17319 2017-09-05  Marek Polacek  <polacek@redhat.com>
17321         PR sanitizer/82072
17322         * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17323         check earlier.
17325 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17327         * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17329 2017-09-05  Richard Biener  <rguenther@suse.de>
17331         PR tree-optimization/82084
17332         * fold-const.c (can_native_encode_string_p): Handle wide characters.
17334 2017-09-05  Richard Biener  <rguenther@suse.de>
17336         PR tree-optimization/82102
17337         * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17339 2017-09-05  Martin Liska  <mliska@suse.cz>
17341         PR tree-optimization/82032
17342         * tree-cfg.c (generate_range_test): New function.
17343         * tree-cfg.h (generate_range_test): Declared here.
17344         * tree-cfgcleanup.c (convert_single_case_switch): New function.
17345         (cleanup_control_expr_graph): Use it.
17346         * tree-switch-conversion.c (try_switch_expansion): Remove
17347         assert.
17348         (emit_case_nodes): Use generate_range_test.
17350 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17352         PR target/82098
17353         * config/i386/i386.md (*<btsc><mode>_mask): Add
17354         TARGET_USE_BT to insn constraint.
17355         (*btr<mode>_mask): Ditto.
17357 2017-09-04  Wilco Dijkstra  <wdijkstr@arm.com>
17359         * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17360         (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17362 2017-09-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17364         PR target/77308
17365         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17366         TARGET_NEON and TARGET_IWMMXT.
17367         (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17368         TARGET_NEON and TARGET_IWMMXT.
17369         (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17371 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17373         * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17374         (ix86_rewrite_tls_address): Ditto.
17375         * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17376         (ix86_rewrite_tls_address_1): Ditto.
17377         (ix86_rewrite_tls_address): Ditto.
17378         * config/i386/predicates.md (tls_address_pattern): New predicate.
17379         * config/i386/i386.md (TLS address splitter): New splitter.
17381 2017-09-04  Richard Biener  <rguenther@suse.de>
17383         PR tree-optimization/82084
17384         * fold-const.h (can_native_encode_string_p): Declare.
17385         * fold-const.c (can_native_encode_string_p): Factor out from ...
17386         (native_encode_string): ... here.
17387         * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17388         vectorizing stores from constants we later cannot handle.
17390 2017-09-04  Marek Polacek  <polacek@redhat.com>
17392         PR c/81783
17393         * doc/invoke.texi: Update -Wtautological-compare documentation.
17395 2017-09-04  Jeff Law  <law@redhat.com>
17397         PR tree-optimization/64910
17398         * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17399         swap the first and last operand if the last is a constant.
17401 2017-09-04  Marek Polacek  <polacek@redhat.com>
17403         PR sanitizer/82072
17404         * convert.c (do_narrow): When sanitizing signed integer overflows,
17405         bail out for signed types.
17406         (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17408 2017-09-04  Richard Biener  <rguenther@suse.de>
17410         PR tree-optimization/82060
17411         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17412         Move devirtualization after stmt folding and before EH/AB/noreturn
17413         cleanup to get the stmt refs canonicalized.  Use a bool instead
17414         of gimple_modified_p since that doesn't work for NOPs.  Schedule
17415         NOPs generated by folding for removal.
17417 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17418             Alan Hayward  <alan.hayward@arm.com>
17419             David Sherwood  <david.sherwood@arm.com>
17421         * coretypes.h (pad_direction): New enum.
17422         * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17423         (FUNCTION_ARG_PADDING): Likewise.
17424         * target.def (function_arg_padding): New hook.
17425         * targhooks.h (default_function_arg_padding): Declare.
17426         * targhooks.c (default_function_arg_padding): New function.
17427         * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17428         (TARGET_FUNCTION_ARG_PADDING): ...this.
17429         * doc/tm.texi: Regenerate.
17430         * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17431         instead of direction.
17432         (compute_argument_addresses): Likewise.
17433         (load_register_parameters): Likewise.
17434         (emit_library_call_value_1): Likewise.
17435         (store_one_arg): Use targetm.calls.function_arg_padding instead
17436         of FUNCTION_ARG_PADDING.
17437         (must_pass_in_stack_var_size_or_pad): Likewise.
17438         * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17439         (emit_group_store): Likewise.
17440         (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17441         instead of FUNCTION_ARG_PADDING.
17442         (emit_push_insn): Likewise, and propagate enum change throughout
17443         function.
17444         * function.h (direction): Delete.
17445         (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17446         of direction.
17447         * function.c (assign_parm_find_stack_rtl): Likewise.
17448         (assign_parm_setup_block_p): Likewise.
17449         (assign_parm_setup_block): Likewise.
17450         (gimplify_parameters): Likewise.
17451         (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17452         instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17453         function.
17454         * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17455         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17456         * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17457         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17458         (aarch64_function_arg_padding): ...this new function.
17459         (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17460         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17461         * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17462         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17463         * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17464         * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17465         (arm_pad_arg_upward): Replace with...
17466         (arm_function_arg_padding): ...this new function.
17467         * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17468         of direction.
17469         * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17470         * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17471         * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17472         (ia64_hpux_function_arg_padding): Replace with...
17473         (ia64_function_arg_padding): ...this new function.  Use pad_direction
17474         instead of direction.  Check for TARGET_HPUX.
17475         * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17476         * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17477         (iq2000_function_arg_padding): New function.
17478         * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17479         * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17480         (mips_function_arg_padding): ...this new function.
17481         (mips_pad_reg_upward): Update accordingly.
17482         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17483         * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17484         targetm.calls.function_arg_padding.
17485         (FUNCTION_ARG_PADDING): Delete.
17486         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17487         * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17488         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17489         * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17490         (nios2_block_reg_padding): Return pad_direction instead of direction.
17491         * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17492         instead of direction.
17493         (nios2_function_arg_padding): Likewise.  Make static.
17494         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17495         * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17496         (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17497         * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17498         * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17499         (pa_function_arg_padding): Make static.  Return pad_direction instead
17500         of direction.
17501         * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17502         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17503         * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17504         instead of direction.  Use targetm.calls.function_arg_padding.
17505         * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17506         * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17507         * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17508         * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17509         * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17510         Redefine.
17511         (function_arg_padding): Rename to...
17512         (rs6000_function_arg_padding): ...this.  Make static.  Return
17513         pad_direction instead of direction.
17514         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17515         * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17516         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17517         * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17518         instead of direction.  Use targetm.calls.function_arg_padding.
17519         * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17520         * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17521         * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17522         * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17523         * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17524         (function_arg_padding): Rename to...
17525         (rs6000_function_arg_padding): ...this.  Make static.  Return
17526         pad_direction instead of direction.
17527         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17528         * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17529         * config/s390/s390.c (s390_function_arg_padding): New function.
17530         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17531         * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17532         * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17533         * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17534         (function_arg_padding): Rename to...
17535         (sparc_function_arg_padding): ...this.  Make static.  Return
17536         pad_direction instead of direction.
17537         * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17538         * config/spu/spu.c (spu_function_arg_padding): New function.
17539         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17540         * system.h (FUNCTION_ARG_PADDING): Poison.
17542 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17543             Alan Hayward  <alan.hayward@arm.com>
17544             David Sherwood  <david.sherwood@arm.com>
17546         * target.def (modes_tieable_p): New hook.
17547         * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17548         (TARGET_MODES_TIEABLE_P): ...this.
17549         * doc/tm.texi.in: Regenerate.
17550         * hooks.h (hook_bool_mode_mode_true): Declare.
17551         * hooks.c (hook_bool_mode_mode_true): New function.
17552         * combine.c (subst): Use targetm.modes_tieable_p instead of
17553         MODES_TIEABLE_P.
17554         * dse.c (find_shift_sequence): Likewise.
17555         * expmed.c (extract_low_bits): Likewise.
17556         * lower-subreg.c: Include target.h.
17557         (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17558         MODES_TIEABLE_P.
17559         * rtlanal.c (rtx_cost): Likewise.
17560         * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17561         * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17562         * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17563         (TARGET_MODES_TIEABLE_P): Redefine.
17564         * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17565         * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17566         (TARGET_MODES_TIEABLE_P): Redefine.
17567         * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17568         * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17569         (arc_modes_tieable_p): New function.
17570         * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17571         * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17572         * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17573         (arm_modes_tieable_p): Make static.
17574         * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17575         * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17576         * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17577         (TARGET_MODES_TIEABLE_P): Redefine.
17578         * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17579         * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17580         (TARGET_MODES_TIEABLE_P): Redefine.
17581         * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17582         * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17583         (cr16_modes_tieable_p): New function.
17584         * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17585         * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17586         * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17587         (TRULY_NOOP_TRUNCATION): Update comment.
17588         * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17589         (TRULY_NOOP_TRUNCATION): Update comment.
17590         * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17591         (frv_modes_tieable_p): New function.
17592         * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17593         * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17594         * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17595         (TARGET_MODES_TIEABLE_P): Redefine.
17596         * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17597         * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17598         * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17599         (TARGET_MODES_TIEABLE_P): Redefine.
17600         * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17601         * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17602         (ia64_modes_tieable_p): New function.
17603         * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17604         * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17605         (iq2000_modes_tieable_p): New function.
17606         * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17607         * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17608         (lm32_modes_tieable_p): New function.
17609         * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17610         * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17611         * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17612         (TARGET_MODES_TIEABLE_P): Redefine.
17613         * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17614         * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17615         (m32r_modes_tieable_p): New function.
17616         * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17617         * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17618         (m68k_modes_tieable_p): New function.
17619         * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17620         * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17621         (mcore_modes_tieable_p): New function.
17622         * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17623         * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17624         function.
17625         (TARGET_MODES_TIEABLE_P): Redefine.
17626         * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17627         * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17628         * config/mips/mips.c (mips_modes_tieable_p): Make static.
17629         (TARGET_MODES_TIEABLE_P): Redefine.
17630         * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17631         * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17632         * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17633         * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17634         (mn10300_modes_tieable_p): ...this and make static.
17635         (TARGET_MODES_TIEABLE_P): Redefine.
17636         * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17637         * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17638         * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17639         * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17640         (msp430_modes_tieable_p): Make static.
17641         * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17642         * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17643         (TARGET_MODES_TIEABLE_P): Redefine.
17644         * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17645         * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17646         * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17647         (TARGET_MODES_TIEABLE_P): Redefine.
17648         * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17649         * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17650         * config/pa/pa.c (pa_modes_tieable_p): Make static.
17651         (TARGET_MODES_TIEABLE_P): Redefine.
17652         * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17653         * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17654         (pdp11_modes_tieable_p): New function.
17655         * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17656         * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17657         (rs6000_modes_tieable_p): New function.
17658         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17659         * config/powerpcspe/powerpcspe.md: Update comment.
17660         * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17661         * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17662         (TARGET_MODES_TIEABLE_P): Redefine.
17663         * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17664         * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17665         (rl78_modes_tieable_p): New function.
17666         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17667         * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17668         (rs6000_modes_tieable_p): New function.
17669         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17670         * config/rs6000/rs6000.md: Update comment.
17671         * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17672         * config/rx/rx.c (rx_modes_tieable_p): New function.
17673         (TARGET_MODES_TIEABLE_P): Redefine.
17674         * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17675         * config/s390/s390.c (s390_modes_tieable_p): New function.
17676         (TARGET_MODES_TIEABLE_P): Redefine.
17677         * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17678         * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17679         (sh_modes_tieable_p): New function.
17680         * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17681         * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17682         * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17683         (sparc_modes_tieable_p): Make static.
17684         * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17685         * config/spu/spu.c (spu_modes_tieable_p): New function.
17686         (TARGET_MODES_TIEABLE_P): Redefine.
17687         * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17688         * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17689         (TARGET_MODES_TIEABLE_P): Redefine.
17690         * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17691         * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17692         * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17693         * config/v850/v850.c (v850_modes_tieable_p): New function.
17694         (TARGET_MODES_TIEABLE_P): Redefine.
17695         * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17696         * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17697         * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17698         (visium_modes_tieable_p): New function.
17699         * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17700         * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17701         (xtensa_modes_tieable_p): New function.
17702         * system.h (MODES_TIEABLE_P): Poison.
17704 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17705             Alan Hayward  <alan.hayward@arm.com>
17706             David Sherwood  <david.sherwood@arm.com>
17708         * target.def (hard_regno_mode_ok): New hook.
17709         * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17710         (TARGET_HARD_REGNO_MODE_OK): ...this.
17711         * doc/tm.texi.in: Regenerate.
17712         * hooks.h (hook_bool_uint_mode_true): Declare.
17713         * hooks.c (hook_bool_uint_mode_true): New function.
17714         * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17715         HARD_REGNO_MODE_OK.
17716         * genpreds.c (write_insn_preds_c): Add an include of target.h.
17717         * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17718         instead of HARD_REGNO_MODE_OK.
17719         * caller-save.c: Include target.h.
17720         (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17721         HARD_REGNO_MODE_OK.
17722         * combine.c (can_combine_p): Likewise.
17723         (combinable_i3pat): Likewise.
17724         (can_change_dest_mode): Likewise.
17725         * expr.c (init_expr_target): Likewise.
17726         (convert_move): Likewise.
17727         (convert_modes): Likewise.
17728         * ira.c (setup_prohibited_class_mode_regs): Likewise.
17729         (setup_prohibited_mode_move_regs): Likewise.
17730         * ira.h (target_ira): Likewise.
17731         * lra-assigns.c (find_hard_regno_for_1): Likewise.
17732         * lra-constraints.c (process_alt_operands): Likewise.
17733         (split_reg): Likewise.
17734         * recog.c (peep2_find_free_register): Likewise.
17735         * ree.c (combine_reaching_defs): Likewise.
17736         * regcprop.c (maybe_mode_change): Likewise.
17737         * reginfo.c (init_reg_sets_1): Likewise.
17738         (choose_hard_reg_mode): Likewise.
17739         (simplifiable_subregs): Likewise.
17740         * regrename.c (check_new_reg_p): Likewise.
17741         * reload.c (find_valid_class): Likewise.
17742         (find_valid_class_1): Likewise.
17743         (reload_inner_reg_of_subreg): Likewise.
17744         (push_reload): Likewise.
17745         (combine_reloads): Likewise.
17746         (find_dummy_reload): Likewise.
17747         (find_reloads): Likewise.
17748         * reload1.c (find_reg): Likewise.
17749         (set_reload_reg): Likewise.
17750         (allocate_reload_reg): Likewise.
17751         (choose_reload_regs): Likewise.
17752         (reload_adjust_reg_for_temp): Likewise.
17753         * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17754         (simplify_subreg_regno): Likewise.
17755         * sel-sched.c (init_regs_for_mode): Likewise.
17756         * varasm.c (make_decl_rtl): Likewise.
17757         * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17758         (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17759         HARD_REGNO_MODE_OK.
17760         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17761         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17762         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17763         * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17764         * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17765         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17766         * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17767         (arc_mode_class): Delete.
17768         (HARD_REGNO_MODE_OK): Delete.
17769         * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17770         (arc_hard_regno_mode_ok): Rename old array to...
17771         (arc_hard_regno_mode_ok_modes): ...this.
17772         (arc_conditional_register_usage): Update accordingly.
17773         (arc_mode_class): Make static.
17774         (arc_hard_regno_mode_ok): New function.
17775         * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17776         * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17777         * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17778         (arm_hard_regno_mode_ok): Make static.
17779         * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17780         HARD_REGNO_MODE_OK.
17781         * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17782         * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17783         * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17784         return a bool.
17785         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17786         * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17787         * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17788         * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17789         (bfin_hard_regno_mode_ok): ...this.  Make static and return a bool.
17790         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17791         * config/bfin/predicates.md (valid_reg_operand): Use
17792         targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17793         * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17794         * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17795         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17796         * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17797         * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17798         * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17799         (cr16_hard_regno_mode_ok): Make static and return a bool.
17800         * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17801         * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17802         (cris_hard_regno_mode_ok): New function.
17803         * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17804         (epiphany_mode_class): Delete.
17805         (HARD_REGNO_MODE_OK): Delete.
17806         * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17807         * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17808         (hard_regno_mode_ok): Rename to...
17809         (epiphany_hard_regno_mode_ok): ...this.  Make static and return a bool.
17810         * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17811         * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17812         HARD_REGNO_MODE_OK.
17813         * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17814         * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17815         * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17816         (frv_hard_regno_mode_ok): Make static and return a bool.
17817         * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17818         HARD_REGNO_MODE_OK.
17819         * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17820         * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17821         * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17822         * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17823         and return a bool.
17824         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17825         * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17826         * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17827         * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17828         return a bool.
17829         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17830         * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17831         * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17832         (ia64_hard_regno_mode_ok): New function.
17833         * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17834         * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17835         (iq2000_hard_regno_mode_ok): New function.
17836         * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17837         * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17838         (lm32_hard_regno_mode_ok): New function.
17839         * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17840         * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17841         * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17842         instead of HARD_REGNO_MODE_OK.
17843         (m32c_hard_regno_ok): Rename to...
17844         (m32c_hard_regno_mode_ok): ...this.  Make static and return a bool.
17845         (m32c_cannot_change_mode_class): Update accordingly.
17846         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17847         * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17848         (m32r_mode_class): Delete.
17849         (HARD_REGNO_MODE_OK): Delete.
17850         * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17851         (m32r_hard_regno_mode_ok): Rename to...
17852         (m32r_hard_regno_modes): ...this.
17853         (m32r_mode_class): Make static.
17854         (m32r_hard_regno_mode_ok): New function.
17855         * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17856         * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17857         * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17858         (m68k_hard_regno_mode_ok): Make static.
17859         * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17860         * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17861         (mcore_hard_regno_mode_ok): New function.
17862         * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17863         (HARD_REGNO_MODE_OK): Delete.
17864         * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17865         Rename to...
17866         (microblaze_hard_regno_mode_ok_p): ...this and make static.
17867         (microblaze_hard_regno_mode_ok): New function.
17868         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17869         * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17870         (mips_hard_regno_mode_ok): Delete.
17871         * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17872         (mips_hard_regno_mode_ok_p): ...this and make static.
17873         (mips_hard_regno_mode_ok_p): Rename to...
17874         (mips_hard_regno_mode_ok_uncached): ...this.
17875         (mips_hard_regno_mode_ok): New function.
17876         (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17877         of HARD_REGNO_MODE_OK.
17878         (mips_option_override): Update after above name changes.
17879         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17880         * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17881         * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17882         * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17883         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17884         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17885         * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17886         * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17887         * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17888         * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17889         (msp430_hard_regno_mode_ok): Make static and return a bool.
17890         * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17891         * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17892         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17893         and return a bool.
17894         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17895         * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17896         * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17897         * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17898         * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17899         (PA_HARD_REGNO_MODE_OK): ...this
17900         * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17901         (PA_HARD_REGNO_MODE_OK): ...this.
17902         * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17903         (pa_hard_regno_mode_ok): New function.
17904         * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17905         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17906         (pdp11_hard_regno_mode_ok): New function.
17907         * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17908         * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17909         Delete.
17910         * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17911         Make static.
17912         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17913         (rs6000_hard_regno_mode_ok): Rename to...
17914         (rs6000_hard_regno_mode_ok_uncached): ...this.
17915         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17916         (rs6000_hard_regno_mode_ok): New function.
17917         * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
17918         * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
17919         * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
17920         (riscv_hard_regno_mode_ok): ...this and make static.
17921         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17922         * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
17923         * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
17924         * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17925         (rl78_hard_regno_mode_ok): Make static and return bool.
17926         * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
17927         * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
17928         Delete.
17929         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
17930         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17931         (rs6000_hard_regno_mode_ok): Rename to...
17932         (rs6000_hard_regno_mode_ok_uncached): ...this.
17933         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17934         (rs6000_hard_regno_mode_ok): New function.
17935         * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
17936         * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
17937         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17938         * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
17939         * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
17940         * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
17941         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17942         * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
17943         * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
17944         * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17945         (sh_hard_regno_mode_ok): Make static.
17946         * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
17947         instead of HARD_REGNO_MODE_OK.
17948         * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
17949         (sparc_mode_class): Delete.
17950         (HARD_REGNO_MODE_OK): Delete.
17951         * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17952         (hard_regno_mode_classes): Make static.
17953         (sparc_mode_class): Likewise.
17954         (sparc_hard_regno_mode_ok): New function.
17955         * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
17956         * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
17957         * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
17958         function.
17959         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17960         * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
17961         * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
17962         * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
17963         * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
17964         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17965         * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
17966         * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
17967         * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17968         (visium_hard_regno_mode_ok): New function.
17969         * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
17970         instead of HARD_REGNO_MODE_OK.
17971         * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
17972         (HARD_REGNO_MODE_OK): Delete.
17973         * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
17974         (xtensa_hard_regno_mode_ok_p): ...this and make static.
17975         (xtensa_option_override): Update accordingly.
17976         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17977         (xtensa_hard_regno_mode_ok): New function.
17978         * system.h (HARD_REGNO_MODE_OK): Poison.
17980 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17981             Alan Hayward  <alan.hayward@arm.com>
17982             David Sherwood  <david.sherwood@arm.com>
17984         * target.def (hard_regno_call_part_clobbered): New hook.
17985         * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
17986         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
17987         * doc/tm.texi: Regenerate.
17988         * hooks.h (hook_bool_uint_mode_false): Declare.
17989         * hooks.c (hook_bool_uint_mode_false): New function.
17990         * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17991         * cselib.c (cselib_process_insn): Use
17992         targetm.hard_regno_call_part_clobbered instead of
17993         HARD_REGNO_CALL_PART_CLOBBERED.
17994         * ira-conflicts.c (ira_build_conflicts): Likewise.
17995         * ira-costs.c (ira_tune_allocno_costs): Likewise.
17996         * lra-constraints.c (need_for_call_save_p): Likewise.
17997         * lra-lives.c: Include target.h.
17998         (check_pseudos_live_through_calls): Use
17999         targetm.hard_regno_call_part_clobbered instead of
18000         HARD_REGNO_CALL_PART_CLOBBERED.
18001         * regcprop.c: Include target.h.
18002         (copyprop_hardreg_forward_1): Use
18003         targetm.hard_regno_call_part_clobbered instead of
18004         HARD_REGNO_CALL_PART_CLOBBERED.
18005         * reginfo.c (choose_hard_reg_mode): Likewise.
18006         * regrename.c (check_new_reg_p): Likewise.
18007         * reload.c (find_equiv_reg): Likewise.
18008         * reload1.c (emit_reload_insns): Likewise.
18009         * sched-deps.c (deps_analyze_insn): Likewise.
18010         * sel-sched.c (init_regs_for_mode): Likewise.
18011         (mark_unavailable_hard_regs): Likewise.
18012         * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
18013         * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18014         * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
18015         New function.
18016         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18017         * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18018         * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
18019         Delete.
18020         * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
18021         and return a bool.
18022         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18023         * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18024         * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
18025         function.
18026         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18027         * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18028         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
18029         function.
18030         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18031         * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
18032         Delete.
18033         * config/powerpcspe/powerpcspe.c
18034         (rs6000_hard_regno_call_part_clobbered): New function.
18035         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18036         * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18037         * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
18038         New function.
18039         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18040         * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18041         * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
18042         function.
18043         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18044         * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18045         * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
18047 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18048             Alan Hayward  <alan.hayward@arm.com>
18049             David Sherwood  <david.sherwood@arm.com>
18051         * rtl.h (subreg_memory_offset): Declare.
18052         * emit-rtl.c (subreg_memory_offset): New function.
18053         * expmed.c (store_bit_field_1): Use it.
18054         * expr.c (undefined_operand_subword_p): Likewise.
18055         * simplify-rtx.c (simplify_subreg): Likewise.
18057 2017-09-04  Alexander Monakov  <amonakov@ispras.ru>
18059         PR rtl-optimization/57448
18060         PR target/67458
18061         PR target/81316
18062         * optabs.c (expand_atomic_load): Place compiler memory barriers if
18063         using atomic_load pattern.
18064         (expand_atomic_store): Likewise.
18066 2017-09-04  Jakub Jelinek  <jakub@redhat.com>
18068         PR sanitizer/81981
18069         * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
18070         and UBSAN_BOUNDS internal calls.  Clean up IFN_UBSAN_OBJECT_SIZE
18071         handling.  Use replace_call_with_value with NULL instead of
18072         gsi_replace, unlink_stmt_vdef and release_defs.
18074         * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
18075         instead of tab.
18077         * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
18079 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18081         PR bootstrap/82045
18082         * rtl.h (emit_library_call_value_1): Declare.
18083         (emit_library_call): Replace declaration with a series of overloads.
18084         Remove the parameter count argument.
18085         (emit_library_call_value): Likewise.
18086         * calls.c (emit_library_call_value_1): Make global.  Replace varargs
18087         with an "rtx_mode_t *".
18088         (emit_library_call_value): Delete.
18089         (emit_library_call): Likewise.
18090         * asan.c (asan_emit_stack_protection): Update calls accordingly.
18091         (asan_emit_allocas_unpoison): Likewise.
18092         * builtins.c (expand_builtin_powi): Likewise.
18093         (expand_asan_emit_allocas_unpoison): Likewise.
18094         * cfgexpand.c (expand_main_function): Likewise.
18095         * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
18096         * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
18097         * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
18098         * config/arm/arm.c (arm_trampoline_init): Likewise.
18099         (arm_call_tls_get_addr): Likewise.
18100         (arm_expand_divmod_libfunc): Likewise.
18101         * config/bfin/bfin.md (umulsi3_highpart): Likewise.
18102         (smulsi3_highpart): Likewise.
18103         * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
18104         (c6x_expand_compare): Likewise.
18105         (c6x_expand_movmem): Likewise.
18106         * config/frv/frv.c (frv_trampoline_init): Likewise.
18107         * config/i386/i386.c (ix86_trampoline_init): Likewise.
18108         (ix86_expand_divmod_libfunc): Likewise.
18109         * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
18110         (ia64_expand_compare): Likewise.
18111         (ia64_profile_hook): Likewise.
18112         * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
18113         (nonlocal_goto): Likewise.
18114         (restore_stack_nonlocal): Likewise.
18115         * config/m32r/m32r.c (block_move_call): Likewise.
18116         (m32r_trampoline_init): Likewise.
18117         * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
18118         * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
18119         (m68k_call_m68k_read_tp): Likewise.
18120         * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
18121         (microblaze_expand_divide): Likewise.
18122         * config/mips/mips.h (mips_args): Likewise.
18123         * config/mips/sdemtk.h (mips_sync_icache): Likewise.
18124         (MIPS_ICACHE_SYNC): Likewise.
18125         * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
18126         (nios2_trampoline_init): Likewise.
18127         * config/pa/pa.c (hppa_tls_call): Likewise.
18128         (pa_trampoline_init): Likewise.
18129         * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
18130         * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
18131         (expand_strn_compare): Likewise.
18132         (rs6000_generate_compare): Likewise.
18133         (rs6000_expand_float128_convert): Likewise.
18134         (output_profile_hook): Likewise.
18135         (rs6000_trampoline_init): Likewise.
18136         * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
18137         * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
18138         * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
18139         * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
18140         (rs6000_generate_compare): Likewise.
18141         (rs6000_expand_float128_convert): Likewise.
18142         (output_profile_hook): Likewise.
18143         (rs6000_trampoline_init): Likewise.
18144         * config/rs6000/rs6000.md (neg<mode>2): Likewise.
18145         * config/sh/sh.c (sh_trampoline_init): Likewise.
18146         * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
18147         (sparc_emit_float_lib_cmp): Likewise.
18148         (sparc32_initialize_trampoline): Likewise.
18149         (sparc64_initialize_trampoline): Likewise.
18150         (sparc_profile_hook): Likewise.
18151         * config/spu/spu.c (ea_load_store): Likewise.
18152         * config/spu/spu.md (floatunssidf2): Likewise.
18153         * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
18154         * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
18155         * config/visium/visium.c (expand_block_move_4): Likewise.
18156         (expand_block_move_2): Likewise.
18157         (expand_block_move_1): Likewise.
18158         (expand_block_set_4): Likewise.
18159         (expand_block_set_2): Likewise.
18160         (expand_block_set_1): Likewise.
18161         (visium_trampoline_init): Likewise.
18162         (visium_profile_hook): Likewise.
18163         * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
18164         (xtensa_setup_frame_addresses): Likewise.
18165         (xtensa_trampoline_init): Likewise.
18166         * except.c (sjlj_emit_function_enter): Likewise.
18167         (sjlj_emit_function_exit): Likewise.
18168         * explow.c (allocate_dynamic_stack_space): Likewise.
18169         (probe_stack_range): Likewise.
18170         * expr.c (convert_mode_scalar): Likewise.
18171         * optabs.c (expand_binop): Likewise.
18172         (expand_twoval_binop_libfunc): Likewise.
18173         (expand_unop): Likewise.
18174         (prepare_cmp_insn): Likewise.
18175         (prepare_float_lib_cmp): Likewise.
18176         (expand_float): Likewise.
18177         (expand_fix): Likewise.
18178         (expand_fixed_convert): Likewise.
18179         (maybe_emit_sync_lock_test_and_set): Likewise.
18180         (expand_atomic_compare_and_swap): Likewise.
18181         (expand_mem_thread_fence): Likewise.
18182         (expand_atomic_fetch_op): Likewise.
18184 2017-09-03  Gerald Pfeifer  <gerald@pfeifer.com>
18186         * doc/generic.texi (OpenACC): Adjust URL.
18187         * doc/invoke.texi (C Dialect Options): Ditto.
18189 2017-09-03  Uros Bizjak  <ubizjak@gmail.com>
18191         * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
18192         predicate for operand 1.  Add (m,<S>) constraint.
18193         (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
18194         Prevent memory operand 1 with register operand 2.
18196 2017-09-01  Segher Boessenkool  <segher@kernel.crashing.org>
18198         PR rtl-optimization/82024
18199         * combine.c (try_combine): If the combination result is a PARALLEL,
18200         and we only need to retain the SET in there that would be placed
18201         at I2, check that we can place that at I3 instead, before doing so.
18203 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18205         PR target/81766
18206         * config/i386/i386.c (ix86_init_large_pic_reg): Return label
18207         instead of void.
18208         (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
18209         if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
18210         and label.
18212 2017-09-01  Joerg Sonnenberger  <joerg@bec.de>
18213             Jeff Law  <law@redhat.com>
18215         * varasm.c (bss_initializer_p): Do not put constants into .bss
18216         (categorize_decl_for_section): Handle bss_initializer_p returning
18217         false when DECL_INITIAL is NULL.
18219 2017-09-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18221         PR target/82012
18222         * config/s390/s390.c (s390_can_inline_p): New function.
18224 2017-09-01  Jeff Law  <law@redhat.com>
18226         PR tree-optimization/82052
18227         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
18228         Always initialize the returned slot after a hash table miss
18229         when INSERT is true.
18231 2017-09-01  Alexander Monakov  <amonakov@ispras.ru>
18233         * config/s390/s390.md (mem_signal_fence): Remove.
18234         * doc/md.texi (mem_signal_fence): Remove.
18235         * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
18236         Update comments.
18237         * target-insns.def (mem_signal_fence): Remove.
18239 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18241         PR sanitizer/81902
18242         * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18244         PR sanitizer/81923
18245         * asan.c (create_odr_indicator): Strip name encoding from assembler
18246         name before appending it after __odr_asan_.
18248 2017-09-01  Martin Liska  <mliska@suse.cz>
18250         PR tree-optimization/82059
18251         * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18252         frequency only when an edge is redirected.
18254 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18256         * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18257         * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18258         (arc_conditional_register_usage): Remove ARC600 lp_count
18259         exception.
18260         (arc_file_start): Emit Tag_ARC_CPU_variation.
18261         (arc_can_use_doloop_p): New conditions to use ZOLs.
18262         (hwloop_fail): New function.
18263         (hwloop_optimize): Likewise.
18264         (hwloop_pattern_reg): Likewise.
18265         (arc_doloop_hooks): New struct, to be used with reorg_loops.
18266         (arc_reorg_loops): New function, calls reorg_loops.
18267         (arc_reorg): Call arc_reorg_loops.  Remove old ZOL handling.
18268         (arc600_corereg_hazard): Remove ZOL checking, case handled by
18269         hwloop_optimize.
18270         (arc_loop_hazard): Remove function, functionality moved into
18271         hwloop_optimize.
18272         (arc_hazard): Remove arc_loop_hazard call.
18273         (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18274         into hwloop_optimize.
18275         (arc_label_align): Remove ZOL handling.
18276         * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18277         * config/arc/arc.md (doloop_begin): Remove pattern.
18278         (doloop_begin_i): Likewise.
18279         (doloop_end_i): Likewise.
18280         (doloop_fallback): Likewise.
18281         (doloop_fallback_m): Likewise.
18282         (doloop_end): Reimplement expand.
18283         (arc_lp): New pattern for LP instruction.
18284         (loop_end): New pattern.
18285         (loop_fail): Likewise.
18286         (decrement_and_branch_until_zero): Likewise.
18287         * config/arc/arc.opt (mlpc-width): New option.
18288         * doc/invoke.texi (mlpc-width): Document option.
18290 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18292         * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18293         (arc_ccfsm_advance): Fix checking for delay slots.
18294         (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18296 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18298         * config/arc/arc.md (movqi_insn): Add stores to save constant long
18299         immediates.
18300         (movhi_insn): Update store instruction constraint which are saving
18301         6-bit short immediates.
18302         (movsi_insn): Consider also short scaled load operations.
18303         (zero_extendhisi2_i): Use Usd constraint instead of T.
18304         (extendhisi2_i): Add q constraint.
18305         (arc_clzsi2): Add type and length attributes.
18306         (arc_ctzsi2): Likewise.
18307         * config/arc/constraints.md (Usc): Update constraint, the
18308         assembler can parse two relocations for a single instruction.
18310 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18312         * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18313         (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18315 2017-08-31  Olivier Hainque  <hainque@adacore.com>
18317         * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18318         match as powerpc-wrs-vxworks*.
18320 2017-08-31  James Greenhalgh  <james.greenhalgh@arm.com>
18322         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18323         register constraint for by-element operand.
18324         (aarch64_mls_elt_merge<mode>): Likewise.
18326 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18328         * config/arc/arc.c (arc_can_follow_jump): Check for short
18329         branches.
18331 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18333         * config.gcc: Use g.opt for arc.
18334         * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18335         functionality moved to ...
18336         (legitimate_scaled_address_p): New function, ...here.
18337         (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18338         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18339         (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18340         condition.
18341         (arc_override_options): Handle G option.
18342         (arc_output_pic_addr_const): Correct function definition.
18343         (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18344         (arc_decl_anon_ns_mem_p): Delete.
18345         (arc_in_small_data_p): Overhaul this function to take into
18346         consideration the value given via G option.
18347         (arc_rewrite_small_data_1): Renamed and corrected old
18348         arc_rewrite_small_data function.
18349         (arc_rewrite_small_data): New function.
18350         (small_data_pattern): Don't use pic_offset_table_rtx.
18351         * config/arc/arc.h (CC1_SPEC): Recognize G option.
18352         * config/arc/simdext.md (movmisalignv2hi): Use
18353         prepare_move_operands function.
18354         (mov*): Likewise.
18355         (movmisalign*): Likewise.
18356         * doc/invoke.texi (ARC options): Document -G option.
18358 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18360         * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18361         prototype.
18362         * config/arc/arc.c (arc_print_operand): Output scalled address for
18363         sdata whenever is possible.
18364         (arc_in_small_data_p): Allow sdata for 64bit datum when double
18365         load/stores are available.
18366         (compact_sda_memory_operand): Check for the alignment required by
18367         code density instructions.
18368         * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18369         constraint.
18370         * config/arc/constraints.md (Usd): Update constraint.
18371         (Us0): New constraint.
18372         (Usc): Update constraint.
18374 2017-08-31  Richard Biener  <rguenther@suse.de>
18376         PR middle-end/82054
18377         * dwarf2out.c (dwarf2out_early_global_decl): Process each
18378         function only once.
18380 2017-08-31  Tamar Christina  <tamar.christina@arm.com>
18382         * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18383         Resize type_signature.
18385 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18386             Alan Hayward  <alan.hayward@arm.com>
18387             David Sherwood  <david.sherwood@arm.com>
18389         * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18390         subregs whose inner modes can be stored in GPRs.
18391         (aarch64_classify_index): Likewise.
18393 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18394             Alan Hayward  <alan.hayward@arm.com>
18395             David Sherwood  <david.sherwood@arm.com>
18397         * config/aarch64/iterators.md (V_cmp_result): Rename to...
18398         (V_INT_EQUIV): ...this.
18399         (v_cmp_result): Rename to...
18400         (v_int_equiv): ...this.
18401         * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18402         * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18403         (copysign<mode>3): Likewise.
18404         (aarch64_simd_bsl<mode>_internal): Likewise.
18405         (aarch64_simd_bsl<mode>): Likewise.
18406         (vec_cmp<mode><mode>): Likewise.
18407         (vcond<mode><mode>): Likewise.
18408         (vcond<v_cmp_mixed><mode>): Likewise.
18409         (vcondu<mode><v_cmp_mixed>): Likewise.
18410         (aarch64_cm<optab><mode>): Likewise.
18411         (aarch64_cmtst<mode>): Likewise.
18412         (aarch64_fac<optab><mode>): Likewise.
18413         (vec_perm_const<mode>): Likewise.
18414         (vcond_mask_<mode><v_cmp_result>): Rename to...
18415         (vcond_mask_<mode><v_int_equiv>): ...this.
18416         (vec_cmp<mode><v_cmp_result>): Rename to...
18417         (vec_cmp<mode><v_int_equiv>): ...this.
18419 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18420             Alan Hayward  <alan.hayward@arm.com>
18421             David Sherwood  <david.sherwood@arm.com>
18423         * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18424         vector modes.
18425         * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18426         * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18427         (UNSPEC_LD4_DREG): New unspecs.
18428         * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18429         (aarch64_ld2<mode>_dreg_be): Replace with...
18430         (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18431         unspec.
18432         (aarch64_ld3<mode>_dreg_le)
18433         (aarch64_ld3<mode>_dreg_be): Replace with...
18434         (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18435         unspec.
18436         (aarch64_ld4<mode>_dreg_le)
18437         (aarch64_ld4<mode>_dreg_be): Replace with...
18438         (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18439         unspec.
18441 2017-08-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18443         PR tree-optimization/81987
18444         * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18445         insert an initializer in a location not dominated by the stride
18446         definition.
18448 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18450         * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18451         on the entire header of the finally block in the fallthru case.
18453 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18455         * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18457 2017-08-30  Pat Haugen  <pthaugen@us.ibm.com>
18459         * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18460         rs6000_emit_move_from_cr and call renamed function.
18461         (rs6000_emit_prologue): Call renamed functions.
18462         * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18463         movesi_from_cr, remove volatile CRs.
18465 2017-08-30  Jon Beniston  <jon@beniston.com>
18466             Richard Biener  <rguenther@suse.de>
18468         * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18469         of VECTOR_MODE_P check.
18470         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18471         element vector types.
18473 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18475         * df.h (df_read_modify_subreg_p): Remove in favor of...
18476         * rtl.h (read_modify_subreg_p): ...this new function.  Take a
18477         const_rtx instead of an rtx.
18478         * cprop.c (local_cprop_find_used_regs): Update accordingly.
18479         * df-problems.c (df_word_lr_mark_ref): Likewise.
18480         * ira-lives.c (mark_pseudo_reg_live): Likewise.
18481         (mark_pseudo_reg_dead): Likewise.
18482         (mark_ref_dead): Likewise.
18483         * reginfo.c (init_subregs_of_mode): Likewise.
18484         * sched-deps.c (sched_analyze_1): Likewise.
18485         * df-scan.c (df_def_record_1): Likewise.
18486         (df_uses_record): Likewise.
18487         (df_read_modify_subreg_p): Remove in favor of...
18488         * rtlanal.c (read_modify_subreg_p): ...this new function.  Take a
18489         const_rtx instead of an rtx.
18491 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18492             Alan Hayward  <alan.hayward@arm.com>
18493             David Sherwood  <david.sherwood@arm.com>
18495         * rtl.h (partial_subreg_p): New function.
18496         * caller-save.c (save_call_clobbered_regs): Use it.
18497         * calls.c (expand_call): Likewise.
18498         * combine.c (combinable_i3pat): Likewise.
18499         (simplify_set): Likewise.
18500         (make_extraction): Likewise.
18501         (make_compound_operation_int): Likewise.
18502         (gen_lowpart_or_truncate): Likewise.
18503         (force_to_mode): Likewise.
18504         (make_field_assignment): Likewise.
18505         (reg_truncated_to_mode): Likewise.
18506         (record_truncated_value): Likewise.
18507         (move_deaths): Likewise.
18508         * cse.c (record_jump_cond): Likewise.
18509         (cse_insn): Likewise.
18510         * cselib.c (cselib_lookup_1): Likewise.
18511         * expmed.c (extract_bit_field_using_extv): Likewise.
18512         * function.c (assign_parm_setup_reg): Likewise.
18513         * ifcvt.c (noce_convert_multiple_sets): Likewise.
18514         * ira-build.c (create_insn_allocnos): Likewise.
18515         * lra-coalesce.c (merge_pseudos): Likewise.
18516         * lra-constraints.c (match_reload): Likewise.
18517         (simplify_operand_subreg): Likewise.
18518         (curr_insn_transform): Likewise.
18519         * lra-lives.c (process_bb_lives): Likewise.
18520         * lra.c (new_insn_reg): Likewise.
18521         (lra_substitute_pseudo): Likewise.
18522         * regcprop.c (mode_change_ok): Likewise.
18523         (maybe_mode_change): Likewise.
18524         (copyprop_hardreg_forward_1): Likewise.
18525         * reload.c (push_reload): Likewise.
18526         (find_reloads): Likewise.
18527         (find_reloads_subreg_address): Likewise.
18528         * reload1.c (alter_reg): Likewise.
18529         (eliminate_regs_1): Likewise.
18530         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18532 2017-08-30  David Edelsohn  <dje.gcc@gmail.com>
18534         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18535         back to if statements, including unpack.
18537 2017-08-30  Martin Liska  <mliska@suse.cz>
18539         PR inline-asm/82001
18540         * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18541         Rename to ...
18542         (func_checker::compare_asm_inputs_outputs): ... this function.
18543         (func_checker::compare_gimple_asm): Use the function to compare
18544         also ASM constrains.
18545         * ipa-icf-gimple.h: Rename the function.
18547 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18548             Alan Hayward  <alan.hayward@arm.com>
18549             David Sherwood  <david.sherwood@arm.com>
18551         * coretypes.h (complex_mode): New type.
18552         * gdbhooks.py (build_pretty_printer): Handle it.
18553         * machmode.h (complex_mode): New class.
18554         (complex_mode::includes_p): New function.
18555         (is_complex_int_mode): Likewise.
18556         (is_complex_float_mode): Likewise.
18557         * genmodes.c (get_mode_class): Handle complex mode classes.
18558         * function.c (expand_function_end): Use is_complex_int_mode.
18560 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18561             Alan Hayward  <alan.hayward@arm.com>
18562             David Sherwood  <david.sherwood@arm.com>
18564         * coretypes.h (scalar_mode_pod): New typedef.
18565         * gdbhooks.py (build_pretty_printer): Handle it.
18566         * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18567         * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18568         * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18569         * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18570         as_a <scalar_mode>.
18572 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18573             Alan Hayward  <alan.hayward@arm.com>
18574             David Sherwood  <david.sherwood@arm.com>
18576         * machmode.h (mode_for_vector): Take a scalar_mode instead
18577         of a machine_mode.
18578         * stor-layout.c (mode_for_vector): Likewise.
18579         * explow.c (promote_mode): Use as_a <scalar_mode>.
18580         * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
18582 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18583             Alan Hayward  <alan.hayward@arm.com>
18584             David Sherwood  <david.sherwood@arm.com>
18586         * target.def (preferred_simd_mode): Take a scalar_mode
18587         instead of a machine_mode.
18588         * targhooks.h (default_preferred_simd_mode): Likewise.
18589         * targhooks.c (default_preferred_simd_mode): Likewise.
18590         * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18591         * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18592         * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18593         * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18594         * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18595         * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18596         * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18597         * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18598         Likewise.
18599         * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18600         * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18601         * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18602         * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18603         (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18604         * doc/tm.texi: Regenerate.
18605         * optabs-query.c (can_vec_mask_load_store_p): Return false for
18606         non-scalar modes.
18608 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18609             Alan Hayward  <alan.hayward@arm.com>
18610             David Sherwood  <david.sherwood@arm.com>
18612         * target.def (scalar_mode_supported_p): Take a scalar_mode
18613         instead of a machine_mode.
18614         * targhooks.h (default_scalar_mode_supported_p): Likewise.
18615         * targhooks.c (default_scalar_mode_supported_p): Likewise.
18616         * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18617         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18618         * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18619         * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18620         * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18621         * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18622         * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18623         * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18624         * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18625         * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18626         * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18627         * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18628         Likewise.
18629         * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18630         * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18631         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18632         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18633         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18634         Likewise.
18635         * doc/tm.texi: Regenerate.
18637 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18638             Alan Hayward  <alan.hayward@arm.com>
18639             David Sherwood  <david.sherwood@arm.com>
18641         * coretypes.h (opt_scalar_mode): New typedef.
18642         * gdbhooks.py (build_pretty_printers): Handle it.
18643         * machmode.h (mode_iterator::get_2xwider): Add overload for
18644         opt_mode<T>.
18645         * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18646         over scalar modes.
18647         * expr.c (convert_mode_scalar): Likewise.
18648         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18649         * optabs.c (expand_float): Likewise.
18650         (expand_fix): Likewise.
18651         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18653 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18654             Alan Hayward  <alan.hayward@arm.com>
18655             David Sherwood  <david.sherwood@arm.com>
18657         * optabs.c (expand_float): Explicitly check for scalars before
18658         using a branching expansion.
18659         (expand_fix): Likewise.
18661 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18662             Alan Hayward  <alan.hayward@arm.com>
18663             David Sherwood  <david.sherwood@arm.com>
18665         * expr.c (convert_mode): Split scalar handling out into...
18666         (convert_mode_scalar): ...this new function.  Treat the modes
18667         as scalar_modes.
18669 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18670             Alan Hayward  <alan.hayward@arm.com>
18671             David Sherwood  <david.sherwood@arm.com>
18673         * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18674         and scalar_mode.
18675         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18677 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18678             Alan Hayward  <alan.hayward@arm.com>
18679             David Sherwood  <david.sherwood@arm.com>
18681         * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18682         rather than a machine_mode.
18683         (fixed_from_string): Likewise.
18684         (fixed_convert): Likewise.
18685         (fixed_convert_from_int): Likewise.
18686         (fixed_convert_from_real): Likewise.
18687         (real_convert_from_fixed): Likewise.
18688         * fixed-value.c (fixed_from_double_int): Likewise.
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         * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
18696 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18697             Alan Hayward  <alan.hayward@arm.com>
18698             David Sherwood  <david.sherwood@arm.com>
18700         * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18701         of separate mode class checks.  Do not allow vector modes here.
18702         (immed_wide_int_const): Use as_a <scalar_mode>.
18703         * explow.c (trunc_int_for_mode): Likewise.
18704         * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18705         (wi::shwi): Likewise.
18706         (wi::min_value): Likewise.
18707         (wi::max_value): Likewise.
18708         * dwarf2out.c (loc_descriptor): Likewise.
18709         * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18710         for CONST_WIDE_INT.
18712 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18713             Alan Hayward  <alan.hayward@arm.com>
18714             David Sherwood  <david.sherwood@arm.com>
18716         * tree.h (SCALAR_TYPE_MODE): New macro.
18717         * expr.c (expand_expr_addr_expr_1): Use it.
18718         (expand_expr_real_2): Likewise.
18719         * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18720         (fold_convert_const_fixed_from_int): Likewise.
18721         (fold_convert_const_fixed_from_real): Likewise.
18722         (native_encode_fixed): Likewise
18723         (native_encode_complex): Likewise
18724         (native_encode_vector): Likewise.
18725         (native_interpret_fixed): Likewise.
18726         (native_interpret_real): Likewise.
18727         (native_interpret_complex): Likewise.
18728         (native_interpret_vector): Likewise.
18729         * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18730         (simd_clone_adjust_argument_types): Likewise.
18731         (simd_clone_init_simd_arrays): Likewise.
18732         (simd_clone_adjust): Likewise.
18733         * stor-layout.c (layout_type): Likewise.
18734         * tree.c (build_minus_one_cst): Likewise.
18735         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18736         * tree-inline.c (estimate_move_cost): Likewise.
18737         * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18738         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18739         (vectorizable_reduction): Likewise.
18740         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18741         (vect_recog_mixed_size_cond_pattern): Likewise.
18742         (check_bool_pattern): Likewise.
18743         (adjust_bool_pattern): Likewise.
18744         (search_type_for_mask_1): Likewise.
18745         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18746         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18747         (vectorizable_load): Likewise.
18748         (vectorizable_store): Likewise.
18749         * ubsan.c (ubsan_encode_value): Likewise.
18750         * varasm.c (output_constant): Likewise.
18752 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18753             Alan Hayward  <alan.hayward@arm.com>
18754             David Sherwood  <david.sherwood@arm.com>
18756         * coretypes.h (scalar_mode): New class.
18757         * machmode.h (scalar_mode): Likewise.
18758         (scalar_mode::includes_p): New function.
18759         (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18760         * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18761         * genmodes.c (get_mode_class): Handle remaining scalar modes.
18762         * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18763         * expmed.c (store_bit_field_1): Likewise.
18764         (extract_bit_field_1): Likewise.
18765         * expr.c (write_complex_part): Likewise.
18766         (read_complex_part): Likewise.
18767         (emit_move_complex_push): Likewise.
18768         (expand_expr_real_2): Likewise.
18769         * function.c (assign_parm_setup_reg): Likewise.
18770         (assign_parms_unsplit_complex): Likewise.
18771         * optabs.c (expand_binop): Likewise.
18772         * rtlanal.c (subreg_get_info): Likewise.
18773         * simplify-rtx.c (simplify_immed_subreg): Likewise.
18774         * varasm.c (output_constant_pool_2): Likewise.
18776 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18777             Alan Hayward  <alan.hayward@arm.com>
18778             David Sherwood  <david.sherwood@arm.com>
18780         * expmed.c (extract_high_half): Use scalar_int_mode and remove
18781         assertion.
18782         (expmed_mult_highpart_optab): Likewise.
18783         (expmed_mult_highpart): Likewise.
18785 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18786             Alan Hayward  <alan.hayward@arm.com>
18787             David Sherwood  <david.sherwood@arm.com>
18789         * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18790         instead of a machine_mode.
18791         (builtin_memset_read_str): Likewise.
18792         * builtins.c (c_readstr): Likewise.
18793         (builtin_memcpy_read_str): Likewise.
18794         (builtin_strncpy_read_str): Likewise.
18795         (builtin_memset_read_str): Likewise.
18796         (builtin_memset_gen_str): Likewise.
18797         (expand_builtin_signbit): Use scalar_int_mode for local variables.
18798         * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18799         instead of a machine_mode.
18800         * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18801         variables.
18802         (make_extraction): Likewise.
18803         (try_widen_shift_mode): Take and return scalar_int_modes instead
18804         of machine_modes.
18805         * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18806         a scalar_int_mode instead of a machine_mode.
18807         * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18808         (avr_addr_space_pointer_mode): Likewise.
18809         * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18810         * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18811         (msp430_unwind_word_mode): Likewise.
18812         * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18813         (spu_addr_space_pointer_mode): Likewise.
18814         (spu_addr_space_address_mode): Likewise.
18815         (spu_libgcc_cmp_return_mode): Likewise.
18816         (spu_libgcc_shift_count_mode): Likewise.
18817         * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18818         (rl78_addr_space_pointer_mode): Likewise.
18819         (fl78_unwind_word_mode): Likewise.
18820         (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18821         machine_mode.
18822         * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18823         * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18824         * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18825         (mips_valid_pointer_mode): Likewise.
18826         * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18827         * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18828         (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18829         of a machine_mode.
18830         (ft32_addr_space_address_mode): Likewise.
18831         * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18832         scalar_int_mode instead of a machine_mode.
18833         (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18834         of a machine_mode.
18835         (m32c_addr_space_address_mode): Likewise.
18836         * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18837         (rs6000_eh_return_filter_mode): Likewise.
18838         * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18839         (rs6000_eh_return_filter_mode): Likewise.
18840         * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18841         (s390_libgcc_shift_count_mode): Likewise.
18842         (s390_unwind_word_mode): Likewise.
18843         (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18844         machine_mode.
18845         * target.def (mode_rep_extended): Likewise.
18846         (valid_pointer_mode): Likewise.
18847         (addr_space.valid_pointer_mode): Likewise.
18848         (eh_return_filter_mode): Return a scalar_int_mode rather than
18849         a machine_mode.
18850         (libgcc_cmp_return_mode): Likewise.
18851         (libgcc_shift_count_mode): Likewise.
18852         (unwind_word_mode): Likewise.
18853         (addr_space.pointer_mode): Likewise.
18854         (addr_space.address_mode): Likewise.
18855         * doc/tm.texi: Regenerate.
18856         * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18857         a machine_mode.
18858         (do_jump): Use scalar_int_mode for local variables.
18859         * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18860         rather than a machine_mode.
18861         * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18862         (scompare_loc_descriptor_wide): Likewise.
18863         (scompare_loc_descriptor_narrow): Likewise.
18864         * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18865         variables.
18866         * except.c (sjlj_emit_dispatch_table): Likewise.
18867         (expand_builtin_eh_copy_values): Likewise.
18868         * explow.c (convert_memory_address_addr_space_1): Likewise.
18869         Take a scalar_int_mode rather than a machine_mode.
18870         (convert_memory_address_addr_space): Take a scalar_int_mode rather
18871         than a machine_mode.
18872         (memory_address_addr_space): Use scalar_int_mode for local variables.
18873         * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18874         rather than a machine_mode.
18875         * expmed.c (mask_rtx): Likewise.
18876         (init_expmed_one_conv): Likewise.
18877         (expand_mult_highpart_adjust): Likewise.
18878         (extract_high_half): Likewise.
18879         (expmed_mult_highpart_optab): Likewise.
18880         (expmed_mult_highpart): Likewise.
18881         (expand_smod_pow2): Likewise.
18882         (expand_sdiv_pow2): Likewise.
18883         (emit_store_flag_int): Likewise.
18884         (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18885         variables.
18886         (extract_low_bits): Likewise.
18887         * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18888         a machine_mode.
18889         * expr.c (pieces_addr::adjust):  Likewise.
18890         (can_store_by_pieces): Likewise.
18891         (store_by_pieces): Likewise.
18892         (clear_by_pieces_1): Likewise.
18893         (expand_expr_addr_expr_1): Likewise.
18894         (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18895         (expand_expr_real_1): Likewise.
18896         (try_casesi): Likewise.
18897         * final.c (shorten_branches): Likewise.
18898         * fold-const.c (fold_convert_const_int_from_fixed): Change the
18899         type of "mode" to machine_mode.
18900         * internal-fn.c (expand_arith_overflow_result_store): Take a
18901         scalar_int_mode rather than a machine_mode.
18902         (expand_mul_overflow): Use scalar_int_mode for local variables.
18903         * loop-doloop.c (doloop_modify): Likewise.
18904         (doloop_optimize): Likewise.
18905         * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18906         than a machine_mode.
18907         (expand_doubleword_shift_condmove): Likewise.
18908         (expand_doubleword_shift): Likewise.
18909         (expand_doubleword_clz): Likewise.
18910         (expand_doubleword_popcount): Likewise.
18911         (expand_doubleword_parity): Likewise.
18912         (expand_absneg_bit): Use scalar_int_mode for local variables.
18913         (prepare_float_lib_cmp): Likewise.
18914         * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
18915         rather than a machine_mode.
18916         (convert_memory_address_addr_space): Likewise.
18917         (get_mode_bounds): Likewise.
18918         (get_address_mode): Return a scalar_int_mode rather than a
18919         machine_mode.
18920         * rtlanal.c (get_address_mode): Likewise.
18921         * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
18922         than a machine_mode.
18923         * targhooks.c (default_mode_rep_extended): Likewise.
18924         (default_valid_pointer_mode): Likewise.
18925         (default_addr_space_valid_pointer_mode): Likewise.
18926         (default_eh_return_filter_mode): Return a scalar_int_mode rather
18927         than a machine_mode.
18928         (default_libgcc_cmp_return_mode): Likewise.
18929         (default_libgcc_shift_count_mode): Likewise.
18930         (default_unwind_word_mode): Likewise.
18931         (default_addr_space_pointer_mode): Likewise.
18932         (default_addr_space_address_mode): Likewise.
18933         * targhooks.h (default_eh_return_filter_mode): Likewise.
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         (default_mode_rep_extended): Take a scalar_int_mode rather than
18940         a machine_mode.
18941         (default_valid_pointer_mode): Likewise.
18942         (default_addr_space_valid_pointer_mode): Likewise.
18943         * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
18944         local variables.
18945         * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
18946         rather than a machine_mode.
18947         * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
18948         for local variables.
18949         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
18950         * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
18951         than a machine_mode.
18953 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18954             Alan Hayward  <alan.hayward@arm.com>
18955             David Sherwood  <david.sherwood@arm.com>
18957         * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
18958         the mode argument to scalar_int_mode.
18959         (do_jump_by_parts_zero_rtx): Likewise.
18960         (do_jump_by_parts_equality_rtx): Likewise.
18961         (do_jump_by_parts_greater): Take a mode argument.
18962         (do_jump_by_parts_equality): Likewise.
18963         (do_jump_1): Update calls accordingly.
18965 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18966             Alan Hayward  <alan.hayward@arm.com>
18967             David Sherwood  <david.sherwood@arm.com>
18969         * is-a.h (safe_dyn_cast): New function.
18970         * rtl.h (rtx_jump_table_data::get_data_mode): New function.
18971         (jump_table_for_label): Likewise.
18972         * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
18973         instead of an rtx_insn *.
18974         (shorten_branches): Use dyn_cast instead of LABEL_P and
18975         JUMP_TABLE_DATA_P.  Use jump_table_for_label and
18976         rtx_jump_table_data::get_data_mode.
18977         (final_scan_insn): Likewise.
18979 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18980             Alan Hayward  <alan.hayward@arm.com>
18981             David Sherwood  <david.sherwood@arm.com>
18983         * combine.c (try_combine): Use is_a <scalar_int_mode> when
18984         trying to combine a full-register integer set with a subreg
18985         integer set.
18987 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18988             Alan Hayward  <alan.hayward@arm.com>
18989             David Sherwood  <david.sherwood@arm.com>
18991         * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
18992         that is always either address_mode or pointer_mode.
18994 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18995             Alan Hayward  <alan.hayward@arm.com>
18996             David Sherwood  <david.sherwood@arm.com>
18998         * expr.c (expand_expr_real_2): Use word_mode instead of innermode
18999         when the two are known to be equal.
19001 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19002             Alan Hayward  <alan.hayward@arm.com>
19003             David Sherwood  <david.sherwood@arm.com>
19005         * simplify-rtx.c (simplify_const_unary_operation): Use
19006         is_a <scalar_int_mode> instead of checking for a nonzero
19007         precision.  Forcibly convert op_mode to a scalar_int_mode
19008         in that case.  More clearly differentiate the operand and
19009         result modes and use the former when deciding what the value
19010         of a count-bits operation should be.  Use is_int_mode instead
19011         of checking for a MODE_INT.  Remove redundant check for whether
19012         this mode has a zero precision.
19014 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19015             Alan Hayward  <alan.hayward@arm.com>
19016             David Sherwood  <david.sherwood@arm.com>
19018         * optabs.c (widen_leading): Change the type of the mode argument
19019         to scalar_int_mode.  Use opt_scalar_int_mode for the mode iterator.
19020         (widen_bswap): Likewise.
19021         (expand_parity): Likewise.
19022         (expand_ctz): Change the type of the mode argument to scalar_int_mode.
19023         (expand_ffs): Likewise.
19024         (epand_unop): Check for scalar integer modes before calling the
19025         above routines.
19027 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19028             Alan Hayward  <alan.hayward@arm.com>
19029             David Sherwood  <david.sherwood@arm.com>
19031         * expr.c (const_scalar_mask_from_tree): Add a mode argument.
19032         Expand commentary.
19033         (expand_expr_real_1): Update call accordingly.
19035 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19036             Alan Hayward  <alan.hayward@arm.com>
19037             David Sherwood  <david.sherwood@arm.com>
19039         * expmed.c (store_bit_field_using_insv): Add op0_mode and
19040         value_mode arguments.  Use scalar_int_mode internally.
19041         (store_bit_field_1): Rename the new integer mode from imode
19042         to op0_mode and use it instead of GET_MODE (op0).  Update calls
19043         to store_split_bit_field, store_bit_field_using_insv and
19044         store_fixed_bit_field.
19045         (store_fixed_bit_field): Add op0_mode and value_mode arguments.
19046         Use scalar_int_mode internally.  Use a bit count rather than a mode
19047         when calculating the largest bit size for get_best_mode.
19048         Update calls to store_split_bit_field and store_fixed_bit_field_1.
19049         (store_fixed_bit_field_1): Add mode and value_mode arguments.
19050         Remove assertion that OP0 has a scalar integer mode.
19051         (store_split_bit_field): Add op0_mode and value_mode arguments.
19052         Update calls to extract_fixed_bit_field.
19053         (extract_bit_field_using_extv): Add an op0_mode argument.
19054         Use scalar_int_mode internally.
19055         (extract_bit_field_1): Rename the new integer mode from imode to
19056         op0_mode and use it instead of GET_MODE (op0).  Update calls to
19057         extract_split_bit_field, extract_bit_field_using_extv and
19058         extract_fixed_bit_field.
19059         (extract_fixed_bit_field): Add an op0_mode argument.  Update calls
19060         to extract_split_bit_field and extract_fixed_bit_field_1.
19061         (extract_fixed_bit_field_1): Add a mode argument.  Remove assertion
19062         that OP0 has a scalar integer mode.  Use as_a <scalar_int_mode>
19063         on the target mode.
19064         (extract_split_bit_field): Add an op0_mode argument.  Update call
19065         to extract_fixed_bit_field.
19067 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19068             Alan Hayward  <alan.hayward@arm.com>
19069             David Sherwood  <david.sherwood@arm.com>
19071         * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
19072         * explow.c (hard_function_value): Likewise.
19073         * expmed.c (extract_fixed_bit_field_1): Likewise.  Move the
19074         convert_to_mode call outside the loop.
19075         * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
19076         for the mode iterator.  Require the mode specified by max_pieces
19077         to exist.
19078         (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
19079         mode iterator.
19080         (copy_blkmode_to_reg): Likewise.
19081         (set_storage_via_setmem): Likewise.
19082         * optabs.c (prepare_cmp_insn): Likewise.
19083         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19084         * stor-layout.c (finish_bitfield_representative): Likewise.
19086 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19087             Alan Hayward  <alan.hayward@arm.com>
19088             David Sherwood  <david.sherwood@arm.com>
19090         * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
19091         * expr.c (convert_move): Use them.
19092         (convert_modes): Likewise.
19093         (store_expr_with_bounds): Likewise.
19095 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19096             Alan Hayward  <alan.hayward@arm.com>
19097             David Sherwood  <david.sherwood@arm.com>
19099         * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
19100         parameter for the mode of "x".  Remove the "known_x", "known_mode"
19101         and "known_ret" arguments.  Change the type of the mode argument
19102         to scalar_int_mode.
19103         (rtl_hooks:reg_num_sign_bit_copies): Likewise.
19104         * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
19105         (reg_num_sign_bit_copies_for_combine): Likewise.
19106         * rtlanal.c (nonzero_bits1): Likewise.
19107         (num_sign_bit_copies1): Likewise.
19108         * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
19109         (reg_num_sign_bit_copies_general): Likewise.
19110         * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
19111         (reg_nonzero_bits_general): Likewise.
19113 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19114             Alan Hayward  <alan.hayward@arm.com>
19115             David Sherwood  <david.sherwood@arm.com>
19117         * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
19118         than in subroutines.  Return 1 for non-integer modes.
19119         (cached_num_sign_bit_copies): Change the type of the mode parameter
19120         to scalar_int_mode.
19121         (num_sign_bit_copies1): Likewise.  Remove early exit for other mode
19122         classes.  Handle CONST_INT_P first and then check whether X also
19123         has a scalar integer mode.  Check the same thing for inner registers
19124         of a SUBREG and for values that are being extended or truncated.
19126 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19127             Alan Hayward  <alan.hayward@arm.com>
19128             David Sherwood  <david.sherwood@arm.com>
19130         * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
19131         in subroutines.  Return the mode mask for non-integer modes.
19132         (cached_nonzero_bits): Change the type of the mode parameter
19133         to scalar_int_mode.
19134         (nonzero_bits1): Likewise.  Remove early exit for other mode
19135         classes.  Handle CONST_INT_P first and then check whether X
19136         also has a scalar integer mode.
19138 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19139             Alan Hayward  <alan.hayward@arm.com>
19140             David Sherwood  <david.sherwood@arm.com>
19142         * expr.c (widest_int_mode_for_size): Make the comment match the code.
19143         Return a scalar_int_mode and assert that the size is greater than
19144         one byte.
19145         (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
19146         (op_by_pieces_d::op_by_pieces_d): Likewise.
19147         (op_by_pieces_d::run): Likewise.
19148         (can_store_by_pieces): Likewise.
19150 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19151             Alan Hayward  <alan.hayward@arm.com>
19152             David Sherwood  <david.sherwood@arm.com>
19154         * combine.c (extract_left_shift): Add a mode argument and update
19155         recursive calls.
19156         (make_compound_operation_int): Change the type of the mode parameter
19157         to scalar_int_mode and update the call to extract_left_shift.
19159 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19160             Alan Hayward  <alan.hayward@arm.com>
19161             David Sherwood  <david.sherwood@arm.com>
19163         * combine.c (simplify_and_const_int): Change the type of the mode
19164         parameter to scalar_int_mode.
19165         (simplify_and_const_int_1): Likewise.  Update recursive call.
19167 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19168             Alan Hayward  <alan.hayward@arm.com>
19169             David Sherwood  <david.sherwood@arm.com>
19171         * combine.c (simplify_compare_const): Check that the mode is a
19172         scalar_int_mode (rather than VOIDmode) before testing its
19173         precision.
19174         (simplify_comparison): Move COMPARISON_P handling out of the
19175         loop and restrict the latter part of the loop to scalar_int_modes.
19176         Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
19177         and when considering SUBREG_REGs.  Use is_int_mode instead of
19178         checking GET_MODE_CLASS against MODE_INT.
19180 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19181             Alan Hayward  <alan.hayward@arm.com>
19182             David Sherwood  <david.sherwood@arm.com>
19184         * combine.c (try_widen_shift_mode): Move check for equal modes to...
19185         (simplify_shift_const_1): ...here.  Use scalar_int_mode for
19186         shift_unit_mode and for modes involved in scalar shifts.
19188 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19189             Alan Hayward  <alan.hayward@arm.com>
19190             David Sherwood  <david.sherwood@arm.com>
19192         * combine.c (force_int_to_mode): New function, split out from...
19193         (force_to_mode): ...here.  Keep xmode up-to-date and use it
19194         instead of GET_MODE (x).
19196 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19197             Alan Hayward  <alan.hayward@arm.com>
19198             David Sherwood  <david.sherwood@arm.com>
19200         * optabs-query.h (extraction_insn::struct_mode): Change type to
19201         opt_scalar_int_mode and update comment.
19202         (extraction_insn::field_mode): Change type to scalar_int_mode.
19203         (extraction_insn::pos_mode): Likewise.
19204         * combine.c (make_extraction): Update accordingly.
19205         * optabs-query.c (get_traditional_extraction_insn): Likewise.
19206         (get_optab_extraction_insn): Likewise.
19207         * recog.c (simplify_while_replacing): Likewise.
19208         * expmed.c (narrow_bit_field_mem): Change the type of the mode
19209         parameter to opt_scalar_int_mode.
19211 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19212             Alan Hayward  <alan.hayward@arm.com>
19213             David Sherwood  <david.sherwood@arm.com>
19215         * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
19216         to a scalar_int_mode instead of a machine_mode.
19217         (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
19218         (get_best_mode): Return a boolean and use a pointer argument to store
19219         the selected mode.  Replace the limit mode parameter with a bit limit.
19220         * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
19221         for the values returned by bit_field_mode_iterator::next_mode.
19222         (store_bit_field): Update call to get_best_mode.
19223         (store_fixed_bit_field): Likewise.
19224         (extract_fixed_bit_field): Likewise.
19225         * expr.c (optimize_bitfield_assignment_op): Likewise.
19226         * fold-const.c (optimize_bit_field_compare): Likewise.
19227         (fold_truth_andor_1): Likewise.
19228         * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
19229         Update for new type of m_mode.
19230         (get_best_mode): As above.
19232 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19233             Alan Hayward  <alan.hayward@arm.com>
19234             David Sherwood  <david.sherwood@arm.com>
19236         * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
19237         to scalar_int_mode.  Remove check for SCALAR_INT_MODE_P.
19238         (store_bit_field): Check is_a <scalar_int_mode> before calling
19239         strict_volatile_bitfield_p.
19240         (extract_bit_field): Likewise.
19242 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19243             Alan Hayward  <alan.hayward@arm.com>
19244             David Sherwood  <david.sherwood@arm.com>
19246         * target.def (cstore_mode): Return a scalar_int_mode.
19247         * doc/tm.texi: Regenerate.
19248         * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19249         * targhooks.h (default_cstore_mode): Likewise.
19250         * targhooks.c (default_cstore_mode): Likewise, using a forced
19251         conversion.
19252         * expmed.c (emit_cstore): Expect the target of the cstore to be
19253         a scalar_int_mode.
19255 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19256             Alan Hayward  <alan.hayward@arm.com>
19257             David Sherwood  <david.sherwood@arm.com>
19259         * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19260         scalar_int_mode.
19261         (niter_desc): Likewise mode.
19262         (iv_analyze): Add a mode parameter.
19263         (biv_p): Likewise.
19264         (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19265         and change its type to scalar_int_mode.
19266         * loop-iv.c: Update commentary at head of file.
19267         (iv_constant): Pass the mode paraeter before the rtx it describes
19268         and change its type to scalar_int_mode.  Remove VOIDmode handling.
19269         (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19270         (iv_extend): Likewise.
19271         (shorten_into_mode): Likewise.
19272         (iv_add): Use scalar_int_mode.
19273         (iv_mult): Likewise.
19274         (iv_shift): Likewise.
19275         (canonicalize_iv_subregs): Likewise.
19276         (get_biv_step_1): Pass the outer_mode parameter before the rtx
19277         it describes and change its mode to scalar_int_mode.   Also change
19278         the type of the returned inner_mode to scalar_int_mode.
19279         (get_biv_step): Likewise, turning outer_mode from a pointer
19280         into a direct parameter.  Update call to get_biv_step_1.
19281         (iv_analyze_biv): Add an outer_mode parameter.  Update calls to
19282         iv_constant and get_biv_step.
19283         (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19284         and change its type to scalar_int_mode.  Don't initialise iv->mode
19285         to VOIDmode and remove later checks for its still being VOIDmode.
19286         Update calls to iv_analyze_op and iv_analyze_expr.  Check
19287         is_a <scalar_int_mode> when changing the mode under consideration.
19288         (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19289         Update call to iv_analyze_expr.
19290         (iv_analyze_op): Add a mode parameter.  Reject subregs whose
19291         inner register is not also a scalar_int_mode.  Update call to
19292         iv_analyze_biv.
19293         (iv_analyze): Add a mode parameter.  Update call to iv_analyze_op.
19294         (biv_p): Add a mode parameter.  Update call to iv_analyze_biv.
19295         (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19296         separate mode class checks.  Update calls to iv_analyze.  Remove
19297         fix-up of VOIDmodes after iv_analyze_biv.
19298         * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19299         don't have a scalar_int_mode.  Update call to biv_p.
19301 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19302             Alan Hayward  <alan.hayward@arm.com>
19303             David Sherwood  <david.sherwood@arm.com>
19305         * cfgexpand.c (convert_debug_memory_address): Use
19306         as_a <scalar_int_mode>.
19307         * combine.c (expand_compound_operation): Likewise.
19308         (make_extraction): Likewise.
19309         (change_zero_ext): Likewise.
19310         (simplify_comparison): Likewise.
19311         * cse.c (cse_insn): Likewise.
19312         * dwarf2out.c (minmax_loc_descriptor): Likewise.
19313         (mem_loc_descriptor): Likewise.
19314         (loc_descriptor): Likewise.
19315         * expmed.c (init_expmed_one_mode): Likewise.
19316         (synth_mult): Likewise.
19317         (emit_store_flag_1): Likewise.
19318         (expand_divmod): Likewise.  Use HWI_COMPUTABLE_MODE_P instead
19319         of a comparison with size.
19320         * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19321         (reduce_to_bit_field_precision): Likewise.
19322         * function.c (expand_function_end): Likewise.
19323         * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19324         * loop-doloop.c (doloop_modify): Likewise.
19325         * optabs.c (expand_binop): Likewise.
19326         (expand_unop): Likewise.
19327         (expand_copysign_absneg): Likewise.
19328         (prepare_cmp_insn): Likewise.
19329         (maybe_legitimize_operand): Likewise.
19330         * recog.c (const_scalar_int_operand): Likewise.
19331         * rtlanal.c (get_address_mode): Likewise.
19332         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19333         (simplify_cond_clz_ctz): Likewise.
19334         * tree-nested.c (get_nl_goto_field): Likewise.
19335         * tree.c (build_vector_type_for_mode): Likewise.
19336         * var-tracking.c (use_narrower_mode): Likewise.
19338 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19339             Alan Hayward  <alan.hayward@arm.com>
19340             David Sherwood  <david.sherwood@arm.com>
19342         * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19343         * builtins.c (expand_builtin_signbit): Use it.
19344         * cfgexpand.c (expand_debug_expr): Likewise.
19345         * dojump.c (do_jump): Likewise.
19346         (do_compare_and_jump): Likewise.
19347         * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19348         * expmed.c (make_tree): Likewise.
19349         * expr.c (expand_expr_real_2): Likewise.
19350         (expand_expr_real_1): Likewise.
19351         (try_casesi): Likewise.
19352         * fold-const-call.c (fold_const_call_ss): Likewise.
19353         * fold-const.c (unextend): Likewise.
19354         (extract_muldiv_1): Likewise.
19355         (fold_single_bit_test): Likewise.
19356         (native_encode_int): Likewise.
19357         (native_encode_string): Likewise.
19358         (native_interpret_int): Likewise.
19359         * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19360         * internal-fn.c (expand_addsub_overflow): Likewise.
19361         (expand_neg_overflow): Likewise.
19362         (expand_mul_overflow): Likewise.
19363         (expand_arith_overflow): Likewise.
19364         * match.pd: Likewise.
19365         * stor-layout.c (layout_type): Likewise.
19366         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19367         * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19368         * tree-ssanames.c (get_range_info): Likewise.
19369         * tree-switch-conversion.c (array_value_type) Likewise.
19370         * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19371         (vect_recog_divmod_pattern): Likewise.
19372         (vect_recog_mixed_size_cond_pattern): Likewise.
19373         * tree-vrp.c (extract_range_basic): Likewise.
19374         (simplify_float_conversion_using_ranges): Likewise.
19375         * tree.c (int_fits_type_p): Likewise.
19376         * ubsan.c (instrument_bool_enum_load): Likewise.
19377         * varasm.c (mergeable_string_section): Likewise.
19378         (narrowing_initializer_constant_valid_p): Likewise.
19379         (output_constant): Likewise.
19381 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19382             Alan Hayward  <alan.hayward@arm.com>
19383             David Sherwood  <david.sherwood@arm.com>
19385         * machmode.h (NARROWEST_INT_MODE): New macro.
19386         * expr.c (alignment_for_piecewise_move): Use it instead of
19387         GET_CLASS_NARROWEST_MODE (MODE_INT).
19388         (push_block): Likewise.
19389         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19390         Likewise.
19391         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19393 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19394             Alan Hayward  <alan.hayward@arm.com>
19395             David Sherwood  <david.sherwood@arm.com>
19397         * postreload.c (move2add_valid_value_p): Change the type of the
19398         mode parameter to scalar_int_mode.
19399         (move2add_use_add2_insn): Add a mode parameter and use it instead
19400         of GET_MODE (reg).
19401         (move2add_use_add3_insn): Likewise.
19402         (reload_cse_move2add): Update accordingly.
19404 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19405             Alan Hayward  <alan.hayward@arm.com>
19406             David Sherwood  <david.sherwood@arm.com>
19408         * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19409         double-word mode.
19410         * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19411         * optabs.c (expand_unop): Likewise.
19413 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19414             Alan Hayward  <alan.hayward@arm.com>
19415             David Sherwood  <david.sherwood@arm.com>
19417         * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19418         (clz_loc_descriptor): Likewise.  Remove SCALAR_INT_MODE_P check.
19419         (popcount_loc_descriptor): Likewise.
19420         (bswap_loc_descriptor): Likewise.
19421         (rotate_loc_descriptor): Likewise.
19422         (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19423         calling the functions above.
19425 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19426             Alan Hayward  <alan.hayward@arm.com>
19427             David Sherwood  <david.sherwood@arm.com>
19429         * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19430         checks.
19431         (try_combine): Likewise.
19432         (simplify_if_then_else): Likewise.
19433         * cse.c (cse_insn): Likewise.
19434         * dwarf2out.c (mem_loc_descriptor): Likewise.
19435         * emit-rtl.c (gen_lowpart_common): Likewise.
19436         * simplify-rtx.c (simplify_truncation): Likewise.
19437         (simplify_binary_operation_1): Likewise.
19438         (simplify_const_relational_operation): Likewise.
19439         (simplify_ternary_operation): Likewise.
19440         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
19442 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19443             Alan Hayward  <alan.hayward@arm.com>
19444             David Sherwood  <david.sherwood@arm.com>
19446         * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19447         * reload.c (push_reload): Likewise.
19448         (find_reloads): Likewise.
19450 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19451             Alan Hayward  <alan.hayward@arm.com>
19452             David Sherwood  <david.sherwood@arm.com>
19454         * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19455         (make_compound_operation_int): Likewise.
19456         (change_zero_ext): Likewise.
19457         * expr.c (convert_move): Likewise.
19458         (convert_modes): Likewise.
19459         * fwprop.c (forward_propagate_subreg): Likewise.
19460         * loop-iv.c (get_biv_step_1): Likewise.
19461         * optabs.c (widen_operand): Likewise.
19462         * postreload.c (move2add_valid_value_p): Likewise.
19463         * recog.c (simplify_while_replacing): Likewise.
19464         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19465         (simplify_binary_operation_1): Likewise.  Remove redundant
19466         mode equality check.
19468 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19469             Alan Hayward  <alan.hayward@arm.com>
19470             David Sherwood  <david.sherwood@arm.com>
19472         * combine.c (combine_simplify_rtx): Add checks for
19473         is_a <scalar_int_mode>.
19474         (simplify_if_then_else): Likewise.
19475         (make_field_assignment): Likewise.
19476         (simplify_comparison): Likewise.
19477         * ifcvt.c (noce_try_bitop): Likewise.
19478         * loop-invariant.c (canonicalize_address_mult): Likewise.
19479         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19481 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19482             Alan Hayward  <alan.hayward@arm.com>
19483             David Sherwood  <david.sherwood@arm.com>
19485         * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19486         is_a <scalar_int_mode> instead of != BLKmode.
19488 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19489             Alan Hayward  <alan.hayward@arm.com>
19490             David Sherwood  <david.sherwood@arm.com>
19492         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19493         instead of != VOIDmode.
19494         * combine.c (if_then_else_cond): Likewise.
19495         (change_zero_ext): Likewise.
19496         * dwarf2out.c (mem_loc_descriptor): Likewise.
19497         (loc_descriptor): Likewise.
19498         * rtlanal.c (canonicalize_condition): Likewise.
19499         * simplify-rtx.c (simplify_relational_operation_1): Likewise.
19501 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19502             Alan Hayward  <alan.hayward@arm.com>
19503             David Sherwood  <david.sherwood@arm.com>
19505         * simplify-rtx.c (simplify_binary_operation_1): Use
19506         is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
19508 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19509             Alan Hayward  <alan.hayward@arm.com>
19510             David Sherwood  <david.sherwood@arm.com>
19512         * wide-int.h (int_traits<unsigned char>) New class.
19513         (int_traits<unsigned short>) Likewise.
19514         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19515         Use GET_MODE_UNIT_PRECISION and remove redundant test for
19516         SCALAR_INT_MODE_P.
19517         * combine.c (set_nonzero_bits_and_sign_copies): Use
19518         is_a <scalar_int_mode>.
19519         (find_split_point): Likewise.
19520         (combine_simplify_rtx): Likewise.
19521         (simplify_logical): Likewise.
19522         (expand_compound_operation): Likewise.
19523         (expand_field_assignment): Likewise.
19524         (make_compound_operation): Likewise.
19525         (extended_count): Likewise.
19526         (change_zero_ext): Likewise.
19527         (simplify_comparison): Likewise.
19528         * dwarf2out.c (scompare_loc_descriptor): Likewise.
19529         (ucompare_loc_descriptor): Likewise.
19530         (minmax_loc_descriptor): Likewise.
19531         (mem_loc_descriptor): Likewise.
19532         (loc_descriptor): Likewise.
19533         * expmed.c (init_expmed_one_mode): Likewise.
19534         * lra-constraints.c (lra_constraint_offset): Likewise.
19535         * optabs.c (prepare_libcall_arg): Likewise.
19536         * postreload.c (move2add_note_store): Likewise.
19537         * reload.c (operands_match_p): Likewise.
19538         * rtl.h (load_extend_op): Likewise.
19539         * rtlhooks.c (gen_lowpart_general): Likewise.
19540         * simplify-rtx.c (simplify_truncation): Likewise.
19541         (simplify_unary_operation_1): Likewise.
19542         (simplify_binary_operation_1): Likewise.
19543         (simplify_const_binary_operation): Likewise.
19544         (simplify_const_relational_operation): Likewise.
19545         (simplify_subreg): Likewise.
19546         * stor-layout.c (bitwise_mode_for_mode): Likewise.
19547         * var-tracking.c (adjust_mems): Likewise.
19548         (prepare_call_arguments): Likewise.
19550 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19551             Alan Hayward  <alan.hayward@arm.com>
19552             David Sherwood  <david.sherwood@arm.com>
19554         * machmode.h (is_int_mode): New fuction.
19555         * combine.c (find_split_point): Use it.
19556         (combine_simplify_rtx): Likewise.
19557         (simplify_if_then_else): Likewise.
19558         (simplify_set): Likewise.
19559         (simplify_shift_const_1): Likewise.
19560         (simplify_comparison): Likewise.
19561         * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19562         * cse.c (notreg_cost): Likewise.
19563         (cse_insn): Likewise.
19564         * cselib.c (cselib_lookup_1): Likewise.
19565         * dojump.c (do_jump_1): Likewise.
19566         (do_compare_rtx_and_jump): Likewise.
19567         * dse.c (get_call_args): Likewise.
19568         * dwarf2out.c (rtl_for_decl_init): Likewise.
19569         (native_encode_initializer): Likewise.
19570         * expmed.c (emit_store_flag_1): Likewise.
19571         (emit_store_flag): Likewise.
19572         * expr.c (convert_modes): Likewise.
19573         (store_field): Likewise.
19574         (expand_expr_real_1): Likewise.
19575         * fold-const.c (fold_read_from_constant_string): Likewise.
19576         * gimple-ssa-sprintf.c (get_format_string): Likewise.
19577         * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19578         * optabs.c (expand_binop): Likewise.
19579         (expand_unop): Likewise.
19580         (expand_abs_nojump): Likewise.
19581         (expand_one_cmpl_abs_nojump): Likewise.
19582         * simplify-rtx.c (mode_signbit_p): Likewise.
19583         (val_signbit_p): Likewise.
19584         (val_signbit_known_set_p): Likewise.
19585         (val_signbit_known_clear_p): Likewise.
19586         (simplify_relational_operation_1): Likewise.
19587         * tree.c (vector_type_mode): Likewise.
19589 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19590             Alan Hayward  <alan.hayward@arm.com>
19591             David Sherwood  <david.sherwood@arm.com>
19593         * machmode.h (smallest_mode_for_size): Fix formatting.
19594         (smallest_int_mode_for_size): New function.
19595         * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19596         instead of smallest_mode_for_size.
19597         * combine.c (make_extraction): Likewise.
19598         * config/arc/arc.c (arc_expand_movmem): Likewise.
19599         * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19600         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19601         * config/s390/s390.c (s390_expand_insv): Likewise.
19602         * config/sparc/sparc.c (assign_int_registers): Likewise.
19603         * config/spu/spu.c (spu_function_value): Likewise.
19604         (spu_function_arg): Likewise.
19605         * coverage.c (get_gcov_type): Likewise.
19606         (get_gcov_unsigned_t): Likewise.
19607         * dse.c (find_shift_sequence): Likewise.
19608         * expmed.c (store_bit_field_1): Likewise.
19609         * expr.c (convert_move): Likewise.
19610         (store_field): Likewise.
19611         * internal-fn.c (expand_arith_overflow): Likewise.
19612         * optabs-query.c (get_best_extraction_insn): Likewise.
19613         * optabs.c (expand_twoval_binop_libfunc): Likewise.
19614         * stor-layout.c (layout_type): Likewise.
19615         (initialize_sizetypes): Likewise.
19616         * targhooks.c (default_get_mask_mode): Likewise.
19617         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19619 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19620             Alan Hayward  <alan.hayward@arm.com>
19621             David Sherwood  <david.sherwood@arm.com>
19623         * machmode.h (opt_mode::else_blk): New function.
19624         (int_mode_for_mode): Declare.
19625         * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19626         * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19627         return type.
19628         * cfgexpand.c (expand_debug_expr): Likewise.
19629         * combine.c (gen_lowpart_or_truncate): Likewise.
19630         (gen_lowpart_for_combine): Likewise.
19631         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19632         * config/avr/avr.c (avr_to_int_mode): Likewise.
19633         (avr_out_plus_1): Likewise.
19634         (avr_out_plus): Likewise.
19635         (avr_out_round): Likewise.
19636         * config/i386/i386.c (ix86_split_to_parts): Likewise.
19637         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19638         (s390_expand_vcond): Likewise.
19639         * config/spu/spu.c (spu_split_immediate): Likewise.
19640         (spu_expand_mov): Likewise.
19641         * dse.c (get_stored_val): Likewise.
19642         * expmed.c (store_bit_field_1): Likewise.
19643         (convert_extracted_bit_field): Use int_mode_for_mode instead of
19644         int_mode_for_size.
19645         (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19646         (extract_low_bits): Likewise.
19647         * expr.c (emit_group_load_1): Likewise.  Separate out the BLKmode
19648         handling rather than repeating the check.
19649         (emit_group_store): Likewise.
19650         (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19651         * optabs.c (expand_absneg_bit): Likewise.
19652         (expand_copysign_absneg): Likewise.
19653         (expand_copysign_bit): Likewise.
19654         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19655         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19656         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19657         * var-tracking.c (prepare_call_arguments):  Likewise.
19658         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19659         int_mode_for_mode instead of mode_for_size.
19660         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19662 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19663             Alan Hayward  <alan.hayward@arm.com>
19664             David Sherwood  <david.sherwood@arm.com>
19666         * machmode.h (int_mode_for_size): New function.
19667         * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19668         instead of mode_for_size.
19669         * calls.c (save_fixed_argument_area): Likewise.  Make use of BLKmode
19670         explicit.
19671         * combine.c (expand_field_assignment): Use int_mode_for_size
19672         instead of mode_for_size.
19673         (make_extraction): Likewise.
19674         (simplify_shift_const_1): Likewise.
19675         (simplify_comparison): Likewise.
19676         * dojump.c (do_jump): Likewise.
19677         * dwarf2out.c (mem_loc_descriptor): Likewise.
19678         * emit-rtl.c (init_derived_machine_modes): Likewise.
19679         * expmed.c (flip_storage_order): Likewise.
19680         (convert_extracted_bit_field): Likewise.
19681         * expr.c (copy_blkmode_from_reg): Likewise.
19682         * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19683         * internal-fn.c (expand_mul_overflow): Likewise.
19684         * lower-subreg.c (simple_move): Likewise.
19685         * optabs-libfuncs.c (init_optabs): Likewise.
19686         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19687         * tree.c (vector_type_mode): Likewise.
19688         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19689         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19690         * tree-vect-generic.c (expand_vector_parallel): Likewise.
19691         * tree-vect-stmts.c (vectorizable_load): Likewise.
19692         (vectorizable_store): Likewise.
19694 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19695             Alan Hayward  <alan.hayward@arm.com>
19696             David Sherwood  <david.sherwood@arm.com>
19698         * coretypes.h (pod_mode): New type.
19699         (scalar_int_mode_pod): New typedef.
19700         * machmode.h (pod_mode): New class.
19701         (int_n_data_t::m): Change type to scalar_int_mode_pod.
19702         * genmodes.c (emit_mode_int_n): Update accordingly.
19703         * lower-subreg.h (target_lower_subreg): Change type to
19704         scalar_int_mode_pod.
19705         * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19706         scalar_int_mode_pod.
19708 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19709             Alan Hayward  <alan.hayward@arm.com>
19710             David Sherwood  <david.sherwood@arm.com>
19712         * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19713         machine_mode to scalar_int_mode.
19714         * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19715         (rs6000_option_override_internal): Remove cast to int.
19716         * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19717         machine_mode to scalar_int_mode.
19718         * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19719         (rs6000_option_override_internal): Remove cast to int.
19720         * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19721         * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19722         to machine_mode.
19723         * config/s390/s390.c (s390_expand_builtin): Likewise.
19724         * coretypes.h (scalar_int_mode): New type.
19725         (opt_scalar_int_mode): New typedef.
19726         * machmode.h (scalar_int_mode): New class.
19727         (scalar_int_mode::includes_p): New function.
19728         (byte_mode): Change type to scalar_int_mode.
19729         (word_mode): Likewise.
19730         (ptr_mode): Likewise.
19731         * emit-rtl.c (byte_mode): Likewise.
19732         (word_mode): Likewise.
19733         (ptr_mode): Likewise.
19734         (init_derived_machine_modes): Update accordingly.
19735         * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19736         and MODE_PARTIAL_INT.
19737         * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19738         opt_scalar_int_mode.
19740 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19741             Alan Hayward  <alan.hayward@arm.com>
19742             David Sherwood  <david.sherwood@arm.com>
19744         * target.def (libgcc_floating_mode_supported_p): Take a
19745         scalar_float_mode.
19746         * doc/tm.texi: Regenerate.
19747         * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19748         scalar_float_mode.
19749         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19750         * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19751         Likewise.
19753 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19754             Alan Hayward  <alan.hayward@arm.com>
19755             David Sherwood  <david.sherwood@arm.com>
19757         * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19758         * doc/tm.texi: Regenerate.
19759         * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19760         * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19761         * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19762         * targhooks.h (default_floatn_mode): Likewise.
19763         * targhooks.c (default_floatn_mode): Likewise.
19764         * tree.c (build_common_tree_nodes): Update accordingly.
19766 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19767             Alan Hayward  <alan.hayward@arm.com>
19768             David Sherwood  <david.sherwood@arm.com>
19770         * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19771         (mode_iterator::iterate_p): Likewise.
19772         (mode_iterator::get_wider): Likewise.
19773         * expr.c (init_expr_target): Use opt_scalar_float_mode.
19775 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19776             Alan Hayward  <alan.hayward@arm.com>
19777             David Sherwood  <david.sherwood@arm.com>
19779         * coretypes.h (opt_scalar_float_mode): New typedef.
19780         * machmode.h (float_mode_for_size): New function.
19781         * emit-rtl.c (double_mode): Delete.
19782         (init_emit_once): Use float_mode_for_size.
19783         * stor-layout.c (layout_type): Likewise.
19784         * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
19786 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19787             Alan Hayward  <alan.hayward@arm.com>
19788             David Sherwood  <david.sherwood@arm.com>
19790         * output.h (assemble_real): Take a scalar_float_mode.
19791         * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19792         * config/arm/arm.md (consttable_4): Likewise.
19793         (consttable_8): Likewise.
19794         (consttable_16): Likewise.
19795         * config/mips/mips.md (consttable_float): Likewise.
19796         * config/s390/s390.c (s390_output_pool_entry): Likewise.
19797         * varasm.c (assemble_real): Take a scalar_float_mode.
19798         (output_constant_pool_2): Update accordingly.
19799         (output_constant): Likewise.
19801 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19802             Alan Hayward  <alan.hayward@arm.com>
19803             David Sherwood  <david.sherwood@arm.com>
19805         * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19806         * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19807         * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19808         (native_encode_real): Likewise.
19809         (native_interpret_real): Likewise.
19810         * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19811         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19813 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19814             Alan Hayward  <alan.hayward@arm.com>
19815             David Sherwood  <david.sherwood@arm.com>
19817         * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19818         <scalar_float_mode>.  Simplify.
19819         (gen_extend_conv_libfunc): Likewise.
19821 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19822             Alan Hayward  <alan.hayward@arm.com>
19823             David Sherwood  <david.sherwood@arm.com>
19825         * coretypes.h (scalar_float_mode): New type.
19826         * machmode.h (mode_traits::from_int): Use machine_mode if
19827         USE_ENUM_MODES is defined.
19828         (is_a): New function.
19829         (as_a): Likewise.
19830         (dyn_cast): Likewise.
19831         (scalar_float_mode): New class.
19832         (scalar_float_mode::includes_p): New function.
19833         (is_float_mode): Likewise.
19834         * gdbhooks.py (MachineModePrinter): New class.
19835         (build_pretty_printer): Use it for scalar_float_mode.
19836         * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19837         (format_helper::format_helper): Turn into a template.
19838         * genmodes.c (get_mode_class): New function.
19839         (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19840         or machine_mode if none.
19841         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19842         as_a <scalar_float_mode>.
19843         * dwarf2out.c (mem_loc_descriptor): Likewise.
19844         (insert_float): Likewise.
19845         (add_const_value_attribute): Likewise.
19846         * simplify-rtx.c (simplify_immed_subreg): Likewise.
19847         * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19848         (expand_unop): Update accordingly.
19849         (expand_abs_nojump): Likewise.
19850         (expand_copysign_absneg): Take a scalar_float_mode.
19851         (expand_copysign_bit): Likewise.
19852         (expand_copysign): Update accordingly.
19854 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19855             Alan Hayward  <alan.hayward@arm.com>
19856             David Sherwood  <david.sherwood@arm.com>
19858         * coretypes.h (opt_mode): New class.
19859         * machmode.h (opt_mode): Likewise.
19860         (opt_mode::else_void): New function.
19861         (opt_mode::require): Likewise.
19862         (opt_mode::exists): Likewise.
19863         (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19864         (GET_MODE_2XWIDER_MODE): Likewise.
19865         (mode_iterator::get_wider): Update accordingly.
19866         (mode_iterator::get_2xwider): Likewise.
19867         (mode_iterator::get_known_wider): Likewise, turning into a template.
19868         * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19869         forcing a wider mode to exist.
19870         * config/cr16/cr16.h (LONG_REG_P): Likewise.
19871         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19872         * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19873         GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19874         * lower-subreg.c (init_lower_subreg): Likewise.
19875         * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19876         on the final iteration.
19877         * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19878         a wider mode exists before asking for a move pattern.
19879         (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19880         forcing a wider mode to exist.
19881         (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19882         returning false if no such mode exists.
19883         * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19884         * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19885         * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19886         Avoid checking for a MODE_INT if we already know the mode is not a
19887         SCALAR_INT_MODE_P.
19888         (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19889         forcing a wider mode to exist.
19890         (expmed_mult_highpart_optab): Likewise.
19891         (expmed_mult_highpart): Likewise.
19892         * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19893         using else_void.
19894         * lto-streamer-in.c (lto_input_mode_table): Likewise.
19895         * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19896         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19897         * internal-fn.c (expand_mul_overflow): Update use of
19898         GET_MODE_2XWIDER_MODE.
19899         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19900         * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19901         GET_MODE_WIDER_MODE.
19902         (convert_plusminus_to_widen): Likewise.
19903         * tree-switch-conversion.c (array_value_type): Likewise.
19904         * var-tracking.c (emit_note_insn_var_location): Likewise.
19905         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19906         Return false inside rather than outside the loop if no wider mode
19907         exists
19908         * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19909         and GET_MODE_2XWIDER_MODE
19910         (can_compare_p): Use else_void.
19911         * gdbhooks.py (OptMachineModePrinter): New class.
19912         (build_pretty_printer): Use it for opt_mode.
19914 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19915             Alan Hayward  <alan.hayward@arm.com>
19916             David Sherwood  <david.sherwood@arm.com>
19918         * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
19919         once.  Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
19921 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19922             Alan Hayward  <alan.hayward@arm.com>
19923             David Sherwood  <david.sherwood@arm.com>
19925         * machmode.h (mode_traits): New structure.
19926         (get_narrowest_mode): New function.
19927         (mode_iterator::start): Likewise.
19928         (mode_iterator::iterate_p): Likewise.
19929         (mode_iterator::get_wider): Likewise.
19930         (mode_iterator::get_known_wider): Likewise.
19931         (mode_iterator::get_2xwider): Likewise.
19932         (FOR_EACH_MODE_IN_CLASS): New mode iterator.
19933         (FOR_EACH_MODE): Likewise.
19934         (FOR_EACH_MODE_FROM): Likewise.
19935         (FOR_EACH_MODE_UNTIL): Likewise.
19936         (FOR_EACH_WIDER_MODE): Likewise.
19937         (FOR_EACH_2XWIDER_MODE): Likewise.
19938         * builtins.c (expand_builtin_strlen): Use new mode iterators.
19939         * combine.c (simplify_comparison): Likewise
19940         * config/i386/i386.c (type_natural_mode): Likewise.
19941         * cse.c (cse_insn): Likewise.
19942         * dse.c (find_shift_sequence): Likewise.
19943         * emit-rtl.c (init_derived_machine_modes): Likewise.
19944         (init_emit_once): Likewise.
19945         * explow.c (hard_function_value): Likewise.
19946         * expmed.c (extract_fixed_bit_field_1): Likewise.
19947         (extract_bit_field_1): Likewise.
19948         (expand_divmod): Likewise.
19949         (emit_store_flag_1): Likewise.
19950         * expr.c (init_expr_target): Likewise.
19951         (convert_move): Likewise.
19952         (alignment_for_piecewise_move): Likewise.
19953         (widest_int_mode_for_size): Likewise.
19954         (emit_block_move_via_movmem): Likewise.
19955         (copy_blkmode_to_reg): Likewise.
19956         (set_storage_via_setmem): Likewise.
19957         (compress_float_constant): Likewise.
19958         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19959         * optabs-query.c (get_best_extraction_insn): Likewise.
19960         * optabs.c (expand_binop): Likewise.
19961         (expand_twoval_unop): Likewise.
19962         (expand_twoval_binop): Likewise.
19963         (widen_leading): Likewise.
19964         (widen_bswap): Likewise.
19965         (expand_parity): Likewise.
19966         (expand_unop): Likewise.
19967         (prepare_cmp_insn): Likewise.
19968         (prepare_float_lib_cmp): Likewise.
19969         (expand_float): Likewise.
19970         (expand_fix): Likewise.
19971         (expand_sfix_optab): Likewise.
19972         * postreload.c (move2add_use_add2_insn): Likewise.
19973         * reg-stack.c (reg_to_stack): Likewise.
19974         * reginfo.c (choose_hard_reg_mode): Likewise.
19975         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19976         * stor-layout.c (mode_for_size): Likewise.
19977         (smallest_mode_for_size): Likewise.
19978         (mode_for_vector): Likewise.
19979         (finish_bitfield_representative): Likewise.
19980         * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
19981         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19982         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
19983         * var-tracking.c (prepare_call_arguments): Likewise.
19985 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19986             Alan Hayward  <alan.hayward@arm.com>
19987             David Sherwood  <david.sherwood@arm.com>
19989         * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
19990         * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
19991         USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
19992         * machmode.h (mode_size): Move earlier in file.
19993         (mode_precision): Likewise.
19994         (mode_inner): Likewise.
19995         (mode_nunits): Likewise.
19996         (mode_unit_size): Likewise.
19997         (unit_unit_precision): Likewise.
19998         (mode_wider): Likewise.
19999         (mode_2xwider): Likewise.
20000         (machine_mode): New class.
20001         (mode_to_bytes): New function.
20002         (mode_to_bits): Likewise.
20003         (mode_to_precision): Likewise.
20004         (mode_to_inner): Likewise.
20005         (mode_to_unit_size): Likewise.
20006         (mode_to_unit_precision): Likewise.
20007         (mode_to_nunits): Likewise.
20008         (GET_MODE_SIZE): Use mode_to_bytes.
20009         (GET_MODE_BITSIZE): Use mode_to_bits.
20010         (GET_MODE_PRECISION): Use mode_to_precision.
20011         (GET_MODE_INNER): Use mode_to_inner.
20012         (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
20013         (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
20014         (GET_MODE_NUNITS): Use mode_to_nunits.
20015         * system.h (ALWAYS_INLINE): New macro.
20016         * config/powerpcspe/powerpcspe-c.c
20017         (altivec_resolve_overloaded_builtin): Use machine_mode instead of
20018         int for arg1_mode and arg2_mode.
20020 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20021             Alan Hayward  <alan.hayward@arm.com>
20022             David Sherwood  <david.sherwood@arm.com>
20024         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
20025         Prefix mode names with E_ in case statements.
20026         * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20027         * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
20028         (aarch64_split_simd_move): Likewise.
20029         (aarch64_gen_storewb_pair): Likewise.
20030         (aarch64_gen_loadwb_pair): Likewise.
20031         (aarch64_gen_store_pair): Likewise.
20032         (aarch64_gen_load_pair): Likewise.
20033         (aarch64_get_condition_code_1): Likewise.
20034         (aarch64_constant_pool_reload_icode): Likewise.
20035         (get_rsqrte_type): Likewise.
20036         (get_rsqrts_type): Likewise.
20037         (get_recpe_type): Likewise.
20038         (get_recps_type): Likewise.
20039         (aarch64_gimplify_va_arg_expr): Likewise.
20040         (aarch64_simd_container_mode): Likewise.
20041         (aarch64_emit_load_exclusive): Likewise.
20042         (aarch64_emit_store_exclusive): Likewise.
20043         (aarch64_expand_compare_and_swap): Likewise.
20044         (aarch64_gen_atomic_cas): Likewise.
20045         (aarch64_emit_bic): Likewise.
20046         (aarch64_emit_atomic_swap): Likewise.
20047         (aarch64_emit_atomic_load_op): Likewise.
20048         (aarch64_evpc_trn): Likewise.
20049         (aarch64_evpc_uzp): Likewise.
20050         (aarch64_evpc_zip): Likewise.
20051         (aarch64_evpc_ext): Likewise.
20052         (aarch64_evpc_rev): Likewise.
20053         (aarch64_evpc_dup): Likewise.
20054         (aarch64_gen_ccmp_first): Likewise.
20055         (aarch64_gen_ccmp_next): Likewise.
20056         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
20057         (alpha_emit_xfloating_libcall): Likewise.
20058         (emit_insxl): Likewise.
20059         (alpha_arg_type): Likewise.
20060         * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
20061         (arc_preferred_simd_mode): Likewise.
20062         (arc_secondary_reload): Likewise.
20063         (get_arc_condition_code): Likewise.
20064         (arc_print_operand): Likewise.
20065         (arc_legitimate_constant_p): Likewise.
20066         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20067         * config/arc/arc.md (casesi_load): Likewise.
20068         (casesi_compact_jump): Likewise.
20069         * config/arc/predicates.md (proper_comparison_operator): Likewise.
20070         (cc_use_register): Likewise.
20071         * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20072         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
20073         (arm_init_iwmmxt_builtins): Likewise.
20074         * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
20075         (neon_expand_vector_init): Likewise.
20076         (arm_attr_length_move_neon): Likewise.
20077         (maybe_get_arm_condition_code): Likewise.
20078         (arm_emit_vector_const): Likewise.
20079         (arm_preferred_simd_mode): Likewise.
20080         (arm_output_iwmmxt_tinsr): Likewise.
20081         (thumb1_output_casesi): Likewise.
20082         (thumb2_output_casesi): Likewise.
20083         (arm_emit_load_exclusive): Likewise.
20084         (arm_emit_store_exclusive): Likewise.
20085         (arm_expand_compare_and_swap): Likewise.
20086         (arm_evpc_neon_vuzp): Likewise.
20087         (arm_evpc_neon_vzip): Likewise.
20088         (arm_evpc_neon_vrev): Likewise.
20089         (arm_evpc_neon_vtrn): Likewise.
20090         (arm_evpc_neon_vext): Likewise.
20091         (arm_validize_comparison): Likewise.
20092         * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
20093         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
20094         * config/avr/avr.c (avr_rtx_costs_1): Likewise.
20095         * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
20096         (c6x_preferred_simd_mode): Likewise.
20097         * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
20098         (epiphany_rtx_costs): Likewise.
20099         * config/epiphany/predicates.md (proper_comparison_operator):
20100         Likewise.
20101         * config/frv/frv.c (condexec_memory_operand): Likewise.
20102         (frv_emit_move): Likewise.
20103         (output_move_single): Likewise.
20104         (output_condmove_single): Likewise.
20105         (frv_hard_regno_mode_ok): Likewise.
20106         (frv_matching_accg_mode): Likewise.
20107         * config/h8300/h8300.c (split_adds_subs): Likewise.
20108         (h8300_rtx_costs): Likewise.
20109         (h8300_print_operand): Likewise.
20110         (compute_mov_length): Likewise.
20111         (output_logical_op): Likewise.
20112         (compute_logical_op_length): Likewise.
20113         (compute_logical_op_cc): Likewise.
20114         (h8300_shift_needs_scratch_p): Likewise.
20115         (output_a_shift): Likewise.
20116         (compute_a_shift_length): Likewise.
20117         (compute_a_shift_cc): Likewise.
20118         (expand_a_rotate): Likewise.
20119         (output_a_rotate): Likewise.
20120         * config/i386/i386.c (classify_argument): Likewise.
20121         (function_arg_advance_32): Likewise.
20122         (function_arg_32): Likewise.
20123         (function_arg_64): Likewise.
20124         (function_value_64): Likewise.
20125         (ix86_gimplify_va_arg): Likewise.
20126         (ix86_legitimate_constant_p): Likewise.
20127         (put_condition_code): Likewise.
20128         (split_double_mode): Likewise.
20129         (ix86_avx256_split_vector_move_misalign): Likewise.
20130         (ix86_expand_vector_logical_operator): Likewise.
20131         (ix86_split_idivmod): Likewise.
20132         (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
20133         (ix86_build_const_vector): Likewise.
20134         (ix86_build_signbit_mask): Likewise.
20135         (ix86_match_ccmode): Likewise.
20136         (ix86_cc_modes_compatible): Likewise.
20137         (ix86_expand_branch): Likewise.
20138         (ix86_expand_sse_cmp): Likewise.
20139         (ix86_expand_sse_movcc): Likewise.
20140         (ix86_expand_int_sse_cmp): Likewise.
20141         (ix86_expand_vec_perm_vpermi2): Likewise.
20142         (ix86_expand_vec_perm): Likewise.
20143         (ix86_expand_sse_unpack): Likewise.
20144         (ix86_expand_int_addcc): Likewise.
20145         (ix86_split_to_parts): Likewise.
20146         (ix86_vectorize_builtin_gather): Likewise.
20147         (ix86_vectorize_builtin_scatter): Likewise.
20148         (avx_vpermilp_parallel): Likewise.
20149         (inline_memory_move_cost): Likewise.
20150         (ix86_tieable_integer_mode_p): Likewise.
20151         (x86_maybe_negate_const_int): Likewise.
20152         (ix86_expand_vector_init_duplicate): Likewise.
20153         (ix86_expand_vector_init_one_nonzero): Likewise.
20154         (ix86_expand_vector_init_one_var): Likewise.
20155         (ix86_expand_vector_init_concat): Likewise.
20156         (ix86_expand_vector_init_interleave): Likewise.
20157         (ix86_expand_vector_init_general): Likewise.
20158         (ix86_expand_vector_set): Likewise.
20159         (ix86_expand_vector_extract): Likewise.
20160         (emit_reduc_half): Likewise.
20161         (ix86_emit_i387_round): Likewise.
20162         (ix86_mangle_type): Likewise.
20163         (ix86_expand_round_sse4): Likewise.
20164         (expand_vec_perm_blend): Likewise.
20165         (canonicalize_vector_int_perm): Likewise.
20166         (ix86_expand_vec_one_operand_perm_avx512): Likewise.
20167         (expand_vec_perm_1): Likewise.
20168         (expand_vec_perm_interleave3): Likewise.
20169         (expand_vec_perm_even_odd_pack): Likewise.
20170         (expand_vec_perm_even_odd_1): Likewise.
20171         (expand_vec_perm_broadcast_1): Likewise.
20172         (ix86_vectorize_vec_perm_const_ok): Likewise.
20173         (ix86_expand_vecop_qihi): Likewise.
20174         (ix86_expand_mul_widen_hilo): Likewise.
20175         (ix86_expand_sse2_abs): Likewise.
20176         (ix86_expand_pextr): Likewise.
20177         (ix86_expand_pinsr): Likewise.
20178         (ix86_preferred_simd_mode): Likewise.
20179         (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
20180         * config/i386/sse.md (*andnot<mode>3): Likewise.
20181         (<mask_codefor><code><mode>3<mask_name>): Likewise.
20182         (*<code><mode>3): Likewise.
20183         * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
20184         (ia64_expand_atomic_op): Likewise.
20185         (ia64_arg_type): Likewise.
20186         (ia64_mode_to_int): Likewise.
20187         (ia64_scalar_mode_supported_p): Likewise.
20188         (ia64_vector_mode_supported_p): Likewise.
20189         (expand_vec_perm_broadcast): Likewise.
20190         * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
20191         (iq2000_function_arg_advance): Likewise.
20192         (iq2000_function_arg): Likewise.
20193         * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
20194         * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
20195         (m68k_libcall_value): Likewise.
20196         (m68k_function_value): Likewise.
20197         (sched_attr_op_type): Likewise.
20198         * config/mcore/mcore.c (mcore_output_move): Likewise.
20199         * config/microblaze/microblaze.c (microblaze_function_arg_advance):
20200         Likewise.
20201         (microblaze_function_arg): Likewise.
20202         * config/mips/mips.c (mips16_build_call_stub): Likewise.
20203         (mips_print_operand): Likewise.
20204         (mips_mode_ok_for_mov_fmt_p): Likewise.
20205         (mips_vector_mode_supported_p): Likewise.
20206         (mips_preferred_simd_mode): Likewise.
20207         (mips_expand_vpc_loongson_even_odd): Likewise.
20208         (mips_expand_vec_unpack): Likewise.
20209         (mips_expand_vi_broadcast): Likewise.
20210         (mips_expand_vector_init): Likewise.
20211         (mips_expand_vec_reduc): Likewise.
20212         (mips_expand_msa_cmp): Likewise.
20213         * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
20214         * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
20215         (cc_flags_for_mode): Likewise.
20216         * config/msp430/msp430.c (msp430_print_operand): Likewise.
20217         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
20218         (nds32_output_casesi_pc_relative): Likewise.
20219         * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20220         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
20221         (nvptx_gen_unpack): Likewise.
20222         (nvptx_gen_pack): Likewise.
20223         (nvptx_gen_shuffle): Likewise.
20224         (nvptx_gen_wcast): Likewise.
20225         (nvptx_preferred_simd_mode): Likewise.
20226         * config/pa/pa.c (pa_secondary_reload): Likewise.
20227         * config/pa/predicates.md (base14_operand): Likewise.
20228         * config/powerpcspe/powerpcspe-c.c
20229         (altivec_resolve_overloaded_builtin): Likewise.
20230         * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
20231         Likewise.
20232         (rs6000_preferred_simd_mode): Likewise.
20233         (output_vec_const_move): Likewise.
20234         (rs6000_expand_vector_extract): Likewise.
20235         (rs6000_split_vec_extract_var): Likewise.
20236         (reg_offset_addressing_ok_p): Likewise.
20237         (rs6000_legitimate_offset_address_p): Likewise.
20238         (rs6000_legitimize_address): Likewise.
20239         (rs6000_emit_set_const): Likewise.
20240         (rs6000_const_vec): Likewise.
20241         (rs6000_emit_move): Likewise.
20242         (spe_build_register_parallel): Likewise.
20243         (rs6000_darwin64_record_arg_recurse): Likewise.
20244         (swap_selector_for_mode): Likewise.
20245         (spe_init_builtins): Likewise.
20246         (paired_init_builtins): Likewise.
20247         (altivec_init_builtins): Likewise.
20248         (do_load_for_compare): Likewise.
20249         (rs6000_generate_compare): Likewise.
20250         (rs6000_expand_float128_convert): Likewise.
20251         (emit_load_locked): Likewise.
20252         (emit_store_conditional): Likewise.
20253         (rs6000_output_function_epilogue): Likewise.
20254         (rs6000_handle_altivec_attribute): Likewise.
20255         (rs6000_function_value): Likewise.
20256         (emit_fusion_gpr_load): Likewise.
20257         (emit_fusion_p9_load): Likewise.
20258         (emit_fusion_p9_store): Likewise.
20259         * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20260         (fusion_gpr_mem_load): Likewise.
20261         (fusion_addis_mem_combo_load): Likewise.
20262         (fusion_addis_mem_combo_store): Likewise.
20263         * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20264         (fusion_gpr_mem_load): Likewise.
20265         (fusion_addis_mem_combo_load): Likewise.
20266         (fusion_addis_mem_combo_store): Likewise.
20267         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20268         Likewise.
20269         * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20270         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20271         (rs6000_preferred_simd_mode): Likewise.
20272         (output_vec_const_move): Likewise.
20273         (rs6000_expand_vector_extract): Likewise.
20274         (rs6000_split_vec_extract_var): Likewise.
20275         (reg_offset_addressing_ok_p): Likewise.
20276         (rs6000_legitimate_offset_address_p): Likewise.
20277         (rs6000_legitimize_address): Likewise.
20278         (rs6000_emit_set_const): Likewise.
20279         (rs6000_const_vec): Likewise.
20280         (rs6000_emit_move): Likewise.
20281         (rs6000_darwin64_record_arg_recurse): Likewise.
20282         (swap_selector_for_mode): Likewise.
20283         (paired_init_builtins): Likewise.
20284         (altivec_init_builtins): Likewise.
20285         (rs6000_expand_float128_convert): Likewise.
20286         (emit_load_locked): Likewise.
20287         (emit_store_conditional): Likewise.
20288         (rs6000_output_function_epilogue): Likewise.
20289         (rs6000_handle_altivec_attribute): Likewise.
20290         (rs6000_function_value): Likewise.
20291         (emit_fusion_gpr_load): Likewise.
20292         (emit_fusion_p9_load): Likewise.
20293         (emit_fusion_p9_store): Likewise.
20294         * config/rx/rx.c (rx_gen_move_template): Likewise.
20295         (flags_from_mode): Likewise.
20296         * config/s390/predicates.md (s390_alc_comparison): Likewise.
20297         (s390_slb_comparison): Likewise.
20298         * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20299         (s390_vector_mode_supported_p): Likewise.
20300         (s390_cc_modes_compatible): Likewise.
20301         (s390_match_ccmode_set): Likewise.
20302         (s390_canonicalize_comparison): Likewise.
20303         (s390_emit_compare_and_swap): Likewise.
20304         (s390_branch_condition_mask): Likewise.
20305         (s390_rtx_costs): Likewise.
20306         (s390_secondary_reload): Likewise.
20307         (__SECONDARY_RELOAD_CASE): Likewise.
20308         (s390_expand_cs): Likewise.
20309         (s390_preferred_simd_mode): Likewise.
20310         * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20311         * config/sh/sh.c (sh_print_operand): Likewise.
20312         (dump_table): Likewise.
20313         (sh_secondary_reload): Likewise.
20314         * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20315         * config/sh/sh.md (casesi_worker_1): Likewise.
20316         (casesi_worker_2): Likewise.
20317         * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20318         (fcc_comparison_operator): Likewise.
20319         * config/sparc/sparc.c (sparc_expand_move): Likewise.
20320         (emit_soft_tfmode_cvt): Likewise.
20321         (sparc_preferred_simd_mode): Likewise.
20322         (output_cbranch): Likewise.
20323         (sparc_print_operand): Likewise.
20324         (sparc_expand_vec_perm_bmask): Likewise.
20325         (vector_init_bshuffle): Likewise.
20326         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20327         (spu_vector_mode_supported_p): Likewise.
20328         (spu_expand_insv): Likewise.
20329         (spu_emit_branch_or_set): Likewise.
20330         (spu_handle_vector_attribute): Likewise.
20331         (spu_builtin_splats): Likewise.
20332         (spu_builtin_extract): Likewise.
20333         (spu_builtin_promote): Likewise.
20334         (spu_expand_sign_extend): Likewise.
20335         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20336         (tilegx_simd_int): Likewise.
20337         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20338         (tilepro_simd_int): Likewise.
20339         * config/v850/v850.c (const_double_split): Likewise.
20340         (v850_print_operand): Likewise.
20341         (ep_memory_offset): Likewise.
20342         * config/vax/vax.c (vax_rtx_costs): Likewise.
20343         (vax_output_int_move): Likewise.
20344         (vax_output_int_add): Likewise.
20345         (vax_output_int_subtract): Likewise.
20346         * config/visium/predicates.md (visium_branch_operator): Likewise.
20347         * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20348         (visium_print_operand_address): Likewise.
20349         * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20350         * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20351         (xtensa_expand_conditional_branch): Likewise.
20352         (xtensa_copy_incoming_a7): Likewise.
20353         (xtensa_output_literal): Likewise.
20354         * dfp.c (decimal_real_maxval): Likewise.
20355         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20357 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20358             Alan Hayward  <alan.hayward@arm.com>
20359             David Sherwood  <david.sherwood@arm.com>
20361         * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20362         (mode_nunits_inline): Likewise.
20363         (mode_inner_inline): Likewise.
20364         (mode_unit_size_inline): Likewise.
20365         (mode_unit_precision_inline): Likewise.
20366         (emit_insn_modes_h): Likewise.  Also emit a #define of the
20367         unprefixed name.
20368         (emit_mode_wider): Add an E_ prefix to mode names.
20369         (emit_mode_complex): Likewise.
20370         (emit_mode_inner): Likewise.
20371         (emit_mode_adjustments): Likewise.
20372         (emit_mode_int_n): Likewise.
20373         * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20374         (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20375         (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20376         (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20377         (CRC32_BUILTIN, ENTRY): Likewise.
20378         * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20379         (aarch64_pop_regs): Likewise.
20380         (aarch64_process_components): Likewise.
20381         * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20382         * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20383         (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20384         (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20385         * config/arm/arm.c (arm_init_libfuncs): Likewise.
20386         * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20387         Likewise.
20388         * config/i386/i386-builtin.def (pcmpestr): Likewise.
20389         (pcmpistr): Likewise.
20390         * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20391         * config/mmix/mmix.c (mmix_output_condition): Likewise.
20392         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20393         Likewise.
20394         * config/rl78/rl78.c (mduc_regs): Likewise.
20395         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20396         (htm_expand_builtin): Likewise.
20397         * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20398         * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20399         * config/xtensa/xtensa.c (print_operand): Likewise.
20400         * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20401         (NUM_MODE_VECTOR_INT): Likewise.
20402         * genoutput.c (null_operand): Likewise.
20403         (output_operand_data): Likewise.
20404         * genrecog.c (print_parameter_value): Likewise.
20405         * lra.c (debug_operand_data): Likewise.
20407 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20409         * dwarf2out.c (scompare_loc_descriptor_wide)
20410         (scompare_loc_descriptor_narrow): New functions, split out from...
20411         (scompare_loc_descriptor): ...here.
20412         * expmed.c (emit_store_flag_int): New function, split out from...
20413         (emit_store_flag): ...here.
20415 2017-08-30  Richard Biener  <rguenther@suse.de>
20417         * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20418         (dwarf2out_early_finish): Move setting of AT_pubnames from
20419         early debug output to early finish.
20421 2017-08-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20423         * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20424         and -mdata-region to the assembler.
20426 2017-08-30  Richard Biener  <rguenther@suse.de>
20428         * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20429         attributes.
20430         (gen_subprogram_die): Add DW_AT_object_pointer only early.
20431         (dwarf2out_early_global_decl): Only generate a DIE for the
20432         abstract origin if it doesn't already exist or is a declaration DIE.
20433         (resolve_addr): Do not add the linkage name twice when
20434         generating a stub DIE for the DW_TAG_GNU_call_site target.
20436 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20438         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20439         Use machine_mode rather than int for arg1_mode.
20441 2017-08-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20443         PR target/82015
20444         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20445         that the second argument of the built-in functions to unpack
20446         128-bit scalar types to 64-bit values is 0 or 1.  Change to use a
20447         switch statement instead a lot of if statements.
20448         * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20449         Allow 64-bit values to be in Altivec registers as well as
20450         traditional floating point registers.
20451         (pack<mode>, FMOVE128_VSX iterator): Likewise.
20453 2017-08-29  Alexander Monakov  <amonakov@ispras.ru>
20455         * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20456         MAX_REGS_PER_ADDRESS == 1.
20458 2017-08-29  Uros Bizjak  <ubizjak@gmail.com>
20460         * config/i386/i386.opt (flag_fentry): Do not init to -1.
20461         * config/i386/i386.c (ix86_option_override_internal): Simplify
20462         setting of opts->x_flag_entry.
20464 2017-08-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20465             Jakub Jelinek  <jakub@redhat.com>
20466             Richard Biener  <rguenther@suse.de>
20468         PR tree-optimization/81503
20469         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20470         folded constant fits in the target type; reorder tests for clarity.
20472 2017-08-29  Martin Liska  <mliska@suse.cz>
20474         * passes.def: Include pass_lower_switch.
20475         * stmt.c (dump_case_nodes): Remove and move to
20476         tree-switch-conversion.
20477         (case_values_threshold): Likewise.
20478         (expand_switch_as_decision_tree_p): Likewise.
20479         (emit_case_decision_tree): Likewise.
20480         (expand_case): Likewise.
20481         (balance_case_nodes): Likewise.
20482         (node_has_low_bound): Likewise.
20483         (node_has_high_bound): Likewise.
20484         (node_is_bounded): Likewise.
20485         (emit_case_nodes): Likewise.
20486         (struct simple_case_node): New struct.
20487         (add_case_node): Remove.
20488         (emit_case_dispatch_table): Use vector instead of case_list.
20489         (reset_out_edges_aux): Remove.
20490         (compute_cases_per_edge): Likewise.
20491         (expand_case): Build list of simple_case_node.
20492         (expand_sjlj_dispatch_table): Use it.
20493         * tree-switch-conversion.c (struct case_node): Moved from
20494         stmt.c and adjusted.
20495         (emit_case_nodes): Likewise.
20496         (node_has_low_bound): Likewise.
20497         (node_has_high_bound): Likewise.
20498         (node_is_bounded): Likewise.
20499         (case_values_threshold): Likewise.
20500         (reset_out_edges_aux): Likewise.
20501         (compute_cases_per_edge): Likewise.
20502         (add_case_node): Likewise.
20503         (dump_case_nodes): Likewise.
20504         (balance_case_nodes): Likewise.
20505         (expand_switch_as_decision_tree_p): Likewise.
20506         (emit_jump): Likewise.
20507         (emit_case_decision_tree): Likewise.
20508         (try_switch_expansion): Likewise.
20509         (do_jump_if_equal): Likewise.
20510         (emit_cmp_and_jump_insns): Likewise.
20511         (fix_phi_operands_for_edge): New function.
20512         (record_phi_operand_mapping): Likewise.
20513         (class pass_lower_switch): New pass.
20514         (pass_lower_switch::execute): New function.
20515         (make_pass_lower_switch): Likewise.
20516         (conditional_probability):
20517         * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20518         * tree-pass.h: Add make_pass_lower_switch.
20520 2017-08-29  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20522         PR target/80993
20523         * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20524         handlers as used.
20526 2017-08-29  Richard Biener  <rguenther@suse.de>
20528         * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20529         we do not add a DW_AT_inline attribute twice.
20530         (gen_subprogram_die): Remove code setting DW_AT_inline on
20531         DECL_ABSTRACT_P nodes.
20533 2017-08-29  Richard Sandiford  <richard.sandiford@linaro.org>
20535         * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20536         calls to internal functions.
20537         (gimplify_modify_expr): Likewise.
20538         * tree-call-cdce.c (use_internal_fn): Likewise.
20539         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20540         (convert_to_divmod): Set the nothrow flag.
20541         * tree-if-conv.c (predicate_mem_writes):  Likewise.
20542         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20543         (vectorizable_call): Likewise.
20544         (vectorizable_store): Likewise.
20545         (vectorizable_load): Likewise.
20546         * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20547         (vect_recog_mask_conversion_pattern): Likewise.
20549 2017-08-29  Martin Liska  <mliska@suse.cz>
20551         PR other/39851
20552         * gcc.c (driver_handle_option): Add new argument.
20553         * opts-common.c (handle_option): Pass
20554         target_option_override_hook.
20555         * opts-global.c (lang_handle_option): Add new option.
20556         (set_default_handlers):  Add new argument.
20557         (decode_options): Likewise.
20558         * opts.c (target_handle_option): Likewise.
20559         (common_handle_option): Call target_option_override_hook.
20560         * opts.h (struct cl_option_handler_func): Add hook for
20561         target option override.
20562         (struct cl_option_handlers): Likewise.
20563         (set_default_handlers): Add new argument.
20564         (decode_options): Likewise.
20565         (common_handle_option): Likewise.
20566         (target_handle_option): Likewise.
20567         * toplev.c (toplev::main): Pass targetm.target_option.override
20568         hook.
20570 2017-08-29  Richard Biener  <rguenther@suse.de>
20571         Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20573         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20574         life to the active subtree.
20576 2017-08-28  Jeff Law  <law@redhat.com>
20578         * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20579         derive_equivalences.
20580         (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20581         Code moved into....
20582         (edge_info::derive_equivalences): New private member function
20584         * tree-ssa-dom.c (class edge_info): Changed from a struct
20585         to a class.  Add ctor/dtor, methods and data members.
20586         (edge_info::edge_info): Renamed from allocate_edge_info.
20587         Initialize additional members.
20588         (edge_info::~edge_info): New.
20589         (free_dom_edge_info): Delete the edge info.
20590         (record_edge_info): Use new class & associated member functions.
20591         Tighten forms for testing for edge equivalences.
20592         (record_temporary_equivalences): Iterate over the simple
20593         equivalences rather than assuming there's only one per edge.
20594         (cprop_into_successor_phis): Iterate over the simple
20595         equivalences rather than assuming there's only one per edge.
20596         (optimize_stmt): Use operand_equal_p rather than pointer
20597         equality for mini-DSE code.
20599 2017-08-28  Nathan Sidwell  <nathan@acm.org>
20601         * gcc.c (execute): Fold SIGPIPE handling into switch
20602         statement.  Adjust internal error message.
20604 2017-08-28  Richard Biener  <rguenther@suse.de>
20606         PR debug/81993
20607         * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20608         Do nothing for removed DIEs.
20610 2017-08-28  Richard Biener  <rguenther@suse.de>
20612         PR tree-optimization/81977
20613         * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20614         memcpy.
20616 2017-08-28  Alexander Monakov  <amonakov@ispras.ru>
20618         PR target/80640
20619         * doc/md.texi (mem_thread_fence): Remove mention of mode.  Rewrite.
20620         * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20621         using targetm.gen_mem_thread_fence.
20623 2017-08-27  Uros Bizjak  <ubizjak@gmail.com>
20625         PR target/81995
20626         * config/i386/i386.md (*<btsc><mode>): Change operand 2
20627         predicate to register_operand.  Reorder operands.
20628         (*btr<mode>): Ditto.
20629         (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20630         (*btr<mode>_mask): Ditto.
20632 2017-08-25  Steven Munroe  <munroesj@gcc.gnu.org>
20634         * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20635         * config/rs6000/xmmintrin.h: New file.
20636         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20638 2017-08-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20640         PR target/81504
20641         * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20642         parameter and_insn and return it.
20643         (recombine_lvx_pattern): Insert a copy to ensure availability of
20644         the base register of the copied masking operation at the point of
20645         the instruction replacement.
20646         (recombine_stvx_pattern): Likewise.
20648 2017-08-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
20650         * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20651         undocumented switches.
20652         (-mpower9-dform-vector): Likewise.
20653         (-mpower9-dform): Likewise.
20654         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20655         comments to delete references to -mpower9-dform* switches.
20656         * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20657         Delete reference to -mpower9-dform* switches, test for
20658         -mpower9-vector instead.
20659         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20660         (OTHER_P9_VECTOR_MASKS): Likewise.
20661         (POWERPC_MASKS): Likewise.
20662         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20663         tests against -mpower9-dform* to -mpower9-vector.  Delete code
20664         that checked for -mpower9-dform* consistancy with other options.
20665         Add test for -mpower9-misc to enable other power9 switches.
20666         (rs6000_init_hard_regno_mode_ok): Likewise.
20667         (rs6000_option_override_internal): Likewise.
20668         (rs6000_emit_prologue): Likewise.
20669         (rs6000_emit_epilogue): Likewise.
20670         (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20671         (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20672         (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20673         -mpower9-vector.
20674         (emit_fusion_p9_store): Likewise.
20675         * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20676         resetting these macros if the assembler does not support ISA 3.0
20677         instructions.
20678         (TARGET_P9_DFORM_VECTOR): Likewise.
20679         * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20680         Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20682 2017-08-25  Alan Modra  <amodra@gmail.com>
20684         PR rtl-optimization/81747
20685         * cse.c (cse_extended_basic_block): Don't attempt to record
20686         equivalences for degenerate conditional branches that branch
20687         to their fall-through.
20689 2017-08-24  Martin Sebor  <msebor@redhat.com>
20691         PR middle-end/81908
20692         * gimple-fold.c (size_must_be_zero_p): New function.
20693         (gimple_fold_builtin_memory_op): Call it.
20695 2017-08-24  Steven Munroe  <munroesj@gcc.gnu.org>
20697         * config/rs6000/mm_malloc.h: New file.
20699 2017-08-24  Bin Cheng  <bin.cheng@arm.com>
20701         PR tree-optimization/81913
20702         * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20703         analysis when either IVs in condition can wrap.
20705 2017-08-24  Uros Bizjak  <ubizjak@gmail.com>
20707         * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20708         * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20710 2017-08-24  Richard Biener  <rguenther@suse.de>
20712         PR target/81921
20713         * targhooks.c (default_target_can_inline_p): Properly
20714         use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20715         is present and always compare.
20716         * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20717         infer -mfpmath=sse from TARGET_SSE_P.
20718         (ix86_can_inline_p): Properly use target_option_default_node when
20719         no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20721 2017-08-24  Richard Biener  <rguenther@suse.de>
20723         PR debug/81936
20724         * dwarf2out.c (output_die): Handle flag_generate_offload like
20725         flag_generate_lto.
20726         (output_comp_unit): Likewise.
20727         (gen_array_type_die): Likewise.
20728         (dwarf2out_early_finish): Likewise.
20729         (note_variable_value_in_expr): Likewise.
20730         (dwarf2out_finish): Likewise.  Adjust assert.
20731         * cgraphunit.c (symbol_table::compile): Move setting of
20732         flag_generate_offload earlier ...
20733         (symbol_table::finalize_compilation_unit): ... here, before
20734         early debug finalization.
20736 2017-08-24  Richard Biener  <rguenther@suse.de>
20738         * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20739         and ipa-fnsummary.h.
20740         (ix86_can_inline_p): When ix86_fpmath flags do not match
20741         check whether the callee uses FP math at all.
20743 2017-08-24  Aldy Hernandez  <aldyh@redhat.com>
20745         PR middle-end/81931
20746         * tree-ssanames.c (get_nonzero_bits): Use element_precision
20747         instead of TYPE_PRECISION.
20749 2017-08-24  Richard Sandiford  <richard.sandiford@linaro.org>
20750             Alan Hayward  <alan.hayward@arm.com>
20751             David Sherwood  <david.sherwood@arm.com>
20753         * combine.c (make_extraction): Use subreg_offset_from_lsb.
20755 2017-08-23  Daniel Santos  <daniel.santos@pobox.com>
20757         * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20758         Remove field.
20759         (ix86_frame::stack_realign_allocate): New field.
20760         (struct machine_frame_state): Modify comments.
20761         (machine_frame_state::sp_realigned_fp_end): New field.
20762         * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20763         layout calculation.
20764         (sp_valid_at): Add assertion to assure no attempt to access invalid
20765         offset of a realigned stack.
20766         (fp_valid_at): Likewise.
20767         (choose_baseaddr): Modify comments.
20768         (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20769         ix86_expand_prologue.
20770         (ix86_expand_prologue): Modify stack realignment and allocation.
20771         (ix86_expand_epilogue): Modify comments.
20772         * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20773         avx2_runtime, avx512f, and avx512f_runtime.
20775 2017-08-23  Uros Bizjak  <ubizjak@gmail.com>
20777         * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20778         (mstackrealign): Do not init to -1.
20779         * config/i386/i386.c (ix86_option_override_internal):
20780         Check opts_set, not opts when setting default value of
20781         opts->x_ix86_force_align_arg_pointer.
20783 2017-08-23  Richard Biener  <rguenther@suse.de>
20785         * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20786         lang_hooks.decl_printable_name.
20787         * print-rtl-function.c (print_rtx_function): Likewise.
20788         * tree-pretty-print.c (dump_function_header): Likewise.
20790 2017-08-23  Richard Biener  <rguenther@suse.de>
20792         PR lto/81940
20793         * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20794         -g0 at compile-time.
20796 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20798         PR middle-end/19706
20799         * doc/sourcebuild.texi (Other hardware attributes):
20800         Document xorsign.
20802 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20804         PR middle-end/19706
20805         * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20806         Fix single-use check.
20808 2017-08-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20810         * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20812 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20814         * doc/install.texi: Modify to add more details on running selected
20815         tests.
20817 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20819         * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20820         is combined with -mabi=ms.
20821         (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20822         ms_abi.
20824 2017-08-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20826         PR tree-optimization/81488
20827         * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20828         and cached_basis fields.
20829         (MAX_SPREAD): New constant.
20830         (alloc_cand_and_find_basis): Initialize new fields.
20831         (clear_visited): New function.
20832         (create_phi_basis_1): Rename from create_phi_basis, set visited
20833         and cached_basis fields.
20834         (create_phi_basis): New wrapper function.
20835         (phi_add_costs_1): Rename from phi_add_costs, add spread
20836         parameter, set visited field, short-circuit when limits reached.
20837         (phi_add_costs): New wrapper function.
20838         (record_phi_increments_1): Rename from record_phi_increments, set
20839         visited field.
20840         (record_phi_increments): New wrapper function.
20841         (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20842         (phi_incr_cost): New wrapper function.
20843         (all_phi_incrs_profitable_1): Rename from
20844         all_phi_incrs_profitable, set visited field.
20845         (all_phi_incrs_profitable): New wrapper function.
20847 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20848             Alan Hayward  <alan.hayward@arm.com>
20849             David Sherwood  <david.sherwood@arm.com>
20851         * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20852         that takes the outer and inner modes.
20853         * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20854         comparison as the canonical test for a paradoxical subreg.
20855         * combine.c (simplify_set): Use paradoxical_subreg_p.
20856         (make_extraction): Likewise.
20857         (force_to_mode): Likewise.
20858         (rtx_equal_for_field_assignment_p): Likewise.
20859         (gen_lowpart_for_combine): Likewise.
20860         (simplify_comparison): Likewise.
20861         * cse.c (equiv_constant): Likewise.
20862         * expmed.c (store_bit_field_1): Likewise.
20863         * final.c (alter_subreg): Likewise.
20864         * fwprop.c (propagate_rtx): Likewise.
20865         (forward_propagate_subreg): Likewise.
20866         * ira-conflicts.c (ira_build_conflicts): Likewise.
20867         * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20868         * lra-constraints.c (curr_insn_transform): Likewise.
20869         (split_reg): Likewise.
20870         * lra-eliminations.c (move_plus_up): Likewise.
20871         (lra_eliminate_regs_1): Likewise.
20872         * recog.c (general_operand): Likewise.
20873         * ree.c (combine_reaching_defs): Likewise.
20874         * reload.c (push_reload): Likewise.
20875         (find_reloads): Likewise.
20876         * reload1.c (elimination_effects): Likewise.
20877         (compute_reload_subreg_offset): Likewise.
20878         (choose_reload_regs): Likewise.
20879         * rtlanal.c (subreg_lsb_1): Likewise.
20880         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20881         (simplify_subreg): Likewise.
20882         * var-tracking.c (track_loc_p): Likewise.
20883         * emit-rtl.c (byte_lowpart_offset): Likewise.
20884         (paradoxical_subreg_p): Delete out-of-line definition.
20886 2017-08-22  Jeff Law  <law@redhat.com>
20888         PR tree-optimization/81741
20889         PR tree-optimization/71947
20890         * tree-ssa-dom.c: Include tree-inline.h.
20891         (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20892         equivalences if one is more expensive to compute than the other.
20893         * tree-ssa-scopedtables.h (class const_or_copies): Make
20894         record_const_or_copy_raw method private.
20895         (class avail_exprs_stack): New method simplify_binary_operation.
20896         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20897         avail_exprs_stack::simplify_binary_operation as needed.
20898         (avail_exprs_stack::simplify_binary_operation): New function.
20900 2017-08-22  Sebastian Huber  <sebastian.huber@embedded-brains.de>
20902         * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20903         * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20904         (DOT_SYMBOLS): Likewise.
20905         (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20906         (RELOCATABLE_NEEDS_FIXUP): Likewise.
20907         (RS6000_ABI_NAME): Likewise.
20908         (TARGET_CMODEL): Likewise.
20909         (TOC_SECTION_ASM_OP): Likewise.
20910         (SET_CMODEL): New macro.
20911         (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
20913 2017-08-22  Richard Biener  <rguenther@suse.de>
20915         * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
20916         to allow for free-lang-data replacements similar to verify_type_variant.
20918 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20919             Alan Hayward  <alan.hayward@arm.com>
20920             David Sherwood  <david.sherwood@arm.com>
20922         * config/aarch64/aarch64.md (casesi): Use DImode rather than
20923         VOIDmode for the LABEL_REF.
20925 2017-08-22  Richard Biener  <rguenther@suse.de>
20927         * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
20929 2017-08-22  Richard Biener  <rguenther@suse.de>
20931         * common.opt (feliminate-dwarf2-dups): Ignore.
20932         * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
20933         * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
20934         same_die_p_wrap, compute_section_prefix,
20935         is_symbol_die, assign_symbol_names, break_out_includes): Remove.
20936         (comdat_symbol_id, comdat_symbol_number): Likewise.
20937         (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
20938         Likewise.
20939         (check_duplicate_cu, record_comdat_symbol_number): Likewise.
20940         (output_die): Mark unreachable path unreachable.
20941         (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
20942         (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
20943         (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
20944         (dwarf2out_early_finish): Likewise.
20946 2017-08-22  Aldy Hernandez  <aldyh@redhat.com>
20948         * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
20950 2017-08-22  Georg-Johann Lay  <avr@gjlay.de>
20952         PR target/81910
20953         * config/avr/avr.c (avr_handle_addr_attribute): Early return if
20954         not VAR_P. Filter attribute warnings with OPT_Wattributes.
20955         (avr_attribute_table) <io, io_low, address>: Initialize
20956         .decl_required with true.
20958 2017-08-21  Michael Meissner  <meissner@linux.vnet.ibm.com>
20960         * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
20961         undocumented debugging options.
20962         (-mvsx-scalar-double): Likewise.
20963         (-mallow-df-permute): Likewise.
20964         (-mvectorize-builtins): Likewise.
20965         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20966         (rs6000_builtin_vectorized_function): Likewise.
20967         (rs6000_builtin_md_vectorized_function): Likewise.
20968         (rs6000_opt_vars): Likewise.
20970 2017-08-21  Uros Bizjak  <ubizjak@gmail.com>
20972         PR target/46091
20973         * config/i386/i386.md (*btsq_imm): Rename from *btsq.
20974         (*btrq_imm): Rename from *btrq.
20975         (*btcq_imm): Rename from *btcq.
20976         (btsc): New code attribute.
20977         (*<btsc><mode>): New insn pattern.
20978         (*btr<mode>): Ditto.
20979         (*<btsc><mode>_mask): New insn_and_split pattern.
20980         (*btr<mode>_mask): Ditto.
20982 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20983             Alan Hayward  <alan.hayward@arm.com>
20984             David Sherwood  <david.sherwood@arm.com>
20986         * function.c (pad_below): Simplify padding calculation.
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         * target.def (function_prologue): Remove frame size argument.
20993         (function_epilogue): Likewise.
20994         * doc/tm.texi: Regenerate.
20995         * final.c (final_start_function): Update call to function_prologue.
20996         (final_end_function): Update call to function_epilogue.
20997         (default_function_pro_epilogue): Remove frame size argument.
20998         * output.h (default_function_pro_epilogue): Likewise.
20999         * config/arm/arm.c (arm_output_function_epilogue): Likewise.
21000         (arm_output_function_prologue): Likewise.
21001         * config/frv/frv.c (frv_function_prologue): Likewise.
21002         (frv_function_epilogue): Likewise.
21003         * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
21004         * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
21005         (ia64_output_function_epilogue): Likewise.
21006         * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
21007         (m32r_output_function_epilogue): Likewise.
21008         * config/microblaze/microblaze.c (microblaze_function_prologue)
21009         (microblaze_function_epilogue): Likewise.
21010         * config/mips/mips.c (mips_output_function_prologue): Likewise.
21011         (mips_output_function_epilogue): Likewise.
21012         * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
21013         (mmix_target_asm_function_epilogue): Likewise.
21014         * config/msp430/msp430.c (msp430_start_function): Likewise.
21015         * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
21016         (nds32_asm_function_epilogue): Likewise.
21017         * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
21018         * config/pa/pa.c (pa_output_function_prologue): Likewise.
21019         (pa_output_function_epilogue): Likewise.
21020         * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
21021         (rs6000_output_function_epilogue): Likewise.
21022         * config/rl78/rl78.c (rl78_start_function): Likewise.
21023         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
21024         (rs6000_output_function_epilogue): Likewise.
21025         * config/rx/rx.c (rx_output_function_prologue): Likewise.
21026         * config/sh/sh.c (sh_output_function_epilogue): Likewise.
21027         * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
21028         (sparc_asm_function_epilogue): Likewise.
21030 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21032         * tree.h (type_has_mode_precision_p): New function.
21033         * convert.c (convert_to_integer_1): Use it.
21034         * expr.c (expand_expr_real_2): Likewise.
21035         (expand_expr_real_1): Likewise.
21036         * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
21037         * match.pd: Likewise.
21038         * tree-ssa-forwprop.c (simplify_rotate): Likewise.
21039         * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
21040         * tree-tailcall.c (process_assignment): Likewise.
21041         * tree-vect-loop.c (vectorizable_reduction): Likewise.
21042         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
21043         (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
21044         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
21045         (vectorizable_assignment): Likewise.
21046         (vectorizable_shift): Likewise.
21047         (vectorizable_operation): Likewise.
21048         * tree-vrp.c (register_edge_assert_for_2): Likewise.
21050 2017-08-21  Wilco Dijkstra  <wdijkstr@arm.com>
21052         * match.pd: Add pow (C, x) simplification.
21054 2017-08-21  Richard Biener  <rguenther@suse.de>
21056         PR tree-optimization/81900
21057         * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
21058         for blocks with abnormal predecessors.
21059         (compute_antic): Do not set visited flag prematurely.
21061 2017-08-21  Georg-Johann Lay  <avr@gjlay.de>
21063         PR target/79883
21064         * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
21066 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21068         * stor-layout.h (vector_type_mode): Move to...
21069         * tree.h (vector_type_mode): ...here.
21070         * stor-layout.c (vector_type_mode): Move to...
21071         * tree.c (vector_type_mode): ...here.  Include rtl.h and regs.h.
21073 2017-08-21  Richard Biener  <rguenther@suse.de>
21075         * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
21076         register_external_die hooks.
21077         (debug_false_tree_charstarstar_uhwistar): Declare.
21078         (debug_nothing_tree_charstar_uhwi): Likewise.
21079         * debug.c (do_nothing_debug_hooks): Adjust.
21080         (debug_false_tree_charstarstar_uhwistar): New do nothing.
21081         (debug_nothing_tree_charstar_uhwi): Likewise.
21082         * dbxout.c (dbx_debug_hooks): Adjust.
21083         (xcoff_debug_hooks): Likewise.
21084         * sdbout.c (sdb_debug_hooks): Likewise.
21085         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
21086         * dwarf2out.c (macinfo_label_base): New global.
21087         (dwarf2out_register_external_die): New function for the
21088         register_external_die hook.
21089         (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
21090         (dwarf2_debug_hooks): Use them.
21091         (dwarf2_lineno_debug_hooks): Adjust.
21092         (struct die_struct): Add with_offset flag.
21093         (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
21094         DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
21095         DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
21096         DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
21097         DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
21098         DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
21099         defining section names for the early LTO debug variants.
21100         (reset_indirect_string): New helper.
21101         (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
21102         (print_dw_val): Add support for offsetted symbol references.
21103         (get_ultimate_context): Split out from is_cxx.
21104         (is_cxx): Use get_ultimate_context.
21105         (is_fortran): Add decl overload.
21106         (compute_comp_unit_symbol): Split out worker from
21107         compute_section_prefix.
21108         (compute_section_prefix): Call compute_comp_unit_symbol and
21109         set comdat_type_p here.
21110         (output_die): Skip DIE symbol output for the LTO added one.
21111         Handle DIE symbol references with offset.
21112         (output_comp_unit): Guard section name mangling properly.
21113         For LTO debug sections emit a symbol at the section beginning
21114         which we use to refer to its DIEs.
21115         (add_abstract_origin_attribute): For DIEs registered via
21116         dwarf2out_register_external_die directly refer to the early
21117         DIE rather than indirectly through the shadow one we created.
21118         Remove obsolete call to dwarf2out_abstract_function for
21119         non-function/block origins.
21120         (gen_array_type_die): When generating early LTO debug do
21121         not emit DW_AT_string_length.
21122         (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
21123         late when in LTO.  As suggested place a gcc_unreachable for
21124         the DECL_ABSTRACT_P case.
21125         (gen_subprogram_die): Avoid another specification DIE
21126         for early built declarations/definitions for the late LTO case.
21127         (gen_variable_die): Add type references for late duplicated VLA dies
21128         when in late LTO.
21129         (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
21130         we have the abstract instance already.
21131         (process_scope_var): Adjust decl DIE contexts in LTO which
21132         first puts them in limbo.
21133         (gen_decl_die): Do not generate type DIEs late apart from
21134         types for VLAs or for decls we do not yet have a DIE.  Do not
21135         call dwarf2out_abstract_function late.
21136         (dwarf2out_early_global_decl): Make sure to create DIEs
21137         for abstract instances of a decl first.
21138         (dwarf2out_late_global_decl): Adjust comment.
21139         (output_macinfo_op): With multiple macro sections use
21140         macinfo_label_base to distinguish labels.
21141         (output_macinfo): Likewise.  Update macinfo_label_base.
21142         Pass in the line info label.
21143         (note_variable_value_in_expr): When generating LTO resolve
21144         all variable values here by generating DIEs as needed.
21145         (init_sections_and_labels): Add early LTO debug flag parameter
21146         and generate different sections and names if set.  Add generation
21147         counter for the labels so we can have multiple of them.
21148         (reset_dies): Helper to allow DIEs to be output multiple times.
21149         (dwarf2out_finish): When outputting DIEs to the fat part of an
21150         LTO object first reset DIEs.
21151         (dwarf2out_early_finish): Output early DIEs when generating LTO.
21152         (modified_type_die): Check for decl_ultimate_origin being self
21153         before recursing.
21154         (gen_type_die_with_usage): Likewise.
21155         (gen_typedef_die): Allow decl_ultimate_origin being self.
21156         (set_decl_abstract_flags): Remove.
21157         (set_block_abstract_flags): Likewise.
21158         (dwarf2out_abstract_function): Treat the early generated DIEs
21159         as the abstract copy and only add DW_AT_inline and
21160         DW_AT_artificial here and call set_decl_origin_self.
21161         If the DIE has an abstract origin don't do anything.
21162         * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
21163         if we have none yet (Go fails to build one, PR78628).
21164         (variably_modified_type_p): Prevent endless recursion for Ada
21165         cyclic pointer types.
21166         * lto-streamer-in.c: Include debug.h.
21167         (dref_queue): New global.
21168         (lto_read_tree_1): Stream in DIE references.
21169         (lto_input_tree): Register DIE references.
21170         (input_function): Stream DECL_DEBUG_ARGS.
21171         * lto-streamer-out.c: Include debug.h.
21172         (lto_write_tree_1): Output DIE references.
21173         (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
21174         Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
21175         (output_function): Stream DECL_DEBUG_ARGS.
21176         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21177         Stream DECL_ABSTRACT_ORIGIN.
21178         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21179         (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
21180         DECL_CONTEXT for file-scope decls.
21181         * lto-streamer.h (struct dref_entry): Declare.
21182         (dref_queue): Likewise.
21183         * cfgexpand.c (pass_expand::execute): Do not call the
21184         outlining_inline_function hook here.
21185         * lto-wrapper.c (debug_obj): New global.
21186         (tool_cleanup): Unlink it if required.
21187         (debug_objcopy): New function.
21188         (run_gcc): Handle early debug sections in the IL files by
21189         extracting them to separate files, partially linkin them and
21190         feeding the result back as result to the linker.
21191         * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
21192         DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
21193         DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
21194         sections into a separate segment.
21195         * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
21196         segments.
21197         (darwin_asm_dwarf_section): Likewise.
21198         (darwin_asm_output_dwarf_offset): Likewise.
21199         * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
21201 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21202             Alan Hayward  <alan.hayward@arm.com>
21203             David Sherwood  <david.sherwood@arm.com>
21205         * read-md.h (md_reader::record_potential_iterator_use): Replace
21206         pointer argument with an rtx and an index.
21207         * read-rtl.c (iterator_group::apply_iterator): Likewise.
21208         (apply_mode_iterator): Likewise.
21209         (apply_code_iterator): Likewise.
21210         (apply_int_iterator): Likewise.
21211         (apply_subst_iterator): Likewise.
21212         (record_iterator_use): Likewise.
21213         (record_attribute_use): Likewise.
21214         (md_reader::record_potential_iterator_use): Likewise.  Update calls
21215         to record_iterator_use and apply_iterator.
21216         (iterator_use): Replace ptr with x and index.
21217         (attribute_use): Likewise.
21218         (apply_attribute_uses): Update calls to apply_iterator.
21219         (apply_iterators): Likewise.  Update initialization of iterator_use.
21220         (rtx_reader::read_rtx_code): Update calls to record_iterator_use
21221         and record_potential_iterator_use.
21222         (rtx_reader::read_rtx_operand): Likewise.
21224 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21225             Alan Hayward  <alan.hayward@arm.com>
21226             David Sherwood  <david.sherwood@arm.com>
21228         * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
21229         CONST_WIDE_INT.
21231 2017-08-21  Richard Biener  <rguenther@suse.de>
21233         PR middle-end/81884
21234         * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
21235         at struct end conservatively when comparing common bases.
21237 2017-08-21  Richard Biener  <rguenther@suse.de>
21239         * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21240         (mem_ref_in_stmt): Remove.
21241         (determine_max_movement): Use ref index to get at the reference.
21242         (invariantness_dom_walker::before_dom_children): Deal with
21243         lim data already initialized.
21244         (gather_mem_refs_stmt): Initialize lim data and record ref index.
21246 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21248         * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21249         (TARGET_ISA_ROUND): Ditto.
21250         (TARGET_ROUND): Ditto.
21251         * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21252         * config/i386/i386.md: Ditto.
21253         * config/i386/sse.md: Ditto.
21254         * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21255         with OPTION_MASK_ISA_SSE4_1.
21257 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21259         PR target/81894
21260         * doc/extend.texi (x86 Built-in Functions): Correct the name of
21261         __builtin_ia32_lzcnt_u16.
21263 2017-08-18  Peter Bergner  <bergner@vnet.ibm.com>
21265         PR target/80210
21266         * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21267         (rs6000_set_current_function): Rewrite function to use it.
21269 2017-08-18  H.J. Lu  <hongjiu.lu@intel.com>
21271         PR c/53037
21272         * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21273         and TYPE_WARN_IF_NOT_ALIGN.
21274         * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21275         (handle_warn_if_not_align): New.
21276         (place_union_field): Call handle_warn_if_not_align.
21277         (place_field): Call handle_warn_if_not_align.
21278         Copy TYPE_WARN_IF_NOT_ALIGN.
21279         (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21280         (layout_type): Likewise.
21281         * tree-core.h (tree_type_common): Add warn_if_not_align.  Set
21282         spare to 18.
21283         (tree_decl_common): Add warn_if_not_align.
21284         * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21285         * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21286         (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21287         (DECL_WARN_IF_NOT_ALIGN): Likewise.
21288         (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21289         * doc/extend.texi: Document warn_if_not_aligned attribute.
21290         * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21292 2017-08-17  Martin Liska  <mliska@suse.cz>
21294         PR bootstrap/81864
21295         * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21296         (get_data_dependence): Use it as pointer type.
21297         (distribute_loop): Likewise.
21299 2017-08-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21301         * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21302         (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21303         (p8_vmrgow_<mode>_direct): New define_insn.
21304         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21305         handle endianness for vmrgew and vmrgow permute patterns.
21307 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21309         * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21310         * config/rs6000/rs6000-cpus.def: Remove comment.
21311         (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21312         (POWERPC_MASKS): Likewise.
21313         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21314         use of TARGET_VSX_TIMODE.
21315         (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21316         (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21317         TARGET_VSX_TIMODE.  Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21318         (rs6000_option_override_internal): Remove dead code.
21319         (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21320         (rs6000_legitimize_reload_address): Likewise.
21321         (rs6000_legitimate_address_p): Likewise.
21322         (rs6000_opt_masks): Delete "vsx-timode".
21323         (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21324         from function comment.
21325         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21326         * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21327         (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21328         condition.
21329         * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21330         * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21331         * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21332         (VSX_TI): Likewise.
21333         (VSX_M): Likewise.
21334         (define_peephole2): Likewise.
21336 2017-08-17  Martin Sebor  <msebor@redhat.com>
21338         PR c/81859
21339         * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21340         past the end of an array.
21341         (test_pp_format): Add test cases.
21343 2017-08-17  Richard Sandiford  <richard.sandiford@linaro.org>
21345         * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21346         missing ECF_NOTHROW flags.
21348 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21350         PR target/72804
21351         * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21352         operands residing in integer registers.
21353         (*vsx_le_perm_load_<mode>): Likewise.
21354         (*vsx_le_perm_store_<mode>): Likewise.
21355         (define_peephole2): Add peepholes to optimize the above.
21357 2017-08-17  Marek Polacek  <polacek@redhat.com>
21359         PR middle-end/81814
21360         * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21361         to mimic what shorten_compare did.  Change the return type to bool.
21362         (fold_cond_expr_with_comparison): Update call to
21363         operand_equal_for_comparison_p.
21364         (fold_ternary_loc): Likewise.
21366 2017-08-17  Jackson Woodruff  <jackson.woodruff@arm.com>
21368         * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21369         register.
21370         (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21372 2017-08-17  Richard Biener  <rguenther@suse.de>
21374         * tree-ssa-structalias.c (solve_graph): When propagating
21375         to successors update the graphs succ edges and avoid duplicate work.
21377 2017-08-17  Maxim Ostapenko  <m.ostapenko@samsung.com>
21379         PR target/81861
21380         * config/i386/i386.c (ix86_option_override_internal): Save target
21381         specific options after ix86_stack_protector_guard_reg was changed.
21383 2017-08-17  Richard Biener  <rguenther@suse.de>
21385         PR tree-optimization/81827
21386         * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21387         (new_var_info): Initialize it conservatively.
21388         (get_call_vi): Mark register vars.
21389         (new_scalar_tmp_constraint_exp): Likewise.
21390         (handle_rhs_call): Likewise.
21391         (handle_const_call): Likewise.
21392         (create_function_info_for): Likewise.
21393         (solve_constraints): Sort varinfos to separate register from
21394         non-register vars to pack points-to solution bitmaps during
21395         iteration.
21397 2017-08-17  Marek Polacek  <polacek@redhat.com>
21399         * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21401 2017-08-17  Richard Biener  <rguenther@suse.de>
21403         * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21404         to true when overflow is undefined and we saturated the result.
21406 2017-08-17  Alan Modra  <amodra@gmail.com>
21408         PR target/80938
21409         * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21410         Don't use store multiple if only one reg needs saving.
21411         (interesting_frame_related_regno): New function.
21412         (rs6000_frame_related): Don't emit frame info for regs that
21413         don't need saving.
21414         (rs6000_emit_epilogue): Likewise.
21416 2017-08-16  Nathan Sidwell  <nathan@acm.org>
21418         * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21419         * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21420         (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21421         * tree.c (free_lang_data_in_type): Use else-if chain.  Always
21422         clear TYPE_LANG_1.  Remove obsolete member-function stripping.
21423         (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21424         (verify_type): Adjust for TYPE_BINFO move.
21425         * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21426         process TYPE_BINFO directly.
21427         (hash_tree): Likewise.
21428         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21429         Likewise.
21430         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21431         Likewise.
21433 2017-08-16  David Malcolm  <dmalcolm@redhat.com>
21435         * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21437 2017-08-16  Uros Bizjak  <ubizjak@gmail.com>
21439         PR target/46091
21440         * config/i386/i386.md (*anddi_1_btr): Change predicates of
21441         operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21442         Add ix86_binary_operator_ok to insn constraint.
21443         (*iordi_1_bts): Ditto.
21444         (*xordi_1_btc): Ditto.
21445         (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21446         Update corresponding peephole2 pattern.
21447         (*btrq): Ditto.
21448         (*btcq): Ditto.
21450 2017-08-16  Bin Cheng  <bin.cheng@arm.com>
21452         PR tree-optimization/81832
21453         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21454         copy loop header which has IFN_LOOP_DIST_ALIAS call.
21456 2017-08-16  Marek Polacek  <polacek@redhat.com>
21458         PR middle/81695
21459         * fold-const.c (fold_indirect_ref_1): Restore original behavior
21460         regarding size_zero_node.
21462 2017-08-16  Martin Liska  <mliska@suse.cz>
21464         PR target/81753
21465         * config.gcc: Respect previously set extra_objs in case
21466         of darwin target.
21468 2017-08-16  Richard Sandiford  <richard.sandiford@linaro.org>
21470         PR tree-optimization/81835
21471         * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21472         the phi SSA_NAME.  Check that the condition in a COND_EXPR does
21473         not depend on the phi.
21475 2017-08-16  Alan Modra  <amodra@gmail.com>
21477         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21478         dead code.
21480 2017-08-16  Alan Modra  <amodra@gmail.com>
21482         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21483         (save_reg_p): ..into this.  Update all callers.
21484         (first_reg_to_save): Simplify.
21486 2017-08-16  Alan Modra  <amodra@gmail.com>
21488         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21489         fixed regs.
21491 2017-08-15  Joseph Myers  <joseph@codesourcery.com>
21493         PR target/78460
21494         PR target/67712
21495         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21496         constant count if that count is less than 32.
21498 2017-08-15  Nathan Sidwell  <nathan@acm.org>
21500         * gcc.c (execute): Emit friendlier message if inferior is killed
21501         by an external cause.
21503 2017-08-15  Richard Biener  <rguenther@suse.de>
21505         PR tree-optimization/81790
21506         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21507         CONSTRUCTORs from simplifying and VN.
21509 2017-08-14  Martin Sebor  <msebor@redhat.com>
21511         * builtin-attrs.def: Add comments.
21513 2017-08-14  Martin Sebor  <msebor@redhat.com>
21515         PR c/81117
21516         * doc/extend.texi (attribute nonstring): Document new attribute.
21518 2017-08-14  Martin Sebor  <msebor@redhat.com>
21520         PR c/81117
21521         * tree-diagnostic.c (default_tree_printer): Handle %G.
21522         * gimple-pretty-print.h (percent_G_format): Declare new function.
21523         * gimple-pretty-print.c (percent_G_format): Define.
21524         * tree-pretty-print.c (percent_K_format): Add argument.
21526 2017-08-14  Martin Sebor  <msebor@redhat.com>
21528         PR translation/79998
21529         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21530         Remove a stray space.
21532 2017-08-14  Uros Bizjak  <ubizjak@gmail.com>
21534         PR target/46091
21535         * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21536         (*iordi_1_bts): Ditto.
21537         (*xordi_1_btc): Ditto.
21539 2017-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21541         PR target/79845
21542         * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21543         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21544         Likewise.
21545         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21546         quoted strings, and make more translator-friendly.
21547         (darwin_rs6000_override_options): Likewise.
21548         (rs6000_option_override_internal): Likewise.
21549         (rs6000_return_in_memory): Fix overlong line.
21550         (init_cmulative_args): Use quoted strings, and make more
21551         translator-friendly.
21552         (rs6000_pass_by_reference): Fix overlong line.
21553         (def_builtin): Use quoted strings.
21554         (altivec_expand_predicate_builtin): Use quoted strings, and make
21555         more translator-friendly.
21556         (htm_expand_builtin): Use quoted strings.
21557         (cpu_expand_builtin): Use quoted strings, and make more
21558         translator-friendly.
21559         (altivec_expand_builtin): Likewise.
21560         (paired_expand_predicate_builtin): Likewise.
21561         (rs6000_invalid_builtin): Likewise.
21562         (builtin_function_type): Use quoted strings.
21563         (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21564         more translator-friendly.
21565         (rs6000_trampoline_init): Likewise.
21566         (rs6000_handle_altivec_attribute): Likewise.
21567         (rs6000_inner_target_options): Use quoted strings.
21568         (rs6000_disable_incompatible_switches): Likewise.
21569         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21570         strings, and make more translator-friendly.
21571         (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21573 2017-08-14  Bin Cheng  <bin.cheng@arm.com>
21575         PR tree-optimization/81799
21576         * tree-loop-distribution.c (version_loop_by_alias_check): Force
21577         cond_expr to simple gimple operand.
21579 2017-08-14  Wilco Dijkstra  <wdijkstr@arm.com>
21581         PR middle-end/46932
21582         * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21584 2017-08-14  Georg-Johann Lay  <avr@gjlay.de>
21586         PR target/81754
21587         PR target/81268
21588         * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21589         * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21590         TARGET_GASISR_PROLOGUES.
21591         * config/avr/avr.c (avr_option_override): Same.
21592         (avr_pass_pre_proep::execute): Same.
21594 2017-08-13  H.J. Lu  <hongjiu.lu@intel.com>
21596         PR target/81820
21597         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21598         frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21600 2017-08-13  Uros Bizjak  <ubizjak@gmail.com>
21602         * config/i386/i386.md (*load_tp_<mode>): Redefine as
21603         define_insn_and_split.  Split to a memory load from 0 in
21604         DEFAULT_TLS_SEG_REG address space.  Merge with *load_tp_x32
21605         using PTR mode iterator.
21606         (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21607         Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21608         (*add_tp_<mode>): Redefine as define_insn_and_split.
21609         Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21610         address space.  Merge with *add_tp_x32 using PTR mode iterator.
21611         (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21612         Split to an add with a  memory load from 0 in
21613         DEFAULT_TLS_SEG_REG address space.
21615 2017-08-12  Andrew Pinski  <apinski@cavium.com>
21617         * config/aarch64/aarch64-option-extensions.def (rdma):
21618         Fix feature string to what Linux prints out in /proc/cpuinfo.
21620 2017-08-12  Pierre-Marie de Rodat  <derodat@adacore.com>
21622         PR ada/79542
21623         * dwarf2out.c (modified_type_die): For C typedef types that have
21624         an ultimate origin, process the ultimate origin instead of the
21625         input type.
21626         (gen_typedef_die): Assert that input DECLs have no ultimate
21627         origin.
21628         (gen_type_die_with_usage): For typedef variants that have an
21629         ultimate origin, just call gen_decl_die on the original DECL.
21630         (process_scope_var): Avoid creating DIEs for local typedefs and
21631         concrete static variables.
21633 2017-08-12  Alan Modra  <amodra@gmail.com>
21635         PR target/81170
21636         PR target/81295
21637         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21638         match gnu-user.h startfile.
21639         (ENDFILE_LINUX_SPEC): Similarly.
21641 2017-08-11  Thomas Schwinge  <thomas@codesourcery.com>
21643         PR lto/81430
21644         * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21645         Remove function.
21646         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21648 2017-08-11  Tamar Christina  <tamar.christina@arm.com>
21649         * config/aarch64/aarch64.md (mov<mode>): Change.
21650         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21651         aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21652         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21654 2017-08-11  Eric Botcazou  <ebotcazou@adacore.com>
21656         * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21657         for storage order barriers.
21659 2017-08-11  Martin Liska  <mliska@suse.cz>
21661         PR tree-opt/79987
21662         * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21663         variables of void type.
21665 2017-08-11  Martin Liska  <mliska@suse.cz>
21667         * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21668         TARGET_SUPPORTS_ALIASES.
21669         * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21670         * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21671         (optimize_weakref): Likewise.
21672         * symtab.c (symtab_node::noninterposable_alias): Likewise.
21673         * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21674         * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21676 2017-08-11  Martin Liska  <mliska@suse.cz>
21678         PR ipa/81213
21679         * config/i386/i386.c (make_resolver_func): Do complete
21680         refactoring of the function.
21682 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21684         PR target/81708
21685         * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21686         * config/i386/i386.c (ix86_stack_protect_guard): Use
21687         ix86_stack_protect_guard_symbol_str to generate varible declaration.
21688         * doc/invoke.texi (x86 Options): Document
21689         -mstack-protector-guard-symbol= option.
21691 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21693         * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21694         * config/i386/i386.c (ix86_split_stack_guard): New function.
21695         (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21696         (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21697         (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21698         (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21699         * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21700         (split_stack_space_check): Call ix86_split_stack_guard.
21702 2017-08-10  Martin Sebor  <msebor@redhat.com>
21704         * print-tree.c (print_node): Print location using the established
21705         format %s:%i%i.
21706         Replace spaces with colons.
21707         (debug_raw, debug): Ditto.
21709 2017-08-10  Martin Sebor  <msebor@redhat.com>
21711         PR c++/81586
21712         * pretty-print.c (pp_format): Correct the handling of %s precision.
21714 2017-08-10  H.J. Lu  <hongjiu.lu@intel.com>
21716         PR target/81736
21717         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21718         to ...
21719         (ix86_finalize_stack_frame_flags): This.  Also clear
21720         frame_pointer_needed if -fno-omit-frame-pointer is used without
21721         stack access.
21722         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21723         with ix86_finalize_stack_frame_flags.
21724         (ix86_expand_epilogue): Likewise.
21725         (ix86_expand_split_stack_prologue): Likewise.
21726         * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21728 2017-08-10  Martin Liska  <mliska@suse.cz>
21730         PR c++/81355
21731         * c-attribs.c (handle_target_attribute):
21732         Report warning for an empty string argument of target attribute.
21734 2017-08-09  Jakub Jelinek  <jakub@redhat.com>
21736         PR c/81687
21737         * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21738         LABEL_DECLs.
21739         * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21740         or DECL_NONLOCAL labels.
21741         (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21742         or DECL_NONLOCAL labels here.
21744 2017-08-09  Will Schmidt  <will_schmidt@vnet.ibm.com>
21746         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21747         to indicate when early gimple folding has been disabled.
21748         (rs6000_gimple_fold_builtin): Add debug content.
21749         (rs6000_invalid_builtin): Fix whitespace.
21750         (rs6000_expand_builtin): Fix whitespace.
21751         * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21753 2017-08-09  Segher Boessenkool  <segher@kernel.crashing.org>
21755         PR target/80938
21756         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21757         SAVE_MULTIPLE if not all the registers that saves, should be saved.
21759 2017-08-09  Jim Wilson  <jim.wilson@linaro.org>
21761         * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21762         (qdf24xx): Likewise.
21763         * config/aarch64/aarch64.md: Include falkor.md.
21764         * config/aarch64/falkor.md: New.
21766 2017-08-09  Marek Polacek  <polacek@redhat.com>
21768         PR c/81233
21769         * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21770         * diagnostic.c (emit_diagnostic): Add a comment.
21771         (emit_diagnostic_valist): New function.
21773 2017-08-09  Marek Polacek  <polacek@redhat.com>
21775         PR c/81417
21776         * input.c (make_location): New overload.
21777         * input.h (make_location): Declare.
21779 2017-08-08  Alan Modra  <amodra@gmail.com>
21780             H.J. Lu  <hongjiu.lu@intel.com>
21782         PR driver/81523
21783         * gcc.c (NO_PIE_SPEC): Delete.
21784         (PIE_SPEC): Define as !no-pie/pie.  Move static|shared|r
21785         exclusion..
21786         (LINK_PIE_SPEC): ..to here.
21787         (LINK_COMMAND_SPEC): Support -no-pie.
21788         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21789         chain of crtbegin*.o selection, update for PIE_SPEC changes and
21790         format.
21791         (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21792         * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21793         (ENDFILE_CRTEND_SPEC): Similarly.
21795 2017-08-08  Uros Bizjak  <ubizjak@gmail.com>
21797         PR target/81708
21798         * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21799         (mstack-protector-guard-offset=): Ditto.
21800         * config/i386/i386.c (ix86_option_override): Handle
21801         -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21802         options.
21803         (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21804         ix86_stack_protect_guard_offset variables.
21805         (TARGET_STACK_PROTECT_GUARD): Always define.
21806         * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21807         and -mstack-protector-guard-offset= options.
21809 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21811         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21812         boundary case for the last candidate.
21814 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21816         * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21817         * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21819 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21821         PR middle-end/19706
21822         * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21823         * config/aarch64/aarch64-builtins.c
21824         (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21825         * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21826         * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21828 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21829             Andrew Pinski <pinskia@gmail.com>
21831         PR middle-end/19706
21832         * internal-fn.def (XORSIGN): New.
21833         * optabs.def (xorsign_optab): New.
21834         * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21835         (convert_expand_mult_copysign): New.
21836         (pass_optimize_widening_mul::execute): Call
21837         convert_expand_mult_copysign.
21839 2017-08-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21841         PR tree-optimization/81354
21842         * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21843         Insert on edges rather than explicitly creating landing pads.
21844         (analyze_candidates_and_replace): Commit edge inserts.
21846 2017-08-08  Richard Biener  <rguenther@suse.de>
21848         PR middle-end/81719
21849         * tree-ssa-loop-niter.c: Include tree-dfa.h.
21850         (expand_simple_operations): Also look through ADDR_EXPRs with
21851         MEM_REF bases treating them as POINTER_PLUS_EXPR.
21853 2017-08-08  Richard Biener  <rguenther@suse.de>
21855         PR tree-optimization/81723
21856         * tree-vect-slp.c (struct bst_traits): New hash traits.
21857         (bst_fail): New global.
21858         (vect_build_slp_tree_2): New worker, split out from ...
21859         (vect_build_slp_tree): ... this now wrapping it with using
21860         bst_fail set to cache SLP tree build fails.  Properly handle
21861         max_tree_size.
21862         (vect_analyze_slp_instance): Allocate and free bst_fail.
21864 2017-08-08  Martin Liska  <mliska@suse.cz>
21866         PR tree-opt/81696
21867         * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21868         LABEL_DECLs that can be from a different function.
21870 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21872         PR tree-optimization/81744
21873         * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21874         loop's number of iterations.
21876 2017-08-08  Martin Liska  <mliska@suse.cz>
21878         * asan.c: Include header files.
21879         * attribs.c (build_decl_attribute_variant): New function moved
21880         from tree.[ch].
21881         (build_type_attribute_qual_variant): Likewise.
21882         (cmp_attrib_identifiers): Likewise.
21883         (simple_cst_list_equal): Likewise.
21884         (omp_declare_simd_clauses_equal): Likewise.
21885         (attribute_value_equal): Likewise.
21886         (comp_type_attributes): Likewise.
21887         (build_type_attribute_variant): Likewise.
21888         (lookup_ident_attribute): Likewise.
21889         (remove_attribute): Likewise.
21890         (merge_attributes): Likewise.
21891         (merge_type_attributes): Likewise.
21892         (merge_decl_attributes): Likewise.
21893         (merge_dllimport_decl_attributes): Likewise.
21894         (handle_dll_attribute): Likewise.
21895         (attribute_list_equal): Likewise.
21896         (attribute_list_contained): Likewise.
21897         * attribs.h (lookup_attribute): New function moved from tree.[ch].
21898         (lookup_attribute_by_prefix): Likewise.
21899         * bb-reorder.c: Include header files.
21900         * builtins.c: Likewise.
21901         * calls.c: Likewise.
21902         * cfgexpand.c: Likewise.
21903         * cgraph.c: Likewise.
21904         * cgraphunit.c: Likewise.
21905         * convert.c: Likewise.
21906         * dwarf2out.c: Likewise.
21907         * final.c: Likewise.
21908         * fold-const.c: Likewise.
21909         * function.c: Likewise.
21910         * gimple-expr.c: Likewise.
21911         * gimple-fold.c: Likewise.
21912         * gimple-pretty-print.c: Likewise.
21913         * gimple.c: Likewise.
21914         * gimplify.c: Likewise.
21915         * hsa-common.c: Likewise.
21916         * hsa-gen.c: Likewise.
21917         * internal-fn.c: Likewise.
21918         * ipa-chkp.c: Likewise.
21919         * ipa-cp.c: Likewise.
21920         * ipa-devirt.c: Likewise.
21921         * ipa-fnsummary.c: Likewise.
21922         * ipa-inline.c: Likewise.
21923         * ipa-visibility.c: Likewise.
21924         * ipa.c: Likewise.
21925         * lto-cgraph.c: Likewise.
21926         * omp-expand.c: Likewise.
21927         * omp-general.c: Likewise.
21928         * omp-low.c: Likewise.
21929         * omp-offload.c: Likewise.
21930         * omp-simd-clone.c: Likewise.
21931         * opts-global.c: Likewise.
21932         * passes.c: Likewise.
21933         * predict.c: Likewise.
21934         * sancov.c: Likewise.
21935         * sanopt.c: Likewise.
21936         * symtab.c: Likewise.
21937         * toplev.c: Likewise.
21938         * trans-mem.c: Likewise.
21939         * tree-chkp.c: Likewise.
21940         * tree-eh.c: Likewise.
21941         * tree-into-ssa.c: Likewise.
21942         * tree-object-size.c: Likewise.
21943         * tree-parloops.c: Likewise.
21944         * tree-profile.c: Likewise.
21945         * tree-ssa-ccp.c: Likewise.
21946         * tree-ssa-live.c: Likewise.
21947         * tree-ssa-loop.c: Likewise.
21948         * tree-ssa-sccvn.c: Likewise.
21949         * tree-ssa-structalias.c: Likewise.
21950         * tree-ssa.c: Likewise.
21951         * tree-streamer-in.c: Likewise.
21952         * tree-vectorizer.c: Likewise.
21953         * tree-vrp.c: Likewise.
21954         * tsan.c: Likewise.
21955         * ubsan.c: Likewise.
21956         * varasm.c: Likewise.
21957         * varpool.c: Likewise.
21958         * tree.c: Remove functions moved to attribs.[ch].
21959         * tree.h: Likewise.
21960         * config/aarch64/aarch64.c: Add attrs.h header file.
21961         * config/alpha/alpha.c: Likewise.
21962         * config/arc/arc.c: Likewise.
21963         * config/arm/arm.c: Likewise.
21964         * config/avr/avr.c: Likewise.
21965         * config/bfin/bfin.c: Likewise.
21966         * config/c6x/c6x.c: Likewise.
21967         * config/cr16/cr16.c: Likewise.
21968         * config/cris/cris.c: Likewise.
21969         * config/darwin.c: Likewise.
21970         * config/epiphany/epiphany.c: Likewise.
21971         * config/fr30/fr30.c: Likewise.
21972         * config/frv/frv.c: Likewise.
21973         * config/ft32/ft32.c: Likewise.
21974         * config/h8300/h8300.c: Likewise.
21975         * config/i386/winnt.c: Likewise.
21976         * config/ia64/ia64.c: Likewise.
21977         * config/iq2000/iq2000.c: Likewise.
21978         * config/lm32/lm32.c: Likewise.
21979         * config/m32c/m32c.c: Likewise.
21980         * config/m32r/m32r.c: Likewise.
21981         * config/m68k/m68k.c: Likewise.
21982         * config/mcore/mcore.c: Likewise.
21983         * config/microblaze/microblaze.c: Likewise.
21984         * config/mips/mips.c: Likewise.
21985         * config/mmix/mmix.c: Likewise.
21986         * config/mn10300/mn10300.c: Likewise.
21987         * config/moxie/moxie.c: Likewise.
21988         * config/msp430/msp430.c: Likewise.
21989         * config/nds32/nds32-isr.c: Likewise.
21990         * config/nds32/nds32.c: Likewise.
21991         * config/nios2/nios2.c: Likewise.
21992         * config/nvptx/nvptx.c: Likewise.
21993         * config/pa/pa.c: Likewise.
21994         * config/pdp11/pdp11.c: Likewise.
21995         * config/powerpcspe/powerpcspe.c: Likewise.
21996         * config/riscv/riscv.c: Likewise.
21997         * config/rl78/rl78.c: Likewise.
21998         * config/rx/rx.c: Likewise.
21999         * config/s390/s390.c: Likewise.
22000         * config/sh/sh.c: Likewise.
22001         * config/sol2.c: Likewise.
22002         * config/sparc/sparc.c: Likewise.
22003         * config/spu/spu.c: Likewise.
22004         * config/stormy16/stormy16.c: Likewise.
22005         * config/tilegx/tilegx.c: Likewise.
22006         * config/tilepro/tilepro.c: Likewise.
22007         * config/v850/v850.c: Likewise.
22008         * config/vax/vax.c: Likewise.
22009         * config/visium/visium.c: Likewise.
22010         * config/xtensa/xtensa.c: Likewise.
22012 2017-08-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
22014         PR target/81593
22015         * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
22016         constraints since the -mupper-regs-* switches have been
22017         eliminated.
22018         (vsx_concat_<mode>_1): New combiner insns to recognize inserting
22019         into a vector from a double word element that was extracted from
22020         another vector, and eliminate extra XXPERMDI instructions.
22021         (vsx_concat_<mode>_2): Likewise.
22022         (vsx_concat_<mode>_3): Likewise.
22023         (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
22024         concat to allow optimizing inserts from previous extracts.
22026 2017-08-07  Uros Bizjak  <ubizjak@gmail.com>
22028         * config/i386/i386.c (ix86_stack_protect_guard): Generate
22029         memory reference to a SSP offset in TLS address space.
22030         (ix86_print_operand) <case '@'>: Remove.
22031         (ix86_print_operand_punct_valid_p): Remove '@' code.
22032         * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
22033         UNSPEC_SP_TLS_TEST.
22034         (stack_tls_protect_set_<mode>): Remove.
22035         (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
22036         (stack_tls_protect_test_<mode>): Remove.
22037         (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
22039 2017-08-07  Olivier Hainque  <hainque@adacore.com>
22041         PR target/81755
22042         * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
22044 2017-08-07  Douglas Rupp  <rupp@adacore.com>
22046         * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
22047         variable was referenced as multidir in command.
22049 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22051         PR c/69389
22052         * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
22053         BIT_FIELD_REF.
22055 2017-08-07  Martin Liska  <mliska@suse.cz>
22057         * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
22058         * config/rl78/rl78.c: Add include of attribs.h.
22059         * config/sh/sh.c: Likewise.
22060         * config/v850/v850.c: Likewise.
22062 2017-08-07  Tom de Vries  <tom@codesourcery.com>
22064         PR middle-end/78266
22065         * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
22067 2017-08-07  Martin Liska  <mliska@suse.cz>
22069         * config/mips/mips.c: Include attribs.h.
22071 2017-08-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
22073         PR fortran/68829
22074         * doc/invoke.texi: Document change in behvaior for -Ofast for
22075         Fortran.
22077 2017-08-07  Wilco Dijkstra  <wdijkstr@arm.com>
22079         * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
22080         Use gen_frame_mem.
22081         (aarch64_pop_regs): Likewise.
22082         (aarch64_gen_load_pair): Likewise.
22083         (aarch64_save_callee_saves): Likewise.
22084         (aarch64_restore_callee_saves): Likewise.
22086 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22088         * config/i386/i386.c: Revert the last change.
22090 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22092         PR target/81736
22093         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
22094         to ...
22095         (ix86_finalize_stack_frame_flags): This.  Also clear
22096         frame_pointer_needed if -fno-omit-frame-pointer is used without
22097         stack access.
22098         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
22099         with ix86_finalize_stack_frame_flags.
22100         (ix86_expand_epilogue): Likewise.
22101         (ix86_expand_split_stack_prologue): Likewise.
22103 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22105         PR target/81743
22106         * config/i386/i386.c (get_builtin_code_for_version): Set priority
22107         to P_AES for Westmere.
22109 2017-08-07  Jonathan Yong  <10walls@gmail.com>
22111         * config/i386/mingw.opt (fset-stack-executable): Removed.
22112         * config/i386/cygming.opt (fset-stack-executable): Moved
22113         from mingw.opt.
22114         * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
22116 2017-08-07  Segher Boessenkool  <segher@kernel.crashing.org>
22118         * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
22120 2017-08-07  Marek Polacek  <polacek@redhat.com>
22122         PR middle-end/81737
22123         * fold-const.c (fold_indirect_ref_1): Check type_domain.
22125 2017-08-07  Martin Liska  <mliska@suse.cz>
22127         * attribs.h (canonicalize_attr_name): New function.
22128         (cmp_attribs): Move from c-format.c and adjusted.
22129         (is_attribute_p): Moved from tree.h.
22130         * tree-inline.c: Add new includes.
22131         * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
22132         (private_is_attribute_p): Remove.
22133         (private_lookup_attribute): Likewise.
22134         (private_lookup_attribute_by_prefix): Simplify.
22135         (remove_attribute): Use is_attribute_p.
22136         * tree.h: Remove removed declarations.
22138 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22140         PR middle-end/81698
22141         * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
22142         instead of computing it in the function.  Formatting fix.
22143         (expand_case): Don't rely on default_edge being the first edge,
22144         clear it if removing it, pass default_edge to
22145         emit_case_dispatch_table.
22146         (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
22147         fix.
22149 2017-08-06  Uros Bizjak  <ubizjak@gmail.com>
22151         * config/alpha/alpha.c (alpha_reorg): If trap is the last active
22152         insn in the function, emit NOP after the insn.
22154 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22156         * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
22157         and element loops.
22159 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22161         * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
22162         loop.
22164 2017-08-04  Yury Gribov  <tetra2005@gmail.com>
22166         PR tree-optimization/57371
22167         * match.pd: New pattern.
22169 2017-08-04  Marek Polacek  <polacek@redhat.com>
22171         PR middle-end/81695
22172         * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
22173         perform the computation in offset_int.
22175 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22177         PR tree-optimization/81136
22178         * tree-vectorizer.h: Include tree-hash-traits.h.
22179         (vec_base_alignments): New typedef.
22180         (vec_info): Add a base_alignments field.
22181         (vect_record_base_alignments): Declare.
22182         * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
22183         field.
22184         (DR_IS_CONDITIONAL_IN_STMT): New macro.
22185         (create_data_ref): Add an is_conditional_in_stmt argument.
22186         * tree-data-ref.c (create_data_ref): Likewise.  Use it to initialize
22187         the is_conditional_in_stmt field.
22188         (data_ref_loc): Add an is_conditional_in_stmt field.
22189         (get_references_in_stmt): Set the is_conditional_in_stmt field.
22190         (find_data_references_in_stmt): Update call to create_data_ref.
22191         (graphite_find_data_references_in_stmt): Likewise.
22192         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
22193         * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
22194         (vect_record_base_alignment): New function.
22195         (vect_record_base_alignments): Likewise.
22196         (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
22197         for nested statements even if we fail to compute a misalignment.
22198         Use pooled base alignments for unconditional references.
22199         (vect_find_same_alignment_drs): Compare base addresses instead
22200         of base objects.
22201         (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
22202         * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
22204 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22206         * tree-vectorizer.h (vec_info): Add a constructor and destructor.
22207         Add an explicit name for the enum.  Use auto_vec for slp_instances
22208         and grouped_stores.
22209         (_loop_vec_info): Add a constructor and destructor.  Use auto_vec
22210         for all vectors.
22211         (_bb_vec_info): Add a constructor and destructor.
22212         (vinfo_for_stmt): Return NULL for uids of -1 as well.
22213         (destroy_loop_vec_info): Delete.
22214         (vect_destroy_datarefs): Likewise.
22215         * tree-vectorizer.c (vect_destroy_datarefs): Delete.
22216         (vec_info::vec_info): New function.
22217         (vec_info::~vec_info): Likewise.
22218         (vectorize_loops): Use delete instead of destroy_loop_vec_info.
22219         * tree-parloops.c (gather_scalar_reductions): Use delete instead of
22220         destroy_loop_vec_info.
22221         * tree-vect-loop.c (new_loop_vec_info): Replace with...
22222         (_loop_vec_info::_loop_vec_info): ...this.
22223         (destroy_loop_vec_info): Replace with...
22224         (_loop_vec_info::~_loop_vec_info): ...this.  Unconditionally delete
22225         the stmt_vec_infos.  Leave handling of vec_info information to its
22226         destructor.  Remove explicit vector releases.
22227         (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
22228         (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
22229         * tree-vect-slp.c (new_bb_vec_info): Replace with...
22230         (_bb_vec_info::_bb_vec_info): ...this.  Don't reserve space in
22231         BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
22232         (destroy_bb_vec_info): Replace with...
22233         (_bb_vec_info::~_bb_vec_info): ...this.  Leave handling of vec_info
22234         information to its destructor.
22235         (vect_slp_analyze_bb_1): Use new and delete instead of
22236         new_bb_vec_info and destroy_bb_vec_info.
22237         (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22238         single delete.
22240 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22242         * tree-data-ref.h (subscript): Add access_fn field.
22243         (data_dependence_relation): Add could_be_independent_p.
22244         (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22245         (same_access_functions): Move to tree-data-ref.c.
22246         * tree-data-ref.c (ref_contains_union_access_p): New function.
22247         (access_fn_component_p): Likewise.
22248         (access_fn_components_comparable_p): Likewise.
22249         (dr_analyze_indices): Add a reference to access_fn_component_p.
22250         (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22251         DR_ACCESS_FN.
22252         (constant_access_functions): Likewise.
22253         (add_other_self_distances): Likewise.
22254         (same_access_functions): Likewise.  (Moved from tree-data-ref.h.)
22255         (initialize_data_dependence_relation): Use XCNEW and remove
22256         explicit zeroing of DDR_REVERSED_P.  Look for a subsequence
22257         of access functions that have the same type.  Allow the
22258         subsequence to end with different bases in some circumstances.
22259         Record the chosen access functions in SUB_ACCESS_FN.
22260         (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22261         a_index and b_index.  Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22262         (subscript_dependence_tester_1): Likewise dra and drb.
22263         (build_classic_dist_vector): Update calls accordingly.
22264         (subscript_dependence_tester): Likewise.
22265         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22266         DDR_COULD_BE_INDEPENDENT_P.
22267         * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22268         comp_alias_ddrs instead of may_alias_ddrs.
22269         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22270         New function.
22271         (vect_analyze_data_ref_dependence): Use it if
22272         DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22273         distance vectors if that fails.
22274         (dependence_distance_ge_vf): New function.
22275         (vect_prune_runtime_alias_test_list): Use it.  Don't clear
22276         LOOP_VINFO_MAY_ALIAS_DDRS.
22278 2017-08-04  Richard Biener  <rguenther@suse.de>
22280         PR middle-end/81705
22281         * fold-const.c (fold_binary_loc): Properly restrict
22282         minus_var0 && minus_var1 case when associating undefined overflow
22283         entities.
22285 2017-08-04  Tom de Vries  <tom@codesourcery.com>
22287         * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22289 2017-08-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22291         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22292         Don't start diagnostic messages with a capital letter.
22293         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22294         Likewise.
22295         (rs6000_invalid_builtin): Likewise.
22296         (rs6000_trampoline_init): Likewise.
22298 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22300         PR target/81621
22301         * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22302         after setting changeable df flags.
22304 2017-08-03  Richard Biener  <rguenther@suse.de>
22306         * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22307         up if the use is in USE - X.
22309 2017-08-03  Alexander Monakov  <amonakov@ispras.ru>
22311         * toplev.c (dumpfile.h): New include.
22312         (internal_error_reentered): New static function.  Use it...
22313         (internal_error_function): ...here to handle reentered internal_error.
22315 2017-08-03  Richard Biener  <rguenther@suse.de>
22317         PR middle-end/81148
22318         * fold-const.c (split_tree): Add minus_var and minus_con
22319         arguments, remove unused loc arg.  Never generate NEGATE_EXPRs
22320         here but always use minus_*.
22321         (associate_trees): Assert we never associate with MINUS_EXPR
22322         and NULL first operand.  Do not recurse for PLUS_EXPR operands
22323         when associating as MINUS_EXPR either.
22324         (fold_binary_loc): Track minus_var and minus_con.
22326 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22328         PR lto/81430
22329         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22330         ACCEL_COMPILER, apply finish_options on
22331         DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22333 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22335         PR target/81662
22336         * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22337         function_entry_patch_area_size > 0.
22339 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22341         PR driver/81650
22342         * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22343         instead of 10??LU, perform unit multiplication in wide_int,
22344         don't change alloc_object_size_limit if the limit is larger
22345         than SSIZE_MAX.
22347         PR tree-optimization/81655
22348         PR tree-optimization/81588
22349         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22350         the case when ranges[i].low and high are 1 for unsigned type with
22351         precision 1.
22353         PR middle-end/81052
22354         * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22355         (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22357 2017-08-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22359         * tree-vrp.h: Add include guard.
22361 2017-08-02  Uros Bizjak  <ubizjak@gmail.com>
22363         PR target/81644
22364         * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22365         (ud2): New insn pattern.
22366         * config/i386/i386.c (ix86_expand_epilogue):
22367         For naked functions, generate ud2 instead of trap insn.
22369 2017-08-02  Marek Polacek  <polacek@redhat.com>
22371         PR other/81667
22372         * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22374 2017-08-02  Tom de Vries  <tom@codesourcery.com>
22375             Cesar Philippidis  <cesar@codesourcery.com>
22377         * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22378         Add missing edge probabilities.
22380 2017-08-02  Tamar Christina  <tamar.christina@arm.com>
22382         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22383         Correct endianness.
22385 2017-08-02  Jakub Jelinek  <jakub@redhat.com>
22387         PR middle-end/79499
22388         * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22389         for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22390         of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22391         prologue_seq sequences - if any.
22393 2017-08-02  Richard Biener  <rguenther@suse.de>
22395         * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22396         via vectors if supported, integer extracts via punning if supported
22397         or otherwise vector extracts.
22399 2017-08-02  Richard Biener  <rguenther@suse.de>
22401         * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22402         into ...
22403         (bitmap_insert_into_set): ... this.
22405 2017-08-02  Richard Biener  <rguenther@suse.de>
22407         PR tree-optimization/81633
22408         Revert
22409         2015-08-17  Alan Hayward  <alan.hayward@arm.com>
22411         PR tree-optimization/71752
22412         * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22414 2017-08-01  Daniel Santos  <daniel.santos@pobox.com>
22416         * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22417         (machine_function::call_ms2sysv_pad_out): Remove field.
22418         * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22419         (ix86_compute_frame_layout): Likewise.
22421 2017-08-01  H.J. Lu  <hongjiu.lu@intel.com>
22423         PR target/81654
22424         * config/i386/i386.c (ix86_set_func_type): Disallow naked
22425         attribute with interrupt attribute.
22427 2017-08-01  Andrew Pinski  <apinski@cavium.com>
22429         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22430         BIT_INSERT_EXPR's operand 1
22431         to see if the types precision matches.
22433 2017-08-01  Martin Liska  <mliska@suse.cz>
22435         PR middle-end/70140
22436         * builtins.c (expand_builtin_memcpy_args): Remove.
22437         (expand_builtin_memcpy): Call newly added function
22438         expand_builtin_memory_copy_args.
22439         (expand_builtin_memcpy_with_bounds): Likewise.
22440         (expand_builtin_mempcpy): Remove last argument.
22441         (expand_builtin_mempcpy_with_bounds): Likewise.
22442         (expand_builtin_memory_copy_args): New function created from
22443         expand_builtin_mempcpy_args with small modifications.
22444         (expand_builtin_mempcpy_args): Remove.
22445         (expand_builtin_stpcpy): Remove unused argument.
22446         (expand_builtin): Likewise.
22447         (expand_builtin_with_bounds): Likewise.
22449 2017-08-01  Martin Liska  <mliska@suse.cz>
22451         Revert r250771
22452         Make mempcpy more optimal (PR middle-end/70140).
22454 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22456         PR target/81622
22457         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22458         __builtin_vec_cmpne verify both arguments are compatible vectors
22459         before looking at TYPE_MODE on the element type.  For __builtin_vec_ld
22460         verify arg1_type is a pointer or array type.  For __builtin_vec_st,
22461         move computation of aligned to after checking the argument types.
22462         Formatting fixes.
22464         PR target/80846
22465         * config/rs6000/vsx.md (vextract_fp_from_shorth,
22466         vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22467         calls.
22469 2017-08-01  Jerome Lambourg  <lambourg@adacore.com>
22470             Doug Rupp  <rupp@adacore.com>
22471             Olivier Hainque  <hainque@adacore.com>
22473         * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22474         well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22475         arm8 (arch v4).
22476         * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22477         for TARGET_OS_CPP_BUILTIN.
22478         (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22479         refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22480         arm_arch7.
22481         (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22482         passing required abi options to the assembler for EABI configurations.
22483         (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22484         of .text.hot and .text.unlikely sections for kernel modules when
22485         using ARM style exceptions.
22486         (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22487         options. Add EXTRA_CC1_SPEC.
22488         (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22489         toolchain options.
22490         (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22491         transition.
22492         (ARM_TARGET2_DWARF_FORMAT): Define.
22493         * config/arm/t-vxworks: Adjust multilib control to removal of the
22494         Diab command line options.
22496 2017-08-01  Martin Liska  <mliska@suse.cz>
22498         PR gcov-profile/81561
22499         * gcov.c (unblock): Make unblocking safe as we need to preserve
22500         index correspondence of blocks and block_lists.
22502 2017-08-01  Richard Biener  <rguenther@suse.de>
22504         PR tree-optimization/81181
22505         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22506         (compute_antic): ... end of iteration here.
22508 2017-08-01  James Greenhalgh  <james.greenhalgh@arm.com>
22510         * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22511         (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22512         (ftree-slp-vectorize): Likewise.
22513         * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22514         can no longer be set independent of flag_tree_loop_vectorize.
22515         * omp-general.c (emp_max_vf): Likewise.
22516         * opts.c (enable_fdo_optimizations): Remove references to
22517         flag_tree_vectorize, these are now implicit.
22518         (common_handle_option): Remove handling for OPT_ftree_vectorize,
22519         and leave it for the options machinery.
22521 2017-08-01  Martin Liska  <mliska@suse.cz>
22523         PR middle-end/70140
22524         * builtins.c (expand_builtin_memcpy_args): Remove.
22525         (expand_builtin_memcpy): Call newly added function
22526         expand_builtin_memory_copy_args.
22527         (expand_builtin_memcpy_with_bounds): Likewise.
22528         (expand_builtin_mempcpy): Remove last argument.
22529         (expand_builtin_mempcpy_with_bounds): Likewise.
22530         (expand_builtin_memory_copy_args): New function created from
22531         expand_builtin_mempcpy_args with small modifications.
22532         (expand_builtin_mempcpy_args): Remove.
22533         (expand_builtin_stpcpy): Remove unused argument.
22534         (expand_builtin): Likewise.
22535         (expand_builtin_with_bounds): Likewise.
22537 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22539         PR target/81641
22540         * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22541         print "ds:" only for immediates in generic address space.
22543 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22545         PR target/81639
22546         * config/i386/i386.c (ix86_funciton_naked): New prototype.
22547         (ix86_function_ok_for_sibcall): Return false for naked functions.
22549 2017-08-01  Richard Biener  <rguenther@suse.de>
22551         * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22552         (compute_antic): Seed worklist with exit block predecessors.
22553         * cfganal.c (dfs_find_deadend): For a cycle return the source
22554         of the edge closing it.
22556 2017-08-01  Tamar Christina  <tamar.christina@arm.com>
22558         * config/aarch64/aarch64.c
22559         (aarch64_can_const_movi_rtx_p): Move 0 check.
22561 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22563         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22564         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22565         above macro.
22566         * match.pd: Ditto in address comparison pattern.
22568 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22570         PR tree-optimization/81627
22571         * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22572         closed ssa form for store-store chain.
22574 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22576         PR tree-optimization/81620
22577         * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22578         for store-store chain.
22580 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22582         PR tree-optimization/81588
22583         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22584         ranges[i].in_p, invert comparison code ccode.  For >/>=,
22585         swap rhs1 and rhs2 and comparison code unconditionally,
22586         for </<= don't do that.  Don't swap rhs1/rhs2 again if
22587         ranges[i].in_p, instead invert comparison code ccode if
22588         opcode or oe->rank is BIT_IOR_EXPR.
22590         PR target/80846
22591         * optabs.def (vec_extract_optab, vec_init_optab): Change from
22592         a direct optab to conversion optab.
22593         * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22594         with GET_MODE_INNER as last argument instead of optab_handler.
22595         * expmed.c (extract_bit_field_1): Likewise.  Use vector from
22596         vector extraction if possible and optab is available.
22597         * expr.c (store_constructor): Use convert_optab_handler instead
22598         of optab_handler.  Use vector initialization from smaller
22599         vectors if possible and optab is available.
22600         * tree-vect-stmts.c (vectorizable_load): Likewise.
22601         * doc/md.texi (vec_extract, vec_init): Document that the optabs
22602         now have two modes.
22603         * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22604         of vec_init from half-sized vectors with the same element mode.
22605         * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22606         (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22607         (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22608         reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22609         reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22610         after mode in gen_vec_extract* calls.
22611         (vec_extract<mode>): Renamed to ...
22612         (vec_extract<mode><ssescalarmodelower>): ... this.
22613         (vec_extract<mode><ssehalfvecmodelower>): New expander.
22614         (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22615         element mode after mode in gen_vec_init* calls.
22616         (VEC_INIT_HALF_MODE): New mode iterator.
22617         (vec_init<mode>): Renamed to ...
22618         (vec_init<mode><ssescalarmodelower>): ... this.
22619         (vec_init<mode><ssehalfvecmodelower>): New expander.
22620         * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22621         (vec_extractv2sfsf): ... this.
22622         (vec_initv2sf): Renamed to ...
22623         (vec_initv2sfsf): ... this.
22624         (vec_extractv2si): Renamed to ...
22625         (vec_extractv2sisi): ... this.
22626         (vec_initv2si): Renamed to ...
22627         (vec_initv2sisi): ... this.
22628         (vec_extractv4hi): Renamed to ...
22629         (vec_extractv4hihi): ... this.
22630         (vec_initv4hi): Renamed to ...
22631         (vec_initv4hihi): ... this.
22632         (vec_extractv8qi): Renamed to ...
22633         (vec_extractv8qiqi): ... this.
22634         (vec_initv8qi): Renamed to ...
22635         (vec_initv8qiqi): ... this.
22636         * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22637         (vec_init<mode>): Renamed to ...
22638         (vec_init<mode><VEC_base_l>): ... this.
22639         (vec_extract<mode>): Renamed to ...
22640         (vec_extract<mode><VEC_base_l>): ... this.
22641         * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22642         (vec_initv2sfsf): ... this.
22643         * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22644         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22645         vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22646         element mode after mode in gen_vec_init* calls.
22647         * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22648         (vec_init<mode><Vel>): ... this.
22649         (vec_extract<mode>): Renamed to ...
22650         (vec_extract<mode><Vel>): ... this.
22651         * config/aarch64/iterators.md (Vel): New mode attribute.
22652         * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22653         Add element mode after mode in gen_vec_extract* calls.
22654         * config/s390/vector.md (non_vec_l): New mode attribute.
22655         (vec_extract<mode>): Renamed to ...
22656         (vec_extract<mode><non_vec_l>): ... this.
22657         (vec_init<mode>): Renamed to ...
22658         (vec_init<mode><non_vec_l>): ... this.
22659         * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22660         s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22661         vec_extract mode.
22662         * config/arm/iterators.md (V_elem_l): New mode attribute.
22663         * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22664         (vec_extract<mode><V_elem_l>): ... this.
22665         (vec_extractv2di): Renamed to ...
22666         (vec_extractv2didi): ... this.
22667         (vec_init<mode>): Renamed to ...
22668         (vec_init<mode><V_elem_l>): ... this.
22669         (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22670         reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22671         reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22672         Add element mode after gen_vec_extract* calls.
22673         * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22674         (vec_init<mode><unitmode>): ... this.
22675         (vec_extract<mode>): Renamed to ...
22676         (vec_extract<mode><unitmode>): ... this.
22677         * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22678         (vec_init<mode><unitmode>): ... this.
22679         * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22680         (vec_initv2sfsf): ... this.
22681         (vec_extractv2sf): Renamed to ...
22682         (vec_extractv2sfsf): ... this.
22683         (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22684         Add element mode after gen_vec_extract* calls.
22685         * config/mips/mips.md (unitmode): New mode iterator.
22686         * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22687         spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22688         * config/spu/spu.md (inner_l): New mode attribute.
22689         (vec_init<mode>): Renamed to ...
22690         (vec_init<mode><inner_l>): ... this.
22691         (vec_extract<mode>): Renamed to ...
22692         (vec_extract<mode><inner_l>): ... this.
22693         * config/sparc/sparc.md (veltmode): New mode iterator.
22694         (vec_init<VMALL:mode>): Renamed to ...
22695         (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22696         * config/ia64/vect.md (vec_initv2si): Renamed to ...
22697         (vec_initv2sisi): ... this.
22698         (vec_initv2sf): Renamed to ...
22699         (vec_initv2sfsf): ... this.
22700         (vec_extractv2sf): Renamed to ...
22701         (vec_extractv2sfsf): ... this.
22702         * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22703         (vec_init<mode>): Renamed to ...
22704         (vec_init<mode><VEC_base_l>): ... this.
22705         (vec_extract<mode>): Renamed to ...
22706         (vec_extract<mode><VEC_base_l>): ... this.
22707         * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22708         (vec_initv2sfsf): ... this.
22709         * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22710         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22711         vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22712         gen_vec_init* calls.
22714 2017-08-01  Richard Biener  <rguenther@suse.de>
22716         PR tree-optimization/81297
22717         * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22718         TREE_OVERFLOW from INTEGER_CSTs.
22720 2017-07-31  Segher Boessenkool  <segher@kernel.crashing.org>
22722         * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22724 2017-07-31  Carl Love  <cel@us.ibm.com>
22726         * config/rs6000/rs6000-c: Add support for built-in functions
22727         vector signed char vec_xl_be (signed long long, signed char *);
22728         vector unsigned char vec_xl_be (signed long long, unsigned char *);
22729         vector signed int vec_xl_be (signed long long, signed int *);
22730         vector unsigned int vec_xl_be (signed long long, unsigned int *);
22731         vector signed long long vec_xl_be (signed long long, signed long long *);
22732         vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22733         vector signed short vec_xl_be (signed long long, signed short *);
22734         vector unsigned short vec_xl_be (signed long long, unsigned short *);
22735         vector double vec_xl_be (signed long long, double *);
22736         vector float vec_xl_be (signed long long, float *);
22737         * config/rs6000/altivec.h (vec_xl_be): Add #define.
22738         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22739         XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22740         for the builtins.
22741         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22742         (altivec_expand_builtin): Add switch statement to call
22743         altivec_expand_xl_be for each builtin.
22744         (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22745         __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22746         __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22747         __builtin_vsx_le_be_v16qi.
22748         * doc/extend.texi: Update the built-in documentation file for the
22749         new built-in functions.
22751 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22753         PR target/25967
22754         * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22755         New function.
22756         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22758 2017-07-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22760         * config.gcc: Add z14.
22761         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22762         CPU model numbers for z13s and z14.
22763         * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22764         arch12 with z14.
22765         * config/s390/s390-opts.h (enum processor_type): Rename
22766         PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22767         * config/s390/s390.c (processor_table): Add field for CPU name to
22768         be passed to Binutils.
22769         (s390_asm_output_machine_for_arch): Use the new field in
22770         processor_table for Binutils.
22771         (s390_expand_builtin): Replace arch12 with z14.
22772         (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22773         (s390_get_sched_attrmask): Likewise.
22774         (s390_get_unit_mask): Likewise.
22775         * config/s390/s390.opt: Add z14 to processor_type enum.
22777 2017-07-31  Martin Jambor  <mjambor@suse.cz>
22779         PR hsa/81477
22780         * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22781         regardless of optimization level.
22783 2017-07-31  Jan Hubicka <hubicka@ucw.cz>
22784             Martin Liska  <mliska@suse.cz>
22786         * predict.def: Remove old comment and adjust probability.
22787         * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22788         PREDICT statements.
22790 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22792         PR target/25967
22793         * config/i386/i386.c (ix86_function_naked): New function.
22794         (ix86_can_use_return_insn_p): Return false for naked functions.
22795         (ix86_expand_prologue): Skip prologue for naked functions.
22796         (ix86_expand_epilogue): Skip epilogue for naked functions
22797         and emit trap instruction.
22798         (ix86_warn_func_return): New function.
22799         (ix86_attribute_table): Add "naked" attribute specification.
22800         (TARGET_WARN_FUNC_RETURN): Define.
22801         * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22803 2017-07-31  Martin Liska  <mliska@suse.cz>
22805         * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22806         (dump_gimple_bb_header): Always dump BB info.
22807         (pp_cfg_jump): Do not append info about BB when dumping a jump.
22809 2017-07-31  Martin Liska  <mliska@suse.cz>
22811         PR sanitize/81530
22812         * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22813         also with current_function_decl non-null equality.
22815 2017-07-31  Jakub Jelinek  <jakub@redhat.com>
22817         PR sanitizer/81604
22818         * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22819         change type to the element type, instead add eltype variable and
22820         use it where we are interested in the element type.
22822         PR tree-optimization/81603
22823         * ipa-polymorphic-call.c
22824         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22825         offset arithmetic in offset_int, bail out if the resulting bit offset
22826         doesn't fit into shwi.
22828 2017-07-31  Martin Liska  <mliska@suse.cz>
22830         * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22831         (gimplify_save_expr): Fix comment.
22833 2017-07-30  H.J. Lu  <hongjiu.lu@intel.com>
22835         PR target/79793
22836         * config/i386/i386.c (ix86_function_arg): Update arguments for
22837         exception handler.
22838         (ix86_compute_frame_layout): Set the initial stack offset to
22839         INCOMING_FRAME_SP_OFFSET.  Update red-zone offset with
22840         INCOMING_FRAME_SP_OFFSET.
22841         (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22842         stack before exception handler returns.
22843         * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22844         the 'ERROR_CODE' for exception handler.
22846 2017-07-30  Uros Bizjak  <ubizjak@gmail.com>
22848         * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22849         (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22850         (ASM_OUTPUT_REG_POP): Ditto.
22851         * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22852         instead of asm_fprintf to output pure string.
22854 2017-07-29  Jakub Jelinek  <jakub@redhat.com>
22856         * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22857         to imported_module_or_decl hook.
22858         (debug_nothing_tree_tree_tree_bool): Remove.
22859         (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22860         * debug.c (do_nothing_debug_hooks): Use
22861         debug_nothing_tree_tree_tree_bool_bool instead of
22862         debug_nothing_tree_tree_tree_bool.
22863         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22864         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22865         * sdbout.c (sdb_debug_hooks): Likewise.
22866         * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22867         (gen_namespace_die): Add DW_AT_export_symbols attribute if
22868         langhook wants it.
22869         (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22870         if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22871         attribute, don't add anything.
22873 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22875         * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22876         (fold_build2_stat_loc): Likewise.
22877         (fold_build3_stat_loc): Likewise.
22878         * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22879         (fold_build1_loc): Remove macro.
22880         (fold_build2_loc): Likewise.
22881         (fold_build3_loc): Likewise.
22883 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22885         * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22886         (gimple_build_debug_bind_source_stat): Likewise.
22887         * gimple.h (gimple_build_debug_bind): Remove macro.
22888         (gimple_build_debug_bind_source): Likewise.
22890 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22892         * bitmap.c (bitmap_alloc): Adjust.
22893         (bitmap_gc_alloc): Likewise.
22894         * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22896 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22898         * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22899         (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22900         * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22901         (bitmap_gc_alloc_stat): Likewise.
22902         (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22904 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22906         * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22907         * rtl.h (shallow_copy_rtx): Remove macro.
22909 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22911         * emit-rtl.c (gen_raw_REG): Adjust.
22912         * gengenrtl.c (gendef): Likewise.
22913         * rtl.c (rtx_alloc_stat): Remove _stat from name.
22914         * rtl.h (rtx_alloc): Remove macro.
22916 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22918         * tree.c (build_tree_list_vec_stat): Remove _stat from name.
22919         (build_tree_list_stat): Likewise.
22920         * tree.h (build_tree_list): Remove macro.
22921         (build_tree_list_vec): Likewise.
22923 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22925         * tree.c (make_vector_stat): Remove _stat from name.
22926         (build_vector_stat): Likewise.
22927         * tree.h (make_vector_stat): Remove macro.
22928         (build_vector_stat): Likewise.
22930 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22932         * tree.h (build_var_debug_value): Remove prototype.
22934 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22936         * tree.c (tree_cons_stat): Remove _stat from name.
22937         * tree.h (tree_cons): Remove macro.
22939 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22941         * tree.c (build_vl_exp_stat): Remove _stat from name.
22942         * tree.h (build_vl_exp): Remove macro.
22944 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22946         * tree.c (build_decl_stat): Remove _stat from name.
22947         * tree.h (build_decl): Remove macro.
22949 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22951         * gimple.c (gimple_build_with_ops_stat): Adjust.
22952         (gimple_alloc_stat): Remove _stat from name.
22953         * gimple.h (gimple_alloc): Remove macro.
22955 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22957         * tree.c (make_tree_vec_stat): Remove _stat from name.
22958         (grow_tree_vec_stat): Likewise.
22959         * tree.h (make_tree_vec_stat): Adjust prototype.
22960         (grow_tree_vec_stat): Likewise.
22961         (make_tree_vec): Remove macro.
22962         (grow_tree_vec): Likewise.
22964 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22966         * fold-const.c (fold_build1_stat_loc): Adjust.
22967         (fold_build2_stat_loc): Likewise.
22968         (fold_build3_stat_loc): Likewise.
22969         * tree.c (build0_stat): Remove _stat from name.
22970         (build1_stat): Likewise.
22971         (build2_stat): Likewise.
22972         (build3_stat): Likewise.
22973         (build4_stat): Likewise.
22974         (build5_stat): Likewise.
22975         * tree.h (build1_loc): Remove macro, and rename _stat function
22976         to this.
22977         (build2_loc): Likewise.
22978         (build3_loc): Likewise.
22979         (build4_loc): Likewise.
22980         (build5_loc): Likewise.
22982 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22984         * tree.c (make_int_cst_stat): Remove _stat from name.
22985         * tree.h (make_int_cst_stat): Adjust prototype.
22986         (make_int_cst): Remove macro.
22988 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22990         * tree.c (make_tre_binfo_stat): Remove _stat from name.
22991         * tree.h (make_tree_binfo_stat): Adjust prototype.
22992         (make_tree_binfo): Remove.
22994 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22996         * tree.c (copy_node_stat): Rename to copy_node.
22997         (build_distinct_type_copy): Adjust.
22998         * tree.h (copy_node_stat): Adjust prototype.
22999         (copy_node): Remove macro.
23001 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
23003         * tree.c (make_node_stat): rename to make_node.
23004         (build_tree_list_stat): Adjust.
23005         (build0_stat): Likewise.
23006         (build2_stat): Likewise.
23007         (build3_stat): Likewise.
23008         (build4_stat): Likewise.
23009         (build5_stat): Likewise.
23010         (build_decl_stat): Likewise.
23011         * tree.h (make_node_stat): Adjust prototype.
23012         (make_node): remove macro.
23014 2017-07-28  Peter Bergner  <bergner@vnet.ibm.com>
23016         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
23017         (PPC_FEATURE2_SCV): Likewise.
23018         * config/rs6000/rs6000.c (cpu_supports_info): Use them.
23020 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23022         * config/aarch64/aarch64.c
23023         (aarch64_internal_mov_immediate): Add new special pattern.
23024         * config/aarch64/aarch64.md (*movdi_aarch64):
23025         Add reg/32bit const mov case.
23027 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23028             Richard Sandiford <richard.sandiford@linaro.org>
23030         * config/aarch64/aarch64.md (mov<mode>): Generalize.
23031         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
23032         Add integer and movi cases.
23033         (movi-split-hf-df-sf split, fp16): New.
23034         (enabled): Added TARGET_FP_F16INST.
23035         * config/aarch64/iterators.md (GPF_HF): New.
23036         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
23038 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23040         * config/aarch64/aarch64.c
23041         (aarch64_simd_container_mode): Add prototype.
23042         (aarch64_expand_mov_immediate): Add HI support.
23043         (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
23044         (aarch64_can_const_movi_rtx_p): New.
23045         (aarch64_preferred_reload_class):
23046         Remove restrictions of using FP registers for certain SIMD operations.
23047         (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
23048         (aarch64_valid_floating_const): Add integer move validation.
23049         (aarch64_simd_imm_scalar_p): Remove.
23050         (aarch64_output_scalar_simd_mov_immediate): Generalize function.
23051         (aarch64_legitimate_constant_p): Expand list of supported cases.
23052         * config/aarch64/aarch64-protos.h
23053         (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
23054         (aarch64_reinterpret_float_as_int): New.
23055         (aarch64_simd_imm_scalar_p): Remove.
23056         * config/aarch64/constraints.md (Uvi): New.
23057         (Dd): Split into Ds and new Dd.
23058         * config/aarch64/aarch64.md (*movsi_aarch64):
23059         Add SIMD mov case.
23060         (*movdi_aarch64): Add SIMD mov case.
23062 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23064         * tree-predcom.c: (struct chain): Handle store-store chain in which
23065         stores for elimination only store loop invariant values.
23066         (execute_pred_commoning_chain): Ditto.
23067         (prepare_initializers_chain_store_elim): Ditto.
23068         (prepare_finalizers): Ditto.
23069         (is_inv_store_elimination_chain): New function.
23070         (initialize_root_vars_store_elim_1): New function.
23072 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23074         * tree-predcom.c: Revise general description of the pass.
23075         (enum chain_type): New enum type for store elimination.
23076         (struct chain): New field supporting store elimination.
23077         (struct component): Ditto.
23078         (dump_chain): Dump store-stores chain.
23079         (release_chain): Release resources.
23080         (split_data_refs_to_components): Compute and create component
23081         contains only stores for elimination.
23082         (get_chain_last_ref_at): New function.
23083         (make_invariant_chain): Initialization.
23084         (make_rooted_chain): Specify chain type in parameter and record it.
23085         (add_looparound_copies): Skip for store-stores chain.
23086         (determine_roots_comp): Compute type of chain and pass it to
23087         make_rooted_chain.
23088         (initialize_root_vars_store_elim_2): New function.
23089         (finalize_eliminated_stores): New function.
23090         (remove_stmt): Handle store for elimination.
23091         (execute_pred_commoning_chain): Execute predictive commoning on
23092         store-store chains.
23093         (determine_unroll_factor): Skip unroll for store-stores chain.
23094         (prepare_initializers_chain_store_elim): New function.
23095         (prepare_initializers_chain): Hanlde store-store chain.
23096         (prepare_finalizers_chain, prepare_finalizers): New function.
23097         (tree_predictive_commoning_loop): Return integer value indicating
23098         if loop is unrolled or lcssa form is corrupted.
23099         (tree_predictive_commoning): Rewrite for lcssa form if necessary.
23101 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23103         * tree-predcom.c (initialize_root): Delete.
23104         (execute_pred_commoning_chain): Initialize root vars and replace
23105         reference of non-combined chain directly, rather than call above
23106         function.
23108 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23110         * tree-predcom.c (ref_at_iteration): Add parameter NITERS.  Compute
23111         memory reference to DR at (NITERS + ITERS)-th iteration of loop.
23113 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23115         * tree-predcom.c (struct chain): New field init_seq.
23116         (release_chain): Release init_seq.
23117         (prepare_initializers_chain): Record intialization stmts in above
23118         field.
23119         (insert_init_seqs): New function.
23120         (tree_predictive_commoning_loop): Call insert_init_seqs.
23122 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23124         * tree-predcom.c (determine_roots_comp): Skip trivial components.
23126 2017-07-28  Richard Biener  <rguenther@suse.de>
23128         * match.pd: Remove superfluous :c.
23129         * genmatch.c (simplify::id): Add member.
23130         (lower_commutative, lower_opt_convert, lower_cond, lower_for):
23131         Copy id.
23132         (current_id): New global.
23133         (dt_node::parent): Move from ...
23134         (dt_operand::parent): ... here.  Add for_id member.
23135         (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
23136         (decision_tree::find_node): Relax order requirement when
23137         merging DT_TRUE nodes to ones inbetween the current simplify
23138         and the one we try to merge with.  Add diagnostic whenever
23139         we need to enforce pattern order by not merging.
23140         (decision_tree::insert): Set current_id.
23141         (decision_tree::print_node): Dump parent node and for_id.
23142         (parser::last_id): Add member.
23143         (parser::push_simplify): Assign unique id.
23144         (parser::parser): Initialize last_id.
23146 2017-07-28  Martin Liska  <mliska@suse.cz>
23148         PR sanitizer/81340
23149         * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
23150         gimple_build_debug_bind.
23152 2017-07-28  Richard Biener  <rguenther@suse.de>
23154         PR tree-optimization/81502
23155         * match.pd: Add pattern combining BIT_INSERT_EXPR with
23156         BIT_FIELD_REF.
23157         * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
23158         size/pos operands.
23159         (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
23160         * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
23161         for BIT_FIELD_REF args.
23162         * fold-const.c (make_bit_field_ref): Likewise.
23163         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23165 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23167         PR sanitizer/80998
23168         * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
23169         * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
23170         * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
23171         Or it into SANITIZER_UNDEFINED.
23172         * ubsan.c: Include gimple-fold.h and varasm.h.
23173         (ubsan_expand_ptr_ifn): New function.
23174         (instrument_pointer_overflow): New function.
23175         (maybe_instrument_pointer_overflow): New function.
23176         (instrument_object_size): Formatting fix.
23177         (pass_ubsan::execute): Call instrument_pointer_overflow
23178         and maybe_instrument_pointer_overflow.
23179         * internal-fn.c (expand_UBSAN_PTR): New function.
23180         * ubsan.h (ubsan_expand_ptr_ifn): Declare.
23181         * sanitizer.def (__ubsan_handle_pointer_overflow,
23182         __ubsan_handle_pointer_overflow_abort): New builtins.
23183         * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
23184         * internal-fn.def (UBSAN_PTR): New internal function.
23185         * opts.c (sanitizer_opts): Add pointer-overflow.
23186         * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
23187         * fold-const.c (build_range_check): Compute pointer range check in
23188         integral type if pointer arithmetics would be needed.  Formatting
23189         fixes.
23191 2017-07-28  Martin Liska  <mliska@suse.cz>
23193         PR sanitizer/81460
23194         * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
23195         parameters that are of a variable-length.
23197 2017-07-28  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23199         * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h.  Add
23200         rs6000/biarch64.h.
23201         * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
23202         (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
23203         (CRT_CALL_STATIC_FUNCTION): Likewise.
23204         (ASM_DEFAULT_SPEC): New define.
23205         (ASM_SPEC32): Likewise.
23206         (ASM_SPEC64): Likewise.
23207         (ASM_SPEC_COMMON): Likewise.
23208         (ASM_SPEC): Likewise.
23209         (INVALID_64BIT): Likewise.
23210         (LINK_OS_DEFAULT_SPEC): Likewise.
23211         (LINK_OS_SPEC32): Likewise.
23212         (LINK_OS_SPEC64): Likewise.
23213         (POWERPC_LINUX): Likewise.
23214         (PTRDIFF_TYPE): Likewise.
23215         (RESTORE_FP_PREFIX): Likewise.
23216         (RESTORE_FP_SUFFIX): Likewise.
23217         (SAVE_FP_PREFIX): Likewise.
23218         (SAVE_FP_SUFFIX): Likewise.
23219         (SIZE_TYPE): Likewise.
23220         (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
23221         (TARGET_64BIT): Likewise.
23222         (TARGET_64BIT): Likewise.
23223         (TARGET_AIX): Likewise.
23224         (WCHAR_TYPE_SIZE): Likewise.
23225         (WCHAR_TYPE): Undefine.
23226         (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
23227         (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
23228         (CPP_OS_RTEMS_SPEC): Delete.
23229         (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems.  Add
23230         asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
23231         link_os_spec64.
23232         * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
23234 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23236         PR tree-optimization/81578
23237         * tree-parloops.c (build_new_reduction): Bail out if
23238         reduction_code isn't one of the standard OpenMP reductions.
23239         Move the details printing after that decision.
23241 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23243         * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23244         related to reload_in_progress.
23245         (splat_input_operand): Likewise.
23246         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23247         Delete prototype.
23248         * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23249         field.
23250         (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23251         (TARGET_INSTANTIATE_DECLS): Likewise.
23252         (legitimate_indexed_address_p): Delete reload_in_progress code.
23253         (rs6000_debug_legitimate_address_p): Likewise.
23254         (rs6000_eliminate_indexed_memrefs): Likewise.
23255         (rs6000_emit_le_vsx_store): Likewise.
23256         (rs6000_emit_move_si_sf_subreg): Likewise.
23257         (rs6000_emit_move): Likewise.
23258         (register_to_reg_type): Likewise.
23259         (rs6000_pre_atomic_barrier): Likewise.
23260         (rs6000_machopic_legitimize_pic_address): Likewise.
23261         (rs6000_allocate_stack_temp): Likewise.
23262         (rs6000_address_for_fpconvert): Likewise.
23263         (rs6000_address_for_altivec): Likewise.
23264         (rs6000_secondary_memory_needed_rtx): Delete function.
23265         (rs6000_check_sdmode): Likewise.
23266         (rs6000_alloc_sdmode_stack_slot): Likewise.
23267         (rs6000_instantiate_decls): Likewise.
23268         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23269         * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23270         Delete reload_in_progress.
23271         (*vec_reload_and_plus_<mptrsize>): Likewise.
23272         * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23273         (vsx_div_v2di): Likewise.
23274         (vsx_udiv_v2di): Likewise.
23276 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23278         * config/rs6000/rs6000.opt (mlra): Replace with stub.
23279         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23280         * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23281         (rs6000_debug_reg_global): Delete print of LRA status.
23282         (rs6000_option_override_internal): Delete dead LRA related code.
23283         (rs6000_lra_p): Delete function.
23284         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23286 2017-07-27  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23288         * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23289         * config/riscv/rtems.h: New file.
23291 2017-07-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23292             Sudakshina Das  <sudi.das@arm.com>
23294         * config/aarch64/aarch64.md
23295         (define_split for and<mode>3nr_compare): Move
23296         non aarch64_logical_operand to a register.
23297         (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23298         register immediate operand to a register.
23299         * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23301 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23303         PR middle-end/81564
23304         * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23306 2017-07-27  Richard Biener  <rguenther@suse.de>
23308         PR tree-optimization/81573
23309         PR tree-optimization/81494
23310         * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23311         multi defuse cycle case.
23313 2017-07-27  Richard Biener  <rguenther@suse.de>
23315         PR tree-optimization/81571
23316         * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23317         PHIs.
23319 2017-07-27  Eric Botcazou  <ebotcazou@adacore.com>
23321         * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23322         earlier and only if MASK_FPU is set.  Adjust formatting.
23324 2017-07-27  Martin Liska  <mliska@suse.cz>
23326         * opt-functions.awk: Add validation of value of Init.
23327         * optc-gen.awk: Pass new argument.
23329 2017-07-27  Martin Liska  <mliska@suse.cz>
23331         * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23332         Fix wrong condition.
23334 2017-07-27  Martin Liska  <mliska@suse.cz>
23336         * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23337         BBs and edges seen by autoFDO.
23339 2017-07-27  Richard Biener  <rguenther@suse.de>
23341         PR tree-optimization/81502
23342         * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23343         with incompatible but same sized type.
23344         (execute_update_addresses_taken): Likewise.
23346 2017-07-27  James Greenhalgh  <james.greenhalgh@arm.com>
23348         * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23349         flag_tree_loop_vectorize rather than flag_tree_vectorize.
23351 2017-07-27  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23353         PR target/81534
23354         * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23355         ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23356         Change s_operand to memory_operand.
23358 2017-07-27  Richard Sandiford  <richard.sandiford@linaro.org>
23360         * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23361         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23362         (rs6000_emit_le_vsx_permute): ...this.  Take the destination as input.
23363         Emit instructions rather than returning an expression.  Handle TFmode
23364         and KFmode by casting to TImode.
23365         (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23366         (rs6000_emit_le_vsx_store): Likewise.
23367         * config/rs6000/vsx.md (VSX_TI): New iterator.
23368         (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23369         (*vsx_le_undo_permute_<mode>): Likewise.
23370         (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23371         emit the split sequence.
23372         (*vsx_le_perm_store_<mode>): Likewise.
23374 2017-07-27  Jakub Jelinek  <jakub@redhat.com>
23376         PR tree-optimization/81555
23377         PR tree-optimization/81556
23378         * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23379         if true, force CHANGED for the recursive invocation.
23380         (reassociate_bb): Remember original length of ops array, pass
23381         len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23383         * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23384         attributes for noipa attribute.  For naked attribute use
23385         lookup_attribute first before lookup_attribute_spec.
23386         * final.c (rest_of_handle_final): Disable IPA RA for functions with
23387         noipa attribute.
23388         * ipa-visibility.c (non_local_p): Fix comment typos.  Return true
23389         for functions with noipa attribute.
23390         (cgraph_externally_visible_p): Return true for functions with noipa
23391         attribute.
23392         * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23393         for functions with noipa attribute.
23394         * doc/extend.texi: Document noipa function attribute.
23395         * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23396         also for functions with noipa attribute.
23397         (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23399 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23401         * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23402         vec_unalign_load_cost and vec_unalign_store_cost.
23404 2017-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
23406         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23407         -mvsx-small-integer option.
23408         (ISA_3_0_MASKS_IEEE): Likewise.
23409         (OTHER_VSX_VECTOR_MASKS): Likewise.
23410         (POWERPC_MASKS): Likewise.
23411         * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23412         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23413         code, only testing for DImode being allowed in non-VSX floating
23414         point registers.
23415         (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23416         to TARGET_P8_VECTOR test.  Remove redundant VSX test inside of
23417         another VSX test.
23418         (rs6000_option_override_internal): Delete -mvsx-small-integer.
23419         (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23420         TARGET_P8_VECTOR test.
23421         (rs6000_secondary_reload_simple_move): Likewise.
23422         (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23423         since TARGET_P9_VECTOR was already tested.
23424         (rs6000_opt_masks): Remove -mvsx-small-integer.
23425         * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23426         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23427         used.
23428         (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23429         test for TARGET_VEXTRACTUB was used, and that uses
23430         TARGET_P9_VECTOR.
23431         (p9 extract splitter): Likewise.
23432         (vsx_extract_<mode>_di_p9): Likewise.
23433         (vsx_extract_<mode>_store_p9): Likewise.
23434         (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23435         for TARGET_P9_VECTOR was used.  Delete code that is now dead with
23436         the elimination of TARGET_VSX_SMALL_INTEGER.
23437         (vsx_extract_<mode>_p8): Likewise.
23438         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23439         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23440         (vsx_set_<mode>_p9): Likewise.
23441         (vsx_set_v4sf_p9): Likewise.
23442         (vsx_set_v4sf_p9_zero): Likewise.
23443         (vsx_insert_extract_v4sf_p9): Likewise.
23444         (vsx_insert_extract_v4sf_p9_2): Likewise.
23445         * config/rs6000/rs6000.md (sign extend splitter): Change
23446         TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23447         (floatsi<mode>2_lfiwax_mem): Likewise.
23448         (floatunssi<mode>2_lfiwzx_mem): Likewise.
23449         (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23450         since a test for TARGET_P9_VECTOR was used.
23451         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23452         (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23453         (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23454         (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23455         TARGET_P8_VECTOR test.
23456         (fix_trunc<mode>si2_stfiwx): Likewise.
23457         (fix_trunc<mode>si2_internal): Likewise.
23458         (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23459         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23460         used.
23461         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23462         (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23463         TARGET_P8_VECTOR test.
23464         (fixuns_trunc<mode>si2_stfiwx): Likewise.
23465         (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23466         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23467         used.
23468         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23469         (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23470         since a test for TARGET_P9_VECTOR was used.
23471         (splitter for loading small constants): Likewise.
23473 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23475         * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23476         vec_fp_stmt_cost.
23478 2017-07-26  H.J. Lu  <hongjiu.lu@intel.com>
23480         PR target/81563
23481         * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23482         (fp_valid_at): Likewise.
23484 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23486         * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23487         (qdf24xx_addrcost_table): Likewise.
23488         (cortexa57_tunings): Update to use generic_branch_cost.
23489         (cortexa72_tunings): Likewise.
23490         (cortexa73_tunings): Likewise.
23491         (qdf24xx_tunings): Likewise.
23493 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23495         * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23496         (thunderx2t99_branch_cost): Likewise.
23497         (cortexa35_tunings): Update to use generic_branch_cost.
23498         (cortexa53_tunings): Likewise.
23499         (cortexa57_tunings): Likewise.
23500         (cortexa72_tunings): Likewise.
23501         (cortexa73_tunings): Likewise.
23502         (thunderx2t99_tunings): Likewise.
23504 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23506         * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23507         (sparc_option_override): Honour MASK_FSMULD.
23508         * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23509         * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23510         * config/sparc/sparc.opt (mfsmuld): New option.
23511         * doc/invoke.texi (mfsmuld): Document option.
23513 2017-07-26  Marek Polacek  <polacek@redhat.com>
23515         PR middle-end/70992
23516         * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23518 2017-07-26  Richard Biener  <rguenther@suse.de>
23520         * gimple-match-head.c (do_valueize): Return OP if valueize
23521         returns NULL_TREE.
23522         (get_def): New helper to get at the def stmt of a SSA name
23523         if valueize allows.
23524         * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23525         do_valueize to get at the def stmt.
23526         (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23528 2017-07-26  Wilco Dijkstra  <wdijkstr@arm.com>
23530         PR middle-end/46932
23531         * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23533 2017-07-26  Martin Liska  <mliska@suse.cz>
23535         PR sanitize/81186
23536         * function.c (expand_function_start): Make expansion of
23537         nonlocal_goto_save_area after parm_birth_insn.
23539 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23541         * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23542         from all CPU target flags enable members.
23544 2017-07-26  Richard Biener  <rguenther@suse.de>
23546         * genmatch.c (dt_simplify::gen): Make iterator vars const.
23547         (decision_tree::gen): Make 'type' const.
23548         (write_predicate): Likewise.
23550 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23552         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23553         Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23554         (rs6000_option_override_internal): Likewise.
23555         (rs6000_expand_vector_set): Likewise.
23556         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23557         (TARGET_UPPER_REGS_SF): Likewise.
23558         (TARGET_UPPER_REGS_DI): Likewise.
23559         (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23560         (TARGET_DIRECT_MOVE_64BIT): Likewise.
23561         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23562         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23563         (Splitters for DI constants in Altivec registers): Likewise.
23564         * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23565         (vsx_set_v4sf_p9): Likewise.
23566         (vsx_set_v4sf_p9_zero): Likewise.
23567         (vsx_insert_extract_v4sf_p9): Likewise.
23568         (vsx_insert_extract_v4sf_p9_2): Likewise.
23570 2017-07-25  Carl Love  <cel@us.ibm.com>
23572         * doc/extend.texi: Update the built-in documentation file for the
23573         existing built-in functions
23574         vector signed char vec_cnttz (vector signed char);
23575         vector unsigned char vec_cnttz (vector unsigned char);
23576         vector signed short vec_cnttz (vector signed short);
23577         vector unsigned short vec_cnttz (vector unsigned short);
23578         vector signed int vec_cnttz (vector signed int);
23579         vector unsigned int vec_cnttz (vector unsigned int);
23580         vector signed long long vec_cnttz (vector signed long long);
23581         vector unsigned long long vec_cnttz (vector unsigned long long);
23583 2017-07-25  Andrew Pinski  <apinski@cavium.com>
23585         * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23586         accesses where the use is for the first operand of a BIT_INSERT.
23588 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23590         PR bootstrap/81521
23591         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23592         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23594 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23596         * config/i386/gstabs.h: Delete.
23597         * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23599 2017-07-25  Uros Bizjak  <ubizjak@gmail.com>
23601         * config/i386/i386.c (ix86_decompose_address): Do not check for
23602         register RTX when looking at index_reg or base_reg.
23603         * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23605 2017-07-25  Eric Botcazou  <ebotcazou@adacore.com>
23607         * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23608         to update EH info here.
23610 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23612         * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23614 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23616         * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23618 2017-07-25  Torsten Duwe  <duwe@suse.de>
23620         * common.opt: Introduce -fpatchable-function-entry
23621         command line option, and its variables function_entry_patch_area_size
23622         and function_entry_patch_area_start.
23623         * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23624         including a two-value parser.
23625         * target.def (print_patchable_function_entry): New target hook.
23626         * targhooks.h (default_print_patchable_function_entry): New function.
23627         * targhooks.c (default_print_patchable_function_entry): Likewise.
23628         * toplev.c (process_options): Switch off IPA-RA if
23629         patchable function entries are being generated.
23630         * varasm.c (assemble_start_function): Look at the
23631         patchable-function-entry command line switch and current
23632         function attributes and maybe generate NOP instructions by
23633         calling the print_patchable_function_entry hook.
23634         * doc/extend.texi: Document patchable_function_entry attribute.
23635         * doc/invoke.texi: Document -fpatchable_function_entry
23636         command line option.
23637         * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23638         New target hook.
23639         * doc/tm.texi: Re-generate.
23641 2017-07-25  Jakub Jelinek  <jakub@redhat.com>
23643         PR target/81532
23644         * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23645         TARGET_AVX512DQ rather than TARGET_AVX512BW.
23647 2017-07-25  Tamar Christina  <tamar.christina@arm.com>
23649         * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23651 2017-07-25  Richard Biener  <rguenther@suse.de>
23653         PR tree-optimization/81455
23654         * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23655         not walk in cycles when looking for guards.
23657 2017-07-25  Richard Biener  <rguenther@suse.de>
23659         PR tree-optimization/81529
23660         * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23661         when optimizing backedge uses.
23663 2017-07-25  David Edelsohn  <dje.gcc@gmail.com>
23665         * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23666         character for AIX.
23667         * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23668         to dl_section_ref.  On AIX, append an expression to subtract
23669         the size of the section length to dl_section_ref.
23671 2017-07-25  Segher Boessenkool  <segher@kernel.crashing.org>
23673         * configure.ac: If any of the config.* scripts fail, exit 1.
23674         * configure: Regenerate.
23676 2017-07-25  Richard Biener  <rguenther@suse.de>
23678         PR middle-end/81546
23679         * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23680         of immediate uses, be more verbose on errors.
23682 2017-07-25  Richard Biener  <rguenther@suse.de>
23684         PR tree-optimization/81510
23685         * tree-vect-loop.c (vect_is_simple_reduction): When the
23686         reduction stmt is not inside the loop bail out.
23688 2017-07-25  Richard Biener  <rguenther@suse.de>
23690         PR tree-optimization/81303
23691         * tree-vect-loop-manip.c (vect_loop_versioning): Build
23692         profitability check against LOOP_VINFO_NITERSM1.
23694 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23696         * domwalk.c (cmp_bb_postorder): Simplify.
23697         (sort_bbs_postorder): New function.  Use it...
23698         (dom_walker::walk): ...here to optimize common cases.
23700 2017-07-25  Martin Liska  <mliska@suse.cz>
23702         PR ipa/81520
23703         * ipa-visibility.c (function_and_variable_visibility): Make the
23704         redirection just on target that supports aliasing.
23705         Fix GNU coding style.
23707 2017-07-25  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23709         PR libgcc/61152
23710         * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23711         Format changes.
23712         * config/arm/rtems.h: Likewise.
23713         * config/bfin/rtems.h: Likewise.
23714         * config/i386/rtemself.h: Likewise.
23715         * config/lm32/rtems.h: Likewise.
23716         * config/m32c/rtems.h: Likewise.
23717         * config/m68k/rtemself.h: Likewise.
23718         * config/microblaze/rtems.h: Likewise.
23719         * config/mips/rtems.h: Likewise.
23720         * config/moxie/rtems.h: Likewise.
23721         * config/nios2/rtems.h: Likewise.
23722         * config/powerpcspe/rtems.h: Likewise.
23723         * config/rs6000/rtems.h: Likewise.
23724         * config/rtems.h: Likewise.
23725         * config/sh/rtems.h: Likewise.
23726         * config/sh/rtemself.h: Likewise.
23727         * config/sparc/rtemself.h: Likewise.
23729 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23731         PR 81487
23732         * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23733         * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23734         * tree-ssa-structalias.c (alias_get_name): Same.
23736 2017-07-25  Bin Cheng  <bin.cheng@arm.com>
23738         PR target/81414
23739         * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23740         instructions if no du chain is found.
23742 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23744         * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23746 2017-07-25  Richard Biener  <rguenther@suse.de>
23748         PR middle-end/81505
23749         * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23750         sticky.
23752 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23754         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23755         upper-regs options.
23756         (ISA_2_7_MASKS_SERVER): Likewise.
23757         (ISA_3_0_MASKS_IEEE): Likewise.
23758         (OTHER_P8_VECTOR_MASKS): Likewise.
23759         (OTHER_VSX_VECTOR_MASKS): Likewise.
23760         (POWERPC_MASKS): Likewise.
23761         (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23762         duplicate list of options.
23763         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23764         explicit -mupper-regs options.
23765         * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23766         -mupper-regs* options.  Delete -mvsx-scalar-memory, which was an
23767         alias for -mupper-regs-df.
23768         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23769         (rs6000_init_hard_regno_mode_ok): Likewise.
23770         (rs6000_option_override_internal): Likewise.
23771         (rs6000_opt_masks): Likewise.
23772         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23773         options in terms of whether -mvsx or -mpower8-vector was used.
23774         (TARGET_UPPER_REGS_DI): Likewise.
23775         (TARGET_UPPER_REGS_SF): Likewise.
23776         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23777         -mupper-regs-* options.
23779 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23781         * passes.c (emergency_dump_function): Print some empty lines and a
23782         header before the RTL dump.
23784 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23786         * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23788 2017-07-24  Wilco Dijkstra  <wdijkstr@arm.com>
23790         PR target/79041
23791         * config/aarch64/aarch64.c (aarch64_classify_symbol):
23792         Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23794 2017-07-24  Carl Love  <cel@us.ibm.com>
23796         * config/rs6000/rs6000-c.c: Add support for built-in functions
23797         vector float vec_extract_fp32_from_shorth (vector unsigned short);
23798         vector float vec_extract_fp32_from_shortl (vector unsigned short);
23799         * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23800         vec_extract_fp_from_shortl): Add defines for the two builtins.
23801         * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23802         VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23803         new builtins.
23804         * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23805         (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23806         * doc/extend.texi: Update the built-in documentation file for the
23807         new built-in function.
23809 2017-07-24  Jakub Jelinek  <jakub@redhat.com>
23811         PR bootstrap/81521
23812         * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23813         documentation.
23814         * doc/generic.texi: Likewise.
23815         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23816         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23818 2017-07-24  Jackson Woodruff  <jackson.woodruff@arm.com>
23820         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23821         (aarch64_mls_elt_merge<mode>): Likewise.
23823 2017-07-23  Krister Walfridsson  <krister.walfridsson@gmail.com>
23825         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23826         having __cxa_atexit.
23828 2017-07-23  Michael Collison  <michael.collison@arm.com>
23830         * config/arm/arm.c (arm_option_override): Deprecate
23831         use of -mstructure-size-boundary.
23832         * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23833         * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23835 2017-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23837         PR target/80695
23838         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23839         Reduce cost estimate for direct moves.
23841 2017-07-23  Uros Bizjak  <ubizjak@gmail.com>
23843         PR target/80569
23844         * config/i386/i386.c (ix86_option_override_internal): Disable
23845         BMI, BMI2 and TBM instructions for -m16.
23847 2017-07-21  Carl Love  <cel@us.ibm.com>
23849         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23850         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23851         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23852         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23853         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23854         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23855         VMULOSW): New enum "unspec" values.
23856         (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23857         altivec_vmulosw): New patterns.
23858         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23859         VMULOSW): Add definitions.
23861 2017-07-21  Jim Wilson  <jim.wilson@linaro.org>
23863         * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23864         (qdf24xx): Likewise.
23865         * config/aarch64/aarch64-options-extensions.def (rdma); New.
23866         * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23867         (AARCH64_FL_V8_1): Renumber.
23868         (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23869         (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23870         * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23871         * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs.  Add
23872         rdma to feature modifiers list.
23874 2017-07-21  Yury Gribov  <tetra2005@gmail.com>
23876         PR middle-end/56727
23877         * ipa-visibility (function_and_variable_visibility): Convert
23878         recursive PLT call to direct call if appropriate.
23880 2017-07-21  Andrew Pinski  <apinski@cavium.com>
23882         * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23883         operand 1 to see if the types precision matches.
23884         * fold-const.c (operand_equal_p): Likewise.
23886 2017-07-21  Richard Biener  <rguenther@suse.de>
23888         PR tree-optimization/81303
23889         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23890         in datarefs vector.  Allow NULL dr0 for no peeling cost estimate.
23891         (vect_peeling_hash_get_lowest_cost): Adjust.
23892         (vect_enhance_data_refs_alignment): Likewise.  Use
23893         vect_get_peeling_costs_all_drs to compute the penalty for no
23894         peeling to match up costs.
23896 2017-07-21  Richard Biener  <rguenther@suse.de>
23898         PR tree-optimization/81500
23899         * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23900         we didn't identify a reduction path.
23902 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23903             Cesar Philippidis  <cesar@codesourcery.com>
23905         PR gcov-profile/81442
23906         * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23907         probabilities.
23909 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23911         PR lto/81430
23912         * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
23913         function.
23914         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
23915         nvptx_override_options_after_change.
23917 2017-07-21  Ulrich Drepper  <drepper@redhat.com>
23919         * dwarf2out.c (output_file_names): Avoid double testing for
23920         dwarf_version >= 5.
23922 2017-07-21  Georg-Johann Lay  <avr@gjlay.de>
23924         * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
23926 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23928         * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
23929         hot/cold regions.
23930         (try_crossjump_to_edge): Do not punt on partitioned functions.
23932 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23934         * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
23935         Put all BBs reachable only via paths crossing cold region to cold
23936         region.
23937         * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
23939 2016-07-21  Richard Biener  <rguenther@suse.de>
23941         PR tree-optimization/81303
23942         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
23943         into account prologue and epilogue iterations when raising
23944         min_profitable_iters to sth at least covering one vector iteration.
23946 2017-07-21  Tamar Christina  <tamar.christina@arm.com>
23948         * config/arm/arm.c (arm_test_cpu_arch_dat):
23949         Check for overlap.
23951 2017-07-20  Nathan Sidwell  <nathan@acm.org>
23953         Remove TYPE_METHODS.
23954         * tree.h (TYPE_METHODS): Delete.
23955         * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
23956         * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
23957         (dbxout_type_methods): Scan TYPE_FIELDS.
23958         (dbxout_type): Don't check TYPE_METHODS here.
23959         * function.c (use_register_for_decl): Always ignore register for
23960         class types when not optimizing.
23961         * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
23962         * tree.c (free_lang_data_in_type): Stitch out member functions and
23963         templates from TYPE_FIELDS.
23964         (build_distinct_type_copy, verify_type_variant,
23965         verify_type): Member fns are on TYPE_FIELDS.
23966         * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
23967         * tree-pretty-print.c (dump_generic_node): Likewise.
23969 2017-07-20  Jakub Jelinek  <jakub@redhat.com>
23971         PR target/80846
23972         * config/i386/i386.c (ix86_expand_vector_init_general): Handle
23973         V2TImode and V4TImode.
23974         (ix86_expand_vector_extract): Likewise.
23975         * config/i386/sse.md (VMOVE): Enable V4TImode even for just
23976         TARGET_AVX512F, instead of only for TARGET_AVX512BW.
23977         (ssescalarmode): Handle V4TImode and V2TImode.
23978         (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
23979         (*vec_extractv2ti, *vec_extractv4ti): New insns.
23980         (VEXTRACTI128_MODE): New mode iterator.
23981         (splitter for *vec_extractv?ti first element): New.
23982         (VEC_INIT_MODE): New mode iterator.
23983         (vec_init<mode>): Consolidate 3 expanders into one using
23984         VEC_INIT_MODE mode iterator.
23986 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23988         * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
23989         non_spilled_static_chain_regno_p.
23991 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23993         * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
23995 2017-07-20  Jan Hubicka  <hubicka@ucw.cz>
23997         * bb-reorder.c (connect_traces): Allow copying of blocks within
23998         single partition.
24000 2017-07-20  Richard Biener  <rguenther@suse.de>
24002         * gimple.h (gimple_phi_result): Add gphi * overload.
24003         (gimple_phi_result_ptr): Likewise.
24004         (gimple_phi_arg): Likewise.  Adjust index assert to only
24005         allow actual argument accesses rather than all slots available
24006         by capacity.
24007         (gimple_phi_arg_def): Add gphi * overload.
24008         * tree-phinodes.c (make_phi_node): Initialize only actual
24009         arguments.
24010         (resize_phi_node): Clear memory not covered by old node,
24011         do not initialize excess argument slots.
24012         (reserve_phi_args_for_new_edge): Initialize new argument slot
24013         completely.
24015 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
24017         PR tree-optimization/81388
24018         Revert r238585:
24019         2016-07-21  Bin Cheng  <bin.cheng@arm.com>
24021         * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
24022         by removing computation of may_be_zero.
24024 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24025             Tom de Vries  <tom@codesourcery.com>
24027         PR middle-end/81030
24028         * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
24029         when gimple level profile disagrees with what RTL expander did.
24031 2017-07-20  Richard Biener  <rguenther@suse.de>
24033         PR tree-optimization/61171
24034         * tree-vectorizer.h (slp_instance): Add reduc_phis member.
24035         (vect_analyze_stmt): Add slp instance parameter.
24036         (vectorizable_reduction): Likewise.
24037         * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
24038         (vect_is_simple_reduction): Deal with chains not detected
24039         as SLP reduction chain, specifically not properly associated
24040         chains containing a mix of plus/minus.
24041         (get_reduction_op): Remove.
24042         (get_initial_defs_for_reduction): Simplify, pass in whether
24043         this is a reduction chain, pass in the SLP node for the PHIs.
24044         (vect_create_epilog_for_reduction): Get the SLP instance as
24045         arg and adjust.
24046         (vectorizable_reduction): Get the SLP instance as arg.
24047         During analysis remember the SLP node with the PHIs in the
24048         instance.  Simplify getting at the vectorized reduction PHIs.
24049         * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
24050         through SLP instance.
24051         (vect_slp_analyze_operations): Likewise.
24052         * tree-vect-stms.c (vect_analyze_stmt): Likewise.
24053         (vect_transform_stmt): Likewise.
24055 2017-07-20  Tom de Vries  <tom@codesourcery.com>
24057         PR tree-optimization/81489
24058         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
24059         read of phi arg location to before loop that modifies phi.
24061 2017-07-20  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
24063         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
24064         New pattern.
24066 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24068         PR middle-end/81331
24069         * except.c (execute): Fix ordering issue.
24071 2018-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24073         PR rtl-optimization/81423
24074         * combine.c (make_compound_operation_int): Don't try to optimize
24075         the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
24077 2017-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24079         PR rtl-optimization/81423
24080         * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
24081         with a constant that is -1 in the truncated to mode.
24083 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24085         * predict.c (propagate_unlikely_bbs_forward): Break out from ...
24086         (determine_unlikely_bbs): ... here.
24087         * predict.h (propagate_unlikely_bbs_forward): Declare.
24088         * cfgexpand.c (pass_expand::execute): Use it.
24089         * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
24090         unlikely edges.
24091         (find_rarely_executed_basic_blocks_and_crossing_edges): Use
24092         propagate_unlikely_bbs_forward.
24094 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24096         PR middle-end/81331
24097         * except.c (maybe_add_nop_after_section_switch): New function.
24098         (execute): Use it.
24100 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24102         * gimple.h (gimple_phi_set_arg): Make assert more strict.
24104 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24106         * gimple.h (gimple_phi_arg): Make assert more strict.
24108 2017-07-19  Steven Munroe  <munroesj@gcc.gnu.org>
24110         * config.gcc (powerpc*-*-*): Add mmintrin.h.
24111         * config/rs6000/mmintrin.h: New file.
24112         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
24114 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24116         PR tree-optimization/81346
24117         * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
24119 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24121         * config/nvptx/nvptx.md (VECIM): Add V2DI.
24123 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24125         * config/nvptx/nvptx-modes.def: Add V2DImode.
24126         * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
24127         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
24128         (nvptx_output_mov_insn): Handle lack of mov.b128.
24129         (nvptx_print_operand): Handle 'H' and 'L' codes.
24130         (nvptx_vector_mode_supported): Allow V2DImode.
24131         (nvptx_preferred_simd_mode): New function.
24132         (nvptx_data_alignment): New function.
24133         (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
24134         nvptx_preferred_simd_mode.
24135         * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
24136         64 to 128 bits.
24137         (DATA_ALIGNMENT): Define.  Set to nvptx_data_alignment.
24139 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24141         * config/nvptx/nvptx-modes.def: New file.  Add V2SImode.
24142         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
24143         (nvptx_vector_mode_supported): New function.  Allow V2SImode.
24144         (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
24145         * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
24146         (mov<VECIM>_insn): New define_insn.
24147         (define_expand "mov<VECIM>): New define_expand.
24149 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24151         * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
24153 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24155         PR tree-optimization/81346
24156         * fold-const.h (fold_div_compare, range_check_type): Declare.
24157         * fold-const.c (range_check_type): New function.
24158         (build_range_check): Use range_check_type.
24159         (fold_div_compare): No longer static, rewritten into
24160         a match.pd helper function.
24161         (fold_comparison): Don't call fold_div_compare here.
24162         * match.pd (X / C1 op C2): New optimization using fold_div_compare
24163         as helper function.
24165 2017-07-19  Nathan Sidwell  <nathan@acm.org>
24167         * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
24168         (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
24169         * tree.c (find_decls_types_r, verify_type): Use
24170         TYPE_{MIN,MAX}_VALUE_RAW.
24171         * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
24172         (hash_tree): Likewise.
24173         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
24174         Likewise.
24175         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
24176         Likewise.
24178 2017-07-18  Tom de Vries  <tom@codesourcery.com>
24180         PR middle-end/81464
24181         * omp-expand.c (expand_omp_for_static_chunk): Handle
24182         equal-argument loop exit phi.
24184 2017-07-18  Uros Bizjak  <ubizjak@gmail.com>
24186         PR target/81471
24187         * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
24188         (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
24189         operand 2 predicate.
24190         (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
24191         operand 2 predicate.
24192         (ror,rol -> rorx splitters): Use const_int_operand as
24193         operand 2 predicate.
24195 2017-06-18  Richard Biener  <rguenther@suse.de>
24197         PR tree-optimization/81410
24198         * tree-vect-stmts.c (vectorizable_load): Properly adjust for
24199         the gap in the ! slp_perm SLP case after each group.
24201 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24203         PR middle-end/81463
24204         * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
24205         again.
24207 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24209         PR middle-end/81462
24210         * predict.c (set_even_probabilities): Cleanup; do not affect
24211         probabilities that are already known.
24212         (combine_predictions_for_bb): Call even when count is set.
24214 2017-07-18  Nathan Sidwell  <nathan@acm.org>
24216         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
24217         TYPE_MAX_VALUE.
24219 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24221         PR target/81408
24222         * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
24223         optimization for loop niter analysis.
24225 2017-07-18  Georg-Johann Lay  <avr@gjlay.de>
24227         PR target/81473
24228         * config/avr/avr.c (avr_optimize_casesi): Don't use
24229         INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
24231 2017-07-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
24233         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
24234         body_cost_vec from _vect_peel_extended_info.
24235         (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
24236         (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24237         npeel.
24239 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24241         * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24243 2017-07-18  Richard Biener  <rguenther@suse.de>
24245         PR tree-optimization/80620
24246         PR tree-optimization/81403
24247         * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24248         info when re-using a VN table entry.
24250 2017-07-18  Richard Biener  <rguenther@suse.de>
24252         PR tree-optimization/81418
24253         * tree-vect-loop.c (vectorizable_reduction): Properly compute
24254         vectype_in.  Verify that with lane-reducing reduction operations
24255         we have a single def-use cycle.
24257 2017-07-17  Carl Love  <cel@us.ibm.com>
24259         Revert commit r249424   2017-06-20  Carl Love  <cel@us.ibm.com>
24261         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24262         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24263         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24264         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24265         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24266         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24267         VMULOSW): New enum "unspec" values.
24268         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24269         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24270         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24271         altivec_vmulosw): New patterns.
24272         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24273         VMULOSW): Add definitions.
24275 2017-07-17  Uros Bizjak  <ubizjak@gmail.com>
24277         * config/alpha/alpha.c: Include predict.h.
24279 2017-07-17  Yury Gribov  <tetra2005@gmail.com>
24281         * tree-vrp.c (compare_assert_loc): Fix comparison function
24282         to return predictable results.
24284 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24286         * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24287         option.
24288         (subdi3): Likewise.
24289         * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24290         * doc/invoke.texi (mexpand-adddi): Update text.
24292 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24294         * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24295         that also clobbers the CC register. The old expand code is moved
24296         to ...
24297         (*arc_clzsi2): ... here.
24298         (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24299         the CC register. The old expand code is moved to ...
24300         (arc_ctzsi2): ... here.
24302 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24304         * config/arc/arc.opt (mindexed-loads): Use initial value
24305         TARGET_INDEXED_LOADS_DEFAULT.
24306         (mauto-modify-reg): Use initial value
24307         TARGET_AUTO_MODIFY_REG_DEFAULT.
24308         * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24309         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24310         * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24311         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24313 2017-07-17  Martin Liska  <mliska@suse.cz>
24315         PR sanitizer/81302
24316         * opts.c (finish_options): Do not allow -fgnu-tm
24317         w/ -fsanitize={kernel-,}address.  Say sorry.
24319 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24321         PR target/81369
24322         * tree-loop-distribution.c (classify_partition): Only assert on
24323         numer of iterations.
24324         (merge_dep_scc_partitions): Delete prameter.  Update function call.
24325         (distribute_loop): Remove code handling loop with unknown niters.
24326         (pass_loop_distribution::execute): Skip loop with unknown niters.
24328 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24330         PR target/81369
24331         * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24332         function sort_partitions_by_post_order.
24334 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24336         PR tree-optimization/81374
24337         * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24338         the max index of basic blocks, rather than number of basic blocks.
24340 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24342         * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24343         proto.
24344         (arc_legitimate_pic_operand_p): Likewise.
24345         * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24346         function.
24347         (arc_needs_pcl_p): Likewise.
24348         (arc_legitimate_pc_offset_p): Likewise.
24349         (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24350         function is also used in constrains.md.
24351         (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24352         validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24353         PLUS.  Only return true/false in known cases, otherwise assert.
24354         (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24355         is already called in arc_legitimate_constant_p.
24356         * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24357         pic addresses.
24358         (LEGITIMATE_PIC_OPERAND_P): Use
24359         arc_raw_symbolic_reference_mentioned_p function.
24360         * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24361         function.
24362         (Cal): Likewise.
24363         (C32): Likewise.
24365 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24366         Andrew Burgess  <andrew.burgess@embecosm.com>
24368         * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24369         (arc_return_address_register): New function.
24370         * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24371         (arc_handle_fndecl_attribute): Add naked attribute.
24372         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24373         (TARGET_WARN_FUNC_RETURN): Likewise.
24374         (arc_allocate_stack_slots_for_args): New function.
24375         (arc_warn_func_return): Likewise.
24376         (machine_function): Change type fn_type.
24377         (arc_compute_function_type): Consider new naked function type,
24378         change function return type.
24379         (arc_must_save_register): Adapt to handle new
24380         arc_compute_function_type's return type.
24381         (arc_expand_prologue): Likewise.
24382         (arc_expand_epilogue): Likewise.
24383         (arc_return_address_regs): Delete.
24384         (arc_return_address_register): New function.
24385         (arc_epilogue_uses): Use above function.
24386         * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24387         (arc_function_type): Change encoding, add naked type.
24388         (ARC_INTERRUPT_P): Change to handle the new encoding.
24389         (ARC_FAST_INTERRUPT_P): Likewise.
24390         (ARC_NORMAL_P): Define.
24391         (ARC_NAKED_P): Likewise.
24392         (arc_compute_function_type): Delete prototype.
24393         * config/arc/arc.md (in_ret_delay_slot): Use
24394         arc_return_address_register function.
24395         (simple_return): Likewise.
24396         (p_return_i): Likewise.
24398 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24400         PR tree-optimization/81428
24401         * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24402         can't be built for those types.
24404 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24406         Remove stuff dead since r239246.
24408         * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24409         * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24410         (avr_inform_devices): Remove dead stuff.
24412 2017-07-17  Tamar Christina  <tamar.christina@arm.com>
24414         * config/arm/arm_neon.h: Fix softp typo.
24416 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24418         PR tree-optimization/81365
24419         * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24420         aggregate moves onto bb predecessor edges, make sure there are no
24421         loads that could alias the lhs in between the start of bb and the
24422         loads from *phi.
24424 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24426         PR 80929
24427         * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24428         (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24429         [LSHIFTRT, outer_code = TRUNCATE]: Same.
24431 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24433         PR tree-optimization/81396
24434         * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24435         (init_symbolic_number): Initialize it to 1.
24436         (perform_symbolic_merge): Add n_ops from both operands into the new
24437         n_ops.
24438         (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24439         without base_addr as useless if they need more than one operation.
24440         (bswap_replace): Handle !bswap case for NULL base_addr.
24442 2017-07-17  Tom de Vries  <tom@codesourcery.com>
24444         PR target/81069
24445         * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24446         as possible.
24448 2017-07-17  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24450         * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24451         conditional builtin define __FIX_LEON3FT_B2BST.
24453 2017-07-17  Daniel Cederman  <cederman@gaisler.com>
24455         * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24456         MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24457         with -mfix-ut700.
24459 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24461         PR rtl-optimization/81424
24462         * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24463         to remove potential trapping from operands if -fnon-call-exceptions.
24465 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24467         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24468         profile_proability for scalling.
24469         * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24471 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24473         * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24475 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24477         * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24478         fixpoint arithmetics.
24480 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24482         * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24483         fixpoint arithmetics.
24485 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24487         * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24488         fixpoint arithmetics.
24490 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24492         * profile-count.h (profile_probability::from_reg_br_prob_note,
24493         profile_probability::to_reg_br_prob_note): New functions.
24494         * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24495         * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24496         * predict.c (probability_reliable_p): Update.
24497         (edge_probability_reliable_p): Update.
24498         (br_prob_note_reliable_p): Update.
24499         (invert_br_probabilities): Update.
24500         (add_reg_br_prob_note): New function.
24501         (combine_predictions_for_insn): Update.
24502         * asan.c (asan_clear_shadow): Update.
24503         * cfgbuild.c (compute_outgoing_frequencies): Update.
24504         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24505         (update_br_prob_note): Update.
24506         (rtl_verify_edges): Update.
24507         (purge_dead_edges): Update.
24508         (fixup_reorder_chain): Update.
24509         * emit-rtl.c (try_split): Update.
24510         * ifcvt.c (cond_exec_process_insns): Update.
24511         (cond_exec_process_if_block): Update.
24512         (dead_or_predicable): Update.
24513         * internal-fn.c (expand_addsub_overflow): Update.
24514         (expand_neg_overflow): Update.
24515         (expand_mul_overflow): Update.
24516         * loop-doloop.c (doloop_modify): Update.
24517         * loop-unroll.c (compare_and_jump_seq): Update.
24518         * optabs.c (emit_cmp_and_jump_insn_1): Update.
24519         * predict.h: Update.
24520         * reorg.c (mostly_true_jump): Update.
24521         * rtl.h: Update.
24522         * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24523         * config/alpha/alpha.c (emit_unlikely_jump): Update.
24524         * config/arc/arc.c: (emit_unlikely_jump): Update.
24525         * config/arm/arm.c: (emit_unlikely_jump): Update.
24526         * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24527         * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24528         * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24529         (ix86_print_operand): Update.
24530         (ix86_split_fp_branch): Update.
24531         (predict_jump): Update.
24532         * config/ia64/ia64.c (ia64_print_operand): Update.
24533         * config/mmix/mmix.c (mmix_print_operand): Update.
24534         * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24535         (rs6000_expand_split_stack_prologue): Update.
24536         * config/rs6000/rs6000.c: Update.
24537         * config/s390/s390.c (s390_expand_vec_strlen): Update.
24538         (s390_expand_vec_movstr): Update.
24539         (s390_expand_cs_tdsi): Update.
24540         (s390_expand_split_stack_prologue): Update.
24541         * config/sh/sh.c (sh_print_operand): Update.
24542         (expand_cbranchsi4): Update.
24543         (expand_cbranchdi4): Update.
24544         * config/sparc/sparc.c (output_v9branch): Update.
24545         * config/spu/spu.c (get_branch_target): Update.
24546         (ea_load_store_inline): Update.
24547         * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24548         * config/tilepro/tilepro.c: Update.
24550 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24552         * gimplify.c (mostly_copy_tree_r): Revert latest change.
24553         (gimplify_save_expr): Likewise.
24555 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24557         * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24559 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24561         * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24562         TV_IPA_FNSUMMARY.
24563         * timevar.def (TV_IPA_FNSUMMARY): Define.
24565 2017-07-16  Daniel Cederman  <cederman@gaisler.com>
24567         * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24568         to back store errata sensitive sequence from being generated.
24569         (sqrtdf2_fix): Likewise.
24571 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24573         * tree-ssa-threadupdate.c (compute_path_counts,
24574         update_joiner_offpath_counts): Use profile_probability.
24576 2017-07-15  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24578         Revert:
24579         2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24581         * config/arm/arm-c.c (arm_cpu_builtins): Define
24582         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24584 2017-07-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24586         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24587         array entries to represent __ieee128 versions of the
24588         scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24589         scalar_extract_sig, and scalar_insert_exp built-in functions.
24590         (altivec_resolve_overloaded_builtin): Add special case handling
24591         for the __builtin_scalar_insert_exp function, as represented by
24592         the P9V_BUILTIN_VEC_VSIEDP constant.
24593         * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24594         exponent support for __ieee128 argument.
24595         (VSESQP): Add scalar extract signature support for __ieee128
24596         argument.
24597         (VSTDCNQP): Add scalar test negative support for __ieee128
24598         argument.
24599         (VSIEQP): Add scalar insert exponent support for __int128 argument
24600         with __ieee128 result.
24601         (VSIEQPF): Add scalar insert exponent support for __ieee128
24602         argument with __ieee128 result.
24603         (VSTDCQP): Add scalar test data class support for __ieee128
24604         argument.
24605         (VSTDCNQP): Add overload support for scalar test negative with
24606         __ieee128 argument.
24607         (VSTDCQP): Add overload support for scalar test data class
24608         __ieee128 argument.
24609         * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24610         UNSPEC_VSX_SXSIGDP.
24611         (UNSPEC_VSX_SIEXPQP): New constant.
24612         (xsxexpqp): New insn for VSX scalar extract exponent quad
24613         precision.
24614         (xsxsigqp): New insn for VSX scalar extract significand quad
24615         precision.
24616         (xsiexpqpf): New insn for VSX scalar insert exponent quad
24617         precision with floating point argument.
24618         (xststdcqp): New expand for VSX scalar test data class quad
24619         precision.
24620         (xststdcnegqp): New expand for VSX scalar test negative quad
24621         precision.
24622         (xststdcqp): New insn to match expansions for VSX scalar test data
24623         class quad precision and VSX scalar test negative quad precision.
24624         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24625         special case operand checking to enforce that second operand of
24626         VSX scalar test data class with quad precision argument is a 7-bit
24627         unsigned literal.
24628         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24629         prototypes and descriptions of __ieee128 versions of
24630         scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24631         scalar_test_data_class, and scalar_test_neg built-in functions.
24633 2016-07-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24635         PR tree-optimization/81162
24636         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24637         replace a negate with an add.
24639 2017-07-14  James Greenhalgh  <james.greenhalgh@arm.com>
24641         * doc/invoke.texi (arm/-mcpu): Document +crypto.
24643 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24645         * config/arm/arm-c.c (arm_cpu_builtins): Define
24646         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24648 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24650         * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24651         (armv8-r): Set ARM Cortex-R52 as default CPU.
24652         * config/arm/arm-tables.opt: Regenerate.
24653         * config/arm/arm-tune.md: Regenerate.
24654         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24655         Cortex-R52.
24656         * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24657         extension for -mcpu=cortex-r52.
24659 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24661         * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24662         (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24663         * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24664         (fp-armv8): Define it as FP_ARMv8 only.
24665         config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24666         (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24667         TARGET_FPU_ARMV8.
24668         config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24669         TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24670         * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24671         first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24672         than TARGET_FPU_ARMV8.
24673         * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24674         __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24675         * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24676         TARGET_FPU_ARMV8.
24677         * config/arm/neon.md (neon_vrint): Likewise.
24678         (neon_vcvt): Likewise.
24679         (neon_<fmaxmin_op><mode>): Likewise.
24680         (<fmaxmin><mode>3): Likewise.
24681         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24682         * config/arm/predicates.md (arm_cond_move_operator): Check against
24683         TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24685 2017-07-14  Jackson Woodruff  <jackson.woodruff@arm.com>
24687         * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24688         to top of function.
24690 2017-07-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24692         * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24693         loop in comment with memset.
24695 2017-07-14  Martin Liska  <mliska@suse.cz>
24697         * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24698         * dwarf2out.c (is_java): Remove the function.
24699         (output_pubname): Remove usage of the function.
24700         (lower_bound_default): Remove usage of DW_LANG_Java.
24701         (gen_compile_unit_die): Likewise.
24702         * gcc.c: Remove compiler defaults for .java and .zip files.
24703         * gimple-expr.c (remove_suffix): Change as there's no longer
24704         extension than 4-letter one.
24705         * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24706         (gimplify_save_expr): Likewise.
24707         * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24708         as it's possible even for other languages than Java.
24709         * langhooks.h (struct lang_hooks): Remove Java from a comment.
24710         * lto-opts.c (lto_write_options): Remove reference to Java.
24711         * opts.c (strip_off_ending): Update file extension handling.
24712         * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24713         * tree-eh.c (lower_resx): Likewise.
24714         * tree.c (free_lang_data_in_type): Remove dead code.
24715         (find_decls_types_r): Likewise.
24716         (build_common_builtin_nodes): Remove Java from a comment.
24717         (verify_type): Remove dead code.
24718         * varasm.c (assemble_external): Remove Java from a comment.
24720 2017-07-14  Martin Liska  <mliska@suse.cz>
24722         * opts.c (finish_options): Add quotes.
24723         (common_handle_option): Likewise.
24725 2017-07-14  Martin Liska  <mliska@suse.cz>
24727         * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24728         * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24729         Remove N_SO_PASCAL.
24730         * dwarf2out.c (lower_bound_default): Do not handle
24731         DW_LANG_Pascal83.
24732         (gen_compile_unit_die): Likewise.
24733         * gcc.c: Remove default extension binding for GNU Pascal.
24734         * stmt.c: Remove Pascal language from a comment.
24735         * xcoffout.c: Likewise.
24737 2017-07-13  David Malcolm  <dmalcolm@redhat.com>
24739         PR c/81405
24740         * diagnostic-show-locus.c (fixit_cmp): New function.
24741         (layout::layout): Sort m_fixit_hints.
24742         (column_range::column_range): Assert that the values are valid.
24743         (struct char_span): New struct.
24744         (correction::overwrite): New method.
24745         (struct source_line): New struct.
24746         (line_corrections::add_hint): Add assertions.  Reimplement memcpy
24747         calls in terms of classes source_line and char_span, and
24748         correction::overwrite.
24749         (selftest::test_overlapped_fixit_printing_2): New function.
24750         (selftest::diagnostic_show_locus_c_tests): Call it.
24752 2017-07-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
24754         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24755         early if there is no lhs.
24757 2017-07-13  Martin Liska  <mliska@suse.cz>
24759         * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24760         (gen_reference_type_die): Likewise.
24761         * stor-layout.c: Remove Pascal-related comment.
24763 2017-07-13  Martin Liska  <mliska@suse.cz>
24765         * opts.c (finish_options): Add quotes to error messages.
24766         (parse_sanitizer_options): Likewise.
24768 2017-07-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24770         * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24772 2017-07-13  Richard Earnshaw  <rearnsha@arm.com>
24774         * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24776 2017-07-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
24778         * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24779         during expansion.
24780         * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24782 2017-07-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
24784         PR target/81193
24785         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24786         provides the hardware capability bits, define the macro
24787         __BUILTIN_CPU_SUPPORTS__.
24788         * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24789         if GLIBC does not provide the hardware capability bits.  Add a
24790         gcc_unreachable call if the built-in cpu function is neither
24791         __builtin_cpu_is nor __builtin_cpu_supports.
24792         (rs6000_get_function_versions_dispatcher): Change the warning
24793         that an old GLIBC is used which does not export the capability
24794         bits to be an error.
24795         * doc/extend.texi (target_clones attribute): Document the
24796         restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24797         (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24798         needed by __builtin_cpu_is and __builtin_cpu_supports.  Document
24799         the macros defined by GCC if the newer GLIBC is available.
24801 2017-07-12  Jeff Law  <law@redhat.com>
24803         * config/riscv/riscv.c: Remove unnecessary includes.  Reorder
24804         remaining includes slightly.
24805         * config/riscv/riscv-builtins.c: Include profile-count.h.
24807 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24809         PR target/79883
24810         * config/avr/avr.c (avr_set_current_function): In diagnostic
24811         messages: Quote keywords and (parts of) identifiers.
24812         [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24813         "INTERUPT".
24815 2017-07-12  Carl Love  <cel@us.ibm.com>
24817         * config/rs6000/rs6000-c.c: Add support for built-in functions
24818         vector bool char vec_revb (vector bool char);
24819         vector bool short vec_revb (vector short char);
24820         vector bool int vec_revb (vector bool int);
24821         vector bool long long vec_revb (vector bool long long);
24822         * doc/extend.texi: Update the built-in documentation file for the
24823         new built-in functions.
24825 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24827         * config/s390/s390.md: Remove movcc splitter.
24829 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24831         * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24832         load/store on condition.
24834 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24836         PR target/81407
24837         * config/avr/avr.c (avr_encode_section_info)
24838         [progmem && !TREE_READONLY]: Error if progmem object needs
24839         constructing.
24841 2017-07-11  Michael Collison  <michael.collison@arm.com>
24843         * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24844         New pattern.
24846 2017-07-11  Carl Love  <cel@us.ibm.com>
24848         * config/rs6000/rs6000-c.c: Add support for builtins
24849         vector unsigned int vec_parity_lsbb (vector signed int);
24850         vector unsigned int vec_parity_lsbb (vector unsigned int);
24851         vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24852         vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24853         vector unsigned long long vec_parity_lsbb (vector signed long long);
24854         vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24855         * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24856         * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24857         * doc/extend.texi: Update the built-in documentation file for the
24858         new built-in functions.
24860 2017-07-11  David Malcolm  <dmalcolm@redhat.com>
24862         * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24863         (layout::m_primary_loc): New field.
24864         (layout::layout): Initialize new field.  Move location filtering
24865         logic from here to...
24866         (layout::maybe_add_location_range): ...this new method.  Add
24867         support for filtering to just the lines already specified by other
24868         locations.
24869         (layout::will_show_line_p): New method.
24870         (gcc_rich_location::add_location_if_nearby): New method.
24871         (selftest::test_add_location_if_nearby): New test function.
24872         (selftest::diagnostic_show_locus_c_tests): Call it.
24873         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24874         New method.
24876 2017-07-11  Tom de Vries  <tom@codesourcery.com>
24878         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24879         (bb_first_real_insn): New function.
24880         (nvptx_single): Add extra initialization of broadcasted condition
24881         variables.
24883 2017-07-11  Nathan Sidwell  <nathan@acm.org>
24885         * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24887 2017-07-11  Georg-Johann Lay  <avr@gjlay.de>
24889         * doc/extend.texi (AVR Function Attributes): Remove weblink to
24890         Binutils doc as TEXI will mess them up.
24891         * doc/invoke.texi (AVR Options): Same here.
24893 2017-07-11  Daniel Cederman  <cederman@gaisler.com>
24895         * config/sparc/sparc.opt (mfix-ut700): New option.
24896         (mfix-gr712rc): Likewise.
24897         (sparc_fix_b2bst): New variable.
24898         * doc/invoke.texi (SPARC options): Document them.
24899         (ARM options): Fix warnings.
24900         * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24901         instructions to prevent sequences that can trigger the store-store
24902         errata for certain LEON3FT processors.
24903         (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24904         (sparc_option_override): Set sparc_fix_b2bst appropriately.
24905         * config/sparc/sparc.md (fix_b2bst): New attribute.
24906         (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24908 2017-07-10  Uros Bizjak  <ubizjak@gmail.com>
24910         PR target/81375
24911         * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
24912         (rcpps): Ditto.
24913         (*rsqrtsf2_sse): Ditto.
24914         (rsqrtsf2): Ditto.
24915         (div<mode>3): Macroize insn from divdf3 and divsf3
24916         using MODEF mode iterator.
24918 2017-07-10  Martin Sebor  <msebor@redhat.com>
24920         PR tree-optimization/80397
24921         * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
24922         instead of testing for equality to INTEGER_TYPE.
24924 2017-07-10  Vineet Gupta <vgupta@synopsys.com>
24926         * config.gcc: Remove uclibc from arc target spec.
24928 2017-07-10  Claudiu Zissulescu  <claziss@synopsys.com>
24930         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
24932 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24934         PR lto/80838
24935         * lto-wrapper.c (remove_option): New function.
24936         (merge_and_complain): Merge PIC/PIE options more realistically.
24938 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24940         Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
24942         PR target/20296
24943         PR target/81268
24944         * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
24945         (HAVE_AS_AVR_MGCCISR_OPTION):  If so, AC_DEFINE it.
24946         * config.in: Regenerate.
24947         * configure: Regenerate.
24948         * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
24949         * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
24950         * config/avr/avr.opt (-mgas-isr-prologues): New option and...
24951         (TARGET_GASISR_PROLOGUES): ...target mask.
24952         * common/config/avr/avr-common.c
24953         (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
24954         Set -mgas-isr-prologues.
24955         * config/avr/avr-passes.def (avr_pass_pre_proep): Add
24956         INSERT_PASS_BEFORE for it.
24957         * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
24958         * config/avr/avr.c (avr_option_override)
24959         [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
24960         (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
24961         (avr_attribute_table) <no_gccisr>: Add new function attribute.
24962         (avr_set_current_function) <is_no_gccisr>: Init machine field.
24963         (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
24964         and rtl_opt_pass.
24965         (make_avr_pass_pre_proep): New function.
24966         (emit_push_sfr) <treg>: Add argument to function and use it
24967         instead of TMP_REG.
24968         (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
24969         and set machine->gasisr.yes.
24970         (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
24971         (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
24972         __gcc_isr.n_pushed to .L__stack_usage.
24973         (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
24974         (avr_asm_final_postscan_insn): ...this new static function.
24975         * config/avr/avr.h (machine_function)
24976         <is_no_gccisr, use_L__stack_usage>: New fields.
24977         <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
24978         * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
24979         (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
24980         (gasisr, *gasisr): New expander and insn.
24981         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
24982         [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
24983         * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
24985 2017-07-10  Richard Earnshaw  <rearnsha@arm.com>
24987         * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
24988         in quoted strings.
24990 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24992         Move jump-tables out of .text again.
24994         PR target/81075
24995         * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
24996         (ASM_OUTPUT_ADDR_VEC): New function.
24997         (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
24998         (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
24999         INSN_ADDRESSes as asm comment.
25000         * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
25001         (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
25002         (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
25003         * config/avr/avr.md (*tablejump): Adjust comment.
25004         * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
25005         * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
25006         New detail.
25007         * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
25008         (avr_output_addr_vec): New proto.
25009         (avr_log_t) <insn_addresses>: New field.
25011 2017-07-09  H.J. Lu  <hongjiu.lu@intel.com>
25013         PR target/81313
25014         * config/i386/i386.c (ix86_function_arg_advance): Set
25015         outgoing_args_on_stack to true if there are outgoing arguments
25016         on stack.
25017         (ix86_function_arg): Likewise.
25018         (ix86_get_drap_rtx): Use DRAP only if there are outgoing
25019         arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
25020         * config/i386/i386.h (machine_function): Add
25021         outgoing_args_on_stack.
25023 2017-07-09  Krister Walfridsson  <krister.walfridsson@gmail.com>
25025         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
25026         supporting pthreds.
25027         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
25029 2017-07-08  Richard Sandiford  <richard.sandiford@linaro.org>
25031         * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
25032         (REAL_H): Remove $(MACHMODE_H).
25033         (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
25034         double-int.h.
25035         (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
25036         $(MACHMODE_H) and double-int.h.
25037         (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
25038         $(MACHMODE_H).
25039         (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
25040         double-int.h.
25042 2017-07-07  Andrew Pinski  <apinski@cavium.com>
25044         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
25045         prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
25047 2017-07-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
25049         * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
25050         Add warning if GCC was not configured to link against a GLIBC that
25051         exports the hardware capability bits.
25052         (make_resolver_func): Make resolver function private and not a
25053         COMDAT function.  Create the name with clone_function_name instead
25054         of make_unique_name.
25056         PR target/81348
25057         * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
25058         correct operand in doing the split.
25060 2017-07-07 Carl Love  <cel@us.ibm.com>
25062         * config/rs6000/rs6000-c: Add support for built-in function
25063         vector unsigned short vec_pack_to_short_fp32 (vector float,
25064                                                       vector float).
25065         * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
25066         BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
25067         * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
25068         * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
25069         (convert_4f32_8i16): Add define_expand.
25070         * doc/extend.texi: Update the built-in documentation file for the
25071         new built-in function.
25073 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25075         * config/sparc/m8.md: New file.
25076         * config/sparc/sparc.md: Include m8.md.
25078 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25080         * config/sparc/sparc.opt: New option -mvis4b.
25081         * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
25082         (sparc_option_override): Handle VIS4B.
25083         (enum sparc_builtins): Define
25084         SPARC_BUILTIN_DICTUNPACK{8,16,32},
25085         SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
25086         SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
25087         SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
25088         SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
25089         (check_constant_argument): New function.
25090         (sparc_vis_init_builtins): Define builtins
25091         __builtin_vis_dictunpack{8,16,32},
25092         __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
25093         __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
25094         __builtin_vis_fpcmpde{8,16,32}shl and
25095         __builtin_vis_fpcmpur{8,16,32}shl.
25096         (sparc_expand_builtin): Check that the constant operands to
25097         __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
25098         constant and in range.
25099         * config/sparc/sparc-c.c (sparc_target_macros): Handle
25100         TARGET_VIS4B.
25101         * config/sparc/sparc.h (SPARC_IMM2_P): Define.
25102         (SPARC_IMM5_P): Likewise.
25103         * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
25104         (enabled): Handle vis4b.
25105         (UNSPEC_DICTUNPACK): New unspec.
25106         (UNSPEC_FPCMPSHL): Likewise.
25107         (UNSPEC_FPUCMPSHL): Likewise.
25108         (UNSPEC_FPCMPDESHL): Likewise.
25109         (UNSPEC_FPCMPURSHL): Likewise.
25110         (cpu_feature): New CPU feature `vis4b'.
25111         (dictunpack{8,16,32}): New insns.
25112         (FPCSMODE): New mode iterator.
25113         (fpcscond): New code iterator.
25114         (fpcsucond): Likewise.
25115         (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
25116         (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
25117         (fpcmpde{8,16,32}{si,di}shl): Likewise.
25118         (fpcmpur{8,16,32}{si,di}shl): Likewise.
25119         * config/sparc/constraints.md: Define constraints `q' for unsigned
25120         2-bit integer constants and `t' for unsigned 5-bit integer
25121         constants.
25122         * config/sparc/predicates.md (imm5_operand_dictunpack8): New
25123         predicate.
25124         (imm5_operand_dictunpack16): Likewise.
25125         (imm5_operand_dictunpack32): Likewise.
25126         (imm2_operand): Likewise.
25127         * doc/invoke.texi (SPARC Options): Document -mvis4b.
25128         * doc/extend.texi (SPARC VIS Built-in Functions): Document the
25129         ditunpack* and fpcmp*shl builtins.
25131 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25133         * config.gcc: Handle m8 in --with-{cpu,tune} options.
25134         * config.in: Add HAVE_AS_SPARC6 define.
25135         * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
25136         M8.
25137         * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
25138         TARGET_CPU_m8.
25139         (ASM_CPU32_DEFAUILT_SPEC): Likewise.
25140         (CPP_CPU_SPEC): Handle m8.
25141         (ASM_CPU_SPEC): Likewise.
25142         * config/sparc/sparc-opts.h (enum processor_type): Add
25143         PROCESSOR_M8.
25144         * config/sparc/sparc.c (m8_costs): New struct.
25145         (sparc_option_override): Handle TARGET_CPU_m8.
25146         (sparc32_initialize_trampoline): Likewise.
25147         (sparc64_initialize_trampoline): Likewise.
25148         (sparc_issue_rate): Likewise.
25149         (sparc_register_move_cost): Likewise.
25150         * config/sparc/sparc.h (TARGET_CPU_m8): Define.
25151         (CPP_CPU64_DEFAULT_SPEC): Define for M8.
25152         (ASM_CPU64_DEFAULT_SPEC): Likewise.
25153         (CPP_CPU_SPEC): Handle M8.
25154         (ASM_CPU_SPEC): Likewise.
25155         (AS_M8_FLAG): Define.
25156         * config/sparc/sparc.md: Add m8 to the cpu attribute.
25157         * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
25158         * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
25159         M8 instructions.
25160         * configure: Regenerate.
25161         * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
25162         -mtune=m8.
25164 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25166         * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
25167         subtypes.
25168         * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
25169         ("*movdi_insn_sp32"): Do not set v3pipe.
25170         ("*movsi_insn"): Likewise.
25171         ("*movdi_insn_sp64"): Likewise.
25172         ("*movsf_insn"): Likewise.
25173         ("*movdf_insn_sp32"): Likewise.
25174         ("*movdf_insn_sp64"): Likewise.
25175         ("*zero_extendsidi2_insn_sp64"): Likewise.
25176         ("*sign_extendsidi2_insn"): Likewise.
25177         ("*mov<VM32:mode>_insn"): Likewise.
25178         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25179         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25180         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25181         ("<vlop:code><VL:mode>3"): Likewise.
25182         ("*not_<vlop:code><VL:mode>3"): Likewise.
25183         ("*nand<VL:mode>_vis"): Likewise.
25184         ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
25185         ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
25186         ("one_cmpl<VL:mode>2"): Likewise.
25187         ("faligndata<VM64:mode>_vis"): Likewise.
25188         ("alignaddrsi_vis"): Likewise.
25189         ("alignaddrdi_vis"): Likweise.
25190         ("alignaddrlsi_vis"): Likewise.
25191         ("alignaddrldi_vis"): Likewise.
25192         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25193         ("bmaskdi_vis"): Likewise.
25194         ("bmasksi_vis"): Likewise.
25195         ("bshuffle<VM64:mode>_vis"): Likewise.
25196         ("cmask8<P:mode>_vis"): Likewise.
25197         ("cmask16<P:mode>_vis"): Likewise.
25198         ("cmask32<P:mode>_vis"): Likewise.
25199         ("pdistn<P:mode>_vis"): Likewise.
25200         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25202 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25204         * config/sparc/sparc.md ("subtype"): New insn attribute.
25205         ("*wrgsr_sp64"): Set insn subtype.
25206         ("*rdgsr_sp64"): Likewise.
25207         ("alignaddrsi_vis"): Likewise.
25208         ("alignaddrdi_vis"): Likewise.
25209         ("alignaddrlsi_vis"): Likewise.
25210         ("alignaddrldi_vis"): Likewise.
25211         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25212         ("fexpand_vis"): Likewise.
25213         ("fpmerge_vis"): Likewise.
25214         ("faligndata<VM64:mode>_vis"): Likewise.
25215         ("bshuffle<VM64:mode>_vis"): Likewise.
25216         ("cmask8<P:mode>_vis"): Likewise.
25217         ("cmask16<P:mode>_vis"): Likewise.
25218         ("cmask32<P:mode>_vis"): Likewise.
25219         ("fchksm16_vis"): Likewise.
25220         ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
25221         ("fmean16_vis"): Likewise.
25222         ("fp<plusminus_insn>64_vis"): Likewise.
25223         ("<plusminus_insn>v8qi3"): Likewise.
25224         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25225         ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
25226         ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
25227         ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
25228         ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
25229         ("*movqi_insn"): Likewise.
25230         ("*movhi_insn"): Likewise.
25231         ("*movsi_insn"): Likewise.
25232         ("movsi_pic_gotdata_op"): Likewise.
25233         ("*movdi_insn_sp32"): Likewise.
25234         ("*movdi_insn_sp64"): Likewise.
25235         ("movdi_pic_gotdata_op"): Likewise.
25236         ("*movsf_insn"): Likewise.
25237         ("*movdf_insn_sp32"): Likewise.
25238         ("*movdf_insn_sp64"): Likewise.
25239         ("*zero_extendhisi2_insn"): Likewise.
25240         ("*zero_extendqihi2_insn"): Likewise.
25241         ("*zero_extendqisi2_insn"): Likewise.
25242         ("*zero_extendqidi2_insn"): Likewise.
25243         ("*zero_extendhidi2_insn"): Likewise.
25244         ("*zero_extendsidi2_insn_sp64"): Likewise.
25245         ("ldfsr"): Likewise.
25246         ("prefetch_64"): Likewise.
25247         ("prefetch_32"): Likewise.
25248         ("tie_ld32"): Likewise.
25249         ("tie_ld64"): Likewise.
25250         ("*tldo_ldub_sp32"): Likewise.
25251         ("*tldo_ldub1_sp32"): Likewise.
25252         ("*tldo_ldub2_sp32"): Likewise.
25253         ("*tldo_ldub_sp64"): Likewise.
25254         ("*tldo_ldub1_sp64"): Likewise.
25255         ("*tldo_ldub2_sp64"): Likewise.
25256         ("*tldo_ldub3_sp64"): Likewise.
25257         ("*tldo_lduh_sp32"): Likewise.
25258         ("*tldo_lduh1_sp32"): Likewise.
25259         ("*tldo_lduh_sp64"): Likewise.
25260         ("*tldo_lduh1_sp64"): Likewise.
25261         ("*tldo_lduh2_sp64"): Likewise.
25262         ("*tldo_lduw_sp32"): Likewise.
25263         ("*tldo_lduw_sp64"): Likewise.
25264         ("*tldo_lduw1_sp64"): Likewise.
25265         ("*tldo_ldx_sp64"): Likewise.
25266         ("*mov<VM32:mode>_insn"): Likewise.
25267         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25268         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25270 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25272         * config/sparc/sparc.md ("type"): New insn type viscmp.
25273         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25274         viscmp.
25275         ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25276         ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25277         ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25278         * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25279         viscmp.
25280         ("n7_vis_logical_11cycle"): Likewise.
25281         * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25282         * config/sparc/niagara2.md ("niag3_vis": Likewise.
25283         * config/sparc/niagara.md ("niag_vis"): Likewise.
25284         * config/sparc/ultra3.md ("us3_fga"): Likewise.
25285         * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25287 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25289         * config/sparc/sparc.md: New instruction type `bmask'.
25290         (bmaskdi_vis): Use the `bmask' type.
25291         (bmasksi_vis): Likewise.
25292         * config/sparc/ultra3.md (us3_array): Likewise.
25293         * config/sparc/niagara7.md (n7_array): Likewise.
25294         * config/sparc/niagara4.md (n4_array): Likewise.
25295         * config/sparc/niagara2.md (niag2_vis): Likewise.
25296         (niag3_vis): Likewise.
25297         * config/sparc/niagara.md (niag_vis): Likewise.
25299 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25301         * ipa-comdats.c: Remove optimize check from gate.
25302         * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25303         for functions not optimized.
25304         (ipa_fn_summary_read): Skip optimize check.
25305         (ipa_fn_summary_write): Likewise.
25306         * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25307         is optimized.
25308         * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25309         uninlinable.
25310         (can_inline_edge_p): Check flag_pcc_struct_return for match.
25311         (check_callers): Give up on caller which is not optimized.
25312         (inline_small_functions): Likewise.
25313         (ipa_inline): Do not give up when not optimizing.
25314         * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25315         away unoptimizes cdtors.
25316         (whole_program_function_and_variable_visibility): Do
25317         ipa_discover_readonly_nonaddressable_vars in LTO mode.
25318         * ipa.c (process_references): Do not check optimize.
25319         (symbol_table::remove_unreachable_nodes): Update optimize check.
25320         (set_writeonly_bit): Update optimize check.
25321         (pass_ipa_cdtor_merge::gate): Do not check optimize.
25322         (pass_ipa_single_use::gate): Remove.
25324 2017-07-06  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25326         * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25327         insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25328         rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25329         mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25330         permute_load, permute_store, adjust_extract, adjust_splat,
25331         adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25332         replace_swap_with_copy, dump_swap_insn_table,
25333         alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25334         recombine_lvx_pattern, recombine_stvx_pattern,
25335         recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25336         make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25337         to file rs6000-p8swap.c.
25338         * config/rs6000/rs6000-p8swap.c: New file.
25339         * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25340         * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25341         and rs6000*-*-* targets.
25343 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25345         * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25347 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25349         * lto-wrapper.c (merge_and_complain): Do not merge
25350         fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25351         fsigned_zeros, ftrapping_math, fwrapv.
25352         (append_compiler_options): Do not track these options.
25353         (append_linker_options): Likewie
25355 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25357         * cgraphunit.c (cgraph_node::finalize_function): When
25358         !flag_toplevel_reorde set no_reorder flag.
25359         (varpool_node::finalize_decl): Likewise.
25360         (symbol_table::compile): Drop no toplevel reorder path.
25362 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25364         * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25365         edges; zero probability is not better than uninitialized.
25367 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25369         * asan.h (asan_sanitize_allocas_p): Declare.
25370         * asan.c (asan_sanitize_allocas_p): New function.
25371         (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25372         (handle_builtin_alloca): Likewise.
25373         * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25374         if !asan_sanitize_allocas_p.
25375         * params.def (asan-instrument-allocas): Add new option.
25376         * params.h (ASAN_PROTECT_ALLOCAS): Define.
25377         * opts.c (common_handle_option): Disable allocas sanitization for
25378         KASan by default.
25380 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25382         * asan.c: Include gimple-fold.h.
25383         (get_last_alloca_addr): New function.
25384         (handle_builtin_stackrestore): Likewise.
25385         (handle_builtin_alloca): Likewise.
25386         (asan_emit_allocas_unpoison): Likewise.
25387         (get_mem_refs_of_builtin_call): Add new parameter, remove const
25388         quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25389         BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25390         (instrument_builtin_call): Pass gimple iterator to
25391         get_mem_refs_of_builtin_call.
25392         (last_alloca_addr): New global.
25393         * asan.h (asan_emit_allocas_unpoison): Declare.
25394         * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25395         (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25396         * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25397         if function calls alloca.
25398         * gimple-fold.c (replace_call_with_value): Remove static keyword.
25399         * gimple-fold.h (replace_call_with_value): Declare.
25400         * internal-fn.c: Include asan.h.
25401         * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25402         BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25404 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25406         * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25407         (C_SELFTEST_FLAGS): New.
25408         (CPP_SELFTEST_FLAGS): New.
25409         (SELFTEST_DEPS): New, from deps of s-selftest.
25410         (C_SELFTEST_DEPS): New, from deps of s-selftest.
25411         (CPP_SELFTEST_DEPS): New.
25412         (selftest): Add dependency on s-selftest-c++.
25413         (s-selftest): Rename to...
25414         (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25415         and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25416         than SELFTEST_FLAGS.
25417         (selftest-gdb): Rename to...
25418         (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25419         C_SELFTEST_FLAGS.
25420         (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25421         (selftest-valgrind): Rename to...
25422         (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25423         C_SELFTEST_FLAGS.
25424         (selftest-valgrind): Reintroduce as an alias for
25425         selftest-c-valgrind.
25426         (s-selftest-c++): New.
25427         (selftest-c++-gdb): New.
25428         (selftest-c++-valgrind): New.
25430 2017-07-06  Olivier Hainque  <hainque@adacore.com>
25432         * gcc.c (process_command): When deciding if undefined variables
25433         should be ignored when processing specs, accept "gcc -v" as well.
25435 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25437         * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25438         afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25440 2017-07-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25442         * config/arm/arm-cpus.in (armv8-r): Add new entry.
25443         * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25444         * config/arm/arm-tables.opt: Regenerate.
25445         * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25446         enumerator.
25447         * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25449 2017-07-06  Carl Love  <cel@us.ibm.com>
25451         * ChangeLog: Clean up from mid air collision
25453 2017-07-06  Carl Love  <cel@us.ibm.com>
25455         * config/rs6000/rs6000-c.c: Add support for built-in functions
25456         vector signed int vec_subc (vector signed int, vector signed int);
25457         vector signed __int128 vec_subc (vector signed __int128,
25458                                          vector signed __int128);
25459         vector unsigned __int128 vec_subc (vector unsigned __int128,
25460                                            vector unsigned __int128);
25461         vector signed int vec_sube (vector signed int, vector signed int,
25462                                     vector signed int);
25463         vector unsigned int vec_sube (vector unsigned int,
25464                                       vector unsigned int,
25465                                       vector unsigned int);
25466         vector signed __int128 vec_sube (vector signed __int128,
25467                                          vector signed __int128,
25468                                          vector signed__int128);
25469         vector unsigned __int128 vec_sube (vector unsigned __int128,
25470                                            vector unsigned __int128,
25471                                            vector unsigned __int128);
25472         vector signed int vec_subec (vector signed int, vector signed int,
25473                                      vector signed int);
25474         vector unsigned int vec_subec (vector unsigned int,
25475                                        vector unsigned int,
25476                                        vector unsigned int);
25477         vector signed __int128 vec_subec (vector signed __int128,
25478                                           vector signed __int128,
25479                                           vector signed__int128);
25480         vector unsigned __int128 vec_subec (vector unsigned __int128,
25481                                             vector unsigned __int128,
25482                                             vector unsigned __int128);
25483         * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25484         ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25485         * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25486         BU_ALTIVEC_OVERLOAD_X definitions.
25487         * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25488         * doc/extend.texi: Update the built-in documentation file for the new
25489         built-in functions.
25491 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25493         PR c++/79300
25494         * diagnostic-show-locus.c (layout::layout): Use start and finish
25495         spelling location for the start and finish of each range.
25496         * genmatch.c (linemap_client_expand_location_to_spelling_point):
25497         Add unused aspect param.
25498         * input.c (expand_location_1): Add "aspect" param, and use it
25499         to access the correct part of the location.
25500         (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25501         expand_location_1.
25502         (expand_location_to_spelling_point): Likewise.
25503         (linemap_client_expand_location_to_spelling_point): Add "aspect"
25504         param, and pass it to expand_location_1.
25506 2017-07-06  Sebastian Peryt  <sebastian.peryt@intel.com>
25508         * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25509         _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25510         _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25511         _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25512         _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25513         _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25514         _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25515         _mm_maskz_getmant_ss): New intrinsics.
25516         (__builtin_ia32_getexpss128_mask): Changed to ...
25517         __builtin_ia32_getexpss128_round ... this.
25518         (__builtin_ia32_getexpsd128_mask): Changed to ...
25519         __builtin_ia32_getexpsd128_round ... this.
25520         * config/i386/i386-builtin-types.def
25521         ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25522         (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25523         * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25524         __builtin_ia32_getexpss_mask_round,     __builtin_ia32_getmantsd_mask_round,
25525         __builtin_ia32_getmantss_mask_round): New builtins.
25526         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25527         V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25528         (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25529         CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25530         * config/i386/sse.md
25531         (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25532         avx512f_sgetexp<mode><mask_scalar_name>
25533         <round_saeonly_scalar_name> ... this.
25534         (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25535         %0, %1, %2<round_saeonly_op3>}): Changed to ...
25536         vgetexp<ssescalarmodesuffix>
25537         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25538         %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25539         (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25540         avx512f_vgetmant<mode><mask_scalar_name>
25541         <round_saeonly_scalar_name> ... this.
25542         (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25543         %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25544         vgetmant<ssescalarmodesuffix>
25545         \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25546         %0<mask_scalar_operand4>, %1, %2
25547         <round_saeonly_scalar_mask_op4>, %3} ... this.
25548         * config/i386/subst.md (mask_scalar_operand4,
25549         round_saeonly_scalar_mask_operand4,     round_saeonly_scalar_mask_op4,
25550         round_saeonly_scalar_nimm_predicate): New subst attributes.
25552 2017-07-06  Julia Koval  <julia.koval@intel.com>
25554         * config/i386/i386.c (ix86_erase_embedded_rounding):
25555         Remove code for old rounding pattern.
25557 2017-07-06  Richard Earnshaw  <rearnsha@arm.com>
25559         * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25561 2017-07-06  Christophe Lyon  <christophe.lyon@linaro.org>
25563         * doc/sourcebuild.texi (Test Directives, Variants of
25564         dg-require-support): Add documentation for dg-require-stack-check.
25566 2017-07-05  Sebastian Peryt  <sebastian.peryt@intel.com>
25568         * config/i386/subst.md (mask_scalar, round_scalar,
25569         round_saeonly_scalar): New meta-templates.
25570         (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25571         round_scalar_mask_operand3, round_scalar_mask_op3,
25572         round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25573         round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25574         round_saeonly_scalar_constraint,
25575         round_saeonly_scalar_prefix): New subst attribute.
25576         * config/i386/sse.md
25577         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25578         <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25579         <round_scalar_name> ... this.
25580         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25581         <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25582         <round_scalar_name> ... this.
25583         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25584         <sse>_vm<code><mode>3<mask_scalar_name>
25585         <round_saeonly_scalar_name> ... this.
25586         (v<plusminus_mnemonic><ssescalarmodesuffix>
25587         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25588         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25589         v<plusminus_mnemonic><ssescalarmodesuffix>
25590         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25591         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25592         (v<multdiv_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<multdiv_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<maxmin_float><ssescalarmodesuffix>
25599         \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25600         %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25601         v<maxmin_float><ssescalarmodesuffix>
25602         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25603         %0<mask_scalar_operand3>, %1, %<iptr>2
25604         <round_saeonly_scalar_mask_op3>} ... this.
25606 2017-07-05  Richard Earnshaw  <rearnsha@arm.com>
25608         * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25609         (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25611 2017-07-05  Richard Sandiford  <richard.sandiford@linaro.org>
25612             Alan Hayward  <alan.hayward@arm.com>
25613             David Sherwood  <david.sherwood@arm.com>
25615         * combine.c (simplify_if_then_else): Remove "enum" before
25616         "machine_mode".
25617         * compare-elim.c (can_eliminate_compare): Likewise.
25618         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25619         Likewise.
25620         (aarch64_lookup_simd_builtin_type): Likewise.
25621         (aarch64_simd_builtin_type): Likewise.
25622         (aarch64_init_simd_builtin_types): Likewise.
25623         (aarch64_simd_expand_args): Likewise.
25624         * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25625         Likewise.
25626         (aarch64_reverse_mask): Likewise.
25627         (aarch64_simd_emit_reg_reg_move): Likewise.
25628         (aarch64_gen_adjusted_ldpstp): Likewise.
25629         (aarch64_ccmp_mode_to_code): Likewise.
25630         (aarch64_operands_ok_for_ldpstp): Likewise.
25631         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25632         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25633         Likewise.
25634         (aarch64_min_divisions_for_recip_mul): Likewise.
25635         (aarch64_reassociation_width): Likewise.
25636         (aarch64_get_condition_code_1): Likewise.
25637         (aarch64_simd_emit_reg_reg_move): Likewise.
25638         (aarch64_simd_attr_length_rglist): Likewise.
25639         (aarch64_reverse_mask): Likewise.
25640         (aarch64_operands_ok_for_ldpstp): Likewise.
25641         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25642         (aarch64_gen_adjusted_ldpstp): Likewise.
25643         * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25644         Likewise.
25645         * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25646         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25647         (arm_lookup_simd_builtin_type): Likewise.
25648         (arm_simd_builtin_type): Likewise.
25649         (arm_init_simd_builtin_types): Likewise.
25650         (arm_expand_builtin_args): Likewise.
25651         * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25652         * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25653         (ft32_setup_incoming_varargs): Likewise.
25654         (ft32_function_arg): Likewise.
25655         (ft32_function_arg_advance): Likewise.
25656         (ft32_pass_by_reference): Likewise.
25657         (ft32_arg_partial_bytes): Likewise.
25658         (ft32_valid_pointer_mode): Likewise.
25659         (ft32_addr_space_pointer_mode): Likewise.
25660         (ft32_addr_space_legitimate_address_p): Likewise.
25661         * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25662         Likewise.
25663         * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25664         (ix86_emit_outlined_ms2sysv_restore): Likewise.
25665         (iamcu_alignment): Likewise.
25666         (canonicalize_vector_int_perm): Likewise.
25667         (ix86_noce_conversion_profitable_p): Likewise.
25668         (ix86_mpx_bound_mode): Likewise.
25669         (ix86_operands_ok_for_move_multiple): Likewise.
25670         * config/microblaze/microblaze-protos.h
25671         (microblaze_expand_conditional_branch_reg): Likewise.
25672         * config/microblaze/microblaze.c
25673         (microblaze_expand_conditional_branch_reg): Likewise.
25674         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25675         Likewise.
25676         (rs6000_reassociation_width): Likewise.
25677         (rs6000_invalid_binary_op): Likewise.
25678         (fusion_p9_p): Likewise.
25679         (emit_fusion_p9_load): Likewise.
25680         (emit_fusion_p9_store): Likewise.
25681         * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25682         Likewise.
25683         (riscv_hard_regno_mode_ok_p): Likewise.
25684         (riscv_address_insns): Likewise.
25685         (riscv_split_symbol): Likewise.
25686         (riscv_legitimize_move): Likewise.
25687         (riscv_function_value): Likewise.
25688         (riscv_hard_regno_nregs): Likewise.
25689         (riscv_expand_builtin): Likewise.
25690         * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25691         (riscv_build_integer): Likewise.
25692         (riscv_split_integer): Likewise.
25693         (riscv_legitimate_constant_p): Likewise.
25694         (riscv_cannot_force_const_mem): Likewise.
25695         (riscv_regno_mode_ok_for_base_p): Likewise.
25696         (riscv_valid_base_register_p): Likewise.
25697         (riscv_valid_offset_p): Likewise.
25698         (riscv_valid_lo_sum_p): Likewise.
25699         (riscv_classify_address): Likewise.
25700         (riscv_legitimate_address_p): Likewise.
25701         (riscv_address_insns): Likewise.
25702         (riscv_load_store_insns): Likewise.
25703         (riscv_force_binary): Likewise.
25704         (riscv_split_symbol): Likewise.
25705         (riscv_force_address): Likewise.
25706         (riscv_legitimize_address): Likewise.
25707         (riscv_move_integer): Likewise.
25708         (riscv_legitimize_const_move): Likewise.
25709         (riscv_legitimize_move): Likewise.
25710         (riscv_address_cost): Likewise.
25711         (riscv_subword): Likewise.
25712         (riscv_output_move): Likewise.
25713         (riscv_canonicalize_int_order_test): Likewise.
25714         (riscv_emit_int_order_test): Likewise.
25715         (riscv_function_arg_boundary): Likewise.
25716         (riscv_pass_mode_in_fpr_p): Likewise.
25717         (riscv_pass_fpr_single): Likewise.
25718         (riscv_pass_fpr_pair): Likewise.
25719         (riscv_get_arg_info): Likewise.
25720         (riscv_function_arg): Likewise.
25721         (riscv_function_arg_advance): Likewise.
25722         (riscv_arg_partial_bytes): Likewise.
25723         (riscv_function_value): Likewise.
25724         (riscv_pass_by_reference): Likewise.
25725         (riscv_setup_incoming_varargs): Likewise.
25726         (riscv_print_operand): Likewise.
25727         (riscv_elf_select_rtx_section): Likewise.
25728         (riscv_save_restore_reg): Likewise.
25729         (riscv_for_each_saved_reg): Likewise.
25730         (riscv_register_move_cost): Likewise.
25731         (riscv_hard_regno_mode_ok_p): Likewise.
25732         (riscv_hard_regno_nregs): Likewise.
25733         (riscv_class_max_nregs): Likewise.
25734         (riscv_memory_move_cost): Likewise.
25735         * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25736         * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25737         (rl78_addr_space_address_mode): Likewise.
25738         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25739         Likewise.
25740         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25741         (rs6000_reassociation_width): Likewise.
25742         (rs6000_invalid_binary_op): Likewise.
25743         (fusion_p9_p): Likewise.
25744         (emit_fusion_p9_load): Likewise.
25745         (emit_fusion_p9_store): Likewise.
25746         * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25747         (ok_for_simple_move_operands): Likewise.
25748         (ok_for_simple_move_strict_operands): Likewise.
25749         (ok_for_simple_arith_logic_operands): Likewise.
25750         (visium_legitimize_reload_address): Likewise.
25751         (visium_select_cc_mode): Likewise.
25752         (output_cbranch): Likewise.
25753         (visium_split_double_move): Likewise.
25754         (visium_expand_copysign): Likewise.
25755         (visium_expand_int_cstore): Likewise.
25756         (visium_expand_fp_cstore): Likewise.
25757         * config/visium/visium.c (visium_pass_by_reference): Likewise.
25758         (visium_function_arg): Likewise.
25759         (visium_function_arg_advance): Likewise.
25760         (visium_libcall_value): Likewise.
25761         (visium_setup_incoming_varargs): Likewise.
25762         (visium_legitimate_constant_p): Likewise.
25763         (visium_legitimate_address_p): Likewise.
25764         (visium_legitimize_address): Likewise.
25765         (visium_secondary_reload): Likewise.
25766         (visium_register_move_cost): Likewise.
25767         (visium_memory_move_cost): Likewise.
25768         (prepare_move_operands): Likewise.
25769         (ok_for_simple_move_operands): Likewise.
25770         (ok_for_simple_move_strict_operands): Likewise.
25771         (ok_for_simple_arith_logic_operands): Likewise.
25772         (visium_function_value_1): Likewise.
25773         (rtx_ok_for_offset_p): Likewise.
25774         (visium_legitimize_reload_address): Likewise.
25775         (visium_split_double_move): Likewise.
25776         (visium_expand_copysign): Likewise.
25777         (visium_expand_int_cstore): Likewise.
25778         (visium_expand_fp_cstore): Likewise.
25779         (visium_split_cstore): Likewise.
25780         (visium_select_cc_mode): Likewise.
25781         (visium_split_cbranch): Likewise.
25782         (output_cbranch): Likewise.
25783         (visium_print_operand_address): Likewise.
25784         * expmed.c (flip_storage_order): Likewise.
25785         * expmed.h (emit_cstore): Likewise.
25786         (flip_storage_order): Likewise.
25787         * genrecog.c (validate_pattern): Likewise.
25788         * hsa-gen.c (gen_hsa_addr): Likewise.
25789         * internal-fn.c (expand_arith_overflow): Likewise.
25790         * ira-color.c (allocno_copy_cost_saving): Likewise.
25791         * lra-assigns.c (find_hard_regno_for_1): Likewise.
25792         * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25793         (process_invariant_for_inheritance): Likewise.
25794         * lra-eliminations.c (move_plus_up): Likewise.
25795         * omp-low.c (lower_oacc_reductions): Likewise.
25796         * simplify-rtx.c (simplify_subreg): Likewise.
25797         * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25798         (TARGET_CHKP_BOUND_MODE): Likewise..
25799         * targhooks.c (default_chkp_bound_mode): Likewise.
25800         (default_setup_incoming_vararg_bounds): Likewise.
25801         * targhooks.h (default_chkp_bound_mode): Likewise.
25802         (default_setup_incoming_vararg_bounds): Likewise.
25803         * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25804         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25805         (have_whole_vector_shift): Likewise.
25806         * tree-vect-stmts.c (vectorizable_load): Likewise.
25807         * doc/tm.texi: Regenerate.
25809 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25811         Graceful degrade if Binutils PR21472 is not available.
25813         PR target/81072
25814         * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25815         .rodata in flash test fails.
25816         (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25817         * confgure: Regenerate.
25818         * config.in: Regenerate.
25819         * config/avr/avr.c (avr_asm_named_section)
25820         [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25821         __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25822         (avr_asm_init_sections): Same.
25824 2017-07-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25826         * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25827         (fma<VH:mode>4_intrinsic): Likewise.
25828         (*fmsub<VCVTF:mode>4): Likewise.
25829         (*fmsub<VH:mode>4_intrinsic): Likewise.
25831 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25833         PR target/81305
25834         * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25835         Don't depend on "optimize > 0".
25836         (out_movhi_r_mr, out_movqi_mr_r): Same.
25837         (out_movhi_mr_r, out_movqi_r_mr): Same.
25838         (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25839         io_address_operand on "optimize > 0".
25841 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25843         * tree-loop-distribution.c: Add general explanantion on the pass.
25844         (generate_loops_for_partition): Mark distributed loop.
25845         (pg_add_dependence_edges): New parameter.  Handle alias data
25846         dependence specially and record it in the parameter if asked.
25847         (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25848         (init_partition_graph_vertices, add_partition_graph_edge): New.
25849         (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25850         (free_partition_graph_vdata, build_partition_graph): New.
25851         (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25852         (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25853         (data_ref_segment_size, latch_dominated_by_data_ref): New.
25854         (compute_alias_check_pairs, version_loop_by_alias_check): New.
25855         (version_for_distribution_p, finalize_partitions): New.
25856         (distribute_loop): Handle alias data dependence specially.  Factor
25857         out loop fusion code as functions and call these functions.
25859 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25861         * tree-loop-distribution.c (classify_partition): New parameter and
25862         better handle reduction statement.
25863         (rdg_build_partitions): Revise comment.
25864         (distribute_loop): Compute statements in all partitions and pass it
25865         to classify_partition.
25867 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25869         * tree-loop-distribution.c (enum partition_type): New.
25870         (struct partition): New field type.
25871         (partition_merge_into): Add parameter.  Update partition type.
25872         (data_dep_in_cycle_p, update_type_for_merge): New functions.
25873         (build_rdg_partition_for_vertex): Compute partition type.
25874         (rdg_build_partitions): Dump partition type.
25875         (distribute_loop): Update calls to partition_merge_into.
25877 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25879         * tree-loop-distribution.c (struct ddr_hasher): New.
25880         (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25881         (ddrs_table): New.
25882         (classify_partition): Call get_data_dependence.
25883         (pg_add_dependence_edges): Ditto.
25884         (distribute_loop): Release data dependence hash table.
25886 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25888         * tree-loop-distribution.c (ref_base_address): Delete.
25889         (similar_memory_accesses): Rename ...
25890         (share_memory_accesses): ... to this.  Check if partitions access
25891         the same memory reference.
25892         (distribute_loop): Call share_memory_accesses.
25894 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25896         * tree-loop-distribution.c (struct partition): New field recording
25897         its data reference.
25898         (partition_alloc, partition_free): Init and release data refs.
25899         (partition_merge_into): Merge data refs.
25900         (build_rdg_partition_for_vertex): Collect data refs for partition.
25901         (pg_add_dependence_edges): Change parameters from vector to bitmap.
25902         Update uses.
25903         (distribute_loop): Remve data refs from vertice data of partition
25904         graph.
25906 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25908         * tree-loop-distribution.c (params.h): Include header file.
25909         (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25910         (datarefs_vec): New global var.
25911         (create_rdg_vertices): Use datarefs_vec directly.
25912         (free_rdg): Don't free data references.
25913         (build_rdg): Update use.  Don't free data references.
25914         (distribute_loop): Compute global variable for data references.
25915         Bail out if there are too many data references.
25917 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25919         * tree-loop-distribution.c (loop_nest): New global var.
25920         (build_rdg): Use loop directly, rather than loop nest.
25921         (pg_add_dependence_edges): Remove loop nest parameter.  Use global
25922         variable directly.
25923         (distribute_loop): Compute global variable loop nest.  Update use.
25925 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25927         * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
25928         (partition_merge_into): New parameter.  Dump reason for fusion.
25929         (distribute_loop): Update use of partition_merge_into.
25931 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25933         * tree-loop-distribution.c (bb_top_order_index): New.
25934         (bb_top_order_index_size, bb_top_order_cmp): New.
25935         (stmts_from_loop): Use topological order.
25936         (pass_loop_distribution::execute): Compute and release topological
25937         order for basic blocks.
25939 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25941         * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
25942         if no loops.
25944 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25946         * cfgloop.h (struct loop): Add comment.  New field orig_loop_num.
25947         * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
25948         * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
25949         * internal-fn.def (LOOP_DIST_ALIAS): New.
25950         * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
25951         (fold_loop_internal_call): ... this.
25952         (vect_loop_dist_alias_call): New function.
25953         (set_uid_loop_bbs): Call fold_loop_internal_call.
25954         (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
25955         internal calls.
25957 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25959         PR target/81300
25960         * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
25961         Require dead FLAGS_REG at the beginning of a peephole.
25963 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25965         PR target/81294
25966         * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
25967         arguments in the call to __builtin_ia32_sbb_u32.
25968         (_subborrow_u64): Swap _X and _Y arguments in the call to
25969         __builtin_ia32_sbb_u64.
25971 2017-07-04  Jakub Jelinek  <jakub@redhat.com>
25973         PR debug/81278
25974         * tree-vrp.c (compare_assert_loc): Turn into a function template
25975         with stable template parameter.  Only test if a->e is NULL,
25976         !a->e == !b->e has been verified already.  Use e == NULL or
25977         e != NULL instead of e or ! e tests.  If stable is true, don't use
25978         iterative_hash_expr, on the other side allow a or b or both NULL
25979         and sort the NULLs last.
25980         (process_assert_insertions): Sort using compare_assert_loc<false>
25981         instead of compare_assert_loc, later sort using
25982         compare_assert_loc<true> before calling process_assert_insertions_for
25983         in a loop.  Use break instead of continue once seen NULL pointer.
25985 2017-07-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25987         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
25988         Cortex-R7 and Cortex-R8 processors.
25990 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25992         * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
25993         uninitialized while src is not.
25995 2017-07-04  Richard Earnshaw  <rearnsha@arm.com>
25997         * common/config/arm/arm-common.c: Adjust include path for
25998         arm-cpu-cdata.h
25999         * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
26000         (arm-cpu.h): Create in build directory.  Adjust dependency rules.
26001         (arm-cpu-data.h): Likewise.
26002         (arm-cpu-cdata.h): Likewise.
26003         * config/arm/arm-cpu.h: Delete.
26004         * config/arm/arm-cpu-cdata.h: Delete.
26005         * config/arm/arm-cpu-data.h: Delete.
26007 2017-07-04  James Greenhalgh  <james.greenhalgh@arm.com>
26009         * config/arm/arm-cpus.in (cortex-a55): New.
26010         (cortex-a75): Likewise.
26011         (cortex-a75.cortex-a55): Likewise.
26012         * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
26013         cortex-a75.
26014         * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
26015         * config/arm/arm-cpu-cdata.h: Regenerate.
26016         * config/arm/arm-cpu-data.h: Regenerate.
26017         * config/arm/arm-cpu.h: Regenerate.
26018         * config/arm/arm-tables.opt: Regenerate.
26019         * config/arm/arm-tune.md: Regenerate.
26021 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26023         * haifa-sched.c (sched_create_recovery_edges): Update profile.
26025 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26027         * bb-reorder.c (better_edge_p): Fix handling of uninitialized
26028         probability.
26030 2017-07-04  Richard Sandiford  <richard.sandiford@linaro.org>
26032         PR tree-optimization/81292
26033         * tree-ssa-strlen.c (handle_builtin_strlen): When setting
26034         full_string_p, also call adjust_related_strinfos if the adjustment
26035         is simple, otherwise invalidate related strinfos.
26037 2017-07-04  Martin Liska  <mliska@suse.cz>
26039         PR sanitizer/81040
26040         * sanopt.c (sanitize_rewrite_addressable_params): Mark the
26041         newly created variable as DECL_IGNORED_P.
26043 2017-07-04  Martin Liska  <mliska@suse.cz>
26045         PR ipa/81293
26046         * ipa-inline.c (inline_small_functions):
26047         Use xstrdup_for_dump.
26049 2017-07-04  Tom de Vries  <tom@codesourcery.com>
26051         * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
26053 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
26055         PR target/81033
26056         * config/darwin.c (darwin_function_switched_text_sections):
26057         Fix spaces.
26059 2017-07-03  Jan Hubicka  <hubicka@ucw.cz>
26061         * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
26063 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26065         * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
26067 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26069         * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
26070         min_profitable_iters, and th as inclusive lower bounds.
26071         Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
26072         (vect_estimate_min_profitable_iters): Return inclusive lower bounds
26073         for min_profitable_iters and min_profitable_estimate.
26074         (vect_transform_loop): Treat th as an inclusive lower bound.
26075         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
26077 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
26079         PR target/81033
26080         * config/darwin.c (darwin_function_switched_text_sections):
26081         Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
26082         in two pieces, and suppress the use of buf.
26084 2017-07-03  Nathan Sidwell  <nathan@acm.org>
26086         * hash-table.h (hash_table_mod1): Fix indentation.
26088 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26090         PR middle-end/81290
26091         * predict.c (force_edge_cold): Be more careful about propagation
26092         backward.
26093         * profile-count.h (profile_probability::guessed,
26094         profile_probability::fdo, profile_count::guessed, profile_count::fdo):
26095         New.
26096         * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
26098 2017-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
26100         * doc/invoke.texi (rcpc architecture extension): Document it.
26102 2017-07-03  Richard Biener  <rguenther@suse.de>
26104         PR tree-optimization/60510
26105         * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
26106         the scalar reduction PHI and use it.
26107         (vectorizable_reduction): Properly guard the single_defuse_cycle
26108         path for non-SLP reduction chains where we cannot use it.
26109         Rework reduc_def/index and vector type deduction.  Rework
26110         vector operand gathering during reduction op code-gen.
26111         * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
26112         chains dissolve the chain and leave it to non-SLP reduction
26113         handling.
26115 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26117         * tree-data-ref.h (dr_alignment): Declare.
26118         * tree-data-ref.c (dr_alignment): New function.
26119         * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
26120         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26121         set it.
26122         * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
26124 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26126         * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
26127         and base_misalignment fields.
26128         (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
26129         * tree-data-ref.c: Include builtins.h.
26130         (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
26131         * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
26132         (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
26133         * tree-vect-data-refs.c: Include tree-cfg.h.
26134         (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
26135         fields instead of calculating an alignment here.
26136         (vect_analyze_data_refs): Use dr_analyze_innermost.  Dump the new
26137         innermost_loop_behavior fields.
26139 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26141         * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
26142         field.
26143         (DR_STEP_ALIGNMENT): New macro.
26144         * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
26145         * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
26146         (create_data_ref): Print it.
26147         * tree-vect-stmts.c (vectorizable_load): Use the step alignment
26148         to tell whether the step preserves vector (mis)alignment.
26149         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26150         Move the check for an integer step and generalise to all INTEGER_CST.
26151         (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
26152         Print the outer step alignment.
26154 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26156         * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
26157         with offset_alignment.
26158         (DR_ALIGNED_TO): Delete.
26159         (DR_OFFSET_ALIGNMENT): New macro.
26160         * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
26161         (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
26162         * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
26163         (create_data_ref): Likewise.
26164         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26165         (vect_analyze_data_refs): Likewise.
26166         * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
26167         creating dummy innermost behavior.
26169 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26171         * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
26172         with a "innermost_loop_behavior *" and refeence tree.
26173         * tree-data-ref.c (dr_analyze_innermost): Likewise.
26174         (create_data_ref): Update call accordingly.
26175         * tree-predcom.c (find_looparound_phi): Likewise.
26177 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26179         * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
26180         fields with dr_wrt_vec_loop.
26181         (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
26182         (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
26183         (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
26184         (vect_dr_behavior): New function.
26185         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26186         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26187         vect_dr_behavior.  Use a step_preserves_misalignment_p boolean to
26188         track whether the step preserves the misalignment.
26189         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26190         Use vect_dr_behavior.
26191         (vect_setup_realignment): Update call accordingly.
26192         (vect_create_data_ref_ptr): Likewise.  Use vect_dr_behavior.
26193         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
26194         call to vect_create_addr_base_for_vector_ref.
26195         (vect_create_cond_for_align_checks): Likewise.
26196         * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
26197         STMT_VINFO_DR_WRT_VEC_LOOP as a block.
26198         (vect_recog_mask_conversion_pattern): Likewise.
26199         * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
26200         (new_stmt_vec_info): Remove redundant zeroing.
26202 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26204         * common/config/arm/arm-common.c (arm_be8_option): New function.
26205         * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
26206         (ISA_ARMv6): Add isa_bit_be8.
26207         * config/arm/arm.h (arm_be8_option): Add prototype.
26208         (BE8_SPEC_FUNCTION): New define.
26209         (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
26210         * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
26211         (mlittle-endian): Similarly.
26212         (mbe8, mbe32): New options.
26213         * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
26214         * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
26216 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26218         * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
26220 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26222         * tree-cfgcleanup.c (want_merge_blocks_p): New function.
26223         (cleanup_tree_cfg_bb): Use it.
26224         * profile-count.h (profile_count::of_for_merging, profile_count::merge):
26225         New functions.
26226         * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
26228 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26230         PR bootstrap/81285
26231         * loop-doloop.c (add_test): Update profile.
26233 2017-07-03  Martin Liska  <mliska@suse.cz>
26235         PR sanitize/81040
26236         * sanopt.c (rewrite_usage_of_param): New function.
26237         (sanitize_rewrite_addressable_params): Likewise.
26238         (pass_sanopt::execute): Call rewrite_usage_of_param.
26240 2017-07-03  Richard Biener  <rguenther@suse.de>
26242         * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26243         back to using VIEW_CONVERT_EXPR.
26245 2017-07-03  Martin Liska  <mliska@suse.cz>
26247         PR other/78366
26248         * doc/extend.texi: Document when a resolver function is
26249         generated for target_clones.
26251 2017-07-03  Martin Liska  <mliska@suse.cz>
26253         * asan.c (asan_emit_stack_protection): Unpoison just red zones
26254         and shadow memory of auto variables which are subject of
26255         use-after-scope sanitization.
26256         (asan_expand_mark_ifn): Add do set only when is_poison.
26258 2016-07-03  Richard Biener  <rguenther@suse.de>
26260         * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26261         reduction PHIs.
26262         (vect_force_simple_reduction): Record reduction def -> phi mapping.
26263         (vectorizable_reduction): Perform reduction PHI creation when
26264         visiting a reduction PHI and adjust and simplify code generation
26265         phase of the reduction op.  Cache dts, use fold_binary, not fold_build2.
26266         (vect_transform_loop): Visit reduction PHIs.
26267         * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26268         defs into the SLP tree.
26269         (vect_build_slp_tree): Reduction defs terminate the recursion.
26270         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26271         of reduction defs.
26272         (vect_get_vec_defs_for_stmt_copy): Export.
26273         (vect_get_vec_defs): Likewise.
26274         * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26275         purpose.
26276         (vect_get_vec_defs_for_stmt_copy): Declare.
26277         (vect_get_vec_defs): Likewise.
26279 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26281         * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26282         parameter with a "loop" parameter and use it instead of the
26283         loop containing DR_STMT.  Don't check simple_iv when doing
26284         BB analysis.  Describe the two analysis modes in the comment.
26286 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26288         PR tree-optimization/69468
26289         * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26290         (find_same_succ_bb): Handle ignore_edge_flags.
26292 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26294         PR tree-optimization/81192
26295         * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26296         hash.
26297         (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26298         differs.
26299         (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26301 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26303         PR tree-optimization/81192
26304         * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26305         BB_SAME_SUCC (bb) == NULL.
26307 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26309         * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26310         consistency.
26312 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26314         * dumpfile.c: Include profile-count.h
26315         * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26316         update profile.
26317         (insert_cond_bb): Update profile.
26318         * tree-cfg.h (insert_cond_bb): Update prototype.
26319         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26320         * tree-dump.c: Do not include tree-cfg.
26322 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26324         * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26326 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26328         * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26329         bb.
26331 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26333         * tree-complex.c (expand_complex_div_wide): update profile.
26335 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26336             Alan Hayward  <alan.hayward@arm.com>
26337             David Sherwood  <david.sherwood@arm.com>
26339         * Makefile.in (MACHMODE_H): Remove insn-modes.h
26340         (CORETYPES_H): New define.
26341         (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26342         (insn-modes-inline.h, s-modes-inline-h): New rules.
26343         (generated_files): Add insn-modes-inline.h.
26344         (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26345         (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26346         (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26347         (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26348         (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26349         (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26350         (build/gencodes.o, build/genconditions.o): Likewise.
26351         (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26352         (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26353         (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26354         (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26355         (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26356         (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26357         * coretypes.h: Include everything up to real.h for generators.
26358         Include insn-modes.h first.  Include wide-int-print.h after
26359         wide-int.h.  Include insn-modes-inline.h and then machmode.h.
26360         * machmode.h: Don't include insn-modes.h here.
26361         * function-tests.c: Remove includes of signop.h, machmode.h,
26362         double-int.h and wide-int.h.
26363         * rtl.h: Likewise.
26364         * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26365         and wide-int.h.
26366         * optc-save-gen.awk: Likewise.
26367         * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26368         * godump.c: Remove include of wide-int-print.h.
26369         * pretty-print.h: Likewise.
26370         * wide-int-print.cc: Likewise.
26371         * wide-int.cc: Likewise.
26372         * hash-map-tests.c: Remove include of signop.h.
26373         * hash-set-tests.c: Likewise.
26374         * rtl-tests.c: Likewise.
26375         * mkconfig.sh: Remove include of machmode.h.
26376         * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26377         into...
26378         (emit_insn_modes_inline_h): ...this new function.  Emit the code
26379         into an insn-modes-inline.h header file, adding appropriate
26380         include guards and end comments.
26381         (emit_insn_modes_c_header): Remove include of machmode.h.
26382         (emit_min_insn_modes_c_header): Include coretypes.h rather than
26383         machmode.h.
26384         (main): Handle -i flag and call emit_insn_modes_inline_h when
26385         it is passed.
26387 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26389         * tree-ssa-strlen.c (strinfo): Rename the length field to
26390         nonzero_chars.  Add a full_string_p field.
26391         (compare_nonzero_chars, zero_length_string_p): New functions.
26392         (get_addr_stridx): Add an offset_out parameter.
26393         Use compare_nonzero_chars.
26394         (get_stridx): Update accordingly.  Use compare_nonzero_chars.
26395         (new_strinfo): Update after above changes to strinfo.
26396         (set_endptr_and_length): Set full_string_p.
26397         (get_string_length): Update after above changes to strinfo.
26398         (unshare_strinfo): Update call to new_strinfo.
26399         (maybe_invalidate): Likewise.
26400         (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26401         Use compare_nonzero_chars and zero_string_p.  Treat nonzero_chars
26402         as a uhwi instead of an shwi.  Update after above changes to
26403         strinfo and new_strinfo.
26404         (zero_length_string): Assert that chainsi contains full strings.
26405         Use zero_length_string_p.  Update call to new_strinfo.
26406         (adjust_related_strinfos): Update after above changes to strinfo.
26407         Copy full_string_p from origsi.
26408         (adjust_last_stmt): Use zero_length_string_p.
26409         (handle_builtin_strlen): Update after above changes to strinfo and
26410         new_strinfo.  Install the lhs as the string length if the previous
26411         entry didn't describe a full string.
26412         (handle_builtin_strchr): Update after above changes to strinfo
26413         and new_strinfo.
26414         (handle_builtin_strcpy): Likewise.
26415         (handle_builtin_strcat): Likewise.
26416         (handle_builtin_malloc): Likewise.
26417         (handle_pointer_plus): Likewise.
26418         (handle_builtin_memcpy): Likewise.  Track nonzero characters
26419         that aren't necessarily followed by a nul terminator.
26420         (handle_char_store): Likewise.
26422 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26424         PR tree-optimization/80769
26425         * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26426         for malloc and calloc.  Document the new invariant that all related
26427         strinfos have delayed lengths or none do.
26428         (verify_related_strinfos): Move earlier in file.
26429         (set_endptr_and_length): New function, split out from...
26430         (get_string_length): ...here.  Also set the lengths of related
26431         strinfos.
26432         (zero_length_string): Assert that chainsi has known (rather than
26433         delayed) lengths.
26434         (adjust_related_strinfos): Likewise.
26436 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26438         PR tree-optimization/81136
26439         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26440         assert that two references with the same misalignment have the same
26441         compile-time misalignment if those compile-time misalignments
26442         are known.
26444 2017-07-01  Andi Kleen  <ak@linux.intel.com>
26446         * print-tree.c (print_node): Print all attributes.
26448 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26450         * cfg.c (scale_bbs_frequencies): New function.
26451         * cfg.h (scale_bbs_frequencies): Declare it.
26452         * cfgloopanal.c (single_likely_exit): Cleanup.
26453         * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26454         as parameter.
26455         (scale_loop_profile): Likewise.
26456         (loop_version): Likewise.
26457         (create_empty_loop_on_edge): Update.
26458         * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26459         scale_loop_frequencies, scale_loop_profile, loopify,
26460         loop_version): Update prototypes.
26461         * modulo-sched.c (sms_schedule): Update.
26462         * predict.c (unlikely_executed_edge_p): Also check probability.
26463         (probably_never_executed_edge_p): Fix typo.
26464         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26465         * tree-parloops.c (gen_parallel_loop): Update.
26466         * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26467         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26468         * tree-ssa-loop-split.c (split_loop): Update.
26469         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26470         * tree-vect-loop-manip.c (vect_do_peeling): Update.
26471         (vect_loop_versioning): Update.
26472         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26474 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26476         * trans-mem.c (split_bb_make_tm_edge): Update profile.
26478 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26480         * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26481         to keep profile consistent.
26483 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26485         * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26486         * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26487         * profile-count.h (max_safe_multiplier): Make unsigned.
26488         (profile_count::guessed_zero): New.
26490 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26492         * bb-reorder.c (fix_up_crossing_landing_pad,
26493         fix_crossing_conditional_branches): Use make_single_succ_edge
26494         to keep profile consistent.
26496 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26498         * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26499         to update profile.
26501 2017-07-01  Jakub Jelinek  <jakub@redhat.com>
26503         PR sanitizer/81262
26504         * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26505         the right scopes, make sure cond_jump isn't preserved between multiple
26506         iterations.  Search for fallthru edge whenever there are 3+ edges and
26507         use find_fallthru_edge for it.
26509 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26511         Patch by Alexander Monakov <amonakov@ispras.ru>
26512         * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26513         probabilities consistently.
26515 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26517         * pa.c (pa_expand_compare_and_swap_loop): Update call of
26518         emit_cmp_and_jump_insns.
26520 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26522         PR ipa/81261
26523         * tree-inline.c (expand_call_inline): Combine profile statuses.
26525 2017-06-30  Andrew Pinski  <apinski@cavium.com>
26527         * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26528         fold_stmt returned true.
26530 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26532         * ggc.h (empty_string): Delete.
26533         * cfgexpand.c (expand_asm_stmt): Use plain "".
26534         * optabs.c (expand_asm_memory_barrier): Likewise.
26535         * stringpool.c (empty_string): Delete.
26536         (digit_vector, digit_string): Delete.
26537         (ggc_alloc_string): Use plain "", don't optimize single digit
26538         strings.  Use ggc_alloc_atomic.
26540 2017-06-30  Richard Earnshaw  <rearnsha@arm.com>
26542         * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26543         comparison set and one other set, use the cost of the non-comparison
26544         set.
26546 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26548         * ggc.h: Replace all 'static inline' with plain 'inline'.  Fix
26549         some formatting.
26551 2017-06-30  Peter Bergner  <bergner@vnet.ibm.com>
26553         * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26554         loops.  Remove now unneeded calls to gimple_switch_set_label() that
26555         just set removed labels to NULL_TREE.
26557 2017-06-30  Aldy Hernandez  <aldyh@redhat.com>
26559         * tree-ssanames.c (set_range_info_raw): Abstract from ...
26560         (set_range_info): ...here.  Only call set_range_info_raw if domain
26561         is useful.
26562         (set_nonzero_bits): Call set_range_info_raw.
26563         * tree-ssanames.h (set_range_info_raw): New.
26565 2017-06-30  Jakub Jelinek  <jakub@redhat.com>
26567         PR target/81225
26568         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26569         V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26570         of nonimmediate_operand and <store_mask_constraint> instead of m
26571         for the input operand.  For V8FI iterator, always split if input
26572         is a MEM.  For V16FI and V8SF_256 iterators, don't test if both
26573         operands are MEM if <mask_applied>.  For VI4F_256 iterator, use
26574         <store_mask_predicate> instead of register_operand and
26575         <store_mask_constraint> instead of v for the input operand.  Make
26576         sure both operands aren't MEMs for if not <mask_applied>.
26578 2017-06-30  Sylvestre Ledru  <sylvestre@debian.org>
26580         * lto-wrapper.c (copy_file) Close both file descriptors before
26581         exiting normally.
26583 2017-06-30  Martin Liska  <mliska@suse.cz>
26585         PR ipa/81214
26586         * multiple_target.c (create_dispatcher_calls): Make ifunc
26587         also for function that don't have calls or are not referenced.
26589 2017-06-30  Richard Biener  <rguenther@suse.de>
26591         * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26592         analyze the first scalar stmt.  Move vector type computation
26593         for the BB case here from ...
26594         * tree-vect-stmts.c (vect_analyze_stmt): ... here.  Guard
26595         live operation processing in the SLP case properly.
26597 2017-06-30  Richard Biener  <rguenther@suse.de>
26599         * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26601 2017-06-30  Martin Liska  <mliska@suse.cz>
26603         PR sanitizer/81021
26604         * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26605         before BUILT_IN_UNWIND_RESUME when ASAN is used.
26607 2017-06-30  Yvan Roux  <yvan.roux@linaro.org>
26609         * doc/invoke.texi (AArch64): Add missing options and remove redundant
26610         ones.
26612 2017-06-30  Richard Biener  <rguenther@suse.de>
26614         PR tree-optimization/81249
26615         * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26616         condition reduction result to original scalar type.
26618 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26620         * profile-count.h (enum profile_quality): Fix typos and whitespace
26621         issues.
26623 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26625         * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26626         type for branch probabilities.
26628 2017-06-29  Julian Brown  <julian@codesourcery.com>
26629             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26631         * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26632         * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26633         (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26634         (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26636 2017-06-29  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26638         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26639         check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26640         * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26641         CC usage from generic code to here.
26642         * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26643         CC usage into the target macros.
26645 2017-06-29  Maya Rashish  <coypu@sdf.org>
26647         * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26648         objects.
26650 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26652         * arm/arm-builtins.c: Include profile-count.h
26653         * except.c (sjlj_emit_function_enter): Use
26654         profile_probability::unlikely.
26656 2017-06-29  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26658         * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26659         and tocrel_offset be pointer args rather than implicitly using
26660         static versions.
26661         (legitimate_constant_pool_address_p, rs6000_emit_move,
26662         const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26663         tocrel_offset and use in toc_relative_expr_p call.
26664         (print_operand, print_operand_address): Use static tocrel_base_oac
26665         and tocrel_offset_oac.
26666         (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26667         tocrel_offset_oac.
26669 2017-06-29  Maya Rashish  <coypu@sdf.org>
26671         * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26673 2017-06-29  Eric Botcazou  <ebotcazou@adacore.com>
26675         * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26676         objects, take into account only the alignment of 'op0' and 'mode1' if
26677         'op0' is a MEM.
26679 2017-06-29  Steve Ellcey  <sellcey@cavium.com>
26681         * ccmp.c (ccmp_tree_comparison_p): New function.
26682         (ccmp_candidate_p): Update to use above function.
26683         (get_compare_parts): New function.
26684         (expand_ccmp_next): Update to use new functions.
26685         (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26686         new functions.
26687         (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26688         take mode as argument.
26689         * ccmp.h (expand_ccmp_expr): Add mode as argument.
26690         * expr.c (expand_expr_real_1): Pass mode as argument.
26692 2017-06-29  Segher Boessenkool  <segher@kernel.crashing.org>
26694         * combine.c (combine_instructions): Print insns to dump_file, together
26695         with their costs.
26697 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26699         * asan.c (asan_emit_stack_protection): Update.
26700         (create_cond_insert_point): Update.
26701         * auto-profile.c (afdo_propagate_circuit): Update.
26702         * basic-block.h (struct edge_def): Turn probability to
26703         profile_probability.
26704         (EDGE_FREQUENCY): Update.
26705         * bb-reorder.c (find_traces_1_round): Update.
26706         (better_edge_p): Update.
26707         (sanitize_hot_paths): Update.
26708         * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26709         (make_single_succ_edge): Update.
26710         (check_bb_profile): Update.
26711         (dump_edge_info): Update.
26712         (update_bb_profile_for_threading): Update.
26713         * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26714         probabilitycount to 0.
26715         * cfgbuild.c (compute_outgoing_frequencies): Update.
26716         * cfgcleanup.c (try_forward_edges): Update.
26717         (outgoing_edges_match): Update.
26718         (try_crossjump_to_edge): Update.
26719         * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26720         (expand_gimple_tailcall): Update.
26721         (construct_init_block): Use make_single_succ_edge.
26722         (construct_exit_block): Use make_single_succ_edge.
26723         * cfghooks.c (verify_flow_info): Update.
26724         (redirect_edge_succ_nodup): Update.
26725         (split_edge): Update.
26726         (account_profile_record): Update.
26727         * cfgloopanal.c (single_likely_exit): Update.
26728         * cfgloopmanip.c (scale_loop_profile): Update.
26729         (set_zero_probability): Remove.
26730         (duplicate_loop_to_header_edge): Update.
26731         * cfgloopmanip.h (loop_version): Update prototype.
26732         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26733         (force_nonfallthru_and_redirect): Update.
26734         (update_br_prob_note): Update.
26735         (rtl_verify_edges): Update.
26736         (purge_dead_edges): Update.
26737         (rtl_lv_add_condition_to_bb): Update.
26738         * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26739         * cgraphunit.c (init_lowered_empty_function): Update.
26740         (cgraph_node::expand_thunk): Update.
26741         * cilk-common.c: Include profile-count.h
26742         * dojump.c (inv): Remove.
26743         (jumpifnot): Update.
26744         (jumpifnot_1): Update.
26745         (do_jump_1): Update.
26746         (do_jump): Update.
26747         (do_jump_by_parts_greater_rtx): Update.
26748         (do_compare_rtx_and_jump): Update.
26749         * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26750         do_jump_1. do_compare_rtx_and_jump): Update prototype.
26751         * dwarf2cfi.c: Include profile-count.h
26752         * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26753         (sjlj_emit_dispatch_table): Likewise.
26754         * explow.c: Include profile-count.h
26755         * expmed.c (emit_store_flag_force): Update.
26756         (do_cmp_and_jump): Update.
26757         * expr.c (compare_by_pieces_d::generate): Update.
26758         (compare_by_pieces_d::finish_mode): Update.
26759         (emit_block_move_via_loop): Update.
26760         (store_expr_with_bounds): Update.
26761         (store_constructor): Update.
26762         (expand_expr_real_2): Update.
26763         (expand_expr_real_1): Update.
26764         * expr.h (try_casesi, try_tablejump): Update prototypes.
26765         * gimple-pretty-print.c (dump_probability): Update.
26766         (dump_profile): New.
26767         (dump_gimple_label): Update.
26768         (dump_gimple_bb_header): Update.
26769         * graph.c (draw_cfg_node_succ_edges): Update.
26770         * hsa-gen.c (convert_switch_statements): Update.
26771         * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26772         (find_if_case_1): Update.
26773         (find_if_case_2): Update.
26774         * internal-fn.c (expand_arith_overflow_result_store): Update.
26775         (expand_addsub_overflow): Update.
26776         (expand_neg_overflow): Update.
26777         (expand_mul_overflow): Update.
26778         (expand_vector_ubsan_overflow): Update.
26779         * ipa-cp.c (good_cloning_opportunity_p): Update.
26780         * ipa-split.c (split_function): Use make_single_succ_edge.
26781         * ipa-utils.c (ipa_merge_profiles): Update.
26782         * loop-doloop.c (add_test): Update.
26783         (doloop_modify): Update.
26784         * loop-unroll.c (compare_and_jump_seq): Update.
26785         (unroll_loop_runtime_iterations): Update.
26786         * lra-constraints.c (lra_inheritance): Update.
26787         * lto-streamer-in.c (input_cfg): Update.
26788         * lto-streamer-out.c (output_cfg): Update.
26789         * mcf.c (adjust_cfg_counts): Update.
26790         * modulo-sched.c (sms_schedule): Update.
26791         * omp-expand.c (expand_omp_for_init_counts): Update.
26792         (extract_omp_for_update_vars): Update.
26793         (expand_omp_ordered_sink): Update.
26794         (expand_omp_for_ordered_loops): Update.
26795         (expand_omp_for_generic): Update.
26796         (expand_omp_for_static_nochunk): Update.
26797         (expand_omp_for_static_chunk): Update.
26798         (expand_cilk_for): Update.
26799         (expand_omp_simd): Update.
26800         (expand_omp_taskloop_for_outer): Update.
26801         (expand_omp_taskloop_for_inner): Update.
26802         * omp-simd-clone.c (simd_clone_adjust): Update.
26803         * optabs.c (expand_doubleword_shift): Update.
26804         (expand_abs): Update.
26805         (emit_cmp_and_jump_insn_1): Update.
26806         (expand_compare_and_swap_loop): Update.
26807         * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26808         * predict.c (predictable_edge_p): Update.
26809         (edge_probability_reliable_p): Update.
26810         (set_even_probabilities): Update.
26811         (combine_predictions_for_insn): Update.
26812         (combine_predictions_for_bb): Update.
26813         (propagate_freq): Update.
26814         (estimate_bb_frequencies): Update.
26815         (force_edge_cold): Update.
26816         * profile-count.c (profile_count::dump): Add missing space into dump.
26817         (profile_count::debug): Add newline.
26818         (profile_count::differs_from_p): Explicitly convert to unsigned.
26819         (profile_count::stream_in): Update.
26820         (profile_probability::dump): New member function.
26821         (profile_probability::debug): New member function.
26822         (profile_probability::differs_from_p): New member function.
26823         (profile_probability::differs_lot_from_p): New member function.
26824         (profile_probability::stream_in): New member function.
26825         (profile_probability::stream_out): New member function.
26826         * profile-count.h (profile_count_quality): Rename to ...
26827         (profile_quality): ... this one.
26828         (profile_probability): New.
26829         (profile_count): Update.
26830         * profile.c (compute_branch_probabilities): Update.
26831         * recog.c (peep2_attempt): Update.
26832         * sched-ebb.c (schedule_ebbs): Update.
26833         * sched-rgn.c (find_single_block_region): Update.
26834         (compute_dom_prob_ps): Update.
26835         (schedule_region): Update.
26836         * sel-sched-ir.c (compute_succs_info): Update.
26837         * stmt.c (struct case_node): Update.
26838         (do_jump_if_equal): Update.
26839         (get_outgoing_edge_probs): Update.
26840         (conditional_probability): Update.
26841         (emit_case_dispatch_table): Update.
26842         (expand_case): Update.
26843         (expand_sjlj_dispatch_table): Update.
26844         (emit_case_nodes): Update.
26845         * targhooks.c: Update.
26846         * tracer.c (better_p): Update.
26847         (find_best_successor): Update.
26848         * trans-mem.c (expand_transaction): Update.
26849         * tree-call-cdce.c: Update.
26850         * tree-cfg.c (gimple_split_edge): Upate.
26851         (move_sese_region_to_fn): Upate.
26852         * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26853         * tree-eh.c (lower_resx): Upate.
26854         (cleanup_empty_eh_move_lp): Upate.
26855         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26856         * tree-inline.c (copy_edges_for_bb): Update.
26857         (copy_cfg_body): Update.
26858         * tree-parloops.c (gen_parallel_loop): Update.
26859         * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26860         (gimple_gen_time_profiler): Update.
26861         * tree-ssa-dce.c (remove_dead_stmt): Update.
26862         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26863         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26864         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26865         (unloop_loops): Update.
26866         (try_peel_loop): Update.
26867         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26868         * tree-ssa-loop-split.c (connect_loops): Update.
26869         (split_loop): Update.
26870         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26871         (hoist_guard): Update.
26872         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26873         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26874         (value_replacement): Update.
26875         * tree-ssa-reassoc.c (branch_fixup): Update.
26876         * tree-ssa-tail-merge.c (replace_block_by): Update.
26877         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26878         (create_edge_and_update_destination_phis): Update.
26879         (compute_path_counts): Update.
26880         (recompute_probabilities): Update.
26881         (update_joiner_offpath_counts): Update.
26882         (freqs_to_counts_path): Update.
26883         (duplicate_thread_path): Update.
26884         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26885         (struct switch_conv_info): Update.
26886         (gen_inbound_check): Update.
26887         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26888         (vect_do_peeling): Update.
26889         (vect_loop_versioning): Update.
26890         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26891         (optimize_mask_stores): Update.
26892         * ubsan.c (ubsan_expand_null_ifn): Update.
26893         * value-prof.c (gimple_divmod_fixed_value): Update.
26894         (gimple_divmod_fixed_value_transform): Update.
26895         (gimple_mod_pow2): Update.
26896         (gimple_mod_pow2_value_transform): Update.
26897         (gimple_mod_subtract): Update.
26898         (gimple_mod_subtract_transform): Update.
26899         (gimple_ic): Update.
26900         (gimple_stringop_fixed_value): Update.
26901         (gimple_stringops_transform): Update.
26902         * value-prof.h: Update.
26904 2017-06-29  Carl Love  <cel@us.ibm.com>
26906         * config/rs6000/rs6000-c.c: Add support for built-in functions
26907         vector signed int vec_signed (vector float);
26908         vector signed long long vec_signed (vector double);
26909         vector signed int vec_signed2 (vector double, vector double);
26910         vector signed int vec_signede (vector double);
26911         vector signed int vec_signedo (vector double);
26912         * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
26913         instruction generator.
26914         * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26915         UNSPEC_VSX_VSIGNED2): Add UNSPECS.
26916         (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
26917         Add define_insn.
26918         (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
26919         vunsignede_v2df): Add define_expands.
26920         * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
26921         VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
26922         VEC_UNSIGNEDO): Add definitions.
26923         * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26924         UNSPEC_VSX_VSIGNED2): Add UNSPECs.
26925         (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
26926         (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
26927         vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
26928         * config/rs6000/altivec.h (vec_signed, vec_signed2,
26929         vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
26930         vec_unsignede, vec_unsignedo): Add builtin defines.
26931         * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
26932         declaration.
26933         * doc/extend.texi: Update the built-in documentation file for the
26934         new built-in functions.
26936 2017-06-29  Richard Biener  <rguenther@suse.de>
26938         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
26939         reduction chains to LOOP_VINFO_REDUCTIONS.
26940         * tree-vect-slp.c (vect_analyze_slp): Continue looking for
26941         SLP reductions after processing reduction chains.
26943 2017-06-29  Nathan Sidwell  <nathan@acm.org>
26945         * builtins.c (fold_builtin_FUNCTION): Use
26946         lang_hooks.decl_printable_name.
26948 2017-06-29  Peter Bergner  <bergner@vnet.ibm.com>
26950         PR middle-end/81194
26951         * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
26952         with only one label.
26953         * stmt.c (expand_case): Assert NCASES is greater than one.
26955 2017-06-29  Richard Biener  <rguenther@suse.de>
26957         * tree-cfg.c (group_case_labels_stmt): Return whether we changed
26958         anything.
26959         (group_case_labels): Likewise.
26960         (find_taken_edge): Push sanity checking on val to workers...
26961         (find_taken_edge_cond_expr): ... here
26962         (find_taken_edge_switch_expr): ... and here, handle cases
26963         with just a default label.
26964         * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
26965         (group_case_labels): Likewise.
26966         * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
26967         group_case_labels does anything cleanup the CFG again.
26969 2017-06-29  Bin Cheng  <bin.cheng@arm.com>
26971         PR tree-optimization/81196
26972         * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
26973         exit condition comparing two IVs.
26975 2017-06-29  Richard Earnshaw  <rearnsha@arm.com>
26977         * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
26978         profile to the dummy entry at the end of the list of architectures.
26979         * config/arm/arm-cpu-cdata.h: Regenerated.
26981 2017-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26982             Michael Collison <michael.collison@arm.com>
26984         PR target/70119
26985         * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
26986         New pattern.
26987         (*aarch64_reg_<mode>3_neg_mask2): New pattern.
26988         (*aarch64_reg_<mode>3_minus_mask): New pattern.
26989         (*aarch64_<optab>_reg_di3_mask2): New pattern.
26990         * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
26991         of shift when the shift amount is masked with constant equal to
26992         the size of the mode.
26993         * config/aarch64/predicates.md (subreg_lowpart_operator): New
26994         predicate.
26996 2017-06-29  Martin Liska  <mliska@suse.cz>
26998         * config/i386/i386.opt: Change range from [1,5] to [0,5].
27000 2017-06-29  Yury Gribov  <tetra2005@gmail.com>
27002         PR bootstrap/80565
27003         * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
27004         code.
27005         * ipa-inline.h
27006         (edge_growth_cache_entry::edge_growth_cache_entry): New
27007         function.
27008         (reset_edge_growth_cache): Update to use constructor.
27010 2017-06-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
27012         * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
27013         (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
27014         (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
27016 2017-06-28  Sebastian Peryt  <sebastian.peryt@intel.com>
27018         * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
27019         (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
27021 2017-06-28  Szabolcs Nagy  <szabolcs.nagy@arm.com>
27023         * config.gcc (*-linux-musl*): Add t-musl tmake_file.
27024         (*-linux-uclibc*): Add t-uclibc tmake_file.
27025         * config/t-musl: New.
27026         * config/t-uclibc: New.
27028 2017-06-28  Richard Earnshaw  <rearnsha@arm.com>
27030         * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
27031         context.
27032         (gen_comm_data): Emit architectural setting of arch_prof.
27033         * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
27034         profile.
27035         (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
27036         (armv8-m.base, armv8-m.main): Likewise.
27037         * arm-protos.h (arm_build_target): Add profile field.
27038         (arch_option): Likewise.
27039         * config/arm/arm.c (arm_configure_build_target): Copy the profile to
27040         the active target.
27041         * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
27042         arm_active_target.profile.
27044 2017-06-28  Richard Biener  <rguenther@suse.de>
27046         PR middle-end/81227
27047         * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
27048         TYPE_OVERFLOW_WRAPS.
27049         * match.pd (negate_expr_p): Likewise.
27050         * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
27051         fold_build2, not fold_binary.
27053 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
27055         * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
27056         Convert memory address to Pmode.
27057         (aarch64_print_operand): Assert MEM operands are always Pmode.
27059 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
27061         PR target/79665
27062         * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
27063         Remove redundant if.
27064         (aarch_forward_to_shift_is_not_shifted_reg): Remove.
27065         * config/arm/aarch-common-protos.h
27066         (aarch_forward_to_shift_is_not_shifted_re): Remove.
27067         * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
27069 2017-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
27071         PR ipa/81238
27072         * multiple_target.c (create_dispatcher_calls): Set the default
27073         clone to be static, not public.
27075 2017-06-28  Richard Biener  <rguenther@suse.de>
27077         * tree-vect-loop.c (vectorizable_reduction): Move special
27078         cond reduction IV var creation ...
27079         (vect_create_epilog_for_reduction): ... here.  Remove induction_index
27080         parameter.  Use STMT_VINFO_VECTYPE.
27081         * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
27082         constant_p.
27084 2017-06-28  Martin Liska  <mliska@suse.cz>
27086         PR ipa/81128
27087         * ipa-visibility.c (non_local_p): Handle visibility.
27089 2017-06-28  Martin Liska  <mliska@suse.cz>
27091         PR driver/79659
27092         * common.opt: Add IntegerRange to various options.
27093         * opt-functions.awk (integer_range_info): New function.
27094         * optc-gen.awk: Add integer_range_info to cl_options struct.
27095         * opts-common.c (decode_cmdline_option): Handle
27096         CL_ERR_INT_RANGE_ARG.
27097         (cmdline_handle_error): Likewise.
27098         * opts.c (print_filtered_help): Show valid interval in
27099         when --help is provided.
27100         * opts.h (struct cl_option): Add range_min and range_max fields.
27101         * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
27103 2017-06-28  Marc Glisse  <marc.glisse@inria.fr>
27105         * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
27106         (x * C EQ/NE y * C): New transformation.
27108 2017-06-28  Christophe Lyon  <christophe.lyon@linaro.org>
27110         * genmultilib (combination_space): Accept '+' in option names.
27112 2017-06-28  Martin Liska  <mliska@suse.cz>
27114         PR sanitizer/81224
27115         * asan.c (instrument_derefs): Bail out inner references
27116         that are hard register variables.
27118 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
27120         PR target/81175
27121         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
27122         rather than def_builtin_pure for __builtin_ia32_gatherpf*.
27124 2017-06-28  Richard Biener  <rguenther@suse.de>
27126         * tree-vectorizer.h (vect_get_vec_defs): Remove.
27127         (vect_get_slp_defs): Adjust.
27128         * tree-vect-loop.c (get_initial_defs_for_reduction): Split
27129         out from ...
27130         * tree-vect-slp.c (vect_get_constant_vectors): ... here and
27131         simplify.
27132         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
27133         get_initial_defs_for_reduction instead of vect_get_vec_defs.
27134         (vectorizable_reduction): Adjust.
27135         * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
27136         handling.
27137         (vect_get_slp_defs): Likewise.
27138         * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
27139         (vectorizable_bswap): Adjust.
27140         (vectorizable_call): Likewise.
27141         (vectorizable_conversion): Likewise.
27142         (vectorizable_assignment): Likewise.
27143         (vectorizable_shift): Likewise.
27144         (vectorizable_operation): Likewise.
27145         (vectorizable_store): Likewise.
27146         (vectorizable_condition): Likewise.
27147         (vectorizable_comparison): Likewise.
27149 2017-06-28  Michael Collison  <michael.collison@arm.com>
27151         PR target/68535
27152         * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
27153         set of base_reg
27154         (arm_gen_movmemqi): Removed unused variable 'i'.
27155         Convert 'for' loop into 'while' loop.
27156         (arm_expand_prologue): Remove last unnecessary set of insn.
27157         (thumb_pop): Remove unused variable 'pushed_words'.
27158         (thumb_exit): Remove last unnecessary set of regs_to_pop.
27160 2017-06-28  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27162         * config/s390/predicates.md: Use s390_rel_address_ok_p.
27163         * config/s390/s390-protos.h: Add prototype of
27164         s390_rel_address_ok_p.
27165         * config/s390/s390.c (s390_got_symbol): New function.
27166         (s390_rel_address_ok_p): New function.
27167         (legitimize_pic_address): Use s390_rel_address_ok_p.
27168         (s390_load_got): Use s390_got_symbol.
27169         (s390_option_override): Issue error if
27170         -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
27171         * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
27172         New macro.
27173         * config/s390/s390.opt: New option mpic-data-is-text-relative.
27175 2017-06-27  Andrew Pinski  <apinski@cavium.com>
27177         * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
27178         (X * copysign (1.0, X)): New pattern.
27179         (X * copysign (1.0, -X)): New pattern.
27180         (copysign (-1.0, CST)): New pattern.
27182 2017-06-27  Joseph Myers  <joseph@codesourcery.com>
27184         * genmultilib (combination_space): Remove variable.
27185         Validate reuse rules against regular expression for any sequence
27186         of multilib options in any order.
27188 2017-06-27  Michael Collison  <michael.collison@arm.com>
27190         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
27191         call aarch64_split_simd_combine.
27192         * (aarch64_combine_internal<mode>): Delete pattern.
27193         * config/aarch64/aarch64.c (aarch64_split_simd_combine):
27194         Allow register and subreg operands.
27196 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27198         * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
27199         specific need, just fallback on defaults.
27200         (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
27202 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27203             Olivier Hainque  <hainque@adacore.com>
27205         * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
27206         map for 64bits.
27207         (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
27208         targets. Pick a default if no particular attempt applied.
27209         (STACK_CHECK_PROTECT): Double for 64bit targets, which have
27210         larger contexts.
27212 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27214         * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
27215         (x86_64-wrs-vxworks7): Likewise.
27217 2017-06-27  Marek Polacek  <polacek@redhat.com>
27219         PR sanitizer/81223
27220         * ubsan.c (instrument_null): Check get_base_address's result for null.
27222 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27224         * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
27226 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27228         * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
27229         BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
27230         (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
27231         BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
27232         New function types.
27233         * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
27234         BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
27235         BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
27236         BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
27237         BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27238         BUILT_IN_FEUPDATEENV): New builtins.
27239         * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27240         TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27241         * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27242         fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27243         macros.
27244         (builtin_structptr_types): Adjust size.
27245         * tree.c (builtin_structptr_types): Add four entries.
27247 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27248             Olivier Hainque  <hainque@adacore.com>
27250         * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27251         (TLS_SYM): New local macro, forcing reference to __tls__ on
27252         link command lines for VxWorks 7 RTPs, triggering initialization
27253         of tlsLib.
27254         (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27255         OS features TLS support, true for RTPs on VxWorks 7.
27256         * config/vxworks.c (vxworks_override_options): Setup emutls
27257         accordingly.
27259 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27261         * predict.c (test_prediction_value_range): Use -1U instead of -1
27262         to avoid narrowing conversion warning.
27263         * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27264         to avoid narrowing conversion warning.
27265         * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27266         -1.
27267         * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27269 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27271         * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27272         64bit configurations.
27273         (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27274         (SIZE_TYPE): Likewise.
27275         * config/vxworks.c (vxworks_emutls_var_fields): Use
27276         long_unsigned_type_node instead of unsigned_type_node as the offset
27277         field type, which is "pointer" mode in emutls.c.
27279 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27281         PR sanitizer/81209
27282         * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27284         PR middle-end/81207
27285         * gimple-fold.c (replace_call_with_call_and_fold): Handle
27286         gimple_vuse copying separately from gimple_vdef copying.
27288 2017-06-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27290         * value-prof.c (free_hist): Remove call to memset and the enclosing if
27291         condition.
27293 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27294             Olivier Hainque  <hainque@adacore.com>
27296         * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27297         for all vxworks7 targets.
27298         * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27299         (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27300         (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27301         variations for VX6/VX7 and 32/64bits later on in ...
27302         (VXWORKS_LIB_SPEC): Leverage new macros.
27303         (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27304         as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27306 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27308         * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27309         _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27311 2017-06-26  Carl Love  <cel@us.ibm.com>
27313         * config/rs6000/rs6000-c.c: Add support for built-in functions
27314         vector bool char vec_reve (vector bool char);
27315         vector signed char vec_reve (vector signed char);
27316         vector unsigned char vec_reve (vector unsigned char);
27317         vector bool int vec_reve (vector bool int);
27318         vector signed int vec_reve (vector signed int);
27319         vector unsigned int vec_reve (vector unsigned int);
27320         vector bool long long vec_reve (vector bool long long);
27321         vector signed long long vec_reve (vector signed long long);
27322         vector unsigned long long vec_reve (vector unsigned long long);
27323         vector bool short vec_reve (vector bool short);
27324         vector signed short vec_reve (vector signed short);
27325         vector double vec_reve (vector double);
27326         vector float vec_reve (vector float);
27327         * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27328         VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27329         * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27330         (altivec_vreve): New pattern.
27331         * config/rs6000/altivec.h (vec_reve): New define.
27332         * doc/extend.texi (vec_rev): Update the built-in documentation file
27333         for the new built-in functions.
27335 2016-06-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27337         PR tree-optimization/71815
27338         * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27339         function.
27340         (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27341         has_single_use.
27342         (slsr_process_phi): Likewise.
27343         (replace_uncond_cands_and_profitable_phis): Don't replace a
27344         multiply candidate with a stride of 1 (copy or cast).
27345         (phi_incr_cost): Call uses_consumed_by_stmt rather than
27346         has_single_use.
27347         (lowest_cost_path): Likewise.
27348         (total_savings): Likewise.
27350 2017-06-26  Richard Biener  <rguenther@suse.de>
27352         PR target/81175
27353         * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27354         Use def_builtin_pure for all gather builtins.
27356 2017-06-26  Richard Biener  <rguenther@suse.de>
27358         PR tree-optimization/81203
27359         * tree-tailcall.c (find_tail_calls): Do not move stmts into
27360         non-dominating BBs.
27362 2017-06-26  Marek Polacek  <polacek@redhat.com>
27364         PR c/80116
27365         * doc/invoke.texi: Document -Wmultistatement-macros.
27367 2017-06-26  Christophe Lyon  <christophe.lyon@linaro.org>
27369         * doc/sourcebuild.texi (ARM-specific attributes): Document new
27370         arm_neon_ok_no_float_abi effective target.
27372 2017-06-26  Richard Biener  <rguenther@suse.de>
27374         PR tree-optimization/80928
27375         * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27376         (copy_bbs): Set BB_DUPLICATED flag early.
27377         (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27378         marked blocks.
27379         (execute_on_shrinking_pred): Likewise.
27380         * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27381         BB_DUPLICATED blocks.
27382         * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27383         iterate over all PHIs considering removal of *gsi.
27385 2017-06-23  Jim Wilson  <jim.wilson@linaro.org>
27387         * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27388         qdf24xx.
27390 2017-06-23  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27392         * config/rs6000/rs6000-string.c: (expand_block_clear,
27393         do_load_for_compare, select_block_compare_mode,
27394         compute_current_alignment, expand_block_compare,
27395         expand_strncmp_align_check, expand_strn_compare,
27396         expand_block_move, rs6000_output_load_multiple)
27397         Move functions related to string/block move/compare
27398         to a separate file.
27399         * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27400         * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27401         for this function which is now used in two files.
27402         * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27403         * config.gcc: Add rs6000-string.o to extra_objs for
27404         targets powerpc*-*-* and rs6000*-*-*.
27406 2017-06-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
27408         PR target/80510
27409         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27410         32-bit, since indexed is not valid for DImode.
27411         (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27412         3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27413         (define_peephole2 for Altivec d-form load): Add 32-bit support.
27414         (define_peephole2 for Altivec d-form store): Likewise.
27416         PR ipa/81185
27417         * multiple_target.c (create_dispatcher_calls): Only create the
27418         dispatcher call if the function is the default clone of a
27419         versioned function.
27421 2017-06-23  Segher Boessenkool  <segher@kernel.crashing.org>
27423         PR middle-end/80902
27424         * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27425         a call, force the call to not be a tail call.
27427 2017-06-23  Jeff Law  <law@redhat.com>
27429         * doc/contrib.texi: Add entry for Steven Pemberton's work on
27430         enquire.
27432 2017-06-23  Will Schmidt  <will_schmidt@vnet.ibm.com>
27434         * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27435         update_call_from_tree().  (rs6000_gimple_fold_builtin): Add
27436         handling for early expansion of vector shifts (sl,sr,sra,rl).
27437         (builtin_function_type): Add vector shift right instructions
27438         to the unsigned argument list.
27440 2017-06-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27442         rtl-optimizatoin/79286
27443         * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27444         * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27445         trap.  PIC register plus a const unspec without offset can never trap.
27447 2017-06-23  Marc Glisse  <marc.glisse@inria.fr>
27449         * tree.h (builtin_structptr_type): New type.
27450         (builtin_structptr_types): Declare new array.
27451         * tree.c (builtin_structptr_types): New array.
27452         (free_lang_data, build_common_tree_nodes): Use it.
27454 2017-06-23  Jonathan Wakely  <jwakely@redhat.com>
27456         PR c++/81187
27457         * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27458         -Wnoexcept.
27460 2017-06-22  Matt Turner  <mattst88@gmail.com>
27462         * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27463         Lake models to skylake case.  Assume skylake for unknown
27464         models with clflushopt.
27466 2017-06-22  Jeff Law  <law@redhat.com>
27468         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27469         frame sizes that do not satisfy aarch64_uimm12_shift.
27471 2017-06-22  Jan Hubicka <hubicka@ucw.cz>
27473         * profile-count.h (apply_probability,
27474         apply_scale, probability_in): Fix checks for zero.
27476 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27478         * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27479         * doc/cppdiropts.texi (-I @var{dir}): Document it.
27481 2016-06-22  Richard Biener  <rguenther@suse.de>
27483         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27484         COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27485         REDUC_MAX_EXPR support.
27486         (vectorizable_reduction): Likewise.
27487         (vect_create_epilog_for_reduction): Likewise.
27489 2017-06-22  James Greenhalgh  <james.greenhalgh@arm.com>
27491         * match.pd (A / (1 << B) -> A >> B): New.
27492         * generic-match-head.c: Include optabs-tree.h.
27493         * gimple-match-head.c: Likewise.
27494         * optabs-tree.h (target_supports_op_p): New.
27495         * optabs-tree.c (target_supports_op_p): New.
27497 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27499         * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27500         $gcc_cv_ld --help output.
27501         (gcc_cv_ld_demangle): Likewise.
27502         (gcc_cv_ld_eh_frame_hdr): Likewise.
27503         (gcc_cv_ld_pie): Likewise.
27504         (gcc_cv_ld_as_needed): Likewise.  Prefer native forms unless $gnu_ld.
27505         (gcc_cv_ld_buildid): Likewise.
27506         (gcc_cv_ld_sysroot): Likewise.
27507         (ld_bndplt_support): Likewise.
27508         (ld_pushpopstate_support): Likewise.
27509         * configure: Regenerate.
27510         * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27512 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27514         PR target/81151
27515         * config/i386/sse.md (round<mode>2): Renumber match_dup and
27516         operands indexes to avoid gap between operands and match_dups.
27518 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27520         * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27521         Increment Arith_shift and Arith_shift_reg by 1.
27522         * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27523         New tuning flag.
27524         * config/aarch64/aarch64.c (thunderx_tunings): Enable
27525         AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27526         (aarch64_strip_extend): Add new argument and test for it.
27527         (aarch64_cheap_mult_shift_p): New function.
27528         (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27529         add a cost if it is true.
27530         Update calls to aarch64_strip_extend.
27531         (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27533 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27535         * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27536         tunings.
27537         (thunderxt88): Likewise.
27538         * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27539         (thunderx_prefetch_tune): New variable.
27540         (thunderx2t99_prefetch_tune): Update for the correct values.
27541         (thunderxt88_tunings): New variable.
27542         (thunderx_tunings): Use thunderx_prefetch_tune instead of
27543         generic_prefetch_tune.
27544         (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27546 2017-06-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27548         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27549         SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27550         (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27551         (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27552         (aarch64_atomic_cas<mode>, GPI): Likewise.
27554 2017-06-21  Martin Liska  <mliska@suse.cz>
27556         * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27557         statements on cold and hot labels.
27558         * predict.c (tree_estimate_probability_bb): Remove the
27559         prediction from this place.
27561 2017-06-21  Martin Liska  <mliska@suse.cz>
27563         PR tree-optimization/79489
27564         * gimplify.c (maybe_add_early_return_predict_stmt): New
27565         function.
27566         (gimplify_return_expr): Call the function.
27567         * predict.c (tree_estimate_probability_bb): Remove handling
27568         of early return.
27569         * predict.def: Update comment about early return predictor.
27570         * gimple-predict.h (is_gimple_predict): New function.
27571         * predict.def: Change default value of early return to 66.
27572         * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27573         statements.
27574         * passes.def: Put pass_strip_predict_hints to the beginning of
27575         IPA passes.
27577 2017-06-21  Pierre-Marie de Rodat  <derodat@adacore.com>
27579         * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27580         FUNCTION_DECL declarations.
27581         (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27582         declarations.
27583         (dwaf2out_decl): Likewise.
27584         * godump.c (go_early_global_decl): Skip call to the real debug hook
27585         for FUNCTION_DECL declarations.
27586         * passes.c (rest_of_decl_compilation): Skip call to the
27587         early_global_decl debug hook for FUNCTION_DECL declarations, unless
27588         -fdump-go-spec is passed.
27590 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27592         * config/i386/i386.c (struct builtin_isa): New field pure_p.
27593         Reorder for compactness.
27594         (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27595         (def_builtin_pure, def_builtin_pure2): New functions.
27596         (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27598 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27600         * match.pd (nop_convert): New predicate.
27601         ((A +- CST1) +- CST2): Allow some NOP conversions.
27603 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27605         PR c++/81130
27606         * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27607         with ctors/dtors if GOVD_SHARED is set.
27609 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27611         * config/aarch64/aarch64.md (movti_aarch64):
27612         Emit mov rather than orr.
27613         (movtf_aarch64): Likewise.
27614         * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27615         Emit mov rather than orr.
27617 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27619         * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27620         Swap alternatives, make integer dup more expensive.
27622 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27624         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27625         Return true for non-tls symbols.
27627 2017-06-21  James Greenhalgh  <james.greenhalgh@arm.com>
27629         * config/aarch64/aarch64-cores.def (cortex-a55): New.
27630         (cortex-a75): Likewise.
27631         (cortex-a75.cortex-a55): Likewise.
27632         * config/aarch64/aarch64-tune.md: Regenerate.
27633         * doc/invoke.texi (-mtune): Document new values for -mtune.
27635 2017-06-21  Tom de Vries  <tom@codesourcery.com>
27637         * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27638         stack_size feature.
27639         (Effective-Target Keywords, Other attributes): Suggest using
27640         dg-add-options stack_size feature to get stack limit in stack_size
27641         effective target documentation.
27643 2017-06-21  Julian Brown  <julian@codesourcery.com>
27644             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
27646         * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27647         (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27648         * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27649         reservation.
27650         * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27651         attribute type list for neon_multiply.
27652         * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27653         attribute type list for neon_multiply.
27654         * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27655         * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27656         attribute type list for neon_multiply.
27657         * config/arm/types.md (crypto_pmull): Add.
27658         * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27659         attribute type list.
27661 2017-06-20  Andreas Tobler  <andreast@gcc.gnu.org>
27663         * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27664         arm1176jzf-s.
27666 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27668         * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27669         to make sure not to dereference a NULL cost_classes_ptr pointer.
27671 2017-06-20  Carl Love  <cel@us.ibm.com>
27673         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27674         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27675         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27676         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27677         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27678         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27679         VMULOSW): New enum "unspec" values.
27680         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27681         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27682         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27683         altivec_vmulosw): New patterns.
27684         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27685         VMULOSW): Add definitions.
27687 2017-06-20  Julia Koval  <julia.koval@intel.com>
27689         * config/i386/i386.c: Fix rounding expand for new pattern.
27690         * config/i386/subst.md: Fix pattern (parallel -> unspec).
27692 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27694         * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27695         * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27697 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27699         * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27700         feature string.
27702 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27704         * config/aarch64/aarch64-cores.def: Rearrange to sort by
27705         architecture, then by implementer ID.
27706         * config/aarch64/aarch64-tune.md: Regenerate.
27708 2017-06-20  Richard Biener  <rguenther@suse.de>
27710         PR middle-end/81097
27711         * fold-const.c (split_tree): Fold to type before negating.
27713 2017-06-20  David Malcolm  <dmalcolm@redhat.com>
27715         * diagnostic-show-locus.c
27716         (selftest::test_fixit_deletion_affecting_newline): New function.
27717         (selftest::diagnostic_show_locus_c_tests): Call it.
27719 2017-06-20  Andreas Schwab  <schwab@suse.de>
27721         PR target/80970
27722         * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27723         instead of "+d".
27725 2017-06-20  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
27727         * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27728         __ARM_FEATURE_COPROC according to support.
27730 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27732         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27733         Rewritten to avoid overflow for > 32-bit pointers.
27735         PR sanitizer/81125
27736         * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27737         by removing enum keyword.
27738         (ubsan_type_descriptor): Likewise.  Formatting fix.
27740         PR target/81121
27741         * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27742         splitter): Require TARGET_SSE2 in the condition.
27744 2017-06-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
27746         PR target/79799
27747         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27748         for doing vector set of SFmode on ISA 3.0.
27749         * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27750         (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27751         element.
27752         (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27753         SFmode value into a V4SF variable that was extracted from another
27754         V4SF variable without converting the element to double precision
27755         and back to single precision vector format.
27756         (vsx_insert_extract_v4sf_p9_2): Likewise.
27758 2017-06-19  Jakub Jelinek  <jakub@redhat.com>
27760         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27761         in UWHI to avoid undefined overflow.
27763         PR sanitizer/81125
27764         * ubsan.h (enum ubsan_encode_value_phase): New.
27765         (ubsan_encode_value): Change second argument to
27766         enum ubsan_encode_value_phase with default value of
27767         UBSAN_ENCODE_VALUE_GENERIC.
27768         * ubsan.c (ubsan_encode_value): Change second argument to
27769         enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27770         adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27771         create_tmp_var_raw instead of create_tmp_var and use a
27772         TARGET_EXPR.
27773         (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27774         instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27775         ubsan_encode_value callers.
27777         PR sanitizer/81111
27778         * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27779         use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27780         just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27782 2017-06-19  Richard Biener  <rguenther@suse.de>
27784         PR middle-end/81118
27785         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27786         estimates if we changed anything.
27788 2017-06-19  Richard Biener  <rguenther@suse.de>
27790         PR tree-optimization/80887
27791         * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27792         (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27793         simplified lookups, then reset mprts_hook.
27794         (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27795         simplifying.
27796         (try_to_simplify): Likewise.
27798 2017-06-19  Martin Liska  <mliska@suse.cz>
27800         PR sanitizer/80879
27801         * gimplify.c (gimplify_switch_expr):
27802         Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27804 2017-06-19  Martin Liska  <mliska@suse.cz>
27806         * doc/install.texi: Document that PGO runs in 4 stages.
27808 2017-06-19  Martin Liska  <mliska@suse.cz>
27810         PR ipa/80732
27811         * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27812         to dispatcher function name.
27813         * multiple_target.c (replace_function_decl): New function.
27814         (create_dispatcher_calls): Redirect both edges and references.
27816 2017-06-19  Jan Hubicka <hubicka@ucw.cz>
27818         * profile-count.c (profile_count::dump): Dump quality.
27819         (profile_count::differs_from_p): Update for unsigned val.
27820         * profile-count.h (profile_count_quality): New enum.
27821         (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27823 2017-06-19  Richard Biener  <rguenther@suse.de>
27825         * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27826         struct function as arg.
27827         (estimate_numbers_of_iterations): Export overload with loop arg.
27828         (free_numbers_of_iterations_estimates_loop): Use an overload of
27829         free_numbers_of_iterations_estimates instead.
27830         * tree-cfg.c (remove_bb): Adjust.
27831         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27832         * tree-parloops.c (gen_parallel_loop): Likewise.
27833         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27834         Likewise.
27835         (tree_unroll_loops_completely): Likewise.
27836         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27837         Use an overload instead and export.
27838         (estimated_loop_iterations): Adjust.
27839         (max_loop_iterations): Likewise.
27840         (likely_max_loop_iterations): Likewise.
27841         (estimate_numbers_of_iterations): Take struct function as arg
27842         and adjust.
27843         (loop_exits_before_overflow): Adjust.
27844         (free_numbers_of_iterations_estimates_loop): Use an overload.
27845         * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27846         * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27848 2017-06-19  Richard Biener  <rguenther@suse.de>
27850         PR ipa/81112
27851         * ipa-prop.c (find_constructor_constant_at_offset): Handle
27852         RANGE_EXPR conservatively.
27854 2017-06-16  Carl Love  <cel@us.ibm.com>
27856         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27857         definitions for vec_float, vec_float2, vec_floato,
27858         vec_floate built-ins.
27859         * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27860         for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27861         floate.
27862         * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27863         FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27864         UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27865         * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27866         define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27867         * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27868         vec_floato): Add builtin defines.
27869         * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27870         Update the built-in documentation file for the new built-in
27871         functions.
27873 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27875         * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27876         (mthumb): Mark as the negative of -marm.
27878 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27880         * doc/invoke.texi (ARM Options, -mcpu): Document supported
27881         extension options.
27882         (ARM Options, -mtune): Document that this accepts the same
27883         extension options as -mcpu.
27884         (ARM Options, -mfpu): Document addition of -mfpu=auto.
27886 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27888         * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27889         permitted extensions.
27891 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27893         * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27894         (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27895         (armv8-m.main): Add option +nodsp.
27896         * config/arm/arm-cpu-cdata.h: Regenerated.
27898 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27900         * config/arm/t-fuchsia: New file.
27901         * config.gcc (arm*-*-fuchsia*): Use it.
27903 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27905         * config/arm/t-symbian: Rewrite for new option infrastructure.
27907 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27909         * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27910         (MULTILIB_REQUIRED): Likewise.
27912 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27914         * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
27915         (MULTILIB_RESUE): Likewise.
27916         (MULTILIB_MATCHES): Likewise.
27917         (MULTLIB_REQUIRED): Likewise.
27919 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27921         * config/arm/t-rtems: Rewrite for new option framework.
27923 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27925         * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
27926         (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
27927         (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
27928         (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
27929         * config/arm/t-multilib: ... here.
27930         (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
27931         (MULTILIB_MATCHES): Use armv7 libraries for armv7-r.  Also use for
27932         armv7-a and armv8*-a when A-profile libraries have not been built.
27933         * config/arm/t-rmprofile: Rewrite.
27935 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27937         * genmultilib (multilib_reuse): Allow an explicit period to be escaped
27938         with a backslash.  Remove the backslash after substituting unescaped
27939         periods.
27940         * doc/fragments.texi (MULTILIB_REUSE): Document it.
27942 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27944         * config.gcc: (arm*-*-*): When building a-profile libraries, force
27945         the driver to pass through the default setting of -mfloat-abi.
27946         * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
27947         rather than NULL.
27948         * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
27949         (all_feat_combs): New rule.
27950         (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories.  Rework
27951         default libraries.
27952         * config/arm/t-aprofile: Rewrite.
27954 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27956         * config/arm/arm.h (FPUTYPE_AUTO): Define.
27957         * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
27958         fpu is not specified by the user/command-line.
27959         * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
27960         * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
27961         * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
27962         * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
27963         * common/config/arm/arm-common.c (arm_canon_arch_option): Use
27964         FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
27966 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27968         * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
27969         * config/arm/t-arm-elf: Rewritten.
27971 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27973         * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
27974         have some floating-point instructions.
27975         (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
27976         (TARGET_MAYBE_HARD_FLOAT): New macro.
27977         * config/arm/arm-builtins.c (arm_init_builtins): Use
27978         TARGET_MAYBE_HARD_FLOAT.
27979         * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
27981 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27983         * common/config/arm/arm-common.c: Define INCLUDE_LIST.
27984         (configargs.h): Include it.
27985         (arm_print_hint_for_fpu_option): New function.
27986         (arm_parse_fpu_option): New function.
27987         (candidate_extension): New class.
27988         (arm_canon_for_multilib): New function.
27989         * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
27990         (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
27991         (ARCH_CANONICAL_SPECS): New macro.
27992         (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
27994 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27996         * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
27997         are set after handling multilib fragments.  Set target_cpu_default2
27998         from with_cpu.
28000 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28002         * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
28003         cpu name.
28004         (arm*-*-*): Set target_cpu_default2 to a quoted string.
28005         * config/arm/parsecpu.awk (check_cpu): Validate any extension
28006         options.
28007         (check_arch): Likewise.
28008         * config/arm/arm.c (arm_configure_build_target): Handle
28009         TARGET_CPU_DEFAULT being a string constant.  Scan any feature
28010         options in the default.
28012 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28014         * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
28015         when an option is an alias of another.
28016         * config/arm/parsecpu.awk (optalias): New parser token.
28017         (gen_comm_data): Mark non-alias options as such.  Emit entries
28018         for extension aliases.
28019         * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
28020         (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
28021         (armv6kz, armv6zk, armv6t2): Likewise.
28022         (armv7): Make vfpv3-d16 an alias.
28023         (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases.  Sort in
28024         canonical order.
28025         (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
28026         Sort in canonical order.
28027         (armv8-a): Sort in canonical order.
28028         (armv8.1-a, armv8.2-a):  Likewise.
28029         (generic-armv7-a): Make neon and neon-vfpv3 aliases.  Sort in
28030         canonical order.
28031         (cortex-a9): Sort in canonical order.
28032         * config/arm/arm.c (selftests.h): Include it.
28033         (arm_test_cpu_arch_data): New function.
28034         (arm_run_self_tests): New function.
28035         (TARGET_RUN_TARGET_SELFTESTS): Redefine.
28036         (targetm): Move declaration to the end of the file.
28037         * arm-cpu-cdata.h: Regenerated.
28039 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28041         * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
28042         call to target_mode_check describing the type of option passed.
28043         * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
28044         (arm_target_thumb_only): Use arm_parse_arch_option_name or
28045         arm_parse_cpu_option_name to match parameters against list of
28046         available targets.
28047         * config/arm/parsecpu.awk (gen_comm_data): Don't generate
28048         arm_arch_core_flags data structure.
28049         * config/arm/arm-cpu_cdata.h: Regenerated.
28051 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28053         * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
28054         config/arm/arm.c.
28055         (arm_print_hint_for_cpu_option): Likewise.
28056         (arm_print_hint_for_arch_option): Likewise.
28057         (arm_parse_cpu_option_name): Likewise.
28058         (arm_parse_arch_option_name): Likewise.
28059         * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
28060         of entries in the all_fpus list.
28061         * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
28062         (arm_parse_cpu_option_name): Declare.
28063         (arm_parse_arch_option_name): Declare.
28064         (arm_parse_option_features): Declare.
28065         (arm_intialize_isa): Declare.
28066         * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
28067         data tables to ...
28068         (gen_comm_data): ... here.  Make definitions non-static.
28069         * config/arm/arm-cpu-data.h: Regenerated.
28070         * config/arm/arm-cpu-cdata.h: Regenerated.
28072 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28074         * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
28075         (cpu_arch_extension): New structure.
28076         (cpu_arch_option, arch_option, cpu_option): New structures.
28077         * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
28078         architecture types.
28079         (gen_data): Generate new format data tables.
28080         * config/arm/arm.c (cpu_tune): New structure.
28081         (cpu_option, processors): Delete.
28082         (arm_print_hint_for_core_or_arch): Delete.  Replace with ...
28083         (arm_print_hint_for_cpu_option): ... this and ...
28084         (arm_print_hint_for_arch_option): ... this.
28085         (arm_parse_arch_cpu_name): Delete.  Replace with ...
28086         (arm_parse_cpu_option_name): ... this and ...
28087         (arm_parse_arch_option_name): ... this.
28088         (arm_unrecognized_feature): Change type of target parameter to
28089         cpu_arch_option.
28090         (arm_parse_arch_cpu_features): Delete.  Replace with ...
28091         (arm_parse_option_features): ... this.
28092         (arm_configure_build_target): Rework to use new configuration data
28093         tables.
28094         (arm_print_tune_info): Rework for new configuration data tables.
28095         * config/arm/arm-cpu-data.h: Regenerated.
28096         * config/arm/arm-cpu.h: Regenerated.
28098 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28100         * Makefile.in (OBJS): Move sbitmap.o from here ...
28101         (OBJS-libcommon): ... to here.
28103 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28105         * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
28106         (ISA_ALL_CRYPTO): New macro.
28107         (ISA_ALL_SIMD): New macro
28108         (ISA_ALL_FP): New macro.
28109         * config/arm/arm.c (fpu_bitlist): Update initializer.
28110         * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
28111         simd or fp.
28112         (arm9e): Add fpu.  Add option for nofp
28113         (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
28114         (arm926ej-s, arm1026ej-s): Likewise.
28115         (generic-armv7-a): Add fpu.  Add options for simd, vfpv3, vfpv3-d16,
28116         vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
28117         neon-fp16, neon-vfpv4, nofp and nosimd.
28118         (cortex-a5, cortex-a7): Add fpu.  Add options for nosimd and nofp.
28119         (cortex-a8): Add fpu.  Add option for nofp.
28120         (cortex-a9): Add fpu.  Add options for nosimd and nofp.
28121         (cortex-a12, cortex-a15, cortex-a17): Add fpu.  Add option for nofp.
28122         (cortex-r4f): Add fpu.
28123         (cortex-r5): Add fpu.  Add options for nofp.dp and nofp.
28124         (cortex-r7): Use idiv option from architecture.  Add fpu.  Add option
28125         for nofp.
28126         (cortex-r8): Likewise.
28127         (cortex-m4): Add fpu.  Add option for nofp.
28128         (cortex-a15.cortex-a7): Add fpu.  Add option for nofp.
28129         (cortex-a17.cortex-a7): Likewise.
28130         (cortex-a32): Add fpu.  Add options for crypto and nofp.
28131         (cortex-a35, cortex-a53): Likewise.
28132         (cortex-a57): Add fpu.  Add option for crypto.
28133         (cortex-a72, cortex-a73): Likewise.
28134         (exynos-m1): Likewise.
28135         (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
28136         (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
28137         (cortex-m33): Add fpu.  Add option for nofp.
28138         * config/arm/arm-cpu-cdata.h: Regenerated
28139         * config/arm/arm-cpu-data.h: Regenerated.
28141 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28143         * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
28144         (armv5te, armv5tej): Likewise.
28145         (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
28146         (armv7): Add options fp and vfpv3-d16.
28147         (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
28148         vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
28149         nofp and nosimd.
28150         (armv7ve): Likewise.
28151         (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
28152         (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
28153         (armv8-a): Add nocrypto option.
28154         (armv8.1-a, armv8.2-a): Likewise.
28155         (armv8-m.main): add options fp, fp.dp and nofp.
28157 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28159         * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
28160         nofp.
28161         (armv8-a+crc): Delete.
28162         (armv8.1-a): Add options simd, crypto and nofp.
28163         (armv8.2-a): Add options fp16, simd, crypto and nofp.
28164         (armv8.2-a+fp16): Delete.
28165         (armv8-m.main): Add option dsp.
28166         (armv8-m.main+dsp): Delete.
28167         (cortex-a8): Add fpu.  Add nofp option.
28168         (cortex-a9): Add fpu.  Add nofp and nosimd options.
28169         * config/arm/parsecpu.awk (gen_data): Generate option tables and
28170         link to main cpu and architecture data structures.
28171         (gen_comm_data): Only put isa attributes from the main architecture
28172         in common tables.
28173         (option): New statement for architecture and CPU entries.
28174         * arm.c (struct cpu_option): New structure.
28175         (struct processors): Add entry for options.
28176         (arm_unrecognized_feature): New function.
28177         (arm_parse_arch_cpu_name): Ignore any characters after the first
28178         '+' character.
28179         (arm_parse_arch_cpu_feature): New function.
28180         (arm_configure_build_target): Separate out any CPU and architecture
28181         features and parse separately.  Don't error out if -mfpu=auto is
28182         used with only an architecture string.
28183         (arm_print_asm_arch_directives): New function.
28184         (arm_file_start): Call it.
28185         * config/arm/arm-cpu-cdata.h: Regenerated.
28186         * config/arm/arm-cpu-data.h: Likewise.
28187         * config/arm/arm-tables.opt: Likewise.
28189 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28191         * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
28192         assembler when it is not -mfpu=auto.
28194 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28196         * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
28197         (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
28198         (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
28199         (ASM_CPU_SPEC): Rewrite.
28200         (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
28201         (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef.  Use
28202         MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS.  Remove
28203         reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
28204         * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
28205         copied string is NUL-terminated.  Also strip any characters prefixed
28206         by '+'.
28207         (arm_rewrite_selected_arch): New function.
28208         (arm_rewrite_march): New function.
28210 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28212         * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
28213         (x_arm_cpu_string, x_arm_tune_string): Likewise.
28214         (march, mcpu, mtune): Convert to string-based options.
28215         * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
28216         (arm_parse_arch_cpu_name): New function.
28217         (arm_configure_build_target): Use arm_parse_arch_cpu_name to
28218         identify selected architecture or CPU.
28219         (arm_option_save): New function.
28220         (TARGET_OPTION_SAVE): Redefine.
28221         (arm_option_restore): Restore string options.
28222         (arm_option_print): Print string options.
28224 2017-06-16  Martin Sebor  <msebor@redhat.com>
28226         PR tree-optimization/80933
28227         PR tree-optimization/80934
28228         * builtins.c (fold_builtin_3): Do not handle bcmp here.
28229         * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
28230         (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
28231         (gimple_fold_builtin): Call them.
28233 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28235         * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
28236         as unlikely; update profile.
28238 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28240         * predict.c (force_edge_cold): Handle declaring edges impossible
28241         more aggresively.
28243 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28245         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28246         profile.
28247         (try_unroll_loop_completely): Fix reporting.
28249 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28251         * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28253 2017-06-16  James Greenhalgh  <james.greenhalgh@arm.com>
28255         PR target/71778
28256         * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28257         if given a non-constant argument for an intrinsic which requires a
28258         constant.
28260 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28262         * profile.c (compare_freqs): New function.
28263         (branch_prob): Sort edge list.
28264         (find_spanning_tree): Assume that the list is priority sorted.
28266 2017-06-16  Richard Biener  <rguenther@suse.de>
28268         PR tree-optimization/81090
28269         * passes.def (pass_record_bounds): Remove.
28270         * tree-pass.h (make_pass_record_bounds): Likewise.
28271         * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28272         make_pass_record_bounds): Likewise.
28273         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28274         not free niter estimates at the beginning but at the end.
28275         * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28277 2017-06-16  Richard Biener  <rguenther@suse.de>
28279         * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28280         initializer to workaround ICE in host GCC 4.8.
28282 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28284         * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28285         counts.
28286         (clone_inlined_nodes): Update.
28288 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28290         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28291         prefetch settings, and enable prefetching by default at -O3.
28293 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28295         * config/aarch64/aarch64.c (aarch64_override_options_internal):
28296         Set flag_prefetch_loop_arrays according to tuning data.
28298 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28300         * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28301         New tune structure.
28302         (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28303         [Unrelated to main purpose of the patch] Place the pointer field last
28304         to enable type checking errors when tune structure are wrongly merged.
28305         * config/aarch64/aarch64.c (generic_prefetch_tune,)
28306         (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28307         (thunderx2t99_prefetch_tune): New tune constants.
28308         (tune_params *_tunings): Update all tunings (no functional change).
28309         (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28310         PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28311         from tunings structures.
28313 2017-06-16  Jakub Jelinek  <jakub@redhat.com>
28315         PR sanitizer/81094
28316         * ubsan.c (instrument_null): Add T argument, use it instead
28317         of computing it based on IS_LHS.
28318         (instrument_object_size): Likewise.
28319         (pass_ubsan::execute): Adjust instrument_null and
28320         instrument_object_size callers to pass gimple_get_lhs or
28321         gimple_assign_rhs1 result to it.  Use instrument_null instead of
28322         calling get_base_address and instrument_mem_ref.  Handle
28323         aggregate call arguments for object-size sanitization.
28325 2017-06-16  Yury Gribov  <tetra2005@gmail.com>
28327         PR tree-optimization/81089
28328         * tree-vrp.c (is_masked_range_test): Validate operands of
28329         subexpression.
28331 2017-06-15  Martin Sebor  <msebor@redhat.com>
28333         PR c++/80560
28334         * dumpfile.c (dump_register): Avoid calling memset to initialize
28335         a class with a default ctor.
28336         * gcc.c (struct compiler): Remove const qualification.
28337         * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28338         * hash-table.h: Ditto.
28339         * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28340           assignment.
28341         * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28342         * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28343         default ctor.
28344         * params.h (struct param_info): Make struct members non-const.
28345         * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28346         with default initialization.
28347         * vec.h (vec_copy_construct, vec_default_construct): New helper
28348         functions.
28349         (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28350         with vec_copy_construct.
28351         (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28352         (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28353         * doc/invoke.texi (-Wclass-memaccess): Document.
28355 2017-06-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28357         * emit-rtl.h (is_leaf): Update comment about local
28358         register allocator.
28360 2017-06-15  Jozef Lawrynowicz   <jozef.l@somniumtech.com>
28362         PR target/78818
28363         * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28364         for a variable to have a section before checking if the section has a
28365         name.
28366         Set section to.persistent if persistent attribute is set.
28367         Warn if .persistent attribute is used on an automatic variable.
28369 2017-06-15  Eric Botcazou  <ebotcazou@adacore.com>
28371         PR rtl-optimization/80474
28372         * reorg.c (update_block): Do not ignore instructions in a delay slot.
28374 2017-06-15  Segher Boessenkool  <segher@kernel.crashing.org>
28376         * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28377         of REGNO.
28379 2017-06-14  Maciej W. Rozycki  <macro@imgtec.com>
28381         * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28382         (casesi): Emit bounds checking as RTL.
28383         (casesi_internal_mips16_<mode>): Remove bounds checking.
28385 2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
28387         * config/xtensa/xtensa.c (xtensa_option_override): Append
28388         MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28389         (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28390          xtensa_doloop_hooks): Define unconditionally.
28391         (xtensa_reorg_loops): Only call reorg_loops in the presence of
28392         TARGET_LOOPS.
28393         * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28394         (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28395         for it in xtensa_option_override.
28396         (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28397          HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28399 2017-06-14  Boris Kolpackov  <boris@codesynthesis.com>
28401         * doc/cppopts.texi: Document '-' special value to -MF.
28403 2017-06-14  Wilco Dijkstra  <wdijkstr@arm.com>
28405         * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28406         (cortex_a53_fconst): Likewise.
28407         (cortex_a53_fpmul): Likewise.
28408         (cortex_a53_f_load_64): Likewise.
28409         (cortex_a53_f_load_many): Likewise.
28410         (cortex_a53_advsimd_alu): Likewise.
28411         (cortex_a53_advsimd_alu_q): Likewise.
28412         (cortex_a53_advsimd_mul): Likewise.
28413         (cortex_a53_advsimd_mul_q): Likewise.
28414         (fpmac bypass): Add new bypass for fpmac-fpmac case.
28415         Add missing fmul, r2f_cvt and fconst cases.
28417 2017-06-14  Richard Biener  <rguenther@suse.de>
28419         PR middle-end/81088
28420         * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28421         literal constants.
28422         (fold_binary_loc): When associating do not treat pre-existing
28423         TREE_OVERFLOW on literal constants as a reason to allow
28424         TREE_OVERFLOW on associated literal constants.
28426 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28428         * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28429         (MASK_FEATURES): New macro.
28430         * config/sparc/sparc.c (sparc_option_override): Remove the special
28431         handling of -mfpu and generalize it to all MASK_FEATURES switches.
28433 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28435         * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28436         a division of 0 if non-call exceptions are enabled.
28438 2017-06-14  Andrew Pinski  <apinski@cavium.com>
28439             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
28441         PR target/71663
28442         * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28443         Improve vector initialization code gen for only variable case.
28445 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28447         * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28449 2017-06-14  Richard Biener  <rguenther@suse.de>
28451         PR tree-optimization/81083
28452         * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28453         as values.
28455 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28457         * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28458         (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28459         * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28460         * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete.  Adjust former use.
28461         * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28462         * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28464 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28466         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28467         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28469 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28471         * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28473 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28475         * config/rs6000/t-rtems: Don't handle SPE.
28477 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28479         * config/rs6000/t-linux: Don't handle SPE.
28481 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28483         * config/rs6000/eabispe.h: Delete file.
28485 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28487         * config/rs6000/t-spe: Delete file.
28489 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28491         * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28492         (rs6000_legitimate_offset_address_p): Return false for anything in
28493         V2SImode or V2SFmode.
28495 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28497         * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28498         except V2SF and V2SI.  Rearrange the vector modes, and add comments.
28499         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28500         and V4HImode.
28501         (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28502         (rs6000_legitimate_offset_address_p): Ditto.
28503         (rs6000_emit_move): Ditto.
28504         (rs6000_init_builtins): Remove V4HI_type_node.
28506 2017-06-13  Martin Liska  <mliska@suse.cz>
28508         PR sanitize/78204
28509         * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28510         (gate_asan): Likewise.
28511         * asan.h (asan_no_sanitize_address_p): Remove the function.
28512         (sanitize_flags_p): New function.
28513         * builtins.def: Fix coding style.
28514         * common.opt: Use renamed enum value.
28515         * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28516         * doc/extend.texi: Document no_sanitize attribute.
28517         * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28518         to SANITIZE_UNDEFINED_NONDEFAULT.
28519         * gcc.c (sanitize_spec_function): Use the renamed enum value.
28520         * gimple-fold.c (optimize_atomic_compare_exchange_p):
28521         Use sanitize_flags_p.
28522         * gimplify.c (gimplify_function_tree): Likewise.
28523         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28524         * opts.c (parse_no_sanitize_attribute): New function.
28525         (common_handle_option): Use renamed enum value.
28526         * opts.h (parse_no_sanitize_attribute): Declare.
28527         * tree.c (sanitize_flags_p): New function.
28528         * tree.h: Declared here.
28529         * tsan.c: Use sanitize_flags_p.
28530         * ubsan.c (ubsan_expand_null_ifn): Likewise.
28531         (instrument_mem_ref): Likewise.
28532         (instrument_bool_enum_load): Likewise.
28533         (do_ubsan_in_current_function): Remove the function.
28534         (pass_ubsan::execute): Use sanitize_flags_p.
28535         * ubsan.h: Remove do_ubsan_in_current_function
28536         * tree-cfg.c (print_no_sanitize_attr_value): New function.
28537         (dump_function_to_file): Use it here.
28539 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28541         PR tree-optimization/80803
28542         PR tree-optimization/81063
28543         * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28544         (propagate_subaccesses_across_link): Enqueue subtree whenever
28545         necessary instead of relying on the caller.
28547 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28549         * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28550         that have a first_link.
28551         (sort_and_splice_var_accesses): Do not check first_link before
28552         enquing.
28553         (subtree_mark_written_and_enqueue): Likewise.
28554         (propagate_all_subaccesses): Likewise and do not stop at first
28555         parent with a first_link.
28557 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28559         * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28560         instead of f.
28562 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28564         * match.pd: New pattern.
28566 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28568         * tree-vrp.c (is_masked_range_test): New function.
28569         (register_edge_assert_for): Determine ranges for
28570         some bit tests.
28572 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28574         PR tree-optimization/67328
28575         * fold-const.c (maskable_range_p): New function.
28576         (build_range_check): Generate bittests if possible.
28578 2017-06-13  Martin Liska  <mliska@suse.cz>
28580         * gimple-pretty-print.c (dump_probability): Add new argument.
28581         (dump_edge_probability): Dump both probability and count.
28582         (dump_gimple_label): Likewise.
28583         (dump_gimple_bb_header): Likewise.
28585 2017-06-13  Georg-Johann Lay  <avr@gjlay.de>
28587         PR target/81072
28588         * config/avr/avr-devices.c: Fix indentation.
28589         * config/avr/gen-avr-mmcu-specs.c: Dito.
28591 2017-06-13  Richard Biener  <rguenther@suse.de>
28593         * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28594         instead get vector type from stmt_info.
28595         (vectorizable_reduction): Adjust.  Remove dead code.
28597 2017-06-13  Richard Biener  <rguenther@suse.de>
28599         PR middle-end/81065
28600         * fold-const.c (extract_muldiv_1): Remove bogus distribution
28601         case of C * (x * C2 + C3).
28602         (fold_addr_of_array_ref_difference): Properly fold index difference.
28604 2017-06-12  David S. Miller  <davem@davemloft.net>
28606         PR target/80968
28607         * config/sparc/sparc.md (return expander): Emit frame blockage if
28608         function uses alloca.
28610 2017-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
28612         * combine.c (make_field_assignment): Check len rather than the mode
28613         precision when calling force_to_mode.
28615 2017-06-12  Georg-Johann Lay  <avr@gjlay.de>
28617         Support multilibs and devices that see flash in RAM address range.
28619         PR target/81072
28620         * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28621         (avr_mcu_t) <flash_pm_offset>: New field.
28622         (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28623         * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28624         (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28625         (AVR_TINY_PM_OFFSET): Remove macro.
28626         * config/avr/avr.opt (-mshort-calls): New option.
28627         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28628         [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28629         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28630         <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28631         <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28632         instead of avr_arch->have_jmp_call.
28633         <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28634         [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28635         avr_arch->flash_pm_offset to define.
28636         * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28637         new field flash_pm_offset.  Add entry for avrxmega3.
28638         (avr_texinfo): Add entry for avrxmega3.
28639         * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28640         attiny212, attiny214,
28641         attiny412, attiny414, attiny416, attiny417,
28642         attiny814, attiny816, attiny817,
28643         attiny1614, attiny1616, attiny1617,
28644         attiny3214, attiny3216, attiny3217.
28645         * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28646         avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28647         (avr_print_operand_address) [AVR_TINY]: Same.
28648         (avr_asm_init_sections) <readonly_data_section>: Only patch
28649         callback if avr_arch->flash_pm_offset = 0.
28650         (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28651         for rodata if avr_arch->flash_pm_offset != 0.
28652         (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28653         * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28654         (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28655         (m_options): Append opt_rcall.
28656         (m_dirnames): Append dir_rcall.
28657         * config/avr/t-multilib: Regenerate.
28659         * configure.ac [target=avr]: Check whether avrxmega3 default
28660         linker description file works as needed.
28661         * configure: Regenerate.
28662         * doc/avr-mmcu.texi: Regenerate.
28663         * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28664         <__AVR_ARCH__>: Document avrxmega3 and 103.
28665         <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28666         <__AVR_SHORT_CALLS__>: Document it.
28667         <__AVR_PM_BASE_ADDRESS__>: Document it.
28668         * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28669         (AVR Variable Attributes) <progmem>: Document this is
28670         not needed for avrxmega3.
28671         (AVR Named Address Spaces) <__flash>: Dito.
28673 2017-06-12  Jan Hubicka  <hubicka@ucw.cz>
28675         * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28677 2017-06-12  Doug Rupp  <rupp@adacore.com>
28679         * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28680         Append vxworks-stdint.h to the tm_file list.
28681         * config/vxworks-stdint.h: New file.
28683 2017-06-12  Martin Liska  <mliska@suse.cz>
28685         PR tree-optimization/81041
28686         * tree-profile.c (gimple_gen_ic_func_profiler):
28687         Create an extra BB in profile-generate
28688         (gimple_gen_time_profiler): Likewise.
28690 2017-06-12  Jakub Jelinek  <jakub@redhat.com>
28692         PR tree-optimization/81003
28693         * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28694         (update_range_test): Use it instead of force_gimple_operand_gsi.
28696 2017-06-12  Richard Biener  <rguenther@suse.de>
28698         PR tree-optimization/81053
28699         * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28700         with backedge value not defined in loop.  Simplify def stmt
28701         compute.
28703 2017-06-11  Tom de Vries  <tom@codesourcery.com>
28705         PR target/79939
28706         * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28707         Return true.
28708         (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28709         nvptx_cannot_force_const_mem.
28711 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28713         * opts.c (finish_options): Move test for flag_split_stack after
28714         it has been initialized.
28716 2017-06-11  Jason Merrill  <jason@redhat.com>
28718         * tree.h (id_equal): New.
28719         * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28720         omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28721         instead of strcmp of IDENTIFIER_POINTER.
28723 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28725         * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28726         (mark_all_inlined_calls_cdtor): Fix formating.
28727         (inline_transform): Rescale profile before inlining.
28729 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28731         * cgraph.h (cgraph_edge::clone): Update prototype.
28732         * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28733         (cgraph_node::create_clone): Update.
28734         (cgraph_node::create_version_clone): Update.
28735         * tree-inline.c (copy_bb): Update.
28736         (expand_call_inline): Update.
28738 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28740         * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28741         factored out from ...
28742         (rs6000_emit_prologue): ... here.
28744 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28746         * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28747         factored out from ...
28748         (rs6000_emit_prologue): ... here.
28750 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28752         * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28753         edge counts.
28754         (handle_missing_profiles): Fix computation of tp_first_run.
28755         (counts_to_freqs): Do not touch freqs when count is 0.
28757 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28759         * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28760         profile.
28762 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28764         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28765         attributes): Document signal effective target.
28767 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28769         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28770         Document effective target stack_size.
28772 2017-06-09  David Malcolm  <dmalcolm@redhat.com>
28774         * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28775         to the edit_context if they can be auto-applied.
28777 2017-06-9  Ian Lance Taylor  <iant@golang.org>
28779         * opts.c (finish_options): If -fsplit-stack, disable implicit
28780         -forder-blocks-and-partition.
28781         * doc/invoke.texi (Optimize Options): Document that when using
28782         -fsplit-stack -forder-blocks-and-partition is not implicitly
28783         enabled.
28785 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28787         * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28788         ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28789         ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28790         * builtins.def (abort, trap, unreachable): Declare cold.
28791         * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28792         * tree-core.h (ECF_COLD): New.
28793         * tree.c (set_call_expr_flags): Handle ECF_COLD.
28794         (build_common_builtin_nodes): Mark unreachable and abort as cold.
28796 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28798         * predict.c (unlikely_executed_stmt_p): Cleanup.
28800 2017-06-09  Richard Biener  <rguenther@suse.de>
28802         * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28803         model if the ref is always written to.
28805 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28807         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28809 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28811         * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28812         than udiv.
28814 2017-06-09  Tom de Vries  <tom@codesourcery.com>
28816         PR target/80855
28817         * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28818         "target cannot support label values" when encountering LABEL_REF.
28820 2017-06-09  Martin Liska  <mliska@suse.cz>
28822         * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28823         (gimple_gen_ic_func_profiler): Emit direct comparison
28824         of __gcov_indirect_call_callee with NULL.
28825         (gimple_gen_time_profiler): Change probability from
28826         PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28828 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28830         * profile.c (edge_gcov_counts): Turn to pointer.
28831         (compute_branch_probabilities, compute_branch_probabilities): Update.
28832         (branch_prob): Do not clear edge_gcov_count.
28833         * profile.h (edge_gcov_counts): Turn to pointer.
28834         (edge_gcov_count): Update.
28836 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28838         * gimple.h (gimple_check_failed): Mark cold.
28840 2017-06-09  Richard Biener  <rguenther@suse.de>
28842         PR tree-optimization/66623
28843         * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28844         refactor check_reduction into two parts, properly computing
28845         whether we have to check reduction validity for outer loop
28846         vectorization.
28848 2017-06-09  Richard Biener  <rguenther@suse.de>
28850         PR tree-optimization/79483
28851         * graphite-scop-detection.c (order): New global.
28852         (get_order): Compute bb to order mapping that satisfies code
28853         generation constraints.
28854         (cmp_pbbs): New helper.
28855         (build_scops): Start domwalk at entry block, sort generated
28856         pbbs.
28858 2017-06-09  Richard Biener  <rguenther@suse.de>
28860         PR middle-end/81007
28861         * ipa-polymorphic-call.c
28862         (ipa_polymorphic_call_context::restrict_to_inner_class):
28863         Skip FIELD_DECLs with error_mark_node type.
28864         * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28865         last again.
28867 2017-06-09  Martin Liska  <mliska@suse.cz>
28869         * predict.c (struct branch_predictor): New struct.
28870         (test_prediction_value_range): New test.
28871         (predict_c_tests): New function.
28872         * selftest-run-tests.c (selftest::run_tests): Run the function.
28873         * selftest.h: Declare new tests.
28875 2017-06-09  Segher Boessenkool  <segher@kernel.crashing.org>
28877         PR target/80966
28878         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28879         gen_add3_insn did not fail.
28880         * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28881         r0, construct that number in a temporary reg and add that reg to r0.
28882         If asked to put the result in r0 as well, fail.
28884 2017-06-08  Will Schmidt  <will_schmidt@vnet.ibm.com>
28886         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28887         for early expansion of vec_eqv.
28889 2017-06-08  Jakub Jelinek  <jakub@redhat.com>
28891         PR middle-end/81005
28892         * ubsan.c (instrument_null): Avoid pointless code temporary.
28893         (pass_ubsan::execute): Instrument aggregate arguments of calls.
28895 2017-06-08  Uros Bizjak  <ubizjak@gmail.com>
28897         PR target/81015
28898         Revert:
28899         2016-12-14  Uros Bizjak  <ubizjak@gmail.com>
28901         PR target/59874
28902         * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28903         (*clzhi2): Ditto.
28905 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28907         * predict.c (unlikely_executed_edge_p): Move ahead.
28908         (probably_never_executed_edge_p): Use it.
28910 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28912         PR middle-end/79988
28913         * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
28914         gimple_call_builtin_p call.
28916 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28918         * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
28919         * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
28920         rtl_check_failed_type2, rtl_check_failed_code1,
28921         rtl_check_failed_code2, rtl_check_failed_code_mode,
28922         rtl_check_failed_block_symbol, cwi_check_failed_bounds,
28923         rtvec_check_failed_bounds, rtl_check_failed_flag,
28924         _fatal_insn_not_found, _fatal_insn): Likewise.
28925         * tree.h (tree_contains_struct_check_failed,
28926         tree_check_failed, tree_not_check_failed,
28927         tree_class_check_failed, tree_range_check_failed,
28928         tree_not_class_check_failed, tree_int_cst_elt_check_failed,
28929         tree_vec_elt_check_failed, phi_node_elt_check_failed,
28930         tree_operand_check_failed, omp_clause_check_failed,
28931         omp_clause_operand_check_failed, omp_clause_range_check_failed):
28932         Likewise.
28934 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28936         * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
28937         flag_branch_probabilities.
28938         * ipa-inline.c (edge_badness): Likewise.
28939         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
28940         * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
28941         * predict.c (maybe_hot_frequency_p): Likewise.
28942         (probably_never_executed): Likewise.
28943         * sched-ebb.c (schedule_ebbs): Likewise.
28944         * sched-rgn.c (find_single_block_region): Likewise.
28945         * tracer.c (tail_duplicate): Likewise.
28947 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28949         * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
28950         longer requires x_flag_profile_use.
28952 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28954         * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
28955         instead of flag_reorder_blocks_and_partition.
28956         * dbxout.c (dbxout_function_end): Likewise.
28957         * dwarf2out.c (gen_subprogram_die): Likewise.
28958         * haifa-sched.c (sched_create_recovery_edges): Likewise.
28959         * hw-doloop.c (reorg_loops): Likewise.
28960         * varasm.c (assemble_start_function,
28961         assemble_end_function): Likewise.
28962         (decide_function_section): Do not check for
28963         flag_reorder_blocks_and_partition.
28965 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28967         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
28968         New function.
28969         (chkp_get_hard_register_fake_addr_expr): Ditto.
28970         (chkp_build_addr_expr): Add check for hard reg case.
28971         (chkp_parse_array_and_component_ref): Ditto.
28972         (chkp_find_bounds_1): Ditto.
28973         (chkp_process_stmt): Don't generate bounds store for
28974         hard reg case.
28976 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28978         * predict.c (maybe_hot_bb_p): Do not check profile status.
28979         (maybe_hot_edge_p): Likewise.
28980         (probably_never_executed): Check for zero counts even if profile
28981         is not read.
28982         (unlikely_executed_edge_p): New function.
28983         (unlikely_executed_stmt_p): New function.
28984         (unlikely_executed_bb_p): New function.
28985         (set_even_probabilities): Use unlikely predicates.
28986         (combine_predictions_for_bb): Likewise.
28987         (predict_paths_for_bb): Likewise.
28988         (predict_paths_leading_to_edge): Likewise.
28989         (determine_unlikely_bbs): New function.
28990         (estimate_bb_frequencies): Use it.
28991         (compute_function_frequency): Use zero counts even if profile is
28992         not read.
28993         * profile-count.h: Fix typo.
28995 2017-08-08  Julia Koval  <julia.koval@intel.com>
28997         * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
28998         _mm512_mask_cvtsepi16_storeu_epi8,
28999         _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
29000         * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
29001         _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
29002         _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
29003         _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
29004         * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
29005         (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
29006         VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
29007         * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
29008         __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
29009         __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
29010         __builtin_ia32_pmovuswb256mem_mask,
29011         __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
29012         __builtin_ia32_pmovwb512mem_mask): New builtins.
29014 2017-08-08  Julia Koval  <julia.koval@intel.com>
29016         PR target/73350,80862
29017         * config/i386/subst.md (round): Fix round pattern.
29018         * config/i386/i386.c (ix86_erase_embedded_rounding):
29019         Fix erasing rounding for the fixed pattern.
29021 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29023         * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
29025 2017-06-08  Martin Liska  <mliska@suse.cz>
29027         PR gcov-profile/80911
29028         * gcov.c (block_info::block_info): New constructor.
29030 2017-06-07  Carl Love  <cel@us.ibm.com>
29032         * config/rs6000/rs6000-c: The return type of the following
29033         built-in functions was implemented as int not long long.  Fix sign
29034         of return value for the unsigned version of vec_mulo and vec_mule.
29035         vector unsigned long long vec_bperm (vector unsigned long long,
29036                                              vector unsigned char)
29037         vector signed long long vec_mule (vector signed int,
29038                                           vector signed int)
29039         vector unsigned long long vec_mule (vector unsigned int,
29040                                             vector unsigned int)
29041         vector signed long long vec_mulo (vector signed int,
29042                                           vector signed int)
29043         vector unsigned long long vec_mulo (vector unsigned int,
29044                                             vector unsigned int)
29045         * doc/extend.texi: Fix the documentation for the built-in
29046         functions.
29048 2017-06-07  Carl Love  <cel@us.ibm.com>
29050         PR target/80982
29051         * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
29052         for BE.
29054 2017-06-07  Carl Love  <cel@us.ibm.com>
29056         * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
29057         support, Generate       doublehv for signed int/float for BE case only.
29059 2017-06-07  Alexander Monakov  <amonakov@ispras.ru>
29061         * doc/invoke.texi (mcx16): Rewrite.
29063 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29065         * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
29066         * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
29067         *mov<mode>_softfloat, and an anonymous splitter): Use
29068         nonimmediate_operand instead of rs6000_nonimmediate_operand.
29070 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29072         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
29073         SPEFSCR registers.
29074         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29075         (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
29076         (rs6000_debug_reg_global): Adjust.
29077         (rs6000_init_hard_regno_mode_ok): Adjust.
29078         (rs6000_dbx_register_number): Adjust.
29079         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
29080         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29081         Remove SPE_ACC and SPEFSCR.
29082         (REG_ALLOC_ORDER): Ditto.
29083         (FRAME_POINTER_REGNUM): Change to 111.
29084         (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
29085         (REG_CLASS_NAMES): Ditto.
29086         (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
29087         (REGISTER_NAMES): Ditto.
29088         (ADDITIONAL_REG_NAMES): Ditto.
29089         (rs6000_reg_names): Ditto.
29090         * config/rs6000/rs6000.md: Renumber some register number
29091         define_constants.
29093 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29095         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
29096         registers.
29097         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29098         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
29099         to 117.
29100         (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
29101         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29102         Delete the SPE high registers.
29103         (REG_ALLOC_ORDER): Ditto.
29104         (enum reg_class): Remove SPE_HIGH_REGS.
29105         (REG_CLASS_NAMES): Ditto.
29106         (REG_CLASS_CONTENTS): Delete the SPE high registers.
29107         (REGISTER_NAMES): Ditto.
29108         (rs6000_reg_names): Ditto.
29109         * doc/tm.texi.in: Remove SPE as example.
29110         * doc/tm.texi: Regenerate.
29112 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29114         * config/rs6000/8540.md (ppc8540_brinc): Delete.
29115         * config/rs6000/e500mc.md (e500mc_brinc): Delete.
29116         * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
29117         * config/rs6000/rs6000.md (type): Remove "brinc".
29119 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29121         * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
29122         (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
29123         * config/rs6000/linuxspe.h: Delete file.
29124         * config/rs6000/rs6000.md: Don't include spe.md.
29125         * config/rs6000/spe.h: Delete file.
29126         * config/rs6000/spe.md: Delete file.
29127         * config/rs6000/t-rs6000: Remove spe.md.
29129 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29131         * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
29132         (reg_or_none500mem_operand): Delete.
29133         * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
29134         instead of reg_or_none500mem_operand.
29136 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29138         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29139         handling of SPE flags.
29140         * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
29142 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29144         * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
29145         SPE ABI handling.
29146         * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
29147         (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
29148         paired_divv2sf3): Similar.
29149         * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
29150         SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
29151         * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
29152         RS6000_BUILTIN_S.
29153         Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
29154         Rename the paired_* instruction patterns.
29155         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
29156         define __SPE__.
29157         * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
29158         * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
29159         (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
29160         spe_padding_size, and spe_64bit_regs_used.  Replace TARGET_SPE and
29161         TARGET_SPE_ABI with 0, simplify.  Replace SPE_VECTOR_MODE with
29162         PAIRED_VECTOR_MODE.
29163         (struct machine_function): Delete field spe_insn_chain_scanned_p.
29164         (spe_func_has_64bit_regs_p): Delete.
29165         (spe_expand_predicate_builtin): Delete.
29166         (spe_expand_evsel_builtin): Delete.
29167         (TARGET_DWARF_REGISTER_SPAN): Do not define.
29168         (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
29169         (invalid_e500_subreg): Delete.
29170         (rs6000_legitimize_address): Always force_reg op2 as well, for
29171         paired single memory accesses.
29172         (rs6000_member_type_forces_blk): Delete.
29173         (rs6000_spe_function_arg): Delete.
29174         (rs6000_expand_unop_builtin): Delete SPE handling.
29175         (rs6000_expand_binop_builtin): Ditto.
29176         (spe_expand_stv_builtin): Delete.
29177         (bdesc_2arg_spe): Delete.
29178         (spe_expand_builtin): Delete.
29179         (spe_expand_predicate_builtin): Delete.
29180         (spe_expand_evsel_builtin): Delete.
29181         (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
29182         (spe_init_builtins): Delete.
29183         (spe_func_has_64bit_regs_p): Delete.
29184         (savres_routine_name): Delete "info" parameter.  Adjust callers.
29185         (rs6000_emit_stack_reset): Ditto.
29186         (rs6000_dwarf_register_span): Delete.
29187         * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
29188         UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
29189         SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
29190         Delete.
29191         * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
29192         Delete.
29193         * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
29194         * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
29195         * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
29196         mulv2sf3, divv2sf3): Delete expanders.
29198 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29200         config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
29202 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29204         * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
29205         * config/rs6000/rs6000.c: Ditto.
29207 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29209         * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
29210         * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
29211         comparison_operator.
29213 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29215         * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
29216         * config/rs6000/rs6000.opt: Ditto.
29217         * config/rs6000/t-rtems: Ditto.
29219 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29221         * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
29222         TARGET_E500_SINGLE by 0, simplify.
29223         * config/rs6000/rs6000.c: Ditto.
29224         (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
29225         (spe_build_register_parallel): Delete.
29226         * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
29227         TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
29228         * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
29229         TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
29230         (E500_CONVERT): Delete.
29231         * config/rs6000/spe.md: Remove many patterns and all define_constants.
29233 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29235         * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
29236         * config/rs6000/dfp.md: Ditto.
29237         (negdd2, *negdd2_fpr): Merge.
29238         (absdd2, *absdd2_fpr): Merge.
29239         (negtd2, *negtd2_fpr): Merge.
29240         (abstd2, *abstd2_fpr): Merge.
29241         * config/rs6000/e500.h: Delete file.
29242         * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29243         TARGET_FPRS by 1 and simplify.
29244         * config/rs6000/rs6000-c.c: Ditto.
29245         * config/rs6000/rs6000.c: Ditto.  Also replace TARGET_SF_SPE and
29246         TARGET_DF_SPE by 0.
29247         * config/rs6000/rs6000.h: Ditto.  Delete TARGET_SF_SPE and
29248         TARGET_DF_SPE.
29249         * config/rs6000/rs6000.md: Ditto.
29250         (floatdidf2, *floatdidf2_fpr): Merge.
29251         (move_from_CR_gt_bit): Delete.
29252         * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29253         (E500_CR_IOR_COMPARE): Delete.
29254         (All patterns that require !TARGET_FPRS): Delete.
29255         * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29257 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29259         * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29261 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29263         * graphds.c (add_edge): Intitialize edge's attached data.
29264         (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29265         pointer parameter.  Call pointed function on each edge during
29266         graph traversing.  Skip traversing the edge when the function
29267         returns true.
29268         (graphds_dfs, graphds_scc): Ditto.
29269         (for_each_edge): New parameter.  Pass the new parameter to callback
29270         function.
29271         * graphds.h (skip_edge_callback): New function pointer type.
29272         (graphds_dfs, graphds_scc): New function pointer parameter.
29273         (graphds_edge_callback, for_each_edge): New parameter.
29275 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29277         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29278         out code checking if runtime alias check is possible to below ...
29279         Call the new function.
29280         * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29281         * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29283 2017-06-07  Marek Polacek  <polacek@redhat.com>
29285         PR sanitizer/80932
29286         * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29287         TYPE_OVERFLOW_WRAPS check.
29289 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29291         * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29292         if versioning is required.
29293         * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29294         peeling with the check for versioning.
29296 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29298         * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29299         * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29300         Set true to new parameter if new ssa variable is defined.
29301         (vect_gen_vector_loop_niters): Refactor.  Set range information
29302         for the new vector loop bound variable.
29303         (vect_do_peeling): Ditto.
29305 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29307         * tree-affine.c (ssa.h): Include header file.
29308         (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29309         has wrapping overflow behavior.
29311 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29313         * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29315 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29317         (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29318         (tree_to_aff_combination): ... here.
29320 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29322         * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29323         reg_pressure model function.
29324         (ivopts_global_cost_for_size): Delete.
29325         (determine_set_costs, iv_ca_recount_cost): Call new model function
29326         ivopts_estimate_reg_pressure.
29328 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29330         * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29331         expensive than udiv.  Remove floating point cases from mod.
29333 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29335         * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29336         Increase idiv cost.
29338 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29340         * config/aarch64/aarch64.md
29341         (copysignsf3): Fix mask generation.
29343 2017-06-07  Jakub Jelinek  <jakub@redhat.com>
29345         * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29346         TDI_gimple.
29347         (class dump_manager): Add register_dumps method.
29348         * dumpfile.c: Include langhooks.h.
29349         (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29350         (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29351         (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29352         (dump_manager::dump_register): Start with 512 entries instead of 32.
29353         (dump_manager::register_dumps): New method.
29354         * toplev.c (general_init): Instead of invoking register_dumps
29355         langhook, invoke register_dumps method on the dump manager.
29356         * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29357         TDI_generic.
29359 2017-06-07  Richard Sandiford  <richard.sandiford@linaro.org>
29361         * doc/md.texi: Clarify the restrictions on a define_insn condition.
29362         Say that # requires an associated define_split to exist, and that
29363         the define_split must be suitable for use after register allocation.
29365 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29367         * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29368         (compute_outgoing_frequencies): Also initialize zero counts.
29369         (find_many_sub_basic_blocks): Do not produce uninitialized profile
29370         around loops; preserve more of profile when nothing changes.
29372 2017-06-06  Jim Wilson  <jim.wilson@linaro.org>
29374         * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29375         here.
29376         * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29377         * config/arm/arm-cpu-cdata.h: Regenerate.
29378         * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29379         * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29380         * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29381         * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29382         * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29383         support.
29384         * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29385         support.
29386         * config/arm/t-rmprofile: Likewise.
29387         * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29389 2017-06-06  David S. Miller  <davem@davemloft.net>
29391         PR target/80968
29392         * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29393         blockage if function uses alloca.
29395 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29397         * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29398         New "uid" fields to hold pretty-print IDs of group and ref.
29399         Memory references are now identified as <group_id>:<ref_id>
29400         instead of using [random] addresses.
29401         (dump_mem_details): Simplify, no functional change.
29402         (dump_mem_ref): Simplify and make output more concise.
29403         Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29404         (find_or_create_group): Initialize group uid.
29405         (record_ref): Initialize ref uid.  Improve debug output.
29406         (prune_group_by_reuse, should_issue_prefetch_p,)
29407         (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29408         (mark_nontemporal_store, determine_loop_nest_reuse):
29409         Improve debug output.
29411 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29413         * dbgcnt.def (prefetch): New debug counter.
29414         * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29415         (schedule_prefetches): Stop issueing prefetches if debug counter
29416         tripped.
29418 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29420         * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29421         gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29423 2017-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29425         * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29426         Use aarch64_reg_or_zero predicate for operand 4.
29427         (aarch64_compare_and_swap<mode> define_insn_and_split):
29428         Use aarch64_reg_or_zero predicate for operand 3.  Add 'Z' constraint.
29429         (aarch64_store_exclusive<mode>): Likewise for operand 2.
29431 2017-06-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29433         * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29434         (arm_compute_save_core_reg_mask): This.
29435         (thumb1_compute_save_reg_mask): Rename into ...
29436         (thumb1_compute_save_core_reg_mask): This.
29437         (arm_compute_save_reg0_reg12_mask): Adapt comment.
29438         (arm_compute_frame_layout): Likewise.
29440 2017-06-06  Richard Biener  <rguenther@suse.de>
29442         PR tree-optimization/80974
29443         * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29444         keep or clear leaders SSA info.
29446 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29448         * config/nvptx/nvptx.c (split_mode_p): New function.
29449         (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29451 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29453         * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29455 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29457         PR bootstrap/80978
29458         * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29459         profile.
29461 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29463         * shrink-wrap.c (handle_simple_exit): Update profile.
29464         (try_shrink_wrapping): Upate profile.
29466 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29468         * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29469         (tree_guess_outgoing_edge_probabilities): New.
29470         * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29471         * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29473 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29475         * ipa-split.c (split_function): Initialize return bb profile.
29477 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29479         * profile.c (compute_branch_probabilities): Also initialize
29480         EXIT_BLOCK profile.
29482 2017-06-06  Richard Biener  <rguenther@suse.de>
29484         PR tree-optimization/80928
29485         * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29486         (vect_analyze_loop_operations): Properly guard analysis for
29487         pure SLP case.
29488         (vect_transform_loop): Likewise.
29489         (vect_analyze_loop_2): Also reset SLP type on PHIs.
29490         (vect_model_induction_cost): Do not cost for pure SLP.
29491         (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29492         of induction in inner loop vectorization.
29493         * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29494         (vect_get_and_check_slp_defs): Handle vect_induction_def.
29495         (vect_build_slp_tree): Likewise.  Handle PHIs as terminating the
29496         recursion.
29497         (vect_analyze_slp_cost_1): Cost induction.
29498         (vect_detect_hybrid_slp_stmts): Handle PHIs.
29499         (vect_get_slp_vect_defs): Likewise.
29500         * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29501         (vect_transform_stmt): Handle SLP reductions.
29502         * tree-vectorizer.h (vectorizable_induction): Adjust.
29504 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29506         * config/rs6000/rs6000.c (make_resolver_func): Update
29507         init_lowered_empty_function call.
29509 2017-06-05  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29511         * doc/invoke.texi: Document the -fprofile-abs-path option.
29512         * common.opt (fprofile-abs-path): New option.
29513         * gcov-io.h (gcov_write_filename): Declare.
29514         * gcov-io.c (gcov_write_filename): New function.
29515         * coverage.c (coverage_begin_function): Use gcov_write_filename.
29516         * profile.c (output_location): Likewise.
29518 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29520         * shring-wrap.c: Revert accidental commit.
29522 2017-06-05  Volker Reichelt  <v.reichelt@netcologne.de>
29524         * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29526 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29528         * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29529         new edge.
29530         * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29531         profile in callgraph edge.
29532         * profile-count.h (apply_probability): If THIS is 0, then result is 0
29533         (apply_scale): Likewise.
29534         * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29535         Also scale profile when inlining function with zero profile.
29536         (initialize_cfun): Update exit block profile even when it is zero.
29537         * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29538         when profile is read.
29540 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29542         * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29543         (CLONE_*): New constants to define the processors we can generate
29544         code for with the target_clone attribute.
29545         (rs6000_clone_map): New array to identify which clone processors
29546         the current program is running on.
29547         (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29548         target_clone attribute.
29549         (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29550         (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29551         (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29552         (cpu_expand_builtin): Add support for target_clone attribute.
29553         (rs6000_valid_attribute_p): Allow "default" attribute.
29554         (get_decl_name): New debug function to simplify printing the
29555         current function name in debugging statements.
29556         (rs6000_clone_priority): New functions to support the target_clone
29557         attribute, and be able to generate code to switch between ISA 2.05
29558         through ISA 3.0 (power6 through power9).
29559         (rs6000_compare_version_priority): Likewise.
29560         (rs6000_get_function_versions_dispatcher): Likewise.
29561         (make_resolver_func): Likewise.
29562         (add_condition_to_bb): Likewise.
29563         (dispatch_function_versions): Likewise.
29564         (rs6000_generate_version_dispatcher_body): Likewise.
29565         (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29566         (fusion_gpr_load_p): Fix a spacing issue.
29567         * doc/extend.texi (Common Function Attributes): Document that the
29568         PowerPC supports the target_clone attribute.
29570 2017-06-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29572         * config/arm/arm.h: explain F symbol found in description of ARM
29573         register allocation in its legend.
29575 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29577         * config/mips/frame-header-opt.c: Include profile-count.h.
29578         * config/riscv/riscv.c: Include profile-count.h
29580 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29582         * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29583         update profile.
29584         (sm_set_flag_if_changed): Add bbs field.
29585         (execute_sm_if_changed_flag_set): Pass BBS.
29586         (execute_sm): Update.
29588 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29590         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29591         New pattern.
29593 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29595         * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29596         (peephole2): New peephole2 to emit the above.
29597         * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29599 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29601         * config/aarch64/aarch64.c (define_peephole2 above
29602         *sub_<shift>_<mode>): New peephole.
29604 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
29606         * config/i386/i386.c (make_resolver_func): Update.
29607         * Makefile.in: Add profile-count.h and profile-count.o
29608         * auto-profile.c (afdo_indirect_call): Update to new API.
29609         (afdo_set_bb_count): Update.
29610         (afdo_propagate_edge): Update.
29611         (afdo_propagate_circuit): Update.
29612         (afdo_calculate_branch_prob): Update.
29613         (afdo_annotate_cfg): Update.
29614         * basic-block.h: Include profile-count.h
29615         (struct edge_def): Turn count to profile_count.
29616         (struct basic_block_def): Likewie.
29617         (REG_BR_PROB_BASE): Move to profile-count.h
29618         (RDIV): Move to profile-count.h
29619         * bb-reorder.c (max_entry_count): Turn to profile_count.
29620         (find_traces): Update.
29621         (rotate_loop):Update.
29622         (connect_traces):Update.
29623         (sanitize_hot_paths):Update.
29624         * bt-load.c (migrate_btr_defs): Update.
29625         * cfg.c (RDIV): Remove.
29626         (init_flow): Use alloc_block.
29627         (alloc_block): Uninitialize count.
29628         (unchecked_make_edge): Uninitialize count.
29629         (check_bb_profile): Update.
29630         (dump_edge_info): Update.
29631         (dump_bb_info): Update.
29632         (update_bb_profile_for_threading): Update.
29633         (scale_bbs_frequencies_int): Update.
29634         (scale_bbs_frequencies_gcov_type): Update.
29635         (scale_bbs_frequencies_profile_count): New.
29636         * cfg.h (update_bb_profile_for_threading): Update.
29637         (scale_bbs_frequencies_profile_count): Declare.
29638         * cfgbuild.c (compute_outgoing_frequencies): Update.
29639         (find_many_sub_basic_blocks): Update.
29640         * cfgcleanup.c (try_forward_edges): Update.
29641         (try_crossjump_to_edge): Update.
29642         * cfgexpand.c (expand_gimple_tailcall): Update.
29643         (construct_exit_block): Update.
29644         * cfghooks.c (verify_flow_info): Update.
29645         (dump_bb_for_graph): Update.
29646         (split_edge): Update.
29647         (make_forwarder_block): Update.
29648         (duplicate_block): Update.
29649         (account_profile_record): Update.
29650         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29651         (get_estimated_loop_iterations): Update.
29652         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29653         (single_likely_exit): Update.
29654         * cfgloopmanip.c (scale_loop_profile): Update.
29655         (loopify): Update.
29656         (set_zero_probability): Update.
29657         (lv_adjust_loop_entry_edge): Update.
29658         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29659         (purge_dead_edges): Update.
29660         (rtl_account_profile_record): Update.
29661         * cgraph.c (cgraph_node::create): Uninitialize count.
29662         (symbol_table::create_edge): Uninitialize count.
29663         (cgraph_update_edges_for_call_stmt_node): Update.
29664         (cgraph_edge::dump_edge_flags): Update.
29665         (cgraph_node::dump): Update.
29666         (cgraph_edge::maybe_hot_p): Update.
29667         * cgraph.h: Include profile-count.h
29668         (create_clone), create_edge, create_indirect_edge): Update.
29669         (cgraph_node): Turn count to profile_count.
29670         (cgraph_edge0: Likewise.
29671         (make_speculative, clone): Update.
29672         (create_edge): Update.
29673         (init_lowered_empty_function): Update.
29674         * cgraphclones.c (cgraph_edge::clone): Update.
29675         (duplicate_thunk_for_node): Update.
29676         (cgraph_node::create_clone): Update.
29677         * cgraphunit.c (cgraph_node::analyze): Update.
29678         (cgraph_node::expand_thunk): Update.
29679         * final.c (dump_basic_block_info): Update.
29680         * gimple-streamer-in.c (input_bb): Update.
29681         * gimple-streamer-out.c (output_bb): Update.
29682         * graphite.c (print_global_statistics): Update.
29683         (print_graphite_scop_statistics): Update.
29684         * hsa-brig.c: Include basic-block.h.
29685         * hsa-dump.c: Include basic-block.h.
29686         * hsa-gen.c (T sum_slice): Update.
29687         (convert_switch_statements):Update.
29688         * hsa-regalloc.c: Include basic-block.h.
29689         * ipa-chkp.c (chkp_produce_thunks): Update.
29690         * ipa-cp.c (struct caller_statistics): Update.
29691         (init_caller_stats): Update.
29692         (gather_caller_stats): Update.
29693         (ipcp_cloning_candidate_p): Update.
29694         (good_cloning_opportunity_p): Update.
29695         (get_info_about_necessary_edges): Update.
29696         (dump_profile_updates): Update.
29697         (update_profiling_info): Update.
29698         (update_specialized_profile): Update.
29699         (perhaps_add_new_callers): Update.
29700         (decide_about_value): Update.
29701         (ipa_cp_c_finalize): Update.
29702         * ipa-devirt.c (struct odr_type_warn_count): Update.
29703         (struct decl_warn_count): Update.
29704         (struct final_warning_record): Update.
29705         (possible_polymorphic_call_targets): Update.
29706         (ipa_devirt): Update.
29707         * ipa-fnsummary.c (redirect_to_unreachable): Update.
29708         * ipa-icf.c (sem_function::merge): Update.
29709         * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29710         * ipa-inline.c (compute_uninlined_call_time): Update.
29711         (compute_inlined_call_time): Update.
29712         (want_inline_small_function_p): Update.
29713         (want_inline_self_recursive_call_p): Update.
29714         (edge_badness): Update.
29715         (lookup_recursive_calls): Update.
29716         (recursive_inlining): Update.
29717         (inline_small_functions): Update.
29718         (dump_overall_stats): Update.
29719         (dump_inline_stats): Update.
29720         * ipa-profile.c (ipa_profile_generate_summary): Update.
29721         (ipa_propagate_frequency): Update.
29722         (ipa_profile): Update.
29723         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29724         * ipa-utils.c (ipa_merge_profiles): Update.
29725         * loop-doloop.c (doloop_modify): Update.
29726         * loop-unroll.c (report_unroll): Update.
29727         (unroll_loop_runtime_iterations): Update.
29728         * lto-cgraph.c (lto_output_edge): Update.
29729         (lto_output_node): Update.
29730         (input_node): Update.
29731         (input_edge): Update.
29732         (merge_profile_summaries): Update.
29733         * lto-streamer-in.c (input_cfg): Update.
29734         * lto-streamer-out.c (output_cfg): Update.
29735         * mcf.c (create_fixup_graph): Update.
29736         (adjust_cfg_counts): Update.
29737         (sum_edge_counts): Update.
29738         * modulo-sched.c (sms_schedule): Update.
29739         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29740         * predict.c (maybe_hot_count_p): Update.
29741         (probably_never_executed): Update.
29742         (dump_prediction): Update.
29743         (combine_predictions_for_bb): Update.
29744         (propagate_freq): Update.
29745         (handle_missing_profiles): Update.
29746         (counts_to_freqs): Update.
29747         (rebuild_frequencies): Update.
29748         (force_edge_cold): Update.
29749         * predict.h: Include profile-count.h
29750         (maybe_hot_count_p, counts_to_freqs): UPdate.
29751         * print-rtl-function.c: Do not include cfg.h
29752         * print-rtl.c: Include basic-block.h
29753         * profile-count.c: New file.
29754         * profile-count.h: New file.
29755         * profile.c (is_edge_inconsistent): Update.
29756         (correct_negative_edge_counts): Update.
29757         (is_inconsistent): Update.
29758         (set_bb_counts): Update.
29759         (read_profile_edge_counts): Update.
29760         (compute_frequency_overlap): Update.
29761         (compute_branch_probabilities): Update; Initialize and deinitialize
29762         gcov_count tables.
29763         (branch_prob): Update.
29764         * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29765         (edge_gcov_count): New.
29766         (bb_gcov_count): New.
29767         * shrink-wrap.c (try_shrink_wrapping): Update.
29768         * tracer.c (better_p): Update.
29769         * trans-mem.c (expand_transaction): Update.
29770         (ipa_tm_insert_irr_call): Update.
29771         (ipa_tm_insert_gettmclone_call): Update.
29772         * tree-call-cdce.c: Update.
29773         * tree-cfg.c (gimple_duplicate_sese_region): Update.
29774         (gimple_duplicate_sese_tail): Update.
29775         (gimple_account_profile_record): Update.
29776         (execute_fixup_cfg): Update.
29777         * tree-inline.c (copy_bb): Update.
29778         (copy_edges_for_bb): Update.
29779         (initialize_cfun): Update.
29780         (freqs_to_counts): Update.
29781         (copy_cfg_body): Update.
29782         (expand_call_inline): Update.
29783         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29784         * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29785         (try_unroll_loop_completely): Update.
29786         (try_peel_loop): Update.
29787         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29788         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29789         * tree-ssa-loop-split.c (connect_loops): Update.
29790         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29791         * tree-ssa-reassoc.c (branch_fixup): Update.
29792         * tree-ssa-tail-merge.c (replace_block_by): Update.
29793         * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29794         (compute_path_counts): Update.
29795         (update_profile): Update.
29796         (recompute_probabilities): Update.
29797         (update_joiner_offpath_counts): Update.
29798         (estimated_freqs_path): Update.
29799         (freqs_to_counts_path): Update.
29800         (clear_counts_path): Update.
29801         (ssa_fix_duplicate_block_edges): Update.
29802         (duplicate_thread_path): Update.
29803         * tree-switch-conversion.c (case_bit_test_cmp): Update.
29804         (struct switch_conv_info): Update.
29805         * tree-tailcall.c (decrease_profile): Update.
29806         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29807         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29808         * value-prof.c (check_counter): Update.
29809         (gimple_divmod_fixed_value): Update.
29810         (gimple_mod_pow2): Update.
29811         (gimple_mod_subtract): Update.
29812         (gimple_ic_transform): Update.
29813         (gimple_stringop_fixed_value): Update.
29814         * value-prof.h (gimple_ic): Update.
29816 2017-06-02  Carl Love  <cel@us.ibm.com>
29818         * config/rs6000/rs6000-c: Add support for built-in functions
29819         vector double vec_doublee (vector signed int);
29820         vector double vec_doublee (vector unsigned int);
29821         vector double vec_doublee (vector float);
29822         vector double vec_doubleh (vector signed int);
29823         vector double vec_doubleh (vector unsigned int);
29824         vector double vec_doubleh (vector float);
29825         vector double vec_doublel (vector signed int);
29826         vector double vec_doublel (vector unsigned int);
29827         vector double vec_doublel (vector float);
29828         vector double vec_doubleo (vector signed int);
29829         vector double vec_doubleo (vector unsigned int);
29830         vector double vec_doubleo (vector float);.
29831         * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29832         DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29833         UNS_DOUBLEL.
29834         * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29835         unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29836         unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29837         VS_sxwsp.
29838         * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29839         vec_doublel, vec_doubleh.
29840         * doc/extend.texi: Update the built-in documentation file for the
29841         new built-in functions.
29843 2017-06-02  David Malcolm  <dmalcolm@redhat.com>
29845         PR jit/80954
29846         * ipa-inline-analysis.c (free_growth_caches): Set
29847         edge_removal_hook_holder to NULL after removing it.
29849 2017-06-02  Sudakshina Das  <sudi.das@arm.com>
29851         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29852         comparision with zero.
29854 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29855         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29856         for early expansion of vec_min and vec_max builtins.
29857         (builtin_function_type): Add min/max unsigned variants to those
29858         identified as having unsigned arguments.
29860 2017-06-02  Olivier Hainque  <hainque@adacore.com>
29862         * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29864 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29866         * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29867         Use VALL_F16 iterator rather than VALL.
29869 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29871         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29872         Emit CBNZ inside loop when doing a strong exchange and comparing
29873         against zero.  Generate the CC flags after the loop.
29875 2017-06-02  David Edelsohn  <dje.gcc@gmail.com>
29877         * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29878         (dl_section_ref): New.
29879         (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29880         On AIX, append an expression to subtract the size of the
29881         section length to dl_section_ref.
29883 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29885         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29886         for early expansion of vector absolute builtins.
29888 2017-06-02  Richard Biener  <rguenther@suse.de>
29890         * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29891         what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29893 2017-06-02  Richard Biener  <rguenther@suse.de>
29895         PR tree-optimization/80948
29896         * tree-tailcall.c (find_tail_calls): Track stmts to move in
29897         stmt order as well.
29899 2017-06-02  Richard Biener  <rguenther@suse.de>
29901         * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29902         PHIs are ok.
29903         * tree-vect-stmts.c (process_use): Do not mark backedge defs
29904         for inductions as relevant.
29906 2017-06-02  Richard Biener  <rguenther@suse.de>
29908         * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29909         (vectorizable_induction): ... this.  Remove dead code.
29911 2017-06-02  Eric Botcazou  <ebotcazou@adacore.com>
29913         * builtins. (expand_builtin_alloca): Remove second parameter and
29914         infer its value from the first parameter instead.
29915         (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
29917 2017-06-02  Jakub Jelinek  <jakub@redhat.com>
29919         PR rtl-optimization/80903
29920         * loop-doloop.c (add_test): Unshare sequence.
29922 2017-06-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29924         * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
29926 2017-06-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29928         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
29929         static.
29930         (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
29931         xlogue_layout::get_instance, logue_layout::xlogue_layout,
29932         sp_valid_at, fp_valid_at, choose_basereg): Formatting.
29933         (xlogue_layout::get_stub_rtx): Make static.
29934         (xlogue_layout::get_stub_name): Avoid const-cast, make static.
29935         (xlogue_layout::compute_stub_managed_regs): Rename to...
29936         (xlogue_layout::count_stub_managed_regs): ...this.
29937         (xlogue_layout::is_stub_managed_reg): New function.
29938         (xlogue_layout::m_stub_names): Rename to...
29939         (xlogue_layout::s_stub_names): ...this, make static.
29940         (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
29941         xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
29942         xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
29943         xlogue_layout::s_stub_names): Instantiate statics.
29944         (stub_managed_regs): Remove.
29945         (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
29946         (disable_call_ms2sysv_xlogues): Rename to...
29947         (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
29948         (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
29949         warning logic.
29950         (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
29951         change after reload_completed.
29952         (ix86_can_use_return_insn_p): Use the ix86_frame data structure
29953         directly.
29954         (ix86_expand_prologue): Likewise.
29955         (ix86_expand_epilogue): Likewise.
29956         (ix86_expand_split_stack_prologue): Likewise.
29957         (ix86_compute_frame_layout): Remove frame parameter ...
29958         (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
29959         (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
29960         only if necessary.
29961         (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
29962         (ix86_frame): Move from here ...
29963         * config/i386/i386.h (ix86_frame): ... to here.
29964         (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
29965         complete ix86_frame data structure instead.  Remove some_ld_name.
29967 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29969         * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
29970         symbols that hold a DECL_VALUE_EXPR.
29972 2017-06-01  Martin Jambor  <mjambor@suse.cz>
29974         PR tree-optimization/80898
29975         * tree-sra.c (process_subtree_disqualification): Removed.
29976         (disqualify_candidate): Do not acll
29977         process_subtree_disqualification.
29978         (subtree_mark_written_and_enqueue): New function.
29979         (propagate_all_subaccesses): Set grp_write of LHS subtree if the
29980         RHS has been disqualified and re-queue LHS if necessary.  Apart
29981         from that, ignore disqualified RHS.
29983 2017-06-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29985         * config/s390/s390.c (s390_emit_epilogue): Disable early return
29986         address fetch for z10 or later.
29988 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29990         * config/arc/arc.md (tst_movb): Add guard when splitting.
29992 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29994         * config/arc/arc.c (arc_can_eliminate): Test against
29995         arc_frame_pointer_needed.
29997 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29999         * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
30000         to prevent store reordering.
30001         * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
30002         (type): Add block type.
30003         (stack_tie): Define special instruction to be used in
30004         expand_prologue.
30006 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30008         * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
30009         constraint. It is not valid for the pattern.
30010         (noncommutative_binary_comparison): Likewise.
30012 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30014         * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
30015         scaled addresses.
30017 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30019         * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
30020         be used by the reg-alloc.
30022 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30024         * config/arc/arc.md (mulsi3): Avoid use of hard registers before
30025         reg-alloc when having mul64 or mul32x16 instructions.
30026         (mulsidi3): Likewise.
30027         (umulsidi3): Likewise.
30028         (mulsi32x16): New pattern.
30029         (mulsi64): Likewise.
30030         (mulsidi64): Likewise.
30031         (umulsidi64): Likewise.
30032         (MUL32x16_REG): Define.
30033         (mul64_600): Use MUL32x16_REG.
30034         (mac64_600): Likewise.
30035         (umul64_600): Likewise.
30036         (umac64_600): Likewise.
30038 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30040         * config/arc/arc.md (mulsi3_700): Make it commutative.
30042 2017-06-01  Jose E. Marchesi  <jose.marchesi@oracle.com>
30044         * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
30045         type for movstouw.
30046         (*sign_extendsidi2_insn): Likewise for movstosw.
30048 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
30050         * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
30051         the type of the input discriminant value.  Convert the
30052         discriminant value of signedness vary.
30054 2017-06-01  Volker Reichelt  <v.reichelt@netcologne.de>
30056         * doc/invoke.texi (-Wcatch-value): Document new shortcut.
30057         Add to -Wall section.
30059 2017-06-01  Richard Biener  <rguenther@suse.de>
30061         PR middle-end/66313
30062         * fold-const.c (fold_plusminus_mult_expr): If the factored
30063         factor may be zero use a wrapping type for the inner operation.
30064         * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
30065         and handle moved defs.
30066         (process_assignment): Properly guard the unary op case.  Return a
30067         tri-state indicating that moving the stmt before the call may allow
30068         to continue.  Pass through to_move.
30069         (find_tail_calls): Handle moving unrelated defs before
30070         the call.
30072 2017-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
30074         PR target/80618
30075         * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
30076         splitter result in the canonical way.
30078 2017-05-31  Uros Bizjak  <ubizjak@gmail.com>
30080         * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
30081         also for 32bit target.  Update insn attributes.
30082         (zero-extendsidi2 splitter): Allow all registers for operand 1.
30084 2017-05-31  Sebastian Peryt  <sebastian.peryt@intel.com>
30086         * config/i386/avx512fintrin.h (_mm_mask_max_sd)
30087         (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
30088         (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
30089         (_mm_maskz_min_ss): New intrinsics.
30091 2017-05-31  Martin Liska  <mliska@suse.cz>
30093         * tree-vect-loop.c (vect_create_epilog_for_reduction):
30094         Change comment style to one we normally use.
30095         (vectorizable_reduction): Likewise.
30096         (vectorizable_induction): Likewise.
30097         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
30098         (vectorizable_call): Likewise.
30099         (vectorizable_simd_clone_call): Likewise.
30100         (vectorizable_conversion): Likewise.
30101         (vectorizable_assignment): Likewise.
30102         (vectorizable_shift): Likewise.
30103         (vectorizable_operation): Likewise.
30104         (vectorizable_store): Likewise.
30105         (vectorizable_load): Likewise.
30106         * tree-vectorizer.h: Likewise.
30108 2017-05-31  Alexander Monakov  <amonakov@ispras.ru>
30110         * passes.c (emergency_dump_function): New.
30111         * tree-pass.h (emergency_dump_function): Declare.
30112         * plugin.c (plugins_internal_error_function): Remove.
30113         * plugin.h (plugins_internal_error_function): Remove declaration.
30114         * toplev.c (internal_error_function): New static function.  Use it...
30115         (general_init): ...here.
30117 2017-05-31  Graham Markall  <graham.markall@embecosm.com>
30119         * config/arc/arc.c (arc_print_operand): Handle constant operands.
30120         (arc_rtx_costs): Add costs for new patterns.
30121         * config/arc/arc.md: Additional *add_n and *sub_n patterns.
30122         * config/arc/predicates.md: Add _1_2_3_operand predicate.
30124 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30126         * tree-ssa-strlen.c (get_next_strinfo): New function.
30127         (get_stridx_plus_constant): Use it.
30128         (zero_length_string): Likewise.
30129         (adjust_related_strinfos): Likewise.
30130         (adjust_last_stmt): Likewise.
30132 2017-05-31  Richard Biener  <rguenther@suse.de>
30134         PR target/80880
30135         * config/i386/i386.c (ix86_expand_builtin): Remove assert
30136         for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
30138 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30140         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
30141         loop_vinfo argument and use of dependence distance vectors.
30142         Check instead whether the two references differ only in their
30143         initial value and assume that they have the same alignment if the
30144         difference is a multiple of the vector alignment.
30145         (vect_analyze_data_refs_alignment): Update call accordingly.
30147 2017-05-31  Martin Liska  <mliska@suse.cz>
30149         PR target/79155
30150         * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
30152 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30154         * tree-vect-loop-manip.c (create_intersect_range_checks_index)
30155         (create_intersect_range_checks): Move from ...
30156         * tree-data-ref.c (create_intersect_range_checks_index)
30157         (create_intersect_range_checks): ... to here.
30158         (create_runtime_alias_checks): New function factored from ...
30159         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
30160         here.  Call above function.
30161         * tree-data-ref.h (create_runtime_alias_checks): New function.
30163 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30165         * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
30166         segment length for dr_b and compute it in wide_int.
30168 2017-05-31  Richard Biener  <rguenther@suse.de>
30170         PR tree-optimization/80906
30171         * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
30172         and pass through iv_map.
30173         (copy_bb_and_scalar_dependences): Adjust.
30174         (translate_pending_phi_nodes): Likewise.
30175         (copy_loop_close_phi_args): Handle code-generating IVs instead
30176         of ICEing.
30178 2017-05-30  David Malcolm  <dmalcolm@redhat.com>
30180         * diagnostic-color.c (color_dict): Add "type-diff".
30181         (parse_gcc_colors): Update comment.
30182         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
30183         -fdiagnostics-show-template-tree and -fno-elide-type.
30184         (GCC_COLORS): Add type-diff to example.
30185         (type-diff=): New.
30186         (-fdiagnostics-show-template-tree): New.
30187         (-fno-elide-type): New.
30188         * pretty-print.c (pp_format): Pass quote and formatters[argno] to
30189         the pp_format_decoder callback.  Call any m_format_postprocessor's
30190         "handle" method.
30191         (pretty_printer::pretty_printer): Initialize
30192         m_format_postprocessor.
30193         (pretty_printer::~pretty_printer): Delete any
30194         m_format_postprocessor.
30195         * pretty-print.h (printer_fn): Add bool and const char ** parameters.
30196         (class format_postprocessor): New class.
30197         (struct pretty_printer::format_decoder): Document the new parameters.
30198         (struct pretty_printer::m_format_postprocessor): New field.
30199         * tree-diagnostic.c (default_tree_printer): Update for new
30200         bool and const char ** params.
30201         * tree-diagnostic.h (default_tree_printer): Likewise.
30203 2017-05-30  Segher Boessenkool  <segher@kernel.crashing.org>
30205         * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
30206         (lwa_operand): Delete rs6000_gen_cell_microcode test.
30207         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30208         rs6000_gen_cell_microcode code.
30209         (rs6000_final_prescan_insn): Delete.
30210         (rs6000_opt_vars): Delete the "gen-cell-microcode" and
30211         "warn-cell-microcode" entries.
30212         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
30213         * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
30214         throughout.  Change cc_reg_not_micro_cr0_operand to
30215         cc_reg_not_cr0_operand throughout.
30216         (*extendhi<mode>2_noload): Delete.
30217         * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
30218         (mwarn-cell-microcode): Delete.
30219         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
30220         -mgen-cell-microcode and -mwarn-cell-microcode.
30222 2017-05-30  Uros Bizjak  <ubizjak@gmail.com>
30224         PR target/80833
30225         * config/i386/constraints.md (Yd): New constraint.
30226         (Ye): Ditto.
30227         * config/i386/i386.md (*movti_internal): Add (?r, Ye)
30228         and (?Yd, r) alternatives.  Update insn attributes.
30229         * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
30230         and (?*Yd, r) alternatives.  Update insn attributes.
30231         (double-mode inter-unit splitters): Add new GR<->XMM splitters.
30233 2017-05-30  Pierre-Marie de Rodat  <derodat@adacore.com>
30235         * gimplify.c (gimplify_modify_expr): Don't create a
30236         DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30237         function.
30239 2017-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
30241         * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30243 2017-05-30  Richard Biener  <rguenther@suse.de>
30245         * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30246         and reduc_def fields.
30247         (STMT_VINFO_REDUC_TYPE): New define.
30248         (STMT_VINFO_REDUC_DEF): Likewise.
30249         (vect_force_simple_reduction): Adjust prototype.
30250         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30251         (vect_is_simple_reduction): Remove check_reduction argument.
30252         (vect_force_simple_reduction): Adjust and set
30253         STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30254         (vectorizable_reduction): Do not re-do reduction analysis
30255         but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30256         * tree-parloops.c (gather_scalar_reductions): Adjust.
30258 2017-05-30  Richard Biener  <rguenther@suse.de>
30260         PR middle-end/80901
30261         * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30262         split_edge code.
30264 2017-05-24  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30266         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30267         Introduce unknown_misalignment parameter and remove vf.
30268         (vect_peeling_hash_get_lowest_cost):
30269         Pass unknown_misalignment parameter.
30270         (vect_enhance_data_refs_alignment):
30271         Fix unsupportable data ref treatment.
30273 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30275         * tree-vect-data-refs.c (vect_get_data_access_cost):
30276         Workaround for SLP handling.
30277         (vect_enhance_data_refs_alignment):
30278         Compute costs for doing no peeling at all, compare to the best
30279         peeling costs so far and avoid peeling if cheaper.
30281 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30283         * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30284         Return peeling info and set costs to zero for unlimited cost
30285         model.
30286         (vect_enhance_data_refs_alignment): Also inspect all datarefs
30287         with unknown misalignment. Compute and costs for unknown
30288         misalignment, compare them to the costs for known misalignment
30289         and choose the cheapest for peeling.
30291 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30293         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30294         (vect_get_peeling_costs_all_drs): Create function.
30295         (vect_peeling_hash_get_lowest_cost):
30296         Use vect_get_peeling_costs_all_drs.
30297         (vect_peeling_supportable): Create function.
30298         (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30300 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30302         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30303         DR_HAS_NEGATIVE_STEP.
30304         (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30305         (vect_enhance_data_refs_alignment): Use.
30306         (vect_duplicate_ssa_name_ptr_info): Use.
30307         * tree-vectorizer.h (dr_misalignment): Use.
30308         (known_alignment_for_access_p): Use.
30310 2017-05-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
30312         PR target/78838
30313         * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30314         .lowtext.
30315         (has_section_name): New function.
30317 2017-05-30  Martin Liska  <mliska@suse.cz>
30319         PR other/80909
30320         * auto-profile.c (get_function_decl_from_block): Fix
30321         parenthesis.
30323 2017-05-30  Richard Biener  <rguenther@suse.de>
30325         PR middle-end/80876
30326         * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30328 2017-05-30  Martin Liska  <mliska@suse.cz>
30330         * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30331         * dumpfile.h (struct dump_file_info): Remove ctors.
30333 2017-05-30  Martin Liska  <mliska@suse.cz>
30335         * predict.def: Fix GNU coding style.
30337 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30339         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30340         Mark 'to' argument with ATTRIBUTE_UNUSED.
30342 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30344         * config/xtensa/xtensa.c (xtensa_emit_call): Use
30345         HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30346         (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30347         format string.
30349 2017-05-29  Eric Botcazou  <ebotcazou@adacore.com>
30351         * doc/install.texi (Options specification): Restore entry of
30352         --enable-sjlj-exceptions.
30354 2017-05-27  Michael Eager  <eager@eagercon.com>
30356         Revert:
30357         2016-01-21  Ajit Agarwal  <ajitkum@xilinx.com>
30359         See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30361         * config/microblaze/microblaze.h
30362         (FIXED_REGISTERS): Update in macro.
30363         (CALL_USED_REGISTERS): Update in macro.
30365 2017-05-27  François-Xavier Coudett  <fxcoudert@gcc.gnu.org>
30367         * doc/install.texi: Add links to macOS binary distributions.
30369 2017-05-27  Jakub Jelinek  <jakub@redhat.com>
30371         PR bootstrap/80887
30372         Revert:
30373         2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30375         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30377 2017-05-26  Martin Liska  <mliska@suse.cz>
30379         * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30381 2017-05-26  Martin Liska  <mliska@suse.cz>
30383         * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30384         always leading ';; '.
30385         (dump_bb_info): Likewise.
30386         (brief_dump_cfg): Likewise.
30387         * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30388         * dumpfile.c: Remove usage of TDF_VERBOSE.
30389         * dumpfile.h (enum dump_kind): Likewise.
30390         (dump_gimple_bb_header): Do not use TDF_COMMENT.
30391         * print-tree.c (debug_verbose): Remove.
30392         * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30393         (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30394         * tree-diagnostic.c (default_tree_printer): Replace
30395         TDF_DIAGNOSTIC with TDF_SLIM.
30397 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30399         * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30400         in parameter loop, rather than loop_vinfo.
30401         (create_intersect_range_checks): Ditto.
30402         (vect_create_cond_for_alias_checks): Update call to above functions.
30404 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30406         PR tree-optimization/80815
30407         * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30408         for merging runtime alias checks.  Handle negative DR_STEPs.
30410 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30412         * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30413         Move from ...
30414         * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30415         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30416         out code pruning runtime alias checks.
30417         * tree-data-ref.c (prune_runtime_alias_test_list): New function
30418         factored out from above.
30419         * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30420         Move from ...
30421         * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30422         ... to here.
30423         (prune_runtime_alias_test_list): New decalaration.
30425 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30427         * tree-vect-data-refs.c (compare_tree): Rename and move ...
30428         * tree-data-ref.c (data_ref_compare_tree): ... to here.
30429         * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30430         * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30431         (operator==, comp_dr_with_seg_len_pair): Ditto.
30432         (vect_prune_runtime_alias_test_list): Ditto.
30434 2017-05-26  Martin Liska  <mliska@suse.cz>
30436         PR ipa/80663
30437         * params.def: Bound partial-inlining-entry-probability param.
30439 2017-05-26  Marek Polacek  <polacek@redhat.com>
30441         PR sanitizer/80875
30442         * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30443         can be negated.
30445 2017-05-26  Richard Biener  <rguenther@suse.de>
30447         PR tree-optimization/80842
30448         * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30449         value.
30451 2017-05-26  Richard Biener  <rguenther@suse.de>
30453         PR tree-optimization/80844
30454         * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30456 2017-05-25  Sebastian Peryt  <sebastian.peryt@intel.com>
30458         * doc/md.texi (Machine Constraints): Update x86 family
30459         machine constraints section to match 'config/i386/constraints.md'.
30461 2017-05-25  Volker Reichelt  <v.reichelt@netcologne.de>
30463         * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30465 2017-05-25  Nathan Sidwell  <nathan@acm.org>
30467         * doc/invoke.texi (--enable-languages): Update documentation.
30469 2017-05-25  Martin Liska  <mliska@suse.cz>
30471         * dumpfile.c: Add TDF_FOLDING.
30472         * dumpfile.h (enum dump_kind): Likewise.
30473         * genmatch.c (dt_simplify::gen_1): Use it.
30475 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30477         * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30479 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30481         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30482         * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30484 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30486         * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30487         * match.pd (X == C): Rewrite it here.
30488         (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30489         with_certain_nonzero_bits2): New predicates.
30490         * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30492 2017-05-24  Nathan Sidwell  <nathan@acm.org>
30494         * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30495         avoid warning.
30497         * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30498         warning.
30500 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30502         * config/powerpcspe: New port.  Files are copied from the rs6000
30503         port, with "rs6000" in filenames replaced by "powerpcspe".
30505 2017-05-24  Wilco Dijkstra  <wdijkstr@arm.com>
30507         PR rtl-optimization/80754
30508         * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30510 2017-05-24  Sheldon Lobo  <smlobo@sheldon.us.oracle.com>
30512         * config/sparc/sparc.md (length): Return the correct value for -mflat
30513         sibcalls to match output_sibcall.
30515 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30517         PR bootstrap/80860
30518         PR bootstrap/80843
30519         * config/rs6000/rs6000.c (struct machine_function): Add new field
30520         n_components.
30521         (rs6000_get_separate_components): Init that field, use it.
30522         (rs6000_components_for_bb): Use the field.
30524 2017-05-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30526         * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30528 2017-05-24  Peter Bergner  <bergner@vnet.ibm.com>
30530         PR middle-end/80823
30531         * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30533 2017-05-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30535         PR target/80725
30536         * config/s390/s390.c (s390_check_qrst_address): Check incoming
30537         address against address_operand predicate.
30538         * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30540 2017-05-24  Eric Botcazou  <ebotcazou@adacore.com>
30542         * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30543         parameters passed indirectly.
30545 2017-05-23  Uros Bizjak  <ubizjak@gmail.com>
30547         * config/i386/i386.md (*movdi_internal): Remove SSE4
30548         alternative 18 (?r, *v).  Update insn attributes.
30549         (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30550         Update insn attributes.
30551         (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30552         Update insn attributes.
30553         * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30554         alternative 1 (r, v). Remove isa attribute.
30555         * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30556         Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30557         and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30559 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30561         * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30562         dg-line directive.
30564 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30566         * cgraphunit.c (symbol_table::process_new_functions): Update.
30567         * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30568         (inline_generate_summary): Rename to ...
30569         (ipa_fn_summary_generate): ... this one.
30570         (inline_read_summary): Rename to ...
30571         (ipa_fn_summary_read): ... this one.
30572         (inline_write_summary): Rename to ...
30573         (ipa_fn_summary_write): ... this one.
30574         (inline_free_summary): Rename to ...
30575         (ipa_free_fn_summary): ... this one.
30576         (pass_data_local_fn_summary, pass_local_fn_summary,
30577         make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30578         pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30579         pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30580         make_pass_ipa_fn_summary): New.
30581         * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30582         inline_write_summary, inline_free_summary): Remove.
30583         (ipa_free_fn_summary) : New.
30584         * ipa-inline.c (ipa_inline): Update.
30585         (pass_ipa_inline): Do not generate summaries.
30586         * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30587         Remove.
30588         * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30589         and add pass_ipa_fn_summary.
30590         * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30591         New.
30592         (make_pass_inline_parameters): Remove.
30594 2017-05-23  Thomas Schwinge  <thomas@codesourcery.com>
30596         * omp-low.c (struct omp_context): Remove "default_kind" member.
30597         Adjust all users.
30599         * omp-offload.c (execute_oacc_device_lower): Remove the
30600         parallelism dimensions function attributes for unparallelized
30601         OpenACC kernels constructs.
30603 2017-05-23  Martin Liska  <mliska@suse.cz>
30605         * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30606         functions.
30607         (cgraph_edge::make_speculative): Likewise.
30608         (cgraph_edge::resolve_speculation): Likewise.
30609         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30610         (cgraph_node::dump): Likewise.
30611         * cgraph.h: Likewise.
30612         * cgraphunit.c (analyze_functions): Likewise.
30613         (symbol_table::compile): Likewise.
30614         * ipa-cp.c (print_all_lattices): Likewise.
30615         (determine_versionability): Likewise.
30616         (initialize_node_lattices): Likewise.
30617         (ipcp_verify_propagated_values): Likewise.
30618         (estimate_local_effects): Likewise.
30619         (update_profiling_info): Likewise.
30620         (create_specialized_node): Likewise.
30621         (perhaps_add_new_callers): Likewise.
30622         (decide_about_value): Likewise.
30623         (decide_whether_version_node): Likewise.
30624         (identify_dead_nodes): Likewise.
30625         (ipcp_store_bits_results): Likewise.
30626         * ipa-devirt.c (dump_targets): Likewise.
30627         (ipa_devirt): Likewise.
30628         * ipa-icf.c (sem_item::dump): Likewise.
30629         (sem_function::equals): Likewise.
30630         (sem_variable::equals): Likewise.
30631         (sem_item_optimizer::read_section): Likewise.
30632         (sem_item_optimizer::execute): Likewise.
30633         (congruence_class::dump): Likewise.
30634         * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30635         (dump_inline_summary): Likewise.
30636         (estimate_node_size_and_time): Likewise.
30637         (inline_analyze_function): Likewise.
30638         * ipa-inline-transform.c (inline_call): Likewise.
30639         * ipa-inline.c (report_inline_failed_reason): Likewise.
30640         (want_early_inline_function_p): Likewise.
30641         (edge_badness): Likewise.
30642         (update_edge_key): Likewise.
30643         (inline_small_functions): Likewise.
30644         * ipa-profile.c (ipa_profile): Likewise.
30645         * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30646         (ipa_make_edge_direct_to_target): Likewise.
30647         (remove_described_reference): Likewise.
30648         (ipa_impossible_devirt_target): Likewise.
30649         (propagate_controlled_uses): Likewise.
30650         (ipa_print_node_params): Likewise.
30651         (ipcp_transform_function): Likewise.
30652         * ipa-pure-const.c (pure_const_read_summary): Likewise.
30653         (propagate_pure_const): Likewise.
30654         * ipa-reference.c (generate_summary): Likewise.
30655         (read_write_all_from_decl): Likewise.
30656         (propagate): Likewise.
30657         (ipa_reference_read_optimization_summary): Likewise.
30658         * ipa-utils.c (ipa_merge_profiles): Likewise.
30659         * ipa.c (walk_polymorphic_call_targets): Likewise.
30660         (symbol_table::remove_unreachable_nodes): Likewise.
30661         (ipa_single_use): Likewise.
30662         * passes.c (execute_todo): Likewise.
30663         * predict.c (drop_profile): Likewise.
30664         * symtab.c (symtab_node::get_dump_name): New function.
30665         (symtab_node::dump_name): Likewise.
30666         (symtab_node::dump_asm_name): Likewise.
30667         (symtab_node::dump_references): Likewise.
30668         (symtab_node::dump_referring): Likewise.
30669         (symtab_node::dump_base): Likewise.
30670         (symtab_node::debug_symtab): Likewise.
30671         * tree-sra.c (convert_callers_for_node): Likewise.
30672         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30673         * value-prof.c (init_node_map): Likewise.
30675 2017-05-23  Martin Liska  <mliska@suse.cz>
30677         * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30678         and symtab_node::debug_symtab to symbol_table::debug.
30679         * cgraphunit.c (analyze_functions): Use the renamed function.
30680         (symbol_table::compile): Likewise.
30681         * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30682         * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30683         * passes.c (execute_todo): Likewise.
30684         * symtab.c (symbol_table::dump): New function.
30685         * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30687 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30689         * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30690         that nonconst implies exec.
30692 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30694         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30695         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30696         (inline_edge_summary_vec): Turn into ...
30697         (ipa_call_summaries): ... this one.
30698         (redirect_to_unreachable, edge_set_predicate,
30699         evaluate_properties_for_edge, inline_summary_alloc,
30700         reset_ipa_call_summary, reset_inline_summary,
30701         inline_summary_t::duplicate): Update.
30702         (inline_edge_duplication_hook): Turn to ...
30703         (ipa_call_summary_t::duplicate): ... this one.
30704         (inline_edge_removal_hook): Turn to ...
30705         (ipa_call_summary_t::remove): ... this one.
30706         (dump_inline_edge_summary): Turn to ...
30707         (dump_ipa_call_summary): ... this one.
30708         (estimate_function_body_sizes): Update.
30709         (inline_update_callee_summaries): Update.
30710         (remap_edge_change_prob): Update.
30711         (remap_edge_summaries): Update.
30712         (inline_merge_summary): Update.
30713         (do_estimate_edge_time): Update.
30714         (inline_generate_summary): Update.
30715         (inline_read_section): Update.
30716         (inline_read_summary): Update.
30717         (inline_free_summary): Update.
30718         * ipa-inline.c (can_inline_edge_p): Update.
30719         (compute_inlined_call_time): Update.
30720         (want_inline_small_function_p): Update.
30721         (edge_badness): Update.
30722         (early_inliner): Update.
30723         * ipa-inline.h (inline_edge_summary): Turn to ...
30724         (ipa_call_summary): ... this one.
30725         (ipa_call_summary_t): New class.
30726         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30727         (ipa_call_summaries): New.
30728         (inline_edge_summary): Remove.
30729         (estimate_edge_growth): Update.
30730         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30731         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30732         * ipa-split.c (execute_split_functions): Update.
30733         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30735 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30737         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30738         attributes): Document rdrand effective target.
30740 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30742         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30743         attributes): Sort alphabetically.
30745 2017-05-23  Georg-Johann Lay  <avr@gjlay.de>
30747         * config/avr/genmultilib.awk: Use gsub instead of gensub.
30749 2017-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30751         PR target/80718
30752         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30753         V2DF/V2DI splat into two separate patterns, one that handles
30754         registers, and the other that only handles memory.  Drop support
30755         for splatting from a GPR on ISA 2.07 and then splitting the
30756         splat into direct move and splat.
30757         (vsx_splat_<mode>_reg): Likewise.
30758         (vsx_splat_<mode>_mem): Likewise.
30760 2017-05-22  Segher Boessenkool  <segher@kernel.crashing.org>
30762         * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30764 2017-05-22  Jakub Jelinek  <jakub@redhat.com>
30766         PR middle-end/80809
30767         * omp-low.c (finish_taskreg_remap): New function.
30768         (finish_taskreg_scan): If unit size of ctx->record_type
30769         is non-constant, unshare the size expression and replace
30770         decls in it with possible outer var refs.
30772         PR middle-end/80809
30773         * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30774         GOVD_SHARED rather than GOVD_PRIVATE with it.
30775         (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30776         GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30778         PR middle-end/80853
30779         * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30780         as last argument to build_outer_var_ref for pointer bases of array
30781         section reductions.
30783 2017-05-19  Martin Sebor  <msebor@redhat.com>
30785         * print-tree.c (print_node): Print DECL_READ_P flag.
30787 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30789         * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30790         * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30791         * cgraph.c: Likewise.
30792         * cgraphunit.c: Likewise.
30793         * gengtype.c: Likewise.
30794         * ipa-cp.c: Likewise.
30795         * ipa-devirt.c: Likewise.
30796         * ipa-icf.c: Likewise.
30797         * ipa-predicate.c: Likewise.
30798         * ipa-profile.c: Likewise.
30799         * ipa-prop.c: Likewise.
30800         * ipa-split.c: Likewise.
30801         * ipa.c: Likewise.
30802         * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30803         edge_predicate_pool, dump_inline_hints,
30804         inline_summary::account_size_time, redirect_to_unreachable,
30805         edge_set_predicate, set_hint_predicate,
30806         evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30807         inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30808         inline_summary_t::remove, remap_hint_predicate_after_duplication,
30809         inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30810         ipa_call_summary_t::remove, initialize_growth_caches,
30811         free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30812         debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30813         mark_modified, unmodified_parm_1, unmodified_parm,
30814         unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30815         set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30816         compute_bb_predicates, will_be_nonconstant_expr_predicate,
30817         will_be_nonconstant_predicate, record_modified_bb_info,
30818         get_minimal_bb, record_modified, param_change_prob,
30819         phi_result_unknown_predicate, predicate_for_phi_result,
30820         array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30821         estimate_function_body_sizes, compute_inline_parameters,
30822         compute_inline_parameters_for_curren, pass_data_inline_parameters,
30823         estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30824         inline_update_callee_summaries, remap_edge_change_prob,
30825         remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30826         inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30827         inline_analyze_function, inline_summary_t::insert,
30828         inline_generate_summary, read_ipa_call_summary, inline_read_section,
30829         inline_read_summary, write_ipa_call_summary, inline_write_summary,
30830         inline_free_summary): Move to ipa-fnsummary.h
30831         (predicate_t): Remove.
30832         * ipa-fnsummary.c: New file.
30833         * ipa-inline.h:  Do not include sreal.h and ipa-predicate.h
30834         (enum inline_hints_vals, inline_hints, agg_position_info,
30835         INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30836         inline_summaries, ipa_call_summary, ipa_call_summary_t,
30837         ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30838         dump_inline_summary, dump_inline_hints, inline_generate_summary,
30839         inline_read_summary, inline_write_summary, inline_free_summary,
30840         inline_analyze_function, initialize_inline_failed,
30841         inline_merge_summary, inline_update_overall_summary,
30842         compute_inline_parameters): Move to ipa-fnsummary.h
30843         * ipa-fnsummary.h: New file.
30844         * ipa-inline-transform.h: Include ipa-inline.h.
30845         * ipa-inline.c: LIkewise.
30847 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30849         * ipa-inline.c (edge_badness): Use inlined_time instead of
30850         inline_summaries->get.
30852 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30854         * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30856 2017-05-22  Nathan Sidwell  <nathan@acm.org>
30858         * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30859         (fdump-lang): Document 'raw' option.
30860         * dumpfile.h (TDI_tu): Delete.
30861         * dumpfile.c (dump_files): Remove translation-unit.
30862         (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30864 2017-05-22  Georg-Johann Lay  <avr@gjlay.de>
30866         * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30867         command option from $(AWK) call.
30868         * config/avr/genmultilib.awk: Simplify and rewrite so that it
30869         generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30870         [FORMAT]: Remove handling of variable.
30871         * config/avr/t-multilib: Regenerate.
30873 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30875         * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30876         self_time.
30877         (dump_inline_summary): Do not print self_time.
30878         (estimate_function_body_sizes): Do not set self_time.
30879         (compute_inline_parameters): Likewise.
30880         (inline_read_section, inline_write_summary): Do not stream self_time.
30881         * ipa-inline.h (inline_summary): Drop self_time.
30883 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30885         * ipa-inline-analysis.c (account_size_time): Rename to ...
30886         (inline_summary::account_size_time): ... this one.
30887         (reset_ipa_call_summary): Turn to ...
30888         (ipa_call_summary::reset): ... this one.
30889         (reset_inline_summary): Turn to ...
30890         (inline_summary::reset): ... this one.
30891         (inline_summary_t::remove): Update.
30892         (inline_summary_t::duplicate): Update.
30893         (ipa_call_summary_t::remove): Update.
30894         (dump_inline_summary): Update.
30895         (estimate_function_body_sizes): Update.
30896         (compute_inline_parameters): Update.
30897         (estimate_node_size_and_time): Update.
30898         (inline_merge_summary): Update.
30899         (inline_update_overall_summary): Update.
30900         (inline_read_section): Update.
30901         (inline_write_summary): Update.
30902         * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30903         add account_size_time and reset member functions.
30904         (ipa_call_summary): Add reset function.
30905         * ipa-predicate.h (predicate::operator &): Constify.
30907 2017-05-22  Richard Biener  <rguenther@suse.de>
30909         * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30911 2017-05-19  Jason Merrill  <jason@redhat.com>
30913         * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
30915 2017-05-19  Marek Polacek  <polacek@redhat.com>
30917         PR sanitizer/80800
30918         * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
30919         TYPE_OVERFLOW_WRAPS checks.
30921 2017-05-19  Thomas Schwinge  <thomas@codesourcery.com>
30923         * tree-core.h (enum omp_clause_default_kind): Add
30924         "OMP_CLAUSE_DEFAULT_PRESENT".
30925         * tree-pretty-print.c (dump_omp_clause): Handle it.
30926         * gimplify.c (enum gimplify_omp_var_data): Add
30927         "GOVD_MAP_FORCE_PRESENT".
30928         (gimplify_adjust_omp_clauses_1): Map it to
30929         "GOMP_MAP_FORCE_PRESENT".
30930         (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
30932         * gimplify.c (oacc_default_clause): Clarify.
30934 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30936         LANG_HOOK_REGISTER_DUMPS
30937         * toplev.c (general_init): Call register dump lang hook.
30938         * doc/invoke.texi: Document -fdump-lang option family.
30939         * dumpfile.c (dump_files): Remove class dump here.
30940         (FIRST_AUTO_NUMBERED_DUMP): Adjust.
30941         * dumpfile.h (tree_dump_index): Remove TDI_class.
30942         * langhooks-def.h (lhd_register_dumps): Declare.
30943         (LANG_HOOKS_REGISTER_DUMPS): Define.
30944         (LANG_HOOKS_INITIALIZER): Add it.
30945         * langhooks.c (lhd_register_dumps): Define.
30946         * langhooks.h (struct lang_hooks): Add register_dumps.
30948 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30950         * context.h (context::set_passes): New.
30951         * context.c (context::context): Do not create pass manager.
30952         * toplev.c (general_init): Create pass manager here.
30954 2017-05-19  Segher Boessenkool  <segher@kernel.crashing.org>
30956         * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
30957         use this splitter if two add or or instructions would also work for
30958         the constant we want to generate.
30960 2017-05-19  Richard Biener  <rguenther@suse.de>
30962         PR build/80821
30963         * genmatch.c (dt_node::gen_kids_1): Add missing scope around
30964         predicate evaluation.
30966 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30968         * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
30969         add ctor.
30970         * ipa-inline.c (want_inline_small_function_p): Do not cast to
30971         unsigned.
30973 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30975         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30976         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30977         (inline_edge_summary_vec): Turn into ...
30978         (ipa_call_summaries): ... this one.
30979         (redirect_to_unreachable, edge_set_predicate,
30980         evaluate_properties_for_edge, inline_summary_alloc,
30981         reset_ipa_call_summary, reset_inline_summary,
30982         inline_summary_t::duplicate): Update.
30983         (inline_edge_duplication_hook): Turn to ...
30984         (ipa_call_summary_t::duplicate): ... this one.
30985         (inline_edge_removal_hook): Turn to ...
30986         (ipa_call_summary_t::remove): ... this one.
30987         (dump_inline_edge_summary): Turn to ...
30988         (dump_ipa_call_summary): ... this one.
30989         (estimate_function_body_sizes): Update.
30990         (inline_update_callee_summaries): Update.
30991         (remap_edge_change_prob): Update.
30992         (remap_edge_summaries): Update.
30993         (inline_merge_summary): Update.
30994         (do_estimate_edge_time): Update.
30995         (inline_generate_summary): Update.
30996         (inline_read_section): Update.
30997         (inline_read_summary): Update.
30998         (inline_free_summary): Update.
30999         * ipa-inline.c (can_inline_edge_p): Update.
31000         (compute_inlined_call_time): Update.
31001         (want_inline_small_function_p): Update.
31002         (edge_badness): Update.
31003         (early_inliner): Update.
31004         * ipa-inline.h (inline_edge_summary): Turn to ...
31005         (ipa_call_summary): ... this one.
31006         (ipa_call_summary_t): New class.
31007         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
31008         (ipa_call_summaries): New.
31009         (inline_edge_summary): Remove.
31010         (estimate_edge_growth): Update.
31011         * ipa-profile.c (ipa_propagate_frequency_1): Update.
31012         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
31013         * ipa-split.c (execute_split_functions): Update.
31014         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
31016 2017-05-19  Richard Biener  <rguenther@suse.de>
31018         PR middle-end/80764
31019         * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
31021 2017-05-18  Segher Boessenkool  <segher@kernel.crashing.org>
31023         * config/rs6000/rs6000.c (struct machine_function): Add field
31024         fpr_is_wrapped_separately.
31025         (rs6000_get_separate_components): Use 64 components.  Handle the
31026         new FPR components.
31027         (rs6000_components_for_bb): Handle the FPR components.
31028         (rs6000_emit_prologue_components): Handle the FPR components.
31029         (rs6000_emit_epilogue_components): Handle the FPR components.
31030         (rs6000_set_handled_components): Handle the FPR components.
31031         (rs6000_emit_prologue): Don't output prologue code for those FPRs
31032         that are already separately shrink-wrapped.
31033         (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
31034         that are already separately shrink-wrapped.
31036 2017-05-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
31038         PR target/80510
31039         * config/rs6000/predicates.md (simple_offsettable_mem_operand):
31040         New predicate.
31042         * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
31043         (define_peephole2 for Altivec d-form load): Add peepholes to catch
31044         cases where the register allocator uses a move and an offsettable
31045         memory operation to/from a FPR register on ISA 2.06/2.07.
31046         (define_peephole2 for Altivec d-form store): Likewise.
31048 2017-05-18  Uros Bizjak  <ubizjak@gmail.com>
31050         PR target/80799
31051         * config/i386/mmx.md (*mov<mode>_internal): Enable
31052         alternatives 11, 12, 13 and 14 also for 32bit targets.
31053         Remove alternatives 15, 16, 17 and 18.
31054         * config/i386/sse.md (vec_concatv2di): Change
31055         alternative (!x, *y) to (x, ?!*Yn).
31057 2017-05-18  Paolo Carlini  <paolo.carlini@oracle.com>
31059         * dumpfile.h (enum dump_kind): Remove stray comma.
31061 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31063         * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
31064         * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
31065         predicate::num_conditions
31066         (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
31067         (CHANGED): turn into predicate::changed.
31068         (agg_position_info): Move to ipa-predicate.h
31069         (add_condition, predicate::add_clause, predicate::operator &=,
31070         predicate::or_with, predicate::evaluate, predicate::probability,
31071         dump_condition, dump_clause, predicate::dump,
31072         predicate::remap_after_duplication, predicate::remap_after_inlining,
31073         predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
31074         (evaluate_conditions_for_known_args): Update.
31075         (set_cond_stmt_execution_predicate): Update.
31076         * ipa-inline.h: Include ipa-predicate.h
31077         (condition, inline_param_summary, conditions, agg_position_info,
31078         predicate): Move to ipa-predicate.h
31079         * ipa-predicate.c: New file.
31080         * ipa-predicate.h: New file.
31082 2017-05-18  Wilco Dijkstra  <wdijkstr@arm.com>
31084         * final.c (leaf_function_p): Check we are not in a sequence.
31086 2017-05-18  Martin Liska  <mliska@suse.cz>
31088         * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
31089         * dumpfile.c (dump_register): Use new enum dump_kind.
31090         (get_dump_file_name): Likewise.
31091         (dump_enable_all): Likewise.
31092         (dump_switch_p_1): Likewise.
31093         (enable_rtl_dump_file): Remove usage of TDF_RTL.
31094         * dumpfile.h (enum dump_kind): New enum type.
31095         (struct dump_file_info): Create constructor and
31096         format fields and comments.
31097         * passes.c (pass_manager::register_one_dump_file):
31098         Use num dump_kind.
31099         * statistics.c (statistics_early_init): Likewise.
31100         * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
31101         TDF_TREE with TDF_SLIM.
31102         (gather_memory_references_ref): Likewise.
31104 2017-05-18  Martin Liska  <mliska@suse.cz>
31106         * vec.h (struct vnull): Use it.
31108 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31110         * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
31111         (true_predicate, false_predicate, true_predicate_p,
31112         false_predicate_p): Remove.
31113         (single_cond_predicate, not_inlined_predicate): Turn to member function
31114         in ipa-inline.h
31115         (add_condition): Update.
31116         (add_clause): Turn to...
31117         (predicate::add_clause): ... this one; update; allow passing NULL
31118         as parameter.
31119         (and_predicates): Turn to ...
31120         (predicate::operator &=): ... this one.
31121         (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
31122         (or_predicates): Turn to ...
31123         (predicate::or_with): ... this one.
31124         (evaluate_predicate): Turn to ...
31125         (predicate::evaluate): ... this one.
31126         (predicate_probability): Turn to ...
31127         (predicate::probability): ... this one.
31128         (dump_condition): Update.
31129         (dump_predicate): Turn to ...
31130         (predicate::dump): ... this one.
31131         (account_size_time): Update.
31132         (edge_set_predicate): Update.
31133         (set_hint_predicate): UPdate.
31134         (evaluate_conditions_for_known_args): Update.
31135         (evaluate_properties_for_edge): Update.
31136         (remap_predicate_after_duplication): Turn to...
31137         (predicate::remap_after_duplication): ... this one.
31138         (remap_hint_predicate_after_duplication): Update.
31139         (inline_summary_t::duplicate): UPdate.
31140         (dump_inline_edge_summary): Update.
31141         (dump_inline_summary): Update.
31142         (set_cond_stmt_execution_predicate): Update.
31143         (set_switch_stmt_execution_predicate): Update.
31144         (compute_bb_predicates): Update.
31145         (will_be_nonconstant_expr_predicate): Update.
31146         (will_be_nonconstant_predicate): Update.
31147         (phi_result_unknown_predicate): Update.
31148         (predicate_for_phi_result): Update.
31149         (array_index_predicate): Update.
31150         (estimate_function_body_sizes): Update.
31151         (estimate_node_size_and_time): Update.
31152         (estimate_ipcp_clone_size_and_time): Update.
31153         (remap_predicate): Rename to ...
31154         (predicate::remap_after_inlining): ... this one.
31155         (remap_hint_predicate): Update.
31156         (inline_merge_summary): Update.
31157         (inline_update_overall_summary): Update.
31158         (estimate_size_after_inlining): Update.
31159         (read_predicate): Rename to ...
31160         (predicate::stream_in): ... this one.
31161         (read_inline_edge_summary): Update.
31162         (write_predicate): Rename to ...
31163         (predicate::stream_out): ... this one.
31164         (write_inline_edge_summary): Update.
31165         * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
31166         (clause_t): Turn to uint32_t
31167         (predicate): Turn to class; implement constructor and operators
31168         ==, !=, &
31169         (size_time_entry): Update.
31170         (inline_summary): Update.
31171         (inline_edge_summary): Update.
31173 2017-05-18  Marc Glisse  <marc.glisse@inria.fr>
31175         * fold-const.c (fold_binary_loc): Move transformation...
31176         * match.pd (C - X CMP X): ... here.
31178 2017-05-18  Sheldon Lobo  <sheldon.lobo@oracle.com>
31180         * config/sparc/sparc.c (sparc_option_override): Set function
31181         alignment for -mcpu=niagara7 to 64 to match the I$ line.
31182         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
31183         latency to 1.
31184         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
31185         latency to 2.
31186         * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
31188 2017-05-18  Marek Polacek  <polacek@redhat.com>
31190         PR sanitizer/80797
31191         * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
31192         (pass_ubsan::execute): Call gimple_assign_single_p instead of
31193         gimple_assign_load_p.
31195 2017-05-17  Segher Boessenkool  <segher@kernel.crashing.org>
31197         PR middle-end/80692
31198         * real.c (do_compare): Give decimal_do_compare preference over
31199         comparing just the signs.
31201 2017-05-17  Uros Bizjak  <ubizjak@gmail.com>
31203         * doc/md.texi (Canonicalization of Instructions): Describe the
31204         canonical form of instructions that inherently set a condition
31205         code register.
31207 2017-05-17  Peter Bergner  <bergner@vnet.ibm.com>
31209         PR middle-end/80775
31210         * tree-cfg.c: Move deletion of unreachable case statements to after
31211         the merging of consecutive case labels.
31213 2017-05-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
31215         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
31216         readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
31217         restoring of callee-saved registers.
31219 2017-05-17  Eric Botcazou  <ebotcazou@adacore.com>
31221         * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
31222         * config/visium/visium.c (single_set_and_flags): Likewise.
31223         * config/visium/visium.md (Substitutions): Likewise.
31225 2017-05-17  Martin Liska  <mliska@suse.cz>
31227         * cfg.c: Introduce dump_flags_t type and
31228         use it instead of int type.
31229         * cfg.h: Likewise.
31230         * cfghooks.c: Likewise.
31231         * cfghooks.h (struct cfg_hooks): Likewise.
31232         * cfgrtl.c: Likewise.
31233         * cfgrtl.h: Likewise.
31234         * cgraph.c (cgraph_node::get_body): Likewise.
31235         * coretypes.h: Likewise.
31236         * domwalk.c: Likewise.
31237         * domwalk.h: Likewise.
31238         * dumpfile.c (struct dump_option_value_info): Likewise.
31239         (dump_enable_all): Likewise.
31240         (dump_switch_p_1): Likewise.
31241         (opt_info_switch_p): Likewise.
31242         * dumpfile.h (enum tree_dump_index): Likewise.
31243         (struct dump_file_info): Likewise.
31244         * genemit.c: Likewise.
31245         * generic-match-head.c: Likewise.
31246         * gengtype.c (open_base_files): Likewise.
31247         * gimple-pretty-print.c: Likewise.
31248         * gimple-pretty-print.h: Likewise.
31249         * graph.c (print_graph_cfg): Likewise.
31250         * graphite-scop-detection.c (dot_all_sese): Likewise.
31251         * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31252         * loop-unroll.c (report_unroll): Likewise.
31253         * passes.c (pass_manager::register_one_dump_file): Likewise.
31254         * print-tree.c: Likewise.
31255         * statistics.c: Likewise.
31256         * tree-cfg.c: Likewise.
31257         * tree-cfg.h: Likewise.
31258         * tree-dfa.c: Likewise.
31259         * tree-dfa.h: Likewise.
31260         * tree-dump.c (dump_function): Likewise.
31261         * tree-dump.h (struct dump_info): Likewise.
31262         * tree-pretty-print.c: Likewise.
31263         * tree-pretty-print.h: Likewise.
31264         * tree-ssa-live.c: Likewise.
31265         * tree-ssa-live.h: Likewise.
31266         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31267         * tree-vect-loop.c: Likewise.
31268         * tree-vect-slp.c: Likewise.
31270 2017-05-16  James Greenhalgh  <james.greenhalgh@arm.com>
31271             Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31273         PR tree-optimization/80457
31274         * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31275         of all arguments to a statement as scalar_to_vec operations.
31276         (vectorizable_call): Adjust call to vect_model_simple_cost for
31277         new parameter.
31278         (vectorizable_conversion): Likewise.
31279         (vectorizable_assignment): Likewise.
31280         (vectorizable_shift): Likewise.
31281         (vectorizable_operation): Likewise.
31282         (vectorizable_comparison): Likewise.
31283         (vect_is_simple_cond): Record the def types for operands.
31284         (vectorizable_condition): Likewise, call vect_model_simple_cost.
31285         * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31286         for statement argument count.
31288 2017-05-16  Carl Love  <cel@us.ibm.com>
31290         * config/rs6000/rs6000-c: Add support for built-in functions
31291         vector unsigned long long vec_bperm (vector unsigned long long,
31292                                              vector unsigned char)
31293         vector signed long long vec_mule (vector signed int,
31294                                           vector signed int)
31295         vector unsigned long long vec_mule (vector unsigned int,
31296                                             vector unsigned int)
31297         vector signed long long vec_mulo (vector signed int,
31298                                           vector signed int)
31299         vector unsigned long long vec_mulo (vector unsigned int,
31300                                             vector unsigned int)
31301         vector signed char vec_sldw (vector signed char,
31302                                      vector signed char,
31303                                      const int)
31304         vector unsigned char vec_sldw (vector unsigned char,
31305                                        vector unsigned char,
31306                                        const int)
31307         vector signed short vec_sldw (vector signed short,
31308                                       vector signed short,
31309                                       const int)
31310         vector unsigned short vec_sldw (vector unsigned short,
31311                                         vector unsigned short,
31312                                         const int)
31313         vector signed int vec_sldw (vector signed int,
31314                                     vector signed int,
31315                                     const int)
31316         vector unsigned int vec_sldw (vector unsigned int,
31317                                       vector unsigned int,
31318                                       const int)
31319         vector signed long long vec_sldw (vector signed long long,
31320                                           vector signed long long,
31321                                           const int)
31322         vector unsigned long long vec_sldw (vector unsigned long long,
31323                                             vector unsigned long long,
31324                                             const int)
31325         * config/rs6000/rs6000-c: Add support for built-in functions
31326         * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31327         * config/rs6000/altivec.h: Add defintion for vec_sldw.
31328         * doc/extend.texi: Update the built-in documentation for the
31329         new built-in functions.
31331 2017-05-16  Marek Polacek  <polacek@redhat.com>
31333         PR sanitizer/80536
31334         PR sanitizer/80386
31335         * tree.c (save_expr): Don't fold the expression.
31337 2017-05-16  Uros Bizjak  <ubizjak@gmail.com>
31339         * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31340         to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31341         and (?*y,m).  Update insn attributes.
31343 2017-05-16  Martin Liska  <mliska@suse.cz>
31345         * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31346         flags argument of print_gimple_stmt, print_gimple_expr,
31347         print_generic_stmt and print_generic_expr.
31348         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31349         * coretypes.h: Likewise.
31350         * except.c (dump_eh_tree): Likewise.
31351         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31352         * gimple-pretty-print.h: Likewise.
31353         * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31354         (backprop::push_to_worklist): Likewise.
31355         (backprop::pop_from_worklist): Likewise.
31356         (backprop::process_use): Likewise.
31357         (backprop::intersect_uses): Likewise.
31358         (note_replacement): Likewise.
31359         * gimple-ssa-store-merging.c
31360         (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31361         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31362         (pass_store_merging::execute): Likewise.
31363         * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31364         (ssa_base_cand_dump_callback): Likewise.
31365         (dump_incr_vec): Likewise.
31366         (replace_refs): Likewise.
31367         (replace_mult_candidate): Likewise.
31368         (create_add_on_incoming_edge): Likewise.
31369         (create_phi_basis): Likewise.
31370         (insert_initializers): Likewise.
31371         (all_phi_incrs_profitable): Likewise.
31372         (introduce_cast_before_cand): Likewise.
31373         (replace_one_candidate): Likewise.
31374         * gimplify.c (gimplify_expr): Likewise.
31375         * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31376         (set_rename): Likewise.
31377         (rename_uses): Likewise.
31378         (copy_loop_phi_nodes): Likewise.
31379         (add_close_phis_to_merge_points): Likewise.
31380         (copy_loop_close_phi_args): Likewise.
31381         (copy_cond_phi_args): Likewise.
31382         (graphite_copy_stmts_from_block): Likewise.
31383         (translate_pending_phi_nodes): Likewise.
31384         * graphite-poly.c (print_pdr): Likewise.
31385         (dump_gbb_cases): Likewise.
31386         (dump_gbb_conditions): Likewise.
31387         (print_scop_params): Likewise.
31388         * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31389         (build_cross_bb_scalars_use): Likewise.
31390         (gather_bbs::before_dom_children): Likewise.
31391         * hsa-dump.c (dump_hsa_immed): Likewise.
31392         * ipa-cp.c (print_ipcp_constant_value): Likewise.
31393         (get_replacement_map): Likewise.
31394         * ipa-inline-analysis.c (dump_condition): Likewise.
31395         (estimate_function_body_sizes): Likewise.
31396         * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31397         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31398         * ipa-prop.c (ipa_dump_param): Likewise.
31399         (ipa_print_node_jump_functions_for_edge): Likewise.
31400         (ipa_modify_call_arguments): Likewise.
31401         (ipa_modify_expr): Likewise.
31402         (ipa_dump_param_adjustments): Likewise.
31403         (ipa_dump_agg_replacement_values): Likewise.
31404         (ipcp_modif_dom_walker::before_dom_children): Likewise.
31405         * ipa-pure-const.c (check_stmt): Likewise.
31406         (pass_nothrow::execute): Likewise.
31407         * ipa-split.c (execute_split_functions): Likewise.
31408         * omp-offload.c (dump_oacc_loop_part): Likewise.
31409         (dump_oacc_loop): Likewise.
31410         * trans-mem.c (tm_log_emit): Likewise.
31411         (tm_memopt_accumulate_memops): Likewise.
31412         (dump_tm_memopt_set): Likewise.
31413         (dump_tm_memopt_transform): Likewise.
31414         * tree-cfg.c (gimple_verify_flow_info): Likewise.
31415         (print_loop): Likewise.
31416         * tree-chkp-opt.c (chkp_print_addr): Likewise.
31417         (chkp_gather_checks_info): Likewise.
31418         (chkp_get_check_result): Likewise.
31419         (chkp_remove_check_if_pass): Likewise.
31420         (chkp_use_outer_bounds_if_possible): Likewise.
31421         (chkp_reduce_bounds_lifetime): Likewise.
31422         * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31423         (chkp_mark_completed_bounds): Likewise.
31424         (chkp_register_incomplete_bounds): Likewise.
31425         (chkp_mark_invalid_bounds): Likewise.
31426         (chkp_maybe_copy_and_register_bounds): Likewise.
31427         (chkp_build_returned_bound): Likewise.
31428         (chkp_get_bound_for_parm): Likewise.
31429         (chkp_build_bndldx): Likewise.
31430         (chkp_get_bounds_by_definition): Likewise.
31431         (chkp_generate_extern_var_bounds): Likewise.
31432         (chkp_get_bounds_for_decl_addr): Likewise.
31433         * tree-chrec.c (chrec_apply): Likewise.
31434         * tree-data-ref.c (dump_data_reference): Likewise.
31435         (dump_subscript): Likewise.
31436         (dump_data_dependence_relation): Likewise.
31437         (analyze_overlapping_iterations): Likewise.
31438         * tree-inline.c (expand_call_inline): Likewise.
31439         (tree_function_versioning): Likewise.
31440         * tree-into-ssa.c (dump_defs_stack): Likewise.
31441         (dump_currdefs): Likewise.
31442         (dump_names_replaced_by): Likewise.
31443         (dump_update_ssa): Likewise.
31444         (update_ssa): Likewise.
31445         * tree-object-size.c (pass_object_sizes::execute): Likewise.
31446         * tree-parloops.c (build_new_reduction): Likewise.
31447         (try_create_reduction_list): Likewise.
31448         (ref_conflicts_with_region): Likewise.
31449         (oacc_entry_exit_ok_1): Likewise.
31450         (oacc_entry_exit_single_gang): Likewise.
31451         * tree-pretty-print.h: Likewise.
31452         * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31453         (get_scalar_evolution): Likewise.
31454         (add_to_evolution): Likewise.
31455         (get_loop_exit_condition): Likewise.
31456         (analyze_evolution_in_loop): Likewise.
31457         (analyze_initial_condition): Likewise.
31458         (analyze_scalar_evolution): Likewise.
31459         (instantiate_scev): Likewise.
31460         (number_of_latch_executions): Likewise.
31461         (gather_chrec_stats): Likewise.
31462         (final_value_replacement_loop): Likewise.
31463         (scev_const_prop): Likewise.
31464         * tree-sra.c (dump_access): Likewise.
31465         (disqualify_candidate): Likewise.
31466         (create_access): Likewise.
31467         (reject): Likewise.
31468         (maybe_add_sra_candidate): Likewise.
31469         (create_access_replacement): Likewise.
31470         (analyze_access_subtree): Likewise.
31471         (analyze_all_variable_accesses): Likewise.
31472         (sra_modify_assign): Likewise.
31473         (initialize_constant_pool_replacements): Likewise.
31474         (find_param_candidates): Likewise.
31475         (decide_one_param_reduction): Likewise.
31476         (replace_removed_params_ssa_names): Likewise.
31477         * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31478         * tree-ssa-copy.c (dump_copy_of): Likewise.
31479         (copy_prop_visit_cond_stmt): Likewise.
31480         * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31481         * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31482         (record_equivalences_from_stmt): Likewise.
31483         * tree-ssa-dse.c (compute_trims): Likewise.
31484         (delete_dead_call): Likewise.
31485         (delete_dead_assignment): Likewise.
31486         * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31487         (forward_propagate_into_cond): Likewise.
31488         (pass_forwprop::execute): Likewise.
31489         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31490         * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31491         Likewise.
31492         (move_computations_worker): Likewise.
31493         (execute_sm): Likewise.
31494         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31495         (remove_exits_and_undefined_stmts): Likewise.
31496         (remove_redundant_iv_tests): Likewise.
31497         * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31498         (adjust_iv_update_pos): Likewise.
31499         * tree-ssa-math-opts.c (bswap_replace): Likewise.
31500         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31501         (value_replacement): Likewise.
31502         * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31503         * tree-ssa-pre.c (print_pre_expr): Likewise.
31504         (get_representative_for): Likewise.
31505         (create_expression_by_pieces): Likewise.
31506         (insert_into_preds_of_block): Likewise.
31507         (eliminate_insert): Likewise.
31508         (eliminate_dom_walker::before_dom_children): Likewise.
31509         (eliminate): Likewise.
31510         (remove_dead_inserted_code): Likewise.
31511         * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31512         * tree-ssa-reassoc.c (get_rank): Likewise.
31513         (eliminate_duplicate_pair): Likewise.
31514         (eliminate_plus_minus_pair): Likewise.
31515         (eliminate_not_pairs): Likewise.
31516         (undistribute_ops_list): Likewise.
31517         (eliminate_redundant_comparison): Likewise.
31518         (update_range_test): Likewise.
31519         (optimize_range_tests_var_bound): Likewise.
31520         (optimize_vec_cond_expr): Likewise.
31521         (rewrite_expr_tree): Likewise.
31522         (rewrite_expr_tree_parallel): Likewise.
31523         (linearize_expr): Likewise.
31524         (break_up_subtract): Likewise.
31525         (linearize_expr_tree): Likewise.
31526         (attempt_builtin_powi): Likewise.
31527         (attempt_builtin_copysign): Likewise.
31528         (transform_stmt_to_copy): Likewise.
31529         (transform_stmt_to_multiply): Likewise.
31530         (dump_ops_vector): Likewise.
31531         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31532         (print_scc): Likewise.
31533         (set_ssa_val_to): Likewise.
31534         (visit_reference_op_store): Likewise.
31535         (visit_use): Likewise.
31536         (sccvn_dom_walker::before_dom_children): Likewise.
31537         (run_scc_vn): Likewise.
31538         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31539         Likewise.
31540         (expr_hash_elt::print): Likewise.
31541         (const_and_copies::pop_to_marker): Likewise.
31542         (const_and_copies::record_const_or_copy_raw): Likewise.
31543         * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31544         * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31545         (dump_predicates): Likewise.
31546         (find_uninit_use): Likewise.
31547         (warn_uninitialized_phi): Likewise.
31548         (pass_late_warn_uninitialized::execute): Likewise.
31549         * tree-ssa.c (verify_vssa): Likewise.
31550         (verify_ssa): Likewise.
31551         (maybe_optimize_var): Likewise.
31552         * tree-vrp.c (dump_value_range): Likewise.
31553         (dump_all_value_ranges): Likewise.
31554         (dump_asserts_for): Likewise.
31555         (register_edge_assert_for_2): Likewise.
31556         (vrp_visit_cond_stmt): Likewise.
31557         (vrp_visit_switch_stmt): Likewise.
31558         (vrp_visit_stmt): Likewise.
31559         (vrp_visit_phi_node): Likewise.
31560         (simplify_cond_using_ranges_1): Likewise.
31561         (fold_predicate_in): Likewise.
31562         (evrp_dom_walker::before_dom_children): Likewise.
31563         (evrp_dom_walker::push_value_range): Likewise.
31564         (evrp_dom_walker::pop_value_range): Likewise.
31565         (execute_early_vrp): Likewise.
31567 2017-05-16  Richard Biener  <rguenther@suse.de>
31569         * dwarf2out.c (loc_list_from_tree_1): Do not create
31570         DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31572 2017-05-16  Richard Biener  <rguenther@suse.de>
31574         * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31575         just generated.
31576         (note_variable_value_in_expr): If we resolved the decl ref
31577         do not push to the stack.
31579 2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
31581         * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31582         operations in fast-math mode.
31583         (vaddq_f16): Likewise.
31584         (vmul_f16): Likewise.
31585         (vmulq_f16): Likewise.
31586         (vsub_f16): Likewise.
31587         (vsubq_f16): Likewise.
31588         * config/arm/neon.md (add<mode>3): New.
31589         (sub<mode>3): New.
31590         (fma:<VH:mode>3): New.  Also remove outdated comment.
31591         (mul<mode>3): New.
31593 2017-05-16  Martin Liska  <mliska@suse.cz>
31595         PR ipa/79849.
31596         PR ipa/79850.
31597         * ipa-devirt.c (warn_types_mismatch): Fix typo.
31598         (odr_types_equivalent_p): Likewise.
31600 2017-05-15  Sylvestre Ledru  <sylvestre@debian.org>
31602         * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31604 2017-05-15  Uros Bizjak  <ubizjak@gmail.com>
31606         PR target/80425
31607         * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31608         non-interunit SSE move alternatives with '?'.
31609         (zero-extendsidi peephole2): New peephole to skip intermediate
31610         general register in SSE zero-extend sequence.
31612 2017-05-15  Jeff Law  <law@redhat.com>
31614         * reorg.c (relax_delay_slots): Create a new variable to hold
31615         the temporary target rather than clobbering TARGET_LABEL.
31617         * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31618         missing argument to extract_bit_field call.
31619         * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31621 2017-05-15  Martin Liska  <mliska@suse.cz>
31623         PR driver/31468
31624         * gcc.c (process_command): Do not allow empty argument of -o option.
31626 2017-05-15  Renlin Li  <renlin.li@arm.com>
31628         * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31629         * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31630         * config/aarch64/constraints.md (Usf): Add long call check.
31631         * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31632         (call_value): Likewise.
31633         (sibcall): Likewise.
31634         (sibcall_value): Likewise.
31635         (call_insn): New.
31636         (call_value_insn): New.
31637         (sibcall_insn): Update rtx pattern.
31638         (sibcall_value_insn): Likewise.
31639         (call_internal): Remove.
31640         (call_value_internal): Likewise.
31641         (sibcall_internal): Likewise.
31642         (sibcall_value_internal): Likewise.
31643         (call_reg): Likewise.
31644         (call_symbol): Likewise.
31645         (call_value_reg): Likewise.
31646         (call_value_symbol): Likewise.
31648 2017-05-14  Krister Walfridsson  <krister.walfridsson@gmail.com>
31650         PR target/80600
31651         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31653 2017-05-14  Uros Bizjak  <ubizjak@gmail.com>
31655         * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31656         compatible with CCGOCmode and with CCZmode.
31658 2017-05-14  Martin Sebor  <msebor@redhat.com>
31660         PR middle-end/77671
31661         * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31662         (gimple_fold_builtin_snprintf): Same.
31663         * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31664         (gimple_fold_builtin_snprintf): Same.
31665         * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31666         of character types.
31667         (is_call_safe): New function.
31668         (try_substitute_return_value): Call it.
31669         (try_simplify_call): New function.
31670         (pass_sprintf_length::handle_gimple_call): Call it.
31672 2017-05-14  Martin Sebor  <msebor@redhat.com>
31674         PR middle-end/80669
31675         * builtins.c (expand_builtin_stpncpy): Simplify.
31677 2017-05-14  Daniel Santos  <daniel.santos@pobox.com>
31679         * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31680         * config/i386/i386.h
31681         (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31682         (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31683         (struct machine_function): Add new members call_ms2sysv,
31684         call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31685         (struct machine_frame_state): New fields sp_realigned and
31686         sp_realigned_offset.
31687         * config/i386/i386.c
31688         (enum xlogue_stub): New enum.
31689         (enum xlogue_stub_sets): New enum.
31690         (class xlogue_layout): New class.
31691         (struct ix86_frame): New fields stack_realign_allocate_offset,
31692         stack_realign_offset and outlined_save_offset.  Modify comments to
31693         detail stack layout when using out-of-line stubs.
31694         (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31695         (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31696         -mcall-ms2sysv-xlogues.
31697         (stub_managed_regs): New static variable.
31698         (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31699         registers managed by out-of-line stub.
31700         (disable_call_ms2sysv_xlogues): New function.
31701         (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31702         m->call_ms2sysv when appropriate and compute frame layout for
31703         out-of-line stubs.
31704         (sp_valid_at, fp_valid_at): New inline functions.
31705         (choose_basereg): New function.
31706         (choose_baseaddr): Add align parameter, use choose_basereg and modify
31707         all callers.
31708         (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31709         Use align parameter of choose_baseaddr to generated aligned SSE movs
31710         when possible.
31711         (pro_epilogue_adjust_stack): Modify to track
31712         machine_frame_state::sp_realigned.
31713         (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31714         (ix86_nsaved_sseregs): Likewise.
31715         (ix86_emit_save_regs): Likewise.
31716         (ix86_emit_save_regs_using_mov): Likewise.
31717         (ix86_emit_save_sse_regs_using_mov): Likewise.
31718         (get_scratch_register_on_entry): Likewise.
31719         (gen_frame_set): New function.
31720         (gen_frame_load): Likewise.
31721         (gen_frame_store): Likewise.
31722         (emit_outlined_ms2sysv_save): Likewise.
31723         (emit_outlined_ms2sysv_restore): Likewise.
31724         (ix86_expand_prologue): Modify stack re-alignment code and call
31725         emit_outlined_ms2sysv_save when appropriate.
31726         (ix86_emit_leave): Clear machine_frame_state::sp_realigned.  Add
31727         parameter rtx_insn *insn, which allows the function to be used to only
31728         generate the notes.
31729         (ix86_expand_epilogue): Modify validity checks of frame and stack
31730         pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31731         (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31732         * config/i386/predicates.md
31733         (save_multiple): New predicate.
31734         (restore_multiple): Likewise.
31735         * config/i386/sse.md
31736         (save_multiple<mode>): New pattern.
31737         (save_multiple_realign<mode>): Likewise.
31738         (restore_multiple<mode>): Likewise.
31739         (restore_multiple_and_return<mode>): Likewise.
31740         (restore_multiple_leave_return<mode>): Likewise.
31741         * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31743 2017-05-14  Julia Koval  <julia.koval@intel.com>
31745         * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31746         * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31747         (__builtin_ia32_xsetbv): New builtins.
31748         * config/i386/i386.c (ix86_expand_special_args_builtin):
31749         Process new types.
31750         (ix86_expand_builtin): Special expand for new intrinsics.
31751         * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31752         (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31753         * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31755 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31757         * cfganal.c (inverted_post_order_compute): Change argument type
31758         to vec *.
31759         * cfganal.h (inverted_post_order_compute): Adjust prototype.
31760         * df-core.c (rest_of_handle_df_initialize): Adjust.
31761         (rest_of_handle_df_finish): Likewise.
31762         (df_analyze_1): Likewise.
31763         (df_analyze): Likewise.
31764         (loop_inverted_post_order_compute): Change argument to be a vec *.
31765         (df_analyze_loop): Adjust.
31766         (df_get_n_blocks): Likewise.
31767         (df_get_postorder): Likewise.
31768         * df.h (struct df_d): Change field to be a vec.
31769         * lcm.c (compute_laterin): Adjust.
31770         (compute_available): Likewise.
31771         * lra-lives.c (lra_create_live_ranges_1): Likewise.
31772         * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31773         * tree-ssa-pre.c (compute_antic): Likewise.
31775 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31777         * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31778         (depth_first_search::depth_first_search): Change structure init
31779         function to this constructor.
31780         (depth_first_search::add_bb): Rename function to this member.
31781         (depth_first_search::execute): Likewise.
31782         (flow_dfs_compute_reverse_finish): Adjust.
31784 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31786         * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31787         (longest_simple_path): Likewise.
31788         * shrink-wrap.c (spread_components): Likewise.
31789         (disqualify_problematic_components): Likewise.
31790         (emit_common_heads_for_components): Likewise.
31791         (emit_common_tails_for_components): Likewise.
31792         (insert_prologue_epilogue_for_components): Likewise.
31794 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31796         * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31797         auto_sbitmap.
31799 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31801         * df-core.c (df_set_blocks): Start using auto_bitmap.
31802         (df_compact_blocks): Likewise.
31803         * df-problems.c (df_rd_confluence_n): Likewise.
31804         * df-scan.c (df_insn_rescan_all): Likewise.
31805         (df_process_deferred_rescans): Likewise.
31806         (df_update_entry_block_defs): Likewise.
31807         (df_update_exit_block_uses): Likewise.
31808         (df_entry_block_bitmap_verify): Likewise.
31809         (df_exit_block_bitmap_verify): Likewise.
31810         (df_scan_verify): Likewise.
31811         * lra-constraints.c (lra_constraints): Likewise.
31812         (undo_optional_reloads): Likewise.
31813         (lra_undo_inheritance): Likewise.
31814         * lra-remat.c (calculate_gen_cands): Likewise.
31815         (do_remat): Likewise.
31816         * lra-spills.c (assign_spill_hard_regs): Likewise.
31817         (spill_pseudos): Likewise.
31818         * tree-ssa-pre.c (bitmap_set_and): Likewise.
31819         (bitmap_set_subtract_values): Likewise.
31821 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31823         * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31824         management with auto_bitmap.
31825         (fix_inter_tick): Likewise.
31826         (fix_recovery_deps): Likewise.
31827         * ira.c (add_store_equivs): Likewise.
31828         (find_moveable_pseudos): Likewise.
31829         (split_live_ranges_for_shrink_wrap): Likewise.
31830         * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31831         (rtx_reuse_manager::seen_def_p): Likewise.
31832         (rtx_reuse_manager::set_seen_def): Likewise.
31833         * print-rtl.h (class rtx_reuse_manager): Likewise.
31835 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31837         * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31838         lifetime.
31839         (migrate_btr_def): Likewise.
31840         * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31841         * df-core.c (loop_post_order_compute): Likewise.
31842         (loop_inverted_post_order_compute): Likewise.
31843         * hsa-common.h: Likewise.
31844         * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31845         * init-regs.c (initialize_uninitialized_regs): Likewise.
31846         * ipa-inline.c (resolve_noninline_speculation): Likewise.
31847         (inline_small_functions): Likewise.
31848         * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31849         * ira.c (combine_and_move_insns): Likewise.
31850         (build_insn_chain): Likewise.
31851         * loop-invariant.c (find_invariants): Likewise.
31852         * lower-subreg.c (propagate_pseudo_copies): Likewise.
31853         * predict.c (tree_predict_by_opcode): Likewise.
31854         (predict_paths_leading_to): Likewise.
31855         (predict_paths_leading_to_edge): Likewise.
31856         (estimate_loops_at_level): Likewise.
31857         (estimate_loops): Likewise.
31858         * shrink-wrap.c (try_shrink_wrapping): Likewise.
31859         (spread_components): Likewise.
31860         * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31861         * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31862         * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31863         * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31864         * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31865         * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31866         * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31867         * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31868         (mark_threaded_blocks): Likewise.
31869         (thread_through_all_blocks): Likewise.
31870         * tree-ssa.c (verify_ssa): Likewise.
31871         (execute_update_addresses_taken): Likewise.
31872         * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31874 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31876         * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31877         auto_vec.
31878         (post_order_compute): Likewise.
31879         (inverted_post_order_compute): Likewise.
31880         (pre_and_rev_post_order_compute_fn): Likewise.
31882 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31884         * genrecog.c (int_set::int_set): Explicitly construct our
31885         auto_vec base class.
31886         * vec.h (auto_vec::auto_vec): New constructor.
31888 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31890         * bitmap.h (class auto_bitmap): New constructor taking
31891         bitmap_obstack * argument.
31893 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31895         * bitmap.h (class auto_bitmap): Change type of m_bits to
31896         bitmap_head, and adjust ctor / dtor and member operators.
31898 2017-05-13  Uros Bizjak  <ubizjak@gmail.com>
31900         * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31901         when returned register mode doesn't match original mode.
31903 2017-05-12  Jeff Law  <law@redhat.com>
31904             Jakub Jelinek  <jakub@redhat.com>
31906         * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31907         we look for cc setter after the compare-elim changes.
31908         * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31909         within the vector to match what compare-elim now expects.
31910         (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31911         (xorsi3_flags, one_cmplsi2_flags): Likewise.
31913         * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
31914         after the compare-elim changes.
31915         * config/rx/rx.md (abssi2_flags): Fix order of patterns within
31916         the vector to match what compare-elim now expects.
31917         (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
31918         (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
31919         (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
31920         (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
31921         (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
31923         * config/visium/visium.c (single_set_and_flags): Fix where
31924         we look for cc setter after the compare-elim changes.
31925         * config/visium/visium.md (flags_subst_logic): Fix order of patterns
31926         with the vector to match what compare-elim now expects.
31927         (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
31928         (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
31929         (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
31930         (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
31931         (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
31932         (neg<mode>2_insn_set_overflow): Likewise.
31934 2017-05-12  Jim Wilson  <jim.wilson@linaro.org>
31936         PR middle-end/79794
31937         * expmed.c (extract_bit_field_1): Add alt_rtl argument.  Before
31938         maybe_expand_insn call, set ops[0].target.  If still set after call,
31939         set alt_rtl.  Add extra arg to recursive calls.
31940         (extract_bit_field): Add alt_rtl argument.  Pass to
31941         extract_bit_field.
31942         * expmed.h (extract_bit_field): Fix prototype.
31943         * expr.c (emit_group_load_1, copy_blkmode_from_reg)
31944         (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
31945         to extract_bit_field_calls.
31946         (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
31947         Pass alt_rtl to extract_bit_field calls.
31948         * calls.c (store_unaligned_arguments_into_psuedos)
31949         load_register_parameters): Pass extra NULL to extract_bit_field calls.
31950         * optabs.c (maybe_legitimize_operand): Clear op->target when call
31951         gen_reg_rtx.
31952         * optabs.h (struct expand_operand): Add target bitfield.
31954 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31956         * compare-elim.c (try_eliminate_compare): Canonicalize
31957         operation with embedded compare to
31958         [(set (reg:CCM) (compare:CCM (operation) (immediate)))
31959          (set (reg) (operation)].
31961         * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
31963 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31965         PR target/80723
31966         * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
31967         cost of adding a carry flag for ADC instruction.
31968         [case MINUS]: Ignore the cost of subtracting a carry flag
31969         for SBB instruction.
31971 2017-05-12  Steven Munroe  <munroesj@gcc.gnu.org>
31973         * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
31974         and x86intrin.h
31975         * config/rs6000/bmiintrin.h: New file.
31976         * config/rs6000/bmi2intrin.h: New file.
31977         * config/rs6000/x86intrin.h: New file.
31979 2017-05-12  Jeff Law  <law@redhat.com>
31981         * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
31982         markers.
31984 2017-05-12  Peter Bergner  <bergner@vnet.ibm.com>
31986         PR middle-end/80707
31987         * tree-cfg.c: Remove cfg edges of unreachable case statements.
31989 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31991         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31992         early expansion of vector divide builtins.
31993         (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
31994         builtins identified as having unsigned arguments.
31996 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31998         * config/rs6000/rs6000.c (gimple-fold.h): New #include.
31999         (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
32000         expansion of vector logical operations (and, andc, or, xor,
32001         nor, orc, nand).
32003 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32005         * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
32006         * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
32008 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
32010         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
32011         early GIMPLE expansion of vector multiplies.
32013 2017-05-12  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32015         * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
32016         TARGET_HAVE_MOVT conditional.
32017         (movt splitter): Likewise.
32019 2017-05-12  Richard Biener  <rguenther@suse.de>
32021         * tree-ssa-sccvn.h (has_VN_INFO): Declare.
32022         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32023         Fold all stmts not inplace.
32025 2017-05-12  Richard Biener  <rguenther@suse.de>
32027         PR tree-optimization/80713
32028         * tree-ssa-pre.c (remove_dead_inserted_code): Clear
32029         inserted_exprs bit for not removed stmts.
32031 2017-05-12  Thomas Schwinge  <thomas@codesourcery.com>
32033         PR middle-end/69921
32034         * tree-parloops.c (create_parallel_loop): Set "oacc kernels
32035         parallelized" attribute for parallelized OpenACC kernels.
32036         * omp-offload.c (execute_oacc_device_lower): Use it.
32038         * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
32039         Set "oacc kernels" attribute.
32040         * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
32041         parameter.  Adjust all users.
32042         (oacc_fn_attrib_kernels_p): Remove function.
32043         * omp-offload.c (execute_oacc_device_lower): Look for "oacc
32044         kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
32045         * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
32046         * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
32047         assert "oacc kernels" attribute is set.
32049 2017-05-11  Carl Love  <cel@us.ibm.com>
32051         * config/rs6000/rs6000-c: Add support for built-in functions
32052         vector unsigned char vec_popcnt (vector signed char)
32053         vector unsigned char vec_popcnt (vector unsigned char)
32054         vector unsigned short vec_popcnt (vector signed short)
32055         vector unsigned short vec_popcnt (vector unsigned short)
32056         vector unsigned int vec_popcnt (vector signed int)
32057         vector unsigned int vec_popcnt (vector unsigned int)
32058         vector unsigned long long vec_popcnt (vector signed long long)
32059         vector unsigned long long vec_popcnt (vector unsigned long long)
32060         vector signed long long vec_slo (vector signed long long,
32061                                          vector signed char)
32062         vector signed long long vec_slo (vector signed long long,
32063                                          vector unsigned char)
32064         vector unsigned long long vec_slo (vector unsigned long long,
32065                                            vector signed char)
32066         vector unsigned long long vec_slo (vector unsigned long long,
32067                                            vector unsigned char)
32068         * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
32069         VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
32070         * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
32071         vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
32072         * doc/extend.texi: Update the built-in documentation file for the
32073         new built-in functions.
32075 2017-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
32077         * attribs.h (sorted_attr_string): Move machine independent
32078         functions for target clone support from the i386 port to common
32079         code.  Rename ix86_function_versions to common_function_versions.
32080         Rename make_name to make_unique_name.
32081         (common_function_versions): Likewise.
32082         (make_unique_name): Likewise.
32083         (make_dispatcher_decl): Likewise.
32084         (is_function_default_version): Likewise.
32085         * attribs.c (attr_strcmp): Likewise.
32086         (sorted_attr_string): Likewise.
32087         (common_function_versions): Likewise.
32088         (make_unique_name): Likewise.
32089         (make_dispatcher_decl): Likewise.
32090         (is_function_default_version): Likewise.
32091         * config/i386/i386.c (attr_strcmp): Likewise.
32092         (sorted_attr_string): Likewise.
32093         (ix86_function_versions): Likewise.
32094         (make_name): Likewise.
32095         (make_dispatcher_decl): Likewise.
32096         (is_function_default_version): Likewise.
32097         (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
32099 2017-05-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32101         PR target/80695
32102         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
32103         Account for direct move costs for vec_construct of integer
32104         vectors.
32106 2017-05-11  Uros Bizjak  <ubizjak@gmail.com>
32108         PR target/80706
32109         * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
32110         (UNSPEC_STX_ATOMIC): Ditto.
32111         (loaddi_via_sse): New insn.
32112         (storedi_via_sse): Ditto.
32113         (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
32114         Update corresponding peephole2 patterns.
32115         (atomic_storedi_fpu): Ditto.
32117 2017-05-11  Julia Koval  <julia.koval@intel.com>
32119         * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
32120         (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
32121         New intrinsics.
32122         * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
32123         (__builtin_ia32_rsqrt14ss_mask): New builtins.
32124         * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
32126 2017-05-11  Nathan Sidwell  <nathan@acm.org>
32128         * graphite-poly.c: Include dumpfile.h.
32130         * dumpfle.h (dump_function): Declare here ...
32131         * tree-dump.h (dump_function): ... not here.
32132         * dumpfile.c: #include tree-cfg.h.
32133         (dump_function): Move here from ...
32134         * tree-dump.c (dump_function): ... here.
32135         * gimplify.c: #include splay-tree.h, not tree-dump.h.
32136         * graphite-poly.c: Don't include tree-dump.h.
32137         * cgraphclones.c: Include dumpfile.h not tree-dump.h.
32138         * print-tree.c: Likewise.
32139         * stor-layout.c: Likewise.
32140         * tree-nested.c: Likewise.
32142         * dumpfile.c (dump_start): Use TDF_FLAGS.
32143         (dump_enable_all): Fix TDF_KIND check thinko.
32145 2017-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32147         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
32148         array entries to represent two legal parameterizations of the
32149         overloaded __builtin_cmpb function, as represented by the
32150         P6_OV_BUILTIN_CMPB constant.
32151         (altivec_resolve_overloaded_builtin): Add special case handling
32152         for the __builtin_cmpb function, as represented by the
32153         P6_OV_BUILTIN_CMPB constant.
32154         * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
32155         (BU_P6_64BIT_2): New macro.
32156         (BU_P6_OVERLOAD_2): New macro
32157         (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
32158         (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
32159         (CMPB): Add overload support to represent both 32-bit and 64-bit
32160         compare-bytes function.
32161         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
32162         support for TARGET_CMPB.
32163         * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
32164         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
32165         documentation of the __builtin_cmpb overloaded built-in function.
32167 2017-05-11  Richard Biener  <rguenther@suse.de>
32169         PR tree-optimization/80705
32170         * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
32171         bases are not vectorizable.
32173 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32175         * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
32176         when counting register pressure.
32178 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32180         * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
32181         (iv_ca_more_deps): Renamed to ...
32182         (iv_ca_compare_deps): ... this.
32183         (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
32185 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32187         * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
32188         to ...
32189         (determine_group_iv_costs): ... here.
32190         (find_inv_vars_cb): Record inv var if it's not recorded before.
32192 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32194         * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
32195         (get_shiftadd_cost): Ditto.
32197 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32199         * tree-ssa-address.c: Include header file.
32200         (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
32201         address.
32202         (add_to_parts): Refactor.
32203         (addr_to_parts): New parameter.  Update use of move_hint_to_base.
32204         (create_mem_ref): Update use of addr_to_parts.  Re-associate addr
32205         in new order.
32207 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32209         PR tree-optimization/53090
32210         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
32211         COMP_IV_EXPR_2.
32212         (extract_cond_operands): Detect condition with IV on both sides
32213         and return COMP_IV_EXPR_2.
32214         (find_interesting_uses_cond): Add iv_use for both IVs in condition.
32215         (rewrite_use_compare): Simplify by removing call to function
32216         extract_cond_operands.
32218 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32220         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
32221         (extract_cond_operands): Detect condition comparing against non-
32222         invariant bound and return appropriate enum value.
32223         (find_interesting_uses_cond): Update use of extract_cond_operands.
32224         Handle its return value accordingly.
32225         (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
32227 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32229         * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
32230         nonlinear iv_use computation in loop invariant sensitive way.
32232 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32234         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
32235         (find_iv_candidates): Call relate_compare_use_with_all_cands.
32237 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32239         * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32240         (dump_cand): Support iv_cand.inv_exprs.
32241         (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32242         for candidates.
32243         (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32244         iv_cand.inv_exprs.
32246 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32248         * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32249         from ...
32250         * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32251         as local function.  Include necessary header files.
32252         * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32254 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32256         * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32258 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32260         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32261         operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32262         RSHIFT_EXPR and BIT_NOT_EXPR.
32264 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32266         * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32267         (adjust_setup_cost): New parameter supporting round up adjustment.
32268         (struct address_cost_data): Delete.
32269         (force_expr_to_var_cost): Don't bound cost with spill_cost.
32270         (split_address_cost, ptr_difference_cost): Delete.
32271         (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32272         (struct ainc_cost_data): New struct.
32273         (get_address_cost_ainc): New function.
32274         (get_address_cost, get_computation_cost): Reimplement.
32275         (determine_group_iv_cost_address): Record inv_expr for all uses of
32276         a group.
32277         (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32278         (iv_ca_has_deps): Reimplemented to ...
32279         (iv_ca_more_deps): ... this.  Check if NEW_CP introduces more deps
32280         than OLD_CP.
32281         (iv_ca_extend): Call iv_ca_more_deps.
32283 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32285         * tree-ssa-address.c (struct mem_address): Move to header file.
32286         (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32287         * tree-ssa-address.h (struct mem_address): Move from C file.
32288         (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32290 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32292         * tree-affine.h (aff_combination_type): New interface.
32293         (aff_combination_zero_p): Remove static.
32294         (aff_combination_const_p): New interface.
32295         (aff_combination_singleton_var_p): New interfaces.
32297 2017-05-11  Richard Biener  <rguenther@suse.de>
32299         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32300         Skip unreachable blocks and destinations.
32301         (eliminate): Move stmt removal and fixup ...
32302         (fini_eliminate): ... here.  Skip inserted exprs.
32303         (pass_pre::execute): Move fini_pre after fini_eliminate.
32304         * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32305         (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32306         PRE to get rid of dead code that has invalid SSA form and
32307         split critical edges again.
32309 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32311         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32313 2017-05-11  Richard Biener  <rguenther@suse.de>
32315         * passes.c (execute_function_todo): Verify loops if they are
32316         said to be up-to-date.
32317         * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32318         * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32320 2017-05-10  John David Anglin  <danglin@gcc.gnu.org>
32322         PR target/80090
32323         * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32324         handle calling assemble_external ourself.
32326         PR target/79027
32327         * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32328         modes with zero size.  Enhance comment.
32330 2017-05-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32332         * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32333         built-ins for vec_xl and vec_xst with short and char pointer
32334         arguments.
32336 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32338         * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32339         (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32340         (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32341         (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32342         (_mm_maskz_min_round_ss): New intrinsics.
32343         * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32344         (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32345         * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32346         (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32347         (__builtin_ia32_minss_mask_round): New builtins.
32348         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32349         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32350         * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32351         Rename to ...
32352         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32353         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32354         Change to ...
32355         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32356         ... this.
32358 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32360         * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32361         (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32362         (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32363         (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32364         (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32365         (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32366         (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32367         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32368         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32369         * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32370         (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32371         (__builtin_ia32_mulss_mask_round): New builtins.
32372         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32373         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32374         * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32375         Rename to ...
32376         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32377         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32378         Change to ...
32379         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32380         ... this.
32382 2017-05-10  Julia Koval  <julia.koval@intel.com>
32384         * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32385         (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32386         (_mm256_setr_m128i): New intrinsics.
32388 2017-05-10  Julia Koval  <julia.koval@intel.com>
32390         * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32391         (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32392         (_mm_maskz_rcp14_ss): New intrinsics.
32393         * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32394         (__builtin_ia32_rcp14ss_mask): New builtins.
32395         * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32397 2017-05-10  Peter Bergner  <bergner@vnet.ibm.com>
32399         PR tree-optimization/51513
32400         * tree-cfg.c (gimple_seq_unreachable_p): New function.
32401         (assert_unreachable_fallthru_edge_p): Use it.
32402         (group_case_labels_stmt): Likewise.
32403         * tree-cfg.h: Prototype it.
32404         * stmt.c: Include cfghooks.h and tree-cfg.h.
32405         (emit_case_dispatch_table) <gap_label>: New local variable.
32406         Use it to fill dispatch table gaps.
32407         Test for default_label before updating probabilities.
32408         (expand_case) <default_label>: Remove unneeded initialization.
32409         Test for unreachable default case statement and remove its edge.
32410         Set default_label accordingly.
32411         * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32413 2017-05-10  Carl Love  <cel@us.ibm.com>
32415         * config/rs6000/rs6000-c: Add support for built-in functions
32416         vector signed char      vec_neg (vector signed char)
32417         vector signed short int vec_neg (vector short int)
32418         vector signed int       vec_neg (vector signed int)
32419         vector signed long long vec_neg (vector signed long long)
32420         vector float            vec_neg (vector float)
32421         vector double           vec_neg (vector double)
32422         * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32423         overload.
32424         * config/rs6000/altivec.h: Add define for vec_neg
32425         * doc/extend.texi: Update the built-in documentation for the
32426         new built-in functions.
32428 2017-05-10  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32430         PR tree-optimization/77644
32431         * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32433 2017-05-10  Nathan Sidwell  <nathan@acm.org>
32435         * dumpfile.h (TDI_lang_all): New.
32436         (TDF_KIND): New. Renumber others
32437         (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32438         than bits.
32439         * dumpfile.c (dump_files): Mark language dumps as TDF_LANG.  add
32440         lang-all.
32441         (get_dump_file_name): Adjust suffix generation.
32442         (dump_enable_all): Use TDF_KIND.
32443         * doc/invoke.texi (-fdump-lang-all): Document.
32445         * dumpfile.h: Tabify.
32447 2017-05-10  Wilco Dijkstra  <wdijkstr@arm.com>
32449         PR target/80671
32450         * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32451         Move member access before delete.
32453 2017-05-10  Alexandre Oliva  <aoliva@redhat.com>
32455         * tree-inline.c (expand_call_inline): Split block at stmt
32456         before the call.
32458 2017-05-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
32460         PR target/68163
32461         * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32462         are now unused after splitting mov{sf,sd}_hardfloat.
32463         (f32_lr2): Likewise.
32464         (f32_lm): Likewise.
32465         (f32_lm2): Likewise.
32466         (f32_li): Likewise.
32467         (f32_li2): Likewise.
32468         (f32_lv): Likewise.
32469         (f32_sr): Likewise.
32470         (f32_sr2): Likewise.
32471         (f32_sm): Likewise.
32472         (f32_sm2): Likewise.
32473         (f32_si): Likewise.
32474         (f32_si2): Likewise.
32475         (f32_sv): Likewise.
32476         (f32_dm): Likewise.
32477         (f32_vsx): Likewise.
32478         (f32_av): Likewise.
32479         (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32480         For movsf, order stores so the VSX stores occur before the GPR
32481         store which encourages the register allocator to use a traditional
32482         FPR instead of a GPR.  For movsd, order the stores so that the GPR
32483         store comes before the VSX stores to allow the power6 to work.
32484         This is due to the power6 not having a 32-bit integer store
32485         instruction from a FPR.
32486         (movsf_hardfloat): Likewise.
32487         (movsd_hardfloat): Likewise.
32489 2017-05-09  Martin Sebor  <msebor@redhat.com>
32491         PR translation/80280
32492         * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32493         added in r247778.
32495         PR translation/80280
32496         * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32497         data member added in r247778.
32498         (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32500 2017-05-09  Nathan Sidwell  <nathan@acm.org>
32502         * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32504         * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32505         typedefs.
32507 2017-05-09  Marek Polacek  <polacek@redhat.com>
32509         * doc/invoke.texi: Fix typo.
32511 2017-05-09  Richard Biener  <rguenther@suse.de>
32513         * tree-vrp.c (vrp_val_is_max): Adjust comment.
32514         (vrp_val_is_min): Likewise.
32515         (set_value_range_to_value): Likewise.
32516         (set_value_range_to_nonnegative): Likewise.
32517         (gimple_assign_nonzero_p): Likewise.
32518         (gimple_stmt_nonzero_p): Likewise.
32519         (vrp_int_const_binop): Likewise.  Remove unreachable case.
32520         (adjust_range_with_scev): Adjust comments.
32521         (compare_range_with_value): Likewise.
32522         (extract_range_from_phi_node): Likewise.
32523         (test_for_singularity): Likewise.
32525 2017-05-09  Richard Biener  <rguenther@suse.de>
32527         * tree-vrp.c (get_single_symbol): Add assert that we don't
32528         get overflowed constants as invariant part.
32529         (compare_values_warnv): Add comment before the TREE_NO_WARNING
32530         checks.  Use wi::cmp instead of recursing for integer constants.
32531         (compare_values): Just ignore whether we assumed undefined
32532         overflow instead of failing the compare.
32533         (extract_range_for_var_from_comparison_expr): Add comment before the
32534         TREE_NO_WARNING sets.
32535         (test_for_singularity): Likewise.
32536         (extract_range_from_comparison): Do not disable optimization
32537         when we assumed undefined overflow.
32538         (extract_range_basic): Remove init of unused var.
32540 2017-05-09  Richard Biener  <rguenther@suse.de>
32542         * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32543         (extract_range_from_multiplicative_op_1): Adjust.
32544         (extract_range_from_binary_expr_1): Use int_const_binop.
32546 2017-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32548         PR target/80101
32549         * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32550         rs6000_store_data_bypass_p in seven define_bypass directives and
32551         in several comments.
32552         * config/rs6000/rs6000-protos.h: Add prototype for
32553         rs6000_store_data_bypass_p function.
32554         * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32555         function implements slightly different (rs6000-specific) semantics
32556         than store_data_bypass_p, returning false rather than aborting
32557         with assertion error when arguments do not satisfy the
32558         requirements of store data bypass.
32559         (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32560         rs6000_store_data_bypass_p.
32562 2017-05-08  Max Filippov  <jcmvbkbc@gmail.com>
32564         * config/xtensa/xtensa-protos.h
32565         (xtensa_initial_elimination_offset): New declaration.
32566         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32567         New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32568         macro definition, add case for FRAME_POINTER_REGNUM when
32569         FRAME_GROWS_DOWNWARD.
32570         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32571         (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32572         xtensa_initial_elimination_offset.
32574 2017-05-08  Nathan Sidwell  <nathan@acm.org>
32576         * doc/invoke.texi: Alphabetize -fdump options.
32578 2017-05-08  Martin Sebor  <msebor@redhat.com>
32580         PR translation/80280
32581         * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32583 2017-05-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
32585         * target.def (compute_frame_layout): New optional target hook.
32586         * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32587         * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32588         * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32589         target hook.
32590         * reload1.c (verify_initial_elim_offsets): Likewise.
32591         * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32592         (use_simple_return_p): Call arm_compute_frame_layout if needed.
32593         (arm_get_frame_offsets): Split up into this ...
32594         (arm_compute_frame_layout): ... and this function.
32596 2017-05-08  Richard Sandiford  <richard.sandiford@arm.com>
32598         * config/aarch64/constraints.md (Usa): New constraint.
32599         * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32601 2017-05-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32603         * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32604         with_multilib_list after it has been checked.
32606 2017-05-08  Richard Biener  <rguenther@suse.de>
32608         * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32609         (bitmap_set_subtract_values): Likewise.
32611 2017-05-08  Richard Biener  <rguenther@suse.de>
32613         * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32614         (gimple_assign_nonzero): ... this and remove strict_overflow_p
32615         argument.
32616         (gimple_stmt_nonzero_warnv_p): Rename to ...
32617         (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32618         argument.
32619         (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32620         (extract_range_basic): Adjust, do not disable propagation on
32621         strict overflow sensitive simplification.
32622         (vrp_visit_cond_stmt): Likewise.
32624 2017-05-05  Jan Hubicka  <hubicka@ucw.cz>
32626         * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32627         body size unconditionally.
32629 2017-05-07  Jeff Law  <law@redhat.com>
32631         Revert:
32632         2017-05-06  Jeff Law  <law@redhat.com>
32633         PR tree-optimization/78496
32634         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32635         code.
32637         PR tree-optimization/78496
32638         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32639         (simplify_stmt_using_ranges): Call it.
32640         (vrp_dom_walker::before_dom_children): Extract equivalences
32641         from an ASSERT_EXPR with an equality comparison against a
32642         constant.
32644 2017-05-06  Jeff Law  <law@redhat.com>
32646         PR tree-optimization/78496
32647         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32648         code.
32650         PR tree-optimization/78496
32651         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32652         (simplify_stmt_using_ranges): Call it.
32653         (vrp_dom_walker::before_dom_children): Extract equivalences
32654         from an ASSERT_EXPR with an equality comparison against a
32655         constant.
32657 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32659         * lra-constraints.c (lra_copy_reg_equiv): New function.
32660         (split_reg): Use it to copy equivalence information from the
32661         original register to the spill register.
32663 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32665         PR rtl-optimization/75964
32666         * simplify-rtx.c (simplify_const_relational_operation): Remove
32667         invalid handling of comparisons of integer ABS.
32669 2017-05-06  Uros Bizjak  <ubizjak@gmail.com>
32671         * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32672         initialize to zero.
32673         (init_regs): Remove declaration.
32674         (function_arg_advance_32): Initialize error_p as boolean variable.
32676 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32678         * store-motion.c (remove_reachable_equiv_notes): Reformat long
32679         lines.  Use for (;;).
32681 2017-05-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32683         * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32684         (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32685         (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32686         (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32687         VF=2 that require versioning.
32689 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32691         * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32692         int.
32694 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32696         * diagnostic.h (diagnostic_override_option_index): Convert from
32697         macro to inline function.
32699 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32701         * diagnostic.c (last_module_changed_p): New function.
32702         (set_last_module): New function.
32703         (diagnostic_report_current_module): Convert macro usage to
32704         the above functions.
32705         * diagnostic.h (diagnostic_context::last_module): Strengthen
32706         from const line_map * to const line_map_ordinary *.
32707         (diagnostic_last_module_changed): Delete macro.
32708         (diagnostic_set_last_module): Delete macro.
32710 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32712         * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32713         with diagnostic_report_diagnostic.
32714         (diagnostic_n_impl_richloc): Likewise.
32715         * diagnostic.h (report_diagnostic): Delete macro.
32716         * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32717         with diagnostic_report_diagnostic.
32718         * substring-locations.c (format_warning_va): Likewise.
32720 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32722         * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32723         save/restor of format_spec.  Move option-printing code to...
32724         (print_option_information): ...this new function, and
32725         reimplement by simply printing to the pretty_printer,
32726         rather than appending to the format string.
32728 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32730         * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32731         handling logic into...
32732         (update_effective_level_from_pragmas): ...this new function.
32734 2017-05-04  Andrew Waterman  <andrew@sifive.com>
32736         * config/riscv/riscv.opt (mstrict-align): New option.
32737         * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it.  Update comment.
32738         (SLOW_UNALIGNED_ACCESS): Define.
32739         (riscv_slow_unaligned_access): Declare.
32740         * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32741         field.
32742         (riscv_slow_unaligned_access): New variable.
32743         (rocket_tune_info): Set slow_unaligned_access to true.
32744         (optimize_size_tune_info): Set slow_unaligned_access to false.
32745         (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32746         (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32747         (riscv_option_override): Set riscv_slow_unaligned_access.
32748         * doc/invoke.texi: Add -mstrict-align to RISC-V.
32750 2017-05-04  Kito Cheng  <kito.cheng@gmail.com>
32752         * config/riscv/riscv.md: Unify indentation.
32754 2017-05-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
32756         PR target/79038
32757         PR target/79202
32758         PR target/79203
32759         * config/rs6000/rs6000.md (u code attribute): Add FIX and
32760         UNSIGNED_FIX.
32761         (extendsi<mode>2): Add support for doing sign extension via
32762         VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32763         don't have ISA 3.0 instructions.
32764         (extendsi<mode>2 splitter): Likewise.
32765         (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32766         generate the normal insns since SImode can now go in vector
32767         registers.  Disallow the special UNSPECs needed for previous
32768         machines to hide SImode being used.  Add new insns
32769         fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32770         (fix_trunc<mode>si2_stfiwx): Likewise.
32771         (fix_trunc<mode>si2_internal): Likewise.
32772         (fixuns_trunc<mode>si2): Likewise.
32773         (fixuns_trunc<mode>si2_stfiwx): Likewise.
32774         (fctiw<u>z_<mode>_smallint): Likewise.
32775         (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32776         of floating point to 32-bit integer from doing a direct move to
32777         the GPR registers to do a store.
32778         (fctiwz_<mode>): Break long line.
32780 2017-05-05  Bin Cheng  <bin.cheng@arm.com>
32782         * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32783         * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32784         (addr_list, addr_offset_valid_p): New.
32785         (split_address_groups): Check offset validity with above function.
32786         (gt-tree-ssa-loop-ivopts.h): Include header file.
32788 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32790         * config.gcc (arm*-*-*): Add missing 'fi'.
32792 2017-05-05  Steve Ellcey  <sellcey@cavium.com>
32794         * doc/invoke.texi (-fopt-info): Explicitly say order of options
32795         included in -fopt-info does not matter.
32796         * doc/optinfo.texi (-fopt-info): Fix description of default
32797         behavour. Explicitly say order of options included in -fopt-info
32798         does not matter.
32800 2017-05-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32802         * config.gcc: Allow combinations of aprofile and rmprofile values for
32803         --with-multilib-list.
32804         * config/arm/t-multilib: New file.
32805         * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32806         variables.  Remove setting of ISA and floating-point ABI in
32807         MULTILIB_OPTIONS and MULTILIB_DIRNAMES.  Set architecture and FPU in
32808         MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32809         and MULTILIB_DIRNAMES respectively.  Add comment to introduce all
32810         matches.  Add architecture matches for marvel-pj4 and generic-armv7-a
32811         CPU options.
32812         * config/arm/t-rmprofile: Likewise except for the matches changes.
32813         * doc/install.texi (--with-multilib-list): Document the combination of
32814         aprofile and rmprofile values and warn about pitfalls in doing that.
32816 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32818         * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32819         (movdi_aarch64): Likewise.
32821 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32823         PR tree-optimization/80632
32824         * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32825         field.
32826         (build_arrays): Initialize it for virtual phis.
32827         (fix_phi_nodes): Use it for virtual phis.
32829         PR tree-optimization/80558
32830         * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32831         [x, y] op z into [x op, y op z] for op & or | if conditions
32832         are met.
32834 2017-05-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
32835             Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32837         PR target/71607
32838         * config/arm/arm.md (use_literal_pool): Remove.
32839         (64-bit immediate split): No longer takes cost into consideration
32840         if arm_disable_literal_pool is enabled.
32841         * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32842         used when arm_disable_literal_pool is enabled.
32843         (arm_max_const_double_inline_cost): Remove use of
32844         arm_disable_literal_pool.
32845         (push_minipool_fix): Add assert.
32846         (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32847         * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32848         (no_literal_pool_sf_immediate): New.
32850 2017-05-05  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32852         PR tree-optimization/80613
32853         * tree-ssa-dce.c (propagate_necessity): Remove cases for
32854         BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32856 2017-05-05  Richard Biener  <rguenther@suse.de>
32858         * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32860 2017-05-05  Georg-Johann Lay  <avr@gjlay.de>
32862         * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32863         of this flag from insn conditions due to removal from r247495.
32865 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32867         * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32868         New function.
32869         (arm_early_store_addr_dep_ptr): Likewise.
32870         * config/arm/aarch-common-protos.h
32871         (arm_early_load_addr_dep_ptr): Add prototype.
32872         (arm_early_store_addr_dep_ptr): Likewise.
32873         * config/arm/cortex-a53.md: Add new bypasses.
32875 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32877         * tree.c (next_type_uid): Change type to unsigned.
32878         (type_hash_canon): Decrement back next_type_uid if
32879         freeing a type node with the highest TYPE_UID.  For INTEGER_TYPEs
32880         also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32881         if possible.
32883 2017-05-04  Martin Sebor  <msebor@redhat.com>
32885         * builtins.c: Fix a trivial typo in a comment.
32887         PR middle-end/79234
32888         * builtins.c (check_sizes): Adjust to handle reading past the end.
32889         Avoid printing excessive upper bound of ranges.  Use %E to print
32890         tree nodes instead of converting them to %wu.
32891         (expand_builtin_memchr): New function.
32892         (compute_dest_size): Rename...
32893         (compute_objsize): ...to this.
32894         (expand_builtin_memcpy): Adjust.
32895         (expand_builtin_mempcpy): Adjust.
32896         (expand_builtin_strcat): Adjust.
32897         (expand_builtin_strcpy): Adjust.
32898         (check_strncat_sizes): Adjust.
32899         (expand_builtin_strncat): Adjust.
32900         (expand_builtin_strncpy): Adjust and simplify.
32901         (expand_builtin_memset): Adjust.
32902         (expand_builtin_bzero): Adjust.
32903         (expand_builtin_memcmp): Adjust.
32904         (expand_builtin): Handle memcmp.
32905         (maybe_emit_chk_warning): Check strncat just once.
32907 2017-05-04  Martin Sebor  <msebor@redhat.com>
32909         PR preprocessor/79214
32910         PR middle-end/79222
32911         PR middle-end/79223
32912         * builtins.c (check_sizes): Add inlining context and issue
32913         warnings even when -Wno-system-headers is set.
32914         (check_strncat_sizes): Same.
32915         (expand_builtin_strncat): Same.
32916         (expand_builtin_memmove): New function.
32917         (expand_builtin_stpncpy): Same.
32918         (expand_builtin): Handle memmove and stpncpy.
32920 2017-05-04  Bin Cheng  <bin.cheng@arm.com>
32922         * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
32923         which is not used any more.
32925 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32927         * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
32929 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32931         * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
32932         (cortexa53_tunings): Likewise.
32933         (cortexa57_tunings): Likewise.
32934         (cortexa72_tunings): Likewise.
32935         (cortexa73_tunings): Likewise.
32937 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32939         * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
32940         Set loop alignment to 8.
32942 2017-05-04  Martin Sebor  <msebor@redhat.com>
32944         PR translation/80280
32945         * builtins.c (expand_builtin_object_size): Add missing quoting to
32946         %D and like directives.
32947         * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
32948         (hsa_type_for_tree_type): Same.
32949         (verify_function_arguments): Same.
32950         * symtab.c (symbol_table::change_decl_assembler_name): Same.
32951         * varasm.c (get_section): Same.
32952         (mark_weak): Same.
32954 2017-05-04  Martin Sebor  <msebor@redhat.com>
32956         PR translation/80280
32957         * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
32959 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32961         * config/aarch64/aarch64.c (generic_addrcost_table):
32962         Change HI/TI mode setting.
32964 2017-05-04  Martin Jambor  <mjambor@suse.cz>
32966         PR tree-optimization/80622
32967         * tree-sra.c (comes_initialized_p): New function.
32968         (build_accesses_from_assign): Only set write lazily when
32969         comes_initialized_p is false.
32970         (analyze_access_subtree): Use comes_initialized_p.
32971         (propagate_subaccesses_across_link): Assert !comes_initialized_p
32972         instead of testing for PARM_DECL.
32974 2017-05-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32976         * config/aarch64/aarch64.md (prefetch); Adjust predicate and
32977         constraint on operand 0 to allow more general addressing modes.
32978         Adjust output template.
32979         * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
32980         New function.
32981         * config/aarch64/aarch64-protos.h
32982         (aarch64_address_valid_for_prefetch_p): Declare prototype.
32983         * config/aarch64/constraints.md (Dp): New address constraint.
32984         * config/aarch64/predicates.md (aarch64_prefetch_operand): New
32985         predicate.
32987 2017-05-04  Jan Hubicka  <hubicka@ucw.cz>
32989         * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
32990         update use of estimate_ipcp_clone_size_and_time.
32991         (estimate_local_effects): Update use of
32992         estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
32993         * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
32994         * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
32995         Return nonspecialized time.
32997 2017-05-04  Richard Biener  <rguenther@suse.de>
32999         * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
33000         for the last VUSE which def dominates the PHI.  Directly call
33001         maybe_skip_until.
33002         (get_continuation_for_phi_1): Remove.
33004 2017-05-04  Richard Sandiford  <richard.sandiford@linaro.org>
33006         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
33007         to explain the use of truncating division.  Cap the number of
33008         iterations to the maximum given by nb_iterations_upper_bound,
33009         if defined.
33011 2017-05-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33013         * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
33014         * configure: Regenerate.
33015         * config.in: Regenerate.
33016         * config/i386/driver-mingw32.c: new file.
33017         * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
33018         * config.host: Link driver-mingw32.o on MinGW host.
33019         * doc/install.texi: Document new --enable-mingw-wildcard configure
33020         option.
33022 2017-05-04  Marek Polacek  <polacek@redhat.com>
33024         PR tree-optimization/80612
33025         * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
33027 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
33028             Andre Simoes Dias Vieira  <andre.simoesdiasvieira@arm.com>
33030         * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
33031         (movt splitter): Likewise.
33032         * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
33033         to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
33034         (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
33035         block for Thumb-1 with MOVT.
33036         (thumb2_legitimate_address_p): Move code block ...
33037         (can_avoid_literal_pool_for_label_p): ... into this new function.
33038         (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
33039         literal pool.
33040         (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
33041         * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
33042         "M-profile targets with the MOVT instruction".
33044 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
33046         * config/arm/arm-builtins.c (arm_init_builtins): Rename
33047         __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
33048         __builtin_arm_stfscr to __builtin_arm_set_fpscr.
33050 2017-05-04  Martin Liska  <mliska@suse.cz>
33052         * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
33053         variable cond_code.
33055 2017-05-04  Richard Biener  <rguenther@suse.de>
33057         * tree.c (array_at_struct_end_p): Handle arrays at struct
33058         end with flexarrays more conservatively.  Refactor and treat
33059         arrays of arrays or aggregates more strict.  Fix
33060         VIEW_CONVERT_EXPR handling.  Remove allow_compref argument.
33061         * tree.c (array_at_struct_end_p): Adjust prototype.
33062         * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
33063         * gimple-fold.c (get_range_strlen): Likewise.
33064         * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
33066 2017-05-04  Richard Biener  <rguenther@suse.de>
33068         PR tree-optimization/31130
33069         * tree-vrp.c (needs_overflow_infinity): Remove as always returning
33070         false.
33071         (supports_overflow_infinity): Likewise.
33072         (is_negative_overflow_infinity): Likewise.
33073         (is_positive_overflow_infinity): Likewise.
33074         (is_overflow_infinity): Likewise.
33075         (stmt_overflow_infinity): Likewise.
33076         (overflow_infinity_range_p): Likewise.
33077         (usable_range_p): Remove as always returning true.
33078         (make_overflow_infinity): Remove.
33079         (negative_overflow_infinity): Likewise.
33080         (positive_overflow_infinity): Likewise.
33081         (avoid_overflow_infinity): Likewise.
33082         (set_value_range): Adjust accordingly.
33083         (set_value_range_to_nonnegative): Likewise, remove now unused
33084         overflow_infinity arg.
33085         (vrp_operand_equal_p): Adjust.
33086         (update_value_range): Likewise.
33087         (range_int_cst_singleton_p): Likewise.
33088         (operand_less_p): Likewise.
33089         (compare_values_warnv): Likewise.
33090         (extract_range_for_var_from_comparison_expr): Likewise.
33091         (vrp_int_const_binop): Likewise.
33092         (zero_nonzero_bits_from_vr): Likewise.
33093         (extract_range_from_multiplicative_op_1): Likewise.
33094         (extract_range_from_binary_expr_1): Likewise.
33095         (extract_range_from_unary_expr): Likewise.
33096         (extract_range_from_comparison): Likewise.
33097         (extract_range_basic): Likewise.
33098         (adjust_range_with_scev): Likewise.
33099         (compare_ranges): Likewise.
33100         (compare_range_with_value): Likewise.
33101         (dump_value_range): Likewise.
33102         (test_for_singularity): Likewise, remove strict_overflow_p parameter
33103         never used.
33104         (simplify_cond_using_ranges): Adjust.
33106 2017-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
33108         * brig-builtins.def: Added a builtin for class_f64.
33109         * builtin-types.def: Added a builtin type needed by class_f64.
33111 2017-05-03  Jason Merrill  <jason@redhat.com>
33113         * timevar.def: Add TV_CONSTEXPR.
33115 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33117         * common.opt (fdiagnostics-parseable-fixits): Fix typo.
33119 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33121         * ipa-prop.c (ipa_update_after_lto_read): Removed.
33122         * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
33123         * ipa-cp.c (ipcp_propagate_stage): Do not call
33124         ipa_update_after_lto_read.
33125         * ipa-inline.c (ipa_inline): Likewise.
33127 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33129         * ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
33130         tag.  Added a default constructor and a destructor.
33131         (ipa_edge_args_sum_t): New class;
33132         (ipa_edge_args_sum): Declare.
33133         (ipa_edge_args_vector): Remove declaration.
33134         (IPA_EDGE_REF): Use ipa_edge_args_sum.
33135         (ipa_free_edge_args_substructures): Remove declaration.
33136         (ipa_check_create_edge_args): Use ipa_edge_args_sum.
33137         (ipa_edge_args_info_available_for_edge_p): Likewise.
33138         * ipa-prop.c (ipa_edge_args_vector): Removed.
33139         (edge_removal_hook_holder): Likewise.
33140         (edge_duplication_hook_holder): Likewise.
33141         (ipa_edge_args_sum): New variable.
33142         (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
33143         ipa_edge_args_vector.
33144         (ipa_free_edge_args_substructures): Likewise.
33145         (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
33146         ipa_edge_args_vector.
33147         (ipa_edge_removal_hook): Turned into method
33148         ipa_edge_args_sum_t::remove.
33149         (ipa_edge_duplication_hook): Turned into method
33150         ipa_edge_args_sum_t::duplicate.
33151         (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
33152         registering edge hooks.
33153         (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
33154         * ipa-inline-analysis.c (estimate_function_body_sizes): Test
33155         ipa_edge_args_sum instead of ipa_edge_args_vector.
33156         * ipa-profile.c (ipa_profile): Likewise.
33158 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33160         * symbol-summary.h (function_summary): New method exists.
33161         (function_summary::symtab_removal): Deallocate through release.
33162         (call_summary): New class.
33163         (gt_ggc_mx): New overload.
33164         (gt_pch_nx): Likewise.
33165         (gt_pch_nx): Likewise.
33167 2017-05-03  Jeff Law  <law@redhat.com>
33169         PR tree-optimization/78496
33170         * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
33171         from simplify_cond_using_ranges.  Split off code to walk
33172         backwards through casts into ...
33173         (simplify_cond_using_ranges_2): New function.
33174         (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
33175         (execute_vrp): After identifying jump threads, call
33176         simplify_cond_using_ranges_2.
33178 2017-05-03  Jan Hubicka  <hubicka@ucw.cz>
33180         PR bootstrap/80609
33181         * ipa-inline.h (inline_summary): Add ctor.
33182         (create_ggc): Do not use ggc_cleared_alloc.
33184 2017-05-03  Jeff Downs  <heydowns@somuchpressure.net>
33185             Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33187         * gcc.c (handle_braces): Support escaping in switch matching
33188         text.
33189         * doc/invoke.texi (Spec Files): Document it.
33190         Remove superfluous @code markup in items.
33192 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33194         * diagnostic-show-locus.c (struct column_range): New struct.
33195         (get_affected_columns): New function.
33196         (get_printed_columns): New function.
33197         (struct correction): New struct.
33198         (correction::ensure_capacity): New function.
33199         (correction::ensure_terminated): New function.
33200         (struct line_corrections): New struct.
33201         (line_corrections::~line_corrections): New dtor.
33202         (line_corrections::add_hint): New function.
33203         (layout::print_trailing_fixits): Reimplement in terms of the new
33204         classes.
33205         (selftest::test_overlapped_fixit_printing): New function.
33206         (selftest::diagnostic_show_locus_c_tests): Call it.
33208 2017-05-03  Nathan Sidwell  <nathan@acm.org>
33210         Canonicalize canonical type hashing
33211         * tree.h (type_hash_canon_hash): Declare.
33212         * tree.c (type_hash_list, attribute_hash_list): Move into
33213         type_hash_canon_hash.
33214         (build_type_attribute_qual_variant): Break out hash code calc into
33215         type_hash_canon_hash.
33216         (type_hash_canon_hash): New.  Generic type hash computation.
33217         (build_range_type_1, build_array_type_1, build_function_type,
33218         build_method_type_directly, build_offset_type, build_complex_type,
33219         make_vector_type): Call it.
33221 2017-05-03  Richard Biener  <rguenther@suse.de>
33223         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
33224         When all DRs have unknown misaligned do not always peel
33225         when there is a store but apply the same costing model as if
33226         there were only loads.
33228 2017-05-03  Richard Biener  <rguenther@suse.de>
33230         Revert
33231         PR tree-optimization/80492
33232         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33233         compare_base_decls returning dont-know properly.
33235 2017-05-03  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33237         * config/arm/iterators.md (CCSI): New mode iterator.
33238         (arch): New mode attribute.
33239         * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33240         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33241         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This.  Use CCSI
33242         code iterator for success result mode.
33243         * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33244         the corresponding new insn generators.
33246 2017-05-03  Bin Cheng  <bin.cheng@arm.com>
33248         Revert r247509
33249         2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33250         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33252 2017-05-03  Richard Sandiford  <richard.sandiford@linaro.org>
33254         * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33255         (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33256         (DDR_A): Wrap DDR argument in brackets.
33257         (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33258         (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33259         (DDR_REVERSED_P): Likewise.
33261 2017-05-03  Jakub Jelinek  <jakub@redhat.com>
33263         PR tree-optimization/79472
33264         * tree-switch-conversion.c (struct switch_conv_info): Add
33265         contiguous_range and default_case_nonstandard fields.
33266         (collect_switch_conv_info): Compute contiguous_range and
33267         default_case_nonstandard fields, don't clear final_bb if
33268         contiguous_range and only the default case doesn't have the required
33269         structure.
33270         (check_all_empty_except_final): Set default_case_nonstandard instead
33271         of failing if contiguous_range and the default case doesn't have empty
33272         block.
33273         (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33274         and only the default case doesn't have the required constants.  Skip
33275         virtual phis.
33276         (gather_default_values): Skip virtual phis.  Allow non-NULL CASE_LOW
33277         if default_case_nonstandard.
33278         (build_constructors): Build constant 1 just once.  Assert that default
33279         values aren't inserted in between cases if contiguous_range.  Skip
33280         virtual phis.
33281         (build_arrays): Skip virtual phis.
33282         (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33283         (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33284         Handle virtual phis.
33285         (gen_inbound_check): Handle default_case_nonstandard case.
33286         (process_switch): Adjust check_final_bb caller.  Call
33287         gather_default_values with the first non-default case instead of
33288         default case if default_case_nonstandard.
33290 2017-05-02  Nathan Sidwell  <nathan@acm.org>
33292         * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33293         check.  Fix formatting.
33295 2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
33297         * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33298         errors when comparing specialized and unspecialized times.
33300 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
33302         * diagnostic-show-locus.c
33303         (layout::should_print_annotation_line_p): Make private.
33304         (layout::print_annotation_line): Make private.
33305         (layout::annotation_line_showed_range_p): Make private.
33306         (layout::show_ruler): Make private.
33307         (layout::print_source_line): Make private.  Pass in line and
33308         line_width, rather than calling location_get_source_line.  Drop
33309         returned value.
33310         (layout::print_leading_fixits): New method.
33311         (layout::print_any_fixits): Rename to...
33312         (layout::print_trailing_fixits): ...this, and make private.
33313         Don't print newline fixits.
33314         (diagnostic_show_locus): Move logic for printing one row into...
33315         (layout::print_line): ...this new function.  Move the
33316         location_get_source_line call and error-handling from
33317         print_source_line to here.  Call print_leading_fixits, and rename
33318         print_any_fixits to print_trailing_fixits.
33319         (selftest::test_fixit_insert_containing_newline): Update now that
33320         newlines are partially supported.
33321         (selftest::test_fixit_insert_containing_newline_2): New test.
33322         (selftest::test_fixit_replace_containing_newline): Update comments.
33323         (selftest::diagnostic_show_locus_c_tests): Call the new test.
33324         * edit-context.c (class added_line): New class.
33325         (class edited_line): Describe newline handling in comment.
33326         (edited_line::actually_edited_p): New method.
33327         (edited_line::print_content): Delete redundant decl.
33328         (edited_line::m_predecessors): New field.
33329         (edited_file::print_content): Call edited_line::print_content.
33330         (edited_file::print_diff): Update to support newlines.
33331         (edited_file::print_diff_hunk): Likewise.
33332         (edited_file::print_run_of_changed_lines): New function.
33333         (edited_file::print_diff_line): Convert to...
33334         (print_diff_line): ...this.
33335         (edited_file::get_effective_line_count): New function.
33336         (edited_line::edited_line): Initialize new field m_predecessors.
33337         (edited_line::~edited_line): Clean up m_predecessors.
33338         (edited_line::apply_fixit): Handle newlines.
33339         (edited_line::get_effective_line_count): New function.
33340         (edited_line::print_content): New function.
33341         (edited_line::print_diff_lines): New function.
33342         (selftest::test_applying_fixits_insert_containing_newline): New
33343         test.
33344         (selftest::test_applying_fixits_replace_containing_newline): New
33345         test.
33346         (selftest::insert_line): New function.
33347         (selftest::test_applying_fixits_multiple_lines): Add example of
33348         inserting a line.
33349         (selftest::edit_context_c_tests): Call the new tests.
33351 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33353         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33354         parameter cand.  Update dump information.
33355         (get_computation_cost): Update uses.
33357 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33359         * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33360         (get_computation_aff): Reorder parameters.  Use get_computation_aff_1.
33361         (get_computation_at, rewrite_use_address): Update use of
33362         get_computation_aff.
33364 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33366         * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33367         (get_computation): Delete.
33368         (get_computation_cost): Implement like get_computation_cost_at.
33369         Use get_computation_at.
33370         (get_computation_cost_at): Delete.
33371         (rewrite_use_nonlinear_expr): Use get_computation_at.
33372         (rewrite_use_compare, remove_unused_ivs): Ditto.
33374 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33376         * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33378 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33380         * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33381         (ivopts_global_cost_for_size): Rename parameter and update uses.
33382         (iv_ca_recount_cost): Update uses.
33383         (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33384         candidates seperately in n_invs and n_cands.
33385         (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33387 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33389         * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33390         (find_inv_vars_cb): New.
33391         (find_depends): Renamed to ...
33392         (find_inv_vars): ... this.
33393         (add_candidate_1, force_var_cost): Call find_inv_vars.
33394         (split_address_cost, determine_group_iv_cost_cond): Ditto.
33396 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33398         * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33399         inv_vars.  Add inv_exprs.
33400         (struct iv_cand): Rename depends_on to inv_vars.
33401         (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33402         max_inv_var_id/n_inv_var_uses.  Move max_inv_expr_id around.
33403         Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33404         (dump_cand): Dump inv_vars.
33405         (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33406         (record_invariant, find_depends, add_candidate_1): Ditto.
33407         (set_group_iv_cost, force_var_cost): Ditto.
33408         (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33409         (get_computation_cost_at, get_computation_cost): Ditto.
33410         (determine_group_iv_cost_generic): Ditto.
33411         (determine_group_iv_cost_address): Ditto.
33412         (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33413         (determine_group_iv_costs): Ditto.
33414         (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33415         (iv_ca_set_remove_invariants): Renamed to ...
33416         (iv_ca_set_remove_invs): ... this.  Support inv_vars and inv_exprs.
33417         (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33418         (iv_ca_set_add_invariants):  Renamed to ...
33419         (iv_ca_set_add_invs): ... this.  Support inv_vars and inv_exprs.
33420         (iv_ca_set_cp): Use iv_ca_set_add_invs.
33421         (iv_ca_has_deps): Support inv_vars and inv_exprs.
33422         (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33423         (create_new_ivs): Remove useless dump.
33425 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33427         * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33428         iv_cand code.
33429         (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33430         (iv_ca_set_no_cp, create_new_iv): Ditto.
33432 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33434         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33436 2017-05-02  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33438         * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33439         function tree_check2.
33441 2017-05-02  Martin Liska  <mliska@suse.cz>
33443         * doc/gcov.texi: Add missing preposition.
33444         * gcov.c (function_info::function_info): Properly fill up
33445         all member variables.
33447 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33449         * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33451 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33453         * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33455 2017-05-02  Martin Liska  <mliska@suse.cz>
33457         PR lto/77954.
33458         * lto-streamer-in.c (lto_read_tree_1): Remove
33459         LTO_STREAMER_DEBUG.
33460         * lto-streamer.c (struct tree_hash_entry): Likewise.
33461         (struct tree_entry_hasher): Likewise.
33462         (tree_entry_hasher::hash): Likewise.
33463         (tree_entry_hasher::equal): Likewise.
33464         (lto_streamer_init): Likewise.
33465         (lto_orig_address_map): Likewise.
33466         (lto_orig_address_get): Likewise.
33467         (lto_orig_address_remove): Likewise.
33468         * lto-streamer.h: Likewise.
33469         * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33470         * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33472 2017-05-02  Sebastian Peryt  <sebastian.peryt@intel.com>
33474         * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33475         (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33476         (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33477         (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33478         (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33479         (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33480         (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33481         (mm_maskz_sub_ss): New intrinsics.
33482         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33483         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33484         * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33485         (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33486         (__builtin_ia32_subss_mask_round): New builtins.
33487         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33488         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33489         * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33490         Renamed to ...
33491         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33492         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33493         Changed to ...
33494         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33495         ... this.
33497 2017-05-02  Martin Jambor  <mjambor@suse.cz>
33499         PR tree-optimization/78687
33500         * tree-sra.c (access): New field parent.
33501         (process_subtree_disqualification): New function.
33502         (disqualify_candidate): Call it.
33503         (build_accesses_from_assign): Reset write flag if creating an
33504         assighnment link.
33505         (build_access_subtree): Fill in parent field and also prpagate
33506         down grp_write flag.
33507         (create_artificial_child_access): New parameter set_grp_write, set
33508         grp_write to its value.
33509         (propagate_subaccesses_across_link): Also propagate grp_write flag
33510         values.
33511         (propagate_all_subaccesses): Push the closest parent back to work
33512         queue if add_access_to_work_queue returned true.
33514 2017-05-02  Richard Biener  <rguenther@suse.de>
33516         * common.opt (fstrict-overflow): Alias negative to fwrapv.
33517         * doc/invoke.texi (fstrict-overflow): Remove all traces of
33518         -fstrict-overflow documentation.
33519         * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33520         (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33521         flag_strict_overflow.
33522         * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33523         * lto-opts.c (lto_write_options): Do not stream it.
33524         * lto-wrapper.c (merge_and_complain): Do not handle it.
33525         * opts.c (default_options_table): Do not set -fstrict-overflow.
33526         (finish_options): Likewise do not clear it when sanitizing.
33527         * simplify-rtx.c (simplify_const_relational_operation): Do not
33528         test flag_strict_overflow.
33530 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33532         * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33533         using enabled attribute.
33534         (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33535         (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33536         (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33537         (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33538         (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33539         (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33540         (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33541         (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33542         (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33543         (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33545 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33547         * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33549 2017-05-02  Richard Biener  <rguenther@suse.de>
33551         PR tree-optimization/80591
33552         Revert
33553         2017-04-10  Richard Biener  <rguenther@suse.de>
33555         * tree-ssa-structalias.c (find_func_aliases): Properly handle
33556         asm inputs.
33558 2017-05-02  Richard Biener  <rguenther@suse.de>
33560         PR tree-optimization/80549
33561         * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33562         (cleanup_tree_cfg_noloop): Create forwarders to known loop
33563         headers if they do not have a preheader.
33565 2017-05-02  Martin Liska  <mliska@suse.cz>
33567         PR other/80589
33568         * common.opt: Fix typo.
33569         * doc/invoke.texi: Likewise.
33571 2017-05-01  Jan Beulich  <jbeulich@suse.com>
33573         * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33574         swapping, add (x,x,m,x,n) alternative.
33576 2017-05-01  Nathan Sidwell  <nathan@acm.org>
33578         * calls.c (combine_pending_stack_adjustment_and_call): Remove
33579         unnecessary unadjusted_alignment check.
33581 2017-05-01  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33583         PR c++/80038
33584         * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33585         operations here.
33586         * gimplify.c (gimplify_cilk_detach): New function.
33587         (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33588         * tree-core.h: Document EXPR_CILK_SPAWN.
33589         * tree.h (EXPR_CILK_SPAWN): Define.
33591 2017-05-01  David Malcolm  <dmalcolm@redhat.com>
33593         * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33594         to use new fixit_hint representation, using the "replace" logic.
33595         (get_line_span_for_fixit_hint): Likewise.
33596         (layout::print_any_fixits): Likewise.
33597         (selftest::test_one_liner_many_fixits): Rename to...
33598         (selftest::test_one_liner_many_fixits_1): ...this, and update
33599         comment and expected output to reflect that the multiple fix-it
33600         hints are now consolidated into one insertion.
33601         (selftest::test_one_liner_many_fixits_2): New test.
33602         (selftest::test_diagnostic_show_locus_one_liner): Update for
33603         above.
33604         (selftest::test_fixit_consolidation): Update for fix-it API
33605         change.
33606         * diagnostic.c (print_parseable_fixits): Likewise.
33607         * edit-context.c (edited_line::m_line_events): Convert from
33608         auto_vec <line_event *> to auto_vec <line_event>.
33609         (class line_event): Convert from abstract base class to a concrete
33610         class, taking over the role of replace_event.
33611         (class insert_event): Delete.
33612         (class replace_event): Rename to class line_event.  Convert to
33613         half-open range.
33614         (edit_context::add_fixits): Reimplement.
33615         (edit_context::apply_insert): Delete.
33616         (edit_context::apply_replace): Rename to...
33617         (edit_context::apply_fixit): ...this.  Convert to half-open range.
33618         (edited_file::apply_insert): Delete.
33619         (edited_file::apply_replace): Rename to...
33620         (edited_file::apply_fixit): ...this.
33621         (edited_line::~edited_line): Drop deletion of events.
33622         (edited_line::apply_insert): Delete.
33623         (edited_line::apply_replace): Rename to...
33624         (edited_line::apply_fixit): ...this.  Convert to half-open range.
33625         Update for change to type of m_line_events.
33626         * edit-context.h (edit_context::apply_insert): Delete.
33627         (edit_context::apply_replace): Rename to...
33628         (edit_context::apply_fixit): ...this.
33630 2017-05-01  Martin Sebor  <msebor@redhat.com>
33632         * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33633         known.
33635 2017-05-01  Uros Bizjak  <ubizjak@gmail.com>
33637         PR target/68491
33638         * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33639         __get_cpuid_max returns 0.
33640         (__get_cpuid_count): Ditto.
33642 2017-05-01  Eric Botcazou  <ebotcazou@adacore.com>
33644         * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33645         replacement expression is another instance of one of its arguments.
33647 2017-05-01  Jakub Jelinek  <jakub@redhat.com>
33649         PR target/79430
33650         * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33651         check for stack push/pop autoinc.
33652         * config/i386/i386.c (ix86_agi_dependent): Return false
33653         if the only reason why modified_in_p returned true is that
33654         addr is SP based and set_insn is a push or pop.
33656 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33658         * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33659         overflow check.
33661 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33663         PR ipa/79224
33664         * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33665         (account_size_time): Use two predicates - exec_pred and
33666         nonconst_pred_ptr.
33667         (evaluate_conditions_for_known_args): Compute both clause and
33668         nonspec_clause.
33669         (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33670         (inline_summary_t::duplicate): Update.
33671         (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33672         separately.
33673         (compute_inline_parameters): Likewise.
33674         (estimate_edge_size_and_time): Update caluclation of time.
33675         (estimate_node_size_and_time): Compute both time and nonspecialized
33676         time.
33677         (estimate_ipcp_clone_size_and_time): Update.
33678         (inline_merge_summary): Update.
33679         (do_estimate_edge_time): Update.
33680         (do_estimate_edge_size): Update.
33681         (do_estimate_edge_hints): Update.
33682         (inline_read_section, inline_write_summary): Stream both new predicates.
33683         * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33684         as argument.
33685         (compute_inlined_call_time): Cleanup.
33686         (big_speedup_p): Update.
33687         (edge_badness): Update.
33688         * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33689         (size_time_entry): Replace predicate by exec_predicate and
33690         nonconst_predicate.
33691         (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33692         (estimate_edge_time): Return also nonspec_time.
33693         (reset_edge_growth_cache): Update.
33695 2017-04-29  Jakub Jelinek  <jakub@redhat.com>
33697         PR rtl-optimization/80491
33698         * ifcvt.c (noce_process_if_block): When looking for x setter
33699         with missing else_bb, don't check only the insn right before
33700         cond_earliest, but look for the last insn that x is modified in
33701         within the same bb.
33703         PR rtl-optimization/80491
33704         * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33706 2017-04-29  Marc Glisse  <marc.glisse@inria.fr>
33708         PR tree-optimization/80487
33709         * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33711 2017-04-29  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33713         PR tree-optimization/79697
33714         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33715         is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33716         (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33717         BUILT_IN_STRNDUP.
33718         * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33719         (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33721 2017-04-28  Martin Sebor  <msebor@redhat.com>
33723         PR tree-optimization/80523
33724         * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33725         (init_target_to_host_charmap, target_to_host, target_strtol10): New
33726         functions.
33727         (maybe_warn, format_directive, parse_directive): Use new functions.
33728         (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33730 2017-04-28  Marc Glisse  <marc.glisse@inria.fr>
33732         * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33734 2017-04-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
33736         * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33737         target_header_dir): Set correctly.
33738         * configure: Regenerated.
33739         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33740         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33741         instead of SYSTEM_HEADER_DIR.
33743 2017-04-28  Jan Hubicka  <hubicka@ucw.cz>
33745         * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33746         (estimate_local_effects): Likewise.
33747         * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33748         edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33749         * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33750         estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33751         do_estimate_edge_time, estimate_edge_time): Likewise.
33752         * ipa-inline-analysis.c (estimate_node_size_and_time,
33753         estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33754         (estimate_time_after_inlining): Remove.
33756 2017-04-28  Martin Liska  <mliska@suse.cz>
33758         * doc/gcov.texi: Enhance documentation of gcov.
33760 2017-04-28  Martin Liska  <mliska@suse.cz>
33762         * doc/gcov.texi: Sort options in alphabetic order.
33763         * doc/gcov-dump.texi: Likewise.
33764         * doc/gcov-tool.texi: Likewise.
33765         * gcov.c (print_usage): Likewise.
33766         * gcov-dump.c (print_usage): Likewise.
33767         * gcov-tool.c (print_merge_usage_message): Likewise.
33768         (print_rewrite_usage_message): Likewise.
33769         (print_overlap_usage_message): Likewise.
33771 2017-04-28  Martin Liska  <mliska@suse.cz>
33773         PR gcov-profile/53915
33774         * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33776 2017-04-28  Martin Liska  <mliska@suse.cz>
33778         PR gcov-profile/79891
33779         * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33780         is marked by compiler as living on a line.
33781         (get_cycles_count): Remove usage of the union.
33782         (output_intermediate_file): Likewise.
33783         (find_source): Fix GNU coding style.
33784         (accumulate_line_counts): Remove old non-all block mode.
33785         (output_lines): Remove usage of the union.
33786         * profile.c (output_location): Include all BBs, even if
33787         belonging to a same line (and file) as a previous BB.
33789 2017-04-28  Martin Liska  <mliska@suse.cz>
33791         * gcov.c (process_args): Handle new argument 'w'.
33792         (read_graph_file): Assign ID to BBs.
33793         (output_branch_count): Display BB # if verbose flag is set.
33794         (output_lines): Likewise for arcs.
33795         (print_usage): Add '--verbose' option help.
33796         * doc/gcov.texi: Document --verbose (-w) option.
33798 2017-04-28  Martin Liska  <mliska@suse.cz>
33800         * gcov.c (struct block_location_info): New struct.
33801         (process_file): Fill up the new structure.
33802         (read_graph_file): Replace usage of encoding by the newly added
33803         struct.
33804         (add_line_counts): Likewise.
33805         (accumulate_line_counts): Remove usage of the union.
33806         (function_info::function_info): New function.
33807         (function_info::~function_info): Likewise.
33808         (process_file): Call delete instead of release_function.
33809         (release_function): Release the function.
33810         (release_structures): Call delete instead of release_function.
33811         (solve_flow_graph): Replace usage of num_blocks.
33812         (find_exception_blocks): Likewise.
33813         (output_lines): Fix GNU coding style.
33815 2017-04-28  Martin Liska  <mliska@suse.cz>
33817         PR driver/56469
33818         * coverage.c (coverage_remove_note_file): New function.
33819         * coverage.h: Declare the function.
33820         * toplev.c (finalize): Clean if an error has been seen.
33822 2017-04-28  Martin Liska  <mliska@suse.cz>
33824         PR gcov-profile/80031
33825         * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33826         * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33827         * gcov.c (read_graph_file): Read just number of blocks.
33828         * profile.c (branch_prob): Do not stream 0 flags per a basic
33829         block.
33831 2017-04-28  Martin Liska  <mliska@suse.cz>
33833         * gcov-dump.c (tag_*): Add new argument to declarations.
33834         (dump_gcov_file): Likewise.
33835         (tag_blocks): Add and use new argument depth.
33836         (tag_arcs): Likewise.
33837         (tag_lines): Likewise.
33838         (tag_counters): Likewise.
33839         (tag_summary): Likewise.
33840         (dump_working_sets): Use depth to do a proper indentation.
33842 2017-04-28  Jakub Jelinek  <jakub@redhat.com>
33844         PR bootstrap/80531
33845         * cgraph.h (symtab_node::debug_symtab): No longer inline.
33846         * symtab.c (symtab_node::debug_symtab): Move definition here.
33848 2017-04-28  Richard Biener  <rguenther@suse.de>
33850         * lto-streamer.h (LTO_major_version): Bump to 7.
33852 2017-04-28  Richard Biener  <rguenther@suse.de>
33854         * tree-vrp.c (assert_info): New struct.
33855         (add_assert_info): New helper.
33856         (register_edge_assert_for_2): Refactor to add asserts to a vector
33857         of assert_info.
33858         (register_edge_assert_for_1): Likewise.
33859         (register_edge_assert_for): Likewise.
33860         (finish_register_edge_assert_for): New helper actually registering
33861         asserts where live on edge.
33862         (find_conditional_asserts): Adjust.
33863         (find_switch_asserts): Likewise.
33864         (evrp_dom_walker::try_find_new_range): Generalize.
33865         (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33867 2017-04-27  Marek Polacek  <polacek@redhat.com>
33869         PR sanitizer/80349
33870         * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33871         arg10 and arg11 to itype.
33873 2017-04-27  Jonathan Wakely  <jwakely@redhat.com>
33875         * doc/extend.texi (Object Size Checking): Improve grammar.
33877 2017-04-27  Richard Earnshaw  <rearnsha@arm.com>
33879         PR target/80530
33880         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33881         that the logic for permitting reciprocal estimates matches that
33882         in use_rsqrt_p.
33884 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33886         PR c++/80534
33887         * tree.c (type_cache_hasher::equal): Only compare
33888         TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33889         (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33890         non-aggregate element types.
33891         * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33892         about the flag on ARRAY_TYPEs in the comment, formatting fix.
33894 2017-04-27  Richard Biener  <rguenther@suse.de>
33896         PR middle-end/80533
33897         * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33898         stripping ARRAY_REFs from MEM_EXPR make sure we're not
33899         keeping a reference to a trailing array.
33901 2017-04-27  Richard Biener  <rguenther@suse.de>
33903         PR middle-end/80539
33904         * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33905         being in loop-closed SSA form conservatively.
33906         (chrec_fold_multiply_poly_poly): Likewise.
33908 2017-04-27  Tamar Christina  <tamar.christina@arm.com>
33910         PR middle-end/79665
33911         * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
33912         CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
33914 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33916         PR target/77728
33917         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
33918         (aarch64_function_arg_alignment): Return unsigned int again, but still
33919         ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
33920         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33921         Don't emit -Wpsabi note.
33922         (aarch64_function_arg_boundary): Likewise.
33923         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33924         caller.
33926 2017-04-26  Nathan Sidwell  <nathan@acm.org>
33928         * tree.h (crc32_unsigned_n): Declare.
33929         (crc32_unsigned, crc32_unsigned): Make inline.
33930         * tree.c (crc32_unsigned_bits): Replace with ...
33931         (crc32_unsigned_n): ... this.
33932         (crc32_unsigned, crc32_byte): Remove.
33933         (crc32_string): Remove unnecessary braces.
33935 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33937         * ipa-cp.c (estimate_local_effects): Convert sreal to int.
33938         * ipa-inline-analysis.c (MAX_TIME): Remove.
33939         (account_size_time): Use sreal for time.
33940         (dump_inline_summary): Update.
33941         (estimate_function_body_sizes): Update.
33942         (estimate_edge_size_and_time): Update.
33943         (estimate_calls_size_and_time): Update.
33944         (estimate_node_size_and_time): Update.
33945         (inline_merge_summary): Update.
33946         (inline_update_overall_summary): Update.
33947         (estimate_time_after_inlining): Update.
33948         (inline_read_section): Update.
33949         (inline_write_summary): Update.
33950         * ipa-inline.c (compute_uninlined_call_time): Update.
33951         (compute_inlined_call_time): Update.
33952         (recursive_inlining): Update.
33953         (inline_small_functions): Update.
33954         (dump_overall_stats): Update.
33955         * ipa-inline.h: Include sreal.h.
33956         (size_time_entry): Turn time to sreal.
33957         (inline_summary): Turn self_time nad time to sreal.
33959 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33961         * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
33962         data-streamer.h
33963         (sreal::stream_out, sreal::stream_in): New.
33964         * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
33966 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
33968         * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
33969         environment.
33971 2017-04-25  Uros Bizjak  <ubizjak@gmail.com>
33973         PR target/70799
33974         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
33975         Handle ASHIFTRT.
33976         (dimode_scalar_chain::compute_convert_gain): Ditto.
33977         (dimode_scalar_chain::make_vector_copies): Ditto.
33978         (dimode_scalar_chain::convert_reg): Ditto.
33979         (dimode_scalar_chain::convert_insn): Ditto.
33980         * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
33981         (VI248_AVX512BW_1): New mode iterator.
33982         (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
33983         <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
33984         mode iterator.
33986 2017-04-25  Martin Sebor  <msebor@redhat.com>
33988         PR tree-optimization/80497
33989         * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
33990         constants are representable in HOST_WIDE_INT.
33991         (parse_directive): Ditto.
33993 2017-04-25  Martin Sebor  <msebor@redhat.com>
33995         PR bootstrap/80486
33996         * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
33997         (new_zero_array): Adjust signature.
33998         (dom_info::dom_init): Used unsigned rather that size_t.
33999         (dom_info::dom_info): Same.
34001 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
34002             Jakub Jelinek  <jakub@redhat.com>
34004         PR target/77728
34005         * config/arm/arm.c: Include gimple.h.
34006         (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
34007         returns negative, increment ncrn only if it returned positive.
34008         (arm_needs_doubleword_align): Return int instead of bool,
34009         ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
34010         members, but if there is any such non-FIELD_DECL
34011         > PARM_BOUNDARY aligned decl, return -1 instead of false.
34012         (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
34013         returns negative, increment nregs only if it returned positive.
34014         (arm_setup_incoming_varargs): Likewise.
34015         (arm_function_arg_boundary): Emit -Wpsabi note if
34016         arm_needs_doubleword_align returns negative, return
34017         DOUBLEWORD_ALIGNMENT only if it returned positive.
34019 2017-04-25  Marek Polacek  <polacek@redhat.com>
34021         PR sanitizer/80349
34022         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34023         first argument to type.
34025 2017-04-25  Bill Seurer  <seurer@linux.vnet.ibm.com>
34027         PR target/80482
34028         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
34029         type checks to test for compatibility instead of equality.
34031 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
34032             Jakub Jelinek  <jakub@redhat.com>
34034         PR target/77728
34035         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
34036         type.
34037         (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
34038         struct.  Ignore DECL_ALIGN of decls other than FIELD_DECL for
34039         the alignment computation, but return their maximum in warn_alignment.
34040         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
34041         Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
34042         is smaller.
34043         (aarch64_function_arg_boundary): Likewise.  Simplify using MIN/MAX.
34044         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
34045         caller.
34047 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34049         * config/arc/simdext.md (dmpyh): Fix typo.
34051 2017-04-25  Richard Biener  <rguenther@suse.de>
34053         PR tree-optimization/80492
34054         * alias.c (compare_base_decls): Handle registers with asm
34055         specification conservatively.
34056         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
34057         compare_base_decls returning dont-know properly.
34059 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34061         * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
34062         (legitimate_offset_address_p): New function.
34063         (arc_legitimate_address_p): Use above function.
34065 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34067         * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
34069 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34071         * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
34072         ACCH registers whenever they are available.
34074 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34076         * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
34077         double regs fix when not used.
34079 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34081         * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
34082         core registers.
34083         (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
34084         (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
34086 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34088         * config/arc/arc.c (arc_output_addsi): Check for h-register class
34089         when emitting short ADD instructions.
34091 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34093         * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
34094         constraint.
34095         (cmpsi_cc_c_insn): Likewise.
34096         (cbranchsi4_scratch): Compute proper instruction length using
34097         compact_hreg_operand.
34098         * config/arc/predicates.md (compact_hreg_operand): New predicate.
34100 2017-04-25  Richard Biener  <rguenther@suse.de>
34102         PR middle-end/80509
34103         * passes.c (pass_manager::pass_manager): Initialize
34104         m_name_to_pass_map.
34106 2017-04-25  Richard Biener  <rguenther@suse.de>
34108         PR tree-optimization/79201
34109         * tree-ssa-sink.c (statement_sink_location): Handle calls.
34111 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34113         PR target/80464
34114         * config/s390/vector.md: Split MEM->GPR vector moves for
34115         non-s_operand addresses.
34117 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34119         PR target/79895
34120         * config/s390/predicates.md (reload_const_wide_int_operand): New
34121         predicate.
34122         * config/s390/s390.md ("movti"): Remove d/P alternative.
34123         ("movti_bigconst"): New pattern definition.
34125 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34127         PR target/80080
34128         * s390-protos.h (s390_expand_cs_hqi): Removed.
34129         (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
34130         * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
34131         modes as well as CCZ1mode and CCZmode.
34132         (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
34133         signature of s390_emit_compare_and_swap.
34134         (s390_expand_cs_hqi): Likewise, make static.
34135         (s390_expand_cs_tdsi): Generate an explicit compare before trying
34136         compare-and-swap, in some cases.
34137         (s390_expand_cs): Wrapper function.
34138         (s390_expand_atomic_exchange_tdsi): New backend specific expander for
34139         atomic_exchange.
34140         (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
34141         * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
34142         patterns for small and large integers.  Forbid symref memory operands.
34143         Move expander to s390.c.  Require cc register.
34144         ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
34145         ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
34146         ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
34147         ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
34148         symref memory operands.  Remove CC mode and call s390_match_ccmode
34149         instead.
34150         ("atomic_exchange<mode>"): Allow and implement all integer modes.
34152 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34154         * config/s390/s390.md (define_peephole2): New peephole to help
34155         combining the load-and-test pattern with volatile memory.
34157 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34159         * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
34160         with CCZmode for TARGET_Z196.
34162 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34164         PR rtl-optimization/80501
34165         * combine.c (make_compound_operation_int): Set subreg_code to SET
34166         even for AND with mask of the sign bit of mode.
34168         PR rtl-optimization/80500
34169         * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
34170         sum's initial value.
34172 2017-04-25  Julian Brown  <julian@codesourcery.com>
34173             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34175         * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
34177 2017-04-25  Marc Glisse  <marc.glisse@inria.fr>
34179         * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
34181 2017-04-25  Julian Brown  <julian@codesourcery.com>
34182             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34184         * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
34185         (thunderx2t99_sha): New Reservation.
34187 2017-04-25  Julian Brown  <julian@codesourcery.com>
34188             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34190         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
34191         type for 1-element load.
34193 2017-04-24  Marc Glisse  <marc.glisse@inria.fr>
34195         * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
34197 2017-04-24  Martin Jambor  <mjambor@suse.cz>
34199         PR tree-optimization/80293
34200         * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
34201         char arrays not totally scalarizable if it is false.
34202         (analyze_all_variable_accesses): Pass correct value in the new
34203         parameter.  Add a statistics counter.
34205 2017-04-24  Jan Hubicka  <hubicka@ucw.cz>
34207         PR middle-end/79931
34208         * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
34210 2017-04-24  Richard Biener  <rguenther@suse.de>
34212         PR tree-optimization/80494
34213         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
34214         out for complex types.
34216 2017-04-24  Richard Biener  <rguenther@suse.de>
34218         * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
34219         * tree-ssa-sccvn.c (print_scc): Print SCC size.
34220         (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
34221         (DFS): Adjust and never fail.
34222         (sccvn_dom_walker::fail): Remove.
34223         (sccvn_dom_walker::before_dom_children): Adjust.
34224         (run_scc_vn): Likewise and never fail.
34225         * tree-ssa-pre.c (pass_pre::execute): Adjust.
34226         (pass_fre::execute): Likewise.
34228 2017-04-24  Richard Biener  <rguenther@suse.de>
34230         PR tree-optimization/79725
34231         * tree-ssa-sink.c (statement_sink_location): Return whether
34232         failure reason was zero uses.  Move that check later.
34233         (sink_code_in_bb): Deal with zero uses by removing the stmt
34234         if possible.
34236 2017-04-24  Richard Biener  <rguenther@suse.de>
34238         PR c++/2972
34239         * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34240         pointer-based references.
34242 2017-04-24  Richard Biener  <rguenther@suse.de>
34244         PR bootstrap/79814
34245         * pass_manager.h (pass_manager::operator new): Remove.
34246         (pass_manager::operator delete): Likewise.
34247         * passes.c (pass_manager::operator new): Remove.
34248         (pass_manager::operator delete): Likewise.
34249         (pass_manager::pass_manager): Zero individual pass members.
34251 2017-04-23  Uros Bizjak  <ubizjak@gmail.com>
34253         PR target/70799
34254         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34255         <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34256         Check "XEXP (src, 1)" operand here.
34257         <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34258         Check "XEXP (src, 1)" operand here.
34259         (dimode_scalar_chain::make_vector_copies): Detect count register
34260         of a shift instruction.  Zero extend count register from QImode
34261         to DImode to satisfy vector shift pattern count operand predicate.
34262         Substitute vector shift count operand with a DImode copy.
34263         (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34264         vector register.
34266 2017-04-21  Uros Bizjak  <ubizjak@gmail.com>
34268         * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34269         Remove UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34270         (*insvqi_1_mem_rex64): Move above insv<mode>_1.  Remove
34271         UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34272         (UNSPEC_NOREX_MEM): Remove definition.
34274 2017-04-21  Richard Biener  <rguenther@suse.de>
34276         PR tree-optimization/79547
34277         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34278         Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34279         bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34280         without any constraints.
34282 2017-04-21  Richard Biener  <rguenther@suse.de>
34284         PR tree-optimization/78847
34285         * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34287 2017-04-21  Richard Biener  <rguenther@suse.de>
34289         * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34290         (build_distinct_type_copy): Likewise.
34291         (build_variant_type_copy): Likewise.
34292         * tree.c (build_qualified_type): Pass down mem-stat info.
34293         (build_distinct_type_copy): Likewise.
34294         (build_variant_type_copy): Likewise.
34296 2017-04-21  Richard Biener  <rguenther@suse.de>
34298         PR tree-optimization/80237
34299         * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34300         defaulted to NULL.
34301         (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34302         for a simplified result.
34304 2016-04-21  Richard Biener  <rguenther@suse.de>
34306         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34307         sth as strict as a simple_iv but a chrec without symbols and an
34308         operand defined in the loop we are peeling (and not some subloop).
34309         (propagate_constants_for_unrolling): Propagate all constants.
34311 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34313         PR target/79804
34314         * config/i386/i386.c (print_reg): Remove assert for disalowed
34315         regno values, call output_operand_lossage instead.
34317 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34319         PR target/78090
34320         * config/i386/constraints.md (Yc): New register constraint.
34321         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34322         Use Yc constraint for alternative 2 of operand 0.  Remove
34323         preferred_for_speed attribute.
34325 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34327         * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34328         lastprivate clauses in SIMT case.
34330 2017-04-20  Volker Reichelt  <v.reichelt@netcologne.de>
34332         * doc/invoke.texi (-Wextra-semi): Document new warning option.
34334 2017-04-20  Richard Biener  <rguenther@suse.de>
34336         PR tree-optimization/57796
34337         * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34338         as N scalar stores.
34339         (vect_model_load_cost): Cost gathers as N scalar loads.
34341 2017-04-20  Richard Biener  <rguenther@suse.de>
34343         * ggc-page.c (ggc_allocated_p): Rename to ...
34344         (safe_lookup_page_table_entry): ... this and return the lookup
34345         result.
34346         (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34348 2017-04-20  Richard Biener  <rguenther@suse.de>
34350         PR tree-optimization/80453
34351         * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34352         * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34353         from the conditions.
34354         (vn_phi_eq): Pass them down.
34355         (vn_phi_lookup): Record them.
34356         (vn_phi_insert): Likewise.
34358 2017-04-20  Matthew Fortune  <matthew.fortune@imgtec.com>
34360         * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34361         uninitialized variable warning to avoid buffer overrun.
34363 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34365         PR other/71250
34366         * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34367         is suppressed for '{ 0 }' in C.
34369 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34371         * BASE-VER: Set to 8.0.0.
34373 2017-04-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34375         * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34376         priority .init_array and .fini_array section with SECTION_NOTYPE
34377         flag.
34379 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34381         PR middle-end/80423
34382         * tree.h (build_array_type): Add typeless_storage default argument.
34383         * tree.c (type_cache_hasher::equal): Also compare
34384         TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34385         (build_array_type): Add typeless_storage argument, set
34386         TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34387         recursive call.
34388         (build_nonshared_array_type): Adjust build_array_type_1 caller.
34389         (build_array_type): Likewise.  Add typeless_storage argument.
34391 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34392             Jakub Jelinek  <jakub@redhat.com>
34394         PR tree-optimization/80426
34395         * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34396         operation on symbolic operands, also compute the overflow for the
34397         invariant part when the operation degenerates into a negation.
34399 2017-04-19  Jakub Jelinek  <jakub@redhat.com>
34401         PR debug/80461
34402         * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34403         Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34405         PR debug/80436
34406         * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34408 2017-04-19  Georg-Johann Lay  <avr@gjlay.de>
34410         PR target/80462
34411         * config/avr/avr.c (tree.h): Include it.
34412         (cgraph.h): Include it.
34413         (avr_encode_section_info): Don't warn for uninitialized progmem
34414         variable if it's just an alias.
34416 2017-04-19  Richard Biener  <rguenther@suse.de>
34418         PR ipa/65972
34419         * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34420         when needed by AutoPGO.
34422 2017-04-19  Paulo J. Matos  <paulo@matos-sorge.com>
34424         PR lto/50345
34425         * doc/lto.texi: Remove an extra 'that'.
34427 2017-04-19  Segher Boessenkool  <segher@kernel.crashing.org>
34429         PR rtl-optimization/80429
34430         * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34431         are only used in debug insns.
34433 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34434             Vladimir Makarov  <vmakarov@redhat.com>
34436         * config/sparc/predicates.md (input_operand): Add comment.  Return
34437         true for any memory operand when LRA is in progress.
34438         * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34440 2017-04-18  Jeff Law  <law@redhat.com>
34442         PR target/74563
34443         * mips.md ({return,simple_return}_internal): Do not overwrite
34444         operands[0].
34446 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34448         PR tree-optimization/80443
34449         * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34450         instead of adding 1, subtract -1 and similarly instead of subtracting
34451         1 add -1.
34453 2017-04-18  Richard Sandiford  <richard.sandiford@arm.com>
34455         PR rtl-optimization/80357
34456         * haifa-sched.c (tmp_bitmap): New variable.
34457         (model_recompute): Handle duplicate use records.
34458         (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34459         (free_global_sched_pressure_data): Free it.
34461 2017-04-18  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34463         Revert:
34464         2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34465         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34466         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34467         instead of SYSTEM_HEADER_DIR.
34469 2017-04-18  Jeff Law  <law@redhat.com>
34471         PR middle-end/80422
34472         * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34473         predecessors after walking up the insn chain.
34475 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34477         PR debug/80263
34478         * dwarf2out.c (modified_type_die): Try harder not to emit internal
34479         sizetype type into debug info.
34481 2017-04-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
34483         PR target/80099
34484         * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34485         unneeded test for TARGET_UPPER_REGS_SF.
34486         * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34488 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34490         PR sanitizer/80444
34491         * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34492         instead of gsi_after_labels.
34494 2017-04-18  Jeff Law  <law@redhat.com>
34496         * regcprop.c (maybe_mode_change): Avoid creating copies of the
34497         stack pointer.
34499         Revert:
34500         2017-04-13  Jeff Law  <law@redhat.com>
34501         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34502         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34504 2017-04-18  Georg-Johann Lay  <avr@gjlay.de>
34506         PR target/79453
34507         * config/avr/avr.c (intl.h): Include it.
34508         (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34510 2017-04-18  Martin Liska  <mliska@suse.cz>
34512         PR gcov-profile/78783
34513         * gcov-tool.c (gcov_output_files): Validate that destination
34514         file is either removed by the tool or by a user.
34516 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34517             Guy Benyei  <guybe@mellanox.com>
34519         * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34520         block, and do not negate it, the stored id is already negative.
34522 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34524         * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34526 2017-04-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
34528         PR target/80098
34529         * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34530         masks of options that should be turned off if the VSX vector
34531         options are turned off.
34532         (OTHER_P8_VECTOR_MASKS): Likewise.
34533         (OTHER_VSX_VECTOR_MASKS): Likewise.
34534         * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34535         rs6000_disable_incompatible_switches to validate no type switches
34536         like -mvsx.
34537         (rs6000_incompatible_switch): New function to disallow turning on
34538         other vector options if -mno-vsx, -mno-power8-vector, or
34539         -mno-power9-vector are specified.
34541 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34543         * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34545 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34547         * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34548         * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34549         * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34550         (ARG_POINTER_CFA_OFFSET): Likewise.
34552 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34554         * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34555         conditions to take advantage of various optimizations.
34557 2017-04-13  Jeff Law  <law@redhat.com>
34559         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34560         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34561         (zero_extendsidi2_dext): Likewise.
34563 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34565         PR sanitizer/80403
34566         * fold-const.c (fold_ternary_loc): Revert
34567         use op0 instead of fold_convert_loc (loc, type, arg0) part of
34568         2017-04-12 change.
34570 2017-04-13  Vladimir Makarov  <vmakarov@redhat.com>
34572         PR rtl-optimization/80343
34573         * lra-remat.c (update_scratch_ops): Assign original hard reg to
34574         new scratch pseudo.
34576 2017-04-13  Denis Khalikov <d.khalikov@partner.samsung.com>
34578         PR sanitizer/80414
34579         * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34580         to ubsan_encode_value.
34582 2017-04-13  Jeff Law  <law@redhat.com>
34584         * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34585         appearing in DEBUG_INSNs.
34587 2017-04-13  Martin Liska  <mliska@suse.cz>
34589         PR gcov-profile/80413
34590         * gcov-io.c (gcov_write_string): Copy to buffer just when
34591         allocated size is greater than zero.
34593 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34595         PR debug/80321
34596         * dwarf2out.c (decls_for_scope): Ignore declarations of
34597         current_function_decl in BLOCK_NONLOCALIZED_VARS.
34599 2017-04-12  Jan Hubicka  <hubicka@ucw.cz>
34601         PR lto/69953
34602         * ipa-visibility.c (non_local_p): Fix typos.
34603         (localize_node): When localizing symbol in same comdat group,
34604         dissolve the group only when we know external symbols are going
34605         to be privatized.
34606         (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34608 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34610         PR tree-optimization/79390
34611         * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34612         order does not result in usable sequence, retry with reversed operand
34613         order.
34615         PR sanitizer/80403
34616         PR sanitizer/80404
34617         PR sanitizer/80405
34618         * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34619         to fold_build2_loc.  Convert TREE_OPERAND (tem, 0) to type.  Use
34620         op0 instead of fold_convert_loc (loc, type, arg0).
34622 2017-04-12  Jeff Law  <law@redhat.com>
34624         * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34625         has a delay slot in the generated code.
34627         * config/cris/cris.md (cris_preferred_reload_class): Return
34628         GENNONACR_REGS rather than GENERAL_REGS.
34630 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34632         PR c/80163
34633         * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34634         vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34635         signedness of the result type.
34637 2017-04-12  Richard Biener  <rguenther@suse.de>
34638             Jeff Law  <law@redhat.com>
34640         PR tree-optimization/80359
34641         * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34642         trim stores to TARGET_MEM_REFs.
34644 2017-04-12  Richard Biener  <rguenther@suse.de>
34646         PR tree-optimization/79390
34647         * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34648         threading case even more.
34650 2017-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
34652         PR target/80382
34653         * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34654         for quad_address_p for TImode, instead of just not indexed_address.
34656 2017-04-12  Richard Biener  <rguenther@suse.de>
34657             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34659         PR middle-end/79671
34660         * alias.c (component_uses_parent_alias_set_from): Handle
34661         TYPE_TYPELESS_STORAGE.
34662         (get_alias_set): Likewise.
34663         * tree-core.h (tree_type_common): Add typeless_storage flag.
34664         * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34665         * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34666         for types containing members with TYPE_TYPELESS_STORAGE.
34667         (place_field): Likewise.
34668         (layout_type): Likewise for ARRAY_TYPE.
34669         * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34670         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34671         TYPE_TYPELESS_STORAGE.
34672         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34674 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34676         PR sanitizer/80349
34677         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34678         first argument to type.
34680 2017-04-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34682         PR target/80376
34683         PR target/80315
34684         * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34685         CONST0_RTX (mode) rather than const0_rtx where appropriate.
34686         (rs6000_expand_binop_builtin): Likewise.
34687         (rs6000_expand_ternop_builtin): Likewise; also add missing
34688         vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34689         vshasigma built-ins.
34690         * doc/extend.texi: Document that vec_xxpermdi's third argument
34691         must be a constant.
34693 2017-04-11  Uros Bizjak  <ubizjak@gmail.com>
34695         * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34696         Use shift_const cost parameter when calculating gain of STV shifts.
34698 2017-04-11  Vladimir Makarov  <vmakarov@redhat.com>
34700         PR rtl-optimization/70478
34701         * lra-constraints.c (process_alt_operands): Check memory for
34702         disfavoring memory insn operand.
34704 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34706         PR middle-end/80100
34707         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34708         left shift in unsigned HOST_WIDE_INT type.
34710         PR rtl-optimization/80385
34711         * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34712         (not (neg X)) into (plus X -1) for complex or non-integral modes.
34714         PR libgomp/80394
34715         * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34716         if they have any depend clauses.
34718 2017-04-11  Martin Liska  <mliska@suse.cz>
34720         PR ipa/80212
34721         * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34722         * ipa-split.c (split_function): Create a local comdat symbol
34723         if caller is in a comdat group.
34725 2017-04-11  Martin Liska  <mliska@suse.cz>
34727         PR ipa/80212
34728         * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34729         flags.
34731 2017-04-11  Martin Sebor  <msebor@redhat.com>
34733         PR middle-end/80364
34734         * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34735         always use the int type.  Use INTEGRAL_TYPE_P() rather than testing
34736         for INTEGER_TYPE.
34737         (directive::set_width, directive::set_precision, format_character):
34738         Adjust.
34739         (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34740         INTEGER_TYPE.
34742 2017-04-11  Richard Earnshaw  <rearnsha@arm.com>
34744         PR target/80389
34745         * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34746         conflict, set target->arch_name instead of target->cpu_name.
34748 2017-04-11  Richard Biener  <rguenther@suse.de>
34750         PR tree-optimization/80374
34751         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34752         build_zero_cst, remove fold_convertible_p check again.
34754 2017-04-11  Martin Liska  <mliska@suse.cz>
34756         PR sanitizer/70878
34757         * ubsan.c (instrument_object_size): Do not instrument register
34758         variables.
34760 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34762         PR target/80381
34763         * config/i386/i386-builtin-types.def
34764         (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34765         V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34766         V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34767         V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34768         V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34769         V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34770         V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34771         V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34772         V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34773         V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34774         V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34775         V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34776         V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34777         V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34778         V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34779         V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34780         V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34781         V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34782         * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34783         __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34784         __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34785         __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34786         __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34787         __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34788         __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34789         __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34790         __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34791         __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34792         __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34793         __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34794         __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34795         __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34796         __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34797         __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34798         __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34799         __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34800         __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34801         __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34802         __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34803         __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34804         __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34805         __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34806         __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34807         __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34808         __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34809         __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34810         aliases.
34811         * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34812         flag to second_arg_count, handle 4 argument function type _COUNT
34813         aliases, handle second_arg_count on second argument rather than last.
34815 2017-04-10  Jeff Law  <law@redhat.com>
34817         PR tree-optimization/80374
34818         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34819         record anything if we can not convert integer_zero_node to the
34820         desired type.
34822 2017-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
34824         PR target/80108
34825         * config/rs6000/rs6000.c (rs6000_option_override_internal):
34826         Enhance special handling given to the TARGET_P9_MINMAX option in
34827         relation to certain other options.
34829 2017-04-10  Bin Cheng  <bin.cheng@arm.com>
34831         PR tree-optimization/80153
34832         * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34833         remove POINTER_PLUS_EXPR's base part directly, rather than through
34834         aff_tree.
34836 2017-04-10  Richard Biener  <rguenther@suse.de>
34837             Bin Cheng  <bin.cheng@arm.com>
34839         PR tree-optimization/80153
34840         * tree-affine.c (aff_combination_to_tree): Get base pointer from
34841         the first element of pointer type aff_tree.  Build result expr in
34842         aff_tree's type.
34843         (add_elt_to_tree): Convert to type unconditionally.  Remove other
34844         fold_convert calls.
34845         * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34846         (rewrite_use_nonlinear_expr): Check invariant using iv information.
34848 2017-04-10  Richard Biener  <rguenther@suse.de>
34850         * tree-ssa-structalias.c (find_func_aliases): Properly handle
34851         asm inputs.
34853 2017-04-10  Vladimir Makarov  <vmakarov@redhat.com>
34855         PR rtl-optimization/70478
34856         * lra-constraints.c (curr_small_class_check): New.
34857         (update_and_check_small_class_inputs): New.
34858         (process_alt_operands): Update curr_small_class_check.  Disfavor
34859         alternative insn memory operands.  Check available regs for small
34860         class operands.
34862 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
34864         PR target/80057
34865         * config/mips/mips.opt (-mvirt): Update description.
34866         * doc/invoke.texi (-mvirt): Likewise.
34868 2017-04-10  Richard Biener  <rguenther@suse.de>
34870         PR middle-end/80362
34871         * fold-const.c (fold_binary_loc): Look at unstripped ops when
34872         looking for NEGATE_EXPR in -A / -B to A / B folding.
34874 2017-04-10  Martin Liska  <mliska@suse.cz>
34876         PR gcov-profile/80224
34877         * gcov.c (print_usage): Fix usage string.
34878         (get_gcov_intermediate_filename): Remove.
34879         (output_gcov_file): Use both for normal and intermediate format.
34880         (generate_results): Do not initialize special file for
34881         intermediate format.
34883 2017-04-10  Richard Biener  <rguenther@suse.de>
34885         PR tree-optimization/80304
34886         * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34887         for safelen.
34889 2017-04-10  Nathan Sidwell  <nathan@acm.org>
34891         PR target/79905
34892         * config/rs6000/rs6000.c (rs6000_vector_type): New.
34893         (rs6000_init_builtins): Use it.
34895 2016-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34897         * config/arm/arm.md (<mrc>): Add mode to SET source.
34898         (<mrrc>): Likewise.
34900 2017-04-10  Richard Biener  <rguenther@suse.de>
34902         PR middle-end/80344
34903         * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34905 2017-04-10  Jakub Jelinek  <jakub@redhat.com>
34907         PR target/80324
34908         * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34909         _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34910         _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34911         _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
34912         _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
34913         _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
34914         _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
34915         _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
34916         _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
34917         _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
34918         _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
34919         _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
34920         _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
34921         _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
34922         _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
34923         _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
34924         _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
34925         _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
34926         _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
34927         _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
34928         _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
34929         _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
34930         _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
34932 2017-04-08  Vladimir Makarov  <vmakarov@redhat.com>
34934         PR rtl-optimization/70478
34935         * lra-constraints.c: Reverse the last patch.
34937 2017-04-08  Andreas Tobler  <andreast@gcc.gnu.org>
34939         * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
34940         Add comment for WCHAR_T.
34942 2017-04-08  Martin Liska  <mliska@suse.cz>
34944         Revert:
34945         2017-04-07  Martin Liska  <mliska@suse.cz>
34947         PR ipa/80212
34948         * ipa-split.c (split_function): Add function part to a same comdat
34949         group.
34951 2017-04-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34953         PR target/80358
34954         * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
34956 2017-04-07  Pat Haugen  <pthaugen@us.ibm.com>
34958         * rs6000/rs6000.c (vec_load_pendulum): Rename...
34959         (vec_pairing): ...to this.
34960         (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
34961         (rs6000_sched_init): Adjust for name change.
34962         (struct rs6000_sched_context): Likewise.
34963         (rs6000_init_sched_context): Likewise.
34964         (rs6000_set_sched_context): Likewise.
34966 2017-04-07  Jakub Jelinek  <jakub@redhat.com>
34968         PR target/80322
34969         PR target/80323
34970         PR target/80325
34971         PR target/80326
34972         * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
34973         intrinsics.
34974         * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
34975         _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
34976         _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
34978 2017-04-07  Andreas Tobler  <andreast@gcc.gnu.org>
34980         * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
34982 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34984         PR rtl-optimization/70703
34985         * ira-color.c (update_conflict_hard_regno_costs): Use
34986         int64_t instead of HOST_WIDE_INT.
34988 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34990         PR rtl-optimization/70478
34991         * lra-constraints.c (process_alt_operands): Disfavor alternative
34992         insn memory operands.
34994 2017-04-07  Jeff Law  <law@redhat.com>
34996         * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
34997         CALL and NOTE_INSN_CALL_ARG_LOCATION.
34999 2017-04-07  Martin Liska  <mliska@suse.cz>
35001         PR target/79889
35002         * config/aarch64/aarch64.c (aarch64_process_target_attr):
35003         Show error message instead of an ICE.
35005 2017-04-07  Martin Liska  <mliska@suse.cz>
35007         PR ipa/80212
35008         * ipa-split.c (split_function): Add function part to a same comdat
35009         group.
35011 2017-04-07  Richard Biener  <rguenther@suse.de>
35013         PR middle-end/80341
35014         * tree.c (get_unwidened): Also handle ! for_type case for
35015         INTEGER_CSTs.
35016         * convert.c (do_narrow): Split out from ...
35017         (convert_to_integer_1): ... here.  Do not pass final truncation
35018         type to get_unwidened for TRUNC_DIV_EXPR.
35020 2017-04-07  Richard Biener  <rguenther@suse.de>
35022         * tree-affine.c (wide_int_ext_for_comb): Take type rather
35023         than aff_tree.
35024         (aff_combination_const): Adjust.
35025         (aff_combination_scale): Likewise.
35026         (aff_combination_add_elt): Likewise.
35027         (aff_combination_add_cst): Likewise.
35028         (aff_combination_convert): Likewise.
35029         (add_elt_to_tree): Likewise.  Remove unused argument.
35030         (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
35032 2017-04-07  Sebastian Huber  <sebastian.huber@embedded-brains.de>
35034         * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
35035         definition.
35036         * config/arm/arm.c (arm_default_short_enums): Use
35037         ARM_DEFAULT_SHORT_ENUMS.
35038         * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
35040 2017-04-06  Jakub Jelinek  <jakub@redhat.com>
35042         PR debug/80234
35043         * dwarf2out.c (gen_member_die): Handle C++17 inline static data
35044         members with redundant out-of-class redeclaration.
35046 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
35048         PR target/80286
35049         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
35050         * config/i386/i386.md (*zero_extendsidi2):
35051         Add (?*x,*x) and (?*v,*v) alternatives.
35053 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
35055         PR target/79733
35056         * config/i386/i386.c (ix86_expand_builtin)
35057         <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
35058         mode from insn data. Convert operands to insn operand mode.
35059         Copy operands that don't satisfy insn predicate to a register.
35061 2017-04-06  Sam Thursfield  <sam.thursfield@codethink.co.uk>
35063         * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
35064         Update comments.
35066 2017-04-06  Richard Biener  <rguenther@suse.de>
35068         PR tree-optimization/80334
35069         * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
35070         preserve alignment of accesses.
35072 2017-04-06  Richard Biener  <rguenther@suse.de>
35074         PR tree-optimization/80262
35075         * tree-sra.c (build_ref_for_offset): Preserve address-space
35076         information.
35077         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
35078         Drop useless address-space information on MEM_REF offsets.
35080 2017-04-05  Andreas Schwab  <schwab@linux-m68k.org>
35082         * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
35084 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35086         PR rtl-optimization/70703
35087         * ira-color.c (update_conflict_hard_regno_costs): Use
35088         HOST_WIDE_INT instead of long.
35090 2017-04-05  Uros Bizjak  <ubizjak@gmail.com>
35092         PR target/80298
35093         * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
35094         not defined for x86_64 target.  Add -mmmx target option when __SSE2__
35095         is not defined.
35096         * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
35097         for x86_64 target.  Handle -m3dnowa option.
35099 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35101         PR rtl-optimization/70703
35102         * ira-color.c (update_costs_from_allocno): Use the smallest mode.
35103         (update_conflict_hard_regno_costs): Use long instead of unsigned
35104         arithmetic for cost calculation.
35106 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35107             Bernd Edlinger  <bernd.edlinger@hotmail.de>
35109         PR sanitizer/80308
35110         * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
35111         for big endian.
35113 2017-04-05  Eric Botcazou  <ebotcazou@adacore.com>
35115         PR target/78002
35116         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
35117         ptr_mode with Pmode throughout.
35118         * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
35119         into probe_stack_range and use DImode.
35121 2017-04-05  Dominik Vogt  <vogt@linux.vnet.ibm.com>
35123         PR target/79890
35124         * config/s390/s390.c (s390_register_info_gprtofpr): Return if
35125         call_eh_return is true.
35127 2017-04-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35129         * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
35130         Initialize last_match_fntype_index.
35132 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35134         PR target/80310
35135         * tree-nvr.c: Include internal-fn.h.
35136         (pass_return_slot::execute): Ignore internal calls without
35137         direct optab.
35139 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35140             Richard Biener  <rguenther@suse.de>
35142         PR c++/80297
35143         * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
35144         captures used multiple times, except for the last use.
35145         * generic-match-head.c: Include gimplify.h.
35147 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35149         PR tree-optimization/79390
35150         * target.h (struct noce_if_info): Declare.
35151         * targhooks.h (default_noce_conversion_profitable_p): Declare.
35152         * target.def (noce_conversion_profitable_p): New target hook.
35153         * ifcvt.h (struct noce_if_info): New type, moved from ...
35154         * ifcvt.c (struct noce_if_info): ... here.
35155         (noce_conversion_profitable_p): Renamed to ...
35156         (default_noce_conversion_profitable_p): ... this.  No longer
35157         static nor inline.
35158         (noce_try_store_flag_constants, noce_try_addcc,
35159         noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
35160         noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
35161         instead of noce_conversion_profitable_p.
35162         * config/i386/i386.c: Include ifcvt.h.
35163         (ix86_option_override_internal): Don't override
35164         PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
35165         (ix86_noce_conversion_profitable_p): New function.
35166         (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
35167         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
35168         * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
35169         * doc/tm.texi: Regenerated.
35171 2017-04-04  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35173         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
35174         correction.
35176 2017-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35178         PR target/80307
35179         * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
35180         instructions for small multiply cores.
35182 2017-04-04  Jeff Law  <law@redhat.com>
35184         * config/mips/mips.c (mips_multi_add): Zero initialize the newly
35185         added member.
35186         (mips_expand_vec_perm_const): Initialize elements in orig_perm
35187         that are not set by the loop over the elements.
35189 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35191         PR target/80286
35192         * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
35193         int mode, convert_modes it to mode as unsigned, otherwise use
35194         lowpart_subreg to mode rather than SImode.
35195         * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
35196         ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
35197         Use DImode instead of SImode for the shift count operand.
35198         * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
35199         Likewise.
35201 2017-04-04  Richard Biener  <rguenther@suse.de>
35203         PR middle-end/80281
35204         * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
35205         arithmetic done for the negate or the plus.  Simplify.
35206         (A - (-B) -> A + B): Likewise.
35207         * fold-const.c (split_tree): Make sure to not negate pointers.
35209 2017-04-04  Segher Boessenkool  <segher@kernel.crashing.org>
35211         PR rtl-optimization/60818
35212         * simplify-rtx.c (simplify_binary_operation_1): Do not replace
35213         a compare of comparisons with the thing compared if this results
35214         in a different machine mode.
35216 2017-04-03  Jonathan Wakely  <jwakely@redhat.com>
35218         * alias.c (base_alias_check): Fix typo in comment.
35219         * cgraph.h (class ipa_polymorphic_call_context): Likewise.
35220         * cgraphunit.c (symbol_table::compile): Likewise.
35221         * collect2.c (maybe_run_lto_and_relink): Likewise.
35222         * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
35223         * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
35224         * config/avr/avr.c (avr_map_op_t): Likewise.
35225         * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
35226         * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
35227         * config/epiphany/epiphany.md (movcc): Likewise.
35228         * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
35229         * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
35230         Likewise.
35231         * config/mips/mips.c (mips_save_restore_reg): Likewise.
35232         * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
35233         * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
35234         * config/sh/sh.c (sh_rtx_costs): Likewise.
35235         * fold-const.c (fold_truth_andor): Likewise.
35236         * genautomata.c (collapse_flag): Likewise.
35237         * gengtype.h (struct type::u::s): Likewise.
35238         * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35239         * input.c (FORMAT_AMOUNT): Likewise.
35240         * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35241         (known_aggs_to_agg_replacement_list): Likewise.
35242         * ipa-inline-analysis.c: Likewise.
35243         * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35244         * ipa-polymorphic-call.c
35245         (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35246         * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35247         * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35248         Likewise.
35249         * modulo-sched.c (apply_reg_moves): Likewise.
35250         * omp-expand.c (build_omp_regions_1): Likewise.
35251         * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35252         * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35253         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35254         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35255         * value-prof.c: Likewise.
35256         * var-tracking.c (val_reset): Likewise.
35258 2017-04-03  Richard Biener  <rguenther@suse.de>
35260         PR tree-optimization/80275
35261         * fold-const.c (split_address_to_core_and_offset): Handle
35262         POINTER_PLUS_EXPR.
35264 2017-04-03  Eric Botcazou  <ebotcazou@adacore.com>
35266         * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35267         descriptors is at least equal to that of functions.
35269 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35271         * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35273 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35275         PR target/80250
35276         * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35277         (mov<IMOD4:mode>): New expander.
35278         (*mov<IMOD4:mode>_internal): New insn and split pattern.
35280 2017-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
35282         PR rtl-optimization/79405
35283         * fwprop.c (propagations_left): New variable.
35284         (forward_propagate_into): Decrement it.
35285         (fwprop_init): Initialize it.
35286         (fw_prop): If the variable has reached zero, stop propagating.
35287         (fwprop_addr): Ditto.
35289 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35291         PR debug/79255
35292         * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35293         a FUNCTION_DECL, pass it as decl instead of origin to
35294         process_scope_var.
35296 2017-03-31  Alexander Monakov  <amonakov@ispras.ru>
35298         * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35299         string.
35301 2017-03-31  Pat Haugen  <pthaugen@us.ibm.com>
35303         PR target/80107
35304         * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35305         TARGET_VSX_SMALL_INTEGER.
35307 2017-03-31  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35309         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35310         reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35312 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
35314         * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35315         extraction from odd-numbered MSA register.
35317 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35319         PR middle-end/80173
35320         * expmed.c (store_bit_field_1): Don't attempt to create
35321         a word subreg out of hard registers wider than word if they
35322         have HARD_REGNO_NREGS of 1 for their mode.
35324         PR middle-end/80163
35325         * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35326         conversions to integer types wider than word and pointer.
35328         PR debug/80025
35329         * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35330         (rtx_equal_for_cselib_p): Pass 0 to it.
35331         * cselib.c (cselib_hasher::equal): Likewise.
35332         (rtx_equal_for_cselib_1): Add depth argument.  If depth
35333         is 128, don't look up VALUE locs and punt.  Increment
35334         depth in recursive calls when walking VALUE locs.
35336 2017-03-31  Bernd Edlinger  <bernd.edlinger@hotmail.de>
35338         * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35339         (make_gcov_file_name): Use the canonical path name for generating
35340         the MD5 value.
35341         (read_line): Fix handling of files with ascii null bytes.
35343 2017-03-30  Matthew Fortune  <matthew.fortune@imgtec.com>
35345         * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35346         to initialise a vector register instead
35347         of using a const_int.
35349 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35351         PR translation/80189
35352         * gimplify.c (omp_default_clause): Use %qs instead of %s in
35353         diagnostic messages.
35355 2017-03-30  Peter Bergner  <bergner@vnet.ibm.com>
35357         PR target/80246
35358         * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35359         (dfp_diex_<mode>): Update mode of operand 1.
35360         * doc/extend.texi (dxex, dxexq): Document change to return type.
35361         (diex, diexq): Document change to argument type.
35363 2017-03-30  Martin Jambor  <mjambor@suse.cz>
35365         PR ipa/77333
35366         * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35367         * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35368         it reflects the signature changes performed at the callee side.
35369         * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35370         to cgraph_build_function_type_skip_args.
35371         (build_function_decl_skip_args): Adjust call to the above function.
35373 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35375         PR target/80206
35376         * config/i386/sse.md
35377         (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35378         register as dest whenever it is a MEM not rtx_equal_p to the
35379         corresponding dup operand, and when forcing into reg move the
35380         reg into the memory afterwards.
35381         (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35382         Likewise.  Use <ssehalfvecmode> instead of <ssequartermode>
35383         for the force_reg mode.
35384         (avx512vl_vextractf128<mode>): Use register as dest either
35385         always when a MEM, or when it is a MEM not rtx_equal_p to the
35386         corresponding dup operand, or even not when it is a CONST_VECTOR
35387         depending on the mode and lo vs. hi.
35388         (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35389         parens.
35390         (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35391         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35392         Likewise.  Require that operands[2] is even.
35393         (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35394         Remove extraneous parens.  Require that operands[2] is a multiple
35395         of 4.
35396         (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35397         operands[0] is a MEM if <mask_applied>, the predicates/constraints
35398         disallow memory then.
35400 2017-03-30  Richard Biener  <rguenther@suse.de>
35402         PR tree-optimization/77498
35403         * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35404         to non-constants over backedges.
35406 2017-03-29  Segher Boessenkool  <segher@kernel.crashing.org>
35408         PR rtl-optimization/80233
35409         * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35410         as last_combined_insn.  Do not test for BARRIER_P separately.
35412 2017-03-29  Andreas Schwab  <schwab@suse.de>
35414         PR ada/80146
35415         * calls.c (prepare_call_address): Convert funexp to Pmode before
35416         copying to temp reg.
35418 2017-03-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35420         PR tree-optimization/80158
35421         * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35422         Handle possible future case of more than one alternate
35423         interpretation.
35424         (replace_rhs_if_not_dup): Likewise.
35425         (replace_one_candidate): Likewise.
35427 2017-03-28  Vladimir Makarov  <vmakarov@redhat.com>
35429         PR rtl-optimization/80193
35430         * ira.c (ira): Do not check allocation for LRA.
35432 2017-03-28  Alexander Monakov  <amonakov@ispras.ru>
35434         * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35435         (nvptx_output_simt_exit): Declare.
35436         * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35437         cfun->machine->unisimt_location.  Handle NULL unisimt_predicate.
35438         (init_softstack_frame): Move initialization of crtl->is_leaf to...
35439         (nvptx_declare_function_name): ...here.  Emit declaration of local
35440         memory space buffer for omp_simt_enter insn.
35441         (nvptx_output_unisimt_switch): New.
35442         (nvptx_output_softstack_switch): New.
35443         (nvptx_output_simt_enter): New.
35444         (nvptx_output_simt_exit): New.
35445         * config/nvptx/nvptx.h (struct machine_function): New fields
35446         has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35447         * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35448         (UNSPECV_SIMT_EXIT): Ditto.
35449         (omp_simt_enter_insn): New insn.
35450         (omp_simt_enter): New expansion.
35451         (omp_simt_exit): New insn.
35452         * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35454         * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35455         (expand_GOMP_SIMT_ENTER_ALLOC): New.
35456         (expand_GOMP_SIMT_EXIT): New.
35457         * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35458         (GOMP_SIMT_ENTER_ALLOC): Ditto.
35459         (GOMP_SIMT_EXIT): Ditto.
35460         * target-insns.def (omp_simt_enter): New insn.
35461         (omp_simt_exit): Ditto.
35462         * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35463         simt_dlist.
35464         (lower_rec_simd_input_clauses): Implement SIMT privatization.
35465         (lower_rec_input_clauses): Likewise.
35466         (lower_lastprivate_clauses): Handle SIMT privatization.
35468         * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35469         (ompdevlow_adjust_simt_enter): New.
35470         (find_simtpriv_var_op): New.
35471         (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35472         IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35474         * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35475         * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35476         (copy_decl_for_dup_finish): Ditto.
35478         * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35480 2017-03-28  Uros Bizjak  <ubizjak@gmail.com>
35482         PR target/53383
35483         * config/i386/i386.c (ix86_option_override_internal): Always
35484         allow -mpreferred-stack-boundary=3 for 64-bit targets.
35486 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35488         * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35490 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35492         * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35493         mark new edge's irreducible flag accordign to it.
35494         (vect_do_peeling): Check loop preheader edge's irreducible flag
35495         and pass it to function slpeel_add_loop_guard.
35497 2017-03-28  Richard Sandiford  <richard.sandiford@arm.com>
35499         PR tree-optimization/80218
35500         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35501         Update block frequencies and counts.
35503 2017-03-28  Richard Biener  <rguenther@suse.de>
35505         PR tree-optimization/78644
35506         * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35507         of a simplification result we may not use it at all.
35509 2017-03-28  Richard Biener  <rguenther@suse.de>
35511         PR ipa/80205
35512         * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35513         without arguments, generate default definition of a SSA name.
35515 2017-03-28  Richard Biener  <rguenther@suse.de>
35517         PR middle-end/80222
35518         * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35519         TYPE_REF_CAN_ALIAS_ALL references.
35520         * fold-const.c (fold_indirect_ref_1): Likewise.
35522 2017-03-28  Martin Liska  <mliska@suse.cz>
35524         PR ipa/80104
35525         * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35526         thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35528 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35529             Thomas Petazzoni  <thomas.petazzoni@free-electrons.com>
35531         * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35532         (EXTRA_SPECS): Define.
35533         (SUBTARGET_EXTRA_SPECS): Likewise.
35534         (SUBTARGET_CPP_SPEC): Likewise.
35535         * config/arc/elf.h (EXTRA_SPECS): Renamed to
35536         SUBTARGET_EXTRA_SPECS.
35537         * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35539 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35541         * config/arc/simdext.md (vst64_insn): Update pattern.
35542         (vld32wh_insn): Likewise.
35543         (vld32wl_insn): Likewise.
35544         (vld64_insn): Likewise.
35545         (vld32_insn): Likewise.
35547 2017-03-28  Marek Polacek  <polacek@redhat.com>
35549         PR sanitizer/80067
35550         * fold-const.c (fold_comparison): Use protected_set_expr_location
35551         instead of SET_EXPR_LOCATION.
35553 2017-03-28  Markus Trippelsdorf  <markus@trippelsdorf.de>
35555         * tree.c (add_expr): Avoid name lookup warning.
35557 2017-03-27  Jeff Law  <law@redhat.com>
35559         PR tree-optimization/80216
35560         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35561         function name.  Limit recursion depth.
35562         (record_temporary_equivalences): Corresponding changes.
35564 2017-03-27  Jonathan Wakely  <jwakely@redhat.com>
35566         * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35567         covered first.
35569 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35571         PR target/80102
35572         * reg-notes.def (REG_CFA_NOTE): Define.  Use it for CFA related
35573         notes.
35574         * cfgcleanup.c (reg_note_cfa_p): New array.
35575         (insns_have_identical_cfa_notes): New function.
35576         (old_insns_match_p): Don't cross-jump in between /f
35577         and non-/f instructions.  If both i1 and i2 are frame related,
35578         verify all CFA notes, their order and content.
35580 2017-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
35582         PR target/78543
35583         * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35584         HImode and SImode with zero extend to DImode to one insn.
35585         (bswap<mode>2_extenddi): Likewise.
35586         (bswapsi2_extenddi): Likewise.
35587         (bswaphi2_extendsi): Likewise.
35588         (bswaphi2): Combine bswap HImode and SImode into one insn.
35589         Separate memory insns from swapping register.
35590         (bswapsi2): Likewise.
35591         (bswap<mode>2): Likewise.
35592         (bswaphi2_internal): Delete, no longer used.
35593         (bswapsi2_internal): Likewise.
35594         (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35595         store, and gpr<-gpr swap insns.
35596         (bswap<mode>2_store): Likewise.
35597         (bswaphi2_reg): Register only splitter, combine with the splitter.
35598         (bswaphi2 splitter): Likewise.
35599         (bswapsi2_reg): Likewise.
35600         (bswapsi2 splitter): Likewise.
35601         (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35602         the insns into load, store, and register/register insns.
35603         (bswapdi2_ldbrx): Likewise.
35604         (bswapdi2_load): Likewise.
35605         (bswapdi2_store): Likewise.
35606         (bswapdi2_reg): Likewise.
35608 2017-03-27  Gunther Nikl  <gnikl@users.sourceforge.net>
35610         * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35611         (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35613 2017-03-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35615         PR target/80103
35616         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35617         add comments.
35618         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35619         special handling for target option conflicts between dform
35620         options (-mpower9-dform, -mpower9-dform-vector,
35621         -mpower9-dform-scalar) and -mno-direct-move.
35623 2017-03-27  Richard Biener  <rguenther@suse.de>
35625         PR tree-optimization/80181
35626         * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35628 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35630         * config/arc/predicates.md (move_double_src_operand): Replace the
35631         call to move_double_src_operand with a call to address_operand.
35633 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35635         * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35636         * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35637         * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35639 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35641         * config/arc/predicates.md (long_immediate_loadstore_operand):
35642         Consider scaled addresses cases.
35644 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35646         * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35647         restored when in interrupt.
35648         * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35649         doesn't have delay slot.
35651 2017-03-27  Richard Biener  <rguenther@suse.de>
35653         PR ipa/79776
35654         * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35655         inlined thunk clones.
35657 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35659         PR sanitizer/80168
35660         * asan.c (instrument_derefs): Copy over last operand from
35661         original COMPONENT_REF to the new COMPONENT_REF with
35662         DECL_BIT_FIELD_REPRESENTATIVE.
35663         * ubsan.c (instrument_object_size): Likewise.
35665 2017-03-27  Richard Biener  <rguenther@suse.de>
35667         PR tree-optimization/80170
35668         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35669         sure DR/SCEV didnt fold in constants we do not see when looking
35670         at the reference base alignment.
35672 2017-03-27  Richard Biener  <rguenther@suse.de>
35674         PR middle-end/80171
35675         * gimple-fold.c (fold_ctor_reference): Properly guard against
35676         NULL return value from canonicalize_constructor_val.
35678 2017-03-25  Uros Bizjak  <ubizjak@gmail.com>
35680         PR target/80180
35681         * config/i386/i386.c (ix86_expand_builtin)
35682         <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35683         flags reg setting and flags reg using instructions.
35684         <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto.  Use non-flags reg
35685         clobbering instructions to zero extend op2.
35687 2017-03-25  Gerald Pfeifer  <gerald@pfeifer.com>
35689         * doc/install.texi (Configuration) <--with-aix-soname>:
35690         Update link to AIX ld.
35692 2017-03-25  Bernd Schmidt  <bschmidt@redhat.com>
35694         PR rtl-optimization/80160
35695         PR rtl-optimization/80159
35696         * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35697         reg_alternate_class into account.
35699 2017-03-24  Vladimir Makarov  <vmakarov@redhat.com>
35701         PR target/80148
35702         * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35703         to consider in curr_insn_transform.
35705 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35707         * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35708         * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35709         and emit_mode_inner.
35711 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35713         * config/s390/s390-builtins.def: Add VXE builtins.  Add a flags
35714         argument to the overloaded builtin variants.  Use the new flag to
35715         deprecate certain builtin variants.
35716         * config/s390/s390-builtin-types.def: Add new builtin types.
35717         * config/s390/s390-builtins.h: Support new flags field for
35718         overloaded builtins.
35719         * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35720         (s390_macro_to_expand): Enable vector float data type.
35721         (s390_cpu_cpp_builtins_internal): Indicate support of the new
35722         builtins by incrementing the __VEC__ version number.
35723         (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35724         vec_xst.
35725         (s390_resolve_overloaded_builtin): Emit error messages depending
35726         on the builtin flags.
35727         * config/s390/s390.c (s390_expand_builtin): Support additional
35728         flags argument.  Change error message to match the messages
35729         emitted in s390-c.c.
35730         * config/s390/s390.md: New UNSPEC_* constants.
35731         (op_type): Add new instruction types.
35732         * config/s390/vecintrin.h: Add new builtins and test data class
35733         constants.
35734         * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35735         (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35736         (VEC_INEXACT, VEC_NOINEXACT): New constants.
35737         ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35738         ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35739         ("vec_mergel<mode>"): V_HW -> VEC_HW.
35741         ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35742         ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35743         ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35744         ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35746         ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35747         ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35748         ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35749         ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35751         ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35752         ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35753         ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35754         ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35755         ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35756         ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35757         ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35759         ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35760         ("vec_scatter_element<V_HW_4:mode>_DI")
35761         ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35762         ("vec_fpint<mode>", "vflls")
35763         ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35764         ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35765         ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35766         ("*vec_cmphe<mode>_cc"): ... these.
35768         ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35769         mode constant instead of magic value.
35771 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35773         * config/s390/s390.c (s390_expand_vec_compare): Support other
35774         vector floating point modes than just V2DF.
35775         (s390_expand_vcond): Likewise.
35776         (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35777         (s390_cannot_change_mode_class): Prevent mode changes between TF
35778         and V1TF in vector registers.
35779         * config/s390/s390.md (DF, SF): New mode attributes.
35780         ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35781         ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35782         SFmode support for VRs.
35783         * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35784         vector fp modes.
35785         (VFT, VF_HW): New mode iterators.
35786         (vw, sdx): New mode attributes.
35787         ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35788         ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35789         ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35790         ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35791         ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35792         also the new vector floating point modes.  Renaming to ...
35794         ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35795         ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35796         ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35797         ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35798         ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35799         ("vec_unordered<mode>"): ... these.
35801         ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35802         ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35803         ("*vec_extendv2df"): New insn definitions.
35805 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35807         * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35808         ("mulditi3_2", "*muldi3_sign"): New patterns.
35809         ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35810         rename the pattern definition.
35812 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35814         * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35815         expander.
35816         ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35818 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35820         * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35821         instruction if possible.
35822         * config/s390/vector.md (vec_halfnumelts): New mode
35823         attribute.
35824         ("*vec_vllezlf<mode>"): New pattern.
35826 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35828         * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35829         ("popcountv4si2", "popcountv2di2"): Rename to ...
35830         ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35831         ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35832         condition.
35833         ("popcount<mode>2_vxe"): New pattern.
35835 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35837         * common/config/s390/s390-common.c (processor_flags_table): Add
35838         arch12.
35839         * config.gcc: Add arch12.
35840         * config/s390/driver-native.c (s390_host_detect_local_cpu):
35841         Default to arch12 for unknown CPU model numbers.
35842         * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35843         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35844         PROCESSOR_max sanity check.
35845         * config/s390/s390-opts.h (enum processor_type): Add
35846         PROCESSOR_ARCH12.
35847         * config/s390/s390.c (processor_table): Add arch12.
35848         (s390_expand_builtin): Add check for B_VXE flag.
35849         (s390_issue_rate): Add PROCESSOR_ARCH12.
35850         (s390_get_sched_attrmask): Likewise.
35851         (s390_get_unit_mask): Likewise.
35852         (s390_sched_score): Enable z13 scheduling for arch12.
35853         (s390_sched_reorder): Likewise.
35854         (s390_sched_variable_issue): Likewise.
35855         * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35856         PF_VXE.
35857         (s390_tune_attr): Use z13 scheduling also for arch12.
35858         (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35859         (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35860         (TARGET_VXE_P): New macros.
35861         * config/s390/s390.md: Add arch12 to cpu attribute.  Add arch12
35862         and vxe to cpu_facility.  Add arch12 and vxe to enabled attribute.
35863         * config/s390/s390.opt: Add arch12 as processor_type.
35865 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35867         * config/s390/s390.md
35868         ("fixuns_truncdddi2", "fixuns_trunctddi2")
35869         ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35870         ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35872         ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35873         Rename expanders to ...
35875         ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35876         ("fixuns_truncdddi2_emu"): ... these.
35878         ("fixuns_trunc<mode>si2_emu"): New expander.
35880         ("*fixuns_truncdfdi2_z13"): Rename to ...
35881         ("*fixuns_truncdfdi2_vx"): ... this.
35883 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35885         * config/s390/2964.md: Remove the single element vector compare
35886         instructions which are no longer used.
35887         * config/s390/s390.c (s390_select_ccmode): Remove handling of
35888         vector CCmodes.
35889         (s390_canonicalize_comparison): Remove handling of DFmode
35890         compares.
35891         (s390_expand_vec_compare_scalar): Remove function.
35892         (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35893         * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35894         pattern.
35895         ("*cmp<mode>_ccs"): Add wfcdb instruction.
35897 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35899         * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35900         FP zero.
35901         ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35902         will anyway by matched by mov<mode>_64dfp.
35904 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35906         * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35907         vlef/vstef.  Add missing operand to vleif.
35909 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35911         * config/s390/s390.c (s390_expand_vec_init): Enable vector load
35912         pair for all vector types with 64 bit elements.
35913         * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
35914         * config/s390/vector.md (V_HW_64): ... here.
35915         (V_128_NOSINGLE): New mode iterator.
35916         ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
35917         ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
35918         ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
35919         ("*vec_load_pairv2di"): Change to ...
35920         ("*vec_load_pair<mode>"): ... this one.
35922 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35924         * config/s390/constraints.md: Add comments.
35925         (jKK): Reject element sizes > 8 bytes.
35926         * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
35927         s_operands.
35928         * config/s390/s390.md: Add the s_operand checks formerly in
35929         s390_split_ok_p to various splitters where they are still
35930         required.
35931         * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
35932         for 128 bit vectors.  Plus two splitters.
35934 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35936         * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
35937         the file.
35939 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35941         PR target/79893
35942         * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
35943         error if the boundary argument is not constant.
35945 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35947         PR rtl-optimization/80112
35948         * loop-doloop.c (doloop_condition_get): Don't check condition
35949         if cmp isn't SET with IF_THEN_ELSE src.
35951 2017-03-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35953         PR tree-optimization/80158
35954         * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
35955         replacing a candidate statement, also replace it for the
35956         candidate's alternate interpretation.
35957         (replace_rhs_if_not_dup): Likewise.
35958         (replace_one_candidate): Likewise.
35960 2017-03-24  Richard Biener  <rguenther@suse.de>
35962         PR tree-optimization/80167
35963         * graphite-isl-ast-to-gimple.c
35964         (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
35965         properly.
35966         (translate_isl_ast_to_gimple::get_rename): Likewise.
35968 2017-03-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35970         * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
35971         handling of certain combinations of target options, including the
35972         combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
35973         -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
35975 2017-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
35977         PR target/71436
35978         * config/arm/arm.md (*load_multiple): Add reload_completed to
35979         matching condition.
35981 2017-03-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35982             Richard Biener  <rguenth@suse.de>
35984         PR tree-optimization/79908
35985         PR tree-optimization/80136
35986         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
35987         been cast away, gimplify_and_add suffices.
35989 2017-03-23  Markus Trippelsdorf  <markus@trippelsdorf.de>
35991         * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
35993 2017-03-23  Richard Biener  <rguenther@suse.de>
35995         PR tree-optimization/80032
35996         * gimplify.c (gimple_push_cleanup): Forced unconditional
35997         cleanups still have to go to the conditional_cleanups
35998         sequence.
36000 2017-03-22  Jakub Jelinek  <jakub@redhat.com>
36002         PR tree-optimization/80072
36003         * tree-ssa-reassoc.c (struct operand_entry): Change id field type
36004         to unsigned int.
36005         (next_operand_entry_id): Change type to unsigned int.
36006         (sort_by_operand_rank): Make sure to return the right return value
36007         even if unsigned fields are bigger than INT_MAX.
36008         (struct oecount): Change cnt and id type to unsigned int.
36009         (oecount_hasher::equal): Formatting fix.
36010         (oecount_cmp): Make sure to return the right return value
36011         even if unsigned fields are bigger than INT_MAX.
36012         (undistribute_ops_list): Change next_oecount_id type to unsigned int.
36014         PR c++/80129
36015         * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
36016         TREE_READONLY on result if writing it more than once.
36018         PR sanitizer/80110
36019         * doc/invoke.texi (-fsanitize=thread): Document that with
36020         -fnon-call-exceptions atomics are not able to throw
36021         exceptions.
36023         PR sanitizer/80110
36024         * tsan.c: Include tree-eh.h.
36025         (instrument_builtin_call): Call maybe_clean_eh_stmt or
36026         maybe_clean_or_replace_eh_stmt where needed.
36027         (instrument_memory_accesses): Add cfg_changed argument.
36028         Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
36029         if it returned true.
36030         (tsan_pass): Adjust caller.  Return TODO_cleanup_cfg if cfg_changed.
36032         PR rtl-optimization/63191
36033         * config/i386/i386.c (ix86_delegitimize_address): Turn into small
36034         wrapper function, moved the whole old content into ...
36035         (ix86_delegitimize_address_1): ... this.  New inline function.
36036         (ix86_find_base_term): Use ix86_delegitimize_address_1 with
36037         true as last argument instead of ix86_delegitimize_address.
36039 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
36041         * config/aarch64/aarch64.c (generic_branch_cost): Copy
36042         cortexa57_branch_cost.
36044 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
36046         * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
36048 2017-03-21  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36050         PR target/80123
36051         * doc/md.texi (Constraints): Document wA constraint.
36052         * config/rs6000/constraints.md (wA): New.
36053         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
36054         (rs6000_init_hard_regno_mode_ok): Init wA constraint.
36055         * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
36056         * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
36058 2017-03-22  Cesar Philippidis  <cesar@codesourcery.com>
36060         PR c++/80029
36061         * gimplify.c (is_oacc_declared): New function.
36062         (oacc_default_clause): Use it to set default flags for acc declared
36063         variables inside parallel regions.
36064         (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
36065         declared variables.
36066         (gimplify_oacc_declare): Gimplify the declare clauses.  Add the
36067         declare attribute to any decl as necessary.
36069 2017-03-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
36071         PR target/80082
36072         * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
36073         (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
36074         * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
36075         (arm_arch_lpae): This.
36076         * config/arm/arm.c (arm_arch7ve): Rename into ...
36077         (arm_arch_lpae): This.  Define it in term of isa_bit_lpae.
36078         * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
36079         arm_arch_lpae.
36081 2017-03-22  Martin Liska  <mliska@suse.cz>
36083         PR target/79906
36084         * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
36085         error message instead of an ICE.
36087 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36089         * doc/extend.texi (6.11 Additional Floating Types): Revise.
36091 2017-03-21  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36093         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
36094         comments.
36095         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
36096         comments.
36098 2017-03-21  Martin Sebor  <msebor@redhat.com>
36100         * doc/extend.texi: Use "cannot" instead of "can't."
36101         * doc/hostconfig.texi: Same.
36102         * doc/install.texi: Same.
36103         * doc/invoke.texi: Same.
36104         * doc/loop.texi: Same.
36105         * doc/md.texi: Same.
36106         * doc/objc.texi: Same.
36107         * doc/rtl.texi: Same.
36108         * doc/tm.texi: Same.
36109         * doc/tm.texi.in: Same.
36110         * doc/trouble.texi: Same.
36112 2017-03-21  Alexandre Oliva  <aoliva@redhat.com>
36114         PR debug/63238
36115         * dwarf2out.c (struct checksum_attributes): Add at_alignment.
36116         (collect_checksum_attributes): Set it.
36117         (die_checksum_ordered): Use it.
36119 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36121         PR tree-optimization/79908
36122         * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
36123         change: For a VA_ARG whose LHS has been cast away, use
36124         force_gimple_operand to construct the side effects.
36126 2017-03-21  David Malcolm  <dmalcolm@redhat.com>
36128         PR translation/80001
36129         * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
36130         more amenable to translation.
36131         (oacc_loop_auto_partitions): Likewise.
36133 2017-03-21  Marek Polacek  <polacek@redhat.com>
36134             Martin Sebor  <msebor@redhat.com>
36136         PR tree-optimization/80109
36137         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
36138         on INTEGRAL_TYPE_P.
36140 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36141             Segher Boessenkool  <segher@kernel.crashing.org>
36143         PR target/80125
36144         * combine.c (can_combine_p): Revert the 2017-03-20 change, only
36145         check reg_used_between_p between insn and one of succ or succ2
36146         depending on if succ is artificial insn not inserted into insn
36147         stream.
36149 2017-03-21  Martin Liska  <mliska@suse.cz>
36151         PR gcov-profile/80081
36152         * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
36153         * doc/gcc.texi: Include gcov-dump stuff.
36154         * doc/gcov-dump.texi: New file.
36156 2017-03-21  Toma Tabacu  <toma.tabacu@imgtec.com>
36158         PR rtl-optimization/79150
36159         * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
36160         conditional jump, if the jump is the last insn of the loop.
36162 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36163             Richard Biener  <rguenth@suse.de>
36165         PR tree-optimization/79908
36166         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36167         been cast away, use force_gimple_operand to construct the side
36168         effects.
36170 2017-03-21  Martin Liska  <mliska@suse.cz>
36172         PR libfortran/79956
36173         * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
36174         to NULL.
36176 2017-03-21  Brad Spengler <spender@grsecurity.net>
36178         PR plugins/80094
36179         * plugin.c (htab_hash_plugin): New function.
36180         (add_new_plugin): Use it and adjust.
36181         (parse_plugin_arg_opt): Adjust.
36182         (init_one_plugin): Likewise.
36184 2017-03-21  Richard Biener  <rguenther@suse.de>
36186         PR tree-optimization/80032
36187         * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
36188         if set force the cleanup to happen unconditionally.
36189         (gimplify_target_expr): Push inserted clobbers with force_uncond
36190         to avoid them being removed by control-dependent DCE.
36192 2017-03-21  Richard Biener  <rguenther@suse.de>
36194         PR tree-optimization/80122
36195         * tree-inline.c (copy_bb): Do not expans va-arg packs or
36196         va_arg_pack_len when the inlined call stmt requires pack
36197         expansion itself.
36198         * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
36200 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36202         PR sanitizer/78158
36203         * tsan.c (instrument_builtin_call): If the memory model argument
36204         is not a constant, assume it is valid.
36206         PR c/67338
36207         * fold-const.c (round_up_loc): Negate divisor in unsigned type to
36208         avoid UB.
36210 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36212         PR rtl-optimization/79910
36213         * combine.c (can_combine_p): Do not allow combining an I0 or I1
36214         if its dest is used by an insn before I2 (other than the combined
36215         insns themselves, which are properly handled already).
36217 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36219         Revert:
36220         2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36222         * combine.c (record_used_regs): New static function.
36223         (try_combine): Handle situations where there is an additional
36224         instruction between I2 and I3 which needs to have a LOG_LINK
36225         updated.
36227         Revert:
36228         2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36230         * combine.c (try_combine): Delete redundant i1 test.  Call
36231         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36233 2017-03-20  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36235         PR target/80083
36236         * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
36237         alternatives 13/14.
36239 2017-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36241         PR tree-optimization/80054
36242         * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36243         the optimization if a PHI or any of its arguments is not dominated
36244         by the candidate's basis.  Use gphi* rather than gimple* as
36245         appropriate.
36246         (replace_profitable_candidates): Clean up a gimple* variable that
36247         should be a gphi* variable.
36249 2017-03-20  Martin Sebor  <msebor@redhat.com>
36251         PR c++/52477
36252         * doc/extend.texi (attribute constructor): Document present limitation.
36254 2017-03-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36256         PR target/79963
36257         * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36258         __POWER9_VECTOR__ #ifdef control, change template definition to
36259         use Power9-specific built-in function.
36260         (vec_any_eq): Likewise.
36261         * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36262         to control outcomes from this test.
36263         (vector_ae_<mode>p): For VEC_F modes, likewise.
36265 2017-03-20  Ian Lance Taylor  <iant@google.com>
36267         * config/i386/i386.c (ix86_function_regparm): Save an extra
36268         register for -fsplit-stack with DECL_STATIC_CHAIN.
36270 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36272         PR target/79912
36273         * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36274         (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36276 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36278         * config/riscv/riscv.c (riscv_print_operand): Use "fence
36279         iorw,ow".
36280         * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36281         iorw,iorw".
36283 2017-03-20  Marek Polacek  <polacek@redhat.com>
36285         PR sanitizer/80063
36286         * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36288 2017-03-20  Richard Biener  <rguenther@suse.de>
36290         PR tree-optimization/80113
36291         * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36292         allocate extra SSA name for PHI def.
36293         (add_close_phis_to_outer_loops): Likewise.
36294         (add_close_phis_to_merge_points): Likewise.
36295         (copy_loop_close_phi_args): Likewise.
36296         (copy_cond_phi_nodes): Likewise.
36298 2017-03-20  Martin Liska  <mliska@suse.cz>
36300         PR middle-end/79753
36301         * tree-chkp.c (chkp_build_returned_bound): Do not build
36302         returned bounds for a LHS that's not a BOUNDED_P type.
36304 2017-03-20  Martin Liska  <mliska@suse.cz>
36306         PR target/79769
36307         PR target/79770
36308         * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36309         COMPLEX_CST and VECTOR_CST.
36311 2017-03-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36313         PR target/78857
36314         * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36315         target operand.  A new splitter adds the clobber statement in case
36316         the target operand is dead anyway.
36318 2017-03-19  Gerald Pfeifer  <gerald@pfeifer.com>
36320         * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36321         to age-old versions of binutils and glibc.
36323 2017-03-18  Segher Boessenkool  <segher@kernel.crashing.org>
36325         * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36327 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36329         * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36331 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36333         * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36334         requirement for binutils 2.13.
36336 2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36338         * combine.c (try_combine): Delete redundant i1 test.  Call
36339         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36341 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com
36343         * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36344         riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36345         contents.
36346         <riscv64-*-elf>: Re-arrange section
36347         <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36348         <riscv32-*-linux>: Likewise.
36349         <riscv64-*-elf>: Likewise
36350         <riscv64-*-linux>: Likewise.
36352 2017-03-17  Richard Earnshaw  <rearnsha@arm.com>
36354         PR target/80052
36355         * aarch64.opt(verbose-cost-dump): Fix typo.
36357 2017-03-17  Pat Haugen  <pthaugen@us.ibm.com>
36359         PR target/79951
36360         * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36361         for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36363 2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36365         * reload.c (find_reloads): When reloading a nonoffsettable address,
36366         use RELOAD_OTHER for it and its address reloads.
36368         PR rtl-optimization/79910
36369         * combine.c (record_used_regs): New static function.
36370         (try_combine): Handle situations where there is an additional
36371         instruction between I2 and I3 which needs to have a LOG_LINK
36372         updated.
36374 2017-03-17  Jeff Law  <law@redhat.com>
36376         PR tree-optimization/71437
36377         * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36378         conditional in the hash table first.
36379         (vrp_dom_walker::before_dom_children): Extract condition from
36380         ASSERT_EXPR.  Record condition, its inverion and any implied
36381         conditions as well.
36383 2017-03-17  Marek Polacek  <polacek@redhat.com>
36384             Markus Trippelsdorf  <markus@trippelsdorf.de>
36386         PR tree-optimization/80079
36387         * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36388         m_stores_head.
36390 2017-03-17  Richard Biener  <rguenther@suse.de>
36392         PR middle-end/80075
36393         * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36394         Properly verify the LHS before the RHS possibly claims to be
36395         handled.
36396         (stmt_could_throw_p): Hande gimple conds fully here.  Clobbers
36397         do not throw.
36399 2017-03-17  Martin Jambor  <mjambor@suse.cz>
36401         * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36402         (List of -O2 options): Likewise.
36403         (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36404         (-fipa-vrp) New.
36406 2017-03-17  Tom de Vries  <tom@codesourcery.com>
36408         * gcov-dump.c (print_usage): Print bug_report_url.
36410 2017-03-17  Richard Biener  <rguenther@suse.de>
36412         PR middle-end/80050
36413         * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36414         (parser::peek): Likewise.
36416 2017-03-17  Richard Biener  <rguenther@suse.de>
36418         PR tree-optimization/80048
36419         * sese.c (free_sese_info): Properly release rename_map and
36420         copied_bb_map elements.
36422 2017-03-16  Alexandre Oliva  <aoliva@redhat.com>
36424         * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36425         Add linked-list forward and backlinks.  Insert on
36426         construction, remove on destruction.
36427         (class pass_store_merging): Add m_stores_head field.
36428         (pass_store_merging::terminate_and_process_all_chains):
36429         Iterate over m_stores_head list.
36430         (pass_store_merging::terminate_all_aliasing_chains):
36431         Likewise.
36432         (pass_store_merging::execute): Check for debug stmts first.
36433         Push new chains onto the m_stores_head stack.
36435 2017-03-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
36437         PR target/71294
36438         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36439         SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36440         but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36442 2017-03-16  Jeff Law  <law@redhat.com>
36444         PR tree-optimization/71437
36445         * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36446         member function.  Implementation moved into after_dom_children
36447         member function and into the threader's thread_outgoing_edges
36448         function.
36449         (dom_opt_dom_walker::after_dom_children): Simplify by moving
36450         some code into new thread_outgoing_edges.
36451         * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36452         definition.  Simplify marker handling (do it here).   Assume we always
36453         have the available expression and the const/copies tables.
36454         (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36455         and tree-vrp.c
36456         * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36457         * tree-vrp.c (equiv_stack): No longer file scoped.
36458         (vrp_dom_walker): New class.
36459         (vrp_dom_walker::before_dom_children): New member function.
36460         (vrp_dom_walker::after_dom_children): Likewise.
36461         (identify_jump_threads):  Setup domwalker.  Use it rather than
36462         walking edges in a random order by hand.  Simplify setup/finalization.
36463         (finalize_jump_threads): Remove.
36464         (vrp_finalize): Do not call identify_jump_threads here.
36465         (execute_vrp): Do it here instead and call thread_through_all_blocks
36466         here too.
36468         PR tree-optimization/71437
36469         * tree-ssa-dom.c (pfn_simplify): Add basic_block argument.  All
36470         callers changed.
36471         (simplify_stmt_for_jump_threading): Add basic_block argument.  All
36472         callers changed.
36473         (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36474         (dom_opt_dom_walker::thread_across_edge): Remove
36475         handle_dominating_asserts argument.  All callers changed.
36476         (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36477         changes.  Remove calls to lhs_of_dominating_assert.  Other
36478         uses of handle_dominating_asserts turn into unconditional code
36479         (simplify_control_stmt_condition_1): Likewise.
36480         (simplify_control_stmt_condition): Likewise.
36481         (thread_through_normal_block, thread_across_edge): Likewise.
36482         * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36483         * tree-vrp.c (lhs_of_dominating_assert): Move here.  Return original
36484         object if it is not an SSA_NAME.
36485         (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36486         before calling into the VRP specific simplifiers.
36487         (identify_jump_threads): Remove handle_dominating_asserts
36488         argument.
36490 2017-03-16  Jakub Jelinek  <jakub@redhat.com>
36492         PR fortran/79886
36493         * tree-diagnostic.c (default_tree_printer): No longer static.
36494         * tree-diagnostic.h (default_tree_printer): New prototype.
36496 2017-03-16  Tamar Christina  <tamar.christina@arm.com>
36498         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36499         Change ins into fmov.
36501 2017-03-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36503         * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36504         * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36505         Use h_con constraint for operand 1.
36506         (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36507         (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36509 2017-03-15  Jeff Law  <law@redhat.com>
36511         PR tree-optimization/71437
36512         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36513         (record_temporary_equivalences): Use it.
36515         PR tree-optimization/71437
36516         * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36517         tree-ssa-scopedtables.
36518         (lookup_avail_expr, build_and_record_new_cond): Likewise.
36519         (record_conditions, record_cond, vuse_eq): Likewise.
36520         (record_edge_info): Adjust to API tweak of record_conditions.
36521         (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36522         (record_temporary_equivalences, optimize_stmt): Likewise.
36523         (eliminate_redundant_computations): Likewise.
36524         (record_equivalences_from_stmt): Likewise.
36525         * tree-ssa-scopedtables.c: Include options.h and params.h.
36526         (vuse_eq): New function, moved from tree-ssa-dom.c
36527         (build_and_record_new_cond): Likewise.
36528         (record_conditions): Likewise.  Accept vector of conditions rather
36529         than edge_equivalence structure for first argument.
36530         for the first argument.
36531         (avail_exprs_stack::lookup_avail_expr): New member function, moved
36532         from tree-ssa-dom.c.
36533         (avail_exprs_stack::record_cond): Likewise.
36534         * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36535         from tree-ssa-dom.c.
36536         (avail_exprs_stack): Add new member functions lookup_avail_expr
36537         and record_cond.
36538         (record_conditions): Declare.
36540 2017-03-15  Vladimir Makarov  <vmakarov@redhat.com>
36542         PR target/80017
36543         * lra-constraints.c (process_alt_operands): Increase reject for
36544         reloading an input/output operand.
36546 2017-03-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
36548         PR target/79038
36549         * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36550         insns to convert from signed/unsigned char/short to IEEE 128-bit
36551         floating point.
36552         (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36554 2017-03-15  Uros Bizjak  <ubizjak@gmail.com>
36556         PR target/80019
36557         * config/i386/i386.c (ix86_vector_duplicate_value): Create
36558         subreg of inner mode for values already in registers.
36560 2017-03-15  Bernd Schmidt  <bschmidt@redhat.com>
36562         * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36563         iteration reg is used after the loop.
36565 2017-03-14  Martin Sebor  <msebor@redhat.com>
36567         PR tree-optimization/79800
36568         * gimple-ssa-sprintf.c (format_floating: Add argument.  Handle
36569         precision in negative-positive range.
36570         (format_floating): Call non-const overload with adjusted precision.
36572 2017-03-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
36574         PR target/79947
36575         * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36576         -mpowerpc-gfxopt.
36578 2017-03-14  Martin Sebor  <msebor@redhat.com>
36580         PR middle-end/80020
36581         * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36582         * builtins.def (aligned_alloc): Use it.
36584         PR c/79936
36585         * Makefile.in (GTFILES): Add calls.c.
36586         * calls.c: Include "gt-calls.h".
36588 2017-03-14  Bernd Schmidt  <bschmidt@redhat.com>
36590         PR rtl-optimization/79728
36591         * regs.h (struct target_regs): New field
36592         x_contains_allocatable_regs_of_mode.
36593         (contains_allocatable_regs_of_mode): New macro.
36594         * reginfo.c (init_reg_sets_1): Initialize it, and change
36595         contains_reg_of_mode so it includes global regs as well.
36596         * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36597         rather than contains_regs_of_mode.
36599 2017-03-14  Martin Liska  <mliska@suse.cz>
36601         * doc/invoke.texi: Document options that can't be combined with
36602         -fcheck-pointer-bounds.
36604 2017-03-14  Martin Liska  <mliska@suse.cz>
36606         PR middle-end/79831
36607         * doc/invoke.texi (-Wchkp): Document the option.
36609 2017-03-14  Martin Liska  <mliska@suse.cz>
36611         * Makefile.in: Install gcov-dump.
36613 2017-03-14  Martin Liska  <mliska@suse.cz>
36615         * multiple_target.c (expand_target_clones): Bail out for
36616         an invalid attribute.
36618 2017-03-14  Richard Biener  <rguenther@suse.de>
36620         * alias.c (struct alias_set_entry): Pack properly.
36621         * cfgloop.h (struct loop): Likewise.
36622         * cse.c (struct set): Likewise.
36623         * ipa-utils.c (struct searchc_env): Likewise.
36624         * loop-invariant.c (struct invariant): Likewise.
36625         * lra-remat.c (struct cand): Likewise.
36626         * recog.c (struct change_t): Likewise.
36627         * rtl.h (struct address_info): Likewise.
36628         * symbol-summary.h (function_summary): Likewise.
36629         * tree-loop-distribution.c (struct partition): Likewise.
36630         * tree-object-size.c (struct object_size_info): Likewise.
36631         * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36632         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36633         * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36634         * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36635         * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36636         * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36637         (struct _stmt_vec_info): Likewise.
36639 2017-03-14  Martin Liska  <mliska@suse.cz>
36641         PR target/79892
36642         * multiple_target.c (create_dispatcher_calls): Check that
36643         a target can create a function dispatcher.
36645 2017-03-14  Martin Liska  <mliska@suse.cz>
36647         PR lto/66295
36648         * multiple_target.c (expand_target_clones): Drop local.local
36649         flag for default implementation.
36651 2017-03-14  Richard Biener  <rguenther@suse.de>
36653         PR tree-optimization/80030
36654         * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36656 2017-03-13  Kito Cheng  <kito.cheng@gmail.com>
36658         * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36659         gcc_fallthrough() instead of __attribute__((fallthrough));
36661 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36663         * doc/gcc.texi: Remove "up" link to (DIR).
36664         * doc/gccint.texi: Ditto.
36666 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36668         * doc/install.texi (Specific) <avr>: Remove reference to
36669         binutils 2.13.
36671 2017-03-13  Jeff Law  <law@redhat.com>
36673         * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36674         attribute rather than comments.
36676         * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36677         match_scratch operand is highest.
36679 2017-03-13  Martin Liska  <mliska@suse.cz>
36681         PR middle-end/78339
36682         * ipa-pure-const.c (warn_function_noreturn): If the declarations
36683         is a CHKP clone, use original declaration.
36685 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36687         * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36688         (arc_conditional_register_usage): Use a different allocation order
36689         when optimizing for size.
36690         * common/config/arc/arc-common.c (arc_option_optimization_table):
36691         Section anchors default on when optimizing for size.
36693 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36695         * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36697 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36699         * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36700         * config/arc/arc.md (cpu_facility): Add cd variant.
36701         (*movqi_insn): Add code density variant.
36702         (*movhi_insn): Likewise.
36703         (*movqi_insn): Likewise.
36704         (*addsi3_mixed): Likewise.
36705         (subsi3_insn): Likewise.
36707 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36709         * config/arc/arc.md (movsi_cond_exec): Update constraint.
36711 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36713         * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36714         expressions with MINUS and UNARY ops.
36716 2017-03-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36718         PR target/79911
36719         * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36720         Rename to...
36721         (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36722         between vec_select and vector argument.
36723         (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36724         (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36725         (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36726         (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36727         (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36728         (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36730 2017-03-13  Richard Biener  <rguenther@suse.de>
36732         PR other/79991
36733         * params.def (vect-max-peeling-for-alignment): Fix typo.
36735 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36737         * doc/install.texi (Specific) <mips-*-*>: Remove description of
36738         issue that only occurred with binutils below 2.18.
36740 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36742         * doc/install.texi (Specific) <cris-axis-elf>: No longer
36743         refer to binutils 2.11/2.12 minimum.
36745 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36747         * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36748         ftp.kernel.org and simplify binutils requirement.
36750 2017-03-11  Gerald Pfeifer  <gerald@pfeifer.com>
36752         * doc/invoke.texi (Warning Options): Fix spelling of link-time
36753         optimization.
36754         (Optimize Options): Ditto.  Also remove redundancy.
36756 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36758         PR translation/79848
36759         * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36760         "%qs".
36761         * ipa-pure-const.c (suggest_attribute): Likewise.  Convert _
36762         to G_ to avoid double translation.
36764 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36766         PR translation/79923
36767         * auto-profile.c (get_combined_location): Convert leading
36768         character of diagnostics to lower case and remove trailing period.
36769         (read_profile): Likewise for various diagnostics.
36770         * config/arm/arm.c (arm_option_override): Remove trailing period
36771         from various diagnostics.
36772         * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36773         (msp430_expand_delay_cycles): Likewise.
36775 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36777         PR target/79925
36778         * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36779         full command-line argument, rather than just "str".
36780         (aarch64_validate_march): Likewise.
36781         (aarch64_validate_mtune): Likewise.
36783 2017-03-10  Bernd Schmidt  <bschmidt@redhat.com>
36785         PR rtl-optimization/78911
36786         * lra-assigns.c (must_not_spill_p): New function.
36787         (spill_for): Use it.
36789 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36791         PR tree-optimization/79981
36792         * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36793         ATOMIC_COMPARE_EXCHANGE ifn result.
36794         (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36795         IFN_ATOMIC_COMPARE_EXCHANGE.
36797 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36799         PR driver/79875
36800         * opts.c (parse_sanitizer_options): Add missing question mark to
36801         "did you mean" message.
36803 2017-03-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36805         * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36806         built-in.
36807         (VMULEUH_UNS): Likewise.
36808         (VMULOUB_UNS): Likewise.
36809         (VMULOUH_UNS): Likewise.
36810         * config/rs6000/rs6000.c (builtin_function_type): Remove
36811         references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36813 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36815         PR bootstrap/79952
36816         * read-rtl-function.c (function_reader::read_rtx_operand): Update
36817         x with result of extra_parsing_for_operand_code_0.
36818         (function_reader::extra_parsing_for_operand_code_0): Convert
36819         return type from void to rtx, returning x.  When reading
36820         SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36821         larger size containing struct block_symbol.
36823 2017-03-10  Segher Boessenkool  <segher@kernel.crashing.org>
36825         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36826         -mfloat128-hardware without -m64.
36828 2017-03-10  Will Schmidt <will_schmidt@vnet.ibm.com>
36830         PR target/79941
36831         * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36832         entries to the case statement that marks unsigned arguments to
36833         overloaded functions.
36835 2017-03-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36837         * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36838         two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36840 2017-03-10  Pat Haugen  <pthaugen@us.ibm.com>
36842         PR target/79907
36843         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36844         TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36846 2017-03-10  Martin Liska  <mliska@suse.cz>
36848         PR target/65705
36849         PR target/69804
36850         * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36851         * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36852         FIELD != NULL.
36854 2017-03-10  Olivier Hainque  <hainque@adacore.com>
36856         * tree-switch-conversion (array_value_type): Start by resetting
36857         candidate type to it's main variant.
36859 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36861         PR rtl-optimization/79909
36862         * combine.c (try_combine): Use simplify_replace_rtx on individual
36863         CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36864         of the whole CALL_INSN_FUNCTION_USAGE.
36866         PR tree-optimization/79972
36867         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36868         get_range_info on SSA_NAMEs.  Formatting fixes.
36870 2017-03-10  Richard Biener  <rguenther@suse.de>
36871             Jakub Jelinek  <jakub@redhat.com>
36873         PR tree-optimization/77975
36874         * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36875         edge to be constant.
36876         (get_val_for): For constant x return it.  Formatting fix.
36877         (loop_niter_by_eval): Avoid pointless looping if the next iteration
36878         would use the same bases as the current one.
36880 2017-03-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36882         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36883         instead of vec_select for V1TImode.
36884         * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36885         longer needed.
36886         (VSX_LE_128): Add V1TI to this mode iterator.
36887         (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36888         (*vsx_le_perm_store_<mode>): Likewise.
36889         (pre-reload splitter for VSX stores): Likewise.
36890         (post-reload splitter for VSX stores): Likewise.
36891         (*vsx_xxpermdi2_le_<mode>): Likewise.
36892         (*vsx_lxvd2x2_le_<mode>): Likewise.
36893         (*vsx_stxvd2x2_le_<mode>): Likewise.
36895 2017-03-09  Michael Eager  <eager@eagercon.com>
36897         Correct failures with --enable-checking=yes,rtl.
36899         * config/microblaze/microblaze.c (microblaze_expand_shift):
36900         Replace GET_CODE test with CONST_INT_P and INTVAL test with
36901         test for const0_rtx.
36902         * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36903         lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36905 2017-03-09  Richard Biener  <rguenther@suse.de>
36907         PR tree-optimization/79977
36908         * graphite-scop-detection.c (scop_detection::merge_sese):
36909         Handle the case of extra exits to blocks dominating the entry.
36911 2017-03-09  Toma Tabacu  <toma.tabacu@imgtec.com>
36913         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
36914         Document rdynamic.
36916 2017-03-09  Vladimir Makarov  <vmakarov@redhat.com>
36918         PR rtl-optimization/79949
36919         * lra-constraints.c (process_alt_operands): Check memory when
36920         trying to predict a cycle.  Print about the overall increase.
36922 2017-03-09  Richard Biener  <rguenther@suse.de>
36924         PR middle-end/79971
36925         * gimple-expr.c (useless_type_conversion_p): Preserve
36926         TYPE_SATURATING for fixed-point types.
36928 2017-03-09  Richard Biener  <rguenther@suse.de>
36930         PR ipa/79970
36931         * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
36932         alignment of BLKmode params.
36934 2017-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36936         PR target/79913
36937         * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
36938         (VALL_NO_V2Q): Likewise.
36939         (VDQF_DF): Delete.
36940         * config/aarch64/aarch64-simd.md
36941         (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
36942         iterator.
36943         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
36944         VALL_NO_V2Q mode iterator.
36945         (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
36947 2017-03-09  Martin Liska  <mliska@suse.cz>
36949         PR tree-optimization/79631
36950         * tree-chkp-opt.c (chkp_is_constant_addr): Call
36951         tree_int_cst_sign_bit just for INTEGER constants.
36953 2017-03-09  Martin Liska  <mliska@suse.cz>
36955         PR target/65705
36956         PR target/69804
36957         * toplev.c (process_options): Disable -fcheck-pointer-bounds with
36958         sanitizers.
36960 2017-03-09  Marek Polacek  <polacek@redhat.com>
36962         PR c++/79672
36963         * tree.c (inchash::add_expr): Handle TREE_VEC.
36965 2017-03-09  Martin Liska  <mliska@suse.cz>
36967         PR ipa/79764
36968         (chkp_narrow_size_and_offset): New function.
36969         (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
36970         (void chkp_parse_bit_field_ref): New function.
36971         (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
36972         (chkp_process_stmt): Use chkp_parse_bit_field_ref.
36974 2017-03-09  Martin Liska  <mliska@suse.cz>
36976         PR ipa/79761
36977         * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
36978         (chkp_find_bounds_1): Remove gcc_unreachable.
36980 2017-03-09  Jakub Jelinek  <jakub@redhat.com>
36982         PR sanitizer/79944
36983         * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
36984         BUILT_IN_SYNC*, determine the access type from the size suffix and
36985         always build a MEM_REF with that type.  Handle forgotten
36986         BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
36988         PR target/79932
36989         * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
36990         _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
36991         _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
36992         _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
36993         _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
36994         _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
36995         _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
36996         _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
36997         _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
36998         _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
36999         _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
37000         _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
37001         _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
37002         _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
37003         _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
37004         _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
37005         _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
37006         _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
37007         _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
37008         _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
37009         _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
37010         _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
37011         _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
37012         _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
37013         _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
37014         _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
37015         _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
37016         _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
37017         _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
37018         _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
37019         definitions outside of __OPTIMIZE__ guarded section.
37021         PR target/79932
37022         * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
37023         _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
37024         _mm512_packus_epi32, _mm512_maskz_packus_epi32,
37025         _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
37026         guarded section.
37028 2017-03-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
37030         * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
37031         ("vfenez<mode>"): Add missing constraints.
37033 2017-03-08  Martin Sebor  <msebor@redhat.com>
37035         PR target/79928
37036         * config/nds32/nds32.c (nds32_option_override):
37037         Fix misspelled diagnostic.
37039 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37041         PR c/79940
37042         * gimplify.c (gimplify_omp_for): Replace index var in outer
37043         taskloop statement with an artificial variable and add
37044         OMP_CLAUSE_PRIVATE clause for it.
37046 2017-03-08  Richard Biener  <rguenther@suse.de>
37048         PR tree-optimization/79955
37049         * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
37050         for accesses that are completely outside of the variable.
37052 2017-03-08  Andrew Haley  <aph@redhat.com>
37054         PR tree-optimization/79943
37055         * tree-ssa-loop-split.c (compute_new_first_bound): When
37056         calculating the new upper bound, (END-BEG) should be added, not
37057         subtracted.
37059 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37061         * config/avr/avr.md (setmemhi): Make sure match_dup
37062         operand number comes before match_scratch.
37064 2017-03-08  Richard Biener  <rguenther@suse.de>
37066         PR tree-optimization/79920
37067         * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
37068         with ncopies == 1 to ...
37069         (vect_transform_slp_perm_load): ... here.  Properly compute
37070         all element loads by iterating VF times over the group.  Do
37071         not handle ncopies (computed in a broken way) in
37072         vect_create_mask_and_perm.
37074 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37076         PR sanitizer/79904
37077         * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
37078         is a uniform vector, use uniform_vector_p return value instead of
37079         building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
37081 2017-03-07  Marek Polacek  <polacek@redhat.com>
37083         PR middle-end/79809
37084         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
37085         (alloca_call_type): Likewise.
37087 2017-03-07  Martin Liska  <mliska@suse.cz>
37089         * gcov.c (process_args): Put comment to correct location.
37091 2017-03-07  Martin Liska  <mliska@suse.cz>
37093         PR middle-end/68270
37094         * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
37095         Use array_at_struct_end_p instead of DECL_CHAIN (field).
37096         (chkp_narrow_bounds_for_field): Likewise.
37097         (chkp_parse_array_and_component_ref): Pass one more argument to
37098         call.
37100 2017-03-07  Richard Biener  <rguenther@suse.de>
37102         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
37103         preheaders.
37105 2017-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
37107         * config/i386/i386.c (ix86_local_alignment): Align most aggregates
37108         of 16 bytes and more to 16 bytes, not those of 16 bits and more.
37110 2017-03-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37112         PR c/79855
37113         * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
37114         to end of description.
37115         (PARAM_MAX_STORES_TO_MERGE): Likewise.
37117 2017-03-07  Jakub Jelinek  <jakub@redhat.com>
37119         PR rtl-optimization/79901
37120         * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
37121         ...
37122         (*avx512f_<code><mode>3<mask_name>): ... this.
37123         (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
37124         iterator instead of VI8_AVX2_AVX512BW.
37126         PR rtl-optimization/79901
37127         * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
37128         min/max expander, expand it using expand_vec_cond_expr.
37130         PR sanitizer/79897
37131         * ubsan.c (ubsan_encode_value): Call mark_addressable on the
37132         temporary.
37134 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37136         PR c++/79821
37137         * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
37138         to void * for PCH reasons.
37139         * dwarf2out.c (output_loc_operands, output_die): Cast
37140         v.val_vec.array to unsigned char *.
37142 2017-03-06  John David Anglin  <danglin@gcc.gnu.org>
37144         PR target/77850
37145         * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
37146         vector types.
37148 2017-03-06  Vladimir Makarov  <vmakarov@redhat.com>
37150         PR rtl-optimization/79571
37151         * lra-constraints.c (process_alt_operands): Calculate static
37152         reject and subtract it from overall when only addresses will be
37153         reloaded.
37155 2017-03-06  Julia Koval  <julia.koval@intel.com>
37157         PR target/79793
37158         * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
37159         incoming stack boundary to 128 for 64-bit targets.
37161 2017-03-06  Richard Biener  <rguenther@suse.de>
37163         PR tree-optimization/79894
37164         * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
37165         to NULL after folding it.
37167 2017-03-06  Richard Biener  <rguenther@suse.de>
37169         PR tree-optimization/79824
37170         * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
37171         check disabling peeling for gaps.
37173 2017-03-06  Toma Tabacu  <toma.tabacu@imgtec.com>
37175         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
37176         attributes): Document gettimeofday.
37178 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37180         * config/s390/s390.c (s390_option_override_internal): Set
37181         PARAM_MIN_VECT_LOOP_BOUND
37183 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37185         * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
37186         * config/s390/s390.md: Likewise.
37188 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37190         PR target/79812
37191         * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
37192         (<avx2_avx512>_perm<mode>): Rename to ...
37193         (avx2_perm<mode>): ... this.  Use VI8F_256 iterator instead
37194         of VI8F_256_512.
37195         (<avx512>_perm<mode>_mask): Rename to ...
37196         (avx512vl_perm<mode>_mask): ... this.  Use VI8F_256 iterator instead
37197         of VI8F_256_512.
37198         (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
37199         (avx2_perm<mode>_1<mask_name): ... this.  Use VI8F_256 iterator
37200         instead of VI8F_256_512.
37201         (avx512f_perm<mode>): New define_expand.
37202         (avx512f_perm<mode>_mask): Likewise.
37203         (avx512f_perm<mode>_1<mask_name>): New define_insn.
37204         (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
37206 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37208         * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
37209         msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
37210         if_then_else.
37211         (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
37213 2017-03-06  Martin Liska  <mliska@suse.cz>
37215         PR sanitize/79783
37216         * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
37217         when having a SSA NAME w/o VAR_DECL assigned to it.
37219 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37221         * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
37222         msa_dpsub_<su>_d): Fix MODE for vec_select.
37224 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37226         * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
37227         argument.
37228         * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
37230 2017-03-06  Richard Biener  <rguenther@suse.de>
37232         * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
37233         * plugin.c (register_plugin_info): Likewise.
37234         * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
37236 2017-03-05  Jakub Jelinek  <jakub@redhat.com>
37238         * config/i386/sse.md (sse_storehps, sse_storelps,
37239         avx_<castmode><avxsizesuffix>_<castmode>,
37240         avx512f_<castmode><avxsizesuffix>_<castmode>,
37241         avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37242         in condition that at least one operand is not a MEM.
37244 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37246         PR middle-end/79805
37247         * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37248         ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37249         ECF_NOTHROW.
37250         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37251         gimple_call_nothrow_p flag based on whether original builtin can throw.
37252         If it can, emit following stmts on the fallthrough edge.
37253         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37254         don't create new bb if inserting just debug stmts on the edge, try to
37255         insert them on the fallthru bb or just reset debug stmts.
37257 2017-03-03  Segher Boesssenkool  <segher@kernel.crashing.org>
37259         PR target/43763
37260         * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37261         restore recog_data (including the operand rtxes inside it) around
37262         the call to get_insn_template.
37264 2017-03-03  Martin Sebor  <msebor@redhat.com>
37266         PR tree-optimization/79699
37267         * context.c (context::~context): Free MPFR caches to avoid
37268         a memory leak on program exit.
37270 2017-03-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37272         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37273         Use wide_int::ulow () instead of .elt (0).
37275 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37277         * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37278         (*pushxf): Limit oF constraint to 32bit targets and add oC
37279         constraint for 64bit targets.
37280         (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37281         (*pushdf): Change rmF constraint to rmC.
37283 2017-03-03  Martin Liska  <mliska@suse.cz>
37285         * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37286         Remove unused variable.
37288 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37290         PR target/79807
37291         * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37292         is a memory operand, increase num_memory.
37293         (ix86_expand_args_builtin): Likewise.
37295 2017-03-03  Jan Hubicka  <jh@suse.cz>
37297         PR lto/79760
37298         * ipa-devirt.c (maybe_record_node): Properly handle
37299         __cxa_pure_virtual visibility.
37301 2017-03-03  Martin Liska  <mliska@suse.cz>
37303         PR tree-optimization/79803
37304         * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37305         assert.
37306         (pass_loop_prefetch::execute): Disabled optimization if an
37307         assumption about L1 cache size is not met.
37309 2017-03-03  Martin Liska  <mliska@suse.cz>
37311         PR rtl-optimization/79574
37312         * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37313         (hash_scan_set): Likewise.
37314         (dump_hash_table): Likewise.
37315         (hoist_code): Likewise.
37317 2017-03-03  Richard Biener  <rguenther@suse.de>
37319         * fixed-value.c (fixed_from_string): Restore use of elt (1)
37320         in place of uhigh ().
37321         (fixed_convert_from_real): Likewise.
37323 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37325         PR target/79514
37326         * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37328 2017-03-03  Richard Biener  <rguenther@suse.de>
37330         PR middle-end/79818
37331         * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37332         TYPE_OVERFLOW_UNDEFINED check.
37334 2017-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37336         * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37337         numbers.
37338         (vector_ae_<mode>_p): Likewise.
37339         (vector_nez_<mode>_p): Likewise.
37340         (vector_ne_v2di_p): Likewise.
37341         (vector_ae_v2di_p): Likewise.
37342         (vector_ne_<mode>_p): Likewise.
37343         * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37344         numbers.
37345         (vsx_tsqrt<mode>2_fe): Likewise.
37347 2017-03-02  Uros Bizjak  <ubizjak@gmail.com>
37349         PR target/79514
37350         * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37352 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37354         PR rtl-optimization/79780
37355         * cprop.c (one_cprop_pass): When second and further conditional trap
37356         in a single basic block is turned into an unconditional trap, turn it
37357         into a deleted note to avoid RTL verification failures.
37359 2017-03-02  Richard Biener  <rguenther@suse.de>
37361         * fold-const.c (const_binop): Use ulow () instead of elt (0).
37363 2017-03-02  Richard Biener  <rguenther@suse.de>
37365         PR tree-optimization/79345
37366         PR c++/42000
37367         * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37368         param and abort the walk, returning -1 if it is hit.
37369         (walk_aliased_vdefs): Take a limit param and pass it on.
37370         * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37371         defaulting to 0 and return a signed int.
37372         * tree-ssa-uninit.c (struct check_defs_data): New struct.
37373         (check_defs): New helper.
37374         (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37375         about uninitialized memory.
37376         * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37377         bogus uninitialized warning.
37378         (fixed_convert_from_real): Likewise.
37380 2017-03-02  Bin Cheng  <bin.cheng@arm.com>
37382         PR tree-optimization/66768
37383         * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37384         iv_use if base object can't be determined.
37386 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37388         PR tree-optimization/79345
37389         * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37390         * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37391         (get_pattern_stats): Initialize it.
37392         * genemit.c (gen_expand): Verify match_scratch numbers come after
37393         match_operand/match_dup numbers.
37394         * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37395         match_scratch numbers.
37396         * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37397         Likewise.
37398         * config/s390/s390.md (trunctdsd2): Likewise.
37400 2017-03-02  Richard Biener  <rguenther@suse.de>
37402         * wide-int.h (wide_int_storage::operator=): Implement in terms
37403         of wi::copy.
37405 2017-03-02  Richard Biener  <rguenther@suse.de>
37407         PR tree-optimization/79777
37408         * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37409         the to insert expression to sth existing.
37411 2017-03-01  Martin Sebor  <msebor@redhat.com>
37413         PR middle-end/79692
37414         * gimple-ssa-sprintf.c
37415         (directive::known_width_and_precision): New function.
37416         (format_integer): Use it.
37417         (get_mpfr_format_length): Consider the full range of precision
37418         when computing %g output with the # flag.  Set the likely byte
37419         count to 3 rather than 1 when precision is indeterminate.
37420         (format_floating): Correct the lower bound of precision.
37422 2017-03-01  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37424         * doc/invoke.texi: Document default code model for 64-bit Linux.
37426 2017-03-01  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
37428         PR target/79752
37429         * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37430         udiv rather than div since input pattern is unsigned.
37432 2017-03-01  Uros Bizjak  <ubizjak@gmail.com>
37434         * config/i386/i386.c (print_reg): Warn for values of
37435         unsupported size in integer register.
37437 2017-03-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
37439         PR target/79439
37440         * config/rs6000/predicates.md (current_file_function_operand): Do
37441         not allow self calls to be local if the function is replaceable.
37443 2017-03-01  Kelvin Nilsen  <kelvin@gcc.gnu.org>
37445         PR target/79395
37446         * config/rs6000/altivec.h (vec_ctz and others): Change the
37447         preprocessor macro that controls conditional compilation from
37448         _ARCH_PWR9 to __POWER9_VECTOR__.
37449         (vec_all_ne): Change parameterization of __altivec_scalar_pred
37450         macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37451         control (instead of _ARCH_PWR9 control) so that template
37452         definition uses power9-specific function.
37453         (vec_any_eq): Likewise.
37454         (vec_all_ne): Change macro definition to use a power9-specific
37455         expansion under #ifdef __POWER9_VECTOR__ control (instead of
37456         _ARCH_PWR9 control).
37457         (vec_any_eq) Likewise.
37458         * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37459         expansion for CMPNEF to remove support for xvcmpnesp instruction.
37460         (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37461         support for xvcmpnedp instruction.
37462         (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37463         macro expansion so that Power9 implementation of vec_all_ne does
37464         not use the AltiVec predicate framework.
37465         (VCMPNEH_P): Likewise.
37466         (VCMPNEW_P): Likewise.
37467         (VCMPNED_P): Likewise.
37468         (VCMPNEFP_P): Likewise.
37469         (VCMPNEDP_P): Likewise.
37470         (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37471         implementation of vec_any_eq to not use AltiVec predicate
37472         framework.
37473         (VCMPAEH_P): Likewise.
37474         (VCMPAEW_P): Likewise.
37475         (VCMPAED_P): Likewise.
37476         (VCMPAEFP_P): Likewise.
37477         (VCMPAEDP_P): Likewise.
37478         (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37479         BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37480         not use the AltiVec predicate framework.
37481         (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37482         of vec_any_eq to not use AltiVec predicate framework.
37483         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37484         support for predefined __POWER9_VECTOR__ macro to indicate that
37485         Power9 instruction selection is enabled.
37486         (altivec_overloaded_builtins): Remove extraneous
37487         ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37488         function argument types RS6000_BTI_bool_V16QI and
37489         RS6000_BTI_bool_V16QI.  Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37490         entry for overloaded function argument types RS6000_BTI_bool_V4SI
37491         andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB.  Remove
37492         two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37493         P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37494         Power9 for implementations of vec_cmpne.  Change the signature for
37495         all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37496         (representing vec_all_ne) to remove the previously described first
37497         argument of type RS6000_BTI_INTSI, as this was an artifact of
37498         reliance on the AltiVec predicate framework, which is no longer
37499         used in the implementation of these functions.  Add
37500         P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37501         function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37502         since, unlike the AltiVec predicate framework implementation, we
37503         do not share function descriptors between vec_alle and vec_anyeq.
37504         (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37505         set of modes that receive special treatment even when
37506         TARGET_P9_VECTOR is true.  The special treatment emits code that
37507         does not depend on Power9 instructions.
37508         * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37509         define_expand to not rely on AltiVec predicate framework.
37510         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37511         function.
37512         (vector_ne_v2di_p): Change this define_expand to not rely on
37513         AltiVec predicate framework.
37514         (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37515         function.
37516         (vector_ne_<mode>_p): Change this define_expand to not rely on
37517         AltiVec predicate framework.
37518         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37519         function.
37520         * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37521         (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37522         define_insn pattern.
37523         (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37524         define_insn pattern because the xvcmpne<VSs>. instruction is not
37525         supported.
37526         (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37527         instruction is not supported.
37529 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37531         * config/nvptx/nvptx.c: Include intl.h.
37533 2017-03-01  Martin Jambor  <mjambor@suse.cz>
37535         PR lto/78140
37536         * ipa-prop.h (ipa_bits): Removed field known.
37537         (ipa_jump_func): Removed field vr_known.  Changed fields bits and m_vr
37538         to pointers.  Adjusted their comments to warn about their sharing.
37539         (ipcp_transformation_summary): Change bits to a vector of pointers.
37540         (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37541         (ipa_get_ipa_bits_for_value): Declare.
37542         * tree-vrp.h (value_range): Mark as GTY((for_user)).
37543         * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37544         (ipa_bits_hash_table): Likewise.
37545         (ipa_vr_ggc_hash_traits): Likewise.
37546         (ipa_vr_hash_table): Likewise.
37547         (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37548         being pointers and vr_known being removed.
37549         (ipa_set_jf_unknown): Likewise.
37550         (ipa_get_ipa_bits_for_value): New function.
37551         (ipa_set_jfunc_bits): Likewise.
37552         (ipa_get_value_range): New overloaded functions.
37553         (ipa_set_jfunc_vr): Likewise.
37554         (ipa_compute_jump_functions_for_edge): Use the above functions to
37555         construct bits and vr parts of jump functions.
37556         (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37557         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37558         exist.
37559         (ipcp_grow_transformations_if_necessary): Also allocate
37560         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37561         exist.
37562         (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37563         them.  Fix too long lines.
37564         (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37565         vr_known being removed.
37566         (ipa_read_jump_function): Use new setter functions to construct bits
37567         and vr parts of jump functions or set them to NULL.
37568         (write_ipcp_transformation_info): Adjust for bits being pointers.
37569         (read_ipcp_transformation_info): Likewise.
37570         (ipcp_update_bits): Likewise.  Fix excessively long lines a trailing
37571         space.
37572         Include gt-ipa-prop.h.
37573         * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37574         being pointers.
37575         (ipcp_store_bits_results): Likewise.
37576         (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37577         Do not write to existing jump functions but use a temporary instead.
37579 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37581         PR c++/79681
37582         * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37583         attempt to use its first operand as BIT_FIELD_REF base.
37585 2017-03-01  Richard Biener  <rguenther@suse.de>
37587         PR middle-end/79721
37588         * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37589         interpolating formula in wrapping arithmetic.
37590         (chrec_apply): Convert chrec_evaluate return value to wanted type.
37592 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37594         PR tree-optimization/79734
37595         * tree-vect-generic.c (expand_vector_condition): Optimize
37596         AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37597         Handle VEC_COND_EXPR where comparison has different inner width from
37598         type's inner width.
37600 2017-02-28  Sandra Loosemore  <sandra@codesourcery.com>
37602         * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37603         markup, and similar issues.  Remove @opindex entries for things
37604         that aren't options.  Add missing -mmpy-option entries.
37606 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37608         PR tree-optimization/79737
37609         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37610         a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37611         tmpbuf[byte_size - 1].  Call natice_encode_expr with byte_size - 1
37612         instead of byte_size.  Formatting fix.
37613         (shift_bytes_in_array_right): Formatting fix.
37615 2017-02-28  Eric Botcazou  <ebotcazou@adacore.com>
37617         PR target/79749
37618         * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37619         condition on optimize for the leaf function test.
37621 2017-02-28  Martin Liska  <mliska@suse.cz>
37623         PR lto/79625
37624         * read-rtl-function.c (function_reader::handle_unknown_directive):
37625         Bail out when one uses -flto.
37627 2017-02-28  Martin Liska  <mliska@suse.cz>
37629         * common.opt: Replace space with tabular for options of <number>
37630         type.
37631         * config/i386/i386.opt: Show <number> value for
37632         -mlarge-data-threshold.
37633         * opts.c (print_filtered_help): Do not display number in hexadecimal
37634         format.
37636 2017-02-28  Martin Liska  <mliska@suse.cz>
37638         * common.opt: Fix --help=option -Q for options which are of
37639         an enum type.
37641 2017-02-28  Uros Bizjak  <ubizjak@gmail.com>
37643         * config/i386/i386.c (print_reg): Error out for values
37644         of 8-bit size in invalid integer register.
37646 2017-02-28  Martin Sebor  <msebor@redhat.com>
37648         PR tree-optimization/79691
37649         * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37651 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37653         PR target/79729
37654         * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37655         gcc_unreachable with output_operand_lossage.
37657 2017-02-28  Richard Biener  <rguenther@suse.de>
37659         PR tree-optimization/79740
37660         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37661         inserts.
37662         (visit_nary_op): Insert the nary into the hashtable if we
37663         pattern-matched sth.
37664         * tree-ssa-pre.c (eliminate_insert): Robustify.
37666 2017-02-28  Richard Biener  <rguenther@suse.de>
37668         PR middle-end/79731
37669         * fold-const.c (decode_field_reference): Reject out-of-bound
37670         accesses.
37672 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37674         * config/i386/i386.c: Include intl.h.
37675         (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37676         instead of just cond ? "..." : "...".
37677         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37678         * coverage.c (read_counts_file): Likewise.
37679         * omp-offload.c: Include intl.h.
37680         (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37681         of just cond ? "..." : "...".
37682         * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37683         of just cond ? "..." : "...".
37685 2017-02-28  Richard Earnshaw  <rearnsha@arm.com>
37687         PR target/79742
37688         * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37689         entry, if present.
37690         * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37691         'tune for' CPU name.
37692         * config/arm/arm-cpu-data.h: Regenerated.
37694 2017-02-28  Richard Biener  <rguenther@suse.de>
37696         PR tree-optimization/79732
37697         * tree-inline.c (expand_call_inline): Do not shadow var.
37699 2017-02-28  Richard Biener  <rguenther@suse.de>
37701         PR tree-optimization/79723
37702         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37703         address-space properly.
37705 2017-02-28  Thomas Schwinge  <thomas@codesourcery.com>
37707         * doc/optinfo.texi (Optimization groups): Fix option used for
37708         OPTGROUP_ALL.
37709         * doc/invoke.texi (-fopt-info): Document "omp".
37710         * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37711         (OPTGROUP_ALL): Add OPTGROUP_OMP.
37712         * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37713         * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37714         * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37716         * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP.  Adjust
37717         all users.
37718         * dumpfile.c (optgroup_options): Instead of "openmp", associate
37719         OPTGROUP_OMP with "omp".
37721 2017-02-27  Pat Haugen  <pthaugen@us.ibm.com>
37723         PR target/79544
37724         * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37725         for arithmetic shift of unsigned V2DI.
37727 2017-02-27  Claudiu Zissulescu  <claziss@synopsys.com>
37729         * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37730         arc/linux.h headers.
37731         * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37732         (LINK_SPEC): Likewise.
37733         (ARC_TLS_EXTRA_START_SPEC): Likewise.
37734         (EXTRA_SPECS): Likewise.
37735         (STARTFILE_SPEC): Likewise.
37736         (ENDFILE_SPEC): Likewise.
37737         (LIB_SPEC): Likewise.
37738         (TARGET_SDATA_DEFAULT): Likewise.
37739         (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37740         (MULTILIB_DEFAULTS): Likewise.
37741         (DWARF2_UNWIND_INFO): Likewise.
37742         * config/arc/big.h: New file.
37743         * config/arc/elf.h: Likewise.
37744         * config/arc/linux.h: Likewise.
37745         * config/arc/t-uClibc: Remove.
37747 2017-02-27  Bin Cheng  <bin.cheng@arm.com>
37749         PR tree-optimization/77536
37750         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37751         (tree_transform_and_unroll_loop): Use above function to compute the
37752         estimated niter of unrolled loop and use it when scaling profile.
37753         Also use count info rather than frequency if it's non-zero.
37754         * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37755         * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37756         (vect_transform_loop): Call above function.
37758 2017-02-27  Richard Biener  <rguenther@suse.de>
37760         PR tree-optimization/45397
37761         * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37762         * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37763         (visit_nary_op): Add pattern matching for CSEing sign-changed
37764         or truncated operations with wider ones.
37766 2017-02-27  Richard Biener  <rguenther@suse.de>
37768         PR tree-optimization/79690
37769         * tree-vect-stmts.c (vectorizable_store): Use vector type
37770         built from the DR with address-space.
37772 2017-02-26  Gerald Pfeifer  <gerald@pfeifer.com>
37774         * doc/invoke.texi (Optimize Options): Refine the description
37775         of asan-use-after-return.
37777 2017-02-25  Alan Modra  <amodra@gmail.com>
37779         PR rtl-optimization/79584
37780         * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37781         base, not ad->base_term, the reg within base.  Remove assertion
37782         that ad->base == ad->base_term.  Replace gen_int_mode using
37783         bogus mode with const0_rtx.
37785 2017-02-25  Jakub Jelinek  <jakub@redhat.com>
37787         PR middle-end/79396
37788         * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37789         FMA_EXPR like tcc_binary or tcc_unary.
37791         * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37793         PR debug/77589
37794         * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37795         bitfield.
37796         (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37797         (output_loc_operands): Handle DW_OP_call_ref and
37798         DW_OP_GNU_variable_value.
37799         (struct variable_value_struct): New type.
37800         (struct variable_value_hasher): Likewise.
37801         (variable_value_hash): New variable.
37802         (string_types): Remove.
37803         (copy_loc_descr): New function.
37804         (add_loc_descr_to_each): Clarify comment.  Use copy_loc_descr.
37805         (prepend_loc_descr_to_each): New function.
37806         (add_loc_list): Fix comment typo.  Use prepend_loc_descr_to_each
37807         instead of add_loc_descr_to_each if the first argument is single
37808         location list and the second has multiple.
37809         (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37810         (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37811         when looking for variable value which doesn't have other location info.
37812         (loc_list_from_tree): Formatting fix.
37813         (gen_array_type_die): Simplify DW_AT_string_length handling.
37814         (adjust_string_types): Remove.
37815         (gen_subprogram_die): Don't call adjust_string_types nor test/set
37816         string_types.  Call resolve_variable_values.
37817         (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37818         (resolve_addr_in_expr): Likewise.  Add A argument.
37819         (copy_deref_exprloc): Remove deref argument.  Adjust for the
37820         original expression being DW_OP_GNU_variable_value with optionally
37821         DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37822         optionally after it.
37823         (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37824         (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37825         callers.  Set remove_AT_byte_size if removing DW_AT_string_length.
37826         (variable_value_hasher::hash, variable_value_hasher::equal): New
37827         methods.
37828         (resolve_variable_value_in_expr, resolve_variable_value,
37829         resolve_variable_values, note_variable_value_in_expr,
37830         note_variable_value): New functions.
37831         (dwarf2out_early_finish): Call note_variable_value on all toplevel
37832         DIEs.
37834 2017-02-24  Jakub Jelinek  <jakub@redhat.com>
37836         PR c/79677
37837         * opts.h (handle_generated_option): Add GENERATED_P argument.
37838         * opts-common.c (handle_option): Adjust function comment.
37839         (handle_generated_option): Add GENERATED_P argument, pass it to
37840         handle_option.
37841         (control_warning_option): Pass false to handle_generated_option
37842         GENERATED_P.
37843         * opts.c (maybe_default_option): Pass true to handle_generated_option
37844         GENERATED_P.
37845         * optc-gen.awk: Likewise.
37847 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37849         * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37850         a REG, look at the REG it is a SUBREG of.
37851         (splitter for cmpeqsi_t): Ditto.
37853 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37855         * config/pa/pa.c (pa_combine_instructions): Do not share RTL.  Make
37856         the special USEs with the pattern of the insn, not the insn itself.
37858 2017-02-24  Matthew Fortune  <matthew.fortune@imgtec.com>
37860         PR target/79473
37861         * doc/invoke.texi: Document -mload-store-pairs.
37863 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37864             Sandra Loosemore  <sandra@codesourcery.com>
37866         * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37867         argument isn't a CONST_INT.
37868         (nios2_alternate_compare_const): Assert op is a CONST_INT.
37869         (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37870         (nios2_validate_compare): Bypass alternate compare logic if *op2
37871         is not a CONST_INT.
37872         (ldstwm_operation_p): Return false if first_base is not a REG or
37873         if first_offset is not a CONST_INT.
37875 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37877         * config/cris/cris.md: Use correct operand in a define_peephole2.
37879 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37881         * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37883 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37885         * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37886         this_insn if it is an INSN or JUMP_INSN.
37887         (force_offsettable): Look at base, not at addr.
37888         * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37889         on things that aren't necessarily CONST_INTs.
37891 2017-02-24  Uros Bizjak  <ubizjak@gmail.com>
37893         * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37894         -mfpmath=sse is the default also for x86-32 targets with SSE2
37895         instruction set when @option{-ffast-math} is enabled
37897 2017-02-24  Jeff Law  <law@redhat.com>
37899         PR rtl-optimizatoin/79286
37900         * ira.c (update_equiv_regs): Drop may_trap_p exception to
37901         dominance test.
37903 2017-02-24  Richard Biener  <rguenther@suse.de>
37905         PR tree-optimization/79389
37906         * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37907         debug insns.
37909 2017-02-24  Aldy Hernandez  <aldyh@redhat.com>
37911         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
37912         function comment to reflect reality.
37913         (loop_exits_before_overflow): Fix typo in function description.
37915 2017-02-24  Richard Biener  <rguenther@suse.de>
37917         PR tree-optimization/79389
37918         * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
37919         properly that a threading opportunity exists.  Detect conditional
37920         copy/constant propagation opportunities.
37922 2017-02-23  Eric Botcazou  <ebotcazou@adacore.com>
37924         * config/visium/visium.md (type): Add trap.
37925         (b): New mode attribute.
37926         (*btst): Rename into...
37927         (*btst<mode>): ...this and adjust.
37928         (*cbranchsi4_btst_insn): Rename into...
37929         (*cbranch<mode>4_btst_insn): ...this and adjust.
37930         (trap): New define_insn.
37932 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37934         PR tree-optimization/79389
37935         * ifcvt.c (struct noce_if_info): Add rev_cond field.
37936         (noce_reversed_cond_code): New function.
37937         (noce_emit_store_flag): Use rev_cond if non-NULL instead of
37938         reversed_comparison_code.  Formatting fix.
37939         (noce_try_store_flag): Test rev_cond != NULL in addition to
37940         reversed_comparison_code.
37941         (noce_try_store_flag_constants): Likewise.
37942         (noce_try_store_flag_mask): Likewise.
37943         (noce_try_addcc): Use rev_cond if non-NULL instead of
37944         reversed_comparison_code.
37945         (noce_try_cmove_arith): Likewise.  Formatting fixes.
37946         (noce_try_minmax, noce_try_abs): Clear rev_cond.
37947         (noce_find_if_block): Initialize rev_cond.
37948         (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
37949         instead of false as last argument never attempt to reverse it
37950         afterwards.
37952 2017-02-23  Bin Cheng  <bin.cheng@arm.com>
37954         PR tree-optimization/79663
37955         * tree-predcom.c (combine_chains): Process refs in reverse order
37956         only for ZERO length chains, and add explaining comment.
37958 2017-02-23  Jeff Law  <law@redhat.com>
37960         PR tree-optimization/79578
37961         * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
37962         in call to operand_equal_p.
37964 2017-01-23  Dominique d'Humieres  <dominiq@lps.ens.fr>
37966         PR target/71017
37967         * config/i386/cpuid.h: Fix another undefined behavior.
37969 2017-02-23  Richard Biener  <rguenther@suse.de>
37971         PR tree-optimization/79683
37972         * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
37973         vector types for data-refs.
37975 2017-02-23  Martin Liska  <mliska@suse.cz>
37977         * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
37979 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37981         PR middle-end/79665
37982         * internal-fn.c (get_range_pos_neg): Moved to ...
37983         * tree.c (get_range_pos_neg): ... here.  No longer static.
37984         * tree.h (get_range_pos_neg): New prototype.
37985         * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
37986         are known to be in between 0 and signed maximum inclusive, try to
37987         expand both unsigned and signed divmod and use the cheaper one from
37988         those.
37990 2017-02-22  Jeff Law  <law@redhat.com>
37992         PR tree-optimization/79578
37993         * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
37994         to compare base operands.
37996 2017-02-22  Segher Boessenkool  <segher@kernel.crashing.org>
37998         PR target/79211
37999         * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
38000         gpc_reg_operand instead of fpr_reg_operand.
38002 2017-02-22  Sameera Deshpande  <sameera.deshpande@imgtec.com>
38004         * config/mips/mips.c (mips_return_in_memory): Force FP
38005         vector types to be returned in memory for o32 ABI.
38007 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
38009         * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
38010         instead of DW_TAG_member for static data member declarations and don't
38011         set no_linkage_name for static inline data members.
38012         (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
38013         to DW_TAG_member.
38015 2017-02-22  Martin Liska  <mliska@suse.cz>
38017         * doc/invoke.texi: Replace inequality signs with square brackets
38018         for -Wnormalized.
38020 2017-02-22  Matthew Fortune  <matthew.fortune@imgtec.com>
38022         PR target/78660
38023         * lra-constraints.c (simplify_operand_subreg): Handle
38024         WORD_REGISTER_OPERATIONS targets.
38026 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
38028         PR target/70465
38029         * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
38030         and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
38031         elimination by swapping fld*.
38033 2017-02-22  Richard Biener  <rguenther@suse.de>
38035         PR tree-optimization/79673
38036         * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
38037         convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
38038         irrelevant address-space qualifiers and avoiding a
38039         ADDR_SPACE_CONVERT_EXPR from fold_convert.
38041 2017-02-22  Richard Biener  <rguenther@suse.de>
38043         PR tree-optimization/79666
38044         * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
38045         to not symbolically negate if that may introduce undefined
38046         overflow.
38048 2017-02-22  Martin Liska  <mliska@suse.cz>
38050         PR lto/79587
38051         * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
38052         * data-streamer-out.c (streamer_write_gcov_count_stream):
38053         Likewise.
38054         * value-prof.c (stream_out_histogram_value): Make assert more
38055         precise based on type of counter.
38057 2017-02-21  Uros Bizjak  <ubizjak@gmail.com>
38059         PR target/79593
38060         * config/i386/i386.md (standard_x87sse_constant_load splitter):
38061         Use nonimmediate_operand instead of memory_operand for operand 1.
38062         (float-extend standard_x87sse_constant_load splitter): Ditto.
38064 2017-02-21  Jeff Law  <law@redhat.com>
38066         PR tree-optimization/79621
38067         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
38068         blocks with edges to themselves.
38070 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38072         PR target/79633
38073         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
38074         is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
38075         Use gimple_call_builtin_p.
38077         PR target/79570
38078         * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
38079         on temporarily removed DEBUG_INSNs.
38081         PR tree-optimization/79649
38082         * tree-loop-distribution.c (classify_partition): Give up on
38083         non-generic address space loads/stores.
38085 2017-02-21  Aldy Hernandez  <aldyh@redhat.com>
38087         * doc/loop.texi (Loop manipulation): Remove nonexistent
38088         tree_ssa_loop_version from the documentation.
38089         * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
38091 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38093         PR target/79494
38094         * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
38095         make_reg_eh_region_note_nothrow_nononlocal on call_insn.
38096         * config/rs6000/rs6000.c: Include except.h.
38097         (rs6000_expand_split_stack_prologue): Call
38098         make_reg_eh_region_note_nothrow_nononlocal on the call insn.
38100 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38102         PR lto/79579
38103         * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
38104         have been analyzed.
38106 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38108         * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
38109         for backward compatibility only.
38110         * doc/invoke.texi (Option Summary): Remove all references to
38111         -fipa-cp-alignment.
38113 2017-02-21  Matthew Fortune  <matthew.fortune@imgtec.com>
38115         PR target/78660
38116         Revert:
38117         2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38119         * lra-constraints.c (curr_insn_transform): Handle
38120         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38122 2017-02-21  Martin Liska  <mliska@suse.cz>
38124         * config/i386/i386.opt: Replace -masm-dialect with -masm.
38126 2017-02-21  Thomas Schwinge  <thomas@codesourcery.com>
38128         PR translation/79638
38129         * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
38131 2017-02-21  Eric Botcazou  <ebotcazou@adacore.com>
38133         PR ada/67205
38134         * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
38135         (arm_function_ok_for_sibcall): Return false for an indirect call by
38136         descriptor if all the argument registers are used.
38137         (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
38138         alignment of the function.
38140 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38142         PR tree-optimization/61441
38143         * simplify-rtx.c (simplify_const_unary_operation): For
38144         -fsignaling-nans and sNaN operand, return NULL_RTX rather than
38145         the sNaN unmodified.
38147 2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
38149         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
38150         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
38151         instead of SYSTEM_HEADER_DIR.
38153 2017-02-20  Gerald Pfeifer  <gerald@pfeifer.com>
38154             Martin LiÅ¡ka  <mliska@suse.cz>
38156         * doc/invoke.texi (use-after-scope-direct-emission-threshold):
38157         Fix typos and grammar, use active voice, and clarify.
38159 2017-02-20  Marek Polacek  <polacek@redhat.com>
38161         PR middle-end/79537
38162         * gimplify.c (gimplify_expr): Handle unused *&&L;.
38164         PR sanitizer/79558
38165         * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
38167 2017-02-20  Jakub Jelinek  <jakub@redhat.com>
38169         PR target/79568
38170         * config/i386/i386.c (ix86_expand_builtin): Handle
38171         OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
38172         ix86_builtins_isa[fcode].isa as a requirement of those
38173         flags and any other flag in the bitmask.
38174         (ix86_init_mmx_sse_builtins): Use 0 instead of
38175         ~OPTION_MASK_ISA_64BIT as mask.
38176         * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
38177         __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
38178         __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
38179         __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
38181 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38183         PR target/78012
38184         * lra-constraints.c (split_reg): Check requested split mode
38185         is supported by the register.
38187 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38189         * lra-constraints.c (simplify_operand_subreg): Remove early
38190         return false.
38192 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38194         PR target/78660
38195         * lra-constraints.c (curr_insn_transform): Tighten condition
38196         for converting SUBREG reloads from OP_OUT to OP_INOUT.
38198 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38200         PR target/78660
38201         * lra-constraints.c (curr_insn_transform): Handle
38202         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38204 2017-02-19  Uros Bizjak  <ubizjak@gmail.com>
38206         Revert:
38207         2016-05-30  Uros Bizjak  <ubizjak@gmail.com>
38209         * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
38211 2017-02-19  Jonathan Wakely  <jwakely@redhat.com>
38213         PR c++/69523
38214         * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
38215         description.
38217 2017-02-19  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38219         * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
38220         for FMA_EXPR.
38222 2017-02-18  Jakub Jelinek  <jakub@redhat.com>
38224         * final.c (last_columnnum, override_columnnum): New variables.
38225         (final_start_function): Set last_columnnum, pass it to begin_prologue
38226         hook and pass 0 to dwarf2out_begin_prologue.
38227         (final_scan_insn): Update override_columnnum.  Pass last_columnnum
38228         to source_line debug hook.
38229         (notice_source_line): Compute last_columnnum and for debug_column_info
38230         return true on column changes.
38231         * debug.h (struct gcc_debug_hooks): Add column argument to
38232         source_line and begin_prologue hooks.
38233         (debug_nothing_int_charstar_int_bool): Remove prototype.
38234         (debug_nothing_int_int_charstar,
38235         debug_nothing_int_int_charstar_int_bool): New prototypes.
38236         (dwarf2out_begin_prologue): Add column argument.
38237         * debug.c (do_nothing_debug_hooks): Adjust source_line and
38238         begin_prologue hooks.
38239         (debug_nothing_int_charstar_int_bool): Remove.
38240         (debug_nothing_int_int_charstar,
38241         debug_nothing_int_int_charstar_int_bool): New functions.
38242         * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38243         through to dwarf2out_source_line.
38244         (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38245         (dwarf2out_source_line): Add column argument, emit it if requested.
38246         * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38247         arguments.
38248         * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38249         * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38250         * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38251         through to dwarf2out_begin_prologue.
38252         (vmsdbgout_source_line): Add column argument, pass it through to
38253         dwarf2out_source_line.
38254         * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38255         dbxout_source_line caller.
38256         (dbxout_source_line): Add column argument.
38258         * common.opt (gno-column-info, gcolumn-info): New options.
38259         * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38260         (check_die): Also test for multiple DW_AT_decl_column attributes.
38261         (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38262         DW_AT_decl_column if requested.
38263         (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38264         if requested.
38265         (gen_variable_die): Likewise.
38266         (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38267         * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38269         PR target/79569
38270         * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38271         * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38272         (ix86_handle_option): Handle OPT_m3dnowa.
38273         * doc/invoke.texi (-m3dnowa): Document.
38274         * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38275         -m3dnowa instead of -m3dnow -march=athlon.
38277         PR target/79559
38278         * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38279         instead of gcc_assert for K, r and R code checks.  Formatting fixes.
38281 2017-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
38283         PR target/79261
38284         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38285         support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38286         * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38287         generator for vsx_xxpermdi_<mode>_be.
38288         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38289         force big-endian semantics.
38290         (vsx_xxpermdi_<mode>_be): New define_expand with same
38291         implementation as previous version of vsx_xxpermdi_<mode>.
38293 2017-02-17  Jakub Jelinek  <jakub@redhat.com>
38295         PR tree-optimization/79327
38296         * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38297         variable, its initialization and use.
38299 2017-02-17  Julia Koval  <julia.koval@intel.com>
38301         * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38302         (OPTION_MASK_ISA_PKU_UNSET): New.
38303         (ix86_handle_option): Handle -mrdpid.
38304         * config/i386/cpuid.h (bit_RDPID): New.
38305         * config/i386/driver-i386.c (host_detect_local_cpu):
38306         Detect RDPID feature.
38307         * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38308         * config/i386/i386-c.c (ix86_target_macros_internal):
38309         Handle RDPID flag.
38310         * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38311         (ix86_valid_target_attribute_inner_p): Add "rdpid".
38312         (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38313         * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38314         * config/i386/i386.md (define_insn "rdpid"): New.
38315         * config/i386/i386.opt Add -mrdpid.
38316         * config/i386/immintrin.h (_rdpid_u32): New.
38318 2017-02-17  Vladimir Makarov  <vmakarov@redhat.com>
38320         PR rtl-optimization/79541
38321         * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38322         instead of transforming it into USE.
38324 2017-02-17  Segher Boessenkool  <segher@kernel.crashing.org>
38326         * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38327         If HONOR_SNANS (SFmode) force the input to a register.
38328         (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38329         (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38330         an frsp or similar insn.
38332 2017-02-17  Martin Liska  <mliska@suse.cz>
38334         PR rtl-optimization/79577
38335         * params.def (selsched-max-sched-times): Increase minimum to 1.
38337 2017-02-17  Martin Liska  <mliska@suse.cz>
38339         PR rtl-optimization/79574
38340         * gcse.c (want_to_gcse_p): Prevent integer overflow.
38342 2017-02-17  Martin Liska  <mliska@suse.cz>
38344         PR tree-optimization/79529
38345         * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38346         ssa_defined_default_def_p to handle cases which are implicitly
38347         defined.
38348         * tree-ssa.c (ssa_defined_default_def_p): New function.
38349         (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38350         which are implicitly defined.
38351         * tree-ssa.h (ssa_defined_default_def_p): Declare.
38353 2017-02-17  Richard Biener  <rguenther@suse.de>
38355         PR middle-end/79576
38356         * params.def (max-ssa-name-query-depth): Limit to 10.
38358 2017-02-17  Richard Biener  <rguenther@suse.de>
38360         PR tree-optimization/79552
38361         * tree-ssa-structalias.c (visit_loadstore): Properly verify
38362         default defs.
38364 2017-02-17  Richard Biener  <rguenther@suse.de>
38366         PR bootstrap/79567
38367         * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38369 2017-02-17  Marek Polacek  <polacek@redhat.com>
38371         PR middle-end/79536
38372         * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38373         (fold_negate_expr): New wrapper.
38375 2017-02-16  Sandra Loosemore  <sandra@codesourcery.com>
38377         * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]: 
38378         Correct terminology and de-emphasize pre-standard behavior.
38380 2017-02-16  Alan Modra  <amodra@gmail.com>
38382         PR rtl-optimization/79286
38383         * ira.c (def_dominates_uses): New function.
38384         (update_equiv_regs): Don't create an equivalence for insns that
38385         may trap where the register def does not dominate the use.
38387 2017-02-16  Vladimir Makarov  <vmakarov@redhat.com>
38389         PR rtl-optimization/78127
38390         * lra.c (lra): Call lra_eliminate before finish the loop after
38391         lra_constraint.
38393 2017-02-16  Richard Biener  <rguenther@suse.de>
38395         * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38396         isl/isl_val.h.
38397         * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38398         (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38399         * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38400         (isl_val_int_from_wi): New function.
38401         (extract_affine_gmp): Rename to ...
38402         (extract_affine_wi): ... this, take a widest_int.
38403         (extract_affine_int): Just wrap extract_affine_wi.
38404         (add_param_constraints): Use isl_val_int_from_wi.
38405         (add_loop_constraints): Likewise, and extract_affine_wi.
38407 2017-02-15  Jeff Law  <law@redhat.com>
38409         PR middle-end/79521
38410         * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38411         ira_init_register_move_cost_if_necessary.
38413 2017-02-15  Martin Sebor  <msebor@redhat.com>
38415         PR middle-end/32003
38416         * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38417         removed in a prior commit.
38419 2017-02-15  Bin Cheng  <bin.cheng@arm.com>
38421         PR tree-optimization/79347
38422         * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38423         counters during peeling.
38425 2017-02-15  Thomas Schwinge  <thomas@codesourcery.com>
38427         * Makefile.in (site.exp): Remove "set ISLVER".
38429 2017-02-15  Jakub Jelinek  <jakub@redhat.com>
38431         PR target/79487
38432         * real.c (real_from_integer): Call real_convert even for decimal.
38434 2017-02-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38436         PR target/79421
38437         * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38439 2017-02-14  Andrew Pinski  <apinski@cavium.com>
38441         * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38442         cores and change the partno/implementer to be correct.
38443         (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38444         the 'B" as the implementer.
38445         * config/aarch64/aarch64-tune.md: Regenerate.
38447 2017-02-14  Carl Love  <cel@us.ibm.com>
38449         * config/rs6000/rs6000.c: Add case statement entry to make the
38450         xvcvuxdsp built-in argument unsigned.
38451         * config/rs6000/vsx.md: Fix the source and return operand types so they
38452         match the instruction definitions from the ISA document.  Fix typo
38453         in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38454         statement.
38456 2017-02-14  Vladimir Makarov  <vmakarov@redhat.com>
38458         PR target/79282
38459         * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38460         member early_clobber_alts.
38461         * lra-lives.c (reg_early_clobber_p): New.
38462         (process_bb_lives): Use it.
38463         * lra.c (new_insn_reg): New arg early_clobber_alts.  Use it.
38464         (debug_operand_data): Initialize early_clobber_alts.
38465         (setup_operand_alternative): Set up early_clobber_alts.
38466         (collect_non_operand_hard_regs): Ditto.  Pass early clobber
38467         alternatives to new_insn_reg.
38468         (add_regs_to_insn_regno_info): Add arg early_clobber_alts.  Use
38469         it.
38470         (lra_update_insn_regno_info): Pass the new arg.
38472 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38474         PR middle-end/79505
38475         * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38476         (new_oacc_loop_raw): Don't clear already cleared fields.
38478         PR target/79481
38479         * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38480         _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38481         _mm512_prefetch_i64gather_ps): New inline functions and macros.
38483 2017-02-14  Uros Bizjak  <ubizjak@gmail.com>
38485         PR target/79495
38486         * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38488 2017-02-14  H.J. Lu  <hongjiu.lu@intel.com>
38490         PR target/79498
38491         * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38492         the extra instruction to the right place to store 128-bit constant
38493         when needed.
38495 2017-02-14  Martin Sebor  <msebor@redhat.com>
38497         PR middle-end/79448
38498         * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38499           warning for strings of unknown length.
38501 2017-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
38503         * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38505 2017-02-14  Jeff Law  <law@redhat.com>
38507         PR target/79404
38508         * ira-costs.c (scan_one_insn): Initialize register move costs
38509         for pseudos seen in USE/CLOBBER insns.
38511         PR tree-optimization/79095
38512         * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38513         if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38514         (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38515         if the operands are known to be not equal, then the resulting range
38516         is ~[0,0].
38517         (intersect_ranges): If the new range is ~[0,0] and the old range is
38518         wide, then prefer ~[0,0].
38519         * tree-vrp.c (overflow_comparison_p_1): New function.
38520         (overflow_comparison_p): New function.
38521         * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38522         if NAME is used in an overflow test.
38523         (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38524         overflow check that can be expressed as an equality test, then adjust
38525         ops to be that equality test.
38527 2017-02-14  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38529         * config/s390/s390-builtin-types.def: Remove flags argument.
38530         * config/s390/s390.c (s390_init_builtins): Likewise.
38532 2017-02-14  Martin Liska  <mliska@suse.cz>
38534         * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38535         vector.  Fix trailing white spaces.
38537 2017-02-14  James Greenhalgh  <james.greenhalgh@arm.com>
38539         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38540         HFmode.
38542 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38544         PR rtl-optimization/68664
38545         * config/arm/arm.c (arm_sched_can_speculate_insn):
38546         New function.  Declare prototype.
38547         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38549 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38551         PR rtl-optimization/68664
38552         * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38553         New function.
38554         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38556 2017-02-14  Amit Pawar  <amit.pawar@amd.com>
38558         * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38559         max skip bytes for function, loop and jump.
38561 2017-02-14  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38563         * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38564         ABS_EXPR for gimple dump.
38566 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38568         PR target/79462
38569         * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38571         PR tree-optimization/79408
38572         * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38573         case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38574         (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38575         also if rhs1 is INTEGER_CST.
38577 2017-02-14  Richard Biener  <rguenther@suse.de>
38579         PR middle-end/79432
38580         * tree-into-ssa.c (insert_phi_nodes): When the function can
38581         have abnormal edges rewrite SSA names with broken use-def
38582         dominance out of SSA and register them for PHI insertion.
38584 2017-02-13  Martin Sebor  <msebor@redhat.com>
38586         PR middle-end/79496
38587         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38588         clearing info.nowrite flag when snprintf size argument is a range.
38590 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38592         * cprop.c (cprop_jump): Add missing space in string literal.
38593         * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38594         (get_constraint_for_component_ref): Likewise.
38595         * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38596         * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38597         * lra-constraints.c (process_alt_operands): Likewise.
38598         * ipa-inline.c (inline_small_functions): Likewise.
38599         * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38600         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38601         * trans-mem.c (diagnose_tm_1_op): Likewise.
38602         * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38603         (grid_parallel_clauses_gridifiable): Likewise.
38605         * config/nvptx/mkoffload.c (process): Add space in between
38606         , and %d.
38608         * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38609         "MOD4_SSE_REGS" and "ALL_REGS".
38611         * spellcheck.c (test_data): Add , in between "foo" and "food".
38613 2017-02-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38615         PR target/79449
38616         * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38617         boundary crossing check and subsequent code generation agree.
38619 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38621         * config/aarch64/aarch64.c (has_memory_op): Delete.
38622         (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38623         has_memory_op.
38625 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38627         PR rtl-optimization/79388
38628         PR rtl-optimization/79450
38629         * combine.c (distribute_notes): When removing TEM_INSN for which
38630         corresponding dest has last value recorded, invalidate that last
38631         value.
38633 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38635         * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38636         of explicit '@'.  Add missing assembly comment marker on branch costs
38637         printout.
38639 2017-02-13  Nathan Sidwell  <nathan@acm.org>
38641         * gengtype-lex.l (<in_struct>): Add '/'.
38643 2017-02-13  Martin Liska  <mliska@suse.cz>
38645         PR c/79471
38646         * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38648 2017-02-13  Richard Biener  <rguenther@suse.de>
38650         * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38651         Remove.
38652         * configure: Re-generate.
38653         * config.in: Likewise.
38654         * graphite-dependences.c: Simplify as if
38655         HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38656         * graphite-isl-ast-to-gimple.c: Likewise.
38657         * graphite-optimize-isl.c: Likewise.
38658         * graphite-poly.c: Likewise.
38659         * graphite-sese-to-poly.c: Likewise.
38660         * graphite.h: Likewise.
38661         * toplev.c: Include isl/version.h and use isl_version () for
38662         printing the ISL version.
38663         * doc/install.texi: Update ISL requirement.
38665 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38667         * doc/standards.texi (Standards): Update reference to
38668         Objective-C 2.0.
38670 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38672         * doc/extend.texi (Named Address Spaces): sourceware.org now
38673         defaults to https.
38674         * doc/install.texi (Binaries): Ditto.
38675         (Specific): Ditto.
38677 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38679         * doc/cpp.texi: Replace "stringify"/"stringification" with C
38680         standard terminology "stringize"/"stringizing" throughout.
38681         * doc/cppinternals.texi: Likewise.
38683 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38685         * doc/extend.texi: Fix some spelling mistakes and typos.
38686         * doc/invoke.texi: Likewise.
38688 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38690         PR ipa/79224
38691         * params.def (inline-min-speedup) Change from 10 to 8.
38693 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38695         * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38696         4.5.
38698 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38700         PR ipa/79224
38701         * ipa-inline-analysis.c (get_minimal_bb): New function.
38702         (record_modified): Use it.
38703         (remap_edge_change_prob): Handle also ancestor functions.
38705 2017-02-11  Gerald Pfeifer  <gerald@pfeifer.com>
38707         * doc/contrib.texi (Contributors): Remove broken link into
38708         the Mauve CVS repository.
38710 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38712         PR middle-end/79454
38713         * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38714         result computation whenever lhs doesn't have vector mode, not
38715         just when it has BLKmode.
38717 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38719         * doc/makefile.texi (profiledbootstrap): Refer to the
38720         installation instructions only in textual form.
38722 2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38724         PR target/79295
38725         * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38727 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38729         * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38730         (Specific): Update mingw-w64 reference.
38731         (Binaries): Ditto.
38732         (Specific): Remove broken link to Renesas RX processor.
38734 2017-02-10  Richard Biener  <rguenther@suse.de>
38736         * toplev.c (process_options): Do not mention obsolete graphite
38737         options when printing sorry message about missing graphite support.
38738         Mention -floop-nest-optimize.
38740 2017-02-10  Christophe Lyon  <christophe.lyon@linaro.org>
38742         * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38743         (vtst_p16): Likewise.
38744         (vtstq_p8): Likewise.
38745         (vtstq_p16): Likewise.
38746         (vtst_p64): New.
38747         (vtstq_p64): Likewise.
38748         * config/arm/arm_neon.h (vgetq_lane_p64): New.
38749         (vset_lane_p64): New.
38750         (vsetq_lane_p64): New.
38752 2017-02-10  Jakub Jelinek  <jakub@redhat.com>
38754         PR tree-optimization/79411
38755         * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38756         stmt operands are SSA_NAMEs used in abnormal phis.
38757         (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38758         phis.
38760 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38762         PR ipa/70795
38763         * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38764         flag if needed.
38766 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38768         * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38770 2017-02-09  Jakub Jelinek  <jakub@redhat.com>
38772         * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38773         to avoid warning.
38775         PR c/79413
38776         * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38777         not arbitrary TREE_CONSTANT.
38779         PR c/79431
38780         * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38781         "omp declare target link" attribute unless is_global_var.
38782         * omp-offload.c (find_link_var_op): Likewise.
38784 2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
38785             Chung-Lin Tang  <cltang@codesourcery.com>
38787         * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38788         OMP_CLAUSE_TILE.
38789         (gimplify_adjust_omp_clauses): Don't delete TILE.
38790         (gimplify_omp_for): Deal with TILE.
38791         * internal-fn.c (expand_GOACC_TILE): New function.
38792         * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38793         (GOACC_TILE): New.
38794         * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38795         (expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
38796         element fields.
38797         (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38798         avoid DIV for outermost collapse var.
38799         (expand_oacc_for): Insert tile element loop as needed.  Adjust.
38800         Remove out of date comments, fix whitespace.
38801         * omp-general.c (omp_extract_for_data): Deal with tiling.
38802         * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38803         adjust OLF_DIM_BASE value.
38804         (struct omp_for_data): Add tiling field.
38805         * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38806         (lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
38807         for auto loops.  Remove default auto determining, moved to
38808         oacc_loop_fixed_partitions.
38809         * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38810         stmts, add e_mask field.
38811         (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38812         (oacc_thread_numbers): Use oacc_dim_call.
38813         (oacc_xform_tile): New.
38814         (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38815         (finish_oacc_loop): Adjust for ifns vector.
38816         (oacc_loop_discover_walk): Append loop abstraction sites to list,
38817         add case for GOACC_TILE fns.
38818         (oacc_loop_xform_loop): Delete.
38819         (oacc_loop_process): Iterate over call list directly, and add
38820         handling for GOACC_TILE fns.
38821         (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38822         dump partitioning.
38823         (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38824         vector partitioning to outer loops.  Assign 2 partitions to loops
38825         when available. Add TILE handling.
38826         (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38827         (execite_oacc_device_lower): Process GOACC_TILE fns,
38828         ignore unknown specs.
38829         * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38830         * tree.c (omp_clause_num_ops): Adjust TILE ops.
38831         * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38833 2017-02-09  Gerald Pfeifer  <gerald@pfeifer.com>
38835         * configure.ac (ACX_BUGURL): Update.
38836         * configure: Regenerate.
38838 2017-02-09  Richard Biener  <rguenther@suse.de>
38840         PR tree-optimization/69823
38841         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38842         Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.
38844 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38846         * config/arc/arc-c.def: Add __NPS400__ definition.
38847         * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38848         (TARGET_NPS400): Define.
38850 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38852         * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38853         file.
38854         (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38855         pointer, arch_info.
38856         (arc_cpu_types): Fill the arch_info field with a pointer into the
38857         arc_arch_types table.
38858         (arc_selected_cpu): Declare.
38859         * config/arc/arc.c (arc_selected_cpu): Make global.
38860         (arc_selected_arch): Delete.
38861         (arc_base_cpu): Delete.
38862         (arc_override_options): Remove references to deleted variables,
38863         update access to arch information.
38864         (ARC_OPT): Update access to arch information.
38865         (ARC_OPTX): Likewise.
38866         * config/arc/arc.h (arc_base_cpu): Remove declaration.
38867         (TARGET_ARC600): Update access to arch information.
38868         (TARGET_ARC601): Likewise.
38869         (TARGET_ARC700): Likewise.
38870         (TARGET_EM): Likewise.
38871         (TARGET_HS): Likewise.
38872         * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38873         information.
38875 2017-02-08  Pat Haugen  <pthaugen@us.ibm.com>
38877         PR target/78604
38878         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38879         condition/operands for integer GE/LE/GEU/LEU operations.
38881 2017-02-08  Segher Boessenkool  <segher@kernel.crashing.org>
38883         PR translation/79397
38884         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38885         of AltiVec.
38887 2017-02-08  Martin Jambor  <mjambor@suse.cz>
38889         PR ipa/79375
38890         * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38891         whether allocation happened.
38892         (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38893         nothing was allocated.
38895 2017-02-08  Jakub Jelinek  <jakub@redhat.com>
38897         PR tree-optimization/79408
38898         * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38899         constant, but SSA_NAME with a known integer range, use the minimum
38900         of that range instead of op1 to determine if modulo can be replaced
38901         with its first operand.
38903 2016-02-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38905         * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38907 2017-02-08  Richard Biener  <rguenther@suse.de>
38909         PR tree-optimization/71824
38910         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38911         Check all loops contained in the merged region.
38913 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38915         * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
38917 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38919         * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
38920         (thunderxt88): Likewise.
38921         (thunderxt81): Disable LSE and change v8.1 to v8.
38922         (thunderxt83): Likewise.
38924 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38925             Richard Biener  <rguenther@suse.de>
38927         PR middle-end/79399
38928         * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
38929         type from int to size_t.
38930         * ira-costs.c (struct_costs_size): Change type from int to size_t.
38932 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38934         PR rtl-optimization/79386
38935         * cprop.c (bypass_conditional_jumps): Initialize
38936         bypass_last_basic_block already before splitting bbs after
38937         unconditional traps...
38938         (bypass_conditional_jumps): ... rather than here.
38940         PR target/79299
38941         * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
38942         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
38943         *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
38944         fix -masm=intel patterns.
38946 2017-02-07  Richard Biener  <rguenther@suse.de>
38948         PR tree-optimization/79256
38949         PR middle-end/79278
38950         * builtins.c (get_object_alignment_2): Use min_align_of_type
38951         to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
38952         and ADJUST_FIELD_ALIGN.
38954         * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
38955         type parameter.
38956         * doc/tm.texi: Regenerate.
38957         * stor-layout.c (layout_decl): Adjust.
38958         (update_alignment_for_field): Likewise.
38959         (place_field): Likewise.
38960         (min_align_of_type): Likewise.
38961         * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
38962         * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
38963         * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
38964         * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
38965         * config/frv/frv.c (frv_adjust_field_align): Likewise.
38966         * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
38967         * config/i386/i386.c (x86_field_alignment): Likewise.
38968         * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
38969         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
38970         * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
38971         * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
38972         * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
38973         * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
38974          Likewise.
38976         Revert
38977         2017-01-30  Richard Biener  <rguenther@suse.de>
38979         PR tree-optimization/79256
38980         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
38981         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
38982         alignment on TYPE.
38984 2017-02-07  Toma Tabacu  <toma.tabacu@imgtec.com>
38986         * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
38987         argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
38988         builtins to SImode and emit a zero-extend, if necessary.
38990 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38992         * docs/invoke.texi (RISC-V Options): Alphabetize.
38994 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38996         * doc/invoke.texi (RISC-V Options): Use two spaces to separate
38997         options.
38999 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
39001         * config/riscv/riscv.c: New file.
39002         * common/config/riscv/riscv-common.c: Likewise.
39003         * config.gcc: Likewise.
39004         * config/riscv/constraints.md: Likewise.
39005         * config/riscv/elf.h: Likewise.
39006         * config/riscv/generic.md: Likewise.
39007         * config/riscv/linux.h: Likewise.
39008         * config/riscv/multilib-generator: Likewise.
39009         * config/riscv/peephole.md: Likewise.
39010         * config/riscv/pic.md: Likewise.
39011         * config/riscv/predicates.md: Likewise.
39012         * config/riscv/riscv-builtins.c: Likewise.
39013         * config/riscv/riscv-c.c: Likewise.
39014         * config/riscv/riscv-ftypes.def: Likewise.
39015         * config/riscv/riscv-modes.def: Likewise.
39016         * config/riscv/riscv-opts.h: Likewise.
39017         * config/riscv/riscv-protos.h: Likewise.
39018         * config/riscv/riscv.h: Likewise.
39019         * config/riscv/riscv.md: Likewise.
39020         * config/riscv/riscv.opt: Likewise.
39021         * config/riscv/sync.md: Likewise.
39022         * config/riscv/t-elf-multilib: Likewise.
39023         * config/riscv/t-linux: Likewise.
39024         * config/riscv/t-linux-multilib: Likewise.
39025         * config/riscv/t-riscv: Likewise.
39026         * configure.ac: Likewise.
39027         * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
39028         Waterman as RISC-V maintainers.
39029         * doc/install.texi: Add RISC-V entries.
39030         * doc/invoke.texi: Add RISC-V options section.
39031         * doc/md.texi: Add RISC-V constraints section.
39032         * configure: Regenerated.
39034 2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
39036         PR target/66144
39037         * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
39038         false values to be constant vectors with all 0 or all 1 bits set.
39039         (vcondu<mode><mode>): Likewise.
39040         * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
39041         predicate.
39042         (fpmask_comparison_operator): Update comment.
39043         (vecint_comparison_operator): New predicate.
39044         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
39045         vector conditionals when the true and false values are constant
39046         vectors with all 0 bits or all 1 bits set.
39048 2017-02-06  Martin Sebor  <msebor@redhat.com>
39050         PR  tree-optimization/79376
39051         * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
39053 2017-02-06  Uros Bizjak  <ubizjak@gmail.com>
39055         * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
39056         explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
39057         to simplify split condition.
39059 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39061         * omp-expand.c (oxpand_omp_atomic_fetch_op,
39062         expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
39063         false.
39065 2017-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
39067         PR rtl-optimization/68664
39068         * target.def (can_speculate_insn): New hook.
39069         * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
39070         * doc/tm.texi: Regenerate.
39071         * sched-rgn.c (can_schedule_ready_p): Use the new hook.
39072         * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
39073         (rs6000_sched_can_speculate_insn): New function.
39075 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39077         PR tree-optimization/79284
39078         * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
39079         * tree-vect-stmts.c (vect_get_vec_def_for_operand,
39080         vectorizable_mask_load_store, vectorizable_operation,
39081         vect_is_simple_cond, get_same_sized_vectype): Use it instead
39082         of comparing TREE_CODE of a type against BOOLEAN_TYPE.
39083         * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
39084         vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
39085         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
39086         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
39087         Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
39088         is_gimple_assign (stmt).  Replace another such test with
39089         is_gimple_assign (stmt).
39091 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39093         PR target/78883
39094         * config/avr/avr.c (rtl-iter.h): Include it.
39095         (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
39096         (avr_legitimate_combined_insn): ...and implementation.
39098 2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39100         * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
39101         * config/s390/s390.c (s390_const_operand_ok)
39102         (s390_canonicalize_comparison, s390_extract_part)
39103         (s390_single_part, s390_contiguous_bitmask_nowrap_p)
39104         (s390_contiguous_bitmask_p, s390_rtx_costs)
39105         (legitimize_pic_address): Likewise.
39106         * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
39107         * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
39108         ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
39109         ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
39110         * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
39112 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39114         * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
39115         REGNO($0) == REGNO($1).
39117 2017-02-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39119         * config/s390/linux.h(SIZE_TYPE): Add comment.
39121 2017-02-06  Julian Brown  <julian@codesourcery.com>
39122             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39123             Virendra Pathak  <virendra.pathak@broadcom.com>
39125         * config/aarch64/aarch64-cores.def: Change the scheduler
39126         to Thunderx2t99.
39127         * config/aarch64/aarch64.md: Include thunderx2t99.md.
39128         * config/aarch64/thunderx2t99.md: New file.
39130 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39132         * doc/standards.texi (Go Language): Update link to language
39133         standard.
39135 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39137         * tree-eh.c (lower_resx): Sanitize profile.
39138         (cleanup_empty_eh_move_lp): Likewise.
39140 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39142         PR tree-ssa/79347
39143         * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
39144         ELSE_PROB.
39145         * cfgloopmanip.h (loop_version): Update prototype.
39146         * modulo-sched.c (sms_schedule): Update call of loop_version.
39147         * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
39148         * tree-parloops.c (gen_parallel_loop): Likewise.
39149         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
39150         * tree-ssa-loop-split.c (split_loop): Likewise.
39151         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
39152         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
39154 2017-02-05  Martin Liska  <mliska@suse.cz>
39156         PR bootstrap/78985
39157         * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
39158         variable to NULL.
39159         (print_operand_address): Initialize a struct to zero.
39161 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39163         * doc/contrib.texi (Contributors): Refer to Hans Boehm's
39164         garbage collector only in textual form.
39166 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39168         * doc/extend.texi (x86 specific memory model extensions for
39169         transactional memory): Simplify a phrase.
39171 2017-02-05  Eric Botcazou  <ebotcazou@adacore.com>
39173         PR target/79353
39174         * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
39175         'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
39176         (atomic_storedi_1): Likewise.
39178 2017-02-04  Jakub Jelinek  <jakub@redhat.com>
39180         PR tree-optimization/79338
39181         * tree-parloops.c (gather_scalar_reductions): Don't call
39182         vect_analyze_loop_form for loop->inner before destroying loop's
39183         loop_vinfo.
39185 2017-02-03  Martin Sebor  <msebor@redhat.com>
39187         PR tree-optimization/79327
39188         * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
39189         when precision has resulted in leading zeros.
39190         (format_integer): Adjust the likely counter to assume an unknown
39191         argument that may be zero is non-zero.
39193 2017-02-03  Jason Merrill  <jason@redhat.com>
39195         PR c++/78689
39196         * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
39197         avoid copying non-taken branch.
39199 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39201         PR tree-optimization/79340
39202         * tree-vect-loop.c (vectorizable_reduction): Release
39203         vec_defs elements after safe_splicing them into other vectors.
39204         Formatting fixes.
39206         PR tree-optimization/79327
39207         * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
39208         true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
39209         dirtype.
39210         (format_integer): Use wide_int_to_tree instead of build_int_cst
39211         + to_?hwi.  If argmin is NULL, just set argmin and argmax to
39212         TYPE_{MIN,MAX}_VALUE of argtype.  Simplify and fix computation
39213         of shortest and longest sequence.
39215 2017-02-03  Uros Bizjak  <ubizjak@gmail.com>
39217         * config/i386/i386.c (dimode_scalar_chain::convert_reg):
39218         Use pextrd for TARGET_SSE4_1 when creating scalar copy.
39220 2017-02-03  Walter Lee  <walt@tilera.com>
39222         PR target/78862
39223         * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
39224         after initial stackframe link reg save.
39225         * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
39227 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39229         PR target/79354
39230         * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
39231         wu for stxssp alternative.
39233 2017-02-03  Martin Sebor  <msebor@redhat.com>
39235         PR tree-optimization/79352
39236         * gimple-fold.c (get_range_strlen): Add argument.
39237         (get_range_strlen): Change return type to bool.
39238         (get_maxval_strlen): Pass in a dummy argument.
39239         * gimple-fold.h (get_range_strlen): Change return type to bool.
39240         * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39241         * tree.h (array_at_struct_end_p): Add argument.
39242         * tree.c (array_at_struct_end_p): Handle it.
39244 2017-02-03  Martin Liska  <mliska@suse.cz>
39246         PR lto/66295
39247         * multiple_target.c (create_dispatcher_calls): Redirect edge
39248         from a caller of a dispatcher.
39249         (expand_target_clones): Make the clones local.
39250         (ipa_target_clone): Do both target clones and resolvers.
39251         (ipa_dispatcher_calls): Remove the pass.
39252         (pass_dispatcher_calls::gate): Likewise.
39253         (make_pass_dispatcher_calls): Likewise.
39254         * passes.def (pass_target_clone): Put as very first IPA early
39255         pass.
39257 2017-02-03  Martin Liska  <mliska@suse.cz>
39259         * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39260         in case of a function with ifunc attribute.
39262 2017-02-03  Martin Liska  <mliska@suse.cz>
39264         * cgraph.c (cgraph_node::dump): Dump function version info.
39265         * symtab.c (symtab_node::dump_base): Add missing new line.
39267 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39269         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39270         (ifcombine_ifandif): Use it.
39272 2017-02-03  Martin Liska  <mliska@suse.cz>
39274         * doc/invoke.texi: Document default value for
39275         use-after-scope-direct-emission-threshold.
39277 2017-02-03  Martin Liska  <mliska@suse.cz>
39279         PR tree-optimization/79339
39280         * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39281         (format_floating): Likewise.
39283 2017-02-03  Martin Liska  <mliska@suse.cz>
39285         PR ipa/79337
39286         * ipa-prop.c (ipa_node_params_t::insert): Remove current
39287         implementation.
39288         (ipa_node_params_t::remove): Likewise.
39289         * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39290         initialization from removed ipa_node_params_t::insert.
39291         (ipa_node_params::~ipa_node_params): Move from removed
39292         ipa_node_params_t::release.
39293         * symbol-summary.h (symbol_summary::m_released): New member.
39294         Do not release a summary twice.  Do not allow to call finalizer
39295         for types of a summary that live in GGC memory.
39297 2017-02-02  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39299         * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39300         cmp_branch fusion.
39302 2017-02-02  Martin Sebor  <msebor@redhat.com>
39304         PR middle-end/79275
39305         * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39306         (format_string): Tighten up the range of output for non-constant
39307         strings and correct the expected range for wide non-constant strings.
39309 2017-02-02  Martin Sebor  <msebor@redhat.com>
39311         * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39313         PR middle-end/32003
39314         * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39315         index.
39316         (-fdump-tree-@var): Add to index and document how to come up
39317         with pass-specific option and dump file names.
39318         (-fdump-passes): Clarify where to look for output.
39320 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39322         PR middle-end/77445
39323         * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39324         statistics of the analyzed path; allow threading for speed when
39325         any of BBs along the path are optimized for speed.
39327 2017-02-02  Eric Botcazou  <ebotcazou@adacore.com>
39329         PR middle-end/78468
39330         * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39331         settings of the virtual registers.
39333         Revert again
39334         2016-08-23  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39336         * explow.c (get_dynamic_stack_size): Take known alignment of stack
39337         pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39338         needed.
39340 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39342         * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39343         ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39345 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39347         * config/s390/s390.md: Add missing comments with the expanded
39348         mnemonics.
39349         * config/s390/vector.md: Likewise.
39350         * config/s390/vx-builtins.md: Likewise.
39352 2017-02-02  Jakub Jelinek  <jakub@redhat.com>
39354         PR target/79197
39355         * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39356         (fixuns_trunc<mode>di2): ... this, remove previous expander.  Put all
39357         conditions on a single line.
39359 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39361         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39362         __S390_VX__ to __VX__.
39364 2017-02-01  Andrew Pinski  <apinski@cavium.com>
39366         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39367         stmt_info to record_stmt_cost.
39368         (vect_get_known_peeling_cost): Pass stmt_info if known to
39369         record_stmt_cost.
39370         * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39371         cpu_vector_cost field into
39372         scalar_int_stmt_cost and scalar_fp_stmt_cost.  Split vec_stmt_cost
39373         field into vec_int_stmt_cost and vec_fp_stmt_cost.
39374         * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39375         splitting of scalar_stmt_cost and vec_stmt_cost.
39376         (thunderx_vector_cost): Likewise.
39377         (cortexa57_vector_cost): LIkewise.
39378         (exynosm1_vector_cost): Likewise.
39379         (xgene1_vector_cost): Likewise.
39380         (thunderx2t99_vector_cost): Improve after the splitting of the two
39381         fields.
39382         (aarch64_builtin_vectorization_cost): Update for the splitting of
39383         scalar_stmt_cost and vec_stmt_cost.
39385 2017-02-01  Torvald Riegel  <triegel@redhat.com>
39386             Richard Henderson  <rth@redhat.com>
39388         * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39389         conditional on existance of a fast atomic load.
39390         * optabs-query.c (can_atomic_load_p): New function.
39391         * optabs-query.h (can_atomic_load_p): Declare it.
39392         * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39393         no fast atomic load is available for the particular size of access.
39394         (expand_atomic_compare_and_swap): Likewise.
39395         (expand_atomic_load): Likewise.
39396         (expand_atomic_store): Likewise.
39397         (expand_atomic_fetch_op): Likewise.
39398         * testsuite/lib/target-supports.exp
39399         (check_effective_target_sync_int_128): Remove x86 because it provides
39400         no fast atomic load.
39401         (check_effective_target_sync_int_128_runtime): Likewise.
39403 2017-02-01  Richard Biener  <rguenther@suse.de>
39405         * graphite.c: Include tree-vectorizer.h for find_loop_location.
39406         (graphite_transform_loops): Provide opt-info for optimized nests.
39407         * tree-parloop.c (parallelize_loops): Provide opt-info for
39408         parallelized loops.
39410 2017-02-01  Richard Biener  <rguenther@suse.de>
39412         PR middle-end/79315
39413         * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39414         was not set before.
39416 2017-02-01  Richard Biener  <rguenther@suse.de>
39418         PR tree-optimization/71824
39419         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39420         Verify the loops are valid in the merged SESE region.
39421         (scop_detection::can_represent_loop_1): Check analyzing the
39422         evolution of the number of iterations in the region succeeds.
39424 2017-01-31  Ian Lance Taylor  <iant@golang.org>
39426         * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39427         REG_ARGS_SIZE note to 32-bit push insns and call insn.
39429 2017-01-31  David Malcolm  <dmalcolm@redhat.com>
39431         PR preprocessor/79210
39432         * input.c (get_substring_ranges_for_loc): Replace line_width
39433         assertion with error-handling.
39435 2017-01-31  Richard Biener  <rguenther@suse.de>
39437         PR tree-optimization/77318
39438         * graphite-sese-to-poly.c (extract_affine): Fix assert.
39439         (create_pw_aff_from_tree): Take loop parameter.
39440         (add_condition_to_pbb): Pass loop of the condition to
39441         create_pw_aff_from_tree.
39443 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39445         * config/s390/s390.c (s390_asan_shadow_offset): New function.
39446         (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39448 2017-01-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
39450         PR target/78597
39451         PR target/79038
39452         * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39453         no longer used.
39454         (convert_int_to_float128): Likewise.
39455         * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39456         (convert_int_to_float128): Likewise.
39457         * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39458         (UNSPEC_IEEE128_CONVERT): Likewise.
39459         (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39460         rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39461         Use local variables for IBM extended format.
39462         (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39463         (fix_trunc<mode>si2_fprs): Likewise.
39464         (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39465         (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39466         (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39467         to know that we can now have integers of all sizes in vector
39468         registers.
39469         (fix<uns>_<mode>di2_hw): Likewise.
39470         (float<uns>_<mode>si2_hw): Likewise.
39471         (fix_<mode>si2_hw): Likewise.
39472         (fixuns_<mode>si2_hw): Likewise.
39473         (float<uns>_<mode>di2_hw): Likewise.
39474         (float_<mode>di2_hw): Likewise.
39475         (float_<mode>si2_hw): Likewise.
39476         (floatuns_<mode>di2_hw): Likewise.
39477         (floatuns_<mode>si2_hw): Likewise.
39478         (xscvqp<su>wz_<mode>): Delete, no longer used.
39479         (xscvqp<su>dz_<mode>): Likewise.
39480         (xscv<su>dqp_<mode>): Likewise.
39481         (ieee128_mfvsrd_64bit): Likewise.
39482         (ieee128_mfvsrd_32bit): Likewise.
39483         (ieee128_mfvsrwz): Likewise.
39484         (ieee128_mtvsrw): Likewise.
39485         (ieee128_mtvsrd_64bit): Likewise.
39486         (ieee128_mtvsrd_32bit): Likewise.
39488 2017-01-31  Martin Liska  <mliska@suse.cz>
39490         PR ipa/79285
39491         * ipa-prop.c (ipa_free_all_node_params): Call release method
39492         instead of ~sumbol_summary to not to trigger double times
39493         dtor of hash_map.
39495 2017-01-31  Aldy Hernandez  <aldyh@redhat.com>
39497         PR tree-optimization/71691
39498         * bitmap.h (class auto_bitmap): New.
39499         * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39500         is_maybe_undefined instead of ssa_undefined_value_p.
39502 2017-01-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39504         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39505         __S390_ARCH_LEVEL__ to __ARCH__.
39507 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39509         PR tree-optimization/79267
39510         * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39511         if should_remove_lhs_p is true.
39513 2017-01-30  Alexandre Oliva  <aoliva@redhat.com>
39515         PR debug/63238
39516         * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39517         (add_alignment_attribute): New.
39518         (base_type_die): Add alignment attribute.
39519         (subrange_type_die): Likewise.
39520         (modified_type_die): Likewise.
39521         (gen_array_type_die): Likewise.
39522         (gen_descr_array_type_die: Likewise.
39523         (gen_enumeration_type_die): Likewise.
39524         (gen_subprogram_die): Likewise.
39525         (gen_variable_die): Likewise.
39526         (gen_field_die): Likewise.
39527         (gen_ptr_to_mbr_type_die): Likewise.
39528         (gen_struct_or_union_type_die): Likewise.
39529         (gen_subroutine_type_die): Likewise.
39530         (gen_typedef_die): Likewise.
39531         (base_type_cmp): Compare alignment attribute.
39533 2017-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
39535         PR target/79170
39536         * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39537         (setb_unsigned) New pattern for setb with CCUNS.
39538         * config/rs6000/rs6000.c (expand_block_compare): Use a different
39539         subfc./subfe sequence to avoid overflow problems.  Generate a
39540         shorter sequence with cmpld/setb for power9.
39541         * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39542         for generating subfc. instruction.
39543         (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39544         now uses this instruction.
39546 2017-01-30  Ian Lance Taylor  <iant@google.com>
39548         PR debug/79289
39549         * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39550         for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39552 2017-01-30  Martin Sebor  <msebor@redhat.com>
39554         * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39555         Move constant to the right of a relational operator.
39556         (get_mpfr_format_length, format_character, format_string): Ditto.
39557         (should_warn_p, maybe_warn): Same.
39559         * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39561 2017-01-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
39563         PR lto/79061
39564         * asan.c (get_translation_unit_decl): Remove function.
39565         (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39567 2017-01-30  Martin Liska  <mliska@suse.cz>
39569         PR gcov-profile/79259
39570         * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39571         -fprofile-generate.
39573 2017-01-30  Martin Liska  <mliska@suse.cz>
39575         PR bootstrap/78985
39576         * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39577         Initialize variables with NULL value.
39579 2017-01-30  Richard Earnshaw  <rearnsh@arm.com>
39581         PR target/79260
39582         * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39583         tm_p_file.
39584         * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39586 2017-01-30  Richard Biener  <rguenther@suse.de>
39588         PR tree-optimization/79276
39589         * tree-vrp.c (process_assert_insertions): Properly adjust common
39590         when removing a duplicate.
39592 2017-01-30  Richard Biener  <rguenther@suse.de>
39594         PR tree-optimization/79256
39595         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39596         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39597         alignment on TYPE.
39598         * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39600 2017-01-30  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39602         PR target/79240
39603         * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39604         ("*r<noxa>sbg_<mode>_sll_bitmask")
39605         ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39606         ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39607         Use contiguous_bitmask_nowrap_operand.
39609 2017-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
39611         PR target/79268
39612         * config/rs6000/altivec.h (vec_xl): Revise #define.
39613         (vec_xst): Likewise.
39615 2017-01-27  Uros Bizjak  <ubizjak@gmail.com>
39617         * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39619 2017-01-27  Bernd Schmidt  <bschmidt@redhat.com>
39621         PR rtl-optimization/79194
39622         * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39623         traps before call to bypass_conditional_jumps.
39625 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39627         PR tree-optimization/71374
39628         * lra-constraints.c (check_conflict_input_operands): New.
39629         (match_reload): Use it.
39631 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39633         PR target/79131
39634         * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39635         account to calculate conflict_set.
39637 2017-01-27  Bin Cheng  <bin.cheng@arm.com>
39639         PR rtl-optimization/78559
39640         * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39641         other_insn in combine.
39643 2017-01-27  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
39645         * builtin-types.def: Use unsigned_char_type_node for BT_UINT8.  Use
39646         uint16_type_node for BT_UINT16.
39648 2017-01-27  David Malcolm  <dmalcolm@redhat.com>
39650         * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39651         "RTL Tests" to menu.
39652         (GIMPLE Tests): New node.
39653         (RTL Tests): New node.
39655 2017-01-27  Richard Biener  <rguenther@suse.de>
39657         PR tree-optimization/79245
39658         * tree-loop-distribution.c (distribute_loop): Apply cost
39659         modeling also to detected patterns.
39661 2017-01-27  Richard Biener  <rguenther@suse.de>
39663         PR tree-optimization/71433
39664         * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39665         (compare_assert_loc): New function.
39666         (process_assert_insertions): Sort and optimize assert locations
39667         to remove duplicates and push down identical assertions on
39668         edges to their destination block.
39670 2017-01-27  Richard Biener  <rguenther@suse.de>
39672         PR tree-optimization/79244
39673         * tree-vrp.c (remove_range_assertions): Forcefully propagate
39674         out SSA names even if abnormal.
39676 2017-01-27  Jakub Jelinek  <jakub@redhat.com>
39678         * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39679         * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39680         instead of MPFR_RNDN.
39682 2017-01-27  Richard Earnshaw  <rearnsha@arm.com>
39684         PR target/79239
39685         * arm.c (arm_option_override): Don't call build_target_option_node
39686         until after doing all option overrides.
39687         (arm_valid_target_attribute_tree): Likewise.
39689 2017-01-27  Martin Liska  <mliska@suse.cz>
39691         * doc/invoke.texi (-fprofile-arcs): Document profiling support
39692         for {cd}tors and C++ {cd}tors.
39694 2017-01-27  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39696         * config/s390/s390.md ("*setmem_long_and")
39697         ("*setmem_long_and_31z"): Use zero_extend instead of and.
39699 2017-01-26  Martin Sebor  <msebor@redhat.com>
39701         * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39702         of precision.
39704 2017-01-26  Martin Sebor  <msebor@redhat.com>
39706         * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39707         HAVE_DFmode before using XFmode or DFmode.
39708         (parse_directive): Avoid using the z length modifier to avoid
39709         the ISO C++98 does not support the â€˜z’ gnu_printf length modifier.
39711         PR middle-end/78703
39712         * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39713         to accept adjustment as an array.
39714         (get_int_range): New function.
39715         (struct directive): Make width and prec arrays.
39716         (directive::set_width, directive::set_precision): Call get_int_range.
39717         (format_integer, format_floating): Handle width and precision ranges.
39718         (format_string, parse_directive): Same.
39720 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39722         PR debug/79129
39723         * dwarf2out.c (generate_skeleton_bottom_up): For children with
39724         comdat_type_p set, just clone them, but keep the children in the
39725         original DIE.
39727         PR debug/78835
39728         * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39729         which have direct callers with -fvar-tracking-assignments enabled
39730         in the current TU.
39731         (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39732         inside of type units.
39734 2017-01-26  Martin Sebor  <msebor@redhat.com>
39736         PR middle-end/78703
39737         * gimple-ssa-sprintf.c (struct result_range): Add likely and
39738         unlikely counters.
39739         (struct format_result): Replace number_chars, number_chars_min,
39740         and number_chars_max with a single member of struct result_range.
39741         Remove bounded.
39742         (format_result::operator+=): Adjust.
39743         (struct fmtresult): Remove bounded.  Handle likely and unlikely
39744         counters.
39745         (fmtresult::adjust_for_width_or_precision): New function.
39746         (fmtresult:type_max_digits): New function.
39747         (bytes_remaining): Handle likely and unlikely counters.
39748         (min_bytes_remaining): Remove.
39749         (format_percent): Simplify.
39750         (format_integer, format_floating): Set likely and unlikely counters.
39751         (get_string_length, format_character, format_string): Same.
39752         (format_plain, should_warn_p): New function.
39753         (maybe_warn): Call should_warn_p.  Update diagnostic messages
39754         and handle those for all directives, including plain strings.
39755         (format_directive): Handle likely and unlikely counters.
39756         Remove unnecessary quoting from diagnostics.  Add an informational
39757         note.
39758         (add_bytes): Remove.
39759         (pass_sprintf_length::compute_format_length): Simplify.
39760         (try_substitute_return_value): Handle likely and unlikely counters.
39762 2017-01-26  Carl Love  <cel@us.ibm.com>
39764         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39765         bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39767 2017-01-26  Vladimir Makarov  <vmakarov@redhat.com>
39769         PR target/79131
39770         * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39771         endianess for subregs into account.
39772         * lra-constraints.c (lra_constraints): Do risky transformations
39773         always on the first iteration.
39774         * lra-lives.c (check_pseudos_live_through_calls): Add arg
39775         last_call_used_reg_set.
39776         (process_bb_lives): Define and use last_call_used_reg_set.
39777         * lra.c (lra): Always continue after lra_constraints on the first
39778         iteration.
39780 2017-01-26  Kirill Yukhin  <kirill.yukhin@gmail.com>
39782         * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39783         constant.
39784         * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39786 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39788         * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39789         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39790         * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39791         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39792         * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39793         __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39794         OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39795         * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39796         (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39798 2017-01-26  Marek Polacek  <polacek@redhat.com>
39800         PR c/79199
39801         * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39802         for the third operand.
39804 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39806         PR middle-end/79236
39807         * omp-low.c (struct omp_context): Add simt_stmt field.
39808         (scan_omp_for): Return omp_context *.
39809         (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39810         context to the _simt_ SIMD stmt.
39811         (lower_omp_for): For combined SIMD with sibling _simt_
39812         SIMD, make sure to use the same decls in _looptemp_
39813         clauses as in the sibling.
39815 2017-01-26  David Sherwood  <david.sherwood@arm.com>
39817         PR middle-end/79212
39818         * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39819         all contexts.
39821 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39823         PR target/70465
39824         * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39825         emit fld b; fld a; if possible.
39827         * brig-builtins.def: Update copyright years.
39828         * config/arm/arm_acle_builtins.def: Update copyright years.
39830 2017-01-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
39832         PR target/79179
39833         * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39834         constraint instead of o for the stxsd instruction.
39836 2017-01-25  Carl Love  <cel@us.ibm.com>
39838         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39839         of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39841 2017-01-25  Jonathan Wakely  <jwakely@redhat.com>
39843         * doc/invoke.texi (C++ Dialect Options): Fix typo.
39845 2017-01-25  Richard Biener  <rguenther@suse.de>
39847         PR tree-optimization/69264
39848         * target.def (vector_alignment_reachable): Improve documentation.
39849         * doc/tm.texi: Regenerate.
39850         * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39851         and add a comment.
39852         * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39853         earlier changes with respect to TYPE_USER_ALIGN.
39854         (vector_alignment_reachable_p): Likewise.  Improve dumping.
39856 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39858         PR target/79145
39859         * config/arm/arm.md (xordi3): Force constant operand into a register
39860         for TARGET_IWMMXT.
39862 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39864         * doc/invoke.texi (-fstore-merging): Correct default optimization
39865         levels at which it is enabled.
39866         (-O): Move -fstore-merging from list to...
39867         (-O2): ... Here.
39869 2017-01-25  Richard Biener  <rguenther@suse.de>
39871         PR debug/78363
39872         * omp-expand.c: Include debug.h.
39873         (expand_omp_taskreg): Make sure to generate early debug before
39874         outlining anything from a function.
39875         (expand_omp_target): Likewise.
39876         (grid_expand_target_grid_body): Likewise.
39878 2017-01-25  Maxim Ostapenko  <m.ostapenko@samsung.com>
39880         PR lto/79061
39881         * asan.c (get_translation_unit_decl): New function.
39882         (asan_add_global): Extract modules file name from globals
39883         TRANSLATION_UNIT_DECL name.
39885 2017-01-24  Eric Botcazou  <ebotcazou@adacore.com>
39887         PR target/77439
39888         * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39889         for long calls with APCS frame and VFP.
39891 2017-01-24  David Malcolm  <dmalcolm@redhat.com>
39893         * cfg.c (original_copy_tables_initialized_p): New function.
39894         * cfg.h (original_copy_tables_initialized_p): New decl.
39895         * cfgrtl.c (relink_block_chain): Guard the call to
39896         free_original_copy_tables with a call to
39897         original_copy_tables_initialized_p.
39898         * cgraph.h (symtab_node::native_rtl_p): New decl.
39899         * cgraphunit.c (symtab_node::native_rtl_p): New function.
39900         (symtab_node::needed_p): Don't assert for early assembly output
39901         for __RTL functions.
39902         (cgraph_node::finalize_function): Set "force_output" for __RTL
39903         functions.
39904         (cgraph_node::analyze): Bail out early for __RTL functions.
39905         (analyze_functions): Update assertion to support __RTL functions.
39906         (cgraph_node::expand): Bail out early for __RTL functions.
39907         * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39908         __RTL functions.
39909         * function.h (struct function): Update comment for field
39910         "pass_startwith".
39911         * gimple-expr.c: Include "tree-pass.h".
39912         (gimple_has_body_p): Return false for __RTL functions.
39913         * Makefile.in (OBJS): Add run-rtl-passes.o.
39914         * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
39915         accessor.
39916         (gcc::pass_manager::get_clean_slate): New accessor.
39917         * passes.c: Include "insn-addr.h".
39918         (should_skip_pass_p): Add logging.  Update logic for running
39919         "expand" to be compatible with both __GIMPLE and __RTL.  Guard
39920         property-provider override so it is only done for gimple passes.
39921         Don't skip dfinit.
39922         (skip_pass): New function.
39923         (execute_one_pass): Call skip_pass when skipping passes.
39924         * read-md.c (md_reader::read_char): Support filtering
39925         the input to a subset of line numbers.
39926         (md_reader::md_reader): Initialize fields
39927         m_first_line and m_last_line.
39928         (md_reader::read_file_fragment): New function.
39929         * read-md.h (md_reader::read_file_fragment): New decl.
39930         (md_reader::m_first_line): New field.
39931         (md_reader::m_last_line): New field.
39932         * read-rtl-function.c (function_reader::create_function): Only
39933         create cfun if it doesn't already exist.  Set PROP_rtl on cfun's
39934         curr_properties.  Set DECL_INITIAL to a dummy block.
39935         (read_rtl_function_body_from_file_range): New function.
39936         * read-rtl-function.h (read_rtl_function_body_from_file_range):
39937         New decl.
39938         * run-rtl-passes.c: New file.
39939         * run-rtl-passes.h: New file.
39941 2017-01-24  Jeff Law  <law@redhat.com>
39943         * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
39944         buffer size.
39946 2017-01-24  Bin Cheng  <bin.cheng@arm.com>
39948         PR tree-optimization/79159
39949         * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
39950         (record_nonwrapping_iv): Improve boundary using above function if no
39951         value range information.
39953 2017-01-24  Pekka Jääskeläinen  <pekka@parmance.com>
39954             Martin Jambor  <mjambor@suse.cz>
39956         * brig-builtins.def: New file.
39957         * builtins.def (DEF_HSAIL_BUILTIN): New macro.
39958         (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
39959         (DEF_HSAIL_SAT_BUILTIN): Likewise.
39960         (DEF_HSAIL_INTR_BUILTIN): Likewise.
39961         (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
39962         * builtin-types.def (BT_INT8): New.
39963         (BT_INT16): Likewise.
39964         (BT_UINT8): Likewise.
39965         (BT_UINT16): Likewise.
39966         (BT_FN_ULONG): Likewise.
39967         (BT_FN_UINT_INT): Likewise.
39968         (BT_FN_UINT_ULONG): Likewise.
39969         (BT_FN_UINT_LONG): Likewise.
39970         (BT_FN_UINT_PTR): Likewise.
39971         (BT_FN_ULONG_PTR): Likewise.
39972         (BT_FN_INT8_FLOAT): Likewise.
39973         (BT_FN_INT16_FLOAT): Likewise.
39974         (BT_FN_UINT32_FLOAT): Likewise.
39975         (BT_FN_UINT16_FLOAT): Likewise.
39976         (BT_FN_UINT8_FLOAT): Likewise.
39977         (BT_FN_UINT64_FLOAT): Likewise.
39978         (BT_FN_UINT16_UINT32): Likewise.
39979         (BT_FN_UINT32_UINT16): Likewise.
39980         (BT_FN_UINT16_UINT16_UINT16): Likewise.
39981         (BT_FN_INT_PTR_INT): Likewise.
39982         (BT_FN_UINT_PTR_UINT): Likewise.
39983         (BT_FN_LONG_PTR_LONG): Likewise.
39984         (BT_FN_ULONG_PTR_ULONG): Likewise.
39985         (BT_FN_VOID_UINT64_UINT64): Likewise.
39986         (BT_FN_UINT8_UINT8_UINT8): Likewise.
39987         (BT_FN_INT8_INT8_INT8): Likewise.
39988         (BT_FN_INT16_INT16_INT16): Likewise.
39989         (BT_FN_INT_INT_INT): Likewise.
39990         (BT_FN_UINT_FLOAT_UINT): Likewise.
39991         (BT_FN_FLOAT_UINT_UINT): Likewise.
39992         (BT_FN_ULONG_UINT_UINT): Likewise.
39993         (BT_FN_ULONG_UINT_PTR): Likewise.
39994         (BT_FN_ULONG_ULONG_ULONG): Likewise.
39995         (BT_FN_UINT_UINT_UINT): Likewise.
39996         (BT_FN_VOID_UINT_PTR): Likewise.
39997         (BT_FN_UINT_UINT_PTR: Likewise.
39998         (BT_FN_UINT32_UINT64_PTR): Likewise.
39999         (BT_FN_INT_INT_UINT_UINT): Likewise.
40000         (BT_FN_UINT_UINT_UINT_UINT): Likewise.
40001         (BT_FN_UINT_UINT_UINT_PTR): Likewise.
40002         (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
40003         (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
40004         (BT_FN_LONG_LONG_UINT_UINT): Likewise.
40005         (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
40006         (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
40007         (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
40008         (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
40009         (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
40010         (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
40011         * doc/frontends.texi: List BRIG FE.
40012         * doc/install.texi (Testing): Add BRIG tesring requirements.
40013         * doc/invoke.texi (Overall Options): Mention BRIG.
40014         * doc/standards.texi (Standards): Doucment BRIG HSA version.
40016 2017-01-24  Richard Biener  <rguenther@suse.de>
40018         PR translation/79208
40019         * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
40021 2017-01-24  Martin Jambor  <mjambor@suse.cz>
40023         PR bootstrap/79198
40024         * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
40025         * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
40026         and known_contexts.
40028 2017-01-24  Aldy Hernandez  <aldyh@redhat.com>
40030         PR middle-end/79123
40031         * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
40032         casts from signed to unsigned really don't have a range.
40034 2017-01-24  Markus Trippelsdorf  <markus@trippelsdorf.de>
40036         * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
40037         GMP_RNDx for compatiblity.
40039 2017-01-24  Martin Liska  <mliska@suse.cz>
40041         PR bootstrap/79132
40042         * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
40043         that would prevent us to call alloca with -1 as argument.
40045 2017-01-24  Jakub Jelinek  <jakub@redhat.com>
40047         * dwarf2out.c (output_compilation_unit_header, output_file_names):
40048         Avoid -Wformat-security warning.
40050 2017-01-23  Andrew Pinski  <apinski@cavium.com>
40052         * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
40053         cost table.
40055 2017-01-23  Martin Sebor  <msebor@redhat.com>
40057         PR middle-end/78703
40058         * gimple-ssa-sprintf.c (warn_level): New global.
40059         (format_integer): Use it here and throughout the rest of the file.
40060         Use the same switch to compute sign as base.
40061         (maybe_warn): New function.
40062         (format_directive): Factor out warnings into maybe_warn.
40063         Add debugging output.  Use warn_level.
40064         (add_bytes): Use warn_level.
40065         (pass_sprintf_length::compute_format_length): Add debugging output.
40066         (try_substitute_return_value): Same.
40067         (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
40069         PR middle-end/78703
40070         * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
40071         (struct fmtresult, format_integer, format_floating): Adjust.
40072         (fmtresult::fmtresult): Set max correctly in two argument ctor.
40073         (get_string_length, format_string,format_directive): Same.
40074         (pass_sprintf_length::compute_format_length): Same.
40075         (try_substitute_return_value): Simplify slightly.
40077         PR middle-end/78703
40078         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
40079         (fmtresult::operator+=): Outlined.
40080         (struct fmtresult): Add ctors.
40081         (struct conversion_spec): Rename...
40082         (struct directive): ...to this.  Add and remove data members.
40083         (directive::set_width, directive::set_precision): New functions.
40084         (format_percent): Use fmtresult ctor.
40085         (get_width_and_precision): Remove.
40086         (format_integer): Make naming changes.  Avoid computing width and
40087         precision.
40088         (format_floating): Same.  Adjust indentation.
40089         (format_character, format_none): New functions.
40090         (format_string): Moved character handling to format_character.
40091         (format_directive): Remove arguments, change return type.
40092         (parse_directive): New function.
40093         (pass_sprintf_length::compute_format_length): Move directive
40094         parsing to parse_directive.
40096 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40098         * tree.h (assign_assembler_name_if_neeeded): Rename to ...
40099         (assign_assembler_name_if_needed): ... this.
40100         * tree.c (assign_assembler_name_if_neeeded): Rename to ...
40101         (assign_assembler_name_if_needed): ... this.
40102         (free_lang_data_in_cgraph): Adjust callers.
40103         * cgraphunit.c (cgraph_node::analyze): Likewise.
40104         * omp-expand.c (expand_omp_taskreg, expand_omp_target):
40105         Likewise.
40107 2017-01-23  Richard Biener  <rguenther@suse.de>
40109         PR tree-optimization/79088
40110         PR tree-optimization/79188
40111         * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
40112         resetting loop bounds after last path deletion.  Reset loop
40113         bounds of the target loop, make code match the comments.
40114         * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
40115         Make sure loops need no fixups.
40117 2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
40119         * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
40120         exponent support with double type for first argument.
40121         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
40122         type returned by __builtin_vec_extract_sig,
40123         __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
40124         functions from "vector int" to "vector unsigned int" or from
40125         "vector long long int" to "vector unsigned long long int".
40126         Changed type returned by __builtin_vec_extract_exp,
40127         __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
40128         functions from "vector int" to "vector unsigned int" or from
40129         "vector long long int" to "vector unsigned long long int".
40130         Changed return type of __builtin_vec_test_data_class,
40131         __builtin_vec_test_data_class_sp, and
40132         __builtin_vec_test_data_class_dp from "vector int" to
40133         "vector bool int" or from "vector long long int" to "vector bool
40134         long long int" and changed second argument type from "unsigned
40135         int" to "int".  Added new overloaded function forms "vector float
40136         __builtin_vec_insert_exp (vector float, vector unsigned int)" and
40137         "vector float __builtin_vec_insert_exp_sp (vector float, vector
40138         unsigned int)" and "vector double __builtin_vec_insert_exp (vector
40139         double, vector unsigned long long int)" and "vector double
40140         __builtin_vec_insert_exp_dp (vector double, vector unsigned long
40141         long int)".  Changed return type of
40142         __builtin_scalar_test_data_class and
40143         __builtin_scalar_test_data_class_sp and
40144         __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
40145         int" and changed second argument from "unsigned int" to "int".
40146         Changed type returned by __builtin_scalar_test_neg,
40147         __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
40148         from "int" to "bool int".  Added new overloaded function form
40149         "double __builtin_scalar_insert_exp (double, unsigned long long int)".
40150         * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
40151         exponent double-precision with floating point first argument.
40152         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
40153         documentation of scalar_test_data_class, scalar_test_neg,
40154         scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
40155         vector_extract_exp, vec_extract_sig, vec_insert_exp, and
40156         vec_test_data_class built-in functions to reflect refinements in
40157         their type signatures.
40159 2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>
40161         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
40162         size of buf.
40163         (aarch64_elf_asm_destructor): Likewise.
40165 2017-01-23  Bernd Schmidt  <bschmidt@redhat.com>
40167         PR rtl-optimization/78634
40168         * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
40169         (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
40170         * ifcvt.c (noce_try_cmove): Add missing cost check.
40172         PR rtl-optimization/71724
40173         * combine.c (if_then_else_cond): Look for situations where it is
40174         beneficial to undo the work of one of the recursive calls.
40176 2017-01-23  Bin Cheng  <bin.cheng@arm.com>
40178         PR tree-optimization/70754
40179         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
40180         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
40181         combined stmt before it if not NULL.
40182         (combine_chains): Process refs reversely and compute dominance point
40183         for root ref.
40185 2017-01-23  Martin Liska  <mliska@suse.cz>
40187         PR tree-optimization/79196
40188         * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
40189         (fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
40190         instead of memcmp.
40191         (strlen_optimize_stmt): Call the renamed function.
40193 2017-01-23  Michael Matz  <matz@suse.de>
40195         PR tree-optimization/78384
40196         * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
40198 2017-01-23  Richard Biener  <rguenther@suse.de>
40200         PR tree-optimization/79186
40201         * tree-vrp.c (register_new_assert_for): Make sure we've seen
40202         both incoming edges before moving an assert.
40204 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40206         * ipa-prop.c (load_from_param_1): Removed.
40207         (load_from_unmodified_param): Bits from load_from_param_1 put back
40208         here.
40209         (load_from_param): Removed.
40210         (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
40211         with stmt.  Reverted back to use of load_from_unmodified_param.
40213 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40215         PR ipa/79108
40216         * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
40217         (ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
40218         field a pointer to garbage collected vector, mark lattices and
40219         ipcp_orig_node with GTY((skip)).
40220         (ipa_get_param_count): Adjust to descriptors being a pointer.
40221         (ipa_get_param): Likewise.
40222         (ipa_get_type): Likewise.
40223         (ipa_get_param_move_cost): Likewise.
40224         (ipa_set_param_used): Likewise.
40225         (ipa_get_controlled_uses): Likewise.
40226         (ipa_set_controlled_uses): Likewise.
40227         (ipa_is_param_used): Likewise.
40228         (ipa_node_params_t): Move into garbage collector.  New methods insert
40229         and remove.
40230         (ipa_node_params_sum): Annotate wth GTY(()).
40231         (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
40232         garbage collected.
40233         (ipa_load_from_parm_agg): Adjust declaration.
40234         * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
40235         * ipa-profile.c (ipa_profile): Likewise.
40236         * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
40237         (ipa_populate_param_decls): Make descriptors parameter garbage
40238         collected.
40239         (ipa_dump_param): Adjust to descriptors being a pointer.
40240         (ipa_alloc_node_params): Likewise.
40241         (ipa_initialize_node_params): Likewise.
40242         (load_from_param_1): Make descriptors parameter garbage collected.
40243         (load_from_unmodified_param): Likewise.
40244         (load_from_param): Likewise.
40245         (ipa_load_from_parm_agg): Likewise.
40246         (ipa_node_params::~ipa_node_params): Removed.
40247         (ipa_free_all_node_params): Remove call to delete operator.
40248         (ipa_node_params_t::insert): New.
40249         (ipa_node_params_t::remove): Likewise.
40250         (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40251         copy known_csts and known_contexts vectors.
40252         (ipa_read_node_info): Adjust to descriptors being a pointer.
40253         (ipcp_modif_dom_walker): Make m_descriptors field garbage
40254         collected.
40255         (ipcp_transform_function): Make descriptors variable garbage
40256         collected.
40258 2017-01-23  Andrew Senkevich  <andrew.senkevich@intel.com>
40260         * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40261         * config/i386/avx512dqintrin.h: Ditto.
40262         * config/i386/avx512fintrin.h: Ditto.
40263         * config/i386/i386.c: Handle new builtins.
40264         * config/i386/i386-builtin.def: Add new builtins.
40265         * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40266         (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40268 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40269             Martin Liska  <mliska@suse.cz>
40271         * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40272         * asan.c (asan_expand_poison_ifn): Support stores and use
40273         appropriate ASAN report function.
40274         * internal-fn.c (expand_ASAN_POISON_USE): New function.
40275         * internal-fn.def (ASAN_POISON_USE): Declare.
40276         * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40277         (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40278         * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40279         ASAN_POISON calls w/o LHS.
40280         * tree-ssa.c (execute_update_addresses_taken): Create clobber
40281         for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40282         from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40283         * gimplify.c (asan_poison_variables): Add attribute
40284         use_after_scope_memory to variables that really needs to live
40285         in memory.
40286         * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40287         having the attribute.
40289 2017-01-23  Martin Liska  <mliska@suse.cz>
40291         * asan.c (create_asan_shadow_var): New function.
40292         (asan_expand_poison_ifn): Likewise.
40293         * asan.h (asan_expand_poison_ifn): New declaration.
40294         * internal-fn.c (expand_ASAN_POISON): Likewise.
40295         * internal-fn.def (ASAN_POISON): New builtin.
40296         * sanopt.c (pass_sanopt::execute): Expand
40297         asan_expand_poison_ifn.
40298         * tree-inline.c (copy_decl_for_dup_finish): Make function
40299         external.
40300         * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40301         * tree-ssa.c (is_asan_mark_p): New function.
40302         (execute_update_addresses_taken): Rewrite local variables
40303         (identified just by use-after-scope as addressable) into SSA.
40305 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40307         * doc/install.texi (Specific): opensource.apple.com uses https
40308         now. Remove trailing slash.
40310 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40312         * README.Portability: Remove note on an Irix compatibility issue.
40314 2017-01-22  Dimitry Andric <dim@FreeBSD.org>
40316         * gcov.c (INCLUDE_ALGORITHM): Define.
40317         (INCLUDE_VECTOR): Define.
40318         No longer include <vector> and <algorithm> directly.
40320 2017-01-21  Gerald Pfeifer  <gerald@pfeifer.com>
40322         * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40323         to https.
40324         * doc/invoke.texi (Code Gen Options): Ditto.
40326 2017-01-21  Jan Hubicka  <hubicka@ucw.cz>
40328         PR lto/78407
40329         * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40331 2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>
40333         rtl-optimization/79125
40334         * cprop.c (local_cprop_pass): Handle cases where we make an
40335         unconditional trap.
40337 2017-01-20  Segher Boessenkool  <segher@kernel.crashing.org>
40339         PR target/61729
40340         PR target/77850
40341         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40342         read from, for big endian.
40344 2017-01-20  Jiong Wang  <jiong.wang@arm.com>
40346         * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40347         register pauth builtins for LP64 only.
40349 2017-01-20  Marek Polacek  <polacek@redhat.com>
40351         PR c/79152
40352         * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40353         non-case labels.
40355 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40357         * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40358         of safelen status.
40359         * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40360         * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40361         * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40363 2017-01-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40365         PR target/71270
40366         * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40367         in big-endian mode when they are not a single duplicated value.
40369 2017-01-20  Richard Biener  <rguenther@suse.de>
40371         * BASE-VER: Bump to 7.0.1.
40373 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40375         * omp-low.c (omplow_simd_context): New struct.  Use it...
40376         (lower_rec_simd_input_clauses): ...here and...
40377         (lower_rec_input_clauses): ...here to hold common data.  Adjust all
40378         references to idx, lane, max_vf, is_simt.
40380 2017-01-20  Graham Markall  <graham.markall@embecosm.com>
40382         * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40383         mcpu=nps400.
40385 2017-01-20  Martin Jambor  <mjambor@suse.cz>
40387         * hsa.h: Renaed to hsa-common.h.  Adjusted a comment.
40388         * hsa.c: Renaed to hsa-common.c.  Change include of gt-hsa.h to
40389         gt-hsa-common.h.
40390         * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40391         (GTFILES): Rename hsa.c to hsa-common.c.
40392         * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40393         * hsa-dump.c: Likewise.
40394         * hsa-gen.c: Likewise.
40395         * hsa-regalloc.c: Likewise.
40396         * ipa-hsa.c: Likewise.
40397         * omp-expand.c: Likewise.
40398         * omp-low.c: Likewise.
40399         * toplev.c: Likewise.
40401 2017-01-20  Marek Polacek  <polacek@redhat.com>
40403         PR c/64279
40404         * doc/invoke.texi: Document -Wduplicated-branches.
40405         * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40406         COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40407         POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40408         STATEMENT_LIST, and RETURN_EXPR.  For non-pure non-const functions
40409         return 0 only when not OEP_LEXICOGRAPHIC.
40410         (fold_build_cleanup_point_expr): Use the expression
40411         location when building CLEANUP_POINT_EXPR.
40412         * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40413         * tree.c (add_expr): Handle error_mark_node.
40415 2017-01-20  Martin Liska  <mliska@suse.cz>
40417         PR lto/69188
40418         * tree-profile.c (init_ic_make_global_vars): Do not call
40419         finalize_decl.
40420         (gimple_init_gcov_profiler): Likewise.
40422 2017-01-20  Martin Liska  <mliska@suse.cz>
40424         PR ipa/71190
40425         * cgraph.h (maybe_create_reference): Remove argument and
40426         update comment.
40427         * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40428         argument.
40429         * ipa-cp.c (create_specialized_node): Likewise.
40430         * symtab.c (symtab_node::maybe_create_reference): Handle
40431         VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40433 2017-01-20  Martin Liska  <mliska@suse.cz>
40435         * read-rtl-function.c (function_reader::create_function): Use
40436         build_decl instread of build_decl_stat.
40438 2017-01-20  Andrew Senkevich  <andrew.senkevich@intel.com>
40440         * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40441         * config/i386/avx512dqintrin.h: Ditto.
40442         * config/i386/avx512fintrin.h: Ditto.
40443         * config/i386/i386-builtin-types.def: Add new types.
40444         * config/i386/i386.c: Handle new types.
40445         * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40446         (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40447         (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40448         (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40449         (__builtin_ia32_kshiftridi): New.
40450         * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40452 2017-01-19  Segher Boessenkool  <segher@kernel.crashing.org>
40454         PR target/78875
40455         PR target/79140
40456         * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40457         define to rs6000_init_stack_protect_guard.
40458         (rs6000_init_stack_protect_guard): New function.
40460 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40461             Yunqiang Su  <yunqiang.su@imgtec.com>
40463         * config.gcc (supported_defaults): Add madd4.
40464         (with_madd4): Add validation.
40465         (all_defaults): Add madd4.
40466         * config/mips/mips.opt (mmadd4): New option.
40467         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40468         mmadd4.
40469         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40470         __mips_no_madd4.
40471         (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40472         (ISA_HAS_FUSED_MADD4): Likewise.
40473         * doc/invoke.texi (-mmadd4): Document the new option.
40474         * doc/install.texi (--with-madd4): Document the new option.
40476 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40478         * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40479         entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40480         AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40481         (aarch64_init_pauth_hint_builtins): New.
40482         (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40483         (aarch64_expand_builtin): Expand new builtins.
40485 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40487         * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40488         * combine-stack-adj.c (no_unhandled_cfa): Handle
40489         REG_CFA_TOGGLE_RA_MANGLE.
40490         * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40491         * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40492         info for return address signing.
40493         (aarch64_expand_epilogue): Likewise.
40495 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40497         * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40498         * config/aarch64/aarch64-protos.h
40499         (aarch64_return_address_signing_enabled): New declaration.
40500         * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40501         New function.
40502         (aarch64_expand_prologue): Sign return address before it's pushed onto
40503         stack.
40504         (aarch64_expand_epilogue): Authenticate return address fetched from
40505         stack.
40506         (aarch64_override_options): Sanity check for ILP32 and ISA level.
40507         (aarch64_attributes): New function attributes for "sign-return-address".
40508         * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40509         UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40510         ("*do_return"): Generate combined instructions according to key index.
40511         ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40512         * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40513         iterators.
40514         (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40515         * config/aarch64/aarch64.opt (msign-return-address=): New.
40516         * doc/extend.texi (AArch64 Function Attributes): Documents
40517         "sign-return-address=".
40518         * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40520 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40522         * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40523         overall option summary.
40525 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40527         * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40528         * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40529         AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40530         * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40532 2017-01-19  Michael Meissner  <meissner@linux.vnet.ibm.com>
40534         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40535         -mpower9-minmax by default for -mcpu=power9.
40536         (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40537         128-bit floating point.
40539 2017-01-20  Alan Modra  <amodra@gmail.com>
40541         * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40542         optimizing for size.
40544 2017-01-20  Alan Modra  <amodra@gmail.com>
40546         PR target/79144
40547         * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40548         for strcmp and strncmp from corresponding builtin decl.
40550 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40552         * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40553         instead of i386/rtems-64.h.
40554         * config/i386/rtems-64.h: Remove.
40556 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40558         PR target/78478
40559         Revert:
40560         2013-11-05  Uros Bizjak  <ubizjak@gmail.com>
40562         * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40564 2017-01-19  Tamar Christina  <tamar.christina@arm.com>
40566         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40567         Change int to HOST_WIDE_INT.
40568         * config/aarch64/aarch64-protos.h
40569         (aarch64_simd_gen_const_vector_dup): Likewise.
40570         * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40572 2017-01-19  David Malcolm  <dmalcolm@redhat.com>
40574         * langhooks-def.h (lhd_type_for_size): New decl.
40575         (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40576         * langhooks.c (lhd_type_for_size): New function, taken from
40577         lto_type_for_size.
40579 2017-01-19  Pat Haugen  <pthaugen@us.ibm.com>
40581         * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40582         define_bypass for CR latency.
40583         (power9-cracked-alu): Update bypass latency and remove power9-branch.
40584         (power9-alu2): Add define_bypass for CR latency.
40585         (power9-cmp): New.
40586         (power9-mul): Update insn latency.
40587         (power9-mul-compare): Update insn latency, bypass latency and remove
40588         power9-branch.
40590 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40592         * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40593         Delete.
40594         * config/aarch64/aarch64.md
40595         (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40596         aarch64_nopcrelative_literal_loads.
40597         (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40599 2017-01-19  Chenghua Xu  <paul.hua.gm@gmail.com>
40601         * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40602         TARGET_LOONGSON_3A.
40603         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40605 2017-01-19  Doug Gilmore  <doug.gilmore@imgtec.com>
40607         PR target/78176
40608         * config.gcc (supported_defaults): Add lxc1-sxc1.
40609         (with_lxc1_sxc1): Add validation.
40610         (all_defaults): Add lxc1-sxc1.
40611         * config/mips/mips.opt (mlxc1-sxc1): New option.
40612         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40613         mlxc1-sxc1.
40614         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40615         __mips_no_lxc1_sxc1.
40616         (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40617         * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40618         * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40620 2017-01-19  Richard Biener  <rguenther@suse.de>
40622         PR tree-optimization/72488
40623         * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40624         sure to restore SSA info.
40625         * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40627 2017-01-19  Richard Earnshaw  <rearnsha@arm.com>
40629         PR rtl-optimization/79121
40630         * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40631         of the inner type when shifting an extended value.
40633 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40635         PR lto/78407
40636         * symtab.c (symtab_node::equal_address_to): Fix comparing of
40637         interposable aliases.
40639 2017-01-18  Peter Bergner  <bergner@vnet.ibm.com>
40641         PR target/78516
40642         * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40643         Use the evmergelohi instruction.
40644         (mov_si<mode>_e500_subreg4_2_le): Likewise.
40645         (mov_sitf_e500_subreg8_2_be): Likewise.
40646         (mov_sitf_e500_subreg12_2_le): Likewise.
40647         (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40648         (mov_si<mode>_e500_subreg4_2_be): Likewise.
40649         (mov_sitf_e500_subreg8_2_le): Likewise.
40650         (mov_sitf_e500_subreg12_2_be): Likewise.
40652 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40654         * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40655         attribute from vecsimple to vecperm.
40656         (altivec_vbpermq2): Likewise.
40658 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40660         PR target/79040
40661         * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40663 2017-01-18  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
40664         * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40665         * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40666         strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40667         case where N arg is SIZE_MAX.
40668         * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40669         (cmpstrsi): Add pattern.
40671 2017-01-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
40673         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40674         __builtin_vec_revb builtins.
40675         * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40676         built-in functions to support generation of the ISA 3.0 XXBR<x>
40677         vector byte reverse instructions.
40678         (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40679         (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40680         (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40681         (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40682         (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40683         (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40684         (P9V_BUILTIN_VEC_REVB): Likewise.
40685         * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40686         generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40687         (p9_xxbrq_v16qi): Likewise.
40688         (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40689         (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40690         (p9_xxbrh_v8hi): Likewise.
40691         * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40692         * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40693         vec_revb built-in functions.
40695 2017-01-18  Uros Bizjak  <ubizjak@gmail.com>
40697         PR rtl-optimization/78952
40698         * config/i386/i386.md (any_extract): New code iterator.
40699         (*insvqi_2): Use any_extract for source operand.
40700         (*insvqi_3): Use any_shiftrt for source operand.
40702 2017-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
40704         * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40705         New function.
40706         (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40708 2017-01-18  Matthias Klose  <doko@ubuntu.com>
40710         * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40712 2016-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40714         * config/rs6000/altivec.h (vec_bperm): Change #define.
40715         * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40716         (altivec_vbpermq2): New define_insn.
40717         (altivec_vbpermd): Likewise.
40718         * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40719         function interface.
40720         (VBPERMD): Likewise.
40721         (VBPERM): New polymorphic function interface.
40722         * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40723         Add entries for P9V_BUILTIN_VEC_VBPERM.
40724         * doc/extend.texi: Add interfaces for vec_bperm.
40726 2017-01-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
40728         * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40729         first letter of error messages.
40730         (s390_resolve_overloaded_builtin): Likewise.
40731         * config/s390/s390.c (s390_expand_builtin): Likewise.
40732         (s390_invalid_arg_for_unprototyped_fn): Likewise.
40733         (s390_valid_target_attribute_inner_p): Likewise.
40734         * config/s390/s390.md ("tabort"): Likewise.
40736 2017-01-18  Toma Tabacu  <toma.tabacu@imgtec.com>
40738         * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40739         (ISA_AVOID_DIV_HILO): New macro.
40740         (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40741         (ISA_HAS_DDIV): Likewise.
40743 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40745         * doc/invoke.texi (fabi-version): Correct number of occurrences.
40747 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40749         * doc/invoke.texi (fabi-version): Spelling fix.
40751 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40753         PR c++/70182
40754         * doc/invoke.texi (fabi-version): Mention mangling fix for
40755         operator names.
40757 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40759         PR c++/77489
40760         * doc/invoke.texi (fabi-version): Document discriminator mangling.
40762 2017-01-17  Segher Boessenkool  <segher@kernel.crashing.org>
40764         PR target/78875
40765         * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40766         * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40767         the new options.
40768         * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40769         flexible settings.
40770         (stack_protect_test): Ditto.
40771         * config/rs6000/rs6000.opt (mstack-protector-guard=,
40772         mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40773         options.
40774         * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40775         Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40776         -mstack-protector-guard-offset=.
40777         (RS/6000 and PowerPC Options): Ditto.
40779 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40781         * config/i386/i386.h (MASK_CLASS_P): New define.
40782         * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40783         there are no registers from different register sets also when
40784         mask registers are used.  Update function comment.
40785         * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40786         to (*k/*r) and (*k/*km) alternatives.
40788 2017-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
40790         * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40791         * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40792         (EH_RETURN_HANDLER_RTX): New define.
40793         * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40794         Force frame pointer in EH return functions.
40795         (aarch64_expand_epilogue): Add barrier for eh_return.
40796         (aarch64_final_eh_return_addr): Remove.
40797         (aarch64_eh_return_handler_rtx): New function.
40798         * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40799         Remove.
40800         (aarch64_eh_return_handler_rtx): New prototype.
40802 2017-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40804         * config/rs6000/altivec.h (vec_rlmi): New #define.
40805         (vec_vrlnm): Likewise.
40806         (vec_rlnm): Likewise.
40807         * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40808         (UNSPEC_VRLNM): Likewise.
40809         (VIlong): New mode iterator.
40810         (altivec_vrl<VI_char>mi): New define_insn.
40811         (altivec_vrl<VI_char>nm): Likewise.
40812         * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40813         function entry.
40814         (VRLDNM): Likewise.
40815         (RLNM): New polymorphic function entry.
40816         (VRLWMI): New monomorphic function entry.
40817         (VRLDMI): Likewise.
40818         (RLMI): New polymorphic function entry.
40819         * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40820         new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40821         * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40822         vec_vrlnm.
40824 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40826         PR debug/78839
40827         * dwarf2out.c (field_byte_offset): Restore the
40828         PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40829         and DECL_FIELD_BIT_OFFSET.  Use fold_build2 instead of build2 + fold.
40830         (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40831         of build2 + fold.
40833 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40835         PR ada/67205
40836         * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40838 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40840         PR debug/71669
40841         * dwarf2out.c (add_data_member_location_attribute): For constant
40842         offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40843         instead of DW_AT_data_member_location, DW_AT_bit_offset and
40844         DW_AT_byte_size attributes.
40846 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40848         * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40849         after forcing to constant memory when the code model is medium.
40851 2017-01-17  Julia Koval  <julia.koval@intel.com>
40853         PR target/76731
40854         * config/i386/avx512fintrin.h
40855         (_mm512_i32gather_ps): Change __addr type to void const*.
40856         (_mm512_mask_i32gather_ps): Ditto.
40857         (_mm512_i32gather_pd): Ditto.
40858         (_mm512_mask_i32gather_pd): Ditto.
40859         (_mm512_i64gather_ps): Ditto.
40860         (_mm512_mask_i64gather_ps): Ditto.
40861         (_mm512_i64gather_pd): Ditto.
40862         (_mm512_mask_i64gather_pd): Ditto.
40863         (_mm512_i32gather_epi32): Ditto.
40864         (_mm512_mask_i32gather_epi32): Ditto.
40865         (_mm512_i32gather_epi64): Ditto.
40866         (_mm512_mask_i32gather_epi64): Ditto.
40867         (_mm512_i64gather_epi32): Ditto.
40868         (_mm512_mask_i64gather_epi32): Ditto.
40869         (_mm512_i64gather_epi64): Ditto.
40870         (_mm512_mask_i64gather_epi64): Ditto.
40871         (_mm512_i32scatter_ps): Change __addr type to void*.
40872         (_mm512_mask_i32scatter_ps): Ditto.
40873         (_mm512_i32scatter_pd): Ditto.
40874         (_mm512_mask_i32scatter_pd): Ditto.
40875         (_mm512_i64scatter_ps): Ditto.
40876         (_mm512_mask_i64scatter_ps): Ditto.
40877         (_mm512_i64scatter_pd): Ditto.
40878         (_mm512_mask_i64scatter_pd): Ditto.
40879         (_mm512_i32scatter_epi32): Ditto.
40880         (_mm512_mask_i32scatter_epi32): Ditto.
40881         (_mm512_i32scatter_epi64): Ditto.
40882         (_mm512_mask_i32scatter_epi64): Ditto.
40883         (_mm512_i64scatter_epi32): Ditto.
40884         (_mm512_mask_i64scatter_epi32): Ditto.
40885         (_mm512_i64scatter_epi64): Ditto.
40886         (_mm512_mask_i64scatter_epi64): Ditto.
40887         * config/i386/avx512pfintrin.h
40888         (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40889         (_mm512_mask_prefetch_i32gather_ps): Ditto.
40890         (_mm512_mask_prefetch_i64gather_pd): Ditto.
40891         (_mm512_mask_prefetch_i64gather_ps): Ditto.
40892         (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40893         (_mm512_prefetch_i32scatter_ps): Ditto.
40894         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40895         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40896         (_mm512_prefetch_i64scatter_pd): Ditto.
40897         (_mm512_prefetch_i64scatter_ps): Ditto.
40898         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40899         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40900         * config/i386/avx512vlintrin.h
40901         (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40902         (_mm_mmask_i32gather_ps): Ditto.
40903         (_mm256_mmask_i32gather_pd): Ditto.
40904         (_mm_mmask_i32gather_pd): Ditto.
40905         (_mm256_mmask_i64gather_ps): Ditto.
40906         (_mm_mmask_i64gather_ps): Ditto.
40907         (_mm256_mmask_i64gather_pd): Ditto.
40908         (_mm_mmask_i64gather_pd): Ditto.
40909         (_mm256_mmask_i32gather_epi32): Ditto.
40910         (_mm_mmask_i32gather_epi32): Ditto.
40911         (_mm256_mmask_i32gather_epi64): Ditto.
40912         (_mm_mmask_i32gather_epi64): Ditto.
40913         (_mm256_mmask_i64gather_epi32): Ditto.
40914         (_mm_mmask_i64gather_epi32): Ditto.
40915         (_mm256_mmask_i64gather_epi64): Ditto.
40916         (_mm_mmask_i64gather_epi64): Ditto.
40917         (_mm256_i32scatter_ps): Change __addr type to void*.
40918         (_mm256_mask_i32scatter_ps): Ditto.
40919         (_mm_i32scatter_ps): Ditto.
40920         (_mm_mask_i32scatter_ps): Ditto.
40921         (_mm256_i32scatter_pd): Ditto.
40922         (_mm256_mask_i32scatter_pd): Ditto.
40923         (_mm_i32scatter_pd): Ditto.
40924         (_mm_mask_i32scatter_pd): Ditto.
40925         (_mm256_i64scatter_ps): Ditto.
40926         (_mm256_mask_i64scatter_ps): Ditto.
40927         (_mm_i64scatter_ps): Ditto.
40928         (_mm_mask_i64scatter_ps): Ditto.
40929         (_mm256_i64scatter_pd): Ditto.
40930         (_mm256_mask_i64scatter_pd): Ditto.
40931         (_mm_i64scatter_pd): Ditto.
40932         (_mm_mask_i64scatter_pd): Ditto.
40933         (_mm256_i32scatter_epi32): Ditto.
40934         (_mm256_mask_i32scatter_epi32): Ditto.
40935         (_mm_i32scatter_epi32): Ditto.
40936         (_mm_mask_i32scatter_epi32): Ditto.
40937         (_mm256_i32scatter_epi64): Ditto.
40938         (_mm256_mask_i32scatter_epi64): Ditto.
40939         (_mm_i32scatter_epi64): Ditto.
40940         (_mm_mask_i32scatter_epi64): Ditto.
40941         (_mm256_i64scatter_epi32): Ditto.
40942         (_mm256_mask_i64scatter_epi32): Ditto.
40943         (_mm_i64scatter_epi32): Ditto.
40944         (_mm_mask_i64scatter_epi32): Ditto.
40945         (_mm256_i64scatter_epi64): Ditto.
40946         (_mm256_mask_i64scatter_epi64): Ditto.
40947         (_mm_i64scatter_epi64): Ditto.
40948         (_mm_mask_i64scatter_epi64): Ditto.
40949         * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
40950         (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
40951         (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
40952         (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
40953         (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
40954         (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
40955         (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
40956         (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
40957         (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
40958         (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
40959         (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
40960         (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
40961         (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
40962         (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
40963         (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
40964         (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
40965         (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
40966         (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
40967         (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
40968         (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
40969         (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
40970         (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
40971         (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
40972         (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
40973         (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
40974         (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
40975         (VOID_QI_V8DI_PCINT_INT_INT): Remove.
40976         (V16SF_V16SF_PCVOID_V16SI_HI_INT,  V8DF_V8DF_PCVOID_V8SI_QI_INT)
40977         (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
40978         (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
40979         (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
40980         (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
40981         (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
40982         (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
40983         (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
40984         (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
40985         (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
40986         (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
40987         (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
40988         (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
40989         (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
40990         (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
40991         (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
40992         (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
40993         (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
40994         (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
40995         (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
40996         (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
40997         (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
40998         (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
40999         (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
41000         (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
41001         (VOID_QI_V8DI_PCVOID_INT_INT): Add.
41002         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
41003         definitions accordingly.
41005 2017-01-17  Kito Cheng  <kito.cheng@gmail.com>
41006             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
41008         PR target/79079
41009         * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
41010         gen_lowpart.
41012 2017-01-17  Vladimir Makarov  <vmakarov@redhat.com>
41014         PR target/79058
41015         * ira-conflicts.c (ira_build_conflicts): Update total conflict
41016         hard regs for inner regno.
41018 2017-01-17  Martin Liska  <mliska@suse.cz>
41020         PR ipa/71207
41021         * ipa-polymorphic-call.c (contains_type_p): Fix wrong
41022         assumption and add comment.
41024 2017-01-17  Nathan Sidwell  <nathan@acm.org>
41026         * ipa-visibility.c (localize_node): New function, broken out of ...
41027         (function_and_variable_visibility): ... here. Call it.
41029 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
41031         PR middle-end/77445
41032         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
41033         correctly set frequency of oudgoing edge.
41034         (duplicate_thread_path): Fix profile updating.
41036 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
41038         PR other/79046
41039         * configure.ac: Add GCC_BASE_VER.
41040         * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
41041         version from BASE-VER file.
41042         (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
41043         (gcc.o): Depend on $(BASEVER).
41044         * common.opt (dumpfullversion): New option.
41045         * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
41046         * doc/invoke.texi: Document -dumpfullversion.
41047         * doc/install.texi: Document --with-gcc-major-version-only.
41048         * configure: Regenerated.
41050 2017-01-17  Richard Biener  <rguenther@suse.de>
41052         PR tree-optimization/71433
41053         * tree-vrp.c (register_new_assert_for): Merge same asserts
41054         on all incoming edges.
41055         (process_assert_insertions_for): Handle insertions at the
41056         beginning of BBs.
41058 2017-01-17  Gerald Pfeifer  <gerald@pfeifer.com>
41060         * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
41061         * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
41063 2017-01-17  Kaz Kojima  <kkojima@gcc.gnu.org>
41065         PR target/78633
41066         * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
41067         RTL sharing.
41069 2017-01-17  Alan Modra  <amodra@gmail.com>
41071         PR target/79066
41072         * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
41073         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
41074         symbolic stack limit when pic.
41076 2017-01-16  Martin Sebor  <msebor@redhat.com>
41078         PR tree-optimization/78608
41079         * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
41081 2017-01-16  Jeff Law  <law@redhat.com>
41083         Revert:
41084         2016-12-02  Tadek Kijkowski  <tkijkowski@gmail.com>
41085         * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
41086         for several include directories that may be relative to sysroot.
41087         * config/i386/x-mingw32 (gplus_includedir): Define.
41088         (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
41089         (native_system_includedir): Likewise.
41090         * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
41091         override if TARGET_SYSTEM_ROOT is defined.
41092         (NATIVE_SYSTEM_HEADER_DIR): Likewise.
41094         PR tree-optimization/79090
41095         PR tree-optimization/33562
41096         PR tree-optimization/61912
41097         PR tree-optimization/77485
41098         * tree-ssa-dse.c (compute_trims): Accept STMT argument.  Dump STMT
41099         and computed trims into the dump file.
41101 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
41103         * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
41105 2017-01-16  Jakub Jelinek  <jakub@redhat.com>
41107         PR c/79089
41108         * gimplify.c (gimplify_init_constructor): If want_value and
41109         object == lhs, unshare lhs to avoid invalid tree sharing.  Formatting
41110         fix.
41112         PR target/79080
41113         * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
41114         sequence.  Formatting fixes.
41115         (doloop_optimize): Formatting fixes.
41117         PR driver/49726
41118         * gcc.c (debug_level_greater_than_spec_func): New function.
41119         (static_spec_functions): Add debug-level-gt spec function.
41120         (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
41121         !g0.
41122         * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41123         * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41124         * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
41125         gpubnames, ggnu-pubnames, gno-record-gcc-switches,
41126         grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
41127         gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
41129 2017-01-16  Uros Bizjak  <ubizjak@gmail.com>
41131         * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
41132         QImode fixups to general and mask registers only.
41134 2017-01-16  Carl Love  <cel@us.ibm.com>
41136         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
41137         for built-in functions
41138         vector signed char vec_nabs (vector signed char)
41139         vector signed short vec_nabs (vector signed short)
41140         vector signed int vec_nabs (vector signed int)
41141         vector signed long long vec_nabs (vector signed long long)
41142         vector float vec_nabs (vector float)
41143         vector double vec_nabs (vector double)
41144         * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
41145         and NABS overload.
41146         * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
41147         * config/rs6000/altivec.h: New define for vec_nabs built-in function.
41148         * doc/extend.texi: Update the documentation file for the new built-in
41149         functions.
41151 2017-01-16  Martin Sebor  <msebor@redhat.com>
41153         * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
41154         message.
41156 2017-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41158         * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
41159         UNSPEC_VSX__XXSPLTD to require special splat handling.
41161 2017-01-16  David Malcolm  <dmalcolm@redhat.com>
41163         PR bootstrap/78616
41164         * system.h: Poison strndup.
41166 2017-01-16  Alan Modra  <amodra@gmail.com>
41168         PR target/79098
41169         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
41170         use a switch.
41172 2017-01-16  Georg-Johann Lay  <avr@gjlay.de>
41174         * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
41176 2017-01-15  Uros Bizjak  <ubizjak@gmail.com>
41178         * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
41179         call recog here.  Assert that INSN_CODE (insn) is non-negative.
41181 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41183         PR target/72749
41184         * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
41185         fallthrough.
41186         * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
41187         in the currently scheduled RTL fragment.
41189 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41191         PR rtl-optimization/78751
41192         * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
41193         give up.
41195 2017-01-14  Jeff Law  <law@redhat.com>
41197         PR tree-optimization/79090
41198         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
41199         variable length stores.
41200         (compute_trims): Delete dead assignment to *trim_tail.
41201         (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
41202         zero length.
41204 2017-01-14  Bernd Schmidt  <bschmidt@redhat.com>
41206         PR rtl-optimization/78626
41207         PR rtl-optimization/78727
41208         * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
41209         of a block, and split such blocks after everything else is finished.
41211 2017-01-14  Alan Modra  <amodra@gmail.com>
41213         PR target/72749
41214         * combine.c (recog_for_combine_1): Set INSN_CODE before calling
41215         target legitimate_combined_insn.
41216         * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
41217         (rs6000_legitimate_combined_insn): New function.
41218         * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
41219         all uses.
41220         (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
41221         (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
41222         (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
41224 2017-01-14  Gerald Pfeifer  <gerald@pfeifer.com>
41226         * doc/frontends.texi (G++ and GCC): Remove references to Java.
41228 2017-01-13  Jeff Law  <law@redhat.com>
41230         PR tree-optimization/33562
41231         PR tree-optimization/61912
41232         PR tree-optimization/77485
41233         * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
41234         a statement.
41235         (delete_dead_assignment): Likewise.
41236         (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
41237         statement to delete_dead_call and delete_dead_assignment.
41239 2017-01-13  David Malcolm  <dmalcolm@redhat.com>
41241         PR c/78304
41242         * substring-locations.c (format_warning_va): Strengthen case 1 so
41243         that both endpoints of the substring must be within the format
41244         range for just the substring to be printed.
41246 2017-01-13  Uros Bizjak  <ubizjak@gmail.com>
41248         * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41249         * config/i386/i386.c (ix86_target_string): Add missing options
41250         to isa_opts and reorder options by implied ISAs.  Rename isa_opts2 to
41251         isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41252         flags_other and ix86_target_other to flags2_other.  Display unknown
41253         isa2 options.
41254         (ix86_valid_target_attribute_inner_p): Add missing options and
41255         reorder options by implied ISAs, as in ix86_target_string.
41257 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41259         * hash-table.h (hash_table::too_empty_p): New function.
41260         (hash_table::expand): Use it.
41261         (hash_table::traverse): Likewise.
41262         (hash_table::empty_slot): Use sizeof (value_type) instead of
41263         sizeof (PTR) to convert bytes to elements.  Shrink the table
41264         if the current size is excessive for the current number of
41265         elements.
41267 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41269         * ira-costs.c (record_reg_classes): Break from the inner loop
41270         early once alt_fail is known to be true.  Update outer loop
41271         handling accordingly.
41273 2017-01-13  Jeff Law  <law@redhat.com>
41275         * tree-ssa-dse.c (decrement_count): New function.
41276         (increment_start_addr, maybe_trim_memstar_call): Likewise.
41277         (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41278         when we know the partially dead statement is a mem* function.
41280         PR tree-optimization/61912
41281         PR tree-optimization/77485
41282         * tree-ssa-dse.c: Include expr.h.
41283         (maybe_trim_constructor_store): New function.
41284         (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41286         PR tree-optimization/33562
41287         PR tree-optimization/61912
41288         PR tree-optimization/77485
41289         * doc/invoke.texi: Document new dse-max-object-size param.
41290         * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41291         * tree-ssa-dse.c: Include params.h.
41292         (dse_store_status): New enum.
41293         (initialize_ao_ref_for_dse): New, partially extracted from
41294         dse_optimize_stmt.
41295         (valid_ao_ref_for_dse, normalize_ref): New.
41296         (setup_live_bytes_from_ref, compute_trims): Likewise.
41297         (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41298         (maybe_trim_partially_dead_store): Likewise.
41299         (maybe_trim_complex_store): Likewise.
41300         (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41301         Track what bytes live from the original store.  Return tri-state
41302         for dead, partially dead or live.
41303         (dse_dom_walker): Add constructor, destructor and new private members.
41304         (delete_dead_call, delete_dead_assignment): New extracted from
41305         dse_optimize_stmt.
41306         (dse_optimize_stmt): Make a member of dse_dom_walker.
41307         Use initialize_ao_ref_for_dse.
41309         PR tree-optimization/33562
41310         PR tree-optimization/61912
41311         PR tree-optimization/77485
41312         * sbitmap.h (bitmap_count_bits): Prototype.
41313         (bitmap_clear_range, bitmap_set_range): Likewise.
41314         * sbitmap.c (bitmap_clear_range): New function.
41315         (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41317 2017-01-13  Martin Liska  <mliska@suse.cz>
41319         PR ipa/79043
41320         * function.c (set_cfun): Add new argument force.
41321         * function.h (set_cfun): Likewise.
41322         * ipa-inline-transform.c (inline_call): Use the function when
41323         strict alising from is dropped for function we inline to.
41325 2017-01-13  Richard Biener  <rguenther@suse.de>
41327         * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41328         for dumping GIMPLE INTEGER_CSTs.
41330 2017-01-13  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
41332         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41333         to 201112L since C++17.
41335 2017-01-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
41337         PR sanitizer/78887
41338         * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41339         if -fsanitize=kernel-address is present.
41341 2017-01-13  Richard Biener  <rguenther@suse.de>
41343         * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41344         as _Literal ( type ) number in case usual suffixes do not
41345         preserve all information.
41347 2017-01-13  Richard Biener  <rguenther@suse.de>
41349         PR tree-optimization/77283
41350         * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41351         and ssa-iterators.h.
41352         (is_feasible_trace): Implement a cost model based on joiner
41353         PHI node uses.
41355 2017-01-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
41357         PR target/79004
41358         * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41359         char or short to __float128/_Float128 directly.
41361 2017-01-12  Martin Sebor  <msebor@redhat.com>
41363         to -Wformat-overflow.
41364         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41365         (min_bytes_remaining): Same.
41366         (get_string_length): Same.
41367         (format_string): Same.
41368         (format_directive): Same.
41369         (add_bytes): Same.
41370         (pass_sprintf_length::handle_gimple_call): Same.
41372 2017-01-12  Jakub Jelinek  <jakub@redhat.com>
41374         * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41375         info.nowrite calls with no lhs that can't throw.  Return bool
41376         whether gsi_remove has been called or not.
41377         (pass_sprintf_length::handle_gimple_call): Return bool whether
41378         try_substitute_return_value called gsi_remove.  Formatting fix.
41379         (pass_sprintf_length::execute): Don't use gsi_remove if
41380         handle_gimple_call returned true.
41382         PR bootstrap/79069
41383         * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41384         be removed due to side-effects, don't remove following barrier nor
41385         turn the successor edge into fallthru edge.
41387 2017-01-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41389         PR target/79044
41390         * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41391         element-reversing loads and stores as not swappable.
41393 2017-01-12  Nathan Sidwell  <nathan@acm.org>
41394             Nicolai Stange  <nicstange@gmail.com>
41396         * combine.c (try_combine): Don't ignore result of overlap checking
41397         loop.  Combine overlap & asm check into single loop.
41399 2017-01-12  Richard Biener  <rguenther@suse.de>
41401         * tree-pretty-print.c (dump_generic_node): Provide -gimple
41402         variant for MEM_REF.  Sanitize INTEGER_CST for -gimple.
41404 2017-01-12  Richard Biener  <rguenther@suse.de>
41406         * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41407         and TS_TARGET_OPTION directly derive from TS_BASE.
41408         * tree-core.h (tree_optimization_option): Derive from tree_base.
41409         (tree_target_option): Likewise.
41411 2017-01-11  Uros Bizjak  <ubizjak@gmail.com>
41413         * config/i386/i386.c (memory_address_length): Increase len
41414         only when rip_relative_addr_p returns false.
41416 2017-01-11  Julia Koval  <julia.koval@intel.com>
41418         * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41419         (OPTION_MASK_ISA_SGX_SET): New.
41420         (ix86_handle_option): Handle OPT_msgx.
41421         * config.gcc: Added sgxintrin.h.
41422         * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41423         * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41424         * config/i386/i386.c (ix86_target_string): Add -msgx.
41425         (PTA_SGX): New.
41426         (ix86_option_override_internal): Handle new options.
41427         (ix86_valid_target_attribute_inner_p): Add sgx.
41428         * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41429         * config/i386/i386.opt: Add msgx.
41430         * config/i386/sgxintrin.h: New file.
41431         * config/i386/x86intrin.h: Add sgxintrin.h.
41433 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41435         PR c++/71537
41436         * fold-const.c (maybe_nonzero_address): Return 1 for function
41437         local objects.
41438         (tree_single_nonzero_warnv_p): Don't handle function local objects
41439         here.
41441         PR c++/72813
41442         * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41443         of c-header.
41445 2017-01-11  David Malcolm  <dmalcolm@redhat.com>
41447         PR driver/78877
41448         * opts.c: Include "spellcheck.h"
41449         (struct string_fragment): New struct.
41450         (struct edit_distance_traits<const string_fragment &>): New
41451         struct.
41452         (get_closest_sanitizer_option): New function.
41453         (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41455 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41457         * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41458         by 12.
41459         (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41460         DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41461         (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41462         (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41463         for initial die_offset if dwarf_split_debug_info.
41464         (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41465         initial next_die_offset if dwo_id is non-NULL.  Don't emit padding
41466         fields.
41467         (output_skeleton_debug_sections): Formatting fix.  Use
41468         DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41469         DWARF_COMPILE_UNIT_HEADER_SIZE.  Don't emit padding.
41471 2017-01-11  Wilco Dijkstra  <wdijkstr@arm.com>
41473         * config/arm/cortex-a53.md: Add bypasses for
41474         cortex_a53_r2f_cvt.
41475         (cortex_a53_r2f): Only use for transfers.
41476         (cortex_a53_f2r): Likewise.
41477         (cortex_a53_r2f_cvt): Add reservation for conversions.
41478         (cortex_a53_f2r_cvt): Likewise.
41480 2017-01-11  Tamar Christina  <tamar.christina@arm.com>
41482         * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41483         to all inlined functions, change static to extern.
41485 2017-01-11  Christophe Lyon  <christophe.lyon@linaro.org>
41487         PR target/78253
41488         * config/arm/arm.c (legitimize_pic_address): Handle reference to
41489         weak symbol.
41490         (arm_assemble_integer): Likewise.
41492 2017-01-11  Richard Earnshaw  <rearnsha@arm.com>
41494         * config.gcc: Use new awk script to check CPU, FPU and architecture
41495         parameters for --with-... options.
41496         * config/arm/parsecpu.awk: New file
41497         * config/arm/arm-cpus.in: New file.
41498         * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41499         files.
41500         * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41501         files.
41502         * config/arm/t-arm: Update dependency rules.
41503         * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41504         of processing .def files.
41505         * config/arm/genopt.sh: Deleted.
41506         * config/arm/gentune.sh: Deleted.
41507         * config/arm/arm-cores.def: Deleted.
41508         * config/arm/arm-arches.def: Deleted.
41509         * config/arm/arm-fpus.def: Deleted.
41510         * config/arm/arm-tune.md: Regenerated.
41511         * config/arm/arm-tables.opt: Regenerated.
41512         * config/arm/arm-cpu.h: New generated file.
41513         * config/arm/arm-cpu-data.h: New generated file.
41514         * config/arm/arm-cpu-cdata.h: New generated file.
41516 2017-01-11  Maxim Ostapenko  <m.ostapenko@samsung.com>
41518         PR lto/79042
41519         * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41520         bit.
41521         (input_varpool_node): Unpack dynamically_initialized bit.
41523 2017-01-11  Eric Botcazou  <ebotcazou@adacore.com>
41525         PR rtl-optimization/79032
41526         * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41527         the alignment of the adjusted memory reference against that of MODE,
41528         instead of the alignment of the original memory reference.
41530 2017-01-11  Martin Jambor  <mjambor@suse.cz>
41532         * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41533         test.
41534         * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41535         decorated functions.
41537 2017-01-11  Richard Biener  <rguenther@suse.de>
41539         * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41540         set range/nonnull info for PHI results.  Do not set it on
41541         stmts marked for removal.
41543 2017-01-10  Eric Botcazou  <ebotcazou@adacore.com>
41545         * expr.c (store_field): In the bitfield case, fetch the return value
41546         from the registers before applying a single big-endian adjustment.
41547         Always do a final load for a BLKmode value not larger than a word.
41549 2017-01-10  David Malcolm  <dmalcolm@redhat.com>
41551         PR c++/77949
41552         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41553         that we correctly handle column numbers greater than
41554         LINE_MAP_MAX_COLUMN_NUMBER.
41556 2017-01-10  Martin Sebor  <msebor@redhat.com>
41558         PR middle-end/78245
41559         * gimple-ssa-sprintf.c (get_destination_size): Call
41560         {init,fini}object_sizes.
41561         * tree-object-size.c (addr_object_size): Adjust.
41562         (pass_through_call): Adjust.
41563         (pass_object_sizes::execute): Adjust.
41564         * tree-object-size.h (fini_object_sizes): Declare.
41566 2017-01-10  Martin Sebor  <msebor@redhat.com>
41568         PR tree-optimization/78775
41569         * builtins.c (get_size_range): Move...
41570         * calls.c: ...to here.
41571         (alloc_max_size): Accept zero argument.
41572         (operand_signed_p): Remove.
41573         (maybe_warn_alloc_args_overflow): Call get_size_range.
41574         * calls.h (get_size_range): Declare.
41576 2017-01-10  Joe Seymour  <joe.s@somniumtech.com>
41578         * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41579         from TI's devices.csv file as of September 2016.
41580         * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41582 2017-01-10  Sandra Loosemore  <sandra@codesourcery.com>
41584         * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41585         * doc/invoke.texi: Likewise.
41586         * doc/md.texi: Likewise.
41587         * doc/objc.texi: Likewise.
41589 2017-01-10  Joshua Conner  <joshconner@google.com>
41591         * config/arm/fuchsia-elf.h: New file.
41592         * config/fuchsia.h: New file.
41593         * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41594         (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41595         targets.
41596         * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41598 2016-01-10  Richard Biener  <rguenther@suse.de>
41600         PR tree-optimization/79034
41601         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41602         Propagate out degenerate PHIs in the joiner.
41604 2017-01-10  Martin Liska  <mliska@suse.cz>
41606         * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41607         (sort_congruence_classes_by_decl_uid): Likewise.
41608         (sort_congruence_class_groups_by_decl_uid): Likewise.
41609         (sem_item_optimizer::merge_classes): Sort class, groups in these
41610         classes and members in the groups by DECL_UID of declarations.
41611         This would make merge operations stable.
41613 2017-01-10  Martin Liska  <mliska@suse.cz>
41615         * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41616         usage of m_classes_vec.
41617         (sem_item_optimizer::~sem_item_optimizer):  Likewise.
41618         (sem_item_optimizer::get_group_by_hash): Likewise.
41619         (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41620         (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41621         (sem_item_optimizer::verify_classes): Likewise.
41622         (sem_item_optimizer::process_cong_reduction): Likewise.
41623         (sem_item_optimizer::dump_cong_classes): Likewise.
41624         (sem_item_optimizer::merge_classes): Likewise.
41625         * ipa-icf.h (congruence_class_hash): Rename from
41626         congruence_class_group_hash.  Remove declaration of m_classes_vec.
41628 2017-01-10  Andrew Senkevich  <andrew.senkevich@intel.com>
41630         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41631         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41632         * config.gcc: Add avx512vpopcntdqintrin.h.
41633         * config/i386/avx512vpopcntdqintrin.h: New.
41634         * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41635         * config/i386/i386-builtin-types.def: Add new types.
41636         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41637         __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41638         __builtin_ia32_vpopcountq_v8di_mask): New.
41639         * config/i386/i386-c.c (ix86_target_macros_internal): Define
41640         __AVX512VPOPCNTDQ__.
41641         * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41642         (PTA_AVX512VPOPCNTDQ): Define.
41643         * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41644         TARGET_AVX512VPOPCNTDQ_P): Define.
41645         * config/i386/i386.opt: Add mavx512vpopcntdq.
41646         * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41647         * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41649 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41651         PR middle-end/77484
41652         * predict.def (PRED_CALL): Set to 67.
41654 2017-01-09  Eric Botcazou  <ebotcazou@adacore.com>
41656         * expr.c (store_field): In the bitfield case, if the value comes from
41657         a function call and is of an aggregate type returned in registers, do
41658         not modify the field mode; extract the value in all cases if the mode
41659         is BLKmode and the size is not larger than a word.
41661 2017-01-09  Dominique d'Humieres  <dominiq@lps.ens.fr>
41663         PR target/71017
41664         * config/i386/cpuid.h: Fix undefined behavior.
41666 2017-01-04  Jeff Law  <law@redhat.com>
41668         PR tree-optimization/79007
41669         PR tree-optimization/67955
41670         * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41671         conservative for pt.null when flag_non_call_exceptions is on.
41673 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41675         PR translation/79019
41676         PR translation/79020
41677         * params.def (PARAM_INLINE_MIN_SPEEDUP,
41678         PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41679         PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41680         in descriptions.
41681         * config/avr/avr.opt (maccumulate-args): Likewise.
41682         * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41683         * common.opt (freport-bug): Likewise.
41684         * cif-code.def (CIF_FINAL_ERROR): Likewise.
41685         * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41686         * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41687         translatable string.
41688         * config/i386/i386.c (function_value_32): Likewise.
41689         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41690         * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41691         Likewise.
41692         * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41693         * common/config/msp430/msp430-common.c (msp430_handle_option):
41694         Likewise.
41695         * symtab.c (symtab_node::verify_base): Likewise.
41696         * opts.c (set_debug_level): Likewise.
41697         * tree.c (verify_type_variant): Likewise.  Fix typo in comment.
41698         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41699         missing whitespace to translatable strings.
41700         * config/avr/avr.md (bswapsi2): Fix typo in comment.
41701         * config/sh/superh.h: Likewise.
41702         * config/i386/xopintrin.h: Likewise.
41703         * config/i386/znver1.md: Likewise.
41704         * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41705         * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41706         * double-int.h (struct double_int): Likewise.
41707         * double-int.c (div_and_round_double): Likewise.
41708         * wide-int.cc: Likewise.
41709         * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41710         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41711         * cfgcleanup.c (crossjumps_occured): Renamed to ...
41712         (crossjumps_occurred): ... this.
41713         (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41714         Adjust all uses.
41716         PR tree-optimization/78899
41717         * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41718         returning bool return struct loop *, NULL for failure and the new
41719         loop on success.
41720         (versionable_outer_loop_p): Don't version outer loop if it has
41721         dont_vectorized bit set.
41722         (tree_if_conversion): When versioning outer loop, ensure
41723         tree_if_conversion is performed also on the inner loop of the
41724         non-vectorizable outer loop copy.
41725         * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix.  Fold
41726         LOOP_VECTORIZED in inner loop of the scalar outer loop and
41727         prevent vectorization of it.
41728         (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41729         the outer loop vectorization of the non-scalar version is attempted
41730         before vectorization of the inner loop in scalar version.  If
41731         outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41732         vectorization of its inner loop.
41733         * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41734         has 2 inner loops, rename also on edges from bb whose single pred
41735         is outer_loop->header.  Fix typo in function comment.
41737 2017-01-09  Martin Sebor  <msebor@redhat.com>
41739         PR bootstrap/79033
41740         * asan.c (asan_emit_stack_protection): Increase local buffer size
41741         to avoid snprintf truncation warning.
41743 2017-01-09  Andrew Pinski  <apinski@cavium.com>
41745         * config/aarch64/aarch64-cores.def: Add thunderx2t99.  Change vulcan
41746         to reference thunderx2t99 for the tuning structure
41747         * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41748         Rename to ...
41749         (thunderx2t99_extra_costs): This.
41750         * config/aarch64/aarch64-tune.md: Regenerate.
41751         * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41752         (vulcan_addrcost_table): This.
41753         (vulcan_regmove_cost): Rename to ...
41754         (thunderx2t99_regmove_cost): This.
41755         (vulcan_vector_cost): Rename to ...
41756         (thunderx2t99_vector_cost): this.
41757         (vulcan_branch_cost): Rename to ...
41758         (thunderx2t99_branch_cost): This.
41759         (vulcan_tunings): Rename to ...
41760         (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41761         * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41763 2017-01-09  Martin Jambor  <mjambor@suse.cz>
41765         PR ipa/78365
41766         PR ipa/78599
41767         * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41768         * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41769         (propagate_vr_accross_jump_function): Use the above function for all
41770         value range computations for pass-through jump functions and type
41771         converasion from explicit value range values.
41772         (ipcp_propagate_stage): Do not attempt to deduce types of formal
41773         parameters from TYPE_ARG_TYPES.
41774         * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41775         (ipa_write_node_info): Stream type of the actual argument.
41776         (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41778 2017-01-09  Martin Liska  <mliska@suse.cz>
41780         PR pch/78970
41781         * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41782         (lookup_compiler): Do not show error message with have_E.
41784 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41786         PR tree-optimization/78938
41787         * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41788         where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41789         BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41790         {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR.  Formatting
41791         fixes.
41793 2017-01-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
41795         * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41796         is const0_rtx.
41798 2017-01-09  Richard Biener  <rguenther@suse.de>
41800         PR tree-optimization/78997
41801         * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41802         name condition properly.
41804 2017-01-09  Richard Biener  <rguenther@suse.de>
41806         PR debug/79000
41807         * dwarf2out.c (is_cxx): New overload with context.
41808         (is_naming_typedef_decl): Use it.
41810 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41812         * invoke.texi (Option Summary): Correct spacing in option lists
41813         and add line breaks to fix over-long lines.
41815 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41817         PR middle-end/17660
41819         * extend.texi (Common Variable Attributes): Add xref to GCC
41820         Internals manual to explain mode attribute keywords.
41822 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41824         PR other/16519
41825         * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41826         and Preprocessor Options.
41827         (Options for Linking): Document -pthread here....
41828         (RS/6000 and PowerPC Options): ...not here.
41829         (Solaris 2 Options): ...or here.
41830         * doc/cppopts.texi: Document -pthread.
41832 2017-01-08  Martin Sebor  <msebor@redhat.com>
41834         PR middle-end/77708
41835         * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41836         * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41837         New member functions.
41838         (format_directive): Used them.
41839         (add_bytes): Same.
41840         (pass_sprintf_length::handle_gimple_call): Same.
41841         * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41842         to avoid truncation for any argument.
41843         (extract_affine_mul): Same.
41844         * tree.c (get_file_function_name): Same.
41846 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41848         PR middle-end/77484
41849         * predict.def (PRED_INDIR_CALL): Set to 86.
41851 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41853         PR preprocessor/54124
41854         * doc/cppopts.texi: Reformat -d subtable to list the full name
41855         of the options.  Add cross-reference to the docs for the general
41856         compiler -d options.
41857         * doc/invoke.texi (Developer Options): Add cross-reference to the
41858         preprocessor-specific -d option documentation.
41860 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41862         PR preprocessor/13498
41863         * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41864         redudant material, and reflect new command-line options.
41865         (System Headers): Likewise.
41867 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41869         * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41870         -isystem, and -idirafter.  Copy-edit.
41871         * doc/cppopts.texi: Copy-edit.  Remove contradiction about
41872         default for -ftrack-macro-expansion.  Delete obsolete and
41873         badly-formatted implementation details about -fdebug-cpp output.
41874         * doc/cppwarnopts.texi: Copy-edit.
41876 2017-01-07  David Malcolm  <dmalcolm@redhat.com>
41878         PR c++/72803
41879         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41880         that the transition from a max line width >= 1<<10 to narrower
41881         lines works correctly.
41883 2017-01-07  Alexandre Oliva  <aoliva@redhat.com>
41885         * doc/options.texi (PerFunction): New.
41886         * opt-functions.awk (switch_flags): Map both Optimization and
41887         PerFunction to CL_OPTIMIZATION.
41888         * opth-gen.awk: Test for PerFunction flag along with
41889         Optimization.
41890         * optc-save-gen.awk: Likewise.  Introduce var_opt_hash and set
41891         it only when the latter is present.  Skip those that don't in
41892         the hash function generator.
41893         * common.opt (fvar-tracking): Mark as PerFunction instead of
41894         Optimization.
41895         (fvar-tracking-assignments): Likewise.
41896         (fvar-tracking-assignments-toggle): Likewise.
41897         (fvar-tracking-uninit): Likewise.
41899 2017-01-07  Jakub Jelinek  <jakub@redhat.com>
41901         PR translation/79018
41902         * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41903         the and store.
41905 2017-01-06  Mikael Pettersson  <mikpelinux@gmail.com>
41907         PR target/57583
41908         * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41909         * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41910         TARGET_LONG_JUMP_TABLE_OFFSETS.
41911         * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
41912         * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
41913         * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
41914         (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
41915         * config/m68k/m68k.md (tablejump expander): Likewise.
41916         (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
41917         TARGET_LONG_JUMP_TABLE_OFFSETS.
41918         (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
41919         * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
41921 2017-01-06  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
41922             David Holsgrove <david.holsgrove@xilinx.com>
41924         * common/config/microblaze/microblaze-common.c
41925         (TARGET_EXCEPT_UNWIND_INFO): Remove.
41926         * config/microblaze/microblaze-protos.h (microblaze_eh_return):
41927         New prototype.
41928         * config/microblaze/microblaze.c (microblaze_must_save_register)
41929         (microblaze_expand_epilogue, microblaze_return_addr): Handle
41930         calls_eh_return.
41931         (microblaze_eh_return): New function.
41932         * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
41933         (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
41934         (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
41935         * config/microblaze/microblaze.md (eh_return): New pattern.
41937 2017-01-06  Jakub Jelinek  <jakub@redhat.com>
41939         * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
41940         GCC_DIAGNOSTIC_STRINGIFY): Define.
41942         * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
41944 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41946         * config/arm/arm.md (<mcrr>): New.
41947         (<mrrc>): New.
41948         * config/arm/arm.c (arm_arch5te): New.
41949         (arm_option_override): Set arm_arch5te.
41950         (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
41951         and mrrc2.
41952         * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
41953         (arm_mcrr_qualifiers): ... this. New.
41954         (MRRC_QUALIFIERS): Define to...
41955         (arm_mrrc_qualifiers): ... this. New.
41956         * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
41957         __arm_mrrc2): New.
41958         * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
41959         * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
41960         (MRRCI, mrrc, MRRC): New.
41961         * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
41962         VUNSPEC_MRRC2): New.
41964 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41966         * config/arm/arm.md (<mcr>): New.
41967         (<mrc>): New.
41968         * config/arm/arm.c (arm_coproc_builtin_available): Add
41969         support for mcr, mrc, mcr2 and mrc2.
41970         * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
41971         (arm_mcr_qualifiers): ... this. New.
41972         (MRC_QUALIFIERS): Define to ...
41973         (arm_mrc_qualifiers): ... this. New.
41974         (MCR_QUALIFIERS): Define to ...
41975         (arm_mcr_qualifiers): ... this. New.
41976         * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
41977         __arm_mrc2): New.
41978         * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
41979         * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
41980         * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
41981         VUNSPEC_MRC2): New.
41983 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41985         * config/arm/arm.md (*ldc): New.
41986         (*stc): New.
41987         (<ldc>): New.
41988         (<stc>): New.
41989         * config/arm/arm.c (arm_coproc_builtin_available): Add
41990         support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
41991         (arm_coproc_ldc_stc_legitimate_address): New.
41992         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
41993         'qualifier_const_pointer'.
41994         (LDC_QUALIFIERS): Define to...
41995         (arm_ldc_qualifiers): ... this. New.
41996         (STC_QUALIFIERS): Define to...
41997         (arm_stc_qualifiers): ... this. New.
41998         * config/arm/arm-protos.h
41999         (arm_coproc_ldc_stc_legitimate_address): New.
42000         * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
42001         __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
42002         * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
42003         stc2, stcl, stc2l): New.
42004         * config/arm/constraints.md (Uz): New.
42005         * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
42006         * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
42007         VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
42008         VUNSPEC_STC2L): New.
42010 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42012         * config/arm/arm.md (<cdp>): New.
42013         * config/arm/arm.c (neon_const_bounds): Rename this ...
42014         (arm_const_bounds): ... this.
42015         (arm_coproc_builtin_available): New.
42016         * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
42017         (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
42018         (CDP_QUALIFIERS): Define to...
42019         (arm_cdp_qualifiers): ... this. New.
42020         (void_UP): Define.
42021         (arm_expand_builtin_args): Add case for 6 arguments.
42022         * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
42023         (arm_const_bounds): ... this.
42024         (arm_coproc_builtin_available): New.
42025         * config/arm/arm_acle.h (__arm_cdp): New.
42026         (__arm_cdp2): New.
42027         * config/arm/arm_acle_builtins.def (cdp): New.
42028         (cdp2): New.
42029         * config/arm/iterators.md (CDPI,CDP,cdp): New.
42030         * config/arm/neon.md: Rename all 'neon_const_bounds' to
42031         'arm_const_bounds'.
42032         * config/arm/types.md (coproc): New.
42033         * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
42034         * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
42035         * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
42036         arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
42038 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42040         * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
42041         (UBINOP_QUALIFIERS): New.
42042         (si_UP): Define.
42043         (acle_builtin_data): New. Change comment.
42044         (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
42045         ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
42046         ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
42047         arm_acle_builtins.def.
42048         (ARM_BUILTIN_ACLE_PATTERN_START): Define.
42049         (arm_init_acle_builtins): New.
42050         (CRC32_BUILTIN): Remove.
42051         (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
42052         crc32cb, crc32ch and crc32cw.
42053         (arm_init_crc32_builtins): Remove.
42054         (arm_init_builtins): Use arm_init_acle_builtins rather
42055         than arm_init_crc32_builtins.
42056         (arm_expand_acle_builtin): New.
42057         (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
42058         * config/arm/arm_acle_builtins.def: New.
42060 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42062         * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
42063         (arm_builtin_datum): ... this.
42064         (arm_init_neon_builtin): Rename to ...
42065         (arm_init_builtin): ... this. Add a new parameters PREFIX
42066         and USE_SIG_IN_NAME.
42067         (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
42068         'arm_init_builtin'. Replace type 'neon_builtin_datum' with
42069         'arm_builtin_datum'.
42070         (arm_init_vfp_builtins): Likewise.
42071         (builtin_arg): Rename enum's replacing 'NEON_ARG' with
42072         'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
42073         (arm_expand_neon_args): Rename to ...
42074         (arm_expand_builtin_args): ... this. Rename builtin_arg
42075         enum values and differentiate between ARG_BUILTIN_MEMORY
42076         and ARG_BUILTIN_NEON_MEMORY.
42077         (arm_expand_neon_builtin_1): Rename to ...
42078         (arm_expand_builtin_1): ... this. Rename builtin_arg enum
42079         values, arm_expand_builtin_args and add bool parameter NEON.
42080         (arm_expand_neon_builtin): Use arm_expand_builtin_1.
42081         (arm_expand_vfp_builtin): Likewise.
42082         (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
42084 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42086         PR middle-end/77484
42087         * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
42088         * predict.c (tree_estimate_probability_bb): Reverse direction of
42089         polymorphic call predictor.
42091 2017-01-06  David Malcolm  <dmalcolm@redhat.com>
42093         * passes.c (execute_one_pass): Split out pass-skipping logic into...
42094         (determine_pass_name_match): ...this new function and...
42095         (should_skip_pass_p): ...this new function.
42097 2017-01-06  Nathan Sidwell  <nathan@acm.org>
42099         * ipa-visibility.c (function_and_variable_visibility): Reformat
42100         comments and long lines.  Remove extrneous if.
42101         * symtab.c (symtab_node::make_decl_local): Fix code format.
42102         (symtab_node::set_section_for_node): Fix comment typo.
42104 2017-01-06  Martin Liska  <mliska@suse.cz>
42106         PR bootstrap/79003
42107         * lra-constraints.c: Rename invariant to lra_invariant.
42108         * predict.c (set_even_probabilities): Initialize e to NULL.
42110 2017-01-05  Martin Sebor  <msebor@redhat.com>
42112         PR tree-optimization/78910
42113         * gimple-ssa-sprintf.c (tree_digits): Add an argument.
42114         (format_integer): Correct off-by-one error in the handling
42115         of precision with negative numbers in signed conversions..
42117 2017-01-05  Eric Botcazou  <ebotcazou@adacore.com>
42119         * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
42121 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42123         PR tree-optimization/71016
42124         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
42125         factor_out_conditional_conversion.  Formatting fix.
42126         (factor_out_conditional_conversion): Add cond_stmt argument.
42127         If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
42128         cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
42129         Formatting fix.
42131 2017-01-05  David Malcolm  <dmalcolm@redhat.com>
42133         * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
42134         read-rtl-function.o, and selftest-rtl.o.
42135         * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
42136         (selftest::aarch64_test_loading_full_dump): New function.
42137         (selftest::aarch64_run_selftests): New function.
42138         (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
42139         selftest::aarch64_run_selftests.
42140         * config/i386/i386.c
42141         (selftest::ix86_test_loading_dump_fragment_1): New function.
42142         (selftest::ix86_test_loading_call_insn): New function.
42143         (selftest::ix86_test_loading_full_dump): New function.
42144         (selftest::ix86_test_loading_unspec): New function.
42145         (selftest::ix86_run_selftests): Call the new functions.
42146         * emit-rtl.c (maybe_set_max_label_num): New function.
42147         * emit-rtl.h (maybe_set_max_label_num): New decl.
42148         * function.c (instantiate_decls): Guard call to
42149         instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
42150         * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
42151         "static".
42152         * gensupport.c (gen_reader::gen_reader): Pass "false"
42153         for new "compact" param of rtx_reader.
42154         * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
42155         rather than an empty string for NULL strings.
42156         * read-md.c: Potentially include config.h rather than bconfig.h.
42157         Wrap include of errors.h with #ifdef GENERATOR_FILE.
42158         (have_error): New global, copied from errors.c.
42159         (md_reader::read_name): Rename to...
42160         (md_reader::read_name_1): ...this, adding "out_loc" param,
42161         and converting "missing name or number" to returning false, rather
42162         than failing.
42163         (md_reader::read_name): Reimplement in terms of read_name_1.
42164         (md_reader::read_name_or_nil): New function.
42165         (md_reader::read_string): Handle "(nil)" by returning NULL.
42166         (md_reader::md_reader): Add new param "compact".
42167         (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
42168         (md_reader::read_file): New method.
42169         * read-md.h (md_reader::md_reader): Add new param "compact".
42170         (md_reader::read_file): New method.
42171         (md_reader::is_compact): New accessor.
42172         (md_reader::read_name): Convert return type from void to file_location.
42173         (md_reader::read_name_or_nil): New decl.
42174         (md_reader::read_name_1): New decl.
42175         (md_reader::m_compact): New field.
42176         (noop_reader::noop_reader): Pass "false" for new "compact" param
42177         of rtx_reader.
42178         (rtx_reader::rtx_reader): Add new "compact" param.
42179         (rtx_reader::read_rtx_operand): Make virtual and convert return
42180         type from void to rtx.
42181         (rtx_reader::read_until): New decl.
42182         (rtx_reader::handle_any_trailing_information): New virtual function.
42183         (rtx_reader::postprocess): New virtual function.
42184         (rtx_reader::finalize_string): New virtual function.
42185         (rtx_reader::m_in_call_function_usage): New field.
42186         (rtx_reader::m_reuse_rtx_by_id): New field.
42187         * read-rtl-function.c: New file.
42188         * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
42189         * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
42190         (selftest::verify_three_block_rtl_cfg): New decl.
42191         * read-rtl-function.h: New file.
42192         * read-rtl.c: Potentially include config.h rather than bconfig.h.
42193         For host, include function.h, memmodel.h, and emit-rtl.h.
42194         (one_time_initialization): New function.
42195         (struct compact_insn_name): New struct.
42196         (compact_insn_names): New array.
42197         (find_code): Handle insn codes in compact dumps.
42198         (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
42199         (bind_subst_iter_and_attr): Likewise.
42200         (add_condition_to_string): Likewise.
42201         (add_condition_to_rtx): Likewise.
42202         (apply_attribute_uses): Likewise.
42203         (add_current_iterators): Likewise.
42204         (apply_iterators): Likewise.
42205         (initialize_iterators): Guard usage of apply_subst_iterator with
42206         #ifdef GENERATOR_FILE.
42207         (read_conditions): Wrap with #ifdef GENERATOR_FILE.
42208         (md_reader::read_mapping): Likewise.
42209         (add_define_attr_for_define_subst): Likewise.
42210         (add_define_subst_attr): Likewise.
42211         (read_subst_mapping): Likewise.
42212         (check_code_iterator): Likewise.
42213         (rtx_reader::read_rtx): Likewise.  Move one-time initialization
42214         logic to...
42215         (one_time_initialization): New function.
42216         (rtx_reader::read_until): New method.
42217         (read_flags): New function.
42218         (parse_reg_note_name): New function.
42219         (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
42220         Handle reuse_rtx ids.
42221         Wrap iterator lookup within #ifdef GENERATOR_FILE.
42222         Add parsing support for RTL dumps, mirroring the special-cases in
42223         print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
42224         values, and calling handle_any_trailing_information.
42225         (rtx_reader::read_rtx_operand): Convert return type from void
42226         to rtx, returning return_rtx.  Handle case 'e'.  Call
42227         finalize_string on XSTR and XTMPL fields.
42228         (rtx_reader::read_nested_rtx):  Handle dumps in which trailing
42229          "(nil)" values were omitted.  Call the postprocess vfunc on the
42230         return_rtx.
42231         (rtx_reader::rtx_reader): Add new "compact" param and pass to base
42232         class ctor.  Initialize m_in_call_function_usage.  Call
42233         one_time_initialization.
42234         * rtl-tests.c (selftest::test_uncond_jump): Call
42235         set_new_first_and_last_insn.
42236         * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
42237         * selftest-rtl.c: New file.
42238         * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42239         (selftest::get_insn_by_uid): New decl.
42240         * selftest-run-tests.c (selftest::run_tests): Call
42241         read_rtl_function_c_tests.
42242         * selftest.h  (selftest::read_rtl_function_c_tests): New decl.
42243         * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42244         dumps.
42246 2017-01-05  Uros Bizjak  <ubizjak@gmail.com>
42248         * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42249         operands in a special way.  Assert that pos+len <= mode precision.
42251 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42253         * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42254         3 argument Alias with unlimited for the negative form.
42255         (fno-vect-cost-model): Removed.
42257 2017-01-05  Martin Liska  <mliska@suse.cz>
42259         * hsa-gen.c (gen_hsa_divmod): New function.
42260         (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42262 2017-01-05  Martin Liska  <mliska@suse.cz>
42264         PR pch/78970
42265         * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42266         header.
42268 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42270         * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42271         small constant length operands.
42273 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42275         * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42276         between loop iterations.
42278 2017-01-05  Martin Liska  <mliska@suse.cz>
42280         PR sanitizer/78815
42281         * gimplify.c (gimplify_decl_expr): Compare to
42282         asan_poisoned_variables instread of checking flags.
42283         (gimplify_target_expr): Likewise.
42284         (gimplify_expr): Likewise.
42285         (gimplify_function_tree): Conditionally initialize
42286         asan_poisoned_variables.
42288 2017-01-04  Jeff Law  <law@redhat.com>
42290         PR tree-optimizatin/78812
42291         * rtl.h (contains_mem_rtx_p): Prototype.
42292         * ifcvt.c (containts_mem_rtx_p): Move from here to...
42293         * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42294         * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42295         and prune MEMs that are not at the toplevel of a SET_SRC rtx.  Look
42296         through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42298 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42300         * input.c (assert_char_at_range): Default-initialize actual_range.
42302 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42304         * df-scan.c (df_ref_create_structure): Make regno unsigned,
42305         to match the caller.
42307 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42309         * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42310         insns after final jump in test to emit dummy move.
42312 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42314         * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42315         * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42317 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42319         * multiple_target.c (create_dispatcher_calls): Init e_next.
42320         * tree-ssa-loop-split.c (split_loop): Init border.
42321         * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42322         scalar_type.
42324 2017-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
42326         PR target/71977
42327         PR target/70568
42328         PR target/78823
42329         * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42330         (altivec_register_operand): Do not return true if the operand
42331         contains a SUBREG mixing SImode and SFmode.
42332         (vsx_register_operand): Likewise.
42333         (vsx_reg_sfsubreg_ok): New predicate.
42334         (vfloat_operand): Do not return true if the operand contains a
42335         SUBREG mixing SImode and SFmode.
42336         (vint_operand): Likewise.
42337         (vlogical_operand): Likewise.
42338         (gpc_reg_operand): Likewise.
42339         (int_reg_operand): Likewise.
42340         * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42341         * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42342         determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42343         SImode and SFmode.
42344         (rs6000_emit_move_si_sf_subreg): New helper function.
42345         (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42346         fixup SUBREGs involving SImode and SFmode.
42347         * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42348         numbers for the new peephole2 optimization.
42349         (peephole2 for SFmode unions): New peephole2 to optimize cases in
42350         the GLIBC math library that do AND/IOR/XOR operations on single
42351         precision floating point.
42352         * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42353         target macros to say whether we need to avoid SUBREGs mixing
42354         SImode and SFmode.
42355         (TARGET_ALLOW_SF_SUBREG): Likewise.
42356         * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42357         (UNSPEC_SI_FROM_SF): Likewise.
42358         (iorxor): Change spacing.
42359         (and_ior_xor): New iterator for AND, IOR, and XOR.
42360         (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42361         (movdi_from_sf_zero_ext): Likewise.
42362         (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42363         instead of gpc_reg_operand.  Add SImode/SFmode SUBREG support.
42364         (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42365         (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42366         (fms<mode>4): Likewise.
42367         (fnma<mode>4): Likewise.
42368         (fnms<mode>4): Likewise.
42369         (nfma<mode>4): Likewise.
42370         (nfms<mode>4): Likewise.
42372 2017-01-04  Marek Polacek  <polacek@redhat.com>
42374         PR c++/64767
42375         * doc/invoke.texi: Document -Wpointer-compare.
42377 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42379         * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42380         RejectNegative.
42382         * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42383         descriptions for -gdwarf-5 and emit them as uleb128 instead of
42384         2-byte data.
42386 2017-01-04  Kelvin Nilsen  <kelvin@gcc.gnu.org>
42388         PR target/78056
42389         * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42390         documentation of the powerpc_popcntb_ok attribute.
42391         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42392         code to issue warning messages if a requested CPU configuration is
42393         not supported by the binary (assembler and loader) toolchain.
42394         (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42395         made to define a built-in function that has been disabled.
42396         (paired_init_builtins): Add assertion to prevent ICE if attempt is
42397         made to define a built-in function that has been disabled.
42398         (altivec_init_builtins): Add comment explaining why definition
42399         of the DST built-in functions is not preceded by an assertion
42400         check.  Add assertions to prevent ICE if attempts are made to
42401         define an altivec predicate or an abs* built-in function that has
42402         been disabled.
42403         (htm_init_builtins): Add comment explaining why definition of the
42404         htm built-in functions is not preceded by an assertion check.
42406 2017-01-04  Jeff Law  <law@redhat.com>
42408         PR tree-optimizatin/67955
42409         * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42410         Allow any SSA_VAR_P as the base objects.  Use integer_zerop.  Verify
42411         the points-to solution does not include pt_null.  Use DECL_PT_UID
42412         unconditionally.
42414 2017-01-04  Uros Bizjak  <ubizjak@gmail.com>
42416         * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42417         Use gen_int_mode instead of gen_lopwart for const_int operands.
42419 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42421         PR tree-optimization/71563
42422         * match.pd: Simplify X << Y into X if Y is known to be 0 or
42423         out of range value - has low bits known to be zero.
42425 2017-01-04  Alan Modra  <amodra@gmail.com>
42427         * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42428         * configure: Regenerate.
42429         * config.in: Regenerate.
42431 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42433         PR bootstrap/77569
42434         * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42435         a substring of the message, but strcmp with the whole message.  Ifdef
42436         ENABLE_NLS, translate the message first using dgettext.
42438 2017-01-03  Jeff Law  <law@redhat.com>
42440         PR tree-optimizatin/78856
42441         * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42442         (mark_threaded_blocks): Remove code to truncate thread paths that
42443         cross multiple loop headers.  Instead invalidate the cached loop
42444         iteration information and handle case of a thread path walking
42445         into an irreducible region.
42447 2017-01-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
42449         PR target/78900
42450         * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42451         assertions.  Add support for doing the signbit if the IEEE 128-bit
42452         floating point value is in a GPR.
42453         * config/rs6000/rs6000.md (Fsignbit): Delete.
42454         (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42455         Update the length attribute if the value is in a GPR.
42456         (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42457         the sign or zero extension instruction, since the value is always 0/1.
42458         (signbit<mode>2_dm2): Delete using <Fsignbit>.
42460         PR target/78953
42461         * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42462         extracting SImode to a GPR register so that we can generate a
42463         store, limit the vector to be in a traditional Altivec register
42464         for the vextuwrx instruction.
42466 2017-01-03  Ian Lance Taylor  <iant@google.com>
42468         * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42470 2017-01-03  Martin Sebor  <msebor@redhat.com>
42472         PR tree-optimization/78696
42473         * gimple-ssa-sprintf.c (format_floating): Correct handling of
42474         precision.  Use MPFR for %f for greater fidelity.  Correct handling
42475         of %g.
42476         (pass_sprintf_length::compute_format_length): Set width and precision
42477         specified by asrerisk to void_node for vararg functions.
42478         (try_substitute_return_value): Adjust dump output.
42480 2017-01-03  David Edelsohn  <dje.gcc@gmail.com>
42482         * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42484 2017-01-03  Eric Botcazou  <ebotcazou@adacore.com>
42486         * doc/invoke.texi (SPARC options): Document -mlra as the default.
42487         * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42488         -mlra/-mno-lra was passed to the compiler.
42490 2017-01-03  James Cowgill  <James.Cowgill@imgtec.com>
42492         PR rtl-optimization/65618
42493         * emit-rtl.c (try_split): Move initialization of "before" and
42494         "after" to just before the call to emit_insn_after_setloc.
42496 2017-01-03  Gerald Pfeifer  <gerald@pfeifer.com>
42498         * doc/md.texi (Standard Names): Remove reference to Java frontend.
42500 2017-01-03  Pierre-Marie de Rodat  <derodat@adacore.com>
42502         * dwarf2out.c (gen_enumeration_type_die): When
42503         -gno-strict-dwarf, add a DW_AT_encoding attribute.
42505 2017-01-03  Jakub Jelinek  <jakub@redhat.com>
42507         PR tree-optimization/78965
42508         * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42509         Change first argument from const call_info & to call_info &.  For %n
42510         set info.nowrite to false.
42512         PR middle-end/78901
42513         * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42514         possibly throwing calls.
42516         * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42517         and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42518         and fns handling, rather than in a separate case SSA_NAME.
42520 2017-01-02  Jeff Law  <law@redhat.com>
42522         * config/darwin-driver.c (darwin_driver_init): Const-correctness
42523         fixes for first_period and second_period variables.
42525 2017-01-02  Uros Bizjak  <ubizjak@gmail.com>
42527         PR target/78967
42528         * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42529         (*insvqi_1): New insn pattern.
42530         (*insvqi_1_mem_rex64): Ditto.
42531         (*insvqi_2): Ditto.
42532         (*insvqi_3): Rename from *insvqi.
42534         (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42536 2017-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
42538         * doc/cfg.texi (Edges): Remove reference to Java.
42539         (Maintaining the CFG): Ditto.
42541 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42543         PR middle-end/77674
42544         * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42545         transparent aliases.
42547 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42549         PR middle-end/77484
42550         * predict.def (PRED_CALL): Update hitrate.
42551         (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42552         * predict.c (tree_estimate_probability_bb): Split CALL predictor
42553         into direct/indirect/polymorphic variants.
42555 2017-01-01  Jakub Jelinek  <jakub@redhat.com>
42557         Update copyright years.
42559         * gcc.c (process_command): Update copyright notice dates.
42560         * gcov-dump.c (print_version): Ditto.
42561         * gcov.c (print_version): Ditto.
42562         * gcov-tool.c (print_version): Ditto.
42563         * gengtype.c (create_file): Ditto.
42564         * doc/cpp.texi: Bump @copying's copyright year.
42565         * doc/cppinternals.texi: Ditto.
42566         * doc/gcc.texi: Ditto.
42567         * doc/gccint.texi: Ditto.
42568         * doc/gcov.texi: Ditto.
42569         * doc/install.texi: Ditto.
42570         * doc/invoke.texi: Ditto.
42572 Copyright (C) 2017 Free Software Foundation, Inc.
42574 Copying and distribution of this file, with or without modification,
42575 are permitted in any medium without royalty provided the copyright
42576 notice and this notice are preserved.