poly_int: expand_vector_ubsan_overflow
[official-gcc.git] / gcc / ChangeLog
blobb215b4f727802dd863d0a71231e0704681cea42a
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         * internal-fn.c (expand_vector_ubsan_overflow): Handle polynomial
6         numbers of elements.
8 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
9             Alan Hayward  <alan.hayward@arm.com>
10             David Sherwood  <david.sherwood@arm.com>
12         * match.pd: Cope with polynomial numbers of vector elements.
14 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
15             Alan Hayward  <alan.hayward@arm.com>
16             David Sherwood  <david.sherwood@arm.com>
18         * fold-const.c (fold_indirect_ref_1): Handle polynomial offsets
19         in a POINTER_PLUS_EXPR.
21 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22             Alan Hayward  <alan.hayward@arm.com>
23             David Sherwood  <david.sherwood@arm.com>
25         * omp-simd-clone.c (simd_clone_subparts): New function.
26         (simd_clone_init_simd_arrays): Use it instead of TYPE_VECTOR_SUBPARTS.
27         (ipa_simd_modify_function_body): Likewise.
29 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
30             Alan Hayward  <alan.hayward@arm.com>
31             David Sherwood  <david.sherwood@arm.com>
33         * tree-vect-generic.c (nunits_for_known_piecewise_op): New function.
34         (expand_vector_piecewise): Use it instead of TYPE_VECTOR_SUBPARTS.
35         (expand_vector_addition, add_rshift, expand_vector_divmod): Likewise.
36         (expand_vector_condition, vector_element): Likewise.
37         (subparts_gt): New function.
38         (get_compute_type): Use subparts_gt.
39         (count_type_subparts): Delete.
40         (expand_vector_operations_1): Use subparts_gt instead of
41         count_type_subparts.
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-data-refs.c (vect_no_alias_p): Replace with...
48         (vect_compile_time_alias): ...this new function.  Do the calculation
49         on poly_ints rather than trees.
50         (vect_prune_runtime_alias_test_list): Update call accordingly.
52 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
53             Alan Hayward  <alan.hayward@arm.com>
54             David Sherwood  <david.sherwood@arm.com>
56         * tree-vect-slp.c (vect_build_slp_tree_1): Handle polynomial
57         numbers of units.
58         (vect_schedule_slp_instance): Likewise.
60 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
61             Alan Hayward  <alan.hayward@arm.com>
62             David Sherwood  <david.sherwood@arm.com>
64         * tree-vect-slp.c (vect_get_and_check_slp_defs): Reject
65         constant and extern definitions for variable-length vectors.
66         (vect_get_constant_vectors): Note that the number of units
67         is known to be constant.
69 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
70             Alan Hayward  <alan.hayward@arm.com>
71             David Sherwood  <david.sherwood@arm.com>
73         * tree-vect-stmts.c (vectorizable_conversion): Treat the number
74         of units as polynomial.  Choose between WIDE and NARROW based
75         on multiple_p.
77 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
78             Alan Hayward  <alan.hayward@arm.com>
79             David Sherwood  <david.sherwood@arm.com>
81         * tree-vect-stmts.c (simd_clone_subparts): New function.
82         (vectorizable_simd_clone_call): Use it instead of TYPE_VECTOR_SUBPARTS.
84 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
85             Alan Hayward  <alan.hayward@arm.com>
86             David Sherwood  <david.sherwood@arm.com>
88         * tree-vect-stmts.c (vectorizable_call): Treat the number of
89         vectors as polynomial.  Use build_index_vector for
90         IFN_GOMP_SIMD_LANE.
92 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
93             Alan Hayward  <alan.hayward@arm.com>
94             David Sherwood  <david.sherwood@arm.com>
96         * tree-vect-stmts.c (get_load_store_type): Treat the number of
97         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
98         for variable-length vectors.
99         (vectorizable_mask_load_store): Treat the number of units as
100         polynomial, asserting that it is constant if the condition has
101         already been enforced.
102         (vectorizable_store, vectorizable_load): Likewise.
104 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
105             Alan Hayward  <alan.hayward@arm.com>
106             David Sherwood  <david.sherwood@arm.com>
108         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
109         of units as polynomial.  Punt if we can't tell at compile time
110         which vector contains the final result.
112 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
113             Alan Hayward  <alan.hayward@arm.com>
114             David Sherwood  <david.sherwood@arm.com>
116         * tree-vect-loop.c (vectorizable_induction): Treat the number
117         of units as polynomial.  Punt on SLP inductions.  Use an integer
118         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
119         cast of such a series for variable-length floating-point
120         reductions.
122 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
123             Alan Hayward  <alan.hayward@arm.com>
124             David Sherwood  <david.sherwood@arm.com>
126         * tree.h (build_index_vector): Declare.
127         * tree.c (build_index_vector): New function.
128         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
129         of units as polynomial, forcibly converting it to a constant if
130         vectorizable_reduction has already enforced the condition.
131         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
132         to create a {1,2,3,...} vector.
133         (vectorizable_reduction): Treat the number of units as polynomial.
134         Choose vectype_in based on the largest scalar element size rather
135         than the smallest number of units.  Enforce the restrictions
136         relied on above.
138 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
139             Alan Hayward  <alan.hayward@arm.com>
140             David Sherwood  <david.sherwood@arm.com>
142         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
143         number of units as polynomial.
145 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
146             Alan Hayward  <alan.hayward@arm.com>
147             David Sherwood  <david.sherwood@arm.com>
149         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
150         * target.def (autovectorize_vector_sizes): Return the vector sizes
151         by pointer, using vector_sizes rather than a bitmask.
152         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
153         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
154         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
155         Likewise.
156         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
157         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
158         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
159         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
160         * omp-general.c (omp_max_vf): Likewise.
161         * omp-low.c (omp_clause_aligned_alignment): Likewise.
162         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
163         * tree-vect-loop.c (vect_analyze_loop): Likewise.
164         * tree-vect-slp.c (vect_slp_bb): Likewise.
165         * doc/tm.texi: Regenerate.
166         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
167         to a poly_uint64.
168         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
169         the vector size as a poly_uint64 rather than an unsigned int.
170         (current_vector_size): Change from an unsigned int to a poly_uint64.
171         (get_vectype_for_scalar_type): Update accordingly.
172         * tree.h (build_truth_vector_type): Take the size and number of
173         units as a poly_uint64 rather than an unsigned int.
174         (build_vector_type): Add a temporary overload that takes
175         the number of units as a poly_uint64 rather than an unsigned int.
176         * tree.c (make_vector_type): Likewise.
177         (build_truth_vector_type): Take the number of units as a poly_uint64
178         rather than an unsigned int.
180 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
181             Alan Hayward  <alan.hayward@arm.com>
182             David Sherwood  <david.sherwood@arm.com>
184         * target.def (get_mask_mode): Take the number of units and length
185         as poly_uint64s rather than unsigned ints.
186         * targhooks.h (default_get_mask_mode): Update accordingly.
187         * targhooks.c (default_get_mask_mode): Likewise.
188         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
189         * doc/tm.texi: Regenerate.
191 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
192             Alan Hayward  <alan.hayward@arm.com>
193             David Sherwood  <david.sherwood@arm.com>
195         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
196         * omp-general.c (omp_max_vf): Likewise.
197         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
198         (expand_omp_simd): Handle polynomial safelen.
199         * omp-low.c (omplow_simd_context): Add a default constructor.
200         (omplow_simd_context::max_vf): Change from int to poly_uint64.
201         (lower_rec_simd_input_clauses): Update accordingly.
202         (lower_rec_input_clauses): Likewise.
204 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
205             Alan Hayward  <alan.hayward@arm.com>
206             David Sherwood  <david.sherwood@arm.com>
208         * tree-vectorizer.h (vect_nunits_for_cost): New function.
209         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
210         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
211         (vect_analyze_slp_cost): Likewise.
212         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
213         (vect_model_load_cost): Likewise.
215 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
216             Alan Hayward  <alan.hayward@arm.com>
217             David Sherwood  <david.sherwood@arm.com>
219         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
220         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
221         from an unsigned int * to a poly_uint64_pod *.
222         (calculate_unrolling_factor): New function.
223         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
225 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
226             Alan Hayward  <alan.hayward@arm.com>
227             David Sherwood  <david.sherwood@arm.com>
229         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
230         from an unsigned int to a poly_uint64.
231         (_loop_vec_info::slp_unrolling_factor): Likewise.
232         (_loop_vec_info::vectorization_factor): Change from an int
233         to a poly_uint64.
234         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
235         (vect_get_num_vectors): New function.
236         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
237         (vect_get_num_copies): Use vect_get_num_vectors.
238         (vect_analyze_data_ref_dependences): Change max_vf from an int *
239         to an unsigned int *.
240         (vect_analyze_data_refs): Change min_vf from an int * to a
241         poly_uint64 *.
242         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
243         than an unsigned HOST_WIDE_INT.
244         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
245         (vect_analyze_data_ref_dependence): Change max_vf from an int *
246         to an unsigned int *.
247         (vect_analyze_data_ref_dependences): Likewise.
248         (vect_compute_data_ref_alignment): Handle polynomial vf.
249         (vect_enhance_data_refs_alignment): Likewise.
250         (vect_prune_runtime_alias_test_list): Likewise.
251         (vect_shift_permute_load_chain): Likewise.
252         (vect_supportable_dr_alignment): Likewise.
253         (dependence_distance_ge_vf): Take the vectorization factor as a
254         poly_uint64 rather than an unsigned HOST_WIDE_INT.
255         (vect_analyze_data_refs): Change min_vf from an int * to a
256         poly_uint64 *.
257         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
258         vfm1 as a poly_uint64 rather than an int.  Make the same change
259         for the returned bound_scalar.
260         (vect_gen_vector_loop_niters): Handle polynomial vf.
261         (vect_do_peeling): Likewise.  Update call to
262         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
263         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
264         be constant.
265         * tree-vect-loop.c (vect_determine_vectorization_factor)
266         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
267         (vect_get_known_peeling_cost): Likewise.
268         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
269         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
270         (vect_transform_loop): Likewise.  Use the lowest possible VF when
271         updating the upper bounds of the loop.
272         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
273         rather than an int.
274         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
275         polynomial unroll factors.
276         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
277         (vect_make_slp_decision): Likewise.
278         (vect_supported_load_permutation_p): Likewise, and polynomial
279         vf too.
280         (vect_analyze_slp_cost): Handle polynomial vf.
281         (vect_slp_analyze_node_operations): Likewise.
282         (vect_slp_analyze_bb_1): Likewise.
283         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
284         than an unsigned HOST_WIDE_INT.
285         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
286         (vectorizable_load): Handle polynomial vf.
287         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
288         a poly_uint64.
289         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
291 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
292             Alan Hayward  <alan.hayward@arm.com>
293             David Sherwood  <david.sherwood@arm.com>
295         * match.pd: Handle bit operations involving three constants
296         and try to fold one pair.
298 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
300         * tree-vect-loop-manip.c: Include gimple-fold.h.
301         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
302         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
303         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
304         Add a path that uses a step of VF instead of 1, but disable it
305         for now.
306         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
307         and niters_no_overflow parameters.  Update calls to
308         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
309         Create a new SSA name if the latter choses to use a ste other
310         than zero, and return it via niters_vector_mult_vf_var.
311         * tree-vect-loop.c (vect_transform_loop): Update calls to
312         vect_do_peeling, vect_gen_vector_loop_niters and
313         slpeel_make_loop_iterate_ntimes.
314         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
315         (vect_gen_vector_loop_niters): Update declarations after above changes.
317 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
319         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
320         128-bit round to integer instructions.
321         (ceil<mode>2): Likewise.
322         (btrunc<mode>2): Likewise.
323         (round<mode>2): Likewise.
325 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
327         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
328         unaligned VSX load/store on P8/P9.
329         (expand_block_clear): Allow the use of unaligned VSX
330         load/store on P8/P9.
332 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
334         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
335         New function.
336         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
337         swap associated with both a load and a store.
339 2018-01-02  Andrew Waterman  <andrew@sifive.com>
341         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
342         * config/riscv/riscv.md (clear_cache): Use it.
344 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
346         * web.c: Remove out-of-date comment.
348 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
350         * expr.c (fixup_args_size_notes): Check that any existing
351         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
352         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
353         (emit_single_push_insn): ...here.
355 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
357         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
358         (const_vector_encoded_nelts): New function.
359         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
360         (const_vector_int_elt, const_vector_elt): Declare.
361         * emit-rtl.c (const_vector_int_elt_1): New function.
362         (const_vector_elt): Likewise.
363         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
364         of CONST_VECTOR_ELT.
366 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
368         * expr.c: Include rtx-vector-builder.h.
369         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
370         directly on the tree encoding.
371         (const_vector_from_tree): Likewise.
372         * optabs.c: Include rtx-vector-builder.h.
373         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
374         sequence of "u" values.
375         * vec-perm-indices.c: Include rtx-vector-builder.h.
376         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
377         directly on the vec_perm_indices encoding.
379 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
381         * doc/rtl.texi (const_vector): Describe new encoding scheme.
382         * Makefile.in (OBJS): Add rtx-vector-builder.o.
383         * rtx-vector-builder.h: New file.
384         * rtx-vector-builder.c: Likewise.
385         * rtl.h (rtx_def::u2): Add a const_vector field.
386         (CONST_VECTOR_NPATTERNS): New macro.
387         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
388         (CONST_VECTOR_DUPLICATE_P): Likewise.
389         (CONST_VECTOR_STEPPED_P): Likewise.
390         (CONST_VECTOR_ENCODED_ELT): Likewise.
391         (const_vec_duplicate_p): Check for a duplicated vector encoding.
392         (unwrap_const_vec_duplicate): Likewise.
393         (const_vec_series_p): Check for a non-duplicated vector encoding.
394         Say that the function only returns true for integer vectors.
395         * emit-rtl.c: Include rtx-vector-builder.h.
396         (gen_const_vec_duplicate_1): Delete.
397         (gen_const_vector): Call gen_const_vec_duplicate instead of
398         gen_const_vec_duplicate_1.
399         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
400         (gen_const_vec_duplicate): Use rtx_vector_builder.
401         (gen_const_vec_series): Likewise.
402         (gen_rtx_CONST_VECTOR): Likewise.
403         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
404         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
405         Build a new vector rather than modifying a CONST_VECTOR in-place.
406         (handle_special_swappables): Update call accordingly.
407         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
408         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
409         Build a new vector rather than modifying a CONST_VECTOR in-place.
410         (handle_special_swappables): Update call accordingly.
412 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
414         * simplify-rtx.c (simplify_const_binary_operation): Use
415         CONST_VECTOR_ELT instead of XVECEXP.
417 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
419         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
420         the selector elements to be different from the data elements
421         if the selector is a VECTOR_CST.
422         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
423         ssizetype for the selector.
425 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
427         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
428         before testing each element individually.
429         * tree-vect-generic.c (lower_vec_perm): Likewise.
431 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
433         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
434         * selftest-run-tests.c (selftest::run_tests): Call it.
435         * vector-builder.h (vector_builder::operator ==): New function.
436         (vector_builder::operator !=): Likewise.
437         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
438         (vec_perm_indices::all_from_input_p): New function.
439         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
440         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
441         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
442         instead of reading the VECTOR_CST directly.  Detect whether both
443         vector inputs are the same before constructing the vec_perm_indices,
444         and update the number of inputs argument accordingly.  Use the
445         utility functions added above.  Only construct sel2 if we need to.
447 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
449         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
450         the broadcast of the low byte.
451         (expand_mult_highpart): Use an explicit encoding for the permutes.
452         * optabs-query.c (can_mult_highpart_p): Likewise.
453         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
454         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
455         (vectorizable_bswap): Likewise.
456         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
457         explicit encoding for the power-of-2 permutes.
458         (vect_permute_store_chain): Likewise.
459         (vect_grouped_load_supported): Likewise.
460         (vect_permute_load_chain): Likewise.
462 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
464         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
465         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
466         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
467         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
468         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
469         (vect_gen_perm_mask_any): Likewise.
471 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
473         * int-vector-builder.h: New file.
474         * vec-perm-indices.h: Include int-vector-builder.h.
475         (vec_perm_indices): Redefine as an int_vector_builder.
476         (auto_vec_perm_indices): Delete.
477         (vec_perm_builder): Redefine as a stand-alone class.
478         (vec_perm_indices::vec_perm_indices): New function.
479         (vec_perm_indices::clamp): Likewise.
480         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
481         (vec_perm_indices::new_vector): New function.
482         (vec_perm_indices::new_expanded_vector): Update for new
483         vec_perm_indices class.
484         (vec_perm_indices::rotate_inputs): New function.
485         (vec_perm_indices::all_in_range_p): Operate directly on the
486         encoded form, without computing elided elements.
487         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
488         encoding.  Update for new vec_perm_indices class.
489         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
490         the given vec_perm_builder.
491         (expand_vec_perm_var): Update vec_perm_builder constructor.
492         (expand_mult_highpart): Use vec_perm_builder instead of
493         auto_vec_perm_indices.
494         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
495         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
496         or double series encoding as appropriate.
497         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
498         vec_perm_indices instead of auto_vec_perm_indices.
499         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
500         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
501         (vect_permute_store_chain): Likewise.
502         (vect_grouped_load_supported): Likewise.
503         (vect_permute_load_chain): Likewise.
504         (vect_shift_permute_load_chain): Likewise.
505         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
506         (vect_transform_slp_perm_load): Likewise.
507         (vect_schedule_slp_instance): Likewise.
508         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
509         (vectorizable_mask_load_store): Likewise.
510         (vectorizable_bswap): Likewise.
511         (vectorizable_store): Likewise.
512         (vectorizable_load): Likewise.
513         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
514         vec_perm_indices instead of auto_vec_perm_indices.  Use
515         tree_to_vec_perm_builder to read the vector from a tree.
516         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
517         vec_perm_builder instead of a vec_perm_indices.
518         (have_whole_vector_shift): Use vec_perm_builder and
519         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
520         truncation to calc_vec_perm_mask_for_shift.
521         (vect_create_epilog_for_reduction): Likewise.
522         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
523         from auto_vec_perm_indices to vec_perm_indices.
524         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
525         instead of changing individual elements.
526         (aarch64_vectorize_vec_perm_const): Use new_vector to install
527         the vector in d.perm.
528         * config/arm/arm.c (expand_vec_perm_d::perm): Change
529         from auto_vec_perm_indices to vec_perm_indices.
530         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
531         instead of changing individual elements.
532         (arm_vectorize_vec_perm_const): Use new_vector to install
533         the vector in d.perm.
534         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
535         Update vec_perm_builder constructor.
536         (rs6000_expand_interleave): Likewise.
537         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
538         (rs6000_expand_interleave): Likewise.
540 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
542         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
543         to qimode could truncate the indices.
544         * optabs.c (expand_vec_perm_var): Likewise.
546 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
548         * Makefile.in (OBJS): Add vec-perm-indices.o.
549         * vec-perm-indices.h: New file.
550         * vec-perm-indices.c: Likewise.
551         * target.h (vec_perm_indices): Replace with a forward class
552         declaration.
553         (auto_vec_perm_indices): Move to vec-perm-indices.h.
554         * optabs.h: Include vec-perm-indices.h.
555         (expand_vec_perm): Delete.
556         (selector_fits_mode_p, expand_vec_perm_var): Declare.
557         (expand_vec_perm_const): Declare.
558         * target.def (vec_perm_const_ok): Replace with...
559         (vec_perm_const): ...this new hook.
560         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
561         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
562         * doc/tm.texi: Regenerate.
563         * optabs.def (vec_perm_const): Delete.
564         * doc/md.texi (vec_perm_const): Likewise.
565         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
566         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
567         expand_vec_perm for constant permutation vectors.  Assert that
568         the mode of variable permutation vectors is the integer equivalent
569         of the mode that is being permuted.
570         * optabs-query.h (selector_fits_mode_p): Declare.
571         * optabs-query.c: Include vec-perm-indices.h.
572         (selector_fits_mode_p): New function.
573         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
574         is defined, instead of checking whether the vec_perm_const_optab
575         exists.  Use targetm.vectorize.vec_perm_const instead of
576         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
577         fit in the vector mode before using a variable permute.
578         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
579         vec_perm_indices instead of an rtx.
580         (expand_vec_perm): Replace with...
581         (expand_vec_perm_const): ...this new function.  Take the selector
582         as a vec_perm_indices rather than an rtx.  Also take the mode of
583         the selector.  Update call to shift_amt_for_vec_perm_mask.
584         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
585         Use vec_perm_indices::new_expanded_vector to expand the original
586         selector into bytes.  Check whether the indices fit in the vector
587         mode before using a variable permute.
588         (expand_vec_perm_var): Make global.
589         (expand_mult_highpart): Use expand_vec_perm_const.
590         * fold-const.c: Includes vec-perm-indices.h.
591         * tree-ssa-forwprop.c: Likewise.
592         * tree-vect-data-refs.c: Likewise.
593         * tree-vect-generic.c: Likewise.
594         * tree-vect-loop.c: Likewise.
595         * tree-vect-slp.c: Likewise.
596         * tree-vect-stmts.c: Likewise.
597         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
598         Delete.
599         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
600         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
601         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
602         (aarch64_vectorize_vec_perm_const): ...this new function.
603         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
604         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
605         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
606         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
607         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
608         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
609         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
610         into...
611         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
612         check for NEON modes.
613         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
614         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
615         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
616         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
617         into...
618         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
619         the old VEC_PERM_CONST conditions.
620         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
621         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
622         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
623         (ia64_vectorize_vec_perm_const_ok): Merge into...
624         (ia64_vectorize_vec_perm_const): ...this new function.
625         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
626         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
627         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
628         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
629         * config/mips/mips.c (mips_expand_vec_perm_const)
630         (mips_vectorize_vec_perm_const_ok): Merge into...
631         (mips_vectorize_vec_perm_const): ...this new function.
632         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
633         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
634         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
635         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
636         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
637         (rs6000_expand_vec_perm_const): Delete.
638         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
639         Delete.
640         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
641         (altivec_expand_vec_perm_const_le): Take each operand individually.
642         Operate on constant selectors rather than rtxes.
643         (altivec_expand_vec_perm_const): Likewise.  Update call to
644         altivec_expand_vec_perm_const_le.
645         (rs6000_expand_vec_perm_const): Delete.
646         (rs6000_vectorize_vec_perm_const_ok): Delete.
647         (rs6000_vectorize_vec_perm_const): New function.
648         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
649         an element count and rtx array.
650         (rs6000_expand_extract_even): Update call accordingly.
651         (rs6000_expand_interleave): Likewise.
652         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
653         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
654         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
655         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
656         (rs6000_expand_vec_perm_const): Delete.
657         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
658         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
659         (altivec_expand_vec_perm_const_le): Take each operand individually.
660         Operate on constant selectors rather than rtxes.
661         (altivec_expand_vec_perm_const): Likewise.  Update call to
662         altivec_expand_vec_perm_const_le.
663         (rs6000_expand_vec_perm_const): Delete.
664         (rs6000_vectorize_vec_perm_const_ok): Delete.
665         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
666         reference to the SPE evmerge intructions.
667         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
668         an element count and rtx array.
669         (rs6000_expand_extract_even): Update call accordingly.
670         (rs6000_expand_interleave): Likewise.
671         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
672         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
673         new function.
674         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
676 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
678         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
679         vector mode and that that mode matches the mode of the data
680         being permuted.
681         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
682         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
683         directly using expand_vec_perm_1 when forcing selectors into
684         registers.
685         (expand_vec_perm_var): New function, split out from expand_vec_perm.
687 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
689         * optabs-query.h (can_vec_perm_p): Delete.
690         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
691         * optabs-query.c (can_vec_perm_p): Split into...
692         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
693         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
694         particular selector is valid.
695         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
696         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
697         (vect_grouped_load_supported): Likewise.
698         (vect_shift_permute_load_chain): Likewise.
699         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
700         (vect_transform_slp_perm_load): Likewise.
701         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
702         (vectorizable_bswap): Likewise.
703         (vect_gen_perm_mask_checked): Likewise.
704         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
705         implementations of variable permutation vectors into account
706         when deciding which selector to use.
707         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
708         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
709         with a false third argument.
710         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
711         to test whether the constant selector is valid and can_vec_perm_var_p
712         to test whether a variable selector is valid.
714 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
716         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
717         * optabs-query.c (can_vec_perm_p): Likewise.
718         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
719         instead of vec_perm_indices.
720         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
721         (vect_gen_perm_mask_checked): Likewise,
722         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
723         (vect_gen_perm_mask_checked): Likewise,
725 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
727         * optabs-query.h (qimode_for_vec_perm): Declare.
728         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
729         (qimode_for_vec_perm): ...this new function.
730         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
732 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
734         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
735         does not have a conditional at the top.
737 2018-01-02  Richard Biener  <rguenther@suse.de>
739         * ipa-inline.c (big_speedup_p): Fix expression.
741 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
743         PR target/81616
744         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
745         for generic 4->6.
747 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
749         PR target/81616
750         Generic tuning.
751         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
752         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
753         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
754         cond_taken_branch_cost 3->4.
756 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
758         PR tree-optimization/83581
759         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
760         TODO_cleanup_cfg if any changes have been made.
762         PR middle-end/83608
763         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
764         convert_modes if target mode has the right side, but different mode
765         class.
767         PR middle-end/83609
768         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
769         last argument when extracting from CONCAT.  If either from_real or
770         from_imag is NULL, use expansion through memory.  If result is not
771         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
772         the parts directly to inner mode, if even that fails, use expansion
773         through memory.
775         PR middle-end/83623
776         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
777         check for bswap in mode rather than HImode and use that in expand_unop
778         too.
780 2017-12-31  Jakub Jelinek  <jakub@redhat.com>
782         PR target/83536
783         * config/i386/i386.c (ix86_attribute_table): Remove excess
784         initializer for "shared" attribute.
786 2017-12-30  Tom de Vries  <tom@codesourcery.com>
788         PR libgomp/83046
789         * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
790         DECL_PRESERVE_P.
791         * lto-streamer-out.c (prune_offload_funcs): New function.  Remove
792         offload_funcs entries that no longer have a corresponding cgraph_node.
793         Mark the remaining ones as DECL_PRESERVE_P.
794         (output_lto): Call prune_offload_funcs.
796 2017-12-30  Jakub Jelinek  <jakub@redhat.com>
798         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
799         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
800         vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
801         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
802         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
803         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
804         vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
805         vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
806         vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
807         vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
808         vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
809         vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
810         vaesenc_<mode>, vpclmulqdq_<mode>,
811         avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
813 2017-12-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
815         * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
816         _Float<N> and _Float<N>X variants for rounding built-in
817         functions.
818         (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
819         (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
820         (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
821         (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
822         (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
823         * builtins.c (mathfn_built_in_2): Likewise.
824         * internal-fn.def (CEIL): Likewise.
825         (FLOOR): Likewise.
826         (NEARBYINT): Likewise.
827         (RINT): Likewise.
828         (ROUND): Likewise.
829         (TRUNC): Likewise.
830         * convert.c (convert_to_integer_1): Likewise.
831         * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
832         (integer_valued_real_call_p): Likewise.
833         * fold-const-call.c (fold_const_call_ss): Likewise.
834         * gencfn-macros.c (print_case_cfn): Change CFN and operator
835         printers to take a const char * suffix instead of a bool.
836         (print_define_operator_list): Likewise.
837         (fltall_suffixes): New list of suffixes, that include the
838         traditional suffixes as well as all of the _Float<N> and
839         _Float<N>X suffixes.
840         (main): For _Float<N> and _Float<N>X functions, emit both
841         <name>_FN and <name>_ALL variants.  The <macro>_FN variant only
842         has the _Float<N> and _Float<N>X case names or operators.  The
843         <name>_ALL variant has both the traditional and the
844         _Float<N>/_Float<N>X case names or operators.
845         * match.pd (COPYSIGN optimizations): Provide optimizations for
846         _Float<N> and _Float<N>X types where possible.
847         (MIN/MAX optimizations): Likewise.
848         (sqrt optimizations): Likewise.
849         (rounding optimizations): Likewise.
851 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
853         * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
854         gen_int_mode rather than GEN_INT.
855         * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
857 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
859         * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
860         instead of CONSTANT_P.
861         (gen_vec_series): Likewise.
862         * simplify-rtx.c (simplify_binary_operation_1): Likewise.
864 2017-12-28  Andreas Schwab  <schwab@linux-m68k.org>
866         * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
867         CC_STATUS_INIT.
869 2017-12-27  Kugan Vivekanandarajah  <kuganv@linaro.org>
871         * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
872         (aarch64_ld1x2<VDC:mode>): Likewise.
873         (aarch64_simd_ld1<mode>_x2): Likewise.
874         (aarch64_simd_ld1<mode>_x2): Likewise.
875         * config/aarch64/arm_neon.h (vld1_u8_x2): New.
876         (vld1_s8_x2): Likewise.
877         (vld1_u16_x2): Likewise.
878         (vld1_s16_x2): Likewise.
879         (vld1_u32_x2): Likewise.
880         (vld1_s32_x2): Likewise.
881         (vld1_u64_x2): Likewise.
882         (vld1_s64_x2): Likewise.
883         (vld1_f16_x2): Likewise.
884         (vld1_f32_x2): Likewise.
885         (vld1_f64_x2): Likewise.
886         (vld1_p8_x2): Likewise.
887         (vld1_p16_x2): Likewise.
888         (vld1_p64_x2): Likewise.
889         (vld1q_u8_x2): Likewise.
890         (vld1q_s8_x2): Likewise.
891         (vld1q_u16_x2): Likewise.
892         (vld1q_s16_x2): Likewise.
893         (vld1q_u32_x2): Likewise.
894         (vld1q_s32_x2): Likewise.
895         (vld1q_u64_x2): Likewise.
896         (vld1q_s64_x2): Likewise.
897         (vld1q_f16_x2): Likewise.
898         (vld1q_f32_x2): Likewise.
899         (vld1q_f64_x2): Likewise.
900         (vld1q_p8_x2): Likewise.
901         (vld1q_p16_x2): Likewise.
902         (vld1q_p64_x2): Likewise.
904 2017-12-27  Martin Liska  <mliska@suse.cz>
906         PR tree-optimization/83552
907         * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
908         of get_string_lenth to a SSA_NAME if not a GIMPLE value.
910 2017-12-27  Tom de Vries  <tom@codesourcery.com>
912         * config/nvptx/nvptx.c (nvptx_option_override): Disable
913         -gstatement-frontiers.
915 2017-12-26  Alexander Monakov  <amonakov@ispras.ru>
917         PR rtl-optimization/83513
918         * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
919         before priority comparison.
921 2017-12-25  Jakub Jelinek  <jakub@redhat.com>
923         PR target/83488
924         * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
925         ix86_isa_flags2 to ix86_isa_flags.
926         * config/i386/i386-c.c (ix86_target_macros_internal): Test
927         OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
928         isa_flags rather than isa_flags2.
929         * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
930         and -mavx512bitalg from isa2_opts to isa_opts.
931         (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
932         in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
933         and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
934         Formatting fixes.
935         (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
936         OPTION_MASK_ISA_AVX512F ored with another option similarly to
937         OPTION_MASK_ISA_AVX512VL.  Even for OPTION_MASK_ISA_AVX512VL don't
938         clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
939         (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
940         OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
941         OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
942         * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
943         builtins from bdesc_args2 to bdesc_args section.
944         (__builtin_ia32_compressstoreuqi512_mask,
945         __builtin_ia32_compressstoreuhi512_mask,
946         __builtin_ia32_compressstoreuqi256_mask,
947         __builtin_ia32_expandloadqi512_mask,
948         __builtin_ia32_expandloadqi512_maskz,
949         __builtin_ia32_expandloadhi512_mask,
950         __builtin_ia32_expandloadhi512_maskz,
951         __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
952         __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
953         __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
954         __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
955         __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
956         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
957         __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
958         __builtin_ia32_vpshldv_v32hi_maskz,
959         __builtin_ia32_vpopcountb_v64qi_mask,
960         __builtin_ia32_vpopcountw_v32hi_mask,
961         __builtin_ia32_vpshufbitqmb512_mask,
962         __builtin_ia32_vpshufbitqmb256_mask): Add
963         " | OPTION_MASK_ISA_AVX512BW".
964         (__builtin_ia32_expandloadqi256_mask,
965         __builtin_ia32_expandloadqi256_maskz,
966         __builtin_ia32_vpopcountb_v32qi_mask): Add
967         " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
968         (__builtin_ia32_expandloadhi256_mask,
969         __builtin_ia32_expandloadhi256_maskz,
970         __builtin_ia32_expandloadqi128_mask,
971         __builtin_ia32_expandloadqi128_maskz,
972         __builtin_ia32_expandloadhi128_mask,
973         __builtin_ia32_expandloadhi128_maskz,
974         __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
975         __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
976         __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
977         __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
978         __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
979         __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
980         __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
981         __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
982         __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
983         __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
984         __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
985         __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
986         __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
987         __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
988         __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
989         __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
990         __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
991         __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
992         __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
993         __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
994         __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
995         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
996         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
997         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
998         __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
999         __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
1000         __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
1001         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
1002         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
1003         __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
1004         __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
1005         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
1006         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1007         __builtin_ia32_vpopcountw_v8hi_mask): Add
1008         " | OPTION_MASK_ISA_AVX512VL".
1009         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
1010         _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
1011         _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
1012         _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
1013         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
1014         _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
1015         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
1016         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
1017         _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
1018         _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
1019         _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
1020         * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
1021         _mm_mask_bitshuffle_epi64_mask): Likewise.
1022         * common/config/i386/i386-common.c
1023         (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
1024         OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
1025         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1026         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
1027         OPTION_MASK_ISA_AVX512BITALG_UNSET.
1028         (OPTION_MASK_ISA2_AVX512F_UNSET,
1029         OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
1030         (ix86_handle_option): For -mno-general-regs-only, clear from
1031         ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
1032         just OPTION_MASK_ISA_MPX.  For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
1033         -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
1034         ix86_isa_flags2.  For -mno-avx512f likewise, instead of masking
1035         individually listed ISAs.  For -m{,no-}avx512{vpopcntdq,bitalg} adjust
1036         for moving from ix86_isa_flags2 to ix86_isa_flags.
1038 2017-12-23  Jakub Jelinek  <jakub@redhat.com>
1040         PR c++/83553
1041         * fold-const.c (struct contains_label_data): New type.
1042         (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
1043         inside of a SWITCH_BODY seen during the walk.
1044         (contains_label_p): Use walk_tree instead of
1045         walk_tree_without_duplicates, prepare data for contains_label_1 and
1046         provide own pset.
1048 2017-12-22  Jakub Jelinek  <jakub@redhat.com>
1050         PR debug/83547
1051         * tree-iterator.c (alloc_stmt_list): Start with cleared
1052         TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
1053         or old one reused.
1055         PR target/83488
1056         * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
1057         enable avx512f explicitly in #pragma GCC target.
1058         * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
1059         __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
1060         __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
1061         __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
1062         __builtin_ia32_vpdpbusds_v8si_mask,
1063         __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
1064         __builtin_ia32_vpdpbusds_v4si_mask,
1065         __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
1066         __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
1067         __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
1068         __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
1069         __builtin_ia32_vpdpwssds_v8si_mask,
1070         __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
1071         __builtin_ia32_vpdpwssds_v4si_mask,
1072         __builtin_ia32_vpdpwssds_v4si_maskz): Use
1073         OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
1074         just OPTION_MASK_ISA_AVX512VNNI.
1076 2017-12-22  Martin Jambor  <mjambor@suse.cz>
1078         PR lto/82027
1079         * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
1080         clones.
1082 2017-12-22  Julia Koval  <julia.koval@intel.com>
1083             Sebastian Peryt  <sebastian.peryt@intel.com>
1085         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
1086         OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
1087         (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
1088         * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
1089         * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
1090         _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
1091         _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
1092         _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
1093         _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
1094         _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
1095         _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
1096         _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
1097         _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
1098         * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
1099         _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
1100         _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
1101         _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
1102         _mm256_maskz_popcnt_epi64): New intrinsics.
1103         * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
1104         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
1105         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
1106         V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
1107         V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
1108         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
1109         __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
1110         __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
1111         __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
1112         __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
1113         __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
1114         __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
1115         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
1116         __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
1117         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1118         __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
1119         __builtin_ia32_vpshufbitqmb256_mask,
1120         __builtin_ia32_vpshufbitqmb512_mask): New builtins.
1121         * config/i386/i386-c.c (__AVX512BITALG__): New.
1122         * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
1123         (ix86_valid_target_attribute_inner_p): Ditto.
1124         (ix86_expand_args_builtin): Handle new types.
1125         * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
1126         * config/i386/i386.opt: Add -mavx512bitalg.
1127         * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
1128         avx512bitalgintrin.h.
1129         * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
1130         (vpopcount<mode><mask_name>): Add more types.
1131         (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
1132         * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
1134 2017-12-22  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
1136         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
1137         Or in OPTION_MASK_ISA_AVX512F_SET.
1138         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1139         OPTION_MASK_ISA_AVX512VNNI_UNSET.
1140         (ix86_handle_option): Adjust for
1141         OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
1142         * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
1143         section to ARGS.
1144         * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
1145         isa_flag instead of isa_flag2.
1146         * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
1147         isa_opts2 to isa_opts.
1148         * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
1149         to ix86_isa_flags.
1151 2017-12-22  Mike Stump  <mikestump@comcast.net>
1152             Eric Botcazou  <ebotcazou@adacore.com>
1154         * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
1156 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1158         PR middle-end/83487
1159         * config/i386/i386.c (ix86_function_arg_boundary): Return
1160         PARM_BOUNDARY for TYPE_EMPTY_P types.
1162         PR c/83448
1163         * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
1164         if navail is >= dir.len.
1166 2017-12-21  Steve Ellcey  <sellcey@cavium.com>
1168         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
1169         triplet for ilp32.
1171 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1173         PR rtl-optimization/80747
1174         PR rtl-optimization/83512
1175         * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
1176         succ edge from ENTRY, copy partition from e->dest to the newly
1177         created bb.
1178         * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
1179         ENTRY, use BB_PARTITION of its successor block as current_partition.
1180         Don't copy partition when splitting succ edge from ENTRY.
1182         PR tree-optimization/83523
1183         * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
1184         for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
1185         (convert_mult_to_fma): Likewise.
1187         PR tree-optimization/83521
1188         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
1189         gimple_build_assign without code on result of
1190         fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
1191         a VIEW_CONVERT_EXPR.
1193 2017-12-21  Andrew Pinski  <apinski@cavium.com>
1194             Steve Ellcey  <sellcey@cavium.com>
1196         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
1197         multi-arch for ilp32.
1199 2017-12-21  Uros Bizjak  <ubizjak@gmail.com>
1201         PR target/83467
1202         * config/i386/i386.md (*ashl<mode>3_mask): Add operand
1203         constraints to operand 2.
1204         (*ashl<mode>3_mask_1): Ditto.
1205         (*<shift_insn><mode>3_mask): Ditto.
1206         (*<shift_insn><mode>3_mask_1): Ditto.
1207         (*<rotate_insn><mode>3_mask): Ditto.
1208         (*<rotate_insn><mode>3_mask_1): Ditto.
1210 2017-12-21  Alexandre Oliva  <aoliva@redhat.com>
1212         * reorg.c (make_return_insns): Reemit each insn with its own location.
1214 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1216         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
1217         generation for cases where splatting a value is not useful.
1218         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
1219         across a vec_duplicate and a paradoxical subreg forming a vector
1220         mode to a vec_concat.
1222 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1224         * combine.c (simplify_set): Do not transform subregs to zero_extends
1225         if the destination is not a scalar int mode.
1227 2017-12-21  Eric Botcazou  <ebotcazou@adacore.com>
1229         PR c++/82872
1230         * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1231         shared zero if the input has overflowed.
1233 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1235         * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1236         support for Cortex-A55 and Cortex-A75.
1238 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1240         * common/config/arm/arm-common.c (compare_opt_names): Add function
1241         comment.  Use strcmp instead of manual loop.
1243 2017-12-21  Martin Liska  <mliska@suse.cz>
1245         PR gcov-profile/83509
1246         * gcov-dump.c (dump_gcov_file): Do not read info about
1247         support_unexecuted_blocks for gcda files.
1249 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1251         PR rtl-optimization/82973
1252         * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1253         (valid_for_const_vector_p): ... this.
1254         * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1255         (valid_for_const_vector_p): ... this.  Adjust function comment.
1256         (gen_vec_duplicate): Adjust caller.
1257         * optabs.c (expand_vector_broadcast): Likewise.
1258         * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1259         CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1260         constant.
1261         (simplify_const_binary_operation): Likewise.  Use CONST_FIXED_P macro
1262         instead of GET_CODE == CONST_FIXED.
1263         (simplify_subreg): Use CONST_FIXED_P macro instead of
1264         GET_CODE == CONST_FIXED.
1266         PR target/83488
1267         * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1268         -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1269         -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1270         (ix86_option_override_internal): Adjust for
1271         OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1272         and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1273         (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1274         (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1275         Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1276         and OPTION_MASK_ISA_CLZERO builtins.  Use def_builtin instead of
1277         def_builtin2 for CET builtins.
1278         (ix86_expand_builtin): Remove bdesc_special_args2 handling.  Fix
1279         up formatting in IX86_BUILTIN_RDPID code.
1280         * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1281         section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1282         * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1283         ix86_isa_flags2 to ix86_isa_flags.
1284         (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1285         ix86_isa_flags2.
1286         * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1287         OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1288         Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1289         of isa_flag2.
1290         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1291         Or in OPTION_MASK_ISA_AVX512F_SET.
1292         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1293         OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1294         (ix86_handle_option): Adjust for
1295         OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1296         and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1298 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1299             Alan Hayward  <alan.hayward@arm.com>
1300             David Sherwood  <david.sherwood@arm.com>
1302         * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1303         factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1304         * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1305         Track polynomial offsets.
1307 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1308             Alan Hayward  <alan.hayward@arm.com>
1309             David Sherwood  <david.sherwood@arm.com>
1311         * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
1312         Treat drb->init as a poly_int.  Fail if its misalignment wrt
1313         vector_alignment isn't known.
1315 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1316             Alan Hayward  <alan.hayward@arm.com>
1317             David Sherwood  <david.sherwood@arm.com>
1319         * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1320         field.
1321         (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1322         (vect_loop_versioning): Take the loop versioning threshold as a
1323         separate parameter.
1324         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1325         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1326         versioning_threshold.
1327         (vect_analyze_loop_2): Compute the loop versioning threshold
1328         whenever loop versioning is needed, and store it in the new
1329         field rather than combining it with the cost model threshold.
1330         (vect_transform_loop): Update call to vect_loop_versioning.
1331         Try to combine the loop versioning and cost thresholds here.
1333 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1334             Alan Hayward  <alan.hayward@arm.com>
1335             David Sherwood  <david.sherwood@arm.com>
1337         * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1338         poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1339         * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1340         * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1341         an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1342         (group_compare_offset): Update accordingly.
1343         (split_small_address_groups_p): Likewise.
1344         (record_use): Take addr_offset as a poly_uint64 rather than
1345         an unsigned HOST_WIDE_INT.
1346         (strip_offset): Return the offset as a poly_uint64 rather than
1347         an unsigned HOST_WIDE_INT.
1348         (record_group_use, split_address_groups): Track polynomial offsets.
1349         (add_iv_candidate_for_use): Likewise.
1350         (addr_offset_valid_p): Take the offset as a poly_int64 rather
1351         than a HOST_WIDE_INT.
1352         (strip_offset_1): Return the offset as a poly_int64 rather than
1353         a HOST_WIDE_INT.
1355 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1356             Alan Hayward  <alan.hayward@arm.com>
1357             David Sherwood  <david.sherwood@arm.com>
1359         * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1360         rather than a HOST_WIDE_INT.
1361         * tree.c (get_binfo_at_offset): Likewise.
1363 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1364             Alan Hayward  <alan.hayward@arm.com>
1365             David Sherwood  <david.sherwood@arm.com>
1367         * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1368         rather than a HOST_WIDE_INT.
1369         * tree-sra.c (build_ref_for_offset): Likewise.
1371 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1372             Alan Hayward  <alan.hayward@arm.com>
1373             David Sherwood  <david.sherwood@arm.com>
1375         * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1376         than an offset_int.
1377         * tree.c (mem_ref_offset): Likewise.
1378         (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1379         * builtins.c (get_object_alignment_2): Likewise.
1380         * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1381         * gimple-fold.c (get_base_constructor): Likewise.
1382         * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1383         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1384         Likewise.
1385         * ipa-polymorphic-call.c
1386         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1387         * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1388         (get_ancestor_addr_info): Likewise.
1389         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1390         * match.pd: Likewise.
1391         * tree-data-ref.c (dr_analyze_innermost): Likewise.
1392         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1393         * tree-eh.c (tree_could_trap_p): Likewise.
1394         * tree-object-size.c (addr_object_size): Likewise.
1395         * tree-ssa-address.c (copy_ref_info): Likewise.
1396         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1397         (indirect_refs_may_alias_p): Likewise.
1398         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1399         * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1400         (non_rewritable_mem_ref_base): Likewise.
1401         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1402         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1403         * varasm.c (decode_addr_const): Likewise.
1405 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1406             Alan Hayward  <alan.hayward@arm.com>
1407             David Sherwood  <david.sherwood@arm.com>
1409         * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1410         offsets for MEM_REFs.
1412 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1413             Alan Hayward  <alan.hayward@arm.com>
1414             David Sherwood  <david.sherwood@arm.com>
1416         * tree-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1417         as a poly_uint64 rather than an unsigned int.
1418         * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1420 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1421             Alan Hayward  <alan.hayward@arm.com>
1422             David Sherwood  <david.sherwood@arm.com>
1424         * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1425         to poly_int64.
1426         (decode_addr_const): Update accordingly.
1428 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1429             Alan Hayward  <alan.hayward@arm.com>
1430             David Sherwood  <david.sherwood@arm.com>
1432         * tree.h (bit_field_size, bit_field_offset): New functions.
1433         * hsa-gen.c (gen_hsa_addr): Use them.
1434         * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1435         (simplify_vector_constructor): Likewise.
1436         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1437         * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1438         BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1439         * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1440         tree_fits_uhwi_p.
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         * expr.h (emit_group_load, emit_group_load_into_temps)
1447         (emit_group_store): Take the size as a poly_int64 rather than an int.
1448         * expr.c (emit_group_load_1, emit_group_load): Likewise.
1449         (emit_group_load_into_temp, emit_group_store): Likewise.
1451 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1452             Alan Hayward  <alan.hayward@arm.com>
1453             David Sherwood  <david.sherwood@arm.com>
1455         * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1456         an unsigned int to a poly_uint64.
1457         * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1458         sizes as poly_uint64s rather than unsigned ints.
1459         * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1460         Likewise.
1462 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1463             Alan Hayward  <alan.hayward@arm.com>
1464             David Sherwood  <david.sherwood@arm.com>
1466         * lra-constraints.c (emit_inc): Change inc_amount from an int
1467         to a poly_int64.
1469 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1470             Alan Hayward  <alan.hayward@arm.com>
1471             David Sherwood  <david.sherwood@arm.com>
1473         * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1474         to a poly_uint64.
1475         (add_stack_var, stack_var_cmp, partition_stack_vars)
1476         (dump_stack_var_partition): Update accordingly.
1477         (alloc_stack_frame_space): Take the size as a poly_int64 rather
1478         than a HOST_WIDE_INT.
1479         (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1480         (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1481         (account_stack_vars, expand_one_var): Likewise.  Return a poly_uint64
1482         rather than a HOST_WIDE_INT.
1484 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1485             Alan Hayward  <alan.hayward@arm.com>
1486             David Sherwood  <david.sherwood@arm.com>
1488         * target.def (return_pops_args): Treat both the input and output
1489         sizes as poly_int64s rather than HOST_WIDE_INTS.
1490         * targhooks.h (default_return_pops_args): Update accordingly.
1491         * targhooks.c (default_return_pops_args): Likewise.
1492         * doc/tm.texi: Regenerate.
1493         * emit-rtl.h (incoming_args): Change pops_args, size and
1494         outgoing_args_size from int to poly_int64_pod.
1495         * function.h (expr_status): Change x_pending_stack_adjust and
1496         x_stack_pointer_delta from int to poly_int64.
1497         (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1498         (ARGS_SIZE_RTX): Update accordingly.
1499         * calls.c (highest_outgoing_arg_in_use): Change from int to
1500         unsigned int.
1501         (stack_usage_watermark, stored_args_watermark): New variables.
1502         (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1503         (emit_call_1): Change the stack_size and rounded_stack_size
1504         parameters from HOST_WIDE_INT to poly_int64.  Track n_popped
1505         as a poly_int64.
1506         (save_fixed_argument_area): Check stack_usage_watermark.
1507         (initialize_argument_information): Change old_pending_adj from
1508         a HOST_WIDE_INT * to a poly_int64_pod *.
1509         (compute_argument_block_size): Return the size as a poly_int64
1510         rather than an int.
1511         (finalize_must_preallocate): Track polynomial argument sizes.
1512         (compute_argument_addresses): Likewise.
1513         (internal_arg_pointer_based_exp): Track polynomial offsets.
1514         (mem_overlaps_already_clobbered_arg_p): Rename to...
1515         (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1516         size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1517         Check stored_args_used_watermark.
1518         (load_register_parameters): Update accordingly.
1519         (check_sibcall_argument_overlap_1): Likewise.
1520         (combine_pending_stack_adjustment_and_call): Take the unadjusted
1521         args size as a poly_int64 rather than an int.  Return a bool
1522         indicating whether the optimization was possible and return
1523         the new adjustment by reference.
1524         (check_sibcall_argument_overlap): Track polynomail argument sizes.
1525         Update stored_args_watermark.
1526         (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1527         (expand_call): Likewise.  Maintain stack_usage_watermark and
1528         stored_args_watermark.  Update calls to
1529         combine_pending_stack_adjustment_and_call.
1530         (emit_library_call_value_1): Handle polynomial argument sizes.
1531         Call stack_region_maybe_used_p and mark_stack_region_used.
1532         Maintain stack_usage_watermark.
1533         (store_one_arg): Likewise.  Update call to
1534         mem_overlaps_already_clobbered_arg_p.
1535         * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1536         HOST_WIDE_INT.
1537         * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1538         * config/microblaze/microblaze.c (microblaze_function_prologue):
1539         Likewise.
1540         * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1541         TARGET_RETURN_POPS_ARGS interface.
1542         (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1543         to HOST_WIDE_INT.
1544         * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1545         * config/i386/i386.c (ix86_return_pops_args): Update for new
1546         TARGET_RETURN_POPS_ARGS interface.
1547         (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1548         * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1549         * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1550         TARGET_RETURN_POPS_ARGS interface.
1551         * config/vax/vax.c (vax_return_pops_args): Likewise.
1552         * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1553         (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1554         * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1555         * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1556         * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1557         * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1558         * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1559         * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1560         * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1561         * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1562         and x_stack_pointer_delta from int to poly_int64.
1563         * dojump.c (do_pending_stack_adjust): Update accordingly.
1564         * explow.c (allocate_dynamic_stack_space): Handle polynomial
1565         stack_pointer_deltas.
1566         * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1567         (pad_to_arg_alignment): Track polynomial offsets.
1568         (assign_parm_find_stack_rtl): Likewise.
1569         (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1570         * toplev.c (output_stack_usage): Update reference to
1571         current_function_pushed_stack_size.
1573 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1574             Alan Hayward  <alan.hayward@arm.com>
1575             David Sherwood  <david.sherwood@arm.com>
1577         * function.c (in_arg_offset, var_offset, dynamic_offset)
1578         (out_arg_offset, cfa_offset): Change from int to poly_int64.
1579         (instantiate_new_reg): Return the new offset as a poly_int64_pod
1580         rather than a HOST_WIDE_INT.
1581         (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1582         (instantiate_virtual_regs_in_insn): Likewise.
1584 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1585             Alan Hayward  <alan.hayward@arm.com>
1586             David Sherwood  <david.sherwood@arm.com>
1588         * rtl.h (get_args_size, add_args_size_note): New functions.
1589         (find_args_size_adjust): Return a poly_int64 rather than a
1590         HOST_WIDE_INT.
1591         (fixup_args_size_notes): Likewise.  Make the same change to the
1592         end_args_size parameter.
1593         * rtlanal.c (get_args_size, add_args_size_note): New functions.
1594         * builtins.c (expand_builtin_trap): Use add_args_size_note.
1595         * calls.c (emit_call_1): Likewise.
1596         * explow.c (adjust_stack_1): Likewise.
1597         * cfgcleanup.c (old_insns_match_p): Update use of
1598         find_args_size_adjust.
1599         * combine.c (distribute_notes): Track polynomial arg sizes.
1600         * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1601         end_true_args_size, beg_delay_args_size and end_delay_args_size
1602         from HOST_WIDE_INT to poly_int64.
1603         (add_cfi_args_size): Take the args_size as a poly_int64 rather
1604         than a HOST_WIDE_INT.
1605         (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1606         (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1607         polynomial arg sizes.
1608         * emit-rtl.c (try_split): Use get_args_size.
1609         * recog.c (peep2_attempt): Likewise.
1610         * reload1.c (reload_as_needed): Likewise.
1611         * expr.c (find_args_size_adjust): Return the adjustment as a
1612         poly_int64 rather than a HOST_WIDE_INT.
1613         (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1614         to a poly_int64 and change the return type in the same way.
1615         (emit_single_push_insn): Track polynomial arg sizes.
1617 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1618             Alan Hayward  <alan.hayward@arm.com>
1619             David Sherwood  <david.sherwood@arm.com>
1621         * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1622         from HOST_WIDE_INT to poly_int64.
1623         * expr.c (push_block, emit_push_insn): Likewise.
1625 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1626             Alan Hayward  <alan.hayward@arm.com>
1627             David Sherwood  <david.sherwood@arm.com>
1629         * function.h (frame_space): Change start and length from HOST_WIDE_INT
1630         to poly_int64.
1631         (get_frame_size): Return the size as a poly_int64 rather than a
1632         HOST_WIDE_INT.
1633         (frame_offset_overflow): Take the offset as a poly_int64 rather
1634         than a HOST_WIDE_INT.
1635         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1636         (assign_stack_temp): Likewise for the size.
1637         * function.c (get_frame_size): Return a poly_int64 rather than
1638         a HOST_WIDE_INT.
1639         (frame_offset_overflow): Take the offset as a poly_int64 rather
1640         than a HOST_WIDE_INT.
1641         (try_fit_stack_local): Take the start, length and size as poly_int64s
1642         rather than HOST_WIDE_INTs.  Return the offset as a poly_int64_pod
1643         rather than a HOST_WIDE_INT.
1644         (add_frame_space): Take the start and end as poly_int64s rather than
1645         HOST_WIDE_INTs.
1646         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1647         (assign_stack_temp): Likewise for the size.
1648         (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1649         to poly_int64.
1650         (find_temp_slot_from_address): Handle polynomial offsets.
1651         (combine_temp_slots): Likewise.
1652         * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1653         to poly_int64.
1654         * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1655         poly_int64 rather than a HOST_WIDE_INT.
1656         (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1657         than a HOST_WIDE_INT.
1658         (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1659         polynomial frame offsets.
1660         * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1661         as a poly_int64 rather than an int.
1662         * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1663         * config/v850/v850-protos.h (compute_frame_size): Likewise.
1664         * config/v850/v850.c (compute_frame_size): Likewise.
1665         * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1666         * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1667         * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1668         * config/pa/pa.c (pa_compute_frame_size): Likewise.
1669         * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1670         rather than a HOST_WIDE_INT.
1671         * explow.c (get_dynamic_stack_base): Likewise.
1672         * final.c (final_start_function): Use the constant lower bound
1673         of the frame size for -Wframe-larger-than.
1674         * ira.c (do_reload): Adjust for new get_frame_size return type.
1675         * lra.c (lra): Likewise.
1676         * reload1.c (reload): Likewise.
1677         * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1678         * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1679         * rtlanal.c (get_initial_register_offset): Return the offset as
1680         a poly_int64 rather than a HOST_WIDE_INT.
1682 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1683             Alan Hayward  <alan.hayward@arm.com>
1684             David Sherwood  <david.sherwood@arm.com>
1686         * reload1.c (elim_table): Change initial_offset, offset and
1687         previous_offset from HOST_WIDE_INT to poly_int64_pod.
1688         (offsets_at): Change the target array's element type from
1689         HOST_WIDE_INT to poly_int64_pod.
1690         (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1691         (elimination_costs_in_insn, update_eliminable_offsets)
1692         (verify_initial_elim_offsets, set_offsets_for_label)
1693         (init_eliminable_invariants): Update after above changes.
1695 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1696             Alan Hayward  <alan.hayward@arm.com>
1697             David Sherwood  <david.sherwood@arm.com>
1699         * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1700         * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1701         (decomposition): Change start and end from HOST_WIDE_INT
1702         to poly_int64_pod.
1703         (decompose, immune_p): Update accordingly.
1704         (find_inc_amount): Return a poly_int64 rather than an int.
1705         * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1706         rather than an int.
1708 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1709             Alan Hayward  <alan.hayward@arm.com>
1710             David Sherwood  <david.sherwood@arm.com>
1712         * tree.h (get_inner_reference): Return the bitsize and bitpos
1713         as poly_int64_pods rather than HOST_WIDE_INT.
1714         * fold-const.h (ptr_difference_const): Return the pointer difference
1715         as a poly_int64_pod rather than a HOST_WIDE_INT.
1716         * expr.c (get_inner_reference): Return the bitsize and bitpos
1717         as poly_int64_pods rather than HOST_WIDE_INT.
1718         (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1719         offsets and sizes.
1720         * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1721         rather than a HOST_WIDE_INT.  Update call to get_inner_reference.
1722         (optimize_bit_field_compare): Update call to get_inner_reference.
1723         (decode_field_reference): Likewise.
1724         (fold_unary_loc): Track polynomial offsets and sizes.
1725         (split_address_to_core_and_offset): Return the bitpos as a
1726         poly_int64_pod rather than a HOST_WIDE_INT.
1727         (ptr_difference_const): Likewise for the pointer difference.
1728         * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1729         * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1730         * dbxout.c (dbxout_expand_expr): Likewise.
1731         * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1732         (loc_list_from_tree_1, fortran_common): Likewise.
1733         * gimple-laddress.c (pass_laddress::execute): Likewise.
1734         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1735         * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1736         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1737         * tree-affine.c (tree_to_aff_combination): Likewise.
1738         (get_inner_reference_aff): Likewise.
1739         * tree-data-ref.c (split_constant_offset_1): Likewise.
1740         (dr_analyze_innermost): Likewise.
1741         * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1742         * tree-sra.c (ipa_sra_check_caller): Likewise.
1743         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1744         * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1745         (instrument_bool_enum_load, instrument_object_size): Likewise.
1746         * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1747         to get_inner_reference.
1748         * hsa-gen.c (gen_hsa_addr): Likewise.
1749         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1750         * tsan.c (instrument_expr): Likewise.
1751         * match.pd: Update call to ptr_difference_const.
1753 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1754             Alan Hayward  <alan.hayward@arm.com>
1755             David Sherwood  <david.sherwood@arm.com>
1757         * fold-const.c (fold_comparison): Track sizes and offsets as
1758         poly_int64s rather than HOST_WIDE_INTs when folding address
1759         comparisons.
1761 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1762             Alan Hayward  <alan.hayward@arm.com>
1763             David Sherwood  <david.sherwood@arm.com>
1765         * expr.h (get_bit_range): Return the bitstart and bitend as
1766         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Return the bitpos
1767         as a poly_int64 rather than a HOST_WIDE_INT.
1768         * expr.c (get_bit_range): Likewise.
1769         (expand_assignment): Update call accordingly.
1770         * fold-const.c (optimize_bit_field_compare): Likewise.
1772 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1773             Alan Hayward  <alan.hayward@arm.com>
1774             David Sherwood  <david.sherwood@arm.com>
1776         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1777         (aarch64_legitimate_address_p): Use it instead of an rtx code,
1778         as an optional final parameter.
1779         * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1780         (aarch64_legitimate_address_p): Likewise.
1781         (aarch64_print_address_internal): Take an aarch64_addr_query_type
1782         instead of an rtx code.
1783         (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1784         (aarch64_legitimate_address_hook_p): Likewise.
1785         (aarch64_print_ldpstp_address): Likewise.
1786         (aarch64_print_operand_address): Likewise.
1787         (aarch64_address_cost): Likewise.
1788         * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1789         * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1790         (aarch64_mem_pair_lanes_operand): Likewise.
1792 2017-12-20  Richard Biener  <rguenther@suse.de>
1794         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1795         update_stmt_if_modified.
1797 2017-12-20  Wilco Dijkstra  <wdijkstr@arm.com>
1799         PR tree-optimization/83491
1800         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1801         before walking uses.  Improve coding style and comments.
1803 2017-12-20  Tom de Vries  <tom@codesourcery.com>
1805         * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1807 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
1809         PR ipa/83506
1810         * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1811         todo_flags_finish.
1812         (pass_ipa_free_fn_summary): Add small_p private data member,
1813         initialize to false in the ctor.
1814         (pass_ipa_free_fn_summary::clone,
1815         pass_ipa_free_fn_summary::set_pass_param,
1816         pass_ipa_free_fn_summary::gate): New methods.
1817         (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1818         | TODO_dump_symtab if small_p.
1819         * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1820         entry and add another instance of the pass with false parm after
1821         ipa-pure-const.
1822         * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1823         ipa_free_fn_summary here.
1825 2017-12-20  Paolo Carlini  <paolo.carlini@oracle.com>
1827         * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1829 2017-12-20  Martin Sebor  <msebor@redhat.com>
1831         PR testsuite/83131
1832         * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1833         (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1834         (expand_builtin_strncmp): Same.
1836 2017-12-20  Alexandre Oliva  <aoliva@redhat.com>
1838         PR bootstrap/83396
1839         * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1840         allowed debug stmts before labels.
1841         (expand_gimple_basic_block): Likewise.
1842         * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1843         * gimple-iterator.h (gsi_after_labels): Likewise.
1844         * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1845         rename reused variable, and simplify using gsi_move_before.
1846         * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1847         * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1848         (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1849         (gimple_verify_flow_info, gimple_block_label): Likewise.
1850         (make_blocks): Move debug markers after adjacent labels.
1851         * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1852         allowed debug insns outside blocks.
1853         * df-scan.c (df_insn_delete): Likewise.
1854         * lra-constraints.c (update_ebb_live_info): Likewise.
1855         * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1856         (vt_initialize, delete_vta_debug_insns): Likewise.
1857         (reemit_marker_as_note): Drop BB parm.  Adjust callers.
1859 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1860             Alan Hayward  <alan.hayward@arm.com>
1861             David Sherwood  <david.sherwood@arm.com>
1863         * poly-int-types.h (round_down_to_byte_boundary): New macro.
1864         (round_up_to_byte_boundary): Likewise.
1865         * expr.h (get_bit_range): Add temporary shim.
1866         * gimple-ssa-store-merging.c (store_operand_info): Change the
1867         bitsize, bitpos, bitregion_start and bitregion_end fields from
1868         unsigned HOST_WIDE_INT to poly_uint64.
1869         (merged_store_group): Likewise load_align_base.
1870         (compatible_load_p, compatible_load_p): Update accordingly.
1871         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1872         (split_group, imm_store_chain_info::output_merged_store): Likewise.
1873         (mem_valid_for_store_merging): Return the bitsize, bitpos,
1874         bitregion_start and bitregion_end as poly_uint64s rather than
1875         unsigned HOST_WIDE_INTs.  Track polynomial offsets internally.
1876         (handled_load): Take the bitsize, bitpos,
1877         bitregion_start and bitregion_end as poly_uint64s rather than
1878         unsigned HOST_WIDE_INTs.
1879         (pass_store_merging::process_store): Update call to
1880         mem_valid_for_store_merging.
1882 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1883             Alan Hayward  <alan.hayward@arm.com>
1884             David Sherwood  <david.sherwood@arm.com>
1886         * builtins.c (get_object_alignment_2): Track polynomial offsets
1887         and sizes.  Update the alignment handling.
1889 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1890             Alan Hayward  <alan.hayward@arm.com>
1891             David Sherwood  <david.sherwood@arm.com>
1893         * tree.h (get_inner_reference): Add a version that returns the
1894         offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1895         * cfgexpand.c (expand_debug_expr): Track polynomial offsets.  Simply
1896         the case in which bitpos is not associated with the first byte.
1898 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1899             Alan Hayward  <alan.hayward@arm.com>
1900             David Sherwood  <david.sherwood@arm.com>
1902         * tree-affine.h (get_inner_reference_aff): Return the size as a
1903         poly_widest_int.
1904         * tree-affine.c (get_inner_reference_aff): Likewise.
1905         * tree-data-ref.c (dr_may_alias_p): Update accordingly.
1906         * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
1908 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1909             Alan Hayward  <alan.hayward@arm.com>
1910             David Sherwood  <david.sherwood@arm.com>
1912         * fold-const.c (pointer_may_wrap_p): Take the offset as a
1913         HOST_WIDE_INT rather than a poly_int64.
1915 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1916             Alan Hayward  <alan.hayward@arm.com>
1917             David Sherwood  <david.sherwood@arm.com>
1919         * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
1920         HOST_WIDE_INT to poly_int64_pod.
1921         (perform_symbolic_merge): Update accordingly.
1922         (bswap_replace): Likewise.
1924 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1925             Alan Hayward  <alan.hayward@arm.com>
1926             David Sherwood  <david.sherwood@arm.com>
1928         * tree-affine.h (aff_tree::offset): Change from widest_int
1929         to poly_widest_int.
1930         (wide_int_ext_for_comb): Delete.
1931         (aff_combination_const, aff_comb_cannot_overlap_p): Take the
1932         constants as poly_widest_int rather than widest_int.
1933         (aff_combination_constant_multiple_p): Return the multiplier
1934         as a poly_widest_int.
1935         (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
1936         polynomial offsets.
1937         * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
1938         version static and add an overload for poly_widest_int.
1939         (aff_combination_const, aff_combination_add_cst)
1940         (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
1941         the constants as poly_widest_int rather than widest_int.
1942         (tree_to_aff_combination): Generalize INTEGER_CST case to
1943         poly_int_tree_p.
1944         (aff_combination_to_tree): Track offsets as poly_widest_ints.
1945         (aff_combination_add_product, aff_combination_mult): Handle
1946         polynomial offsets.
1947         (aff_combination_constant_multiple_p): Return the multiplier
1948         as a poly_widest_int.
1949         * tree-predcom.c (determine_offset): Return the offset as a
1950         poly_widest_int.
1951         (split_data_refs_to_components, suitable_component_p): Update
1952         accordingly.
1953         (valid_initializer_p): Update call to
1954         aff_combination_constant_multiple_p.
1955         * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
1956         * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
1957         as a poly_int64 rather than a HOST_WIDE_INT.
1958         (get_address_cost): Handle polynomial offsets.
1959         (iv_elimination_compare_lt): Likewise.
1960         (rewrite_use_nonlinear_expr): Likewise.
1962 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1963             Alan Hayward  <alan.hayward@arm.com>
1964             David Sherwood  <david.sherwood@arm.com>
1966         * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
1967         as a poly_int64_pod rather than a HOST_WIDE_INT.
1968         (get_addr_base_and_unit_offset): Likewise.
1969         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1970         (get_addr_base_and_unit_offset): Likewise.
1971         * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
1972         to poly_int64 in example.
1973         * fold-const.c (fold_binary_loc): Update call to
1974         get_addr_base_and_unit_offset.
1975         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1976         (maybe_canonicalize_mem_ref_addr): Likewise.
1977         (gimple_fold_stmt_to_constant_1): Likewise.
1978         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1979         Likewise.
1980         * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
1981         * match.pd: Likewise.
1982         * omp-low.c (lower_omp_target): Likewise.
1983         * tree-sra.c (build_ref_for_offset): Likewise.
1984         (build_debug_ref_for_model): Likewise.
1985         * tree-ssa-address.c (maybe_fold_tmr): Likewise.
1986         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
1987         * tree-ssa-ccp.c (optimize_memcpy): Likewise.
1988         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
1989         (constant_pointer_difference): Likewise.
1990         * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
1991         * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
1992         * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
1993         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
1994         (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
1995         (set_ssa_val_to): Likewise.
1996         * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
1997         (maybe_diag_stxncpy_trunc): Likewise.
1998         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1999         * tree.c (build_simple_mem_ref_loc): Likewise.
2000         (array_at_struct_end_p): Likewise.
2002 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2003             Alan Hayward  <alan.hayward@arm.com>
2004             David Sherwood  <david.sherwood@arm.com>
2006         * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
2007         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
2008         (get_ref_base_and_extent_hwi): Declare.
2009         * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
2010         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
2011         (get_ref_base_and_extent_hwi): New function.
2012         * cfgexpand.c (expand_debug_expr): Update call to
2013         get_ref_base_and_extent.
2014         * dwarf2out.c (add_var_loc_to_decl): Likewise.
2015         * gimple-fold.c (get_base_constructor): Return the offset as a
2016         poly_int64_pod rather than a HOST_WIDE_INT.
2017         (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
2018         * ipa-polymorphic-call.c
2019         (ipa_polymorphic_call_context::set_by_invariant)
2020         (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
2021         (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
2022         (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
2023         rather than get_ref_base_and_extent.
2024         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
2025         * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
2026         (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
2027         Likewise.
2028         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
2029         call to get_ref_base_and_extent.
2030         * tree-sra.c (create_access, get_access_for_expr): Likewise.
2031         * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
2032         (stmt_kills_ref_p): Likewise.
2033         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2034         * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
2035         Likewise.
2036         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
2037         Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
2038         when calling native_encode_expr.
2039         * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
2040         call to get_ref_base_and_extent.
2041         (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
2042         get_ref_base_and_extent.
2043         * var-tracking.c (track_expr_p): Likewise.
2045 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2046             Alan Hayward  <alan.hayward@arm.com>
2047             David Sherwood  <david.sherwood@arm.com>
2049         * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
2050         HOST_WIDE_INT to poly_int64_pod.
2051         * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
2052         polynomail parameter offsets.
2054 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2055             Alan Hayward  <alan.hayward@arm.com>
2056             David Sherwood  <david.sherwood@arm.com>
2058         * gengtype.c (main): Handle poly_int64_pod.
2059         * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
2060         (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
2061         (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
2062         from HOST_WIDE_INT to poly_int64_pod.
2063         * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
2064         (copy_cfa): New function.
2065         (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
2066         (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
2067         (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
2068         (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
2069         (initial_return_save): Treat offsets as poly_ints.
2070         (def_cfa_0): Likewise.  Cache the CFA in dw_cfi_cfa_loc if either
2071         offset is nonconstant.
2072         (reg_save): Take the offset as a poly_int64.  Fall back to
2073         DW_CFA_expression for nonconstant offsets.
2074         (queue_reg_save): Take the offset as a poly_int64.
2075         * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
2077 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2078             Alan Hayward  <alan.hayward@arm.com>
2079             David Sherwood  <david.sherwood@arm.com>
2081         * rtl.h (operand_subword, operand_subword_force): Take the offset
2082         as a poly_uint64 an unsigned int.
2083         * emit-rtl.c (operand_subword, operand_subword_force): Likewise.
2085 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2086             Alan Hayward  <alan.hayward@arm.com>
2087             David Sherwood  <david.sherwood@arm.com>
2089         * doc/rtl.texi: Update documentation of SUBREG_BYTE.  Document the
2090         'p' format code.  Use INT_LIST rather than SUBREG as the example of
2091         a code with an XINT and an XEXP.  Remove the implication that
2092         accessing an rtx field using XINT is expected to work.
2093         * rtl.def (SUBREG): Change format from "ei" to "ep".
2094         * rtl.h (rtunion::rt_subreg): New field.
2095         (XCSUBREG): New macro.
2096         (SUBREG_BYTE): Use it.
2097         (subreg_shape): Change offset from an unsigned int to a poly_uint16.
2098         Update constructor accordingly.
2099         (subreg_shape::operator ==): Update accordingly.
2100         (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
2101         than an unsigned int.
2102         (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
2103         a poly_uint64 rather than an unsigned int.
2104         (subreg_lsb_1): Likewise.  Take the offset as a poly_uint64 rather
2105         than an unsigned int.
2106         (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
2107         (subreg_size_highpart_offset): Return a poly_uint64 rather than
2108         an unsigned int.  Take the sizes as poly_uint64s.
2109         (subreg_offset_from_lsb): Return a poly_uint64 rather than
2110         an unsigned int.  Take the shift as a poly_uint64 rather than
2111         an unsigned int.
2112         (subreg_regno_offset, subreg_offset_representable_p): Take the offset
2113         as a poly_uint64 rather than an unsigned int.
2114         (simplify_subreg_regno): Likewise.
2115         (byte_lowpart_offset): Return the memory offset as a poly_int64
2116         rather than an int.
2117         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2118         poly_uint64 rather than an unsigned int.
2119         (simplify_subreg, simplify_gen_subreg, subreg_get_info)
2120         (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
2121         poly_uint64 rather than an unsigned int.
2122         * rtl.c (rtx_format): Describe 'p' in comment.
2123         (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
2124         * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
2125         offset as a poly_uint64 rather than an unsigned int.
2126         (byte_lowpart_offset): Return the memory offset as a poly_int64
2127         rather than an int.
2128         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2129         poly_uint64 rather than an unsigned int.
2130         (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
2131         mode sizes as poly_uint64s rather than unsigned ints.  Return a
2132         poly_uint64 rather than an unsigned int.
2133         (subreg_lowpart_p): Treat subreg offsets as poly_ints.
2134         (copy_insn_1): Handle 'p'.
2135         * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
2136         (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
2137         an unsigned int.  Return the shift in the same way.
2138         (subreg_lsb): Return the shift as a poly_uint64 rather than an
2139         unsigned int.
2140         (subreg_size_offset_from_lsb): Take the sizes and shift as
2141         poly_uint64s rather than unsigned ints.  Return the offset as
2142         a poly_uint64.
2143         (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
2144         (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
2145         an unsigned int.
2146         * rtlhash.c (add_rtx): Handle 'p'.
2147         * genemit.c (gen_exp): Likewise.
2148         * gengenrtl.c (type_from_format, gendef): Likewise.
2149         * gensupport.c (subst_pattern_match, get_alternatives_number)
2150         (collect_insn_data, alter_predicate_for_insn, alter_constraints)
2151         (subst_dup): Likewise.
2152         * gengtype.c (adjust_field_rtx_def): Likewise.
2153         * genrecog.c (find_operand, find_matching_operand, validate_pattern)
2154         (match_pattern_2): Likewise.
2155         (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
2156         (rtx_test::subreg_field): New function.
2157         (operator ==, safe_to_hoist_p, transition_parameter_type)
2158         (print_nonbool_test, print_test): Handle SUBREG_FIELD.
2159         * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
2160         * genpeep.c (match_rtx): Likewise.
2161         * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
2162         (rtx_writer::print_rtx_operand): Handle 'p'.
2163         (print_value): Handle SUBREG.
2164         * read-rtl.c (apply_int_iterator): Likewise.
2165         (rtx_reader::read_rtx_operand): Handle 'p'.
2166         * alias.c (rtx_equal_for_memref_p): Likewise.
2167         * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
2168         * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
2169         as poly_ints.
2170         * calls.c (expand_call): Likewise.
2171         * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
2172         (make_extraction, gen_lowpart_for_combine): Likewise.
2173         * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
2174         Likewise.
2175         * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
2176         rather than an unsigned int.  Treat subreg offsets as poly_ints.
2177         (exp_equiv_p): Handle 'p'.
2178         (hash_rtx_cb): Likewise.  Treat subreg offsets as poly_ints.
2179         (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
2180         * dse.c (find_shift_sequence): Likewise.
2181         * dwarf2out.c (rtl_for_decl_location): Likewise.
2182         * expmed.c (extract_low_bits): Likewise.
2183         * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
2184         (expand_expr_real_2): Likewise.
2185         * final.c (alter_subreg): Likewise.
2186         (leaf_renumber_regs_insn): Handle 'p'.
2187         * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
2188         Treat subreg offsets as poly_ints.
2189         * fwprop.c (forward_propagate_and_simplify): Likewise.
2190         * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
2191         * ira.c (get_subreg_tracking_sizes): Likewise.
2192         * ira-conflicts.c (go_through_subreg): Likewise.
2193         * ira-lives.c (process_single_reg_class_operands): Likewise.
2194         * jump.c (rtx_renumbered_equal_p): Likewise.  Handle 'p'.
2195         * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
2196         as a poly_uint64 rather than an unsigned int.
2197         (simplify_gen_subreg_concatn, resolve_simple_move): Treat
2198         subreg offsets as poly_ints.
2199         * lra-constraints.c (operands_match_p): Handle 'p'.
2200         (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
2201         * lra-spills.c (assign_mem_slot): Likewise.
2202         * postreload.c (move2add_valid_value_p): Likewise.
2203         * recog.c (general_operand, indirect_operand): Likewise.
2204         * regcprop.c (copy_value, maybe_mode_change): Likewise.
2205         (copyprop_hardreg_forward_1): Likewise.
2206         * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
2207         (record_subregs_of_mode): Likewise.
2208         * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
2209         * reload.c (operands_match_p): Handle 'p'.
2210         (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
2211         * reload1.c (alter_reg, choose_reload_regs): Likewise.
2212         (compute_reload_subreg_offset): Likewise, and return an poly_int64.
2213         * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
2214         (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
2215         (simplify_const_poly_int_tests<N>::run): Likewise.
2216         (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
2217         a poly_uint64 rather than an unsigned int.
2218         * valtrack.c (debug_lowpart_subreg): Likewise.
2219         * var-tracking.c (var_lowpart): Likewise.
2220         (loc_cmp): Handle 'p'.
2222 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2223             Alan Hayward  <alan.hayward@arm.com>
2224             David Sherwood  <david.sherwood@arm.com>
2226         * ira.c (get_subreg_tracking_sizes): New function.
2227         (init_live_subregs): Take an integer size rather than a register.
2228         (build_insn_chain): Use get_subreg_tracking_sizes.  Update calls
2229         to init_live_subregs.
2231 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2232             Alan Hayward  <alan.hayward@arm.com>
2233             David Sherwood  <david.sherwood@arm.com>
2235         * expr.c (store_constructor_field): Change bitsize from a
2236         unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2237         HOST_WIDE_INT to a poly_int64.
2238         (store_constructor): Change size from a HOST_WIDE_INT to
2239         a poly_int64.
2240         (store_field): Likewise bitsize and bitpos.
2242 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2243             Alan Hayward  <alan.hayward@arm.com>
2244             David Sherwood  <david.sherwood@arm.com>
2246         * expmed.h (store_bit_field): Change bitregion_start and
2247         bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2248         * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2249         (store_bit_field_1, store_integral_bit_field, store_bit_field)
2250         (store_fixed_bit_field, store_split_bit_field): Likewise.
2251         * expr.c (store_constructor_field, store_field): Likewise.
2252         (optimize_bitfield_assignment_op): Likewise.  Make the same change
2253         to bitsize and bitpos.
2254         * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2255         and m_bitregion_end from HOST_WIDE_INT to poly_int64.  Make the
2256         same change in the constructor arguments.
2257         (get_best_mode): Change bitregion_start and bitregion_end from
2258         unsigned HOST_WIDE_INT to poly_uint64.
2259         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2260         Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2261         poly_int64.
2262         (bit_field_mode_iterator::next_mode): Update for new types
2263         of m_bitregion_start and m_bitregion_end.
2264         (get_best_mode): Change bitregion_start and bitregion_end from
2265         unsigned HOST_WIDE_INT to poly_uint64.
2267 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2268             Alan Hayward  <alan.hayward@arm.com>
2269             David Sherwood  <david.sherwood@arm.com>
2271         * rtl.h (simplify_gen_subreg): Add a temporary overload that
2272         accepts poly_uint64 offsets.
2273         * expmed.h (extract_bit_field): Take bitsize and bitnum as
2274         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2275         * expmed.c (lowpart_bit_field_p): Likewise.
2276         (extract_bit_field_as_subreg): New function, split out from...
2277         (extract_bit_field_1): ...here.  Take bitsize and bitnum as
2278         poly_uint64s rather than unsigned HOST_WIDE_INTs.  For vector
2279         extractions, check that BITSIZE matches the size of the extracted
2280         value and that BITNUM is an exact multiple of that size.
2281         If all else fails, try forcing the value into memory if
2282         BITNUM is variable, and adjusting the address so that the
2283         offset is constant.  Split the part that can only handle constant
2284         bitsize and bitnum out into...
2285         (extract_integral_bit_field): ...this new function.
2286         (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2287         rather than unsigned HOST_WIDE_INTs.
2289 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2290             Alan Hayward  <alan.hayward@arm.com>
2291             David Sherwood  <david.sherwood@arm.com>
2293         * expmed.h (store_bit_field): Take bitsize and bitnum as
2294         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2295         * expmed.c (simple_mem_bitfield_p): Likewise.  Add a parameter
2296         that returns the byte size.
2297         (store_bit_field_1): Take bitsize and bitnum as
2298         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Update call
2299         to simple_mem_bitfield_p.  Split the part that can only handle
2300         constant bitsize and bitnum out into...
2301         (store_integral_bit_field): ...this new function.
2302         (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2303         than unsigned HOST_WIDE_INTs.
2304         (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2306 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2307             Alan Hayward  <alan.hayward@arm.com>
2308             David Sherwood  <david.sherwood@arm.com>
2310         * lra-int.h (lra_reg): Change offset from int to poly_int64.
2311         (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2312         to poly_int64.
2313         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2314         update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2315         (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2316         offset as a poly_int64 rather than an int.
2317         * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2318         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2319         * lra-constraints.c (equiv_address_substitution): Track offsets
2320         as poly_int64s.
2321         (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2322         (curr_insn_transform): Handle the new form of sp_offset.
2323         * lra-eliminations.c (lra_elim_table): Change previous_offset
2324         and offset from HOST_WIDE_INT to poly_int64.
2325         (print_elim_table, update_reg_eliminate): Update accordingly.
2326         (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2327         (get_elimination): Update accordingly.
2328         (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2329         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2330         update_sp_offset from a HOST_WIDE_INT to a poly_int64.  Handle
2331         poly_int64 offsets generally.
2332         (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2333         (mark_not_eliminable, init_elimination): Update accordingly.
2334         (remove_reg_equal_offset_note): Return a bool and pass the new
2335         offset back by pointer as a poly_int64.
2336         * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2337         rather than a HOST_WIDE_INT.
2338         (do_remat): Track offsets poly_int64s.
2339         * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2341 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2342             Alan Hayward  <alan.hayward@arm.com>
2343             David Sherwood  <david.sherwood@arm.com>
2345         * rtl.h (mem_attrs): Add a default constructor.  Change size and
2346         offset from HOST_WIDE_INT to poly_int64.
2347         * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2348         (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2349         (widen_memory_access): Take the sizes and offsets as poly_int64s
2350         rather than HOST_WIDE_INTs.
2351         * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2352         (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2353         and ints.
2354         (adjust_offset_for_component_ref): Change the offset from a
2355         HOST_WIDE_INT to a poly_int64.
2356         (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2357         * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2358         * dce.c (find_call_stack_args): Likewise.
2359         * dse.c (record_store): Likewise.
2360         * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2361         * print-rtl.c (rtx_writer::print_rtx): Likewise.
2362         * read-rtl-function.c (test_loading_mem): Likewise.
2363         * rtlanal.c (may_trap_p_1): Likewise.
2364         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2365         * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2366         * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2367         (mem_attrs::mem_attrs): New function.
2368         (set_mem_attributes_minus_bitpos): Change bitpos from a
2369         HOST_WIDE_INT to poly_int64.
2370         (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2371         (clear_mem_offset, clear_mem_size, change_address)
2372         (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2373         initialize mem_attrs.
2374         (set_mem_offset, set_mem_size, adjust_address_1)
2375         (adjust_automodify_address_1, offset_address, widen_memory_access):
2376         Likewise.  Take poly_int64s rather than HOST_WIDE_INT.
2378 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2379             Alan Hayward  <alan.hayward@arm.com>
2380             David Sherwood  <david.sherwood@arm.com>
2382         * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2383         as poly_int64s rather than HOST_WIDE_INTs.  Use a size of -1
2384         rather than 0 to represent an unknown size.  Assert that the size
2385         is known when the mode isn't BLKmode.
2386         (may_trap_p_1): Use -1 for unknown sizes.
2387         (rtx_addr_can_trap_p): Likewise.  Pass BLKmode rather than VOIDmode.
2389 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2390             Alan Hayward  <alan.hayward@arm.com>
2391             David Sherwood  <david.sherwood@arm.com>
2393         * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2394         to poly_int64.  Update commentary for positions_needed.large.
2395         (read_info_type): Change offset and width from HOST_WIDE_INT
2396         to poly_int64.
2397         (set_usage_bits): Likewise.
2398         (canon_address): Return the offset as a poly_int64 rather than
2399         a HOST_WIDE_INT.  Use strip_offset_and_add.
2400         (set_all_positions_unneeded, any_positions_needed_p): Use
2401         positions_needed.large to track stores with non-constant widths.
2402         (all_positions_needed_p): Likewise.  Take the offset and width
2403         as poly_int64s rather than ints.  Assert that rhs is nonnull.
2404         (record_store): Cope with non-constant offsets and widths.
2405         Nullify the rhs of an earlier store if we can't tell which bytes
2406         of it are needed.
2407         (find_shift_sequence): Take the access_size and shift as poly_int64s
2408         rather than ints.
2409         (get_stored_val): Take the read_offset and read_width as poly_int64s
2410         rather than HOST_WIDE_INTs.
2411         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2412         non-constant offsets and widths.
2414 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2415             Alan Hayward  <alan.hayward@arm.com>
2416             David Sherwood  <david.sherwood@arm.com>
2418         * inchash.h (inchash::hash::add_poly_int): New function.
2419         * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2420         Use poly_int64 rather than HOST_WIDE_INT.
2421         (ao_ref::max_size_known_p): New function.
2422         * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2423         rather than HOST_WIDE_INT.
2424         * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2425         to temporaries until its interface is adjusted to match.
2426         (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2427         (aliasing_component_refs_p, decl_refs_may_alias_p)
2428         (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2429         the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2430         (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2431         ao_ref fields.
2432         * alias.c (ao_ref_from_mem): Likewise.
2433         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2434         * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2435         (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2436         (maybe_trim_complex_store, maybe_trim_constructor_store)
2437         (live_bytes_read, dse_classify_store): Likewise.
2438         * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2439         (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2440         (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2441         (vn_reference_lookup_3): Likewise.
2442         * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
2444 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2445             Alan Hayward  <alan.hayward@arm.com>
2446             David Sherwood  <david.sherwood@arm.com>
2448         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2449         (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2450         instead of ranges_overlap_p.
2452 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2453             Alan Hayward  <alan.hayward@arm.com>
2454             David Sherwood  <david.sherwood@arm.com>
2456         * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2457         sizes as poly_int64s rather than HOST_WIDE_INTs.
2459 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2460             Alan Hayward  <alan.hayward@arm.com>
2461             David Sherwood  <david.sherwood@arm.com>
2463         * gimple-fold.h (fold_ctor_reference): Take the offset and size
2464         as poly_uint64 rather than unsigned HOST_WIDE_INT.
2465         * gimple-fold.c (fold_ctor_reference): Likewise.
2467 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2468             Alan Hayward  <alan.hayward@arm.com>
2469             David Sherwood  <david.sherwood@arm.com>
2471         * target.def (dwarf_poly_indeterminate_value): New hook.
2472         * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2473         * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2474         * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2475         * doc/tm.texi: Regenerate.
2476         * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2477         offset as a poly_int64.
2478         * dwarf2out.c (new_reg_loc_descr): Move later in file.  Take the
2479         offset as a poly_int64.
2480         (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2481         Take the offset as a poly_int64.
2482         (build_cfa_loc): Likewise.  Use loc_descr_plus_const.
2483         (frame_pointer_fb_offset): Change to a poly_int64.
2484         (int_loc_descriptor): Take the offset as a poly_int64.  Use
2485         targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2486         (based_loc_descr): Take the offset as a poly_int64.
2487         Use strip_offset_and_add to handle (plus X (const)).
2488         Use new_reg_loc_descr instead of an open-coded version of the
2489         previous implementation.
2490         (mem_loc_descriptor): Handle CONST_POLY_INT.
2491         (compute_frame_pointer_to_fb_displacement): Take the offset as a
2492         poly_int64.  Use strip_offset_and_add to handle (plus X (const)).
2494 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2495             Alan Hayward  <alan.hayward@arm.com>
2496             David Sherwood  <david.sherwood@arm.com>
2498         * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2499         (gen_rtx_REG_offset): Take the offset as a poly_int64.
2500         * inchash.h (inchash::hash::add_poly_hwi): New function.
2501         * gengtype.c (main): Register poly_int64.
2502         * emit-rtl.c (reg_attr_hasher::hash): Use inchash.  Treat the
2503         offset as a poly_int.
2504         (reg_attr_hasher::equal): Use must_eq to compare offsets.
2505         (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2506         offset as a poly_int64.
2507         (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2508         * print-rtl.c (print_poly_int): New function.
2509         (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2510         a poly_int.
2511         * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2512         functions.
2513         (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2514         (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2515         (vt_get_decl_and_offset): Return the offset as a poly_int64.
2516         Enforce track_offset_p for parts of a PARALLEL.
2517         (vt_add_function_parameter): Use const_offset for the final
2518         offset to track.  Use get_tracked_reg_offset for the parts
2519         of a PARALLEL.
2521 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2522             Alan Hayward  <alan.hayward@arm.com>
2523             David Sherwood  <david.sherwood@arm.com>
2525         * target.def (truly_noop_truncation): Take poly_uint64s instead of
2526         unsigned ints.  Change default to hook_bool_puint64_puint64_true.
2527         * doc/tm.texi: Regenerate.
2528         * hooks.h (hook_bool_uint_uint_true): Delete.
2529         (hook_bool_puint64_puint64_true): Declare.
2530         * hooks.c (hook_bool_uint_uint_true): Delete.
2531         (hook_bool_puint64_puint64_true): New function.
2532         * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2533         instead of unsigned ints.
2534         * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2535         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
2537 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2538             Alan Hayward  <alan.hayward@arm.com>
2539             David Sherwood  <david.sherwood@arm.com>
2541         * optabs.h (expand_operand): Add an int_value field.
2542         (create_expand_operand): Add an int_value parameter and use it
2543         to initialize the new expand_operand field.
2544         (create_integer_operand): Replace with a declaration of a function
2545         that accepts poly_int64s.  Move the implementation to...
2546         * optabs.c (create_integer_operand): ...here.
2547         (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2548         the mode preserves the value of int_value, instead of calling
2549         const_int_operand on the rtx.  Use gen_int_mode to generate
2550         the new rtx.
2552 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2553             Alan Hayward  <alan.hayward@arm.com>
2554             David Sherwood  <david.sherwood@arm.com>
2556         * dumpfile.h (dump_dec): Declare.
2557         * dumpfile.c (dump_dec): New function.
2558         * pretty-print.h (pp_wide_integer): Turn into a function and
2559         declare a poly_int version.
2560         * pretty-print.c (pp_wide_integer): New function for poly_ints.
2562 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2563             Alan Hayward  <alan.hayward@arm.com>
2564             David Sherwood  <david.sherwood@arm.com>
2566         * doc/generic.texi (POLY_INT_CST): Document.
2567         * tree.def (POLY_INT_CST): New tree code.
2568         * treestruct.def (TS_POLY_INT_CST): New tree layout.
2569         * tree-core.h (tree_poly_int_cst): New struct.
2570         (tree_node): Add a poly_int_cst field.
2571         * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2572         (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2573         instead of a wide_int_ref.
2574         (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2575         of a HOST_WIDE_INT.
2576         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2577         instead of an unsigned HOST_WIDE_INT.
2578         (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2579         (ptrdiff_tree_p): Declare.
2580         (tree_to_poly_int64, tree_to_poly_uint64): Likewise.  Provide
2581         extern inline implementations if the target doesn't use POLY_INT_CST.
2582         (poly_int_tree_p): New function.
2583         (wi::unextended_tree): New class.
2584         (wi::int_traits <unextended_tree>): New override.
2585         (wi::extended_tree): Add a default constructor.
2586         (wi::extended_tree::get_tree): New function.
2587         (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2588         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2589         (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2590         (wi::tree_to_poly_wide_ref): New typedefs.
2591         (wi::ints_for): Provide overloads for extended_tree and
2592         unextended_tree.
2593         (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2594         (wi::to_wide): New functions.
2595         (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2596         * tree.c (poly_int_cst_hasher): New struct.
2597         (poly_int_cst_hash_table): New variable.
2598         (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2599         (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2600         POLY_INT_CST.
2601         (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2602         (init_ttree): Initialize poly_int_cst_hash_table.
2603         (build_int_cst, build_int_cst_type, build_invariant_address): Take
2604         a poly_int64 instead of a HOST_WIDE_INT.
2605         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2606         instead of an unsigned HOST_WIDE_INT.
2607         (wide_int_to_tree): Rename to...
2608         (wide_int_to_tree_1): ...this.
2609         (build_new_poly_int_cst, build_poly_int_cst): New functions.
2610         (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2611         (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2612         (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2613         (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2614         * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2615         TS_POLY_INT_CST.
2616         * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2617         (streamer_read_tree_body): Likewise.
2618         * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2619         (streamer_write_tree_body): Likewise.
2620         * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2621         * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2622         * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2623         * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2624         * gimple-expr.h (is_gimple_constant): Likewise.
2625         * gimplify.c (maybe_with_size_expr): Likewise.
2626         * print-tree.c (print_node): Likewise.
2627         * tree-data-ref.c (data_ref_compare_tree): Likewise.
2628         * tree-pretty-print.c (dump_generic_node): Likewise.
2629         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2630         * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2631         * tree-vrp.c (compare_values_warnv): Likewise.
2632         * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2633         (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2634         (force_expr_to_var_cost): Likewise.
2635         * tree-ssa-loop.c (for_each_index): Likewise.
2636         * fold-const.h (build_invariant_address, size_int_kind): Take a
2637         poly_int64 instead of a HOST_WIDE_INT.
2638         * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2639         (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2640         POLY_INT_CST.
2641         (size_binop_loc): Likewise.  Allow int_const_binop_1 to fail.
2642         (int_const_binop_2): New function, split out from...
2643         (int_const_binop_1): ...here.  Handle POLY_INT_CST.
2644         (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2645         * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2646         * gimple-ssa-strength-reduction.c (slsr_process_add)
2647         (slsr_process_mul): Check for INTEGER_CSTs before using them
2648         as candidates.
2649         * stor-layout.c (bits_from_bytes): New function.
2650         (bit_from_pos): Use it.
2651         (layout_type): Likewise.  For vectors, multiply the TYPE_SIZE_UNIT
2652         by BITS_PER_UNIT to get the TYPE_SIZE.
2653         * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2654         MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2656 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2657             Alan Hayward  <alan.hayward@arm.com>
2658             David Sherwood  <david.sherwood@arm.com>
2660         * doc/rtl.texi (const_poly_int): Document.  Also document the
2661         rtl sharing behavior.
2662         * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2663         * rtl.h (const_poly_int_def): New struct.
2664         (rtx_def::u): Add a cpi field.
2665         (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2666         (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2667         (wi::rtx_to_poly_wide_ref): New typedef
2668         (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2669         (poly_int_rtx_p): New functions.
2670         (trunc_int_for_mode): Declare a poly_int64 version.
2671         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2672         (immed_wide_int_const): Take a poly_wide_int_ref rather than
2673         a wide_int_ref.
2674         (strip_offset): Declare.
2675         (strip_offset_and_add): New function.
2676         * rtl.def (CONST_POLY_INT): New rtx code.
2677         * rtl.c (rtx_size): Handle CONST_POLY_INT.
2678         (shared_const_p): Use poly_int_rtx_p.
2679         * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2680         HOST_WIDE_INT.
2681         (gen_int_shift_amount): Likewise.
2682         * emit-rtl.c (const_poly_int_hasher): New class.
2683         (const_poly_int_htab): New variable.
2684         (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2685         (const_poly_int_hasher::hash): New function.
2686         (const_poly_int_hasher::equal): Likewise.
2687         (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2688         (immed_wide_int_const): Rename to...
2689         (immed_wide_int_const_1): ...this and make static.
2690         (immed_wide_int_const): New function, taking a poly_wide_int_ref
2691         instead of a wide_int_ref.
2692         (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2693         (gen_lowpart_common): Handle CONST_POLY_INT.
2694         * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2695         * cselib.c (cselib_hash_rtx): Likewise.
2696         * dwarf2out.c (const_ok_for_output_1): Likewise.
2697         * expr.c (convert_modes): Likewise.
2698         * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2699         * rtlhash.c (add_rtx): Likewise.
2700         * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2701         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2702         Handle existing CONST_POLY_INT rtxes.
2703         * expmed.h (expand_shift): Take a poly_int64 instead of a
2704         HOST_WIDE_INT.
2705         * expmed.c (expand_shift): Likewise.
2706         * rtlanal.c (strip_offset): New function.
2707         (commutative_operand_precedence): Give CONST_POLY_INT the same
2708         precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2709         and CONST_INT.
2710         * rtl-tests.c (const_poly_int_tests): New struct.
2711         (rtl_tests_c_tests): Use it.
2712         * simplify-rtx.c (simplify_const_unary_operation): Handle
2713         CONST_POLY_INT.
2714         (simplify_const_binary_operation): Likewise.
2715         (simplify_binary_operation_1): Fold additions of symbolic constants
2716         and CONST_POLY_INTs.
2717         (simplify_subreg): Handle extensions and truncations of
2718         CONST_POLY_INTs.
2719         (simplify_const_poly_int_tests): New struct.
2720         (simplify_rtx_c_tests): Use it.
2721         * wide-int.h (storage_ref): Add default constructor.
2722         (wide_int_ref_storage): Likewise.
2723         (trailing_wide_ints): Use GTY((user)).
2724         (trailing_wide_ints::operator[]): Add a const version.
2725         (trailing_wide_ints::get_precision): New function.
2726         (trailing_wide_ints::extra_size): Likewise.
2728 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2729             Alan Hayward  <alan.hayward@arm.com>
2730             David Sherwood  <david.sherwood@arm.com>
2732         * emit-rtl.h (gen_int_shift_amount): Declare.
2733         * emit-rtl.c (gen_int_shift_amount): New function.
2734         * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2735         instead of GEN_INT.
2736         * calls.c (shift_return_value): Likewise.
2737         * cse.c (fold_rtx): Likewise.
2738         * dse.c (find_shift_sequence): Likewise.
2739         * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2740         (expand_shift, expand_smod_pow2): Likewise.
2741         * lower-subreg.c (shift_cost): Likewise.
2742         * optabs.c (expand_superword_shift, expand_doubleword_mult)
2743         (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2744         (expand_vec_perm_var): Likewise.
2745         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2746         (simplify_binary_operation_1): Likewise.
2747         * combine.c (try_combine, find_split_point, force_int_to_mode)
2748         (simplify_shift_const_1, simplify_shift_const): Likewise.
2749         (change_zero_ext): Likewise.  Use simplify_gen_binary.
2751 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2753         * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2755 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2757         * doc/invoke.texi (ARM Options): Document accepted extension options
2758         for -march=armv8.3-a.
2760 2017-12-20  Richard Earnshaw  <rearnsha@arm.com>
2762         PR target/83105
2763         * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2764         or --with-float=softfp, set the default CPU to arm10e.
2766 2017-12-20  Eric Botcazou  <ebotcazou@adacore.com>
2768         * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2769         * config/visium/predicates.md (const_shift_operand): Likewise.
2770         * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2771         (visium_legitimize_reload_address): Likewise.
2773 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2775         PR target/82975
2776         * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2778 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
2780         PR c++/83490
2781         * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2783 2017-12-20  Julia Koval  <julia.koval@intel.com>
2785         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2786         OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2787         (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2788         * config.gcc: Include vpclmulqdqintrin.h.
2789         * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2790         * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2791         -mvpclmulqdq.
2792         * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2793         __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2794         * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2795         * config/i386/i386.c (isa2_opts): Add -mcx16.
2796         (isa_opts): Add -mpclmulqdq, remove -mcx16.
2797         (ix86_option_override_internal): Move mcx16 to flags2.
2798         (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2799         (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2800         * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2801         * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2802         * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2803         * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2804         * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2805         _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2806         * doc/invoke.texi: Add -mvpclmulqdq.
2808 2017-12-20  Tom de Vries  <tom@codesourcery.com>
2810         PR middle-end/83423
2811         * config/i386/i386.c (ix86_static_chain): Move
2812         DECL_STATIC_CHAIN test ...
2813         * calls.c (rtx_for_static_chain): ... here.  New function.
2814         * calls.h (rtx_for_static_chain): Declare.
2815         * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2816         instead of targetm.calls.static_chain.
2817         * df-scan.c (df_get_entry_block_def_set): Same.
2819 2017-12-19  Tom de Vries  <tom@codesourcery.com>
2821         PR tree-optimization/83493
2822         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2823         and lb.
2825 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2827         * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2828         inform with hardcoded english plural handling.
2830 2017-12-18  Jeff Law  <law@redhat.com>
2832         PR tree-optimization/83477
2833         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2834         a non-virtual PHI, always push a new range.
2836 2017-12-19  Martin Sebor  <msebor@redhat.com>
2838         PR middle-end/77608
2839         * builtins.c (compute_objsize): Handle non-constant offsets.
2841 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2843         PR tree-optimization/83444
2844         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2845         character load case, if get_stridx on MEM_REF's operand doesn't
2846         look usable, retry with get_addr_stridx.
2848 2017-12-19  Alexandre Oliva  <aoliva@redhat.com>
2850         PR debug/83422
2851         * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2852         (variable_tracking_main_1): Keep markers even when VTA fails.
2854         PR bootstrap/83396
2855         * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2856         even if there are markers before it.
2857         * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2859 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2861         * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2862         typical order conditions.
2863         * sel-sched.c (extract_new_fences_from): Likewise.
2864         * config/visium/constraints.md (J, K, L): Likewise.
2865         * config/visium/predicates.md (const_shift_operand): Likewise.
2866         * config/visium/visium.c (visium_legitimize_address,
2867         visium_legitimize_reload_address): Likewise.
2868         * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2869         * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2870         * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2871         * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2872         SET_DUMP_DETAIL): Likewise.
2873         * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2874         * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2875         avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2876         avr_set_core_architecture, avr_set_current_function,
2877         avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2878         output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2879         avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2880         avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2881         avr_map_decompose, avr_fold_builtin): Likewise.
2882         * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2883         * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2884         * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2885         * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2886         * config/m32c/m32c.c (m32c_conditional_register_usage,
2887         m32c_address_cost): Likewise.
2888         * config/m32c/predicates.md (shiftcount_operand,
2889         longshiftcount_operand): Likewise.
2890         * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2891         * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2892         can_use_cdx_ldstw): Likewise.
2893         * config/nios2/nios2.h (CDX_REG_P): Likewise.
2894         * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2895         Likewise.
2896         * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2897         * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
2898         * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
2899         * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
2900         * config/vax/vax.c (adjacent_operands_p): Likewise.
2901         * config/ft32/constraints.md (L, b, KA): Likewise.
2902         * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
2903         Likewise.
2904         * cfgexpand.c (expand_stack_alignment): Likewise.
2905         * gcse.c (insert_expr_in_table): Likewise.
2906         * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
2907         * cgraphunit.c (cgraph_node::expand): Likewise.
2908         * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
2909         * emit-rtl.c (add_insn): Likewise.
2910         * input.c (dump_location_info): Likewise.
2911         * passes.c (NEXT_PASS): Likewise.
2912         * read-rtl-function.c (parse_note_insn_name,
2913         function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
2914         Likewise.
2915         * sched-rgn.c (sched_rgn_init): Likewise.
2916         * diagnostic-show-locus.c (layout::show_ruler): Likewise.
2917         * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
2918         if_then_else_cond, simplify_shift_const_1, simplify_comparison):
2919         Likewise.
2920         * explow.c (eliminate_constant_term): Likewise.
2921         * final.c (leaf_renumber_regs_insn): Likewise.
2922         * cfgrtl.c (print_rtl_with_bb): Likewise.
2923         * genhooks.c (emit_init_macros): Likewise.
2924         * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
2925         * tree-data-ref.c (conflict_fn): Likewise.
2926         * selftest.c (assert_streq): Likewise.
2927         * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
2928         * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
2929         fold_binary_loc, multiple_of_p): Likewise.
2930         * reload.c (push_reload, find_equiv_reg): Likewise.
2931         * et-forest.c (et_nca, et_below): Likewise.
2932         * dbxout.c (dbxout_symbol_location): Likewise.
2933         * reorg.c (relax_delay_slots): Likewise.
2934         * dojump.c (do_compare_rtx_and_jump): Likewise.
2935         * gengtype-parse.c (type): Likewise.
2936         * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
2937         simplify_const_relational_operation): Likewise.
2938         * reload1.c (do_output_reload): Likewise.
2939         * dumpfile.c (get_dump_file_info_by_switch): Likewise.
2940         * gengtype.c (type_for_name): Likewise.
2941         * gimple-ssa-sprintf.c (format_directive): Likewise.
2943 2017-12-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2945         PR target/82975
2946         * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
2947         accessing it.  Adjust comment.
2949 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2951         PR middle-end/81914
2952         * predict.c (zero_one_minusone): New function.
2953         (apply_return_prediction): Avoid return prediction for functions
2954         returning only -1, 0 and 1 values, unless they only return -1 and 0
2955         or 0 and 1.
2957 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2959         * config/arc/arc.c (legitimate_scaled_address_p): Clean
2960         fall-through warning.
2961         (arc_compute_frame_size): Remove unused variables.
2962         (arc_print_operand): Fix fprintif format.
2963         (arc_can_follow_jump): Clean fall-through warning.
2965 2017-12-19  Marek Polacek  <polacek@redhat.com>
2967         PR c++/83489
2968         * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
2969         on an error node.
2971 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2973         * config/arc/arc.c (overriderregs): New variable.
2974         (arc_override_options): Track fixed/call saved/call options.
2975         (arc_conditional_register_usage): Check against overrideregs
2976         variable whenever we change register properties.
2978 2017-12-19  Nathan Sidwell  <nathan@acm.org>
2980         * opts.c (finish_options): Don't prefix dump_base_name if it
2981         already contains directories.
2983 2017-12-19  Martin Liska  <mliska@suse.cz>
2985         PR rtl-optimization/82675
2986         * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
2987         more element in sbitmap.
2989 2017-12-19  Paolo Carlini  <paolo.carlini@oracle.com>
2991         * gimplify.c (gimplify_expr): Use error_operand_p.
2993 2017-12-19  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2995         PR target/83387
2996         * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
2997         multilib.
2999 2017-12-19  Daniel Cederman  <cederman@gaisler.com>
3001         * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
3002         the jump is to a label.
3004 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
3006         PR tree-optimization/83444
3007         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
3008         character loads.
3010         PR ipa/82801
3011         PR ipa/83346
3012         * ipa-inline.c (flatten_remove_node_hook): New function.
3013         (ipa_inline): Keep only nodes with flatten attribute at the end of
3014         the array in the order from ipa_reverse_postorder, only walk that
3015         portion of array for flattening, if there is more than one such
3016         node, temporarily register a removal hook and ignore removed nodes.
3018         PR tree-optimization/80631
3019         * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
3020         induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
3021         IFN_REDUC_MAX or IFN_REDUC_MIN.
3023 2017-12-18  Jeff Law  <law@redhat.com>
3025         * tree-ssa-dom.c (record_equivalences_from_phis): Do not
3026         record symbolic equivalences from backedges in the CFG.
3028         Revert
3029         2017-11-19  Jeff Law  <law@redhat.com>
3031         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
3032         of degenerates resulting from ignoring an edge.
3034 2017-12-18  Martin Sebor  <msebor@redhat.com>
3036         PR middle-end/83373
3037         PR tree-optimization/78450
3038         * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
3039         (handle_builtin_strlen): Call it.
3041 2017-12-18  Segher Boessenkool  <segher@kernel.crashing.org>
3043         PR rtl-optimization/83424
3044         * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
3046 2017-12-18  Marek Polacek  <polacek@redhat.com>
3048         PR middle-end/83463
3049         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
3050         Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
3051         values.
3053 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3055         * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
3056         (umaddsidi4, umaddsidi_split): Likewise.
3058 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
3060         * config/arc/arc.c (arc_legitimate_constant_p): Always check all
3061         constants.
3063 2017-12-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
3065         PR target/83420
3066         * config/s390/s390.c (s390_option_override): Avoid strncpy.
3068 2017-12-18  Richard Biener  <rguenther@suse.de>
3070         PR tree-optimization/81877
3071         * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
3072         (outermost_indep_loop): Adjust.
3073         (ref_indep_loop_p_1): Likewise.  Remove safelen handling again.
3074         (can_sm_ref_p): Adjust.
3076 2017-12-18  Richard Biener  <rguenther@suse.de>
3078         PR middle-end/77291
3079         * tree.c (array_at_struct_end_p): Return true if the underlying
3080         object has space for at least one element in excess of what
3081         the array domain specifies.
3083 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3085         * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
3086         example to avoid overfull hbox.
3087         * doc/invoke.texi (Option Summary): Add missing @gol.
3088         (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
3089         overfull hbox.
3091 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3092             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3094         * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
3095         to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
3096         (AArch64 Options): Document -mverbose-cost-dump.
3097         (ARM Options): Likewise, plus -mflip-thumb.
3099 2017-12-17  Martin Sebor  <msebor@redhat.com>
3101         PR bootstrap/83446
3102         * gimple-ssa-warn-restrict.c
3103         (builtin_memref::offset_out_of_bounds): Correct the handling of
3104         anti-ranges.
3106 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3108         * doc/invoke.texi (ARC Options): Add missing -mlra entry.
3110 2017-12-17  John David Anglin  <danglin@gcc.gnu.org>
3112         * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
3114 2017-12-17  Markus Trippelsdorf  <markus@trippelsdorf.de>
3116         * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
3117         latencies.
3119 2017-12-16  Sandra Loosemore  <sandra@codesourcery.com>
3121         * doc/invoke.texi: Fix some typos.
3123 2017-12-16  Martin Sebor  <msebor@redhat.com>
3125         PR tree-optimization/78918
3126         * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
3127         * builtins.c (check_sizes): Rename...
3128         (check_access): ...to this.  Rename function arguments for clarity.
3129         (check_memop_sizes): Adjust names.
3130         (expand_builtin_memchr, expand_builtin_memcpy): Same.
3131         (expand_builtin_memmove, expand_builtin_mempcpy): Same.
3132         (expand_builtin_strcat, expand_builtin_stpncpy): Same.
3133         (check_strncat_sizes, expand_builtin_strncat): Same.
3134         (expand_builtin_strncpy, expand_builtin_memset): Same.
3135         (expand_builtin_bzero, expand_builtin_memcmp): Same.
3136         (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
3137         (maybe_emit_sprintf_chk_warning): Same.
3138         (expand_builtin_strcpy): Adjust.
3139         (expand_builtin_stpcpy): Same.
3140         (expand_builtin_with_bounds): Detect out-of-bounds accesses
3141         in pointer-checking forms of memcpy, memmove, and mempcpy.
3142         (gcall_to_tree_minimal, max_object_size): Define new functions.
3143         * builtins.h (max_object_size): Declare.
3144         * calls.c (alloc_max_size): Call max_object_size instead of
3145         hardcoding ssizetype limit.
3146         (get_size_range): Handle new argument.
3147         * calls.h (get_size_range): Add a new argument.
3148         * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
3149         * doc/invoke.texi (-Wrestrict): Adjust, add example.
3150         * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
3151         operations.
3152         (gimple_fold_builtin_memory_chk): Same.
3153         (gimple_fold_builtin_stxcpy_chk): New function.
3154         * gimple-ssa-warn-restrict.c: New source.
3155         * gimple-ssa-warn-restrict.h: New header.
3156         * gimple.c (gimple_build_call_from_tree): Propagate location.
3157         * passes.def (pass_warn_restrict): Add new pass.
3158         * tree-pass.h (make_pass_warn_restrict): Declare.
3159         * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
3160         operations.
3161         (handle_builtin_strcat): Same.
3162         (strlen_optimize_stmt): Rename...
3163         (strlen_check_and_optimize_stmt): ...to this.  Handle strncat,
3164         stpncpy, strncpy, and their checking forms.
3166 2017-12-16  Jan Hubicka  <hubicka@ucw.cz>
3168         PR rtl-optimization/82849
3169         * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
3170         and get_max_loop_iterations_int.
3172 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3173             Alan Hayward  <alan.hayward@arm.com>
3174             David Sherwood  <david.sherwood@arm.com>
3176         * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
3177         (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
3178         size as a poly_uint64.
3179         (mode_for_vector, mode_for_int_vector): Take the number of vector
3180         elements as a poly_uint64.
3181         * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
3182         size as a poly_uint64.
3183         (mode_for_vector, mode_for_int_vector): Take the number of vector
3184         elements as a poly_uint64.
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 (MACRO_MODE): New macro.
3191         * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
3192         * targhooks.c (default_libcall_value, default_secondary_reload)
3193         (default_memory_move_cost, default_register_move_cost)
3194         (default_class_max_nregs): Likewise.
3196 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3197             Alan Hayward  <alan.hayward@arm.com>
3198             David Sherwood  <david.sherwood@arm.com>
3200         * doc/sourcebuild.texi: Document IN_TARGET_CODE.
3201         * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
3202         target C file.
3203         * genautomata.c (main): Likewise.
3204         * genconditions.c (write_header): Likewise.
3205         * genemit.c (main): Likewise.
3206         * genextract.c (print_header): Likewise.
3207         * genopinit.c (main): Likewise.
3208         * genoutput.c (output_prologue): Likewise.
3209         * genpeep.c (main): Likewise.
3210         * genpreds.c (write_insn_preds_c): Likewise.
3211         * genrecog.c (writer_header): Likewise.
3212         * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
3213         * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
3214         * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
3215         * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
3216         * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
3217         * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
3218         * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
3219         * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
3220         * config/arc/arc.c (IN_TARGET_CODE): Likewise.
3221         * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
3222         * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
3223         * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3224         * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3225         * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3226         * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3227         * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3228         * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3229         * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3230         * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3231         * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3232         * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3233         * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3234         * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3235         * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3236         * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3237         * config/darwin.c (IN_TARGET_CODE): Likewise.
3238         * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3239         * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3240         * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3241         * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3242         * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3243         * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3244         * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3245         * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3246         * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3247         * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3248         * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3249         * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3250         * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3251         * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3252         * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3253         * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3254         * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3255         * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3256         * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3257         * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3258         * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3259         * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3260         * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3261         * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3262         * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3263         * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3264         * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3265         * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3266         * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3267         * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3268         * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3269         * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3270         * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3271         * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3272         * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3273         * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3274         * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3275         * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3276         * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3277         * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3278         * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3279         * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3280         * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3281         * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3282         * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3283         * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3284         * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3285         * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3286         * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3287         * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3288         * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3289         * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3290         * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3291         * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3292         * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3293         * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3294         * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3295         * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3296         * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3297         * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3298         * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3299         * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3300         * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3301         * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3302         * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3303         * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3304         * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3305         * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3306         * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3307         * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3308         * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3309         * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3310         * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3311         * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3312         * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3313         * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3314         * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3315         * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3316         * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3317         * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3318         * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3319         * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3320         * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3321         * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3322         * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3323         * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3324         * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3325         * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3326         * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3327         * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3328         * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3329         * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3330         * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3331         * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3332         * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3333         * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3334         * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3335         * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3336         * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3337         * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3338         * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3339         * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3340         * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3341         * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3342         * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3343         * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3345 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3346             Alan Hayward  <alan.hayward@arm.com>
3347             David Sherwood  <david.sherwood@arm.com>
3349         * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3350         checks for MEM_REF.
3352 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3353             Alan Hayward  <alan.hayward@arm.com>
3354             David Sherwood  <david.sherwood@arm.com>
3356         * doc/generic.texi (VEC_SERIES_EXPR): Document.
3357         * doc/md.texi (vec_series@var{m}): Document.
3358         * tree.def (VEC_SERIES_EXPR): New tree code.
3359         * tree.h (build_vec_series): Declare.
3360         * tree.c (build_vec_series): New function.
3361         * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3362         * tree-pretty-print.c (dump_generic_node): Likewise.
3363         * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3364         * tree-inline.c (estimate_operator_cost): Likewise.
3365         * expr.c (expand_expr_real_2): Likewise.
3366         * optabs-tree.c (optab_for_tree_code): Likewise.
3367         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3368         * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3369         * expmed.c (make_tree): Handle VEC_SERIES.
3370         * optabs.def (vec_series_optab): New optab.
3371         * optabs.h (expand_vec_series_expr): Declare.
3372         * optabs.c (expand_vec_series_expr): New function.
3373         * tree-vect-generic.c (expand_vector_operations_1): Check that
3374         the operands also have vector type.
3376 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3377             Alan Hayward  <alan.hawyard@arm.com>
3378             David Sherwood  <david.sherwood@arm.com>
3380         * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3381         (VEC_COND_EXPR): Add missing @tindex.
3382         * doc/md.texi (vec_duplicate@var{m}): Document.
3383         * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3384         * tree.c (build_vector_from_val): Add stubbed-out handling of
3385         variable-length vectors, using VEC_DUPLICATE_EXPR.
3386         (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3387         * cfgexpand.c (expand_debug_expr): Likewise.
3388         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3389         * tree-inline.c (estimate_operator_cost): Likewise.
3390         * tree-pretty-print.c (dump_generic_node): Likewise.
3391         * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3392         * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3393         (test_vec_duplicate_folding): New function.
3394         (fold_const_c_tests): Call it.
3395         * optabs.def (vec_duplicate_optab): New optab.
3396         * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3397         * optabs.h (expand_vector_broadcast): Declare.
3398         * optabs.c (expand_vector_broadcast): Make non-static.  Try using
3399         vec_duplicate_optab.
3400         * expr.c (store_constructor): Try using vec_duplicate_optab for
3401         uniform vectors.
3402         (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3404 2017-12-15  Markus Trippelsdorf  <markus@trippelsdorf.de>
3406         PR target/83358
3407         * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3408         div/mod latencies a bit.
3410 2017-12-15  Jeff Law  <law@redhat.com>
3412         PR tree-optimization/36550
3413         * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3414         (mark_threaded_blocks): Rewrite code to avoid block copying when
3415         optimizing for size.  Don't pessimize blocks which will be
3416         copied, but all the statements will be dead.
3418 2017-12-15  Alexandre Oliva  <aoliva@redhat.com>
3420         PR tree-optimization/81165
3421         * tree-ssa-threadupdate.c (uses_in_bb): New.
3422         (estimate_threading_killed_stmts): New.
3423         * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3424         * tree-ssa-threadedge.c
3425         (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3426         when its hit.
3428 2017-12-15  Jeff Law  <law@redhat.com>
3430         PR tree-optimization/83410
3431         * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3432         threads when parallelizing loops.
3434 2017-12-15  Jakub Jelinek  <jakub@redhat.com>
3436         * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3437         handler fields.
3438         * config/alpha/alpha.c (vms_attribute_table): Swap
3439         affects_type_identity and handler fields, adjust comments.
3440         * config/mips/mips.c (mips_attribute_table): Likewise.
3441         * config/visium/visium.c (visium_attribute_table): Likewise.
3442         * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3443         * config/microblaze/microblaze.c (microblaze_attribute_table):
3444         Likewise.
3445         * config/spu/spu.c (spu_attribute_table): Likewise.
3446         * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3447         * config/arc/arc.c (arc_attribute_table): Likewise.
3448         * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3449         * config/v850/v850.c (v850_handle_interrupt_attribute,
3450         v850_handle_data_area_attribute): Formatting fixes.
3451         (v850_attribute_table): Swap affects_type_identity and handler
3452         fields, adjust comments.
3453         * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3454         * config/arm/arm.c (arm_attribute_table): Likewise.
3455         * config/avr/avr.c (avr_attribute_table): Likewise.
3456         * config/s390/s390.c (s390_attribute_table): Likewise.
3457         * config/sh/sh.c (sh_attribute_table): Likewise.
3458         * config/i386/i386.c (ix86_handle_cconv_attribute,
3459         ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3460         (ix86_attribute_table): Swap affects_type_identity and handler
3461         fields, adjust comments.
3462         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3463         * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3464         * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3465         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3466         * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3467         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3468         * config/rx/rx.c (rx_attribute_table): Likewise.
3469         * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3470         * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3471         * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3472         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3473         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3474         * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3475         * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3476         * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3477         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3478         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3479         * doc/plugins.texi (user_attr): Likewise.  Add NULL for
3480         exclude.
3481         * attribs.c (empty_attribute_table): Swap affects_type_identity and
3482         handler fields.
3483         (register_scoped_attributes, decl_attributes): Formatting fixes.
3485         PR tree-optimization/83269
3486         * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3487         subtraction in arg0's type if type is signed and arg0 is unsigned.
3488         Formatting fix.
3490         PR sanitizer/81281
3491         * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3492         convert? on @0 instead of convert.  Check type of @1, not @0.
3493         ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3494         convert? on @0 instead of convert.  Check type of @1, not @0.
3495         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3496         only optimize if either both @1 and @2 types are narrower
3497         precision, or both are wider or equal precision, and in the former
3498         case only if both have undefined overflow.
3500 2017-12-15  Richard Biener  <rguenther@suse.de>
3502         PR lto/83388
3503         * internal-fn.def (IFN_NOP): Add.
3504         * internal-fn.c (expand_NOP): Do nothing.
3505         * lto-streamer-in.c (input_function): Instead of removing
3506         sanitizer calls replace them with IFN_NOP calls.
3508 2017-12-15  Richard Sandiford  <richard.sandiford@linaro.org>
3509             Alan Hayward  <alan.hayward@arm.com>
3510             David Sherwood  <david.sherwood@arm.com>
3512         * dse.c (store_info, read_info_type): Replace begin and end with
3513         offset and width.
3514         (print_range): New function.
3515         (set_all_positions_unneeded, any_positions_needed_p)
3516         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3517         accordingly.
3518         (record_store): Likewise.  Optimize the case in which all positions
3519         are unneeded.
3520         (get_stored_val): Replace read_begin and read_end with read_offset
3521         and read_width.
3522         (replace_read): Update call accordingly.
3524 2017-12-15  Bin Cheng  <bin.cheng@arm.com>
3526         * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3527         (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3528         (loop_cand::loop_cand): Initialize above members.
3529         (loop_cand::supported_operations): Delete.
3530         (loop_cand::can_interchange_p): Inline above function.
3531         (loop_cand::classify_simple_reduction): Record number of constant
3532         initialized simple reductions.
3533         (should_interchange_loops): New parameters.  Check stmt cost of loops
3534         to be interchange.
3535         (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3536         Update call to should_interchange_loops.
3537         (should_interchange_loop_nest): Update call to
3538         should_interchange_loops.
3540 2017-12-15  Eric Botcazou  <ebotcazou@adacore.com>
3542         PR target/66488
3543         * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3544         * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3545         * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3547 2017-12-15  Julia Koval  <julia.koval@intel.com>
3549         * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3550         __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3551         New.
3552         * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3553         * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3554         _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3556 2017-12-15  Julia Koval  <julia.koval@intel.com>
3558         * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3559         __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3560         * config/i386/sse.md (vaesenc_<mode>): New pattern.
3561         * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3562         _mm_aesenc_epi128): New intrinsics.
3564 2017-12-15  Julia Koval  <julia.koval@intel.com>
3566         * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3567         __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3568         New.
3569         * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3570         * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3571         _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3573 2017-12-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3575         * gimple-ssa-strength-reduction.c (analyze_increments):
3576         Distinguish replacement costs for constant strides from those for
3577         unknown strides.
3579 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3581         * var-tracking.c (variable_tracking_main_1): Formatting fix.
3583 2017-12-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3585         * doc/invoke.texi: Document -Wcast-function-type.
3586         * recog.h (stored_funcptr): Change signature.
3587         * tree-dump.c (dump_node): Avoid warning.
3588         * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3590 2017-12-14  Qing Zhao  <qing.zhao@oracle.com>
3592         PR middle_end/79538
3593         * gimple-fold.c (get_range_strlen): Add the handling of non-member 
3594         array.
3596 2017-12-14  David Malcolm  <dmalcolm@redhat.com>
3598         PR tree-optimization/83312
3599         * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3600         * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3601         "val" param, and to cope with arbitrary basic blocks.
3602         (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3603         handle NULL_TREE for "val", dropping "bb" param.
3604         (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3605         drop "bb" param.  Handle NULL_TREE for "val".
3606         (find_case_label_for_value): Make "switch_stmt" param const.
3607         * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3608         of dom_walker.
3609         (vrp_prop::check_all_array_refs): Reimplement as...
3610         (check_array_bounds_dom_walker::before_dom_children): ...this new
3611         vfunc.  Replace linear search through BB block list, excluding
3612         those with non-executable in-edges via dominator walk.
3614 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3616         * config/arm/arm.opt (mverbose-cost-dump): New option.
3617         * config/arm/arm.c (arm_rtx_costs): Use it.
3619 2017-12-14  Andreas Schwab  <schwab@linux-m68k.org>
3621         PR bootstrap/83396
3622         * reload1.c (emit_input_reload_insns): Skip debug markers.
3624 2017-12-14  Alexandre Oliva  <aoliva@redhat.com>
3626         * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3627         for its nop_endbr.
3629         PR bootstrap/83396
3630         * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3631         * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3632         * config/sh/sh.c: Skip debug insns besides notes.
3633         * config/sh/sh.md: Likewise.
3634         * config/sh/sh_treg_combine.cc: Likewise.
3635         * config/sh/sync.md: Likewise.
3637 2017-12-14  Tom de Vries  <tom@codesourcery.com>
3639         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3640         Add item for weak_undefined.
3642 2017-12-14  Richard Biener  <rguenther@suse.de>
3644         PR tree-optimization/67842
3645         * sese.h (bb_in_region): Remove #if 0'ed code.
3647 2017-12-14  Richard Biener  <rguenther@suse.de>
3649         PR tree-optimization/83326
3650         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3651         may_be_zero parameter and handle it by not marking the first
3652         peeled copy as not exiting the loop.
3653         (try_peel_loop): Likewise.
3654         (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3655         to handle the case of constant or zero iterations and perform
3656         loop header copying on-the-fly.
3658 2017-12-14  Richard Biener  <rguenther@suse.de>
3660         PR tree-optimization/83418
3661         * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3662         Instead of asserting we don't get unfolded comparisons deal with them.
3664 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3666         PR bootstrap/83396
3667         * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3668         basic blocks.  Assert debug bind insns don't appear outside of bbs,
3669         don't reset them.  Assert insns without BLOCK_FOR_INSN are outside of
3670         bb.  Simplify.
3672         PR tree-optimization/83198
3673         * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3674         dir.modifier, regardless of TREE_TYPE (arg).  Assume non-REAL_CST
3675         value if arg is a REAL_CST with incompatible type.
3677 2017-12-14  Sudakshina Das  <sudi.das@arm.com>
3678             Bin Cheng  <bin.cheng@arm.com>
3680         PR target/81228
3681         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3682         to CCFPEmode.
3683         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3684         LTGT.
3686 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3688         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3689         cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3691 2017-12-14  Richard Sandiford  <richard.sandiford@linaro.org>
3692             Alan Hayward  <alan.hayward@arm.com>
3693             David Sherwood  <david.sherwood@arm.com>
3695         * poly-int.h: New file.
3696         * poly-int-types.h: Likewise.
3697         * coretypes.h: Include them.
3698         (POLY_INT_CONVERSION): Define.
3699         * target.def (estimated_poly_value): New hook.
3700         * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3701         * doc/tm.texi: Regenerate.
3702         * doc/poly-int.texi: New file.
3703         * doc/gccint.texi: Include it.
3704         * doc/rtl.texi: Describe restrictions on subreg modes.
3705         * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3706         * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3707         (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3708         * targhooks.h (default_estimated_poly_value): Declare.
3709         * targhooks.c (default_estimated_poly_value): New function.
3710         * target.h (estimated_poly_value): Likewise.
3711         * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3712         (wi::unary_traits): Delete.
3713         (wi::binary_traits::signed_shift_result_type): Define for
3714         offset_int << HOST_WIDE_INT, etc.
3715         (generic_wide_int::operator <<=): Define for all types and use
3716         wi::lshift instead of <<.
3717         (wi::hwi_with_prec): Add a default constructor.
3718         (wi::ints_for): New class.
3719         (operator <<): Define for all wide-int types.
3720         (operator /): New function.
3721         (operator %): Likewise.
3722         * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3723         (ASSERT_MAYBE_NE_AT): New macros.
3725 2017-12-13  Eric Botcazou  <ebotcazou@adacore.com>
3726             Dominik Vogt  <vogt@linux.vnet.ibm.com>
3728         PR middle-end/78468
3729         * emit-rtl.c (init_emit): Remove ??? comment.
3730         * explow.c (get_dynamic_stack_size): Take known alignment of stack
3731         pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3732         * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3733         alignment of 3 virtual registers to BITS_PER_WORD.
3735         * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3737 2017-12-13  Peter Bergner  <bergner@vnet.ibm.com>
3739         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3740         * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3742 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3744         PR bootstrap/83396
3745         * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3747 2017-12-13  Jakub Jelinek  <jakub@redhat.com>
3749         * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3750         with the exception of debug begin stmt markers appear before
3751         labels.
3753         PR bootstrap/83396
3754         * final.c (rest_of_handle_final): Call variable_tracking_main only
3755         if !flag_var_tracking.
3757 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3758             Jakub Jelinek  <jakub@redhat.com>
3760         PR bootstrap/83396
3761         PR debug/83391
3762         * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3763         labels debug stmts that can only appear after labels.
3765 2017-12-13  Alexander Monakov  <amonakov@ispras.ru>
3767         PR rtl-optimization/82398
3768         * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3769         EXPR_USEFULNESS in priority comparison.
3771 2017-12-13  Segher Boessenkool  <segher@kernel.crashing.org>
3773         PR rtl-optimization/83393
3774         * combine.c (move_deaths): If reg_stat points to a too new insn in
3775         last_death, do not use it: find the proper insn instead.
3777 2017-12-12  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3779         PR target/82190
3780         * config/rs6000/rs6000-string.c (expand_block_compare,
3781         expand_strn_compare): Fix set_mem_size() calls.
3783 2017-12-12  Jeff Law  <law@redhat.com>
3785         PR tree-optimization/83298
3786         PR tree-optimization/83362
3787         PR tree-optimization/83383
3788         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3789         push_value_range a public interface.  Add new argument to
3790         record_ranges_from_stmt.
3791         * gimple-ssa-evrp-analyze.c
3792         (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3793         Update comments.  Handle recording temporary equivalences.
3794         * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3795         new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3796         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3797         * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3798         gimple-ssa-evrp-analyze.h.
3799         (record_temporary_equivalences_from_phis): Add new argument.  When
3800         the PHI arg is an SSA_NAME, set the result's range to the range
3801         of the PHI arg.
3802         (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3803         from statements too.
3804         (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3805         Pass it down to children as needed.
3806         (thread_outgoing_edges): Likewise.
3807         (thread_across_edge): Likewise.   Push/pop range state as needed.
3808         * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3810 2017-12-12  Julia Koval  <julia.koval@intel.com>
3812         * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3813         (PTA_CANNONLAKE): Remove PTA_CLWB.
3815 2017-12-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3817         PR target/83332
3818         * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3819         (vcondv2div2df): Likewise.
3820         (vconduv2dfv2di): Likewise.
3822 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3824         * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3825         BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3826         BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3827         ATTR_MATHFN_FPROUNDING_ERRNO.
3829 2017-12-12  Richard Biener  <rguenther@suse.de>
3831         PR tree-optimization/83385
3832         * graphite-scop-detection.c (get_order, order): Remove.
3833         (bb_to_rpo): New global.
3834         (cmp_pbbs): Adjust.
3835         (build_scops): Sort pbbs in RPO order.
3837 2017-12-12  James Greenhalgh  <james.greenhalgh@arm.com>
3839         * combine.c (simplify_set): Do not transform subregs to zero_extends
3840         if the destination mode is a vector mode.
3842 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3844         PR tree-optimization/83359
3845         * tree-cfg.h (fold_loop_internal_call): Declare.
3846         * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3847         * tree-cfg.c (fold_loop_internal_call): ... here.  No longer static.
3848         (find_loop_dist_alias): New function.
3849         (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3850         updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3851         calls.
3853         PR tree-optimization/80631
3854         * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3855         (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3856         arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3857         hardcoding zero as the value if COND_EXPR is never true.  For
3858         INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3859         INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3860         hardcoding MAX_EXPR as the reduction operation.
3861         (is_nonwrapping_integer_induction): Allow negative step.
3862         (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3863         vect_create_epilog_for_reduction, if no value is suitable, don't
3864         use INTEGER_INDUC_COND_REDUCTION for now.  Formatting fixes.
3866 2017-12-12  Richard Biener  <rguenther@suse.de>
3868         PR tree-optimization/81889
3869         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3870         range info from the non-wrapping IV instead of just the range
3871         of the type.
3873 2017-12-12  Julia Koval  <julia.koval@intel.com>
3875         * config.gcc: Add vaesintrin.h.
3876         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3877         New type.
3878         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3879         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3880         New builtins.
3881         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3882         * config/i386/immintrin.h: Include vaesintrin.h.
3883         * config/i386/sse.md (vaesdec_<mode>): New pattern.
3884         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3885         _mm_aesdec_epi128): New intrinsics.
3887 2017-12-12  Julia Koval  <julia.koval@intel.com>
3889         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3890         OPTION_MASK_ISA_VAES_UNSET): New.
3891         (ix86_handle_option): Handle -mvaes.
3892         * config/i386/cpuid.h: Define bit_VAES.
3893         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3894         * config/i386/i386-c.c (__VAES__): New.
3895         * config/i386/i386.c (ix86_target_string): Add -mvaes.
3896         (ix86_valid_target_attribute_inner_p): Ditto.
3897         * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
3898         * config/i386/i386.opt: Add -mvaes.
3899         * doc/invoke.texi: Ditto.
3901 2017-12-12  Alexandre Oliva  <aoliva@redhat.com>
3903         * debug.h (gcc_debug_hooks): Add inline_entry.
3904         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
3905         * debug.c (do_nothing_debug_hooks): Likewise.
3906         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
3907         * dwarf2out.c (dwarf2_debug_hooks): Likewise.
3908         (dwarf2_lineno_debug_hooks): Likewise.
3910         * common.opt (gstatement-frontiers): New, setting
3911         debug_nonbind_markers_p.
3912         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
3913         * toplev.c (process_options): Autodetect value for debug statement
3914         frontiers option.
3915         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
3916         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
3918         * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
3919         markers.  Integrate source bind into debug stmt expand loop.
3920         (pass_expand::execute): Check debug marker limit.  Avoid deep
3921         TER and expand debug locations for debug bind insns only.
3922         * cse.c (insn_live_p): Keep nonbind markers and debug bindings
3923         followed by them.
3924         * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
3925         * final.c (reemit_insn_block_notes): Take current block from
3926         nonbind markers.  Declare note where it's first set.
3927         (final_scan_insn): Handle begin stmt notes.  Emit is_stmt according to
3928         begin stmt markers if enabled.
3929         (notice_source_line): Handle nonbind markers.  Fail if their
3930         location is unknown or that of builtins.
3931         (rest_of_handle_final): Convert begin stmt markers to notes if
3932         var-tracking didn't run.
3933         (rest_of_clean_state): Skip begin stmt markers.
3934         * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
3935         markers.
3936         * function.c (allocate_struct_function): Set begin_stmt_markers.
3937         * function.h (struct function): Add debug_marker_count counter
3938         and debug_nonbind_markers flag.
3939         * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
3940         * gimple-low.c (lower_function_body): Adjust
3941         debug_nonbind_markers.
3942         (lower_stmt): Drop or skip gimple debug stmts.
3943         (lower_try_catch): Skip debug stmts.
3944         * gimple.c (gimple_build_debug_begin_stmt): New.
3945         (gimple_copy): Increment debug_marker_count if copying one.
3946         * gimple.h (gimple_build_debug_begin_stmt): Declare.
3947         * gimplify.c (rexpr_location): New.
3948         (rexpr_has_location): New.
3949         (warn_switch_unreachable_r): Handle gimple debug stmts.
3950         (shortcut_cond_r): Call expr_location.
3951         (find_goto): New.
3952         (find_goto_label): New.
3953         (shortcut_cond_expr): Call expr_has_location, expr_location, and
3954         find_goto_label.
3955         (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
3956         expr_location.
3957         (gimplify_expr): Handle begin stmt markers.  Reject debug expr decls.
3958         * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New.  Add to...
3959         (LANG_HOOKS_INITIALIZER): ... this.
3960         * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
3961         * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
3962         debug insns.
3963         (update_ebb_live_info): Skip debug insn markers.
3964         * lra.c (debug_insn_static_data): Rename to...
3965         (debug_bind_static_data): ... this.
3966         (debug_marker_static_data): New.
3967         (lra_set_insn_recog_data): Select one of the above depending
3968         on debug insn kind.
3969         (lra_update_isn_regno_info): Don't assume debug insns have
3970         freqs.
3971         (push_insns): Skip debug insns.
3972         * lto-streamer-in.c (input_function): Drop debug stmts
3973         depending on active options.  Adjust debug_nonbind_markers.
3974         * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
3975         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3976         begin stmt marker notes.
3977         (print_insn): Likewise.
3978         * recog.c (extract_insn): Recognize rtl for debug markers.
3979         * rtl.def (DEBUG_MARKER): New.
3980         * tree-inline.c: Include params.h.
3981         (remap_gimple_stmt): Handle nonbind markers.
3982         (maybe_move_debug_stmts_to_successors): Likewise.
3983         (copy_debug_stmt): Likewise.
3984         * tree-iterator.c (append_to_statement_list_1): Append begin stmt
3985         markers regardless of no side effects.
3986         (tsi_link_before): Don't update container's side effects when adding
3987         a begin stmt marker.
3988         (tsi_link_after): Likewise.
3989         (expr_first): Skip begin stmt markers.
3990         (expr_last): Likewise.
3991         * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
3992         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
3993         Disregard nonbind markers.
3994         * tree.c (make_node_stat): Don't set side effects for begin stmt
3995         markers.
3996         (build1_stat): Likewise.
3997         * tree.def (DEBUG_BEGIN_STMT): New.
3998         * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
3999         * var-tracking.c (delete_debug_insns): Renamed to...
4000         (delete_vta_debug_insns): ... this.
4001         (reemit_marker_as_note): New.
4002         (vt_initialize): Reemit markers.
4003         (delete_vta_debug_insns): Likewise.
4004         (vt_debug_insns_local): Reemit or delete markers.
4005         (variable_tracking_main_1): Likewise.
4006         * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
4007         * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
4008         (gimple_debug_nonbind_marker_p): New.
4009         (gimple_build_debug_bind): Adjust.
4010         (gimple_build_debug_begin_stmt): New.
4011         * doc/invoke.texi (max-debug-marker-count): New param.
4012         * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
4013         (debug_parameter_ref, debug_marker): New.
4014         (NOTE_INSN_BEGIN_STMT): New.
4015         (DEBUG_INSN): Describe begin stmt markers.
4017         * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
4018         without debug insns, we wouldn't, but clean up debug insns
4019         after a control flow insn nevertheless.
4021         * cfgcleanup.c (delete_unreachable_blocks): Use alternate
4022         block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
4023         * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
4024         * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
4025         (rtl_tidy_fallthru_edge): Likewise.
4026         (rtl_verify_fallthru): Likewise.
4027         (rtl_verify_bb_layout): Likewise.
4028         (skip_insns_after_block): Likewise.
4029         (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
4030         * dwarf2out.c: Include print-rtl.h.
4031         (dwarf2out_next_real_insn): New.
4032         (dwarf2out_var_location): Call it.  Disregard begin stmt markers.
4033         Dump debug binds in asm comments.
4034         * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
4035         * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
4036         callers to use gsi_start_nondebug_bb instead.
4037         (gsi_after_labels): Skip gimple debug stmts.
4038         (gsi_start_nondebug): New.
4039         * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
4040         (proper_loop_form_for_interchange): Adjust.
4041         * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
4042         * gimple.h (gimple_seq_last_nondebug_stmt): New.
4043         * gimplify.c (last_stmt_in_scope): Skip debug stmts.
4044         (collect_fallthrough_labels): Likewise.
4045         (should_warn_for_implicit_fallthrough): Likewise.
4046         (warn_implicit_fallthrough_r): Likewise.
4047         (expand_FALLTHROUGH_r): Likewise.
4048         * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
4049         (graphite_copy_stmts_from_block): Skip nonbind markers.
4050         * haifa-sched.c (sched_extend_bb): Skip debug insns.
4051         * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
4052         * jump.c (clean_barriers): Skip debug insns.
4053         * omp-expand.c (expand_parallel_call): Skip debug insns.
4054         (expand_task_call): Likewise.
4055         (remove_exit_barrier): Likewise.
4056         (expand_omp_taskreg): Likewise.
4057         (expand_omp_for_init_counts): Likewise.
4058         (expand_omp_for_generic): Likewise.
4059         (expand_omp_for_static_nochunk): Likewise.
4060         (expand_omp_for_static_chunk): Likewise.
4061         (expand_omp_simd): Likewise.
4062         (expand_omp_taskloop_for_outer): Likewise.
4063         (expand_omp_taskloop_for_inner): Likewise.
4064         (expand_oacc_for): Likewise.
4065         (expand_omp_sections): Likewise.
4066         (expand_omp_single): Likewise.
4067         (expand_omp_synch): Likewise.
4068         (expand_omp_atomic_load): Likewise.
4069         (expand_omp_atomic_store): Likewise.
4070         (expand_omp_atomic_fetch_op): Likewise.
4071         (expand_omp_atomic_pipeline): Likewise.
4072         (expand_omp_atomic_mutex): Likewise.
4073         (expand_omp_target): Likewise.
4074         (grid_expand_omp_for_loop): Likewise.
4075         (grid_expand_target_grid_body): Likewise.
4076         (build_omp_regions_1): Likewise.
4077         * omp-low.c (check_combined_parallel): Skip debug stmts.
4078         * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
4079         * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
4080         * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
4081         test.
4082         * tree-cfg.c (make_blobs_1): Skip debug stmts.
4083         (make_edges): Likewise.
4084         (cleanup_dead_labels): Likewise.
4085         (gimple_can_merge_blocks_p): Likewise.
4086         (stmt_starts_bb_p): Likewise.
4087         (gimple_block_label): Likewise.
4088         (gimple_redirect_edge_and_branch): Likewise.
4089         * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
4090         of debug stmts.
4091         (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
4092         TDF_SLIM.
4093         * tree-pretty-print (print_declaration): Omit initializer in slim
4094         dumps.
4095         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
4096         markers.
4097         (eliminate_unnecessary_stmts): Stabilize block removal order.
4098         * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
4099         * var-tracking.c (get_first_insn): New.
4100         (vt_emit_notes): Call it.
4101         (vt_initialize): Walk any insns before the first BB.
4102         (delete_debug_insns): Likewise.
4104         * gimple.h (enum gimple_debug_subcode): Add
4105         GIMPLE_DEBUG_BEGIN_STMT.
4106         (gimple_debug_begin_stmt_p): New.
4107         (gimple_debug_nonbind_marker_p): New.
4108         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
4109         (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
4110         (MAY_HAVE_DEBUG_STMTS): ... this.  Check both.
4111         * insn-notes.def (BEGIN_STMT): New.
4112         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
4113         (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
4114         (MAY_HAVE_DEBUG_INSNS): ... this.  Check both.
4115         (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
4116         (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
4117         (INSN_DEBUG_MARKER_KIND): New.
4118         (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
4119         (INSN_VAR_LOCATION): Check for VAR_LOCATION.
4120         (INSN_VAR_LOCATION_PTR): New.
4121         * cfgexpand.c (expand_debug_locations): Handle debug bind insns
4122         only.
4123         (expand_gimple_basic_block): Likewise.  Emit debug temps for TER
4124         deps only if debug bind insns are enabled.
4125         (pass_expand::execute): Avoid deep TER and expand
4126         debug locations for debug bind insns only.
4127         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
4128         debug stmts special handling down to debug bind stmts.
4129         * combine.c (try_combine): Narrow debug insns special handling
4130         down to debug bind insns.
4131         * cse.c (delete_trivially_dead_insns): Handle debug bindings.
4132         Narrow debug insns preexisting special handling down to debug
4133         bind insns.
4134         * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
4135         handling down to debug bind insns.
4136         * function.c (instantiate_virtual_regs): Skip debug markers,
4137         adjust handling of debug binds.
4138         * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
4139         temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
4140         * haifa-sched.c (schedule_insn): Narrow special handling of debug
4141         insns to debug bind insns.
4142         * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
4143         special handling of debug stmts to debug bind stmts.
4144         * ipa-split.c (split_function): Likewise.
4145         * ira.c (combine_and_move_insns): Adjust debug bind insns only.
4146         * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
4147         debug insns.
4148         * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
4149         * regrename.c (build_def_use): Likewise.
4150         * regcprop.c (copyprop_hardreg_forward_1): Likewise.
4151         (pass_cprop_hardreg): Narrow special casing of debug insns to
4152         debug bind insns.
4153         * regstat.c (regstat_init_n_sets_and_refs): Likewise.
4154         * reload1.c (reload): Likewise.
4155         * sese.c (sese_insert_phis_for_liveouts): Narrow special
4156         casing of debug stmts to debug bind stmts.
4157         * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
4158         * ssa-iterators.h (num_imm_uses): Likewise.
4159         * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
4160         debug stmts to debug bind stmts.
4161         * tree-inline.c (tree_function_versioning): Narrow special casing
4162         of debug stmts to debug bind stmts.
4163         * tree-loop-distribution.c (generate_loops_for_partition):
4164         Narrow special casing of debug stmts to debug bind stmts.
4165         * tree-sra.c (analyze_access_subtree): Narrow special casing
4166         of debug stmts to debug bind stmts.
4167         * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
4168         stmts to debug bind stmts.
4169         * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
4170         casing of debug stmts to debug bind stmts.
4171         * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
4172         * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
4173         casing of debug stmts to debug bind stmts.
4174         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4175         Likewise.
4176         * tree-ssa.c (flush_pending_stmts): Narrow special casing of
4177         debug stmts to debug bind stmts.
4178         (gimple_replace_ssa_lhs): Likewise.
4179         (insert_debug_temp_for_var_def): Likewise.
4180         (insert_debug_temps_for_defs): Likewise.
4181         (reset_debug_uses): Likewise.
4182         * tree-ssanames.c (release_ssa_name_fn): Likewise.
4183         * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
4184         (adjust_debug_stmts): Likewise.
4185         (adjust_phi_and_debug_stmts): Likewise.
4186         (vect_do_peeling): Likewise.
4187         * tree-vect-loop.c (vect_transform_loop): Likewise.
4188         * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
4189         * var-tracking.c (adjust_mems): Narrow special casing of debug
4190         insns to debug bind insns.
4191         (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
4192         (compute_bb_dataflow, vt_find_locations): Likewise.
4193         (vt_expand_loc, emit_notes_for_changes): Likewise.
4194         (vt_init_cfa_base): Likewise.
4195         (vt_emit_notes): Likewise.
4196         (vt_initialize): Likewise.
4197         (vt_finalize): Likewise.
4199         * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
4200         (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
4201         (next_nonnote_nondebug_insn_bb): New.
4202         (prev_nonnote_nondebug_insn_bb): New.
4203         (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
4204         * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
4205         (prev_nonnote_nondebug_insn_bb): Declare.
4206         (next_nonnote_nondebug_insn_bb): Declare.
4207         * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
4208         * cfgrtl.c (get_last_bb_insn): Likewise.
4209         * lra.c (push_insns): Likewise.
4211 2017-12-11  David Malcolm  <dmalcolm@redhat.com>
4213         PR c/82050
4214         * selftest-run-tests.c (selftest::run_tests): Move start/finish code
4215         to...
4216         * selftest.c (selftest::test_runner::test_runner): New ctor.
4217         (selftest::test_runner::~test_runner): New dtor.
4218         * selftest.h (class selftest::test_runner): New class.
4220 2017-12-11  Carl Love  <cel@us.ibm.com>
4222         * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
4223         vec_extract_fp32_from_shortl]): Add #defines.
4224         * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4225         * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4226         ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4227         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4228         ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4229         ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4230         * doc/extend.texi: Add documentation for the added builtins.
4232 2017-12-11  Alexandre Oliva  <aoliva@redhat.com>
4234         PR rtl-optimization/80693
4235         PR rtl-optimization/81019
4236         PR rtl-optimization/81020
4237         * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4238         are not mentioned in i3.  Place the REG_UNUSED note on i2,
4239         possibly modified to REG_DEAD, if it did not originate in i3.
4241 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4243         * recog.c (store_data_bypass_p_1): New function.
4244         (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER.  Use
4245         store_data_bypass_p_1 to avoid code duplication.  Formatting fixes.
4247 2017-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
4249         PR rtl-optimization/83361
4250         * ifcvt.c (if_convert): Call fixup_partitions.
4252 2017-12-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
4254         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4255         early folding of splat_u{8,16,32}.
4257 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4259         * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4260         output_operand_lossage first argument with capital letter.
4261         (aarch64_override_options): Don't start error and sorry first argument
4262         with capital letter.
4264 2017-12-11  Andi Kleen  <ak@linux.intel.com>
4266         PR gcov-profile/83355
4267         * auto-profile.c (string_table::get_index_by_decl): Don't
4268         recurse when abstract origin points to itself.
4270 2017-12-11  Bin Cheng  <bin.cheng@arm.com>
4272         PR tree-optimization/83320
4273         * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4274         (prune_datarefs_not_in_loop): Ditto.
4276 2017-12-10  Gerald Pfeifer  <gerald@pfeifer.com>
4278         * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4280 2017-12-10  Jakub Jelinek  <jakub@redhat.com>
4282         PR tree-optimization/83337
4283         * gimple-loop-interchange.cc (compute_access_stride): Handle
4284         bitfield DRs properly.
4286 2017-12-09  Jakub Jelinek  <jakub@redhat.com>
4288         PR tree-optimization/83338
4289         * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4290         vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4291         vector of unsigned integers to vector of signed integers.
4293 2017-12-08  Vladimir Makarov  <vmakarov@redhat.com>
4295         PR rtl-optimization/83317
4296         * lra-constraints.c (process_address_1): Add insn code check.
4298 2017-12-08  Michael Matz  <matz@suse.de>
4300         Fix PR tree-optimization/83323
4301         * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4302         head-controlled loops and loop BBs.
4303         * common.opt (funroll-and-jam): Remove, instead ...
4304         (floop-unroll-and-jam): ... reuse this option.
4305         * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4306         * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4307         (-floop-unroll-and-jam): ... this option.
4309 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4311         * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4312         a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4313         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4315 2017-12-08  Richard Biener  <rguenther@suse.de>
4317         PR middle-end/81782
4318         * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4319         handle accesses outside of zero-sized vars.
4321 2017-12-08  Martin Jambor  <mjambor@suse.cz>
4323         PR tree-optimization/83141
4324         * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4325         test for MEM_REFs implicitely changing types with padding.  Remove
4326         inline keyword.
4327         (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4329 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4331         * config/arc/arc.c (arc_attribute_table): Add exclusions to
4332         the comment.
4333         * config/avr/avr.c (avr_attribute_table): Likewise.
4334         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4335         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4336         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4337         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4338         of struct attribute_spec.
4339         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4341 2017-12-08  Ulrich Weigand  <uweigand@de.ibm.com>
4343         PR target/82960
4344         * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4346 2017-12-08  Jan Hubicka  <hubicka@ucw.cz>
4348         PR middle-end/83609
4349         * profile-count.c (profile_count::from_gcov_type): Move from
4350         profile-count.h; handle overflow.
4351         * profile-count.h (profile_count::from_gcov_type): Move offline.
4353 2017-12-08  Segher Boessenkool  <segher@kernel.crashing.org>
4355         PR rtl-optimization/83304
4356         * combine.c (move_deaths): If we do not know where a register died,
4357         search for it.
4359 2017-12-08  Richard Biener  <rguenther@suse.de>
4361         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4362         Provide -fopt-info-loop feedback when we interchange in a nest.
4364 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4366         * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4367         for armv6 ARM CPU IDs.
4369 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4371         * common/config/arm/arm-common.c: Include <algorithm>.
4372         (INCLUDE_VECTOR): Define.
4373         (compare_opt_names): New function.
4374         (arm_rewrite_selected_arch): Only strip out extensions that can be
4375         expressed through -mfpu.  Sort the remaining extensions
4376         alphabetically.
4378 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4380         * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4381         (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4382         (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4383         * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4385 2017-06-08  Tristan Gingold  <gindold@adacore.com>
4387         PR ada/81470
4388         * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4389         .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4391 2017-12-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4393         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4394         of struct attribute_spec.
4396 2017-12-08  Julia Koval  <julia.koval@intel.com>
4398         * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4399         _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4400         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4401         _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4402         _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4403         _mm_maskz_dpwssds_epi32): Ditto.
4405 2017-12-08  Richard Biener  <rguenther@suse.de>
4407         PR tree-optimization/81303
4408         * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4409         conditions try to create a comparison vector type matching
4410         the data vector type.
4411         (vectorizable_condition): Adjust.
4412         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4413         Leave invariant conditions alone in case we can vectorize those.
4415 2017-12-08  Julia Koval  <julia.koval@intel.com>
4417         * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4418         _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4419         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4420         _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4421         _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4423 2017-12-08  Julia Koval  <julia.koval@intel.com>
4425         * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4426         _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4427         * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4428         _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4429         _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4430         _mm_maskz_dpbusds_epi32): New intrinsics.
4432 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4434         * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4435         operand.
4436         (ld<bh>io_signed): Likewise.
4437         (st<bhw>io): Likewise.
4438         * config/nios2/predicates.md (ldstio_memory_operand): Allow
4439         SMALL_INT12 constant integer operand.
4441 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4443         * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4444         Rename to...
4445         (nios2_large_constant_allowed): ...this.  Adjust uses.
4446         (nios2_plus_symbolic_constant_p): Rename to...
4447         (nios2_plus_large_constant_p): ...this.  Adjust uses.
4448         (nios2_legitimate_address_p): Correct CONST_INT handling.
4449         (nios2_symbolic_memory_operand_p): Rename to...
4450         (nios2_large_constant_memory_operand_p): ...this.  Adjust uses.
4451         (nios2_large_constant_p): Check for large constant integers too.
4452         (nios2_split_large_constant): Handle constant integers.
4453         (nios2_split_symbolic_memory_operand): Rename to...
4454         (nios2_split_large_constant_memory_operand): ...this.  Adjust uses.
4455         (nios2_legitimize_constant_address): Handle constant integers.
4456         (r0rel_constant_p): Handle small constant integers.
4457         (nios2_print_operand_address): Handle r0-relative integer addresses.
4458         * config/nios2/nios2-protos.h: Adjust for renamed functions.
4459         * config/nios2/nios2.md: Adjust for renamed functions.
4461 2017-12-07  Andrew Waterman  <andrew@sifive.com>
4463         * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4464         (TARGET_HAVE_SRODATA_SECTION): New define.
4465         (riscv_select_section): New function.
4467 2017-12-08  Joseph Myers  <joseph@codesourcery.com>
4468             Alexander Monakov  <amonakov@ispras.ru>
4469             Jakub Jelinek  <jakub@redhat.com>
4471         PR target/81906
4472         * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4474 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4476         * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4477         a bool success value.  Don't call output_operand_lossage here.
4478         (aarch64_print_ldpstp_address): Return a bool success value.
4479         (aarch64_print_operand_address): Call output_addr_const if
4480         aarch64_print_address_internal fails.
4481         (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4482         'y'; call output_operand_lossage instead.  Call output_operand_lossage
4483         if aarch64_print_ldpstp_address fails.
4485 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4487         * tree-vector-builder.h
4488         (tree_vector_builder::binary_encoded_nelts): Declare.
4489         * tree-vector-builder.c
4490         (tree_vector_builder::binary_encoded_nelts): New function.
4491         * fold-const.c (negate_expr_p): Likewise.
4492         (operand_equal_p, fold_checksum_tree): Likewise.
4493         * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4494         * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4495         (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4496         (uniform_vector_p): Likewise.
4497         * varasm.c (const_hash_1, compare_constant): Likewise.
4498         * tree-ssa-ccp.c: Include tree-vector-builder.h.
4499         (valid_lattice_transition): Operate directly on the VECTOR_CST
4500         encoding.
4501         * ipa-icf.c: Include tree-vector-builder.h.
4502         (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4503         * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4505 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4507         * tree.c (build_vector): Delete.
4508         * tree.h (build_vector): Make static and move into the self-testing
4509         block.
4511 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4513         * vector-builder.h (vector_builder::derived): New const overload.
4514         (vector_builder::elt): New function.
4515         * tree-vector-builder.h (tree_vector_builder::type): New function.
4516         (tree_vector_builder::apply_step): Declare.
4517         * tree-vector-builder.c (tree_vector_builder::apply_step): New
4518         function.
4519         * gimple-fold.h (tree_vector_builder): Declare.
4520         (gimple_build_vector): Take a tree_vector_builder instead of a
4521         type and vector of elements.
4522         * gimple-fold.c (gimple_build_vector): Likewise.
4523         * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4524         accordingly.
4525         (get_initial_defs_for_reduction): Likewise.
4526         (vectorizable_induction): Likewise.
4528 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4530         * tree-vector-builder.h
4531         (tree_vector_builder::new_binary_operation): Declare.
4532         * tree-vector-builder.c
4533         (tree_vector_builder::new_binary_operation): New function.
4534         * fold-const.c (fold_relational_const): Use it.
4535         (const_binop): Likewise.  Check that both input vectors have
4536         the same number of elements, thus excluding things like WIDEN_SUM.
4537         Check whether it is possible to operate directly on the encodings
4538         of stepped inputs.
4540 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4542         * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4543         new_unary_operation, operating only on the encoded elements.
4544         (const_unop): Likewise.
4545         (exact_inverse): Likewise.
4546         (distributes_over_addition_p): New function.
4547         (const_binop): Use tree_vector_builder and new_unary_operation
4548         for combinations of VECTOR_CST and INTEGER_CST.  Operate only
4549         on the encoded elements unless the encoding is strided and the
4550         operation does not distribute over addition.
4551         (fold_convert_const):  Use tree_vector_builder and
4552         new_unary_operation.  Operate only on the encoded elements
4553         for truncating integer conversions, or for non-stepped encodings.
4555 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4557         * config/sparc/sparc.c: Include tree-vector-builder.h.
4558         (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4559         * expmed.c: Include tree-vector-builder.h.
4560         (make_tree): Use tree_vector_builder instead of build_vector.
4561         * fold-const.c: Include tree-vector-builder.h.
4562         (const_binop): Use tree_vector_builder instead of build_vector.
4563         (const_unop): Likewise.
4564         (native_interpret_vector): Likewise.
4565         (fold_vec_perm): Likewise.
4566         (fold_ternary_loc): Likewise.
4567         * gimple-fold.c: Include tree-vector-builder.h.
4568         (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4569         of build_vector.
4570         * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4571         (simplify_vector_constructor): Use tree_vector_builder instead
4572         of build_vector.
4573         * tree-vect-generic.c: Include tree-vector-builder.h.
4574         (add_rshift): Use tree_vector_builder instead of build_vector.
4575         (expand_vector_divmod): Likewise.
4576         (optimize_vector_constructor): Likewise.
4577         * tree-vect-loop.c: Include tree-vector-builder.h.
4578         (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4579         of build_vector.  Explicitly use a stepped encoding for
4580         { 1, 2, 3, ... }.
4581         * tree-vect-slp.c: Include tree-vector-builder.h.
4582         (vect_get_constant_vectors): Use tree_vector_builder instead
4583         of build_vector.
4584         (vect_transform_slp_perm_load): Likewise.
4585         (vect_schedule_slp_instance): Likewise.
4586         * tree-vect-stmts.c: Include tree-vector-builder.h.
4587         (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4588         (vect_gen_perm_mask_any): Likewise.
4589         (vectorizable_call): Likewise.  Explicitly use a stepped encoding.
4590         * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4591         of build_vector.
4592         (build_vector_from_val): Likewise.  Explicitly use a duplicate
4593         encoding.
4595 2017-12-07  Richard Sandiford  <richard.sandiford@arm.com>
4597         * doc/generic.texi (VECTOR_CST): Describe new representation of
4598         vector constants.
4599         * vector-builder.h: New file.
4600         * tree-vector-builder.h: Likewise.
4601         * tree-vector-builder.c: Likewise.
4602         * Makefile.in (OBJS): Add tree-vector-builder.o.
4603         * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4604         * tree-core.h (tree_base): Add a vector_cst field to the u union.
4605         (tree_vector): Change the number of elements to
4606         vector_cst_encoded_nelts.
4607         * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4608         (VECTOR_CST_ELTS): Delete.
4609         (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4610         (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4611         (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4612         (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4613         (VECTOR_CST_ENCODED_ELT): Likewise.
4614         (vector_cst_encoded_nelts): New function.
4615         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4616         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4617         (vector_cst_int_elt, vector_cst_elt): Declare.
4618         * tree.c: Include tree-vector-builder.h.
4619         (tree_code_size): Abort if passed VECTOR_CST.
4620         (tree_size): Update for new VECTOR_CST layout.
4621         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4622         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4623         (build_vector): Use tree_vector_builder.
4624         (vector_cst_int_elt, vector_cst_elt): New functions.
4625         (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4626         encoded elements and then create the vector in the canonical form.
4627         (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4628         (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4629         (tree_c_tests): Call test_vector_cst_patterns.
4630         * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4631         VECTOR_CST fields.
4632         (hash_tree): Likewise.
4633         * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4634         (streamer_write_tree_header): Likewise.
4635         * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4636         (streamer_alloc_tree): Likewise.  Update call to make_vector.
4637         * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4639 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4641         * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4642         (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4643         * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4645 2017-12-07  Bin Cheng  <bin.cheng@arm.com>
4646             Richard Biener  <rguenther@suse.de>
4648         PR tree-optimization/81303
4649         * Makefile.in (gimple-loop-interchange.o): New object file.
4650         * common.opt (floop-interchange): Reuse the option from graphite.
4651         * doc/invoke.texi (-floop-interchange): Ditto.  New document for
4652         -floop-interchange and mention it for -O3.
4653         * opts.c (default_options_table): Enable -floop-interchange at -O3.
4654         * gimple-loop-interchange.cc: New file.
4655         * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4656         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4657         * passes.def (pass_linterchange): New pass.
4658         * timevar.def (TV_LINTERCHANGE): New time var.
4659         * tree-pass.h (make_pass_linterchange): New declaration.
4660         * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4661         interchange.  Record IV before/after increment in new parameters.
4662         * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4663         * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4664         path check into...
4665         (check_reduction_path): ...New function here.
4666         * tree-vectorizer.h (check_reduction_path): New declaration.
4668 2017-12-07  Vladimir Makarov  <vmakarov@redhat.com>
4670         PR target/83252
4671         PR rtl-optimization/80818
4672         * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4673         always early clobbered.
4674         * lra-lives.c (process_bb_lives): Check input hard regs for early
4675         clobbered non-operand hard reg.
4677 2017-12-07  Jakub Jelinek  <jakub@redhat.com>
4679         PR middle-end/83164
4680         * tree-cfg.c (verify_gimple_assign_binary): Don't require
4681         types_compatible_p, just that TYPE_MODE is the same.
4683 2017-12-07  Martin Sebor  <msebor@redhat.com>
4685         PR c/81544
4686         * attribs.c (empty_attribute_table): Initialize new member of
4687         struct attribute_spec.
4688         (decl_attributes): Add argument.  Handle mutually exclusive
4689         combinations of attributes.
4690         (selftests::test_attribute_exclusions): New function.
4691         (selftests::attribute_c_tests): Ditto.
4692         * attribs.h (decl_attributes): Add default argument.
4693         * selftest.h (attribute_c_tests): Declare.
4694         * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4695         * tree-core.h (attribute_spec::exclusions, exclude): New type and
4696         member.
4697         * doc/extend.texi (Common Function Attributes): Update const and pure.
4698         * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4699         of struct attribute_spec.
4700         * config/arc/arc.c (arc_attribute_table): Same.
4701         * config/arm/arm.c (arm_attribute_table): Same.
4702         * config/avr/avr.c ( avr_attribute_table): Same.
4703         * config/bfin/bfin.c (bfin_attribute_table): Same.
4704         * config/cr16/cr16.c (cr16_attribute_table): Same.
4705         * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4706         * config/h8300/h8300.c (h8300_attribute_table): Same.
4707         * config/i386/i386.c (ix86_attribute_table): Same.
4708         * config/ia64/ia64.c (ia64_attribute_table): Same.
4709         * config/m32c/m32c.c (m32c_attribute_table): Same.
4710         * config/m32r/m32r.c (m32r_attribute_table): Same.
4711         * config/m68k/m68k.c (m68k_attribute_table): Same.
4712         * config/mcore/mcore.c (mcore_attribute_table): Same.
4713         * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4714         * config/mips/mips.c (mips_attribute_table): Same.
4715         * config/msp430/msp430.c (msp430_attribute_table): Same.
4716         * config/nds32/nds32.c (nds32_attribute_table): Same.
4717         * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4718         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4719         * config/rl78/rl78.c (rl78__attribute_table): Same.
4720         * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4721         * onfig/rx/rx.c (rx_attribute_table): Same.
4722         * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4723         * config/sh/sh.c (sh_attribute_table): Same.
4724         * config/sparc/sparc.c (sparc_attribute_table): Same.
4725         * config/spu/spu.c (spu_attribute_table): Same.
4726         * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4727         * config/v850/v850.c (v850_attribute_table): Same.
4728         * config/visium/visium.c (visium_attribute_table): Same.
4730 2017-12-07  Tamar Christina  <tamar.christina@arm.com>
4732         PR target/82641
4733         * config/arm/arm.c (INCLUDE_STRING): Define.
4734         (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4735         (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4736         and .fpu.
4738 2017-12-07  Michael Matz  <matz@suse.de>
4740         Add unroll and jam pass
4742         * gimple-loop-jam.c: New file.
4743         * Makefile.in (OBJS): Add gimple-loop-jam.o.
4744         * common.opt (funroll-and-jam): New option.
4745         * opts.c (default_options_table): Add unroll-and-jam at -O3.
4746         * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4747         (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4748         * passes.def: Add pass_loop_jam.
4749         * timevar.def (TV_LOOP_JAM): Add.
4750         * tree-pass.h (make_pass_loop_jam): Declare.
4751         * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4752         * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4753         * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4754         to flow_loop_tree_node_add.
4755         (duplicate_subloops, copy_loops_to): Append to sibling list.
4756         * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4757         * doc/invoke.texi (-funroll-and-jam): Document new option.
4758         (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4760 2017-12-07  Richard Biener  <rguenther@suse.de>
4762         PR tree-optimization/83296
4763         PR tree-optimization/67769
4764         * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4765         flow sensitive info in an unrelated BB.
4766         (value_replacement): Use reset_flow_sensitive_info.
4767         (minmax_replacement): Reset flow sensitive info on the def
4768         we move.  Do not reset flow sensitive info in the whole BB
4769         we move the stmt to.
4770         (abs_replacement): Likewise.
4772 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4774         PR target/43871
4775         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4776         rs6000_cpu to the given -mcpu=, or to the default processor.
4778 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4780         * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4781         * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4782         instead of rs6000_cpu_attr.
4783         (rs6000_adjust_cost): Ditto.
4784         (is_microcoded_insn): Ditto.
4785         (rs6000_adjust_priority): Ditto.
4786         (rs6000_issue_rate): Ditto.
4787         (rs6000_use_sched_lookahead): Ditto.
4788         (rs6000_use_sched_lookahead_guard): Ditto.
4789         (rs6000_sched_reorder): Ditto.
4790         (force_new_group): Ditto.
4791         * config/rs6000/rs6000.md (cpu attribute): Ditto.
4792         (group_ending_nop): Ditto.
4794 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4796         * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4797         * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4798         rs6000_tune.  Use rs6000_tune instead of rs6000_cpu where appropriate.
4799         (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4800         appropriate.
4801         (rs6000_reassociation_width): Ditto.
4802         (rs6000_emit_epilogue): Ditto.
4803         (rs6000_adjust_cost): Ditto.
4804         (is_microcoded_insn): Ditto.
4805         (is_cracked_insn): Ditto.
4806         (rs6000_adjust_priority): Ditto.
4807         (rs6000_sched_reorder): Ditto.
4808         (rs6000_sched_reorder2): Ditto.
4809         (insn_must_be_first_in_group): Ditto.
4810         (insn_must_be_last_in_group): Ditto.
4811         (rs6000_register_move_cost): Ditto.
4812         * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4813         rs6000_cpu.
4815 2017-12-07  Julia Koval  <julia.koval@intel.com>
4817         * config.gcc: Add vaesintrin.h.
4818         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4819         New type.
4820         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4821         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4822         New builtins.
4823         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4824         * config/i386/immintrin.h: Include vaesintrin.h.
4825         * config/i386/sse.md (vaesdec_<mode>): New pattern.
4826         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4827         _mm_aesdec_epi128): New intrinsics.
4829 2017-12-06  David Malcolm  <dmalcolm@redhat.com>
4831         * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4832         * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4833         c-family/c-spellcheck.cc.
4834         (best_macro_match::best_macro_match): Likewise.
4835         * spellcheck-tree.h
4836         (struct edit_distance_traits<cpp_hashnode *>): Move to
4837         c-family/c-spellcheck.h.
4838         (class best_macro_match): Likewise.
4840 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4842         PR tree-optimization/83293
4843         * gimple-ssa-strength-reduction.c (insert_initializers): Use
4844         GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4845         might insert into empty bb.
4847         PR sanitizer/81281
4848         * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4849         simplify for plus with :c added, and pointer_plus without that.
4850         ((T)P - (T)(P + A) -> -(T) A): Likewise.  If type is integral
4851         with undefined overflow and the conversion is not widening,
4852         perform negation in utype and only convert to type afterwards.
4853         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4854         simplify for plus with :c added, and pointer_plus without that.
4855         If type is integral with undefined overflow and the conversion is
4856         not widening, perform minus in utype and only convert to type
4857         afterwards.  Move the last pointer_diff_expr simplify into the
4858         two outermost ifs.
4860 2017-12-06  Martin Sebor  <msebor@redhat.com>
4862         PR tree-optimization/82646
4863         * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4864         strncpy, not maxlen.
4866 2017-12-06  Martin Sebor  <msebor@redhat.com>
4868         * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4869         nonstring.
4871         PR tree-optimization/83075
4872         * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4873         strncat/strncpy don't change length of source string.
4875 2017-12-06  Eric Botcazou  <ebotcazou@adacore.com>
4877         Revert
4878         2017-11-29  Martin Aberg  <maberg@gaisler.com>
4880         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4881         to prevent b2bst errata sequence.
4882         (sqrtdf2_fix): Likewise.
4884 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4886         PR tree-optimization/81945
4887         * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4888         * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4889         to dest_cfun has orig_loop_num set, either remap it to the new
4890         loop number if the loop got moved too, or clear it.
4892 2017-12-05  Steve Ellcey  <sellcey@cavium.com>
4894         * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4895         to reservation.
4896         (thunderx2t99_nothing): New insn reservation.
4897         (thunderx2t99_mrs): New insn reservation.
4898         (thunderx2t99_multiple): New insn reservation.
4899         (thunderx2t99_alu_basi): Add bfx to reservation.
4900         (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
4902 2017-12-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4904         PR target/82248
4905         * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
4907 2017-12-05  Bin Cheng  <bin.cheng@arm.com>
4909         * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
4910         tree-ssa-pre.c::remove_dead_inserted_code.
4911         * tree-ssa-dce.h: New file.
4912         * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
4913         (remove_dead_inserted_code): Move and rename to function
4914         tree-ssa-dce.c::simple_dce_from_worklist.
4915         (pass_pre::execute): Update use.
4917 2017-12-05  Richard Biener  <rguenther@suse.de>
4919         PR tree-optimization/83277
4920         * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
4921         to code-gen liveout vars.
4923 2017-12-05  Richard Sandiford  <richard.sandiford@linaro.org>
4925         * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
4926         (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
4927         (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
4928         split condition.
4930 2017-12-05  Max Filippov  <jcmvbkbc@gmail.com>
4932         * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
4933         function.
4934         (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
4935         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
4936         ASAN is enabled.
4938 2017-12-05  Richard Biener  <rguenther@suse.de>
4940         * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
4941         * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
4942         (pass_data_cse_sincos): Use TV_TREE_SINCOS.
4943         (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
4945 2017-12-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4947         * dbxout.c (dbxout_block): Grow buf to 30 bytes.
4949 2017-12-05  Martin Liska  <mliska@suse.cz>
4950             Jakub Jelinek  <jakub@redhat.com>
4952         * doc/invoke.texi: Document the options.
4953         * flag-types.h (enum sanitize_code): Add
4954         SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4955         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
4956         of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4957         * opts.c: Define new sanitizer options.
4958         * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
4959         (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
4961 2017-12-05  Julia Koval  <julia.koval@intel.com>
4963         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
4964         OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
4965         (ix86_handle_option): Handle -mavx512vnni.
4966         * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
4967         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
4968         * config/i386/i386-c (__AVX512VNNI__): New.
4969         * config/i386/i386.c (ix86_target_string): Handle new option.
4970         (ix86_valid_target_attribute_inner_p): Handle new option.
4971         * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
4972         * config/i386/i386.opt (mavx512vnni): New option.
4974 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4976         PR target/81616
4977         * athlon.md: Disable for generic.
4978         * haswell.md: Enable for generic.
4979         * i386.c (ix86_sched_init_global): Add core hooks for generic.
4980         * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
4981         to 4.
4982         (ix86_adjust_cost): Move generic to haswell path.
4984 2017-12-04  Eric Botcazou  <ebotcazou@adacore.com>
4986         * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
4987         instead of MEM_P in a couple more places.  Fix formatting issues.
4989 2017-12-04  Jim Wilson  <jimw@sifive.com>
4991         * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
4992         instead of GP_REG_LAST-1.
4993         (riscv_adjust_libcall_cfi_prologue): Likewise.
4994         (riscv_adjust_libcall_cri_epilogue): Likewise.
4995         * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
4996         comment.
4998 2017-12-04  Luis Machado  <luis.machado@linaro.org>
5000         * ipa-pure-const.c (check_decl): Add missing newline.
5001         (state_from_flags): Likewise.
5003 2017-12-04  Jeff Law  <law@redhat.com>
5005         PR tree-optimizatin/78496
5006         * gimple-ssa-evrp-analyze.h
5007         (evrp_range_analyzer::get_vr_values): Simplify.
5008         * gimple-ssa-evrp-analyze.c: Corresponding changes.
5009         * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
5010         and gimple-ssa-evrp-analyze.h.
5011         (dom_opt_dom_walker class): Add evrp_range_analyzer member.
5012         (simplify_stmt_for_jump_threading): Copy a blob of code from
5013         tree-vrp.c to use ranges to simplify statements.
5014         (dom_opt_dom_walker::before_dom_children): Call
5015         evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
5016         (dom_opt_dom_walker::after_dom_children): Similarly for
5017         evrp_range_analyzer::leave.
5018         (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
5019         conditionals.
5021         * gimple-ssa-evrp-analyze.c
5022         (evrp_range_analyzer::extract_range_from_stmt):  Always use
5023         vr_values::update_value_range so preexisting range info is
5024         medged with new range info, even if the new range is VR_VARYING.
5026 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5028         * combine.c: Adjust comment.
5029         (use_crosses_set_p): Delete.
5030         (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
5031         (try_combine): Ditto.
5033 2017-12-04  Richard Biener  <rguenther@suse.de>
5035         PR tree-optimization/83255
5036         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
5037         Re-add zero-iteration check.
5039 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5041         PR rtl-optimization/83245
5042         * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
5043         hard registers as earlyclobber, also if not in an asm.
5045 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
5047         PR bootstrap/83265
5048         Revert
5049         2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5051         PR target/43871
5052         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5053         rs6000_cpu based on cpu_index, not tune_index.
5055 2017-12-04  Richard Biener  <rguenther@suse.de>
5057         PR tree-optimization/83238
5058         * graphite-scop-detection.c (scop_detection::merge_sese): Make
5059         code match comment, rejecting invalid SESE regions.
5061 2017-12-03  John David Anglin  <danglin@gcc.gnu.org>
5063         * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
5064         require base operand is a REG_POINTER prior to reload on targets
5065         with non-equivalent space registers.
5067 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5069         * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
5070         (update_specialized_profile): Fix updating of counts.
5071         (perhaps_add_new_callers): Likewise.
5073 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
5075         PR target/81616
5076         * x86-tune.def: Remove obsolette FIXMEs.
5077         (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
5078         (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
5079         X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
5080         Enable for generic.
5081         (X86_TUNE_PAD_RETURNS): Disable for generic.
5083 2017-12-02  Jakub Jelinek  <jakub@redhat.com>
5085         PR tree-optimization/83170
5086         PR tree-optimization/83241
5087         * gimple-ssa-store-merging.c
5088         (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
5089         gimple_vuse (ins_stmt) in case it has changed.
5090         (imm_store_chain_info::output_merged_store): Likewise.
5092         * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
5093         POINTER_DIFF_EXPR.
5095         PR c++/81212
5096         * tree-cfg.c (pass_warn_function_return::execute): Handle
5097         __builtin_ubsan_handle_missing_return like __builtin_unreachable
5098         with BUILTINS_LOCATION.
5100         PR target/78643
5101         PR target/80583
5102         * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
5103         is BLKmode for vector field with vector raw mode, use TYPE_MODE
5104         instead of DECL_MODE.
5106         * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
5107         last argument to rtx pointer.
5108         * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
5109         with OPERANDS.  For AVX+ 128-bit VEX encoded instructions over 256-bit
5110         or 512-bit.  If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
5111         depending on the chosen ISAs.
5112         * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
5113         *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
5114         *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
5115         callers.
5116         * config/i386/sse.md (mov<mode>_internal): Likewise.
5117         * config/i386/mmx.md (*mov<mode>_internal): Likewise.
5119 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5121         * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
5123 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5125         * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
5126         parameters from prototype.
5127         * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
5128         parameters.  Don't print a comment.
5129         (emit_fusion_gpr_load): Adjust.
5130         (emit_fusion_load_store): Adjust.
5131         * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
5132         * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
5133         comment on the second line.
5135 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5137         PR target/43871
5138         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5139         rs6000_cpu based on cpu_index, not tune_index.
5141 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5143         * final.c (output_asm_name): Print insn_cost.  Shorten output.  Print
5144         which_alternative instead of which_alternative + 1.
5145         (output_asm_insn): Print an extra tab if the template is short.
5147 2017-12-01  Jim Wilson  <jimw@sifive.com>
5149         * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
5150         comment.
5151         * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
5152         * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
5153         reference.
5154         * doc/tm.texi: Regenerate.
5156 2017-12-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
5158         PR target/81959
5159         * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
5160         whether we can allocate pseudos before trying to fix an address.
5161         * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
5162         memory address is indexed or indirect.
5163         (floatuns_<mode>si2_hw2): Likewise.
5165 2017-12-01  Jason Merrill  <jason@redhat.com>
5167         * Makefile.in (TAGS): Add c-family/*.cc.
5169 2017-12-01  Wilco Dijkstra  <wdijkstr@arm.com>
5171         * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
5172         (call_value_insn): Likewise.
5173         (sibcall_insn): Likewise.
5174         (sibcall_value_insn): Likewise.
5175         (movsi_aarch64): Likewise.
5176         (movdi_aarch64): Likewise.
5177         (add_losym_): Likewise.
5178         (ldr_got_small_): Likewise.
5179         (ldr_got_small_sidi): Likewise.
5180         (ldr_got_small_28k_): Likewise.
5181         (ldr_got_small_28k_sidi): Likewise.
5182         * config/aarch64/aarch64.c (aarch64_print_address_internal):
5183         Move output_addr_const to symbolic case. Add error check.
5185 2017-12-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5187         * config/s390/predicates.md (plus16_Q_operand): New predicate.
5188         * config/s390/s390.md: Disable MVC merging peephole if it would
5189         disable operand forwarding.
5190         (new peephole2): Split MVCs if it would turn them into up to 2
5191         forwardable MVCs.
5193 2017-12-01  Richard Biener  <rguenther@suse.de>
5195         PR tree-optimization/83232
5196         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
5197         detection of same access. Instead of breaking the group here
5198         do not consider the duplicate.  Add comment explaining real fix.
5200 2017-12-01  Jonathan Wakely  <jwakely@redhat.com>
5202         * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
5204 2017-12-01  Sudakshina Das  <sudi.das@arm.com>
5206         * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
5208 2017-12-01  Jakub Jelinek  <jakub@redhat.com>
5210         * function.h (struct function): Remove cilk_frame_decl,
5211         is_cilk_function and calls_cilk_spawn fields.
5212         * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
5213         field.
5214         * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
5215         cilk_elemental field.
5216         * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
5217         * target.def: Adjust comment.
5218         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5219         Don't test cilk_elemental.
5221         PR tree-optimization/83233
5222         * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
5223         bswap_stat name for the struct.
5225         PR c/79153
5226         * tree.h (SWITCH_BREAK_LABEL_P): Define.
5227         * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5228         starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5229         SWITCH_BREAK_LABEL_P set on the label.
5230         (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5231         added for default case if it was missing and not all cases covered.
5232         Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5233         switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5234         set on the label.
5235         * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5236         to avoid -Wimplicit-fallthrough warning.
5237         * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5238         FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5240         PR tree-optimization/83221
5241         * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5242         down by 16.
5243         (init_reassoc): Formatting fix.
5245         PR sanitizer/81275
5246         * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5247         C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5248         is diagnosed.
5250         PR sanitizer/83219
5251         * tree-cfg.c: Include asan.h.
5252         (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5254 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5256         * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5258 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5260         * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5261         vpcmpeqd instruction.
5263 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5265         * config/i386/i386.c (standard_sse_constant_opcode): Fix
5266         registers type for 128bit mode.
5268 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5270         * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5271         of ;;.
5272         * gengtype-state.c (read_state_pair): Likewise.
5273         * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5274         * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5275         * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5276         * ifcvt.c (noce_try_store_flag_constants): Likewise.
5277         * tree-ssa-ccp.c (ccp_finalize): Likewise.
5278         * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5279         * builtins.c (fold_builtin_3): Likewise.
5280         * graphite-scop-detection.c
5281         (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5282         * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5283         Likewise.
5285 2017-12-01  Maxim Ostapenko  <m.ostapenko@samsung.com>
5287         PR sanitizer/81697
5288         * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5289         parameter. Return true if ignore_decl_rtl_set_p is true and other
5290         conditions are satisfied.
5291         * asan.h (asan_protect_global): Add new parameter.
5292         * varasm.c (categorize_decl_for_section): Pass true as second parameter
5293         to asan_protect_global calls.
5295 2017-11-30  Jim Wilson  <jimw@sifive.com>
5297         * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5298         -mno-memcpy options.  For -mplt, -mfdiv, -mdiv, -msave-restore, and
5299         -mstrict-align, add info on default value.  Delete redundant lines for
5300         -mabi.  Add missing -mexplicit-relocs docs.
5302 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5304         * config/arc/arc.md (trap): New pattern.
5306 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5308         * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5309         instruction to end into a delay slot.
5310         * config/arc/arc.md (cond_delay_insn): Check if the instruction
5311         can be placed into a delay slot against reg_note.
5313 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5315         * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5316         labels number of usages.
5318 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5320         * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5321         function.
5322         (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5324 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5326         PR target/83210
5327         * internal-fn.c (expand_mul_overflow): Optimize unsigned
5328         multiplication by power of 2 constant into two shifts + comparison.
5330 2017-11-30  Jan Hubicka  <hubicka@ucw.cz>
5332         PR target/81616
5333         * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5335 2017-11-30  Richard Biener  <rguenther@suse.de>
5337         PR tree-optimization/83202
5338         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5339         allow_peel argument and guard peeling.
5340         (canonicalize_loop_induction_variables): Likewise.
5341         (canonicalize_induction_variables): Pass false.
5342         (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5343         peeling from cunrolli.
5345 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5347         * combine.c (try_combine): Print a message to dump file whenever
5348         I0, I1, or I2 cannot be combined into I3.
5350 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5352         PR rtl-optimization/83156
5353         PR rtl-optimization/82621
5354         * combine.c (try_combine): Don't split an I2 if one of the dests is
5355         set again before I3.  Allow unused dests.
5357 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5359         * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5361 2017-11-29  Vladimir Makarov  <vmakarov@redhat.com>
5363         PR rtl-optimization/80818
5364         * lra.c (collect_non_operand_hard_regs): New arg insn.  Pass it
5365         recursively.  Use insn code for clobber.
5366         (lra_set_insn_recog_data): Pass the new arg to
5367         collect_non_operand_hard_regs.
5368         (add_regs_to_insn_regno_info): Pass insn instead of uid.  Use insn
5369         code for clobber.
5370         (lra_update_insn_regno_info): Pass insn to
5371         add_regs_to_insn_regno_info.
5373 2017-11-29  Jim Wilson  <jimw@sifive.com>
5374             Andrew Waterman  <andrew@sifive.com>
5376         * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5377         (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5378         SINGLE_SHIFT_COST.  Case LT and ZERO_EXTEND, likewise.  Case ASHIFT,
5379         use SINGLE_SHIFT_COST.
5380         * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5381         (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5383 2017-11-29  Julia Koval  <julia.koval@intel.com>
5385         * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5386         _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5387         _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5388         _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5389         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5390         _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5391         _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5392         _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5393         _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5394         _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5395         _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5396         * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5397         __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5398         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5399         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5400         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5401         __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5402         __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5403         __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5404         __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5405         __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5406         __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5407         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5408         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5409         __builtin_ia32_vpshldv_v2di_mask,
5410         __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5411         * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5412         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5414 2017-11-29  Julia Koval  <julia.koval@intel.com>
5416         * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5417         _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5418         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5419         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5420         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5421         _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5422         _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5423         _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5424         _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5425         _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5426         _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5427         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5428         V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5429         V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5430         V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5431         V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5432         V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5433         V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5434         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5435         * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5436         vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5438 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5440         * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5441         movsi_pic_gotdata_op instruction as a load for the UT699 errata
5442         workaround.
5444 2017-11-29  Martin Aberg  <maberg@gaisler.com>
5446         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5447         to prevent b2bst errata sequence.
5448         (sqrtdf2_fix): Likewise.
5450 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5452         * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5453         (div_sqrt_insn_p): New function.
5454         (sparc_do_work_around_errata): Insert NOP instructions to
5455         prevent sequences that could trigger the TN-0013 errata for
5456         certain LEON3 processors.
5457         (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5458         (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5459         * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5460         (in_branch_delay): Prevent div and sqrt in delay slot if
5461         fix_lost_divsqrt.
5462         * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5464 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5466         * config/sparc/sparc.c (atomic_insn_p): New function.
5467         (sparc_do_work_around_errata): Insert NOP instructions to
5468         prevent sequences that could trigger the TN-0010 errata for
5469         UT700.
5470         * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5471         instruction referable in atomic_insns_p.
5473 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5475         * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5476         (atomic_compare_and_swap_leon3_1): Likewise.
5477         (ldstub): Likewise.
5479 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5481         * config/sparc/sparc.c (fpop_insn_p): New function.
5482         (sparc_do_work_around_errata): Insert NOP instructions to
5483         prevent sequences that could trigger the TN-0012 errata for
5484         GR712RC.
5485         (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5486         * config/sparc/sparc.md (fix_gr712rc): New attribute.
5487         (in_branch_annul_delay): Prevent floating-point instructions
5488         in delay slot of annulled integer branch.
5490 2017-11-29  Richard Biener  <rguenther@suse.de>
5492         PR tree-optimization/83202
5493         * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5494         (bst_fail): Use it.
5495         (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5496         nodes vectorized to the same stmts multiple times.
5497         (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5498         (vect_analyze_slp_instance): Adjust.
5499         (scalar_stmts_to_slp_tree_map_t): New typedef.
5500         (vect_schedule_slp_instance): Add a map recording the SLP node
5501         representing the vectorized stmts for a set of scalar stmts.
5502         Avoid code-generating redundancies.
5503         (vect_schedule_slp): Allocate map and pass it down.
5505 2017-11-29  Nathan Sidwell  <nathan@acm.org>
5507         PR c++/83187
5508         * tree.c (build_complex_type): Fix canonicalization.  Only fill in
5509         type if it is new.
5511 2017-11-29  Wilco Dijkstra  <wdijkstr@arm.com>
5513         * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5514         cases for printing LDP/STP memory addresses.
5515         (aarch64_print_address_internal): Renamed from
5516         aarch64_print_operand_address, added parameter, add Pmode check.
5517         (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5518         (aarch64_print_operand_address): Indirect to
5519         aarch64_print_address_internal.
5520         * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5521         'y' operand output specifier.
5523 2017-11-29  Jakub Jelinek  <jakub@redhat.com>
5525         PR middle-end/83185
5526         * tree.c (build_simple_mem_ref_loc): Handle
5527         get_addr_base_and_unit_offset returning a MEM_REF.
5529         PR middle-end/80929
5530         * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5532         PR target/80819
5533         * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5534         alternative.
5536 2017-11-28  David Malcolm  <dmalcolm@redhat.com>
5538         * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5539         m_x_offset.
5540         (layout::move_to_column): Likewise.
5542 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5544         PR sanitizer/81275
5545         * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5546         is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5548 2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5549             Martin Jambor  <mjambor@suse.cz>
5551         PR ipa/82808
5552         * tree.h (expr_type_first_operand_type_p): Declare
5553         * tree.c (expr_type_first_operand_type_p): New function.
5554         * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5555         (ipa_value_from_jfunc): Adjust declaration.
5556         * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5557         Use it as result type for arithmetics, unless it is NULL in which case
5558         be more conservative.
5559         (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5560         ipa_get_jf_pass_through_result.
5561         (propagate_vals_across_pass_through): Likewise.
5562         (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5563         is to propagate_vals_across_pass_through.
5564         (propagate_constants_across_call): Pass PARM_TYPE to
5565         propagate_scalar_across_jump_function.
5566         (find_more_scalar_values_for_callers_subset): Pass parameter type to
5567         ipa_value_from_jfunc.
5568         (cgraph_edge_brings_all_scalars_for_node): Likewise.
5569         * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5570         to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5571         * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5572         target_type, pass it to ipa_value_from_jfunc.
5573         (update_indirect_edges_after_inlining): Pass parameter type to
5574         try_make_edge_direct_simple_call.
5576 2017-11-28  Jeff Law  <law@redhat.com>
5578         * gimple-ssa-evrp-analyze.c
5579         (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5580         refine ranges if scev_initialized_p returns true.
5581         * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5583 2017-11-28  Julia Koval  <julia.koval@intel.com>
5585         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5586         _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5587         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5588         _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5589         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5590         _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5591         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5592         _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5593         _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5594         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5595         _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5596         * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5597         __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5598         __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5599         __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5600         __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5601         __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5602         __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5603         __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5604         __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5605         __builtin_ia32_vpshrd_v2di_mask): New builtins.
5606         * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5608 2017-11-28  Julia Koval  <julia.koval@intel.com>
5610         * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5611         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5612         _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5613         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5614         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5615         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5616         _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5617         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5618         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5619         _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5620         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5621         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5622         V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5623         V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5624         V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5625         V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5626         V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5627         V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5628         V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5629         V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5630         * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5631         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5632         __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5633         __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5634         __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5635         __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5636         __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5637         __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5638         __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5639         __builtin_ia32_vpshld_v2di_mask): New builtins.
5640         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5641         * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5643 2017-11-28  Richard Biener  <rguenther@suse.de>
5645         PR tree-optimization/80776
5646         * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5647         Declare.
5648         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5649         New function.
5650         (evrp_range_analyzer::record_ranges_from_incoming_edges):
5651         If the incoming edge is an effective fallthru because the other
5652         edge only reaches a __builtin_unreachable () then record ranges
5653         derived from the controlling condition in SSA info.
5654         (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5655         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5657 2017-11-28  Olivier Hainque  <hainque@adacore.com>
5659         * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5660         on mingw build hosts.
5662 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5664         PR debug/81307
5665         * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5667 2017-11-28  Richard Biener  <rguenther@suse.de>
5669         PR middle-end/83141
5670         * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5671         copies generated from memcpy use a character array as reference
5672         type.
5674 2017-11-28  Julia Koval  <julia.koval@intel.com>
5675             Sebastian Peryt  <sebastian.peryt@intel.com>
5677         * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5678         c-family/c-cilkplus.o, c-family/array-notation-common.o,
5679         cilk-common.o, cilk.h, cilk-common.c): Remove.
5680         * builtin-types.def
5681         (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5682         * builtins.c (is_builtin_name): Remove cilkplus condition.
5683         (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5684         * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5685         cilk-builtins.def, cilkplus.def): Remove.
5686         * cif-code.def (CILK_SPAWN): Remove.
5687         * cilk-builtins.def: Delete.
5688         * cilk-common.c: Ditto.
5689         * cilk.h: Ditto.
5690         * cilkplus.def: Ditto.
5691         * config/darwin.h (fcilkplus): Delete.
5692         * cppbuiltin.c: Ditto.
5693         * doc/extend.texi: Remove cilkplus doc.
5694         * doc/generic.texi: Ditto.
5695         * doc/invoke.texi: Ditto.
5696         * doc/passes.texi: Ditto.
5697         * gcc.c (fcilkplus): Remove.
5698         * gengtype.c (cilk.h): Remove.
5699         * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5700         support.
5701         * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5702         Remove.
5703         * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5704         gimplify_call_expr,
5705         is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5706         gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5707         cilkplus conditions.
5708         * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5709         inline_read_section): Ditto.
5710         * ipa-inline-analysis.c (cilk.h): Remove.
5711         * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5712         * lto-wrapper.c (merge_and_complain, append_compiler_options,
5713         append_linker_options): Remove condition for fcilkplus.
5714         * lto/lto-lang.c (cilk.h): Remove.
5715         (lto_init): Remove condition for fcilkplus.
5716         * omp-expand.c (expand_cilk_for_call): Delete.
5717         (expand_omp_taskreg, expand_omp_for_static_chunk,
5718         expand_omp_for): Remove cilkplus
5719         conditions.
5720         (expand_cilk_for): Delete.
5721         * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5722         * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5723         execute_lower_omp, diagnose_sb_0): Ditto.
5724         * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5725         * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5726         * tree-nested.c: Ditto.
5727         * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5728         (dump_generic_node): Ditto.
5729         * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5730         * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5731         Delete.
5732         * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5734 2017-11-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5736         * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5737         * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5738         for vector mode and !TARGET_SIMD.
5740 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5742         * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5743         Adjust comment.
5744         * tree.h (SWITCH_LABELS): Remove.
5745         * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5746         assert SWITCH_BODY is non-NULL.
5747         * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5748         handling.
5749         * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5751         PR tree-optimization/80788
5752         * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5753         has TREE_OVERFLOW set, call drop_tree_overflow.
5755 2017-11-28  Richard Biener  <rguenther@suse.de>
5757         PR tree-optimization/83158
5758         * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5760 2017-11-28  Segher Boessenkool  <segher@kernel.crashing.org>
5762         PR 81288/target
5763         * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5764         TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.
5766 2017-11-27  Segher Boessenkool  <segher@kernel.crashing.org>
5768         * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5769         LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5771 2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5773         PR middle_end/82333
5774         * varasm.c (compare_constant): Take the mode of the constants into
5775         account when comparing floating point constants.
5777 2017-11-27  Gerald Pfeifer  <gerald@pfeifer.com>
5779         * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5780         from explicit instantiation of debug_helper.
5781         * vec.h (DEFINE_DEBUG_VEC): Ditto.
5783 2017-11-27  Richard Biener  <rguenther@suse.de>
5785         * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5786         refactor a bit.
5788 2017-11-27  Richard Biener  <rguenther@suse.de>
5790         * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5791         (type_hash_canon): Also clear int_cst_hash_table entry for
5792         TYPE_MIN/MAX_VALUE.
5793         (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5795 2017-11-27  Tamar Christina  <tamar.christina@arm.com>
5797         * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5798         * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5799         * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5801 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5803         * hash-map.h (gt_cleare_cache): Avoid UB.
5805 2017-11-27  Eric Botcazou  <ebotcazou@adacore.com>
5807         * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5808         * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5809         (decide_unroll_runtime_iterations): Likewise.
5810         (decide_unroll_stupid): Likewise.
5812 2017-11-27  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5814         PR target/83109
5815         * config/i386/i386.md: Add a loop with incssp.
5817 2017-11-27  Martin Jambor  <mjambor@suse.cz>
5819         PR tree-optimization/81248
5820         * tree-sra.c (splice_param_accesses): Remove size check.
5821         (decide_one_param_reduction): Fix size check.
5822         * gimple-pretty-print.c (dump_profile): Silence warning.
5823         * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5825 2017-11-27  Jakub Jelinek  <jakub@redhat.com>
5827         PR debug/81307
5828         * dbxout.c (lastlineno): New variable.
5829         (dbx_debug_hooks): Use dbxout_switch_text_section as
5830         switch_text_section debug hook.
5831         (dbxout_function_end): Switch to current_function_section
5832         rather than function_section.  If crtl->has_bb_partition,
5833         output just one N_FUN, depending on in_cold_section_p.
5834         (dbxout_source_line): Remember last lineno in lastlineno.
5835         (dbxout_switch_text_section): New function.
5836         (dbxout_function_decl): Adjust dbxout_block caller.
5837         (dbx_block_with_cold_children): New function.
5838         (dbxout_block): Return true if any LBRAC/RBRAC have been
5839         emitted.  Use dbx_block_with_cold_children at depth == 0
5840         in second partition.  Add PARENT_BLOCKNUM argument, pass
5841         it optionally adjusted to children.  Output LBRAC/RBRAC
5842         around recursive call only if the block is in the current
5843         partition, if not and anything was output, emit empty
5844         range LBRAC/RBRAC.
5845         * final.c (final_scan_insn): Compute cold_function_name
5846         before calling switch_text_section debug hook.  Call
5847         that hook even if dwarf2out_do_frame if not emitting
5848         dwarf debug info.
5850         PR target/83100
5851         * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5852         TREE_READONLY decls.
5854 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5856         PR rtl-optimization/82488
5857         * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5859 2017-11-26  Julia Koval  <julia.koval@intel.com>
5861         * config/i386/i386.c (processor_target_table): Add skylake_cost for
5862         skylake-avx512.
5863         * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5864         skylake_cost): New.
5866 2017-11-26  Julia Koval  <julia.koval@intel.com>
5868         * config/i386/driver-i386.c (host_detect_local_cpu):
5869         Detect skylake-avx512.
5871 2017-11-26  Julia Koval  <julia.koval@intel.com>
5873         * config.gcc: Add -march=cannonlake.
5874         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5875         * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5876         * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5877         (PTA_CANNONLAKE): New.
5878         (processor_target_table): Add cannonlake.
5879         (ix86_option_override_internal): Ditto.
5880         (fold_builtin_cpu): Ditto.
5881         (get_builtin_code_for_version): Handle cannonlake.
5882         (M_INTEL_COREI7_CANNONLAKE): New.
5883         * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5884         * doc/invoke.texi: Add -march=cannonlake.
5886 2017-11-14  Boris Kolpackov  <boris@codesynthesis.com>
5888         * plugin.c (add_new_plugin): Use platform-specific library extensions.
5889         (try_init_one_plugin): Alternative implementation for MinGW.
5890         * Makefile.in (plugin_implib): New.
5891         (gengtype-lex.c): Fix broken AIX workaround.
5892         * configure: Regenerate.
5893         * doc/plugins.texi: Document support for MinGW.
5895 2017-11-25  Jakub Jelinek  <jakub@redhat.com>
5897         PR rtl-optimization/81553
5898         * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
5899         to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
5900         is a shift where C1 has different mode than the whole shift, use C1's
5901         mode for MULT rather than the shift's mode.
5903         PR target/82848
5904         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
5905         builtins not enabled in the currently selected ISA.
5907 2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>
5909         PR tree-optimization/71026
5910         * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
5911         (insert_reciprocals): Change to insert reciprocals before a division
5912         by a square and to insert the square of a reciprocal.
5913         (execute_cse_reciprocals_1): Change to consider division by a square.
5914         (register_division_in): Add importance parameter.
5916 2017-11-24  Richard Biener  <rguenther@suse.de>
5918         PR tree-optimization/82402
5919         * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
5920         set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
5922 2017-11-24  Marc Glisse  <marc.glisse@inria.fr>
5924         * match.pd (0-ptr): New transformation.
5926 2017-11-24  Jan Hubicka  <hubicka@ucw.cz>
5928         PR bootstrap/83015
5929         * ipa-inline.c (inline_small_functions): Set current badnes correctly
5930         when skipping checking.
5932 2017-11-24  Richard Biener  <rguenther@suse.de>
5934         PR tree-optimization/83128
5935         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
5936         (vn_reference_lookup_3): Likewise.
5938 2017-11-24  Jakub Jelinek  <jakub@redhat.com>
5940         PR sanitizer/83014
5941         * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
5942         instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
5943         tree_to_uhwi twice.
5945         * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
5946         ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
5947         check for the latter with gimple_call_builtin_p.  Do not handle
5948         BUILT_IN_STPNCPY_CHK which is not a pass through call.
5950 2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>
5952         * config/arm/arm_neon.h: Fix pragma GCC push_options before
5953         vdot_u32.
5955 2017-11-23  Julia Koval  <julia.koval@intel.com>
5957         * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
5958         _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
5959         _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
5960         _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
5961         _mm512_maskz_expandloadu_epi16): New intrinsics.
5962         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
5963         _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
5964         _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
5965         _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
5966         _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
5967         _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
5968         _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
5969         _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
5970         _mm256_maskz_expandloadu_epi8): New intrinsics.
5971         * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
5972         V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
5973         V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
5974         V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
5975         * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
5976         * config/i386/sse.md (VI248_VLBW): New iterator.
5977         (expand<mode>_mask, expand<mode>_maskz): New patterns.
5979 2017-11-23  Julia Koval  <julia.koval@intel.com>
5981         * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
5982         * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
5983         _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
5984         _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
5985         _mm512_mask_compressstoreu_epi16): New.
5986         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
5987         _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
5988         _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
5989         _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
5990         _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
5991         _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
5992         _mm256_mask_compressstoreu_epi8): New.
5993         * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
5994         VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
5995         VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
5996         VOID_FTYPE_PV8HI_V8HI_UQI): New types.
5997         * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
5998         __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
5999         __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
6000         __builtin_ia32_compresshi128_mask,
6001         __builtin_ia32_compressstoreuqi512_mask,
6002         __builtin_ia32_compressstoreuhi512_mask,
6003         __builtin_ia32_compressstoreuqi256_mask,
6004         __builtin_ia32_compressstoreuqi128_mask,
6005         __builtin_ia32_compressstoreuhi256_mask,
6006         __builtin_ia32_compressstoreuhi128_mask): New builtins.
6007         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
6008         array for flags2.
6009         (ix86_expand_special_args_builtin): Handle new types.
6010         (s4fma_expand): Handle new builtin array.
6011         * config/i386/immintrin.h: Include new headers.
6012         * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
6013         (compress<mode>_mask, compressstore<mode>_mask): New patterns.
6015 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6017         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
6018         fixes.  Declare temp and g variables at the top in order to avoid
6019         {} in most of the cases.
6021 2017-11-23  Marc Glisse  <marc.glisse@inria.fr>
6023         * match.pd (ptr-0): New transformation.
6025 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
6027         * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
6028         (struct addr_mode_cost_table): New.
6029         (struct tune_params): Add field addr_mode_costs.
6030         * config/arm/arm.c (generic_addr_mode_costs): New.
6031         (arm_slowmul_tune): Initialise addr_mode_costs field.
6032         (arm_fastmul_tune): Likewise.
6033         (arm_strongarm_tune): Likewise.
6034         (arm_xscale_tune): Likewise.
6035         (arm_9e_tune): Likewise.
6036         (arm_marvell_pj4_tune): Likewise.
6037         (arm_v6t2_tune): Likewise.
6038         (arm_cortex_tune): Likewise.
6039         (arm_cortex_a8_tune): Likewise.
6040         (arm_cortex_a7_tune): Likewise.
6041         (arm_cortex_a15_tune): Likewise.
6042         (arm_cortex_a35_tune): Likewise.
6043         (arm_cortex_a53_tune): Likewise.
6044         (arm_cortex_a57_tune): Likewise.
6045         (arm_exynosm1_tune): Likewise.
6046         (arm_xgene1_tune): Likewise.
6047         (arm_cortex_a5_tune): Likewise.
6048         (arm_cortex_a9_tune): Likewise.
6049         (arm_cortex_a12_tune): Likewise.
6050         (arm_cortex_a73_tune): Likewise.
6051         (arm_v7m_tune): Likewise.
6052         (arm_cortex_m7_tune): Likewise.
6053         (arm_v6m_tune): Likewise.
6054         (arm_fa726te_tune): Likewise.
6055         (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
6057 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
6059         * config/arm/arm.c (arm_mem_costs): New function.
6060         (arm_rtx_costs_internal): Use arm_mem_costs.
6062 2017-11-23  Mark Wielaard  <mark@klomp.org>
6064         * dwarf2out.c (init_sections_and_labels): Use generation to create
6065         unique ranges_section_label and ranges_base_label. Return generation.
6066         (output_rnglists): Add generation argument. Use generation to create
6067         unique ranges labels.
6068         (dwarf2out_finish): Get generation from init_sections_and_labels
6069         and pass generation to output_rnglists.
6071 2017-11-23  Mike Stump  <mikestump@comcast.net>
6072             Eric Botcazou  <ebotcazou@adacore.com>
6074         * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
6075         * cfgloop.h (struct loop): Add unroll field.
6076         * function.h (struct function): Add has_unroll bitfield.
6077         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
6078         (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
6079         * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
6080         (pass_rtl_unroll_loops::gate): Likewise.
6081         * loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
6082         for which loop->unroll==1.
6083         (decide_unroll_constant_iterations): Use note for consistency and
6084         take loop->unroll into account.  Return early if loop->unroll is set.
6085         Fix thinko in existing test.
6086         (decide_unroll_runtime_iterations): Use note for consistency and
6087         take loop->unroll into account.
6088         (decide_unroll_stupid): Likewise.
6089         * lto-streamer-in.c (input_cfg): Read loop->unroll.
6090         * lto-streamer-out.c (output_cfg): Write loop->unroll.
6091         * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
6092         New case.
6093         (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
6094         (print_loop): Print loop->unroll if set.
6095         * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
6096         * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
6097         * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
6098         New case.
6099         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
6100         loop->unroll is set and smaller than the trip count.  Otherwise bypass
6101         entirely the heuristics if loop->unroll is set.  Remove dead note.
6102         Fix off-by-one bug in other note.
6103         (try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
6104         (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
6105         is greater than 1.
6106         (tree_unroll_loops_completely): Make static.
6107         (pass_complete_unroll::execute): Use correct type for variable.
6108         (pass_complete_unrolli::execute): Fix formatting.
6109         * tree.def (ANNOTATE_EXPR): Add 3rd operand.
6111 2017-11-23  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6113         * config/i386/i386.h (TARGET_PREFER_AVX256): Also
6114         enable when TARGET_PREFER_AVX128 is set.
6116 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6118         * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
6119         frequency.
6120         * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
6121         (cgraph_edge::maybe_hot_p): Use sreal frequencies.
6123 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6125         * ipa-fnsummary.c (record_modified_bb_info): Use sreal
6126         frequencies. Fix estimation of aggregate parameters.
6128 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6130         * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
6131         when inlining.
6133 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6135         * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
6137 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6139         * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
6140         macro.
6142 2017-11-23  Oleg Endo  <olegendo@gcc.gnu.org>
6144         PR target/83111
6145         * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
6146         sibcall_value_pcrel_fdpic): Use local variable instead of
6147         operands[3].
6148         (calli_tbr_rel): Add missing operand 2.
6149         (call_valuei_tbr_rel): Add missing operand 3.
6151 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6153         PR middle-end/82253
6154         * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
6155         bitpos/bitsize covering the whole destination, use store_expr only if
6156         the complex mode is the same.  Otherwise, use expand_normal and if
6157         it returns CONCAT, subreg each part separately instead of trying to
6158         subreg the whole result.
6160 2017-11-23  Richard Biener  <rguenther@suse.de>
6162         PR tree-optimization/23094
6163         * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
6164         come here from walking over backedges in the first iteration.
6165         (vn_reference_lookup_3): Skip clobbers that store the same value.
6167 2017-11-23  Richard Biener  <rguenther@suse.de>
6169         PR tree-optimization/81403
6170         * tree-ssa-pre.c (get_representative_for): Add parameter specifying
6171         a block we need a leader relative to.
6172         (phi_translate_1): For nary processing require a leader from
6173         get_representative_for given we run expression simplification
6174         using match-and-simplify.  Remove previous fix.
6176 2017-11-22  Jeff Law  <law@redhat.com>
6178         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
6179         Use new method allocate_value_range rather than accessing the
6180         vrp_value_range_pool data member directly.
6181         * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
6182         to use extract_range_from_stmt method to avoid need for
6183         extract_range_from_assignment method.
6184         (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
6185         method rather than setting values_propgated data member directly.
6186         * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
6187         and values propagated data members and extract_range_from_assignment
6188         method.  Reorder private data members to conform to standards.
6189         Add new methods set_lattice_propagation_complete and
6190         allocate_value_range.
6192 2017-11-22  Eric Botcazou  <ebotcazou@adacore.com>
6194         PR rtl-optimization/83030
6195         * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
6196         for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
6197         (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
6198         * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
6199         CROSSING_JUMP_P flag is already set before setting it.
6200         * cfgrtl.c (fixup_partition_crossing): Likewise.
6201         * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
6202         insn as useless.
6204 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6206         * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
6207         Handle the case where both arguments are using gen_const_vec_series.
6209 2017-11-22  David Malcolm  <dmalcolm@redhat.com>
6211         PR c++/62170
6212         * pretty-print.c (pp_format): Move quoting implementation to
6213         pp_begin_quote and pp_end_quote.  Update pp_format_decoder call
6214         to pass address of "quote" local.
6215         (pp_begin_quote): New function.
6216         (pp_end_quote): New function.
6217         * pretty-print.h (printer_fn): Convert penultimate param from bool
6218         to bool *.
6219         (pp_begin_quote): New decl.
6220         (pp_end_quote): New decl.
6221         * tree-diagnostic.c (default_tree_printer): Convert penultimate
6222         param from bool to bool *.
6223         * tree-diagnostic.h (default_tree_printer): Likewise.
6225 2017-11-22  Jeff Law  <law@redhat.com>
6227         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6228         Initialize vr_values.
6229         (evrp_range_analyzer::try_find_new_range): Call methods attached to
6230         vr_values via vr_values class instance rather than delegators.
6231         (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6232         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6233         (evrp_range_analyzer::push_value_range): Likewise.
6234         (evrp_range_analyzer::pop_value_range): Likewise.
6235         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6236         most delegators.  Those remaining are exposed as public interfaces.
6237         Make vr_values a pointer and private.
6238         (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6239         vr_values.
6240         (evrp_range_analyzer::get_vr_value): New method.
6241         * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6242         (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6243         (class evrp_dom_walker): Attach evrp_folder class, initialize
6244         it in the ctor.  Remove temporary delegators.
6245         (evrp_dom_walker::before_dom_children): Call methods in attached
6246         evrp_range_analyzer class via class instance pointer.  Use
6247         free value_range_constant_singleton to remove need for
6248         op_with_constant_singleton_value delegator method.  Do not
6249         create a vrp_prop class instance for every call!  Narrow
6250         scope of a couple variables.
6251         (evrp_dom_walker::cleanup): Call methods in attached
6252         evrp_range_analyzer class via class instance pointer.
6253         * vr-values.h (class vr_values): Privatize many methods and
6254         data members.
6256 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6258         * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6259         * cfgexpand.c (expand_debug_expr): Remove handling for them.
6260         * expr.c (expand_expr_real_2): Likewise.
6261         * fold-const.c (const_unop): Likewise.
6262         * optabs-tree.c (optab_for_tree_code): Likewise.
6263         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6264         * tree-inline.c (estimate_operator_cost): Likewise.
6265         * tree-pretty-print.c (dump_generic_node): Likewise.
6266         (op_code_prio): Likewise.
6267         (op_symbol_code): Likewise.
6268         * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6269         (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6270         * internal-fn.c (direct_internal_fn_optab): New function.
6271         (direct_internal_fn_array, direct_internal_fn_supported_p
6272         (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6273         * fold-const-call.c (fold_const_reduction): New function.
6274         (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6275         CFN_REDUC_MIN.
6276         * tree-vect-loop.c: Include internal-fn.h.
6277         (reduction_code_for_scalar_code): Rename to...
6278         (reduction_fn_for_scalar_code): ...this and return an internal
6279         function.
6280         (vect_model_reduction_cost): Take an internal_fn rather than
6281         a tree_code.
6282         (vect_create_epilog_for_reduction): Likewise.  Build calls rather
6283         than assignments.
6284         (vectorizable_reduction): Use internal functions rather than tree
6285         codes for the reduction operation.  Update calls to the functions
6286         above.
6287         * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6288         Use calls to internal functions rather than REDUC tree codes.
6289         * config/aarch64/aarch64-simd.md: Update comment accordingly.
6291 2017-11-22  Olivier Hainque  <hainque@adacore.com>
6293         * config/vxworks.c (vxworks_override_options): Pick default
6294         dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6295         * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6296         DWARF_GNAT_ENCODINGS_DEFAULT.
6297         * config/vxworksae.h: Likewise.
6299 2017-11-22  Marc Glisse  <marc.glisse@inria.fr>
6301         PR tree-optimization/83104
6302         * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6303         not result.
6305 2017-11-22  Marek Polacek  <polacek@redhat.com>
6306             H.J. Lu  <hongjiu.lu@intel.com>
6307             Jason Merrill  <jason@redhat.com>
6309         PR c++/60336
6310         PR middle-end/67239
6311         PR target/68355
6312         * calls.c (initialize_argument_information): Call
6313         warn_parameter_passing_abi target hook.
6314         (store_one_arg): Use 0 for empty record size.  Don't push 0 size
6315         argument onto stack.
6316         (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6317         * common.opt: Update -fabi-version description.
6318         * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6319         (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6320         int_size_in_bytes.
6321         (ix86_is_empty_record): New function.
6322         (ix86_warn_parameter_passing_abi): New function.
6323         (TARGET_EMPTY_RECORD_P): Redefine.
6324         (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6325         * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6326         * doc/tm.texi: Regenerated.
6327         * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6328         TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6329         * dwarf2out.c (get_ultimate_context): Move to tree.c.
6330         * explow.c (hard_function_value): Call arg_int_size_in_bytes
6331         instead of int_size_in_bytes.
6332         * expr.c (copy_blkmode_to_reg): Likewise.
6333         * function.c (aggregate_value_p): Return 0 for empty types.
6334         (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6335         target hook.
6336         (locate_and_pad_parm): Call arg size_in_bytes instead
6337         size_in_bytes.
6338         * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6339         * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6340         * target.def (empty_record_p, warn_parameter_passing_abi): New target
6341         hooks.
6342         * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6343         (std_gimplify_va_arg_expr): Skip empty records.  Call
6344         arg_size_in_bytes instead size_in_bytes.
6345         * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6346         * tree-core.h (tree_type_common): Add empty_flag.
6347         (tree_decl_common): Update comments.
6348         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6349         DECL_PADDING_P.
6350         (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6351         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6352         DECL_PADDING_P.
6353         (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6354         * tree.c (default_is_empty_type): New function.
6355         (default_is_empty_record): New function.
6356         (arg_int_size_in_bytes): New function.
6357         (arg_size_in_bytes): New function.
6358         (get_ultimate_context): New function.
6359         * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6360         TRANSLATION_UNIT_WARN_EMPTY_P.
6361         (default_is_empty_record, arg_int_size_in_bytes,
6362         arg_size_in_bytes, get_ultimate_context): Declare.
6364 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6366         * config/arm/arm.c (cmse_clear_registers): New function.
6367         (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6368         code by call to cmse_clear_registers.
6369         (cmse_nonsecure_entry_clear_before_return): Likewise.
6371 2017-11-22  Tamar Christina  <tamar.christina@arm.com>
6373         * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6374         (vdot_s32, vdotq_s32): New.
6375         (vdot_lane_u32, vdotq_lane_u32): New.
6376         (vdot_lane_s32, vdotq_lane_s32): New.
6379 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6381         PR middle-end/82547
6382         * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6383         for unsigned values with fewer HWIs than the precision.
6384         (test_overflow): New function.
6385         (wide_int_cc_tests): Call it.
6387 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6389         * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6390         has the same mode class as Pmode.
6392 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6394         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6395         padding_bits_to_clear_ptr.
6396         (cmse_nonsecure_entry_clear_before_return): Likewise.
6398 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6400         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6401         auto_sbitap instead of integer bitfield to control register needing
6402         clearing.
6404 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6406         PR tree-optimization/83044
6407         * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6408         INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6409         that need the upper bound.  Subtract offset from
6410         get_addr_base_and_unit_offset only if positive and subtract it
6411         before division by eltsize rather than after it.
6413         PR debug/83084
6414         * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6415         debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6416         (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6417         asm.
6419         PR middle-end/82875
6420         * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6421         expand_binop with *mul_widen_optab, make sure at least one of the
6422         operands doesn't have VOIDmode.
6424         PR debug/83034
6425         * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6427         PR rtl-optimization/82044
6428         PR tree-optimization/82042
6429         * dse.c (record_store): Check for overflow.
6430         (check_mem_read_rtx): Properly check for overflow if width == -1, call
6431         add_wild_read instead of clear_rhs_from_active_local_stores on
6432         overflow and log it into dump_file.
6434 2017-11-22  Richard Biener  <rguenther@suse.de>
6436         * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6437         fake edges to exit when looking for a place to insert.
6438         * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6439         and remove.
6440         (insert_into_preds_of_block): Commit edge insertion immediately,
6441         assert that doesn't require new BBs.
6442         (fini_pre): Release expressions.
6443         (pass_pre::execute): Shuffle things around a bit, if the fn
6444         is too large do not compute AVAIL either as this is really the
6445         quadratic bit.
6447 2017-11-22  Richard Biener  <rguenther@suse.de>
6449         PR tree-optimization/83089
6450         * tree-if-conv.c (pass_if_conversion::execute): If anything
6451         changed reset SCEV and free the number of iteration estimates.
6453 2017-11-21  Martin Sebor  <msebor@redhat.com>
6455         PR tree-optimization/82945
6456         * calls.h (warn_nonstring_bound): Remove unused function.
6458 2017-11-21  Martin Sebor  <msebor@redhat.com>
6460         PR tree-optimization/82945
6461         * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6462         * calls.h (maybe_warn_nonstring_arg): Declare new function.
6463         * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6464         functions.
6465         (initialize_argument_information): Call maybe_warn_nonstring_arg.
6466         * calls.h (get_attr_nonstring_decl): Declare new function.
6467         * doc/extend.texi (attribute nonstring): Update.
6468         * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6469         get_attr_nonstring_decl and handle it.
6470         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same.  Improve
6471         detection of nul-termination.
6472         (strlen_to_stridx): Change to a pointer.
6473         (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6474         (pass_strlen::execute): Same.
6476 2017-11-21  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6478         * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6479         for the new option -mprefer-vector-width=[none|128|256|512].
6480         * config/i386/i386.c (ix86_target_string): remove old style options
6481         -mprefer-avx256 and make -mprefer-avx128 as alias.
6482         (ix86_option_override_internal):  Apply defaults for the
6483         -mprefer-vector-width=[128|256] option.
6484         * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6485         Implement macros to work with -mprefer-vector-width=.
6486         * config/i386/i386.opt: Implemented option
6487         -mprefer-vector-width=[none|128|256|512].
6488         * doc/invoke.texi: Documentation for
6489         -mprefer-vector-width=[none|128|256|512].
6491 2017-11-21  Pat Haugen  <pthaugen@us.ibm.com>
6493         * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6495 2017-11-21  Jonathan Wakely  <jwakely@redhat.com>
6497         * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6499 2017-11-21  Marc Glisse  <marc.glisse@inria.fr>
6501         * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6502         POINTER_PLUS_EXPR.
6503         * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6504         * expr.c (expand_expr_real_2): Likewise.
6505         * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6506         fold_binary_loc): Likewise.
6507         * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6508         P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6509         (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6510         MINUS_EXPR transformations.
6511         * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6512         * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6513         * tree-inline.c (estimate_operator_cost): Likewise.
6514         * tree-pretty-print.c (dump_generic_node, op_code_prio,
6515         op_symbol_code): Likewise.
6516         * tree-vect-stmts.c (vectorizable_operation): Likewise.
6517         * vr-values.c (extract_range_from_binary_expr): Likewise.
6518         * varasm.c (initializer_constant_valid_p_1): Likewise.
6519         * tree.def: New tree code POINTER_DIFF_EXPR.
6521 2017-11-21  Uros Bizjak  <ubizjak@gmail.com>
6523         * config/i386/i386.md (*bswap<mode>2_movbe): Add
6524         integer suffix to movbe mnemonic.
6525         (*bswaphi2_movbe): Ditto.
6526         (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6528 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6530         PR c++/83045
6531         * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6532         Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6533         Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6535 2017-11-21  Martin Liska  <mliska@suse.cz>
6537         * tree-inline.c (expand_call_inline): Remove not needed
6538         xstrdup_for_dump.
6540 2017-11-21  James Cowgill  <James.Cowgill@imgtec.com>
6541             Jakub Jelinek  <jakub@redhat.com>
6543         PR target/82880
6544         * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6545         Remove static keyword from f variable.
6547 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6549         PR tree-optimization/83086
6550         * gimple-ssa-store-merging.c
6551         (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6552         rather than n.base_addr.
6554 2017-11-21  Martin Liska  <mliska@suse.cz>
6556         PR rtl-optimization/82044
6557         PR tree-optimization/82042
6558         * dse.c (check_mem_read_rtx): Check for overflow.
6560 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6562         * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6563         typo in comment.
6565 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6567         * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6568         and its corresponding call arg location note.
6570 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6571             Andrew Burgess  <andrew.burgess@embecosm.com>
6573         * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6574         declaration.
6575         (arc_return_slot_offset): Likewise.
6576         (arc_eh_return_address_location): New declaration.
6577         * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6578         (MUST_SAVE_REGISTER): Add exception handler case.
6579         (MUST_SAVE_RETURN_ADDR): Likewise.
6580         (arc_frame_pointer_required): Likewise.
6581         (arc_frame_pointer_needed): New function.
6582         (arc_compute_frame_size): Changed.
6583         (arc_expand_prologue): Likewise.
6584         (arc_expand_epilogue): Likewise.
6585         (arc_initial_elimination_offset): Likewise.
6586         (arc_return_slot_offset): Delete.
6587         (arc_eh_return_address_location): New function.
6588         (arc_builtin_setjmp_frame_value): Likewise.
6589         * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6590         (EH_RETURN_STACKADJ_RTX): Define.
6591         (EH_RETURN_HANDLER_RTX): Likewise.
6592         * config/arc/arc.md (eh_return): Delete.
6594 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6596         * print-tree.h (debug_vec_tree): Remove prototype.
6597         * gdbinit.in (pvt): Remove macro.
6599 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6601         PR tree-optimization/83047
6602         * gimple-ssa-store-merging.c
6603         (imm_store_chain_info::output_merged_store): If the loads with the
6604         same vuse are in different basic blocks, for load_gsi pick a load
6605         location that is dominated by the other loads.
6607         PR c++/83059
6608         * config/i386/i386.c (ix86_memmodel_check): Start
6609         -Winvalid-memory-model diagnostics with lowercase letter.
6611         PR debug/82718
6612         * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6613         set in_cold_section_p to the partition containing loc_list->first.
6614         When seeing loc_list->last_before_switch node, update secname and
6615         perform range_across_switch second partition handling only after that.
6617         PR debug/82933
6618         * run-rtl-passes.c: Include debug.h.
6619         (run_rtl_passes): Call debug_hooks->assembly_start.
6620         * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6621         multiple times.
6623         PR target/82981
6624         * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6625         OPTAB_DIRECT in calls to expand_simple_binop.
6627 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6629         PR c/81404
6630         * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6632 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6634         PR c++/72786
6635         * spellcheck.h (best_match::blithely_get_best_candidate): New
6636         accessor.
6638 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6640         * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6641         with lower case letter.
6643 2017-11-20  Uros Bizjak  <ubizjak@gmail.com>
6645         * config/i386/i386.md (bswaphi2): New expander.
6646         (*bswaphi2_movbe): New insn pattern.
6647         (bswaphi -> rorhi pepehole2): New peephole pattern.
6649 2017-11-20  Jeff Law  <law@redhat.com>
6651         * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6652         * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6653         * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6654         * gimple-ssa-evrp.c: Remove bits moved into new files.  Include
6655         gimple-ssa-evrp-analyze.h.
6657         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6658         set BB_VISITED here.
6659         (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6661 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6663         * tree-predcom.c: Add general comment on Store-Store chains.
6664         (split_data_refs_to_components): Postpone clearing eliminate_store_p
6665         flag in component.
6666         (get_chain_last_ref_at): Rename into...
6667         (get_chain_last_write_at): ...this.
6668         (get_chain_last_write_before_load): New function.
6669         (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6670         CT_STORE_STORE when write reference is added.
6671         (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6672         (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6673         (initialize_root_vars_store_elim_1): Ditto.
6674         (initialize_root_vars_store_elim_2): Ditto.  Replace rhs once default
6675         definition is created.
6676         (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6677         chain by replacing it with dominant stored value.
6679 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6681         * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6683 2017-11-20  Marc Glisse  <marc.glisse@inria.fr>
6685         * vr-values.c (extract_range_from_binary_expr): Use a full range
6686         for VR_VARYING.
6688 2017-11-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6690         * config/arm/arm.md (R4_REGNUM): Define constant.
6691         (nonsecure_call_internal): Remove r4 clobber.
6692         (nonsecure_call_value_internal): Likewise.
6693         * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6694         clobber and resequence match_operands.
6695         (nonsecure_call_value_reg_thumb1_v5): Likewise.
6696         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6697         (nonsecure_call_value_reg_thumb2): Likewise.
6699 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6701         PR tree-optimization/78821
6702         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6703         if base is TARGET_MEM_REF.  If base is not MEM_REF, set base_addr
6704         to the address of the base rather than the base itself.
6705         (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6706         (find_bswap_or_nop_finalize): New function.
6707         (find_bswap_or_nop): Use it.
6708         (bswap_replace): Return a tree rather than bool, change first
6709         argument from gimple * to gimple_stmt_iterator, allow inserting
6710         into an empty sequence, allow ins_stmt to be NULL - then emit
6711         all stmts into gsi.  Fix up MEM_REF address gimplification.
6712         (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6713         (struct store_immediate_info): Add N and INS_STMT non-static
6714         data members.
6715         (store_immediate_info::store_immediate_info): Initialize them
6716         from newly added ctor args.
6717         (merged_store_group::apply_stores): Formatting fixes.  Sort by
6718         bitpos at the end.
6719         (stmts_may_clobber_ref_p): For stores call also
6720         refs_anti_dependent_p.
6721         (gather_bswap_load_refs): New function.
6722         (imm_store_chain_info::try_coalesce_bswap): New method.
6723         (imm_store_chain_info::coalesce_immediate_stores): Use it.
6724         (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6725         (imm_store_chain_info::output_merged_store): Fail if number of
6726         new estimated stmts is bigger or equal than old.  Handle LROTATE_EXPR
6727         and NOP_EXPR rhs_code.
6728         (pass_store_merging::process_store): Compute n and ins_stmt, if
6729         ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6730         LROTATE_EXPR rhs_code.  Pass n and ins_stmt to store_immediate_info
6731         ctor.
6732         (pass_store_merging::execute): Calculate dominators.
6734         * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6735         BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6736         CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6737         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6738         find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6739         class pass_optimize_bswap, bswap_replace,
6740         pass_optimize_bswap::execute): Moved to ...
6741         * gimple-ssa-store-merging.c: ... this file.
6742         Include optabs-tree.h.
6743         (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6744         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6745         find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6746         anonymous namespace, remove static keywords.
6747         (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6748         (pass_optimize_bswap::execute): ... rather than here.  Formatting fix.
6750 2017-11-20  Jan Hubicka  <hubicka@ucw.cz>
6752         PR bootstrap/83062
6753         * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6755 2017-11-20  Aldy Hernandez  <aldyh@redhat.com>
6757         * vec.h (debug_helper): New function.
6758         (DEFINE_DEBUG_VEC): New macro.
6759         * hash-set.h (debug_helper): New function.
6760         (DEFINE_DEBUG_HASH_SET): New macro.
6761         * cfg.c (debug_slim (edge)): New function.
6762         Call DEFINE_DEBUG_VEC for edges.
6763         Call DEFINE_DEBUG_HASH_SET for edges.
6764         * cfghooks.c (debug_slim (basic_block)): New function.
6765         Call DEFINE_DEBUG_VEC for basic blocks.
6766         Call DEFINE_DEBUG_HASH_SET for basic blocks.
6767         * print-tree.c (debug_slim): New function to handle trees.
6768         Call DEFINE_DEBUG_VEC for trees.
6769         Call DEFINE_DEBUG_HASH_SET for trees.
6770         (debug (vec<tree, va_gc>) &): Remove.
6771         (debug (<vec<tree, va_gc>) *): Remove.
6772         * print-rtl.c (debug_slim): New function to handle const_rtx.
6773         Call DEFINE_DEBUG_VEC for rtx_def.
6774         Call DEFINE_DEBUG_VEC for rtx_insn.
6775         Call DEFINE_DEBUG_HASH_SET for rtx_def.
6776         Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6777         * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6778         (debug (vec<rtx_insn *> *ptr): Remove.
6779         (debug_insn_vector): Remove.
6780         * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6782 2017-11-20  Tom de Vries  <tom@codesourcery.com>
6784         PR rtl-optimization/82020
6785         * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6786         IF_THEN_ELSE condition.
6788 2017-11-19  Jeff Law  <law@redhat.com>
6790         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6791         of degenerates resulting from ignoring an edge.
6793 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6795         PR ipa/81360
6796         * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6798 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6800         PR ipa/83001
6801         * profile-count.c (profile_count::to_sreal_scale): Fix return value
6802         for uninitialied counts.
6804 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6806         PR ipa/60243
6807         * tree-inline.c (estimate_num_insns): Set to 1 at least.
6809 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6811         PR target/82713
6812         * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6814 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6816         * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6817         "do while (0)".
6819 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6821         * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6822         "do {} while (0)".
6824 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6826         * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6827         macro body.
6829 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6831         * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6832         "do {} while (0)".
6833         * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6835 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6837         * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6838         "do {} while (0)".
6839         * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6840         semicolon after MCORE_EXPORT_NAME call.
6842 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6844         PR target/82961
6845         * vmsdbgout.c (vmsdbgout_early_finish): New function.
6846         (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6848 2017-11-18  Jan Hubicka  <hubicka@ucw.cz>
6850         * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6851         (cgraph_edge::clone): Cleanup updating of profile.
6852         * ipa-cp.c (update_profiling_info): Likewise.
6853         * ipa-inline-transform.c (inline_transform): Likewise.
6854         * ipa-inline.c (inline_small_functions): Add missing space to dump.
6855         * ipa-split.c (execute_split_functions): Do not split when function
6856         is cold.
6857         * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6858         * profile-count.c (profile_count::dump): Add global0.
6859         (profile_count::to_cgraph_frequency): Do not ICE when entry is
6860         undefined.
6861         (profile_count::to_sreal_scale): Likewise.
6862         (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6863         (profile_count::combine_with_ipa_count): New function.
6864         * profile-count.h (profile_guessed_global0adjusted): New.
6865         (profile_count::adjusted_zero): New.
6866         (profile_count::global0adjusted): New.
6867         (profile_count::combine_with_ipa_count): New.
6868         * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6869         correct profile of return block of split functions.
6870         (copy_cfg_body): Remove unused profile_count.
6871         (copy_body): Likewise.
6872         (expand_call_inline): Update.
6873         (tree_function_versioning): Update.
6875 2017-11-18  Aldy Hernandez  <aldyh@redhat.com>
6877         * hash-set.h (hash_set::empty): New.
6878         * tree-ssa-threadbackward.h: Delete.
6879         * tree-ssa-threadbackward.c (class thread_jumps): New.
6880         Move max_threaded_paths into class.
6881         (fsm_find_thread_path): Remove arguments that are now in class.
6882         (profitable_jump_thread_path): Rename to...
6883         (thread_jumps::profitable_jump_thread_path): ...this.
6884         (convert_and_register_jump_thread_path): Rename to...
6885         (thread_jumps::convert_and_register_current_path): ...this.
6886         (check_subpath_and_update_thread_path): Rename to...
6887         (thread_jumps::check_subpath_and_update_thread_path): ...this.
6888         (register_jump_thread_path_if_profitable): Rename to...
6889         (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6890         (handle_phi): Rename to...
6891         (thread_jumps::handle_phi): ...this.
6892         (handle_assignment): Rename to...
6893         (thread_jumps::handle_assignment): ...this.
6894         (fsm_find_control_statement_thread_paths): Rename to...
6895         (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6896         (find_jump_threads_backwards): Rename to...
6897         (thread_jumps::find_jump_threads_backwards): ...this.
6898         Initialize path local data.
6899         (pass_thread_jumps::execute): Call find_jump_threads_backwards
6900         from within thread_jumps class.
6901         (pass_early_thread_jumps::execute): Same.
6903 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6905         * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
6907 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6909         * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
6910         to canonical location.
6912 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6914         PR target/81356
6915         * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
6916         Remove.
6917         (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
6919 2017-11-17  H.J. Lu  <hongjiu.lu@intel.com>
6921         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
6922         rcrt1.o%s/grcrt1.o%s for -static-pie.
6924 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6926         * i386.c (ix86_multiplication_cost, ix86_division_cost,
6927         ix86_shift_rotate_cost): Break out from ...
6928         (ix86_rtx_costs): ... here.
6929         (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
6930         vector operations.
6932 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6934         * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
6935         when entry block was promoted unlikely.
6936         (estimate_bb_frequencies): Increase frequency scale.
6937         * profile-count.h (profile_count): Export precision info.
6939 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6941         * tree-tailcall.c (eliminate_tail_call): Be more careful about not
6942         disturbin profile of entry block.
6944 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6946         * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
6947         roundoff errors.
6949 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6951         * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
6952         * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
6954 2017-11-17  Jeff Law  <law@redhat.com>
6956         * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
6957         from evrp_dom_walker class.  Various methods moved into new class.
6958         (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
6959         (evrp_range_analyzer::enter): New method.
6960         (evrp_range_analyzer::leave): New method.
6961         (evrp_dom_walker): Remove delegators no longer needed by this class.
6962         Replace vr_values data member with evrp_range_analyzer
6964         * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
6965         method extracted from evrp_dom_walker::before_dom_children.
6966         (evrp_dom_walker::record_ranges_from_stmt): Likewise.
6967         (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
6969         * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
6970         Add private copy constructor and move assignment operators.
6971         Privatize methods and class data where trivially possible.
6972         (evrp_dom_walker::cleanup): New function, extracted from
6973         execute_early_vrp.  Simplify access to class data.
6975         * vr-values.h (get_output_for_vrp): Prototype.
6976         * vr-values.c (get_output_for_vrp): New function extracted from
6977         vrp_visit_assignment_or_call and extract_range_from_stmt.
6978         (vrp_visit_assignment_or_call): Use get_output_for_vrp.  Simplify.
6980 2017-11-17  Luis Machado  <luis.machado@linaro.org>
6982         * config/aarch64/aarch64.c
6983         (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
6984         (qdf24xx_tunings) <autoprefetcher_model>: Set to
6985         tune_params::AUTOPREFETCHER_WEAK.
6987 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6989         PR target/82641
6990         * config/arm/arm.c (arm_valid_target_attribute_rec):
6991         Parse "arch=" and "+<ext>".
6992         (arm_valid_target_attribute_tree): Re-init global options.
6993         (arm_option_override): Make non-static.
6994         (arm_options_perform_arch_sanity_checks): Make errors fatal.
6995         * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
6996         (__ARM_FEATURE_CRC32): Support undef.
6997         * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
6998         * doc/extend.texi (ARM Function Attributes): Add pragma and target.
7000 2017-11-17  David Malcolm  <dmalcolm@redhat.com>
7002         * gdbinit.in (break-on-diagnostic): New command.
7004 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7006         * config/i386/i386.c (ix86_expand_epilogue): Change simple
7007         return to indirect jump for EH return if control-flow
7008         protection is enabled. Change explicit 'false' argument in
7009         pro_epilogue_adjust_stack with a value of
7010         flag_cf_protection.
7011         * config/i386/i386.md (simple_return_indirect_internal):
7012         Remove SImode restriction to support 64-bit.
7014 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
7016         * combine.c (added_notes_insn): New.
7017         (try_combine): Handle added_notes_insn like added_links_insn.
7018         Rewrite return value code.
7019         (distribute_notes): Set added_notes_insn to the earliest insn we added
7020         a note to.
7022 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
7024         PR rtl-optimization/82621
7025         * combine.c (try_combine): Do not split PARALLELs of two SETs if the
7026         dest of one of those SETs is unused.
7028 2017-11-17  Richard Biener  <rguenther@suse.de>
7030         PR fortran/83017
7031         * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
7032         * tree-pretty-print.c (dump_generic_node): Handle
7033         annot_expr_parallel_kind.
7034         * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
7035         * gimplify.c (gimple_boolify): Likewise.
7037 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
7039         * config.gcc (extra_headers): Add cet.h for x86 targets.
7040         * config/i386/cet.h: New file.
7041         * doc/install.texi: Add --enable-cet/--disable-cet.
7043 2017-11-17  Richard Biener  <rguenther@suse.de>
7045         PR tree-optimization/83017
7046         * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
7047         (gen_parallel_loop): Properly count iterations.
7048         (parallelize_loops): Handle loop->can_be_parallel independent
7049         of flag_loop_parallelize_all.  Make static profitability test match
7050         the runtime one.
7051         * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
7052         * invoke.texi (parloops-min-per-thread): Document.
7054 2017-11-17  Vineet Gupta  <vgupta@synopsys.com>
7056         * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
7057         upstreaming review comments.
7059 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
7061         * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
7062         with fast unaligned access.
7063         * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
7065 2017-11-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
7067         * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
7068         to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
7070 2017-11-17  Richard Biener  <rguenther@suse.de>
7072         * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
7073         folding of references.
7075 2017-11-17  Qing Zhao  <qing.zhao@oracle.com>
7077         PR middle-end/78809
7078         * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
7079         of replacing call to strncmp with corresponding call to strcmp when
7080         meeting conditions.
7082 2017-11-17  Sergey Shalnov  <Sergey.Shalnov@intel.com>
7084         * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
7085         option prefer-avx256 for skylake-avx512 configuration.
7086         * config/i386/i386.c (ix86_option_override_internal): Ditto.
7087         (get_builtin_code_for_version): Ditto.
7089 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7090             Monk Chiang  <sh.chiang04@gmail.com>
7092         * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
7093         (FIXED_REGISTERS): Reserve more register numbers.
7094         (CALL_USED_REGISTERS): Likewise.
7095         (REG_ALLOC_ORDER): Likewise.
7096         (REG_CLASS_CONTENTS): Likewise.
7097         (REGISTER_NAMES): Likewise.
7099 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7100             Kito Cheng  <kito.cheng@gmail.com>
7102         * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
7103         V2SI.
7104         * config/nds32/iterators.md: Add vector mode iterators and attributes.
7106 2017-11-16  Steven Munroe  <munroesj@gcc.gnu.org>
7108         * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
7109         parameter list for vec_splats.
7111 2017-11-16  Joseph Myers  <joseph@codesourcery.com>
7113         * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
7114         date of C17.
7115         (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
7117 2017-11-16  Thomas Koenig  <tkoenig@gcc.gnu.org>
7119         PR bootstrap/82856
7120         * doc/install.texi: Document incompatibility of Perl >=5.6.26
7121         with the required version of automake 1.11.6.
7123 2017-11-16  Pat Haugen  <pthaugen@us.ibm.com>
7125         * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
7126         for it.
7127         (DU_C2_3_power9): Correct reservation combinations.
7128         (FP_DIV_power9, VEC_DIV_power9): New.
7129         (power9-alu): Split out rotate/shift...
7130         (power9-rot): ...to here, correct dispatch resource.
7131         (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
7132         resource.
7133         (power9-fp): Correct latency.
7134         (power9-sdiv): Add div/sqrt resource.
7135         (power9-ddiv): Correct latency, add div/sqrt resource.
7136         (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
7137         (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
7138         resource.
7139         (power9-qpdiv, power9-qpmul): Adjust resource usage.
7141 2017-11-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
7143         * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
7144         switch statement mapping KF built-ins to TF built-ins if we don't
7145         have the proper ISA 3.0 assembler support.
7147 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7149         * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
7150         (lower_emutls_function_body): Do not compute it.
7152 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7154         * ipa-split.c (split_bb_info): Turn time to sreal.
7155         (split_point): Likewise.
7156         (dump_split_point): Likewise.
7157         (fine_split_points): Likewise.
7158         (execute_split_functions): Only zero split_bbs; turn time to sreals.
7160 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7162         * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
7163         in sreal.
7165 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7167         * predict.c (combine_predictions_for_bb): Preserve zero predicted
7168         edges.
7169         (expensive_function_p): Remove useless assert.
7170         (determine_unlikely_bbs): Propagate also forward; determine cold blocks
7172 2017-11-16  Martin Sebor  <msebor@redhat.com>
7174         PR tree-optimization/82588
7175         PR tree-optimization/82583
7176         * tree-vrp.c (check_array_ref): Handle flexible array members,
7177         string literals, and inner indices.
7178         (search_for_addr_array): Add detail to diagnostics.
7180 2017-11-16  Nathan Sidwell  <nathan@acm.org>
7182         PR c++/82836
7183         PR c++/82737
7184         * tree.h (COPY_DECL_RTL): Rename parms for clarity.
7185         (SET_DECL_ASSEMBLER_NAME): Forward to
7186         overwrite_decl_assembler_name.
7187         (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
7188         (overwrite_decl_assembler_name): Declare.
7189         * tree.c (overwrite_decl_assembler_name): New.
7190         * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
7191         (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
7192         (LANG_HOOKS_INITIALIZER): Add it.
7193         * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
7194         * langhooks.c (lhd_set_decl_assembler_name): Use
7195         SET_DECL_ASSEMBLER_NAME.
7196         (lhd_overwrite_decl_assembler_name): Default implementation.
7198 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7199             Jackson Woodruff  <jackson.woodruff@arm.com>
7201         PR tree-optimization/71026
7202         * match.pd: Canonicalize constant multiplies in division.
7204 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7206         * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
7207         * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
7208         Remove OPT_fomit_frame_pointer entry.
7209         * common/config/arc/arc-common.c: Likewise.
7210         * common/config/arm/arm-common.c: Likewise.
7211         * common/config/avr/avr-common.c: Likewise.
7212         * common/config/c6x/c6x-common.c: Likewise.
7213         * common/config/cr16/cr16-common.c: Likewise.
7214         * common/config/cris/cris-common.c: Likewise.
7215         * common/config/epiphany/epiphany-common.c: Likewise.
7216         * common/config/fr30/fr30-common.c: Likewise.
7217         * common/config/frv/frv-common.c: Likewise.
7218         * common/config/ia64/ia64-common.c: Likewise.
7219         * common/config/iq2000/iq2000-common.c: Likewise.
7220         * common/config/lm32/lm32-common.c: Likewise.
7221         * common/config/m32r/m32r-common.c: Likewise.
7222         * common/config/mcore/mcore-common.c: Likewise.
7223         * common/config/microblaze/microblaze-common.c: Likewise.
7224         * common/config/mips/mips-common.c: Likewise.
7225         * common/config/mmix/mmix-common.c: Likewise.
7226         * common/config/mn10300/mn10300-common.c: Likewise.
7227         * common/config/nios2/nios2-common.c: Likewise.
7228         * common/config/pa/pa-common.c: Likewise.
7229         * common/config/pdp11/pdp11-common.c: Likewise.
7230         * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7231         * common/config/riscv/riscv-common.c: Likewise.
7232         * common/config/rs6000/rs6000-common.c: Likewise.
7233         * common/config/rx/rx-common.c: Likewise.
7234         * common/config/s390/s390-common.c: Likewise.
7235         * common/config/sh/sh-common.c: Likewise.
7236         * common/config/sparc/sparc-common.c: Likewise.
7237         * common/config/tilegx/tilegx-common.c: Likewise.
7238         * common/config/tilepro/tilepro-common.c: Likewise.
7239         * common/config/v850/v850-common.c: Likewise.
7240         * common/config/visium/visium-common.c: Likewise.
7241         * common/config/xstormy16/xstormy16-common.c: Likewise.
7242         * common/config/xtensa/xtensa-common.c: Likewise.
7243         * invoke.texi (-fomit-frame-pointer): Update documentation.
7245 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7247         * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7249 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7251         * cfg.c (scale_bbs_frequencies_int,
7252         cale_bbs_frequencies_gcov_type): Remove.
7253         * cfg.h (scale_bbs_frequencies_int,
7254         cale_bbs_frequencies_gcov_type): Remove.
7256 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7258         * tree-ssa-loop-manip.c
7259         (scale_dominated_blocks_in_loop): Update to profile counts.
7260         (tree_transform_and_unroll_loop): Likewise.
7262 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7264         * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7265         scale_bbs_frequencies_int.
7267 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7269         * final.c (compute_alignments): Use counts rather than frequencies.
7271 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7273         * cfgloopanal.c: Include sreal.h
7274         (average_num_loop_insns): Use counts and sreal for accounting.
7276 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7278         * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7279         manipulation.
7281 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7282             Kito Cheng  <kito.cheng@gmail.com>
7284         * config/nds32/constraints.md: Provide more constraints.
7285         * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7286         REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7287         support constraints usage.
7289 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7291         * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7293 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7294             Kito Cheng  <kito.cheng@gmail.com>
7296         * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7297         * config/nds32/nds32.opt: Refine the layout.
7298         * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7299         TARGET_EXT_STRING): Support new options.
7300         * config/nds32/nds32.h: Likewise.
7301         * config/nds32/nds32.md: Likewise.
7302         * config/nds32/nds32-predicates.c: Likewise.
7303         * config/nds32/constraints.md: Likewise.
7304         * common/config/nds32/nds32-common.c: Likewise.
7306 2017-11-16  Julia Koval  <julia.koval@intel.com>
7308         PR target/82983
7309         * config/i386/gfniintrin.h: Add sse check.
7310         * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7312 2017-11-16  Julia Koval  <julia.koval@intel.com>
7314         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7315         OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7316         (ix86_handle_option): Handle -mavx512vbmi2.
7317         * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7318         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7319         * config/i386/i386-c.c (__AVX512VBMI2__): New.
7320         * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7321         (ix86_valid_target_attribute_inner_p): Ditto.
7322         * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7323         * config/i386/i386.opt (mavx512vbmi2): New option.
7324         * doc/invoke.texi: Add new option.
7326 2017-11-16  Julia Koval  <julia.koval@intel.com>
7328         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7329         _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7330         _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7331         _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7332         _mm512_gf2p8mul_epi8): New intrinsics.
7333         * config/i386/i386-builtin-types.def
7334         (V64QI_FTYPE_V64QI_V64QI): New type.
7335         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7336         __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7337         __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7338         __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7339         * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7340         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7342 2017-11-15  Uros Bizjak  <ubizjak@gmail.com>
7344         * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7345         explicitly as a stream of bytes.
7347 2017-11-15  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7349         * config/rs6000/altivec.h (vec_xst_be): New #define.
7350         * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7351         and externalize from *altivec_vperm_<mode>_internal.
7352         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7353         instantiation.
7354         (XL_BE_V8HI): Likewise.
7355         (XL_BE_V4SI): Likewise.
7356         (XL_BE_V4SI): Likewise.
7357         (XL_BE_V2DI): Likewise.
7358         (XL_BE_V4SF): Likewise.
7359         (XL_BE_V2DF): Likewise.
7360         (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7361         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7362         all array entries with these keys: VSX_BUILTIN_VEC_XL,
7363         VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST.  Add entries for key
7364         VSX_BUILTIN_VEC_XST_BE.
7365         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7366         (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7367         built-ins.
7368         (altivec_init_builtins): Replace conditional calls to def_builtin
7369         for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7370         __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7371         with unconditional calls.  Remove calls to def_builtin for
7372         __builtin_vsx_le_be_<mode>.  Add a call to def_builtin for
7373         __builtin_vec_xst_be.
7374         * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7375         to define_expand, and add alternate RTL generation for P8.
7376         (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7377         vsx_ld_elemrev_v8hi.
7378         (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7379         add alternate RTL generation for P8.
7380         (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7381         vsx_ld_elemrev_v16qi.
7382         (vsx_st_elemrev_v8hi): Convert define_insn
7383         to define_expand, and add alternate RTL generation for P8.
7384         (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7385         vsx_st_elemrev_v8hi.
7386         (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7387         add alternate RTL generation for P8.
7388         (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7389         vsx_st_elemrev_v16qi.
7391 2017-11-15  H.J. Lu  <hongjiu.lu@intel.com>
7393         PR target/82990
7394         * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7395         TARGET_AVX512ER check.
7396         (ix86_option_override_internal): Set MASK_VZEROUPPER if
7397         neither -mzeroupper nor -mno-zeroupper is used and
7398         TARGET_EMIT_VZEROUPPER is set.
7399         * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7400         * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7402 2017-11-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7404         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7405         folding of vector compares.
7406         (fold_build_vec_cmp): New helper function.
7407         (fold_compare_helper): New helper function.
7408         (builtin_function_type): Add compare builtins to the list of functions
7409         having unsigned arguments.  Cosmetic updates to comment indentation.
7410         * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7411         the not+eq combination.
7413 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7415         PR tree-optimization/82726
7416         PR tree-optimization/70754
7417         * tree-predcom.c (order_drefs_by_pos): New function.
7418         (combine_chains): Move code setting has_max_use_after to...
7419         (try_combine_chains): ...here.  New parameter.  Sort combined chains
7420         according to position information.
7421         (tree_predictive_commoning_loop): Update call to above function.
7422         (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7424 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7426         PR tree-optimization/82726
7427         Revert
7428         2017-01-23  Bin Cheng  <bin.cheng@arm.com>
7430         PR tree-optimization/70754
7431         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7432         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
7433         combined stmt before it if not NULL.
7434         (combine_chains): Process refs reversely and compute dominance point
7435         for root ref.
7437         Revert
7438         2017-02-23  Bin Cheng  <bin.cheng@arm.com>
7440         PR tree-optimization/79663
7441         * tree-predcom.c (combine_chains): Process refs in reverse order
7442         only for ZERO length chains, and add explaining comment.
7444 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7446         * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7447         * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7448         * doc/invoke.texi (ARM Options): Add armv8.3-a.
7450 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7452         * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7454 2017-11-15  Martin Liska  <mliska@suse.cz>
7456         * tree-cfg.c (pass_warn_function_return::execute):
7457         Compare warn_return_type for greater than zero.
7459 2017-11-15  Sebastian Peryt  <sebastian.peryt@intel.com>
7461         PR target/82941
7462         PR target/82942
7463         * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7464         to return true on Xeon and not on Xeon Phi.
7465         (ix86_check_avx256_register): Changed to ...
7466         (ix86_check_avx_upper_register): ... this. Add extra check for
7467         VALID_AVX512F_REG_OR_XI_MODE.
7468         (ix86_avx_u128_mode_needed): Changed
7469         ix86_check_avx256_register to ix86_check_avx_upper_register.
7470         (ix86_check_avx256_stores): Changed to ...
7471         (ix86_check_avx_upper_stores): ... this. Changed
7472         ix86_check_avx256_register to ix86_check_avx_upper_register.
7473         (ix86_avx_u128_mode_after): Changed
7474         avx_reg256_found to avx_upper_reg_found. Changed
7475         ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7476         (ix86_avx_u128_mode_entry): Changed
7477         ix86_check_avx256_register to ix86_check_avx_upper_register.
7478         (ix86_avx_u128_mode_exit): Ditto.
7479         * config/i386/i386.h: (host_detect_local_cpu): New define.
7481 2017-11-15  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
7483         * config/arm/xgene1.md (xgene1): Split into automatons
7484         xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7485         (xgene1_f_load): Adjust reservations and/or types.
7486         (xgene1_f_store): Likewise.
7487         (xgene1_load_pair): Likewise.
7488         (xgene1_store_pair): Likewise.
7489         (xgene1_fp_load1): Likewise.
7490         (xgene1_load1): Likewise.
7491         (xgene1_store1): Likewise.
7492         (xgene1_move): Likewise.
7493         (xgene1_alu): Likewise.
7494         (xgene1_simd): Likewise.
7495         (xgene1_bfm): Likewise.
7496         (xgene1_neon_load1): Likewise.
7497         (xgene1_neon_store1): Likewise.
7498         (xgene1_neon_logic): Likewise.
7499         (xgene1_neon_st1): Likewise.
7500         (xgene1_neon_ld1r): Likewise.
7501         (xgene1_alu_cond): Added.
7502         (xgene1_shift_reg): Likwise.
7503         (xgene1_bfx): Likewise.
7504         (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7506 2017-11-15  Jakub Jelinek  <jakub@redhat.com>
7508         PR target/82981
7509         * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7510         ssa-iterators.h.
7511         (can_widen_mult_without_libcall): New function.
7512         (expand_mul_overflow): If only checking unsigned mul overflow,
7513         not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7514         Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7515         no other way works.  Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7516         (expand_DIVMOD): Formatting fix.
7517         * expmed.h (expand_mult): Add NO_LIBCALL argument.
7518         * expmed.c (expand_mult): Likewise.  Use OPTAB_WIDEN rather
7519         than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7521         PR tree-optimization/82977
7522         * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7523         constructed temporary to strlen_to_stridx.put.
7525 2017-11-15  Martin Liska  <mliska@suse.cz>
7527         * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7528         * configure: Regenerate.
7530 2017-11-15  Martin Liska  <mliska@suse.cz>
7532         PR target/82927
7533         * config/sh/sh-mem.cc: Use proper probability for
7534         REG_BR_PROB_NOTE.
7536 2017-11-14  Jeff Law  <law@redhat.com>
7538         * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7539         the red zone for stack_clash_protection_final_dynamic_probe targets
7540         when the total dynamic stack size is zero bytes.
7542         * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7543         blocks is post order.
7545 2017-11-15  Alexandre Oliva  <aoliva@redhat.com>
7547         * dumpfile.h (TDF_COMPARE_DEBUG): New.
7548         * final.c (rest_of_clean_state): Set it for the
7549         -fcompare-debug dump.
7550         * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7551         class when TDF_COMPARE_DEBUG is set.
7553         * dwarf2out.c (gen_producer_string): Discard
7554         OPT_fcompare_debug.
7556 2017-11-15  Joseph Myers  <joseph@codesourcery.com>
7558         PR c/81156
7559         * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7560         * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7561         (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7562         (__TGMATH_REAL_2_3): Remove macros.
7563         (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7564         (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7565         __builtin_tgmath.
7566         (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7567         __TGMATH_REAL_2.
7568         (remquo): Define using __TGMATH_REAL_3.
7570 2017-11-14  Jeff Law  <law@redhat.com>
7572         * vr-values.c: New file with contents extracted from tree-vrp.c.
7573         * Makefile.in (OBJS): Add vr-values.o
7574         * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7575         (set_value_range, set_and_canonicalize_value_range): Likewise.
7576         (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7577         (value_range_constant_singleton, symbolic_range_p): Likewise.
7578         (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7579         (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7580         (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7581         (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7582         (find_case_label_range, find_case_label_index): Likewise.
7583         (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7584         (range_int_cst_singleton_p, value_inside_range): Likewise.
7585         (get_single_symbol): Likewise.
7586         (switch_update): Move structure definition here.
7587         (to_remove_edges, to_update_switch_stmts): Provide externs.
7588         * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7589         (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7590         (vrp_val_is_min, set_value_range): Likewise.
7591         (set_and_canonicalize_value_range, copy_value_range): Likewise.
7592         (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7593         (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7594         (range_is_nonnull, range_int_cst_p): Likewwise.
7595         (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7596         (get_single_symbol, operand_less_p): Likewise
7597         (compare_values_warnv, compare_values): Likewise.
7598         (value_inside_range, value_range_constant_singleton): Likewise.
7599         (zero_nonzero_bitgs_from_vr): Likewise.
7600         (extract_range_from_binary_expr_1): Likewise.
7601         (overflow_comparison_p): Likewise.
7602         (to_remove_edges, to_update_switch_stmts): Likewise.
7603         (find_case_label-index, find_case_label_range): Likewise.
7604         (switch_update, set_value_range_to_nonnegative): Remove.
7605         (set_value_range_to_truthvalue): Likewise.
7606         (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7607         (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7608         (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7609         (find_case_label_ranges, test_for_singularity): Likewise.
7610         (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7611         (x_vr_values): Move to its remaining use site.
7613 2017-11-10  Jeff Law  <law@redhat.com>
7615         * vr-values.h (VR_INITIALIZER): Move #define here.
7616         * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7617         * Makefile.in (OBJS): Add tree-evrp.o
7618         * tree-vrp.h (assert_info): Move structure definition here.
7619         (set_value_range_to_varying): Prototype.
7620         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7621         (infer_value_range, register_edge_assert_for): Likewise.
7622         (stmt_interesting_for_vrp): Likewise.
7623         * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7624         (set_value_range_to_varying): No longer static.
7625         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7626         (infer_value_range, register_edge_assert_for): Likewise.
7628 2017-11-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
7630         * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7631         generate the XXBRD instruction.
7633         * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7634         (rs6000_builtin_type_compatible): Treat _Float128 and long double
7635         as being compatible if -mabi=ieeelongdouble.
7636         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7637         to setup float128 built-ins with hardware support.
7638         (BU_FLOAT128_HW_2): Likewise.
7639         (BU_FLOAT128_HW_3): Likewise.
7640         (BU_FLOAT128_HW_VSX_1): Likewise.
7641         (BU_FLOAT128_HW_VSX_2): Likewise.
7642         (scalar_extract_expq): Change float128 built-in functions to
7643         accommodate having both KFmode and TFmode functions.  Use the
7644         KFmode variant as the default.
7645         (scalar_extract_sigq): Likewise.
7646         (scalar_test_neg_qp): Likewise.
7647         (scalar_insert_exp_q): Likewise.
7648         (scalar_insert_exp_qp): Likewise.
7649         (scalar_test_data_class_qp): Likewise.
7650         (sqrtf128_round_to_odd): Delete processing the round to odd
7651         built-in functions as special built-in functions, and define them
7652         as float128 built-ins.  Use the KFmode variant as the default.
7653         (truncf128_round_to_odd): Likewise.
7654         (addf128_round_to_odd): Likewise.
7655         (subf128_round_to_odd): Likewise.
7656         (mulf128_round_to_odd): Likewise.
7657         (divf128_round_to_odd): Likewise.
7658         (fmaf128_round_to_odd): Likewise.
7659         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7660         support for KFmode and TFmode xststdcqp calls.
7661         (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7662         point, switch the built-in handlers for the get/set float128
7663         exponent, get float128 mantissa, float128 test built-ins, and the
7664         float128 round to odd built-in functions.  Eliminate creating the
7665         float128 round to odd built-in functions as special built-ins.
7666         (rs6000_init_builtins): Eliminate special creation of the float128
7667         round to odd built-in functions.
7668         * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7669         function insns to support both TFmode and KFmode variants.
7670         (xsxsigqp_<mode>): Likewise.
7671         (xsiexpqpf_<mode>): Likewise.
7672         (xsiexpqp_<mode>): Likewise.
7673         (xststdcqp_<mode>): Likewise.
7674         (xststdcnegqp_<mode>): Likewise.
7675         (xststdcqp_<mode>): Likewise.
7677 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7679         * tree-ssa-threadupdate.c (compute_path_counts): Remove
7680         unused path_in_freq_ptr parameter.
7681         (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7683 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7685         * ipa-inline.c (edge_badness): Dump sreal frequency.
7686         (compute_inlined_call_time): Match natural implementaiton ...
7687         * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7688         forgotten division by CGRAPH_FREQ_BASE.
7690 2017-11-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7692         * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7693         Solaris 11.  Update comment.
7694         * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7695         renaming.
7696         * config/sol2.h (STARTFILE_SPEC): Likewise.
7697         * configure: Regenerate.
7699 2017-11-14  Carl Love  <cel@us.ibm.com>
7701         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7702         le_ and be_ prefixes to swap* variables.  Remove
7703         if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7704         statements.
7706 2017-11-14  Jason Merrill  <jason@redhat.com>
7708         Support GTY((cache)) on hash_map.
7709         * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7710         (ggc_cache_remove): Override it instead of ggc_mx.
7711         * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7712         (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7713         * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7714         (simple_cache_map_traits): Override maybe_mx.
7715         * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7716         (hash_map): Friend gt_cleare_cache.
7717         (gt_cleare_cache): New.
7718         * tree.h (tree_cache_traits): New hash_map traits class.
7719         (tree_cache_map): New typedef.
7721 2017-11-14  Richard Biener  <rguenther@suse.de>
7723         * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7724         paramter and handling.
7725         (cleanup_control_flow_bb): Likewise.
7726         (cleanup_control_flow_pre): New helper performing a DFS walk
7727         to call cleanup_control_flow_bb in PRE order.
7728         (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7729         via cleanup_control_flow_pre.
7731 2017-11-14  James Greenhalgh  <james.greenhalgh@arm.com>
7733         * config/aarch64/aarch64-simd.md
7734         (aarch64_simd_bsl<mode>_internal): Remove DImode.
7735         (*aarch64_simd_bsl<mode>_alt): Likewise.
7736         (aarch64_simd_bsldi_internal): New.
7737         (aarch64_simd_bsldi_alt): Likewise.
7739 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7741         * tracer.c (better_p): Do not compare frequencies.
7742         * reg-stack.c (better_edge): Likewise.
7743         * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7744         and back.
7746 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7748         * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7749         * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7750         * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7751         * lto-streamer-in.c (input_function): Use update_max_bb_count.
7752         * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7753         * predict.c (maybe_hot_frequency_p): Inline to ...
7754         (maybe_hot_count_p): ... here; rewrite to counts.
7755         (counts_to_freqs): Rename to ...
7756         (update_max_bb_count): ... this one.
7757         (expensive_function_p): Use counts.
7758         (estimate_bb_frequencies): Update.
7759         (rebuild_frequencies): Update.
7760         * predict.h (counts_to_freqs): Rename to ...
7761         (update_max_bb_count): ... this one.
7762         * profile.c (compute_branch_probabilities): Add debug info
7763         * tree-inline.c (expand_call_inline): Update debug info.
7764         (optimize_inline_calls): Use update_max_bb_count..
7765         (tree_function_versioning): Use update_max_bb_count..
7766         * value-prof.c (gimple_value_profile_transformations):
7767         Do not use update_max_bb_count.
7769 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7771         * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7772         always use frequencies.
7774 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7776         * bb-reorder.c: Remove frequencies from comments.
7777         (better_edge_p): Use profile counts.
7778         (find_traces): Dump profile counts.
7779         (rotate_loop): Use profile counts.
7780         (find_traces_1_round): Likewise.
7781         (connect_better_edge_p): Use counts instead of probabilities for
7782         reverse walk.
7783         (copy_bb_p): Drop early check for non-0 frequency.
7784         (sanitize_hot_paths): Update comments.
7786 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7788         * ipa-split.c (struct split_point): Add count.
7789         (consider_split): Do not compute incoming frequency; compute incoming
7790         count and store it to split_point.
7791         (split_function): Set count of the call to split part correctly.
7793 2017-11-13  Carl Love  <cel@us.ibm.com>
7795         * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7797 2017-11-13  Tom Tromey  <tom@tromey.com>
7799         * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7801 2017-11-13  Carl Love  <cel@us.ibm.com>
7803         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7804         Add support for builtins:
7805         unsigned int vec_first_{,miss}_match_{,or_eos}index,
7806         vector {un,}signed {char,int,short},
7807         vector {un,}signed {char,int,short}) arguments.
7808         * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7809         VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7810         Add BU_P9V_AV_2 expansions for the builtins.
7811         * config/rs6000/altivec.h (vec_first_match_index,
7812         vec_first_mismatch_index, vec_first_match_or_eos_index,
7813         vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7814         * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7815         new extern declaration.
7816         * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7817         * config/rs6000/vsx.md (first_match_index_<mode>,
7818         first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7819         first_mismatch_or_eos_index_<mode>): Add define expand.
7820         (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7821         * doc/extend.texi: Update the built-in documenation file for the new
7822         built-in functions.
7824 2017-11-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
7826         * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7827         into the min/max operations for _Float<N> and _Float<N>X types.
7829 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7831         PR lto/81351
7832         * dwarf2out.c (do_eh_frame): New static variable.
7833         (dwarf2out_begin_prologue): Set it.
7834         (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7836 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7838         * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7840         * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7842 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7844         PR lto/81351
7845         * debug.h (dwarf2out_do_eh_frame): Declare.
7846         * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7847         (dwarf2out_do_frame): Use it.
7848         (dwarf2out_do_cfi_asm): Likewise.
7849         * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7850         (dwarf2out_assembly_start): Likewise.
7851         (dwarf2out_begin_prologue): Fix comment.
7852         * toplev.c (compile_file): Always call dwarf2out_frame_finish
7853         if the target needs either debug or unwind DWARF2 info.
7854         * lto-opts.c (lto_write_options): Do not save -fexceptions,
7855         -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7856         -ftrapping-math, -ftrapv and -fwrapv.
7858 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7860         * cgraph.c (cgraph_edge::sreal_frequency): New function.
7861         * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7862         * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7863         (estimate_edge_size_and_time): Likewise.
7864         (ipa_merge_fn_summary_after_inlining): Likewise.
7865         * ipa-inline.c (cgraph_freq_base_rec): Remove.
7866         (compute_uninlined_call_time): Use sreal_frequency.
7867         (compute_inlined_call_time): Likewise.
7868         (ipa_inline): Do not initialize cgraph_freq_base_rec.
7869         * profile-count.c: Include sreal.h.
7870         (profile_count::to_sreal_scale): New.
7871         * profile-count.h: Forward declare sreal.
7872         (profile_count::to_sreal_scale): Declare.
7874 2017-11-13  Nathan Sidwell  <nathan@acm.org>
7876         * diagnostic.c (maybe_line_and_column): New.
7877         (diagnostic_get_location_text): Use it.
7878         (diagnostic_report_current_module): Likewise.
7879         (test_diagnostic_get_location_text): Add tests.
7881 2017-11-13  Luis Machado  <luis.machado@linaro.org>
7883         * doc/md.texi (Specifying processor pipeline description): Fix
7884         incorrect latency for the div instruction example.
7886 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7888         PR tree-optimization/78821
7889         * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7890         that bit_not_p is the same.
7891         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7892         (split_group): Count precisely bit_not_p bits in each statement.
7893         (invert_op): New function.
7894         (imm_store_chain_info::output_merged_store): Use invert_op to
7895         emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7896         but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
7898 2017-11-13  Martin Liska  <mliska@suse.cz>
7900         * gcov.c (struct coverage_info): Remove typedef of coverage_t.
7901         (struct source_info): Likewise.
7902         (add_branch_counts): Likewise.
7903         (add_line_counts): Likewise.
7904         (function_summary): Likewise.
7905         (output_intermediate_line): Likewise.
7906         (generate_results): Likewise.
7908 2017-11-13  Martin Liska  <mliska@suse.cz>
7910         * gcov.c (struct block_info): Remove typedef for block_t.
7911         (struct line_info): Likewise.
7912         (line_info::has_block): Likewise.
7913         (EXIT_BLOCK): Likewise.
7914         (unblock): Likewise.
7915         (circuit): Likewise.
7916         (get_cycles_count): Likewise.
7917         (process_file): Likewise.
7918         (read_graph_file): Likewise.
7919         (solve_flow_graph): Likewise.
7920         (find_exception_blocks): Likewise.
7921         (add_line_counts): Likewise.
7922         (accumulate_line_info): Likewise.
7923         (output_line_details): Likewise.
7925 2017-11-13  Martin Liska  <mliska@suse.cz>
7927         * gcov.c (struct arc_info): Remove typedef for arc_t.
7928         (struct line_info): Likewise.
7929         (add_branch_counts): Likewise.
7930         (output_branch_count): Likewise.
7931         (function_info::~function_info): Likewise.
7932         (circuit): Likewise.
7933         (output_intermediate_line): Likewise.
7934         (read_graph_file): Likewise.
7935         (solve_flow_graph): Likewise.
7936         (find_exception_blocks): Likewise.
7937         (add_line_counts): Likewise.
7938         (accumulate_line_info): Likewise.
7939         (output_line_details): Likewise.
7940         (output_function_details): Likewise.
7942 2017-11-13  Martin Liska  <mliska@suse.cz>
7944         * gcov.c (struct function_info): Remove typedef for function_t.
7945         (struct source_info): Likewise.
7946         (source_info::get_functions_at_location): Likewise.
7947         (solve_flow_graph): Likewise.
7948         (find_exception_blocks): Likewise.
7949         (add_line_counts): Likewise.
7950         (output_intermediate_file): Likewise.
7951         (process_file): Likewise.
7952         (generate_results): Likewise.
7953         (release_structures): Likewise.
7954         (read_graph_file): Likewise.
7955         (read_count_file): Likewise.
7956         (accumulate_line_counts): Likewise.
7957         (output_lines): Likewise.
7959 2017-11-13  Martin Liska  <mliska@suse.cz>
7961         * gcov.c (function_info::function_info): Remove num_counts
7962         and add vector<gcov_type>.
7963         (function_info::~function_info): Use the vector.
7964         (process_file): Likewise.
7965         (read_graph_file): Likewise.
7966         (read_count_file): Likewise.
7967         (solve_flow_graph): Likewise.
7969 2017-11-13  Martin Liska  <mliska@suse.cz>
7971         * gcov.c (function_info::is_artificial): New function.
7972         (process_file): Erase all artificial early.
7973         (generate_results): Skip as all artificial are already
7974         removed.
7976 2017-11-13  Martin Liska  <mliska@suse.cz>
7978         * gcov.c (read_graph_file): Store to global vector of functions.
7979         (read_count_file): Iterate the vector.
7980         (process_file): Likewise.
7981         (generate_results): Likewise.
7982         (release_structures): Likewise.
7984 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7986         PR tree-optimization/82954
7987         * gimple-ssa-store-merging.c
7988         (imm_store_chain_info::coalesce_immediate_stores): If
7989         !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
7991 2017-11-13  Richard Sandiford  <richard.sandiford@linaro.org>
7993         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
7994         Upddate call to ENDIAN_LANE_N.
7995         (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
7996         (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
7997         (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
7998         and use aarch64_endian_lane_rtx.
7999         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
8001 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8003         * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
8005 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8007         * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
8009 2017-11-12  Tom de Vries  <tom@codesourcery.com>
8011         * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
8012         body.
8013         * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
8014         * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
8015         * defaults.h (ASM_OUTPUT_LABELREF): Same.
8017 2017-11-11  Martin Sebor  <msebor@redhat.com>
8019         PR c/81117
8020         * doc/extend.texi (attribute nonstring): Remove spurious argument.
8022         PR bootstrap/82948
8023         * prefic.c (translate_name): Replace strncpy with memcpy to
8024         avoid -Wstringop-truncation.
8026 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8028         * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
8030 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8032         * predict.c (maybe_hot_frequency_p): Do not use cfun.
8034 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8036         * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
8037         merging.
8039 2017-11-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
8041         * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
8042         enable generating XXBRH if the value is in a vector register.
8043         (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
8044         value is in a vector register.
8045         (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
8046         register to register bswap64's instead of doing the GPR sequence
8047         used on previous machines.
8048         (bswapdi2_xxbrd): New insn.
8049         (bswapdi2_reg): Disallow on ISA 3.0.
8050         (register to register bswap64 splitter): Do not split the insn on
8051         ISA 3.0 systems that use XXBRD.
8053 2017-11-10  Martin Sebor  <msebor@redhat.com>
8055         PR c/81117
8056         * config/darwin-c.c (framework_construct_pathname): Replace strncpy
8057         with memcpy.
8058         (find_subframework_file): Same.
8060 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
8062         * auto-profile.c (afdo_indirect_call): Drop frequency.
8063         * cgraph.c (symbol_table::create_edge): Drop frequency argument.
8064         (cgraph_node::create_edge): Drop frequency argument.
8065         (cgraph_node::create_indirect_edge): Drop frequency argument.
8066         (cgraph_edge::make_speculative): Drop frequency arguments.
8067         (cgraph_edge::resolve_speculation): Do not update frequencies
8068         (cgraph_edge::dump_edge_flags): Do not dump frequency.
8069         (cgraph_node::dump): Check consistency in IPA mode.
8070         (cgraph_edge::maybe_hot_p): Use IPA counter.
8071         (cgraph_edge::verify_count_and_frequency): Rename to ...
8072         (cgraph_edge::verify_count): ... this one; drop frequency checking.
8073         (cgraph_node::verify_node): Update.
8074         * cgraph.h (struct cgraph_edge): Drop frequency.
8075         (cgraph_edge::frequency): New function.
8076         * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
8077         frequencies.
8078         (cgraph_edge::rebuild_edges): Likewise.
8079         * cgraphclones.c (cgraph_edge::clone): Scale only counts.
8080         (duplicate_thunk_for_node): Do not pass frequency.
8081         (cgraph_node::create_clone): Scale only counts.
8082         (cgraph_node::create_virtual_clone): Do not pass frequency.
8083         (cgraph_node::create_edge_including_clones): Do not pass frequency.
8084         (cgraph_node::create_version_clone): Do not pass frequency.
8085         * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
8086         (cgraph_node::expand_thunk): Do not pass frequency.
8087         (cgraph_node::create_wrapper): Do not pass frequency.
8088         * gimple-iterator.c (update_call_edge_frequencies): Do not pass
8089         frequency.
8090         * gimple-streamer-in.c (input_bb): Scale only IPA counts.
8091         * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
8092         * ipa-cp.c (ipcp_lattice::print): Use frequency function.
8093         (gather_caller_stats): Use frequency function.
8094         (ipcp_cloning_candidate_p): Use frequency function.
8095         (ipcp_propagate_stage): Use frequency function.
8096         (get_info_about_necessary_edges): Use frequency function.
8097         (update_profiling_info): Update only IPA profile.
8098         (update_specialized_profile): Use frequency functoin.
8099         (perhaps_add_new_callers): Update only IPA profile.
8100         * ipa-devirt.c (ipa_devirt): Use IPA profile.
8101         * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
8102         (dump_ipa_call_summary): Use frequency function.
8103         (estimate_edge_size_and_time): Use frequency function.
8104         (ipa_merge_fn_summary_after_inlining): Use frequency function.
8105         * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
8106         * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
8107         (update_noncloned_counts): ... ths one; scale counts only.
8108         (clone_inlined_nodes): Do not scale frequency.
8109         (inline_call): Do not pass frequency.
8110         * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
8111         (compute_inlined_call_time): Use IPA profile.
8112         (want_inline_small_function_p): Use IPA profile.
8113         (want_inline_self_recursive_call_p): Use IPA profile.
8114         (edge_badness): Use IPA profile.
8115         (lookup_recursive_calls): Use IPA profile.
8116         (recursive_inlining): Do not pass frequency.
8117         (resolve_noninline_speculation): Do not update frequency.
8118         (inline_small_functions): Collect max of IPA profile.
8119         (dump_overall_stats): Dump IPA porfile.
8120         (dump_inline_stats): Dump IPA porfile.
8121         (ipa_inline): Collect IPA stats.
8122         * ipa-inline.h (clone_inlined_nodes): Update prototype.
8123         * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
8124         (ipa_propagate_frequency): Use frequency function.
8125         (ipa_profile): Cleanup.
8126         * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
8127         * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
8128         * lto-cgraph.c (lto_output_edge): Do not stream frequency.
8129         (input_node): Do not stream frequency.
8130         (input_edge): Do not stream frequency.
8131         (merge_profile_summaries): Scale only IPA profiles.
8132         * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
8133         * predict.c (drop_profile): Do not recompute frequency.
8134         * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
8135         (ipa_tm_insert_gettmclone_call): Do not pass frequency.
8136         * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
8137         * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
8138         * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
8139         * tree-inline.c (copy_bb): Do not scale frequency.
8140         (expand_call_inline): Do not scale frequency.
8141         (tree_function_versioning): Do not scale frequency.
8142         * ubsan.c (ubsan_create_edge): Do not pass frequency.
8144 2017-11-10  Julia Koval  <julia.koval@intel.com>
8146         * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
8147         (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
8148         (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
8149         (_mm256_maskz_gf2p8affine_epi64_epi8)
8150         (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
8151         (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
8152         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
8153         (__builtin_ia32_vgf2p8affineqb_v32qi)
8154         (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
8155         * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
8157 2017-11-10  Tamar Christina  <tamar.christina@arm.com>
8159         PR target/82641
8160         * config/arm/arm.c
8161         (arm_option_override): Refactor.
8162         (arm_option_reconfigure_globals): New.
8163         (arm_options_perform_arch_sanity_checks): New.
8164         * config/arm/arm-protos.h (arm_option_reconfigure_globals):
8165         New prototype.
8166         (arm_options_perform_arch_sanity_checks): Likewise
8168 2017-11-10  Pat Haugen  <pthaugen@us.ibm.com>
8170         * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
8171         (power9-qpmul): New.
8172         * rs6000/rs6000.md ("type" attr): Add qmul.
8173         (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
8174         *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
8175         *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
8177 2017-11-10  Martin Sebor  <msebor@redhat.com>
8179         PR c/81117
8180         * builtins.c (compute_objsize): Handle arrays that
8181         compute_builtin_object_size likes to fail for.  Make extern.
8182         * builtins.h (compute_objsize): Declare.
8183         (check_strncpy_sizes): New function.
8184         (expand_builtin_strncpy): Call check_strncpy_sizes.
8185         * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
8186         -Wstringop-truncation.
8187         (gimple_fold_builtin_strncat): Same.
8188         * gimple.c (gimple_build_call_from_tree): Set call location.
8189         * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
8190         (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
8191         (handle_builtin_stxncpy, handle_builtin_strncat): Same.
8192         (handle_builtin_strlen): Use strlen_to_stridx.
8193         (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
8194         stpncpy.
8195         Use strlen_to_stridx.
8196         (pass_strlen::execute): Release strlen_to_stridx.
8197         * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
8198         (-Wstringop-truncation): Document new option.
8200 2017-11-10  Martin Liska  <mliska@suse.cz>
8202         PR gcov-profile/82702
8203         * gcov.c (main): Handle intermediate files in a different
8204         way.
8205         (get_gcov_intermediate_filename): New function.
8206         (output_gcov_file): Remove support of intermediate files.
8207         (generate_results): Allocate intermediate file.
8208         (release_structures): Clean-up properly fn_end.
8209         (output_intermediate_file): Start iterating with line 1.
8211 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8213         PR tree-optimization/82929
8214         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8215         ops_swapped_p non-static data member.
8216         (store_immediate_info::store_immediate_info): Clear it.
8217         (imm_store_chain_info::coalesce_immediate_stores): If swapping
8218         ops set ops_swapped_p.
8219         (count_multiple_uses): Handle ops_swapped_p.
8221 2017-11-10  Martin Liska  <mliska@suse.cz>
8223         * coverage.c (coverage_init): Stream information about
8224         support of has_unexecuted_blocks.
8225         * doc/gcov.texi: Document that.
8226         * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8227         * gcov.c (read_graph_file): Likewise.
8228         (output_line_beginning): Fix a small issue with
8229         color output.
8231 2017-11-10  Bin Cheng  <bin.cheng@arm.com>
8233         * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8234         reference of trivial component.
8236 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8238         PR bootstrap/82916
8239         * gimple-ssa-store-merging.c
8240         (pass_store_merging::terminate_all_aliasing_chains): For
8241         gimple_store_p stmts also call refs_output_dependent_p.
8243         PR rtl-optimization/82913
8244         * compare-elim.c (try_merge_compare): Punt if def_insn is not
8245         single set.
8247 2017-11-09  Jeff Law  <law@redhat.com>
8249         * vr-values.h: New file with vr_values class.
8250         * tree-vrp.c: Include vr-values.h
8251         (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8252         data objects into the vr_values class.
8253         (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8254         (get_value_range): Make it a member function within vr_values class.
8255         (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8256         (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8257         (op_with_constant_singleton_value_range): Likewise.
8258         (extract_range_for_var_from_comparison_expr): Likewise.
8259         (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8260         (extract_range_from_binary_expr): Likewise.
8261         (extract_range_from_unary_expr): Likewise.
8262         (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8263         (check_for_binary_op_overflow, extract_range_basic): Likewise.
8264         (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8265         (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8266         (compare_name_with_value, compare_names): Likewise.
8267         (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8268         (vrp_evaluate_conditional_warnv_with_ops): Likewise.  Remove prototype.
8269         (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8270         (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8271         (extract_range_from_phi_node): Likewise.
8272         (simplify_truth_ops_using_ranges): Likewise.
8273         (simplify_div_or_mod_using_ranges): Likewise.
8274         (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8275         (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8276         (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8277         (simplify_float_conversion_using_ranges): Likewise.
8278         (simplify_internal_call_using_ranges): Likewise.
8279         (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8280         (vrp_visit_assignment_or_call): Likewise.  Smuggle class instance
8281         poitner via x_vr_values for calls into gimple folder.
8282         (vrp_initialize_lattice): Make this the vr_values ctor.
8283         (vrp_free_lattice): Make this the vr_values dtor.
8284         (set_vr_value): New function.
8285         (class vrp_prop): Add vr_values data member.  Add various member
8286         functions as well as member functions that delegate to vr_values.
8287         (check_array_ref): Make a member function within vrp_prop class.
8288         (search_for_addr_array, vrp_initialize): Likewise.
8289         (vrp_finalize): Likewise.  Revamp to avoid direct access to
8290         vr_value, values_propagated, etc.
8291         (check_array_bounds): Extract vrp_prop class instance pointer from
8292         walk info structure.  Use it to call member functions.
8293         (check_all_array_refs): Make a member function within vrp_prop class.
8294         Smuggle class instance pointer via walk info structure.
8295         (x_vr_values): New local static.
8296         (vrp_valueize): Use x_vr_values to get class instance.
8297         (vr_valueize_1): Likewise.
8298         (class vrp_folder): Add vr_values data member.  Add various member
8299         functions as well as member functions that delegate to vr_values.
8300         (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8301         (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8302         class instance from vr_values.  Use it to call member functions.
8303         (vrp_dom_walker): Add vr_values data member.
8304         (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8305         instance via x_vr_values.
8306         (identify_jump_threads): Accept vr_values as argument.  Store
8307         it into the walker structure.
8308         (evrp_dom_walker): Add vr_values class data member.  Add various
8309         delegators.
8310         (evrp_dom_walker::try_find_new_range): Use vr_values data
8311         member to access the memory allocator.
8312         (evrp_dom_walker::before_dom_children): Store vr_values class
8313         instance into the vrp_folder class.
8314         (evrp_dom_walker::push_value_range): Rework to avoid direct
8315         access to num_vr_values and vr_value.
8316         (evrp_dom_walker::pop_value_range): Likewise.
8317         (execute_early_vrp): Remove call to vrp_initialize_lattice.
8318         Use vr_values to get to dump_all_value_ranges member function.
8319         Remove call to vrp_free_lattice.  Call vrp_initialize, vrp_finalize,
8320         and simplify_cond_using_ranges_2 via vrp_prop class instance.
8321         Pass vr_values class instance down to identify_jump_threads.
8322         Remove call to vrp_free_lattice.
8323         (debug_all_value_ranges): Remove.
8325         * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8326         (vrp_folder): Likewise.
8328         * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8329         Get it from the existing bitmap instead.
8330         (vrp_intersect_ranges_1): Likewise.
8332 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8334         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8335         bit_not_p field.
8336         (store_immediate_info::store_immediate_info): Add bitnotp argument,
8337         set bit_not_p to it.
8338         (imm_store_chain_info::coalesce_immediate_stores): Break group
8339         if bit_not_p is different.
8340         (count_multiple_uses, split_group,
8341         imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8342         (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8343         (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8344         result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8345         to store_immediate_info ctor.
8347 2017-11-09  Jim Wilson  <jimw@sifive.com>
8349         * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8350         (scan_prog_file): Likewise.
8352 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8354         * bb-reorder.c (max_entry_frequency): Remove.
8355         (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8356         connect_traces, push_to_next_round_p): Remove prototypes.
8357         (find_traces_1_round): Use counts only.
8358         (push_to_next_round_p): Likewise.
8359         (find_traces): Likewise.
8360         (rotate_loop): Likewise.
8361         (find_traces_1_round): Likewise.
8362         (connect_traces): Likewise.
8363         (edge_order): Likewise.
8365 2017-11-09  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8367         * config/arm/arm.c (output_return_instruction): Add comments to
8368         indicate requirement for cmse_nonsecure_entry return to account
8369         for the size of clearing instruction output here.
8370         (thumb_exit): Likewise.
8371         * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8372         return in hardfloat mode.
8374 2017-11-09  Segher Boessenkool  <segher@kernel.crashing.org>
8376         * config/rs6000/rs6000.c (machine_function): Add a bool,
8377         "toc_is_wrapped_separately".
8378         (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8379         if it wasn't explicitly set or unset, we are optimizing for speed, and
8380         doing separate shrink-wrapping.
8381         (rs6000_get_separate_components): Enable the TOC component if
8382         saving the TOC register in the prologue.
8383         (rs6000_components_for_bb): Handle the TOC component.
8384         (rs6000_emit_prologue_components): Store the TOC register where needed.
8385         (rs6000_set_handled_components): Mark TOC as handled, if handled.
8386         (rs6000_emit_prologue): Don't save the TOC if that is already done.
8388 2017-11-09  Martin Jambor  <mjambor@suse.cz>
8390         * ipa-param-manipulation.c: New file.
8391         * ipa-param-manipulation.h: Likewise.
8392         * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8393         (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8394         * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8395         (ipa_parm_adjustment): Likewise.
8396         (ipa_parm_adjustment_vec): Likewise.
8397         (ipa_get_vector_of_formal_parms): Moved declaration to
8398         ipa-param-manipulation.h.
8399         (ipa_get_vector_of_formal_parm_types): Likewise.
8400         (ipa_modify_formal_parameters): Likewise.
8401         (ipa_modify_call_arguments): Likewise.
8402         (ipa_combine_adjustments): Likewise.
8403         (ipa_dump_param_adjustments): Likewise.
8404         (ipa_modify_expr): Likewise.
8405         (ipa_get_adjustment_candidate): Likewise.
8406         * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8407         ipa-param-manipulation.c.
8408         (ipa_get_vector_of_formal_parm_types): Likewise.
8409         (ipa_modify_formal_parameters): Likewise.
8410         (ipa_modify_call_arguments): Likewise.
8411         (ipa_modify_expr): Likewise.
8412         (get_ssa_base_param): Likewise.
8413         (ipa_get_adjustment_candidate): Likewise.
8414         (index_in_adjustments_multiple_times_p): Likewise.
8415         (ipa_combine_adjustments): Likewise.
8416         (ipa_dump_param_adjustments): Likewise.
8417         * tree-sra.c: Also include ipa-param-manipulation.h
8418         * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8419         ipa-param.h.
8421 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8422             Alan Hayward  <alan.hayward@arm.com>
8423             David Sherwood  <david.sherwood@arm.com>
8425         * doc/sourcebuild.texi (vect_masked_store): Document.
8427 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8428             Alan Hayward  <alan.hayward@arm.com>
8429             David Sherwood  <david.sherwood@arm.com>
8431         * doc/sourcebuild.texi (vect_align_stack_vars): Document.
8433 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8434             Alan Hayward  <alan.hayward@arm.com>
8435             David Sherwood  <david.sherwood@arm.com>
8437         * doc/sourcebuild.texi (vect_variable_length): Document.
8439 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8440             Alan Hayward  <alan.hayward@arm.com>
8441             David Sherwood  <david.sherwood@arm.com>
8443         * doc/sourcebuild.texi (vect_unaligned_possible): Document.
8445 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8446             Alan Hayward  <alan.hayward@arm.com>
8447             David Sherwood  <david.sherwood@arm.com>
8449         * doc/sourcebuild.texi (vect_element_align_preferred): Document.
8451 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8452             Alan Hayward  <alan.hayward@arm.com>
8453             David Sherwood  <david.sherwood@arm.com>
8455         * doc/sourcebuild.texi (vect_perm_short, vect_perm_byte): Document
8456         previously undocumented selectors.
8457         (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8459 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8461         * doc/rtl.texi (const_vector): Say that elements can be
8462         const_wide_ints too.
8463         * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8464         * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8465         (gen_vec_duplicate): Use it instead of CONSTANT_P.
8466         * optabs.c (expand_vector_broadcast): Likewise.
8468 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8469             Alan Hayward  <alan.hayward@arm.com>
8470             David Sherwood  <david.sherwood@arm.com>
8472         * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8473         scaled index even if the unscaled address was invalid.
8474         Don't increase the complexity of using a scale in that case.
8476 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8477             Alan Hayward  <alan.hayward@arm.com>
8478             David Sherwood  <david.sherwood@arm.com>
8480         * doc/rtl.texi: Rewrite the subreg rules so that they partition
8481         the inner register into REGMODE_NATURAL_SIZE bytes rather than
8482         UNITS_PER_WORD bytes.
8483         * emit-rtl.c (validate_subreg): Divide subregs into blocks
8484         based on REGMODE_NATURAL_SIZE of the inner mode.
8485         (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8486         !SCALAR_FLOAT_MODE_P cases.  Use REGMODE_NATURAL_SIZE for the latter.
8487         * expmed.c (lowpart_bit_field_p): Divide the value up into
8488         chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8489         * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8490         whether something is likely to occupy more than one register.
8492 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8494         PR ipa/82879
8495         * ipa-inline-transform.c (update_noncloned_frequencies): Use
8496         profile_count::adjust_for_ipa_scaling.
8497         * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8498         * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8499         function.
8500         * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8502 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8504         * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8505         (split_group): Add total_orig and total_new arguments, estimate the
8506         number of statements related to the store group without store merging
8507         and with store merging.
8508         (imm_store_chain_info::output_merged_store): Adjust split_group
8509         callers, punt if estimated number of statements with store merging
8510         is not smaller than estimated number of statements without it.
8511         Formatting fix.
8512         (handled_load): Remove has_single_use checks.
8513         (pass_store_merging::process_store): Likewise.
8515 2017-11-09  Richard Biener  <rguenther@suse.de>
8517         PR tree-optimization/82902
8518         * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8520 2017-11-09  Martin Liska  <mliska@suse.cz>
8522         PR target/82863
8523         * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8524         uninitialized.
8526 2017-11-09  Martin Liska  <mliska@suse.cz>
8528         PR tree-optimization/82669
8529         * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8531 2017-11-09  Martin Liska  <mliska@suse.cz>
8533         PR gcov-profile/48463
8534         * coverage.c (coverage_begin_function): Output also end locus
8535         of a function and information whether the function is
8536         artificial.
8537         * gcov-dump.c (tag_function): Parse and print the information.
8538         * gcov.c (INCLUDE_MAP): Add include.
8539         (INCLUDE_SET): Likewise.
8540         (struct line_info): Move earlier in the source file because
8541         of vector<line_info> in function_info structure.
8542         (line_info::line_info): Likewise.
8543         (line_info::has_block): Likewise.
8544         (struct source_info): Add new member index.
8545         (source_info::get_functions_at_location): New function.
8546         (function_info::group_line_p): New function.
8547         (output_intermediate_line): New function.
8548         (output_intermediate_file): Use the mentioned function.
8549         (struct function_start): New.
8550         (struct function_start_pair_hash): Likewise.
8551         (process_file): Add code that identifies group functions.
8552         Assign lines either to global or function scope.
8553         (generate_results): Skip artificial functions.
8554         (find_source): Assign index for each source file.
8555         (read_graph_file): Read new flag artificial and end_line.
8556         (add_line_counts): Assign it either to global of function scope.
8557         (accumulate_line_counts): Isolate core of the function to
8558         accumulate_line_info and call it for both function and global
8559         scope lines.
8560         (accumulate_line_info): New function.
8561         (output_line_beginning): Fix GNU coding style.
8562         (print_source_line): New function.
8563         (output_line_details): Likewise.
8564         (output_function_details): Likewise.
8565         (output_lines): Iterate both source (global) scope and function
8566         scope.
8567         (struct function_line_start_cmp): New class.
8568         * doc/gcov.texi: Reflect changes in documentation.
8570 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8572         PR debug/82837
8573         * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8574         (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8575         and similarly for not instead of neg.
8577 2017-11-08  Andi Kleen  <ak@linux.intel.com>
8579         * config/i386/i386.opt: Add -mforce-indirect-call.
8580         * config/i386/predicates.md: Check for flag_force_indirect_call.
8581         * doc/invoke.texi: Document -mforce-indirect-call
8583 2017-11-08  Kito Cheng  <kito.cheng@gmail.com>
8585         * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8586         New extern.
8587         (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8588         config/riscv/riscv.c (predict.h): New include.
8589         (riscv_slow_unaligned_access_p): No longer static.
8590         (riscv_block_move_straight): Add require.
8591         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8593 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8595         PR target/82855
8596         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8597         <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8598         nonimmediate_operand predicate for operand 1 instead of
8599         register_operand.
8601 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8603         * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8604         New pattern.
8605         * config/aarch64/constraints.md (Uml): New constraint.
8606         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8607         predicate.
8609 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8611         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8612         of two vec_duplicates into a vec_concat.
8614 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8616         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8617         Simplify vec_merge of vec_duplicate and vec_concat.
8618         * config/aarch64/constraints.md (Utq): New constraint.
8619         * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8620         define_insn.
8622 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8624         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8625         Simplify vec_merge of vec_duplicate and const_vector.
8626         * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8627         New predicate.
8628         * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8629         mode iterator.  Update predicate on operand 1 to
8630         handle non-const_vec constants.  Delete constraints.
8631         (*aarch64_combinez_be<mode>): Likewise for operand 2.
8633 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8635         PR tree-optimization/78821
8636         * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8637         data member.
8638         (store_operand_info::store_operand_info): Initialize it to false.
8639         (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8640         ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8641         store in the group, and if chain_info is non-NULL, to ignore altogether
8642         that chain.
8643         (compatible_load_p): Fail if bit_not_p does not match.
8644         (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8645         (handled_load): Fill in bit_not_p.  Handle BIT_NOT_EXPR.
8646         (pass_store_merging::process_store): Adjust
8647         terminate_all_aliasing_chains calls to pass NULL in all current spots,
8648         call terminate_all_aliasing_chains newly when adding a store into
8649         a chain with non-NULL chain_info.
8651 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8653         * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8655 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8657         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8658         Remove.
8659         (aarch64_layout_frame): Initialise emit_frame_chain.
8660         (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8661         (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8663 2017-11-08  Martin Liska  <mliska@suse.cz>
8665         * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8666         of gimple_call_internal_p.
8668 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8670         * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8672 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8674         * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8675         "do {} while (0)".
8677 2017-11-08  Martin Liska  <mliska@suse.cz>
8679         PR sanitizer/82792
8680         * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8682 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8684         * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8686 2017-11-07  Jakub Jelinek  <jakub@redhat.com>
8688         PR target/82855
8689         * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8690         (*cmp<mode>_ccz_1): New insn with $k alternative.
8692         PR target/82855
8693         * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8694         RTX_COMM_COMPARE as commutative as well.
8695         (ix86_binary_operator_ok): Formatting fix.
8696         * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8697         *<code><mode>3<mask_name><round_saeonly_name>,
8698         *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8699         *<s>mul<mode>3_highpart<mask_name>,
8700         *vec_widen_umult_even_v16si<mask_name>,
8701         *vec_widen_umult_even_v8si<mask_name>,
8702         *vec_widen_umult_even_v4si<mask_name>,
8703         *vec_widen_smult_even_v16si<mask_name>,
8704         *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8705         *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8706         *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8707         *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8708         *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8709         <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8710         *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8711         *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8712         *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8713         !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8714         ix86_binary_operator_ok.  Formatting fixes.
8715         (*<plusminus_insn><mode>3<mask_name><round_name>,
8716         *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8717         fixes.
8719 2017-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
8721         * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8722         ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8723         eq and ne if TARGET_ISEL.
8724         (cmp): New code_iterator.
8725         (UNS, UNSU_, UNSIK): New code_attrs.
8726         (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8727         ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8728         to...
8729         ("eq<mode>3"): ... this.
8730         ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8731         to...
8732         ("ne<mode>3"): ... this.
8734 2017-11-07  Julia Koval  <julia.koval@intel.com>
8736         PR target/82812
8737         * common/config/i386/i386-common.c
8738         (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8739         (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8740         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8741         * config/i386/i386.opt: Ditto.
8742         * config/i386/i386.c (ix86_target_string): Ditto.
8743         (ix86_option_override_internal): Ditto.
8744         (ix86_init_mpx_builtins): Move MPX to args2.
8745         (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8746         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8747         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8748         __builtin_ia32_vgf2p8affineinvqb_v32qi,
8749         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8750         __builtin_ia32_vgf2p8affineinvqb_v16qi,
8751         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8753 2017-11-07  Uros Bizjak  <ubizjak@gmail.com>
8755         PR target/80425
8756         * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8757         and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8758         (zero-extendsidi peephole2): Remove peephole.
8760 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8762         PR c/53037
8763         * stor-layout.c: Include attribs.h.
8764         (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8765         explicit lookup of "aligned" attribute.
8767 2017-11-07  Andrew Waterman  <andrew@sifive.com>
8769         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8770         (riscv_expand_block_move): Likewise.
8771         * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8772         implementation.
8773         (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8774         (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8775         * config/riscv/riscv.c (riscv_block_move_straight): New function.
8776         (riscv_adjust_block_mem): Likewise.
8777         (riscv_block_move_loop): Likewise.
8778         (riscv_expand_block_move): Likewise.
8779         * config/riscv/riscv.md (movmemsi): New pattern.
8781 2017-11-07  Michael Clark  <michaeljclark@mac.com>
8783         * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8784         (MUSL_DYNAMIC_LINKER): Likewise.
8786 2017-11-07  Richard Sandiford  <richard.sandiford@linaro.org>
8788         * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8789         (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8790         nonmmory_operand.
8792 2017-11-07  Richard Biener  <rguenther@suse.de>
8794         * match.pd: Fix build.
8796 2017-11-07  Wilco Dijkstra  <wdijkstr@arm.com>
8797             Jackson Woodruff  <jackson.woodruff@arm.com>
8799         PR tree-optimization/71026
8800         * match.pd: Canonicalize negate in division.
8802 2017-11-07  Sudakshina Das  <sudi.das@arm.com>
8804         PR middle-end/80131
8805         * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8807 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8809         * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8810         (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8812 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8814         * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8815         non-scalar integral types.
8816         * match.pd (negate_expr_p): Handle MINUS_EXPR.
8817         (-(A-B), -(~A)): New transformations.
8819 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8821         * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8822         semicolon after "do {} while (0)".
8823         * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8824         * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8825         * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8826         * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8827         * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8829 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8831         * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8832         after "do {} while (0)".
8833         * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8834         * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8835         * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8836         * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8837         * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8839 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8841         PR other/82784
8842         * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8843         "while {} do (0)".
8844         (arm_rtx_costs_internal): Add missing semicolon after
8845         HANDLE_NARROW_SHIFT_ARITH call.
8847 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8849         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8850         disable isel if it was not set explicitly.
8852 2017-11-06  James Bowman  <james.bowman@ftdichip.com>
8854         * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8855         (add_type_attribute) likewise.
8857 2017-11-06  H.J. Lu  <hongjiu.lu@intel.com>
8859         * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8860         of struct ix86_frame.
8861         (ix86_initial_elimination_offset): Likewise.
8862         (ix86_expand_split_stack_prologue): Likewise.
8864 2017-11-06  Marc Glisse  <marc.glisse@inria.fr>
8866         * tree-vrp.h (enum value_range_type): Update stale comment.
8868 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8869             Alan Hayward  <alan.hayward@arm.com>
8870             David Sherwood  <david.sherwood@arm.com>
8872         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8873         (aarch64_expand_vec_perm_const): Take the number of units too.
8874         * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8875         (aarch64_expand_vec_perm_const): Likewise.
8876         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8877         (vec_perm<mode>): Update accordingly.
8879 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8880             Alan Hayward  <alan.hayward@arm.com>
8881             David Sherwood  <david.sherwood@arm.com>
8883         * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8884         Take the number of units too.
8885         * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8886         (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8887         but check for a vector mode before rather than after the call.
8888         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8889         (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8890         (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8891         (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8892         (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8893         (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8894         (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8895         (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8896         (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8897         (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
8898         (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
8899         (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
8900         (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
8901         (aarch64_sqdmull2_n<mode>): Update accordingly.
8903 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8904             Alan Hayward  <alan.hayward@arm.com>
8905             David Sherwood  <david.sherwood@arm.com>
8907         * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
8908         the number of units too.
8909         * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
8910         * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
8911         (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
8912         (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
8913         (vec_store_lanesxi<mode>): Update accordingly.
8915 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8916             Alan Hayward  <alan.hayward@arm.com>
8917             David Sherwood  <david.sherwood@arm.com>
8919         * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
8920         * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
8921         * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
8922         of units rather than the mode.
8923         * config/aarch64/iterators.md (nunits): New mode attribute.
8924         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
8925         Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
8926         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
8927         (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
8928         (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
8929         (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
8930         (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
8931         (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
8932         (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
8933         (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
8934         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
8935         (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
8936         (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
8937         (reduc_<maxmin_uns>_scal_<mode>): Likewise.
8938         (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
8939         (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
8940         (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
8941         (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
8942         (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
8943         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
8944         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
8945         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
8946         (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
8947         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
8948         (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
8949         (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
8950         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
8951         (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
8952         (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
8953         (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
8954         (aarch64_vec_load_lanesci_lane<mode>): Likewise.
8955         (aarch64_vec_store_lanesci_lane<mode>): Likewise.
8956         (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
8957         (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
8958         (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
8959         (aarch64_simd_vec_setv2di): Likewise.
8961 2017-11-06  Carl Love  <cel@us.ibm.com>
8963         * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
8964         definitions.
8965         (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
8966         * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
8967         to power 8.
8968         * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
8969         extern declaration.
8970         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
8971         * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
8972         Add power 8 macro expansions.
8973         (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
8974         * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
8975         power 8 instructions.  (VSX_XXBR): Add iterator.
8977 2017-11-06  Wilco Dijkstra  <wdijkstr@arm.com>
8979         * config/arm/arm.md (predicable_short_it): Change default to "no",
8980         improve documentation, remove uses that are identical to the default.
8981         (enabled_for_depr_it): Rename to enabled_for_short_it.
8982         * gcc/config/arm/arm-fixed.md (predicable_short_it):
8983         Remove default uses.
8984         * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
8985         * gcc/config/arm/sync.md (predicable_short_it): Likewise.
8986         * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
8987         * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
8989 2017-11-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
8991         PR target/82748
8992         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
8993         float128 helper macros, which are no longer used after deleting
8994         the old 'q' built-in functions, and moving the round to odd
8995         built-in functions to being special built-in functions.
8996         (BU_FLOAT128_2): Likewise.
8997         (BU_FLOAT128_1_HW): Likewise.
8998         (BU_FLOAT128_2_HW): Likewise.
8999         (BU_FLOAT128_3_HW): Likewise.
9000         (FABSQ): Delete old 'q' built-in functions.
9001         (COPYSIGNQ): Likewise.
9002         (SQRTF128_ODD): Move round to odd built-in functions to be
9003         special built-in functions, so that we can handle
9004         -mabi=ieeelongdouble.
9005         (TRUNCF128_ODD): Likewise.
9006         (ADDF128_ODD): Likewise.
9007         (SUBF128_ODD): Likewise.
9008         (MULF128_ODD): Likewise.
9009         (DIVF128_ODD): Likewise.
9010         (FMAF128_ODD): Likewise.
9011         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
9012         built-in names to 'f128'.
9013         * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
9014         old 'q' built-in functions, as the machine independent code for
9015         'f128' built-in functions handles this.
9016         (rs6000_expand_builtin): Add expansion for float128 round to odd
9017         functions, keying off on -mabi=ieeelongdouble of whether to use
9018         the KFmode or TFmode variant.
9019         (rs6000_init_builtins): Initialize the _Float128 round to odd
9020         built-in functions.
9021         * doc/extend.texi (PowerPC Built-in Functions): Document the old
9022         _Float128 'q' built-in functions are now mapped into the new
9023         'f128' built-in functions.
9025 2017-11-06  David Edelsohn  <dje.gcc@gmail.com>
9027         * collect2.c (add_lto_object): Compile for OBJECT_COFF.
9028         (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
9030 2017-11-06  David Malcolm  <dmalcolm@redhat.com>
9032         PR jit/82826
9033         * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
9034         * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
9035         * toplev.c: Include "ipa-fnsummary.h".
9036         (toplev::finalize): Call ipa_fnsummary_c_finalize.
9038 2017-11-06  Jakub Jelinek  <jakub@redhat.com>
9040         PR tree-optimization/82838
9041         * gimple-ssa-store-merging.c
9042         (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
9043         on a separate gimple_seq which is then appended to seq.
9045 2017-11-06  Jeff Law  <law@redhat.com>
9047         PR target/82788
9048         * config/i386/i386.c (PROBE_INTERVAL): Remove.
9049         (get_probe_interval): New functions.
9050         (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
9051         (ix86_adjust_stack_and_probe): Likewise.
9052         (output_adjust_stack_and_probe): Likewise.
9053         (ix86_emit_probe_stack_range): Likewise.
9054         (ix86_expand_prologue): Likewise.
9056 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9058         PR tree-optimization/82816
9059         * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
9060         if the modes of the two types are the same.
9061         (convert_plusminus_to_widen): Likewise.
9063 2017-11-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
9065         * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
9066         p9_vadu<mode>3.
9067         (usadv16qi): New define_expand.
9068         (usadv8hi): New define_expand.
9070 2017-11-06  Jan Hubicka  <hubicka@ucw.cz>
9072         PR bootstrap/82832
9073         * ipa-inline-transform.c (update_noncloned_frequencies): Always
9074         scale.
9075         (inline_transform): Likewise.
9076         * predict.c (counts_to_freqs): Remove useless conditional.
9077         * profile-count.h (profile_count::apply_scale): Move sanity check.
9078         * tree-inline.c (copy_bb): Always scale.
9079         (copy_cfg_body): Likewise.
9081 2017-11-06  Christophe Lyon  <christophe.lyon@linaro.org>
9083         PR target/67591
9084         * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
9085         attribute.
9086         (*cmp_ite0): Add enabled_for_depr_it attribute.
9087         (*cmp_ite1): Likewise.
9089 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
9091         * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
9092         TYPE_MFCRF.
9094 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9096         * tree-vrp.c (vrp_int_const_binop): Return true on success and
9097         return the value by pointer.
9098         (extract_range_from_multiplicative_op_1): Update accordingly.
9099         Return as soon as an operation fails.
9101 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9103         PR other/82784
9104         * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
9105         (DEF_SANITIZER_BUILTIN): ... here.
9106         (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
9107         DEF_SANITIZER_BUILTIN in if stmt.  Add missing semicolon.
9109 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9111         PR other/82784
9112         * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
9113         macro body.
9114         (ASM_OUTPUT_CASE_LABEL): Add semicolon after
9115         ASM_OUTPUT_BEFORE_CASE_LABEL call.
9116         * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
9117         after macro body.
9118         * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9119         * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9120         * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9122 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9124         PR other/82784
9125         * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
9126         "do {} while (0)".
9128 2017-11-04  Michael Clark  <michaeljclark@mac.com>
9130         * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
9131         config/riscv/riscv.md (addsi3): Use 'i' for immediates.
9132         (adddi3): Likewise.
9133         (*addsi3_extended): Likewise.
9134         (*addsi3_extended2): Likewise.
9135         (<optab>si3): Likewise.
9136         (<optab>di3): Likewise.
9137         (<optab><mode>3): Likewise.
9138         (<*optabe>si3_internal): Likewise.
9139         (zero_extendqi<SUPERQI:mode>2): Likewise.
9140         (*add<mode>hi3): Likewise.
9141         (*xor<mode>hi3): Likewise.
9142         (<optab>di3): Likewise.
9143         (*<optab>si3_extend): Likewise.
9144         (*sge<u>_<X:mode><GPR:mode>): Likewise.
9145         (*slt<u>_<X:mode><GPR:mode>): Likewise.
9146         (*sle<u>_<X:mode><GPR:mode>): Likewise.
9148 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9150         * config/riscv/riscv.c (riscv_option_override): Conditionally set
9151         TARGET_STRICT_ALIGN based upon -mtune argument.
9153 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9155         * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
9157 2017-11-04  Daniel Santos  <daniel.santos@pobox.com>
9159         * config/i386/i386.c (choose_basereg): Use optional scratch
9160         register and add assertion.
9161         (x86_emit_outlined_ms2sysv_save): Use scratch register when
9162         needed, and don't allocate stack.
9163         (ix86_expand_prologue): Rearrange where SSE saves/stub call is
9164         emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
9165         (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
9167 2017-11-03  Jeff Law  <law@redhat.com>
9169         * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
9170         (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
9171         to probe at the start of a noreturn function.
9173 2017-11-03  Jakub Jelinek  <jakub@redhat.com>
9175         PR tree-optimization/78821
9176         * gimple-ssa-store-merging.c: Update the file comment.
9177         (MAX_STORE_ALIAS_CHECKS): Define.
9178         (struct store_operand_info): New type.
9179         (store_operand_info::store_operand_info): New constructor.
9180         (struct store_immediate_info): Add rhs_code and ops data members.
9181         (store_immediate_info::store_immediate_info): Add rhscode, op0r
9182         and op1r arguments to the ctor, initialize corresponding data members.
9183         (struct merged_store_group): Add load_align_base and load_align
9184         data members.
9185         (merged_store_group::merged_store_group): Initialize them.
9186         (merged_store_group::do_merge): Update them.
9187         (merged_store_group::apply_stores): Pick the constant for
9188         encode_tree_to_bitpos from one of the two operands, or skip
9189         encode_tree_to_bitpos if neither operand is a constant.
9190         (class pass_store_merging): Add process_store method decl.  Remove
9191         bool argument from terminate_all_aliasing_chains method decl.
9192         (pass_store_merging::terminate_all_aliasing_chains): Remove
9193         var_offset_p argument and corresponding handling.
9194         (stmts_may_clobber_ref_p): New function.
9195         (compatible_load_p): New function.
9196         (imm_store_chain_info::coalesce_immediate_stores): Terminate group
9197         if there is overlap and rhs_code is not INTEGER_CST.  For
9198         non-overlapping stores terminate group if rhs is not mergeable.
9199         (get_alias_type_for_stmts): Change first argument from
9200         auto_vec<gimple *> & to vec<gimple *> &.  Add IS_LOAD, CLIQUEP and
9201         BASEP arguments.  If IS_LOAD is true, look at rhs1 of the stmts
9202         instead of lhs.  Compute *CLIQUEP and *BASEP in addition to the
9203         alias type.
9204         (get_location_for_stmts): Change first argument from
9205         auto_vec<gimple *> & to vec<gimple *> &.
9206         (struct split_store): Remove orig_stmts data member, add orig_stores.
9207         (split_store::split_store): Create orig_stores rather than orig_stmts.
9208         (find_constituent_stmts): Renamed to ...
9209         (find_constituent_stores): ... this.  Change second argument from
9210         vec<gimple *> * to vec<store_immediate_info *> *, push pointers
9211         to info structures rather than the statements.
9212         (split_group): Rename ALLOW_UNALIGNED argument to
9213         ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
9214         it.  Adjust find_constituent_stores caller.
9215         (imm_store_chain_info::output_merged_store): Handle rhs_code other
9216         than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
9217         get_location_for_stmts callers.  Set MR_DEPENDENCE_CLIQUE and
9218         MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
9219         (mem_valid_for_store_merging): New function.
9220         (handled_load): New function.
9221         (pass_store_merging::process_store): New method.
9222         (pass_store_merging::execute): Use process_store method.  Adjust
9223         terminate_all_aliasing_chains caller.
9225 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9227         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9228         Return true for more constants, symbols and label references.
9229         (aarch64_valid_floating_const): Remove unused function.
9231 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9233         PR target/82786
9234         * config/aarch64/aarch64.c (aarch64_layout_frame):
9235         Undo forcing of LR at bottom of frame.
9237 2017-11-03  Jeff Law  <law@redhat.com>
9239         PR target/82823
9240         * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9241         for int_registers_saved.
9243         * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9244         extracted from tree-ssa-dom.c.
9245         * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9246         * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9247         (record_equivalences_from_incoming_edge): Add additional argument
9248         to single_pred_edge_ignoring_loop_edges call.
9249         * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9250         (uncprop_dom_walker::before_dom_children): Add additional argument
9251         to single_pred_edge_ignoring_loop_edges call.
9252         * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9253         single_pred_edge_ignoring_loop_edges rather than open coding.
9254         * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9256 2017-11-03  Marc Glisse  <marc.glisse@inria.fr>
9258         * match.pd (-(-A)): Rewrite.
9260 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9262         * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9263         (rs6000_emit_int_cmove): New declaration.
9264         * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9265         (rs6000_emit_sISEL): Delete.
9266         (rs6000_emit_int_cmove): Make non-static.
9267         * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9268         instead of rs6000_emit_sISEL.
9270 2017-11-03  Jan Hubicka  <hubicka@ucw.cz>
9272         * asan.c (create_cond_insert_point): Maintain profile.
9273         * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9274         merged.
9275         * basic-block.h (struct basic_block_def): Remove frequency.
9276         (EDGE_FREQUENCY): Use to_frequency
9277         * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9278         heuristics.
9279         (find_traces): Update to use to_frequency.
9280         (find_traces_1_round): Likewise; use only IPA counts.
9281         (bb_to_key): Likewise.
9282         (connect_traces): Use IPA counts only.
9283         (copy_bb_p): Update to use to_frequency.
9284         (fix_up_crossing_landing_pad): Likewise.
9285         (sanitize_hot_paths): Likewise.
9286         * bt-load.c (basic_block_freq): Likewise.
9287         * cfg.c (init_flow): Set count_max to uninitialized.
9288         (check_bb_profile): Remove frequencies; check counts.
9289         (dump_bb_info): Do not dump frequencies.
9290         (update_bb_profile_for_threading): Update counts only.
9291         (scale_bbs_frequencies_int): Likewise.
9292         (MAX_SAFE_MULTIPLIER): Remove.
9293         (scale_bbs_frequencies_gcov_type): Update counts only.
9294         (scale_bbs_frequencies_profile_count): Update counts only.
9295         (scale_bbs_frequencies): Update counts only.
9296         * cfg.h (struct control_flow_graph): Add count-max.
9297         (update_bb_profile_for_threading): Update prototype.
9298         * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9299         (find_many_sub_basic_blocks): Likewise.
9300         * cfgcleanup.c (try_forward_edges): Likewise.
9301         (try_crossjump_to_edge): Likewise.
9302         * cfgexpand.c (expand_gimple_cond): Likewise.
9303         (expand_gimple_tailcall): Likewise.
9304         (construct_init_block): Likewise.
9305         (construct_exit_block): Likewise.
9306         * cfghooks.c (verify_flow_info): Check consistency of counts.
9307         (dump_bb_for_graph): Do not dump frequencies.
9308         (split_block_1): Do not update frequencies.
9309         (split_edge): Do not update frequencies.
9310         (make_forwarder_block): Do not update frequencies.
9311         (duplicate_block): Do not update frequencies.
9312         (account_profile_record): Do not update frequencies.
9313         * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9314         for global heuristics.
9315         * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9316         (expected_loop_iterations_unbounded): Use counts only.
9317         * cfgloopmanip.c (scale_loop_profile): Simplify.
9318         (create_empty_loop_on_edge): Simplify
9319         (loopify): Simplify
9320         (duplicate_loop_to_header_edge): Simplify
9321         * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9322         (update_br_prob_note): Take care of removing note when profile
9323         becomes undefined.
9324         (relink_block_chain): Do not dump frequency.
9325         (rtl_account_profile_record): Use to_frequency.
9326         * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9327         (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9328         (cgraph_update_edges_for_call_stmt_node): Likewise.
9329         (cgraph_edge::verify_count_and_frequency): Update.
9330         (cgraph_node::verify_node): Temporarily disable frequency verification.
9331         * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9332         to_cgraph_frequency.
9333         (cgraph_edge::rebuild_edges): Convert to ipa counts.
9334         * cgraphunit.c (init_lowered_empty_function): Do not initialize
9335         frequencies.
9336         (cgraph_node::expand_thunk): Update profile.
9337         * except.c (dw2_build_landing_pads): Do not update frequency.
9338         * final.c (compute_alignments): Use to_frequency.
9339         (dump_basic_block_info): Do not dump frequency.
9340         * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9341         (dump_gimple_bb_header): Do not dump frequency.
9342         * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9343         do update count.
9344         * gimple-streamer-in.c (input_bb): Do not stream frequency.
9345         * gimple-streamer-out.c (output_bb): Do not stream frequency.
9346         * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9347         (init_before_recovery): Do not update frequency.
9348         (sched_create_recovery_edges): Do not update frequency.
9349         * hsa-gen.c (convert_switch_statements): Do not update frequency.
9350         * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9351         (ipa_cp_c_finalize): Set max_count to uninitialized.
9352         * ipa-fnsummary.c (get_minimal_bb): Use counts.
9353         (param_change_prob): Use counts.
9354         * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9355         local profiles.
9356         * ipa-split.c (consider_split): Use to_frequency.
9357         (split_function): Use to_frequency.
9358         * ira-build.c (loop_compare_func): Likewise.
9359         (mark_loops_for_removal): Likewise.
9360         (mark_all_loops_for_removal): Likewise.
9361         * loop-doloop.c (doloop_modify): Do not update frequency.
9362         * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9363         frequency.
9364         * lto-streamer-in.c (input_function): Update count_max.
9365         * omp-expand.c (expand_omp_taskreg): Update count_max.
9366         * omp-simd-clone.c (simd_clone_adjust): Update profile.
9367         * predict.c (maybe_hot_frequency_p): Use to_frequency.
9368         (maybe_hot_count_p): Use ipa counts only.
9369         (maybe_hot_bb_p): Simplify.
9370         (maybe_hot_edge_p): Simplify.
9371         (probably_never_executed): Do not take frequency argument.
9372         (probably_never_executed_bb_p): Do not pass frequency.
9373         (probably_never_executed_edge_p): Likewise.
9374         (combine_predictions_for_bb): Check that profile is nonzero.
9375         (propagate_freq): Do not set frequency.
9376         (drop_profile): Simplify.
9377         (counts_to_freqs): Simplify.
9378         (expensive_function_p): Use to_frequency.
9379         (propagate_unlikely_bbs_forward): Simplify.
9380         (determine_unlikely_bbs): Simplify.
9381         (estimate_bb_frequencies): Add hack to silence graphite issues.
9382         (compute_function_frequency): Use ipa counts.
9383         (pass_profile::execute): Update.
9384         (rebuild_frequencies): Use counts only.
9385         (force_edge_cold): Use counts only.
9386         * profile-count.c (profile_count::dump): Dump new count types.
9387         (profile_count::differs_from_p): Check compatiblity.
9388         (profile_count::to_frequency): New function.
9389         (profile_count::to_cgraph_frequency): New function.
9390         * profile-count.h (struct function): Declare.
9391         (enum profile_quality): Add profile_guessed_local and
9392         profile_guessed_global0.
9393         (class profile_proability): Decrease number of bits to 29;
9394         update from_reg_br_prob_note and to_reg_br_prob_note.
9395         (class profile_count: Update comment; decrease number of bits
9396         to 61. Check compatibility.
9397         (profile_count::compatible_p): New private member function.
9398         (profile_count::ipa_p): New member function.
9399         (profile_count::operator<): Handle global zero correctly.
9400         (profile_count::operator>): Handle global zero correctly.
9401         (profile_count::operator<=): Handle global zero correctly.
9402         (profile_count::operator>=): Handle global zero correctly.
9403         (profile_count::nonzero_p): New member function.
9404         (profile_count::force_nonzero): New member function.
9405         (profile_count::max): New member function.
9406         (profile_count::apply_scale): Handle IPA scalling.
9407         (profile_count::guessed_local): New member function.
9408         (profile_count::global0): New member function.
9409         (profile_count::ipa): New member function.
9410         (profile_count::to_frequency): Declare.
9411         (profile_count::to_cgraph_frequency): Declare.
9412         * profile.c (OVERLAP_BASE): Delete.
9413         (compute_frequency_overlap): Delete.
9414         (compute_branch_probabilities): Do not use compute_frequency_overlap.
9415         * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9416         * sched-ebb.c (rank): Use counts only.
9417         * shrink-wrap.c (handle_simple_exit): Use counts only.
9418         (try_shrink_wrapping): Use counts only.
9419         (place_prologue_for_one_component): Use counts only.
9420         * tracer.c (find_best_predecessor): Use to_frequency.
9421         (find_trace): Use to_frequency.
9422         (tail_duplicate): Use to_frequency.
9423         * trans-mem.c (expand_transaction): Do not update frequency.
9424         * tree-call-cdce.c: Do not update frequency.
9425         * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9426         (gimple_merge_blocks): Likewise.
9427         (gimple_split_edge): Likewise.
9428         (gimple_duplicate_sese_region): Likewise.
9429         (gimple_duplicate_sese_tail): Likewise.
9430         (move_sese_region_to_fn): Likewise.
9431         (gimple_account_profile_record): Likewise.
9432         (insert_cond_bb): Likewise.
9433         * tree-complex.c (expand_complex_div_wide): Likewise.
9434         * tree-eh.c (lower_resx): Update profile.
9435         * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9436         frequencies.
9437         (initialize_cfun): Do not initialize frequencies
9438         (freqs_to_counts): Delete.
9439         (copy_cfg_body): Ignore count parameter.
9440         (copy_body): Update.
9441         (expand_call_inline): Update count_max.
9442         (optimize_inline_calls): Update count_max.
9443         (tree_function_versioning): Update count_max.
9444         * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9445         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9446         frequency.
9447         * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9448         * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9449         (try_peel_loop): Likewise.
9450         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9451         to_frequency.
9452         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9453         (tree_transform_and_unroll_loop): Do not use frequencies
9454         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9455         Use reliable prediction only.
9456         * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9457         * tree-ssa-sink.c (select_best_block): Use to_frequency.
9458         * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9459         probability scaling.
9460         * tree-ssa-threadupdate.c (create_block_for_threading): Do
9461         not update frequency
9462         (any_remaining_duplicated_blocks): Likewise.
9463         (update_profile): Likewise.
9464         (estimated_freqs_path): Delete.
9465         (freqs_to_counts_path): Delete.
9466         (clear_counts_path): Delete.
9467         (ssa_fix_duplicate_block_edges): Likewise.
9468         (duplicate_thread_path): Likewise.
9469         * tree-switch-conversion.c (gen_inbound_check): Use counts.
9470         * tree-tailcall.c (decrease_profile): Do not update frequency.
9471         (eliminate_tail_call): Likewise.
9472         * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9473         * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9474         (optimize_mask_stores): Likewise.
9475         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9476         * ubsan.c (ubsan_expand_null_ifn): Update profile.
9477         (ubsan_expand_ptr_ifn): Update profile.
9478         * value-prof.c (gimple_ic): Simplify.
9479         * value-prof.h (gimple_ic): Update prototype.
9480         * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9481         * ipa-inline.c (compute_uninlined_call_time): Be sure that
9482         counts are nonzero.
9483         (want_inline_self_recursive_call_p): Likewise.
9484         (resolve_noninline_speculation): Only cummulate defined counts.
9485         (inline_small_functions): Use nonzero_p.
9486         (ipa_inline): Do not access freed node.
9488 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9490         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9491         Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9493 2017-11-03  Kito Cheng  <kito.cheng@gmail.com>
9495         * config/riscv/riscv.c (riscv_legitimize_move): Handle
9496         non-legitimate address.
9498 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9500         * config/rs6000/rs6000.md (*lt0_disi): Delete.
9501         (*lt0_<mode>di, *lt0_<mode>si): New.
9503 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9505         * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9506         TARGET_PAIRED_FLOAT.
9508 2017-11-03  Siddhesh Poyarekar  <siddhesh.poyarekar@linaro.org>
9509             Jim Wilson  <jim.wilson@linaro.org>
9511         * config/aarch64/aarch64-cores.def (saphira): New CPU.
9512         * config/aarch64/aarch64-tune.md: Regenerated.
9513         * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9514         * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9516 2017-11-03  Cupertino Miranda  <cmiranda@synopsys.com>
9518         * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9519         (arc_expand_prologue): Restore blink for millicode.
9520         * config/arc/linux.h (LINK_EH_SPEC): Defined.
9522 2017-11-03  Richard Sandiford  <richard.sandiford@linaro.org>
9524         PR target/82809
9525         * config/i386/i386.c (ix86_vector_duplicate_value): Use
9526         gen_vec_duplicate after forcing the scalar into a register.
9528 2017-11-02  Segher Boessenkool  <segher@kernel.crashing.org>
9530         * combine (try_combine): Print the insns input to try_combine to the
9531         dump file.
9533 2017-11-02  Steve Ellcey  <sellcey@cavium.com>
9535         PR target/79868
9536         * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9537         Remove second argument from aarch64_process_target_attr call.
9538         * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9539         Ditto.
9540         * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9541         field type.
9542         (aarch64_handle_attr_arch): Remove second argument.
9543         (aarch64_handle_attr_cpu): Ditto.
9544         (aarch64_handle_attr_tune): Ditto.
9545         (aarch64_handle_attr_isa_flags): Ditto.
9546         (aarch64_process_one_target_attr): Ditto.
9547         (aarch64_process_target_attr): Ditto.
9548         (aarch64_option_valid_attribute_p): Remove second argument.
9549         on aarch64_process_target_attr call.
9551 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9553         * diagnostic.c: Include "selftest-diagnostic.h".
9554         (selftest::assert_location_text): New function.
9555         (selftest::test_diagnostic_get_location_text): New function.
9556         (selftest::diagnostic_c_tests): Call it.
9558 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9560         * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9561         * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9562         (class selftest::test_diagnostic_context): Move to...
9563         * selftest-diagnostic.c: New file.
9564         * selftest-diagnostic.h: New file.
9566 2017-11-02  James Bowman  <james.bowman@ftdichip.com>
9568         * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9569         offset range for FT32B.
9570         * config/ft32/ft32.h: option "mcompress" enables relaxation.
9571         * config/ft32/ft32.md: Add TARGET_NOPM.
9572         * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9573         * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9575 2017-11-02  Wilco Dijkstra  <wdijkstr@arm.com>
9577         * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9579 2017-11-02  Jeff Law  <law@redhat.com>
9581         * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9582         virtual keyword on FINAL OVERRIDE members.
9584         * tree-ssa-propagate.h (ssa_propagation_engine): Group
9585         virtuals together.  Add virtual destructor.
9586         (substitute_and_fold_engine): Similarly.
9588 2017-11-02  Jan Hubicka  <hubicka@ucw.cz>
9590         * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9592 2017-11-02  Richard Biener  <rguenther@suse.de>
9594         PR tree-optimization/82795
9595         * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9597 2017-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9599         * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9600         gcc_SUN_LD_VERSION.
9601         (gcc_GAS_CHECK_FEATURE): Remove.
9602         * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9603         gcc_AC_INITFINI_ARRAY here.  Update for Solaris 11.4 changes.
9604         * configure: Regenerate.
9606 2017-11-02  Claudiu Zissulescu <claziss@synopsys.com>
9608         * config/arc/arc.c (hwloop_optimize): Account for empty
9609         body loops.
9611 2017-11-02  Richard Biener  <rguenther@suse.de>
9613         PR middle-end/82765
9614         * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9615         Truncate ARRAY_REF index and element size.
9617 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9619         * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9621 2017-11-01  Jeff Law  <law@redhat.com>
9623         * tree-ssa-ccp.c (ccp_folder): New class derived from
9624         substitute_and_fold_engine.
9625         (ccp_folder::get_value): New member function.
9626         (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9627         (ccp_fold_stmt): Remove prototype.
9628         (ccp_finalize): Call substitute_and_fold from the ccp_class.
9629         * tree-ssa-copy.c (copy_folder): New class derived from
9630         substitute_and_fold_engine.
9631         (copy_folder::get_value): Renamed from get_value.
9632         (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9633         * tree-vrp.c (vrp_folder): New class derived from
9634         substitute_and_fold_engine.
9635         (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9636         (vrp_folder::get_value): New member function.
9637         (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9638         (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9639         * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9640         provide a class interface to folder/substitute routines.
9641         (ssa_prop_fold_stmt_fn): Remove typedef.
9642         (ssa_prop_get_value_fn): Likewise.
9643         (subsitute_and_fold): Remove prototype.
9644         (replace_uses_in): Likewise.
9645         * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9646         Renamed from replace_uses_in.  Call the virtual member function
9647         (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9648         (substitute_and_fold_dom_walker): Remove initialization of
9649         data member entries for calbacks.  Add substitute_and_fold_engine
9650         member and initialize it.
9651         (substitute_and_fold_dom_walker::before_dom_children0: Use the
9652         member functions for get_value, replace_phi_args_in c
9653         replace_uses_in, and fold_stmt calls.
9654         (substitute_and_fold_engine::substitute_and_fold): Renamed from
9655         substitute_and_fold.  Remove assert.   Update ctor call.
9657         * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9658         (ssa_prop_visit_phi_fn): Likewise.
9659         (class ssa_propagation_engine): New class to provide an interface
9660         into ssa_propagate.
9661         * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9662         variable.
9663         (ssa_prop_visit_phi): Likewise.
9664         (ssa_propagation_engine::simulate_stmt): Moved into class.
9665         Call visit_phi/visit_stmt from the class rather than via
9666         file scoped static variables.
9667         (ssa_propagation_engine::simulate_block): Moved into class.
9668         (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9669         (ssa_propagation_engine::ssa_propagate): Similarly.  No longer
9670         set file scoped statics for the visit_stmt/visit_phi callbacks.
9671         * tree-complex.c (complex_propagate): New class derived from
9672         ssa_propagation_engine.
9673         (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9674         (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9675         (tree_lower_complex): Call ssa_propagate via the complex_propagate
9676         class.
9677         * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9678         ssa_propagation_engine.
9679         (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9680         (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9681         (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9682         * tree-ssa-copy.c (copy_prop): New class derived from
9683         ssa_propagation_engine.
9684         (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9685         (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9686         (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9687         * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9688         (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9689         (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9690         (execute_vrp): Call ssa_propagate from the vrp_prop class.
9692 2017-11-01  Jakub Jelinek  <jakub@redhat.com>
9694         PR rtl-optimization/82778
9695         PR rtl-optimization/82597
9696         * compare-elim.c (struct comparison): Add in_a_setter field.
9697         (find_comparison_dom_walker::before_dom_children): Remove killed
9698         bitmap and df_simulate_find_defs call, instead walk the defs.
9699         Compute last_setter and initialize in_a_setter.  Merge definitions
9700         with first initialization for a few variables.
9701         (try_validate_parallel): Use insn_invalid_p instead of
9702         recog_memoized.  Return insn rather than just the pattern.
9703         (try_merge_compare): Fix up comment.  Don't uselessly test if
9704         in_a is a REG_P.  Use cmp->in_a_setter instead of walking UD
9705         chains.
9706         (execute_compare_elim_after_reload): Remove df_chain_add_problem
9707         call.
9709 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9710             Alan Hayward  <alan.hayward@arm.com>
9711             David Sherwood  <david.sherwood@arm.com>
9713         * config/aarch64/aarch64.c (aarch64_rtx_costs): Use
9714         aarch64_hard_regno_nregs to get the number of registers
9715         in a mode.
9717 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9718             Alan Hayward  <alan.hayward@arm.com>
9719             David Sherwood  <david.sherwood@arm.com>
9721         * config/aarch64/constraints.md (Upl): Rename to...
9722         (Uaa): ...this.
9723         * config/aarch64/aarch64.md
9724         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9725         Update accordingly.
9727 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9728             Alan Hayward  <alan.hayward@arm.com>
9729             David Sherwood  <david.sherwood@arm.com>
9731         * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9732         (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9733         earlier in file.
9735 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9736             Alan Hayward  <alan.hayward@arm.com>
9737             David Sherwood  <david.sherwood@arm.com>
9739         * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9740         (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9741         (aarch64_evpc_dup): Generate rtl direcly, rather than using
9742         named expanders.
9743         (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9744         of a single element.
9745         * config/aarch64/iterators.md: Add a comment above the permute
9746         unspecs to say that they are generated directly by
9747         aarch64_expand_vec_perm_const.
9748         * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9750 2017-11-01  Nathan Sidwell  <nathan@acm.org>
9752         * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9754 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9756         * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9757         and medany code models, and describe what they do.
9759 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9761         Revert accidental duplicate:
9763         * combine.c (can_change_dest_mode): Reject changes in
9764         REGMODE_NATURAL_SIZE.
9766 2017-11-01  Segher Boessenkool  <segher@kernel.crashing.org>
9768         PR rtl-optimization/64682
9769         PR rtl-optimization/69567
9770         PR rtl-optimization/69737
9771         PR rtl-optimization/82683
9772         * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9773         register mentioned in the note, drop the note, unless it came from I3,
9774         in which case it should go to I3 again.
9776 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9778         * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9779         and return false if not.
9780         (clear_bytes_written_by, live_bytes_read): Update accordingly.
9782 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9784         * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9785         range is known to be empty.
9787 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9788             Alan Hayward  <alan.hayward@arm.com>
9789             David Sherwood  <david.sherwood@arm.com>
9791         * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9792         and CONST_VECTOR_NUNITS instead of computing the number of units from
9793         the byte sizes of the vector and element.
9794         (simplify_binary_operation_1): Likewise.
9795         (simplify_const_binary_operation): Likewise.
9796         (simplify_ternary_operation): Likewise.
9798 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9799             Alan Hayward  <alan.hayward@arm.com>
9800             David Sherwood  <david.sherwood@arm.com>
9802         * var-tracking.c (INT_MEM_OFFSET): Replace with...
9803         (int_mem_offset): ...this new function.
9804         (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9805         (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9806         (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9807         Update accordingly.
9809 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9810             Alan Hayward  <alan.hayward@arm.com>
9811             David Sherwood  <david.sherwood@arm.com>
9813         * lower-subreg.c (interesting_mode_p): New function.
9814         (compute_costs, find_decomposable_subregs, decompose_register)
9815         (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9816         (resolve_clobber, dump_choices): Use it.
9818 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9819             Alan Hayward  <alan.hayward@arm.com>
9820             David Sherwood  <david.sherwood@arm.com>
9822         * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
9824 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9825             Alan Hayward  <alan.hayward@arm.com>
9826             David Sherwood  <david.sherwood@arm.com>
9828         * alias.c (find_base_value, find_base_term): Only process integer
9829         truncations.  Check the precision rather than the size.
9831 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9832             Alan Hayward  <alan.hayward@arm.com>
9833             David Sherwood  <david.sherwood@arm.com>
9835         * machmode.h (is_narrower_int_mode): New function
9836         * optabs.c (expand_float, expand_fix): Use it.
9837         * dwarf2out.c (rotate_loc_descriptor): Likewise.
9839 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9840             Alan Hayward  <alan.hayward@arm.com>
9841             David Sherwood  <david.sherwood@arm.com>
9843         * rtl.h (narrower_subreg_mode): New function.
9844         * ira-color.c (update_costs_from_allocno): Use it.
9846 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9847             Alan Hayward  <alan.hayward@arm.com>
9848             David Sherwood  <david.sherwood@arm.com>
9850         * optabs-query.h (convert_optab_p): New function, split out from...
9851         (convert_optab_handler): ...here.
9852         (widening_optab_handler): Delete.
9853         (find_widening_optab_handler): Remove permit_non_widening parameter.
9854         (find_widening_optab_handler_and_mode): Likewise.  Provide an
9855         override that operates on mode class wrappers.
9856         * optabs-query.c (widening_optab_handler): Delete.
9857         (find_widening_optab_handler_and_mode): Remove permit_non_widening
9858         parameter.  Assert that the two modes are the same class and that
9859         the "from" mode is narrower than the "to" mode.  Use
9860         convert_optab_handler instead of widening_optab_handler.
9861         * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9862         instead of widening_optab_handler.
9863         * expr.c (expand_expr_real_2): Update calls to
9864         find_widening_optab_handler.
9865         * optabs.c (expand_widen_pattern_expr): Likewise.
9866         (expand_binop_directly): Take the insn_code as a parameter.
9867         (expand_binop): Only call find_widening_optab_handler for
9868         conversion optabs; use optab_handler otherwise.  Update calls
9869         to find_widening_optab_handler and expand_binop_directly.
9870         Use convert_optab_handler instead of widening_optab_handler.
9871         * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9872         find_widening_optab_handler and use scalar_mode rather than
9873         machine_mode.
9874         (convert_plusminus_to_widen): Likewise.
9876 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9877             Alan Hayward  <alan.hayward@arm.com>
9878             David Sherwood  <david.sherwood@arm.com>
9880         * machmode.h (fixed_size_mode): New class.
9881         * rtl.h (get_pool_mode): Return fixed_size_mode.
9882         * gengtype.c (main): Add fixed_size_mode.
9883         * target.def (get_raw_result_mode): Return a fixed_size_mode.
9884         (get_raw_arg_mode): Likewise.
9885         * doc/tm.texi: Regenerate.
9886         * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9887         * targhooks.c (default_get_reg_raw_mode): Likewise.
9888         * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9889         * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9890         * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9891         (msp430_get_raw_result_mode): Likewise.
9892         * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9893         * dbxout.c (dbxout_parms): Require fixed-size modes.
9894         * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9895         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9896         * omp-low.c (lower_oacc_reductions): Likewise.
9897         * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
9898         (simplify_subreg): Update accordingly.
9899         * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
9900         (force_const_mem): Update accordingly.  Return NULL_RTX for modes
9901         that aren't fixed-size.
9902         (get_pool_mode): Return a fixed_size_mode.
9903         (output_constant_pool_2): Take a fixed_size_mode.
9905 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9906             Alan Hayward  <alan.hayward@arm.com>
9907             David Sherwood  <david.sherwood@arm.com>
9909         * doc/rtl.texi (vec_series): Document.
9910         (const): Say that the operand can be a vec_series.
9911         * rtl.def (VEC_SERIES): New rtx code.
9912         * rtl.h (const_vec_series_p_1): Declare.
9913         (const_vec_series_p): New function.
9914         * emit-rtl.h (gen_const_vec_series): Declare.
9915         (gen_vec_series): Likewise.
9916         * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
9917         (gen_vec_series): Likewise.
9918         * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
9919         * simplify-rtx.c (simplify_unary_operation): Handle negations
9920         of vector series.
9921         (simplify_binary_operation_series): New function.
9922         (simplify_binary_operation_1): Use it.  Handle VEC_SERIES.
9923         (test_vector_ops_series): New function.
9924         (test_vector_ops): Call it.
9925         * config/powerpcspe/altivec.md (altivec_lvsl): Use
9926         gen_const_vec_series.
9927         (altivec_lvsr): Likewise.
9928         * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
9930 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9931             Alan Hayward  <alan.hayward@arm.com>
9932             David Sherwood  <david.sherwood@arm.com>
9934         * doc/rtl.texi (const): Update description of address constants.
9935         Say that vector constants are allowed too.
9936         * common.md (E, F): Use CONSTANT_P instead of checking for
9937         CONST_VECTOR.
9938         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9939         checking for CONST_VECTOR.
9940         * expmed.c (make_tree): Use build_vector_from_val for a CONST
9941         VEC_DUPLICATE.
9942         * expr.c (expand_expr_real_2): Check for vector modes instead
9943         of checking for CONST_VECTOR.
9944         * rtl.h (const_vec_p): New function.
9945         (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
9946         (unwrap_const_vec_duplicate): Handle them here too.
9948 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9949             David Malcolm  <dmalcolm@redhat.com>
9950             Alan Hayward  <alan.hayward@arm.com>
9951             David Sherwood  <david.sherwood@arm.com>
9953         * rtl.h (vec_duplicate_p): New function.
9954         * selftest-rtl.c (assert_rtx_eq_at): New function.
9955         * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
9956         (assert_rtx_eq_at): Declare.
9957         * selftest.h (selftest::simplify_rtx_c_tests): Declare.
9958         * selftest-run-tests.c (selftest::run_tests): Call it.
9959         * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
9960         (simplify_unary_operation_1): Recursively handle vector duplicates.
9961         (simplify_binary_operation_1): Likewise.  Handle VEC_SELECTs of
9962         vector duplicates.
9963         (simplify_subreg): Handle subregs of vector duplicates.
9964         (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
9965         (selftest::simplify_rtx_c_tests): New functions.
9967 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9968             Alan Hayward  <alan.hayward@arm.com>
9969             David Sherwood  <david.sherwood@arm.com>
9971         * emit-rtl.h (gen_const_vec_duplicate): Declare.
9972         (gen_vec_duplicate): Likewise.
9973         * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
9974         out from...
9975         (gen_const_vector): ...here.
9976         (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
9977         (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
9978         whose elements are all equal.
9979         * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
9980         * simplify-rtx.c (simplify_const_unary_operation): Likewise.
9981         (simplify_relational_operation): Likewise.
9982         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
9983         Likewise.
9984         (aarch64_simd_dup_constant): Use gen_vec_duplicate.
9985         (aarch64_expand_vector_init): Likewise.
9986         * config/arm/arm.c (neon_vdup_constant): Likewise.
9987         (neon_expand_vector_init): Likewise.
9988         (arm_expand_vec_perm): Use gen_const_vec_duplicate.
9989         (arm_block_set_unaligned_vect): Likewise.
9990         (arm_block_set_aligned_vect): Likewise.
9991         * config/arm/neon.md (neon_copysignf<mode>): Likewise.
9992         * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
9993         (expand_vec_perm_even_odd_pack): Likewise.
9994         (ix86_vector_duplicate_value): Use gen_vec_duplicate.
9995         * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
9996         * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
9997         gen_const_vec_duplicate.
9998         * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
9999         * config/mips/mips.c (mips_gen_const_int_vector): Use
10000         gen_const_vec_duplicate.
10001         (mips_expand_vector_init): Use CONST0_RTX.
10002         * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
10003         (define_split): Use gen_const_vec_duplicate.
10004         * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
10005         (define_split): Use gen_const_vec_duplicate.
10006         * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
10007         (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
10008         * config/spu/spu.c (spu_const): Likewise.
10010 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
10011             Alan Hayward  <alan.hayward@arm.com>
10012             David Sherwood  <david.sherwood@arm.com>
10014         * combine.c (can_change_dest_mode): Reject changes in
10015         REGMODE_NATURAL_SIZE.
10017 2017-10-31  Sandra Loosemore  <sandra@codesourcery.com>
10019         * configure.ac (--enable-libssp): New.
10020         (gcc_cv_libc_provides_ssp): Check for explicit setting before
10021         trying to determine target-specific default.  Adjust indentation.
10022         * configure: Regenerated.
10023         * doc/install.texi (Configuration): Expand --disable-libssp
10024         documentation.
10026 2017-10-31  Daniel Santos  <daniel.santos@pobox.com>
10028         config/i386/i386.c (ix86_expand_epilogue): Correct stack
10029         calculation.
10031 2017-10-31  Martin Jambor  <mjambor@suse.cz>
10033         PR c++/81702
10034         * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
10036 2017-10-31  David Malcolm  <dmalcolm@redhat.com>
10038         * auto-profile.c (autofdo_source_profile::read): Use
10039         UNKNOWN_LOCATION rather than 0.
10040         * diagnostic-core.h (warning_at_rich_loc): Rename to...
10041         (warning_at): ...this overload.
10042         (warning_at_rich_loc_n): Rename to...
10043         (warning_n): ...this overload.
10044         (error_at_rich_loc): Rename to...
10045         (error_at): ...this overload.
10046         (pedwarn_at_rich_loc): Rename to...
10047         (pedwarn): ...this overload.
10048         (permerror_at_rich_loc): Rename to...
10049         (permerror): ...this overload.
10050         (inform_at_rich_loc): Rename to...
10051         (inform): ...this overload.
10052         * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
10053         (diagnostic_n_impl_richloc): Rename to...
10054         (diagnostic_n_impl): ...this rich_location *-based decl.
10055         (inform_at_rich_loc): Rename to...
10056         (inform): ...this, and add an assertion.
10057         (inform_n): Update for removal of location_t-based diagnostic_n_impl.
10058         (warning_at_rich_loc): Rename to...
10059         (warning_at): ...this, and add an assertion.
10060         (warning_at_rich_loc_n): Rename to...
10061         (warning_n): ...this, and add an assertion.
10062         (warning_n): Update location_t-based implementation for removal of
10063         location_t-based diagnostic_n_impl.
10064         (pedwarn_at_rich_loc): Rename to...
10065         (pedwarn): ...this, and add an assertion.
10066         (permerror_at_rich_loc): Rename to...
10067         (permerror): ...this, and add an assertion.
10068         (error_n): Update for removal of location_t-based diagnostic_n_impl.
10069         (error_at_rich_loc): Rename to...
10070         (error_at): ...this, and add an assertion.
10071         * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
10072         (driver::do_spec_on_infiles): Likewise.
10073         * substring-locations.c (format_warning_va): Update for renaming
10074         of inform_at_rich_loc.
10076 2017-10-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
10078         * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
10079         _Float<N>X built-in functions so that the variant without the
10080         "__builtin_" prefix is only enabled for the GNU C and Objective C
10081         languages when they are in non-strict ANSI/ISO mode.
10082         (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
10083         * target.def (floatn_builtin_p): Add a target hook to control
10084         whether _Float<N> and _Float<N>X built-in functions without the
10085         "__builtin_" prefix are enabled, and return true for C and
10086         Objective C in the default hook.  Include langhooks.h in
10087         targhooks.c.
10088         * targhooks.h (default_floatn_builtin_p): Likewise.
10089         * targhooks.c (default_floatn_builtin_p): Likewise.
10090         * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
10091         floatn_builtin_p target hook.
10092         * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
10094 2017-10-31  Matthew Fortune  <matthew.fortune@imgtec.com>
10095             Eric Botcazou  <ebotcazou@adacore.com>
10097         PR rtl-optimization/81803
10098         * lra-constraints.c (curr_insn_transform): Also reload the whole
10099         register for a strict subreg no wider than a word if this is for
10100         a WORD_REGISTER_OPERATIONS target.
10102 2017-10-31  Jason Merrill  <jason@redhat.com>
10104         * gdbinit.in: Skip over inlines from timevar.h.
10106 2017-10-31  Martin Liska  <mliska@suse.cz>
10108         * doc/gcov.texi: Document new option.
10109         * gcov.c (print_usage): Likewise print it.
10110         (process_args): Support the argument.
10111         (format_count): New function.
10112         (format_gcov): Use the function.
10114 2017-10-31  Martin Liska  <mliska@suse.cz>
10116         * gcov.c (struct name_map): do not use typedef.
10117         Define operator== and operator<.
10118         (name_search): Remove.
10119         (name_sort): Remove.
10120         (main): Do not allocate names.
10121         (process_file): Add vertical space.
10122         (generate_results): Use std::find.
10123         (release_structures): Do not release memory.
10124         (find_source): Use std::find.
10126 2017-10-31  Martin Liska  <mliska@suse.cz>
10128         * gcov.c (struct line_info): Remove it's typedef.
10129         (line_info::line_info): Add proper ctor.
10130         (line_info::has_block): Do not use a typedef.
10131         (struct source_info): Do not use typedef.
10132         (circuit): Likewise.
10133         (get_cycles_count): Likewise.
10134         (output_intermediate_file): Iterate via vector iterator.
10135         (add_line_counts): Use std::vector methods.
10136         (accumulate_line_counts): Likewise.
10137         (output_lines): Likewise.
10139 2017-10-31  Martin Liska  <mliska@suse.cz>
10141         * gcov.c (struct source_info): Remove typedef.
10142         (source_info::source_info): Add proper ctor.
10143         (accumulate_line_counts): Use struct, not it's typedef.
10144         (output_gcov_file): Likewise.
10145         (output_lines): Likewise.
10146         (main): Do not allocate an array.
10147         (output_intermediate_file): Use size of vector container.
10148         (process_file): Resize the vector.
10149         (generate_results): Do not preallocate, use newly added vector
10150         lines.
10151         (release_structures): Do not release sources.
10152         (find_source): Use vector methods.
10153         (add_line_counts): Do not use typedef.
10155 2017-10-31  Martin Liska  <mliska@suse.cz>
10157         * doc/gcov.texi: Document that.
10158         * gcov.c (add_line_counts): Mark lines with a non-executed
10159         statement.
10160         (output_line_beginning): Handle such lines.
10161         (output_lines): Pass new argument.
10162         (output_intermediate_file): Print it in intermediate format.
10164 2017-10-31  Martin Liska  <mliska@suse.cz>
10166         * color-macros.h: New file.
10167         * diagnostic-color.c: Factor out color related to macros to
10168         color-macros.h.
10169         * doc/gcov.texi: Document -k option.
10170         * gcov.c (INCLUDE_STRING): Include string.h.
10171         (print_usage): Add -k option.
10172         (process_args): Parse it.
10173         (pad_count_string): New function.
10174         (output_line_beginning): Likewise.
10175         (DEFAULT_LINE_START): New macro.
10176         (output_lines): Support color output.
10178 2017-10-31  Martin Liska  <mliska@suse.cz>
10180         PR gcov-profile/82633
10181         * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
10182         their interaction with GCOV infrastructure.
10183         * configure.ac: Add -fkeep-{inline,static}-functions to
10184         coverage_flags.
10185         * configure: Regenerate.
10187 2017-10-31  Uros Bizjak  <ubizjak@gmail.com>
10189         PR target/82772
10190         * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
10192 2017-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
10194         PR target/82674
10195         * config/rs6000/rs6000.md (allocate_stack): Force update interval
10196         into a register if it does not fit into an immediate offset field.
10198 2017-10-31  Olivier Hainque  <hainque@adacore.com>
10200         * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
10202 2017-10-31  Julia Koval  <julia.koval@intel.com>
10204         * config.gcc: Add gfniintrin.h.
10205         * config/i386/gfniintrin.h: New.
10206         * config/i386/i386-builtin-types.def
10207         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10208         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
10209         __builtin_ia32_vgf2p8affineinvqb_v32qi,
10210         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
10211         __builtin_ia32_vgf2p8affineinvqb_v16qi,
10212         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
10213         * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
10214         V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
10215         V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
10216         V64QI_FTYPE_V64QI_V64QI_INT): New types.
10217         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
10218         * config/i386/immintrin.h: Include gfniintrin.h.
10219         * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
10221 2017-10-30  Eric Botcazou  <ebotcazou@adacore.com>
10223         * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10225 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10227         * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10228         (arm_ashldi3_1bit): Remove pattern.
10229         (ashrdi3): Remove shift by 1 expansion.
10230         (arm_ashrdi3_1bit): Remove pattern.
10231         (lshrdi3): Remove shift by 1 expansion.
10232         (arm_lshrdi3_1bit): Remove pattern.
10233         * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10234         cost of ashldi3 by 1.
10235         * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10236         (<shift>di3_neon): Likewise.
10238 2017-10-30  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
10240         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10241         both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10242         and (*aarch64_simd_mov<VQ:mode>).
10243         (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10244         pattern alternative.
10245         (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10246         attributes to match pattern alternative.
10248 2017-10-30  Steven Munroe  <munroesj@gcc.gnu.org>
10250         * config.gcc (powerpc*-*-*): Add emmintrin.h.
10251         * config/rs6000/emmintrin.h: New file.
10252         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10254 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10256         * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10257         * (movdi_vfp_cortexa8): Remove pattern.
10259 2017-10-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10261         * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10262         etc. reference.
10263         (Specific, alpha*-dec-osf5.1): Remove.
10264         (Specific, mips-sgi-irix5): Remove.
10265         (Specific, mips-sgi-irix6): Remove.
10267 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10269         PR middle-end/22141
10270         * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10271         arguments to clear_bit_region_be.
10273 2017-10-30  Jim Wilson  <wilson@tuliptree.org>
10275         * gimplify.c: Include memmodel.h.
10277 2017-10-30  Martin Jambor  <mjambor@suse.cz>
10279         * omp-grid.c (grid_attempt_target_gridification): Also insert a
10280         condition whether loop should be executed at all.
10282 2017-10-30  Will Schmidt  <will_schmidt@vnet.ibm.com>
10284         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10285         gimple folding of vec_madd() intrinsics.
10286         * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10287         fmav8hi4.  (altivec_vmladduhm): Rename to fmav8hi4.
10288         * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10290 2017-10-30  Richard Biener  <rguenther@suse.de>
10292         PR tree-optimization/82762
10293         Revert
10294         2017-10-23  Richard Biener  <rguenther@suse.de>
10296         PR tree-optimization/82129
10297         Revert
10298         2017-08-01  Richard Biener  <rguenther@suse.de>
10300         PR tree-optimization/81181
10301         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10302         (compute_antic): ... end of iteration here.
10304 2017-10-30  Joseph Myers  <joseph@codesourcery.com>
10306         * doc/invoke.texi (C Dialect Options): Document -std=c17,
10307         -std=iso9899:2017 and -std=gnu17.
10308         * doc/standards.texi (C Language): Document C17 support.
10309         * doc/cpp.texi (Overview): Mention -std=c17.
10310         (Standard Predefined Macros): Document C11 and C17 values of
10311         __STDC_VERSION__.  Do not refer to C99 support as incomplete.
10312         * doc/extend.texi (Inline): Do not list individual options for
10313         standards newer than C99.
10314         * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10315         "GNU C17".
10316         * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10317         language name.
10319 2017-10-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
10321         * asan.c (asan_finish_file): Align asan globals array by shadow
10322         granularity.
10324 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10326         PR middle-end/22141
10327         * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10328         (struct store_immediate_info): Add bitregion_start and bitregion_end
10329         fields.
10330         (store_immediate_info::store_immediate_info): Add brs and bre
10331         arguments and initialize bitregion_{start,end} from those.
10332         (struct merged_store_group): Add bitregion_start, bitregion_end,
10333         align_base and mask fields.  Drop unnecessary struct keyword from
10334         struct store_immediate_info.  Add do_merge method.
10335         (clear_bit_region_be): Use memset instead of loop storing zeros.
10336         (merged_store_group::do_merge): New method.
10337         (merged_store_group::merge_into): Use do_merge.  Allow gaps in between
10338         stores as long as the surrounding bitregions have no gaps.
10339         (merged_store_group::merge_overlapping): Use do_merge.
10340         (merged_store_group::apply_stores): Test that bitregion_{start,end}
10341         is byte aligned, rather than requiring that start and width are
10342         byte aligned.  Drop unnecessary struct keyword from
10343         struct store_immediate_info.  Allocate and populate also mask array.
10344         Make start of the arrays relative to bitregion_start rather than
10345         start and size them according to bitregion_{end,start} difference.
10346         (struct imm_store_chain_info): Drop unnecessary struct keyword from
10347         struct store_immediate_info.
10348         (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10349         (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10350         struct keyword from struct store_immediate_info.
10351         (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10352         between stores as long as the surrounding bitregions have no gaps.
10353         Formatting fixes.
10354         (struct split_store): Add orig non-static data member.
10355         (split_store::split_store): Initialize orig to false.
10356         (find_constituent_stmts): Return store_immediate_info *, non-NULL
10357         if there is exactly a single original stmt.  Change stmts argument
10358         to pointer from reference, if NULL, don't push anything to it.  Add
10359         first argument, use it to optimize skipping over orig stmts that
10360         are known to be before bitpos already.  Simplify.
10361         (split_group): Return unsigned int count how many stores are or
10362         would be needed rather than a bool.  Add allow_unaligned argument.
10363         Change split_stores argument from reference to pointer, if NULL,
10364         only do a dry run computing how many stores would be produced.
10365         Rewritten algorithm to use both alignment and misalign if
10366         !allow_unaligned and handle bitfield stores with gaps.
10367         (imm_store_chain_info::output_merged_store): Set start_byte_pos
10368         from bitregion_start instead of start.  Compute allow_unaligned
10369         here, if true, do 2 split_group dry runs to compute which one
10370         produces fewer stores and prefer aligned if equal.  Punt if
10371         new count is bigger or equal than original before emitting any
10372         statements, rather than during that.  Remove no longer needed
10373         new_ssa_names tracking.  Replace num_stmts with
10374         split_stores.length ().  Use 32-bit stack allocated entries
10375         in split_stores auto_vec.  Try to reuse original store lhs/rhs1
10376         if possible.  Handle bitfields with gaps.
10377         (pass_store_merging::execute): Ignore bitsize == 0 stores.
10378         Compute bitregion_{start,end} for the stores and construct
10379         store_immediate_info with that.  Formatting fixes.
10381 2017-10-30  Uros Bizjak  <ubizjak@gmail.com>
10383         PR target/82725
10384         * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10385         UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10387 2017-10-29  Jim Wilson  <wilson@tuliptree.org>
10389         * gimplify.c: Include tm_p.h.
10391         * common.opt (gcoff): Re-add as ignored option.
10392         (gcoff1, gcoff2, gcoff3): Likewise.
10394         * Makefile.in (OBJS): Delete sdbout.o.
10395         (GTFILES): Delete $(srcdir)/sdbout.c.
10396         * debug.h: Delete sdb_debug_hooks.
10397         * final.c: Delete sdbout.h include.
10398         (final_scan_insn): Delete SDB_DEBUG check.
10399         (rest_of_clean_state): Likewise.
10400         * output.h: Delete sdb_begin_function_line.
10401         * sdbout.c: Delete.
10402         * sdbout.h: Delete.
10403         * toplev.c: Delete sdbout.h include.
10404         (process_options): Delete SDB_DEBUG check.
10405         * tree-core.h (tree_type_common): Delete pointer field of
10406         tree_type_symtab.
10407         * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10408         TYPE_SYMTAB_POINTER.
10409         * tree.h (TYPE_SYMTAB_POINTER): Delete.
10410         (TYPE_SYMTAB_IS_POINTER): Delete.
10411         (TYPE_SYMTAB_IS_DIE): Renumber.
10412         * xcoffout.c: Refer to former sdbout.c file.
10413         (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10415         * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10416         * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10417         (Debugging Options): Delete -gcoff.
10418         (-gstabs): Delete SDB reference.
10419         (-gcoff): Delete.
10420         (-gcoff@var{level}): Delete.
10421         * doc/passes.texi (Debugging information output): Delete SDB and
10422         sdbout.c references.
10423         * doc/tm.texi: Regenerate.
10424         * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10425         (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10426         references.
10427         (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10428         (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10429         (SDB_DEBUGGING_INFO): Delete.
10430         (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10431         SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10432         * target.def (output_source_filename): Delete COFF reference.
10434         * common.opt (gcoff): Delete.
10435         (gxcoff+): Update Negative chain.
10436         * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10437         SDB_DEBUG.
10438         * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10439         * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10440         * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10441         and SDB references.
10442         (expand_function_start): Change sdb reference to past tense.
10443         (expand_function_end): Change sdb reference to past tense.
10444         * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10445         * opts.c (debug_type_names): Delete coff entry.
10446         (common_handle_option): Delete OPT_gcoff case.
10447         * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10449         * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10450         * config/cris/cris.h: Delete SDB reference in comment.
10451         * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10452         (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10453         * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10454         * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10455         to past tense.
10456         (ix86_expand_prologue): Likewise.
10457         * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10458         * config/ia64/ia64.h: Likewise.
10459         * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10460         * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10461         support.
10462         * config/mmix/mmix.h: Likewise.
10463         * config/nds32/nds32.c: Likewise.
10464         * config/stormy/storym16.h: Likewise.
10465         * config/visium/visium.h: Likewise.
10466         * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10468 2017-10-28  Sandra Loosemore  <sandra@codesourcery.com>
10470         * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10471         * config/nios2/nios2.c (nios2_initial_elimination_offset):  Make
10472         FRAME_POINTER_REGNUM point at high end of local var area.
10474 2017-10-27  Eric Botcazou  <ebotcazou@adacore.com>
10476         * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10477         Move comment around.  Do not reset best_edge for a copiable
10478         destination if the copy would cause a partition change.
10479         (better_edge_p): Remove redundant check.
10481 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10483         * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10485 2017-10-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
10487         * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10488         for math functions that have _Float<N> and _Float<N>X variants.
10489         (mathfn_built_in_2): Add support for math functions that have
10490         _Float<N> and _Float<N>X variants.
10491         (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10492         (expand_builtin_mathfn_ternary): Add support for fma with
10493         _Float<N> and _Float<N>X variants.
10494         (expand_builtin): Likewise.
10495         (fold_builtin_3): Likewise.
10496         * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10497         create math function _Float<N> and _Float<N>X variants as external
10498         library builtins.
10499         (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10500         DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10501         the __builtin_ prefix and if not strict ansi, without the prefix.
10502         (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10503         (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10504         (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10505         (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10506         (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10507         (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10508         * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10509         function signatures for fma _Float<N> and _Float<N>X variants.
10510         (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10511         (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10512         (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10513         (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10514         (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10515         (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10516         * gencfn-macros.c (print_case_cfn): Add support for math functions
10517         that have _Float<N> and _Float<N>X variants.
10518         (print_define_operator_list): Likewise.
10519         (fltfn_suffixes): Likewise.
10520         (main): Likewise.
10521         * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10522         for math functions that have _Float<N> and _Float<N>X variants.
10523         (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10524         and _Float<N>X variants.
10525         (COPYSIGN): Likewise.
10526         (FMIN): Likewise.
10527         (FMAX): Likewise.
10528         * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10529         copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10530         variants.
10531         (integer_valued_read_call_p): Likewise.
10532         * fold-const-call.c (fold_const_call_ss): Likewise.
10533         (fold_const_call_sss): Add support for copysign, fmin, and fmax
10534         _Float<N> and _Float<N>X variants.
10535         (fold_const_call_ssss): Add support for fma _Float<N> and
10536         _Float<N>X variants.
10537         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10538         support for copysign and fma _Float<N> and _Float<N>X variants.
10539         (backprop::process_builtin_call_use): Likewise.
10540         * tree-call-cdce.c (can_test_argument_range); Add support for
10541         sqrt _Float<N> and _Float<N>X variants.
10542         (edom_only_function): Likewise.
10543         (get_no_error_domain): Likewise.
10544         * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10545         * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10546         copysign _Float<N> and _Float<N>X variants.
10547         * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10548         handled by machine independent code.
10549         (FMAF128): Likewise.
10550         * doc/cpp.texi (Common Predefined Macros): Document defining
10551         __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10552         fma _Float<N> and _Float<N>X variants.
10554 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10556         PR target/82692
10557         * config/i386/i386-modes.def (CCFPU): Remove definition.
10558         * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10559         (ix86_cc_modes_compatible): Ditto.
10560         (ix86_expand_carry_flag_compare): Ditto.
10561         (ix86_expand_int_movcc): Ditto.
10562         (ix86_expand_int_addcc): Ditto.
10563         (ix86_reverse_condition): Ditto.
10564         (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10565         Return true/false for unordered/ordered fp comparisons.
10566         (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10567         (ix86_prepare_fp_compare_args): Update for rename.
10568         (ix86_expand_fp_compare): Update for rename.  Generate unordered
10569         compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10570         (ix86_expand_sse_compare_and_jump): Ditto.
10571         * config/i386/predicates.md (fcmov_comparison_operator):
10572         Remove CCFPU mode handling.
10573         (ix86_comparison_operator): Ditto.
10574         (ix86_carry_flag_operator): Ditto.
10575         * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10576         (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10577         (*cmpu<mode>_cc_i387): Ditto.
10578         (FPCMP): Remove mode iterator.
10579         (unord): Remove mode attribute.
10580         (unord_subst): New define_subst transformation
10581         (unord): New define_subst attribute.
10582         (unordered): Ditto.
10583         (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10584         (*cmpi<unord>xf_i387): Ditto.
10585         * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10586         from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10587         using unord_subst transformation.
10588         * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10589         (round_saeonly): Also handle CCFP mode.
10590         * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10591         Remove UNSPEC_SAHF unspec handling.
10593 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10595         * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10597 2017-10-27  Jeff Law  <law@redhat.com>
10599         * gimple-ssa-sprintf.c: Include domwalk.h.
10600         (class sprintf_dom_walker): New class, derived from dom_walker.
10601         (sprintf_dom_walker::before_dom_children): New function.
10602         (struct call_info): Moved into sprintf_dom_walker class
10603         (compute_formath_length, handle_gimple_call): Likewise.
10604         (sprintf_length::execute): Call the dominator walker rather
10605         than walking the statements.
10607         * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10608         gimple statement locations.
10609         (check_array_bounds): Corresponding changes.  Get the statement's
10610         location directly from wi->stmt.
10612 2017-10-27  Palmer Dabbelt  <palmer@dabbelt.com>
10614         PR target/82717
10615         * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10617 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10619         * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10620         X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10622 2017-10-27  Jakub Jelinek  <jakub@redhat.com>
10624         PR target/82703
10625         * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10626         * config/i386/i386.c (maybe_get_pool_constant): Removed.
10627         (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10628         maybe_get_pool_constant.
10629         * config/i386/predicates.md (zero_extended_scalar_load_operand):
10630         Likewise.
10632 2017-10-27  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10634         * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10635         2.26 caveat.  Update gas and gld versions.
10636         (Specific, *-*-solaris2*): Update binutils version.  Remove caveat
10637         reference.
10639 2017-10-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10641         * cgraph.h (set_malloc_flag): Declare.
10642         * cgraph.c (set_malloc_flag_1): New function.
10643         (set_malloc_flag): Likewise.
10644         * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10645         * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10646         false.
10647         (read_ipa_call_summary): Add support for reading is_return_callee.
10648         (write_ipa_call_summary): Stream is_return_callee.
10649         * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10650         * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10651         ipa-prop.h, ipa-fnsummary.h.
10652         (pure_const_names): Change to static.
10653         (malloc_state_e): Define.
10654         (malloc_state_names): Define.
10655         (funct_state_d): Add field malloc_state.
10656         (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10657         (check_retval_uses): New function.
10658         (malloc_candidate_p): Likewise.
10659         (analyze_function): Add support for malloc attribute.
10660         (pure_const_write_summary): Stream malloc_state.
10661         (pure_const_read_summary): Add support for reading malloc_state.
10662         (dump_malloc_lattice): New function.
10663         (propagate_malloc): New function.
10664         (warn_function_malloc): New function.
10665         (ipa_pure_const::execute): Call propagate_malloc and
10666         ipa_free_fn_summary.
10667         (pass_local_pure_const::execute): Add support for malloc attribute.
10668         * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10669         * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10671 2017-10-27  Martin Liska  <mliska@suse.cz>
10673         PR gcov-profile/82457
10674         * doc/invoke.texi: Document that one needs a non-strict ISO mode
10675         for fork-like functions to be properly instrumented.
10677 2017-10-27  Richard Biener  <rguenther@suse.de>
10679         PR middle-end/81659
10680         * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10681         info when we redirected EH.
10683 2017-10-26  Michael Collison  <michael.collison@arm.com>
10685         * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10686         New pattern.
10687         (<optab>_trunchf<GPI:mode>2: New pattern.
10688         (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10689         * config/aarch64/iterators.md (wv): New mode attribute.
10690         (vf, VF): New mode attributes.
10691         (vgp, VGP): New mode attributes.
10692         (s): Update attribute with SImode and DImode prefixes.
10694 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10696         * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10697         * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10698         * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10699         (nios2_option_overide): Initialize it.  Don't allow R0-relative
10700         addressing with PIC.
10701         (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10702         (nios2_symbolic_constant_p): Likewise.
10703         (nios2_legitimate_address_p): Likewise.
10704         (nios2_r0rel_section_name_p): New.
10705         (nios2_symbol_ref_in_r0rel_data_p): New.
10706         (nios2_emit_move_sequence): Handle r0rel_constant_p.
10707         (r0rel_constant_p): New.
10708         (nios2_print_operand_address): Handle r0rel_constant_p.
10709         (nios2_cdx_narrow_form_p): Likewise.
10710         * config/nios2/nios2.opt (mr0rel-sec=): New option.
10711         * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10712         (Nios II Options): Document -mr0rel-sec.
10714 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10716         * config/nios2/nios2.c: Include xregex.h.
10717         (nios2_gprel_sec_regex): New.
10718         (nios2_option_overide): Initialize it.  Don't allow GP-relative
10719         addressing with PIC.
10720         (nios2_small_section_name_p): Check for regex match.
10721         * config/nios2/nios2.opt (mgprel-sec=): New option.
10722         * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10723         (Nios II Options): Document -mgprel-sec.
10725 2017-10-26  Jim Wilson  <wilson@tuliptree.org>
10727         * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10729 2017-10-26  Tom de Vries  <tom@codesourcery.com>
10731         PR tree-optimization/82707
10732         * gimple.c (gimple_copy): Fix unsharing of
10733         GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10735 2017-10-26  Olga Makhotina  <olga.makhotina@intel.com>
10737         * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10738         _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10739         _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10740         _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10741         _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10742         _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10743         _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10744         _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10745         _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10746         _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10747         _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10748         _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10749         _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10750         _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10751         _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10752         _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10753         _mm512_mask_cmpunord_ps_mask): New intrinsics.
10755 2017-10-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
10757         * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10758         default to IBM.
10759         * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10760         * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10761         warning to rs6000.c.  Remove the Undocumented flag, since it has
10762         been documented.
10763         (-mabi=ibmlongdouble): Likewise.
10764         * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10765         already set, set the default format for long double.
10766         (rs6000_debug_reg_global): Print whether long double is IBM or
10767         IEEE.
10768         (rs6000_option_override_internal): Rework setting long double
10769         format.  Only warn if the user is changing the long double default
10770         and they did not use -Wno-psabi.
10771         * doc/invoke.texi (PowerPC options): Update the documentation for
10772         -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10774 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10775             Alan Hayward  <alan.hayward@arm.com>
10776             David Sherwood  <david.sherwood@arm.com>
10778         * rtl.h (wider_subreg_mode): New function.
10779         * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10780         rather than an unsigned int *.
10781         * ira-color.c (regno_max_ref_width): Replace with...
10782         (regno_max_ref_mode): ...this new variable.
10783         (coalesced_pseudo_reg_slot_compare): Update accordingly.
10784         Use wider_subreg_mode.
10785         (ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
10786         rather than an unsigned int *.
10787         * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10788         (process_alt_operands): Likewise.
10789         (invariant_p): Likewise.
10790         * lra-spills.c (assign_mem_slot): Likewise.
10791         (add_pseudo_to_slot): Likewise.
10792         * lra.c (collect_non_operand_hard_regs): Likewise.
10793         (add_regs_to_insn_regno_info): Likewise.
10794         * reload1.c (regno_max_ref_width): Replace with...
10795         (regno_max_ref_mode): ...this new variable.
10796         (reload): Update accordingly.  Update call to
10797         ira_sort_regnos_for_alter_reg.
10798         (alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
10799         (init_eliminable_invariants): Update to use regno_max_ref_mode.
10800         (scan_paradoxical_subregs): Likewise.
10802 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10804         * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10805         (aarch64_frame): Add emit_frame_chain boolean.
10806         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10807         Move eh_return case to aarch64_layout_frame.
10808         (aarch64_layout_frame): Initialize emit_frame_chain.
10809         (aarch64_expand_prologue): Use emit_frame_chain.
10811 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10813         * config/aarch64/aarch64.c (aarch64_layout_frame):
10814         Ensure LR is always stored at the bottom of the callee-saves.
10815         Remove rarely used frame layout which saves callee-saves at top of
10816         frame, so the store of LR can be used as a valid probe in all cases.
10818 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10820         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10821         Improve unaligned TImode/TFmode base/offset split.
10823 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10824             Alan Hayward  <alan.hayward@arm.com>
10825             David Sherwood  <david.sherwood@arm.com>
10827         * caller-save.c (mark_referenced_regs):  Use read_modify_subreg_p.
10828         * combine.c (find_single_use_1): Likewise.
10829         (expand_field_assignment): Likewise.
10830         (move_deaths): Likewise.
10831         * lra-constraints.c (simplify_operand_subreg): Likewise.
10832         (curr_insn_transform): Likewise.
10833         * lra.c (collect_non_operand_hard_regs): Likewise.
10834         (add_regs_to_insn_regno_info): Likewise.
10835         * rtlanal.c (reg_referenced_p): Likewise.
10836         (covers_regno_no_parallel_p): Likewise.
10838 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10840         * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10841         Don't print any bits outside the precision of the value.
10842         * wide-int.cc (test_printing): Add some new tests.
10844 2017-10-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10846         * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10847         supports -xbrace_comment option.
10848         * configure: Regenerate.
10849         * config.in: Regenerate.
10850         * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10851         (ASM_CPU_SPEC): Use it.
10853 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10855         * target.def (static_rtx_alignment): New hook.
10856         * targhooks.h (default_static_rtx_alignment): Declare.
10857         * targhooks.c (default_static_rtx_alignment): New function.
10858         * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10859         * doc/tm.texi: Regenerate.
10860         * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10861         instead of targetm.constant_alignment.  Remove call to
10862         set_mem_attributes.
10863         * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10864         (cris_preferred_mininum_alignment): New function, split out from...
10865         (cris_constant_alignment): ...here.
10866         (cris_static_rtx_alignment): New function.
10867         * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10868         split out from...
10869         (ix86_constant_alignment): ...here.
10870         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10871         * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10872         (mmix_static_rtx_alignment): New function.
10873         * config/spu/spu.c (spu_static_rtx_alignment): New function.
10874         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10876 2017-10-26  Tamar Christina  <tamar.christina@arm.com>
10878         PR target/81800
10879         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10880         Add flag_trapping_math and flag_fp_int_builtin_inexact.
10882 2017-10-25  Palmer Dabbelt  <palmer@dabbelt.com>
10884         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10885         * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10886         mark as a sign-extending load.
10887         (local_pic_load_u): Define.
10889 2017-10-25  Eric Botcazou  <ebotcazou@adacore.com>
10891         PR middle-end/82062
10892         * fold-const.c (operand_equal_for_comparison_p): Also return true
10893         if ARG0 is a simple variant of ARG1 with narrower precision.
10894         (fold_ternary_loc): Always pass unstripped operands to the predicate.
10896 2017-10-25  Jan Hubicka  <hubicka@ucw.cz>
10898         * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
10899         cost correctly.
10900         * i386.h (processor_costs): Add gather_static, gather_per_elt,
10901         scatter_static, scatter_per_elt.
10902         * x86-tune-costs.h: Add new cost entries.
10904 2017-10-25  Richard Biener  <rguenther@suse.de>
10906         * tree-ssa-sccvn.h (vn_eliminate): Declare.
10907         * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
10908         class pass_fre): Move to ...
10909         * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
10910         class pass_fre): ... here and adjust for statistics.
10912 2017-10-25  Jakub Jelinek  <jakub@redhat.com>
10914         PR libstdc++/81706
10915         * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
10916         for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
10917         (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
10918         * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
10919         declarations.
10921 2017-10-25  Richard Biener  <rguenther@suse.de>
10923         * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
10924         el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
10925         eliminate_push_avail, eliminate_insert): Move inside...
10926         (class eliminate_dom_walker): ... this class in preparation
10927         of move.
10928         (fini_eliminate): Remove by merging with ...
10929         (eliminate): ... this function.  Adjust for class changes.
10930         (pass_pre::execute): Remove fini_eliminate call.
10931         (pass_fre::execute): Likewise.
10933 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10935         PR target/82460
10936         * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
10937         (VPERMI2, VPERMI2I): New mode iterators.
10938         (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
10939         (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
10940         patterns.
10941         (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
10942         mode iterator.  Remove 3 old define_insn patterns.
10943         (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
10944         (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
10945         VPERMI2 mode iterator, remove the other two expanders.
10946         (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
10947         to use VPERMI2 mode iterator, add another alternative for vpermi2*
10948         instructions, remove the other two patterns.
10949         (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
10950         mode iterator, remove the other two patterns.
10951         * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
10952         (ix86_expand_vec_perm_vpermt2): ... this.  Swap mask and op0
10953         arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
10954         and adjust argument order accordingly.
10955         (ix86_expand_vec_perm): Adjust caller.
10956         (expand_vec_perm_1): Likewise.
10957         (expand_vec_perm_vpermi2_vpshub2): Rename to ...
10958         (expand_vec_perm_vpermt2_vpshub2): ... this.
10959         (ix86_expand_vec_perm_const_1): Adjust caller.
10960         (ix86_vectorize_vec_perm_const_ok): Adjust comments.
10962         PR target/82370
10963         * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
10964         (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
10965         (vec_shl_<mode>): Remove unused expander.
10966         (avx512bw_<shift_insn><mode>3): New define_insn.
10967         (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
10968         (<sse2_avx2>_<shift_insn><mode>3): ... this.  New define_insn.
10970 2017-10-24  Paolo Carlini  <paolo.carlini@oracle.com>
10972         PR c++/82466
10973         * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
10974         description.
10976 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10978         PR rtl-optimization/82396
10979         * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
10980         (autopref_multipass_init): Simplify initialization.
10981         (autopref_rank_data): Simplify sort order.
10982         * gcc/sched-int.h (autopref_multipass_data_): Remove
10983         multi_mem_insn_p, min_offset and max_offset.
10985 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10987         PR middle-end/60580
10988         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10989         Check special value of flag_omit_frame_pointer.
10990         (aarch64_can_eliminate): Likewise.
10991         (aarch64_override_options_after_change_1): Simplify handling of
10992         -fomit-frame-pointer and -fomit-leaf-frame-pointer.
10994 2017-10-24  Richard Biener  <rguenther@suse.de>
10996         PR tree-optimization/82697
10997         * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
10998         zero for conditional load and unconditional store.
11000 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
11002         * doc/install.texi: Document bootstrap-cet.
11004 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
11006         PR target/82659
11007         * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
11008         ENDBR instruction at function entrance if function is only
11009         called directly.
11011 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
11013         PR target/82628
11014         * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
11015         patterns to better describe from which operation the CF is computed.
11016         (addcarry<mode>_0, subborrow<mode>_0): New patterns.
11017         * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
11018         one LTU with [DT]Imode and another one with [SD]Imode.  If arg0
11019         is 0, use _0 suffixed expanders instead of emitting a comparison
11020         before it.
11022 2017-10-06  Sergey Shalnov  <Sergey.Shalnov@intel.com>
11024         * config/i386/i386.md(*movsf_internal, *movdf_internal):
11025         Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
11027 2017-10-24  Eric Botcazou  <ebotcazou@adacore.com>
11029         PR middle-end/82569
11030         * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
11031         * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
11032         * loop-iv.c (iv_get_reaching_def): Likewise.
11033         * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
11034         variable is promoted and the partition contains undefined values.
11036 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11038         * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
11039         reflect reality.
11040         (nios2_address_cost): Define.
11041         (nios2_legitimize_address): Recognize (exp + constant) directly.
11042         (TARGET_ADDRESS_COST): Define.
11044 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11046         * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
11047         (nios2_symbolic_memory_operand_p): Declare.
11048         (nios2_split_large_constant): Declare.
11049         (nios2_split_symbolic_memory_operand): Declare.
11050         * config/nios2/nios2.c: Adjust includes.
11051         (nios2_symbolic_constant_allowed): New.
11052         (nios2_symbolic_constant_p): New.
11053         (nios2_plus_symbolic_constant_p): New.
11054         (nios2_valid_addr_expr_p): Recognize addresses involving
11055         symbolic constants.
11056         (nios2_legitimate_address_p): Likewise, also LO_SUM.
11057         (nios2_symbolic_memory_operand_p): New.
11058         (nios2_large_constant_p): New.
11059         (nios2_split_large_constant): New.
11060         (nios2_split_plus_large_constant): New.
11061         (nios2_split_symbolic_memory_operand): New.
11062         (nios2_legitimize_address): Code refactoring.  Handle addresses
11063         involving symbolic constants.
11064         (nios2_emit_move_sequence): Likewise.
11065         (nios2_print_operand): Improve error output.
11066         (nios2_print_operand_address): Handle LO_SUM.
11067         (nios2_cdx_narrow_form_p): Likewise.
11068         * config/nios2/nios2.md (movqi_internal): Add splitter for memory
11069         operands involving symbolic constants.
11070         (movhi_internal, movsi_internal): Likewise.
11071         (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
11072         (extendhisi2, extendqi<mode>2): Likewise.
11074 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11076         * tree-pass.h (PROP_rtl_split_insns): Define.
11077         * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
11079 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
11081         * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
11083 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11085         PR debug/82630
11086         * target.def (const_not_ok_for_debug_p): Default to
11087         default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
11088         * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
11089         * targhooks.c (default_const_not_ok_for_debug_p): New function.
11090         * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
11091         which targetm.const_not_ok_for_debug_p returned true.
11092         * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
11093         for UNSPECs.
11094         * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
11095         Likewise.
11096         * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
11097         * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
11098         UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
11099         if !base_term_p.
11100         (ix86_const_not_ok_for_debug_p): New function.
11101         (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
11102         (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
11104 2017-10-23  David Malcolm  <dmalcolm@redhat.com>
11106         PR bootstrap/82610
11107         * system.h: Conditionally include "unique-ptr.h" if
11108         INCLUDE_UNIQUE_PTR is defined.
11109         * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
11110         of defining INCLUDE_UNIQUE_PTR before including "system.h".
11112 2017-10-23  Sebastian Perta  <sebastian.perta@renesas.com>
11114         * config/rl78/rl78.md: New define_expand "subdi3".
11116 2017-10-23  H.J. Lu  <hongjiu.lu@intel.com>
11118         PR target/82673
11119         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
11120         DF_REF_INSN if DF_REF_INSN_INFO is false.
11122 2017-10-23  Jan Hubicka  <hubicka@ucw.cz>
11124         * i386.c (dimode_scalar_chain::compute_convert_gain): Use
11125         xmm_move instead of sse_move.
11126         (sse_store_index): New function.
11127         (ix86_register_move_cost): Be more sensible about mismatch stall;
11128         model AVX moves correctly; make difference between sse->integer and
11129         integer->sse.
11130         (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
11131         moves; make difference between SSE and AVX.
11132         * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
11133         and zmm_move. Increase size of sse load and store tables;
11134         add unaligned load and store tables; add ssemmx_to_integer.
11135         * x86-tune-costs.h: Update all entries according to real
11136         move latencies from Agner Fog's manual and chip documentation.
11138 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11140         PR target/82628
11141         * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
11142         * config/i386/constraints.md (Wf): New constraint.
11143         * config/i386/i386.md (UNSPEC_SBB): New unspec.
11144         (cmp<dwi>_doubleword): Removed.
11145         (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
11146         (sub<mode>3_carry_ccgz): Use unspec instead of compare.
11147         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
11148         expand with cmp<dwi>_doubleword.  For LTU and GEU use
11149         sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
11151         * common.opt (gcolumn-info): Enable by default.
11152         * doc/invoke.texi (gcolumn-info): Document new default.
11154 2017-10-23  Richard Biener  <rguenther@suse.de>
11156         PR tree-optimization/82672
11157         * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
11158         Fold the stmt if we propagated into it.
11160 2017-10-23  Richard Biener  <rguenther@suse.de>
11162         * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
11163         (bitmap_remove_expr_from_set): ... this.  All callers call this
11164         for non-constant values.
11165         (bitmap_set_subtract): Rename to...
11166         (bitmap_set_subtract_expressions): ... this.  Adjust and
11167         optimize.
11168         (bitmap_set_contains_value): Remove superfluous check.
11169         (bitmap_set_replace_value): Inline into single caller ...
11170         (bitmap_value_replace_in_set): ... here and simplify.
11171         (dependent_clean): Merge into ...
11172         (clean): ... this using an overload.  Adjust.
11173         (prune_clobbered_mems): Adjust.
11174         (compute_antic_aux): Likewise.
11175         (compute_partial_antic_aux): Likewise.
11177 2017-10-23  Richard Biener  <rguenther@suse.de>
11179         PR tree-optimization/82129
11180         Revert
11181         2017-08-01  Richard Biener  <rguenther@suse.de>
11183         PR tree-optimization/81181
11184         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
11185         (compute_antic): ... end of iteration here.
11187 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11189         * target.def (starting_frame_offset): New hook.
11190         * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
11191         (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
11192         * doc/tm.texi.in: Regenerate.
11193         * hooks.h (hook_hwi_void_0): Declare.
11194         * hooks.c (hook_hwi_void_0): New function.
11195         * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
11196         STARTING_FRAME_OFFSET.
11197         * builtins.c (expand_builtin_setjmp_receiver): Likewise.
11198         * reload1.c (reload): Likewise.
11199         * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
11200         instead of STARTING_FRAME_OFFSET.
11201         * function.c (try_fit_stack_local): Likewise.
11202         (assign_stack_local_1): Likewise
11203         (instantiate_virtual_regs): Likewise.
11204         * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
11205         * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
11206         * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
11207         * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
11208         * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
11209         * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
11210         * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
11211         * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
11212         * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
11213         * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
11214         * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
11215         * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
11216         * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
11217         * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
11218         * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
11219         * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
11220         * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
11221         * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
11222         * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
11223         * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11224         * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11225         * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11226         * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11227         * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11228         * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11229         * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11230         * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11231         * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11232         * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11233         * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11234         * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11235         * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11236         * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11237         * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11238         * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11239         * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11240         * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11241         * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11242         * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11243         * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11244         * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11245         * config/avr/avr.c (avr_starting_frame_offset): Make static and
11246         return a HOST_WIDE_INT.
11247         (avr_builtin_setjmp_frame_value): Use it instead of
11248         STARTING_FRAME_OFFSET.
11249         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11250         * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11251         * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11252         New function.
11253         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11254         * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11255         * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11256         (TARGET_CONSTANT_ALIGNMENT): Redefine.
11257         * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11258         * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11259         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11260         * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11261         * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11262         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11263         * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11264         * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11265         New function.
11266         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11267         * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11268         * config/mips/mips.c (mips_compute_frame_info): Refer to
11269         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11270         (mips_starting_frame_offset): New function.
11271         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11272         * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11273         * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11274         * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11275         and return a HOST_WIDE_INT.
11276         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11277         (mmix_initial_elimination_offset): Refer to
11278         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11279         * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11280         * config/pa/pa.c (pa_starting_frame_offset): New function.
11281         (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11282         (pa_expand_prologue): Likewise.
11283         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11284         * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11285         !FRAME_GROWS_DOWNWARD handling to...
11286         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11287         * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11288         !FRAME_GROWS_DOWNWARD handling to...
11289         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11290         * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11291         !FRAME_GROWS_DOWNWARD handling to...
11292         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11293         * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11294         Redefine.
11295         (rs6000_starting_frame_offset): New function.
11296         * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11297         !FRAME_GROWS_DOWNWARD handling to...
11298         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11299         * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11300         !FRAME_GROWS_DOWNWARD handling to...
11301         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11302         * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11303         !FRAME_GROWS_DOWNWARD handling to...
11304         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11305         * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11306         (rs6000_starting_frame_offset): New function.
11307         * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11308         * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11309         * config/vax/vax.c (vax_starting_frame_offset): New function.
11310         (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11311         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11312         * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11313         * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11314         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11315         * system.h (STARTING_FRAME_OFFSET): Poison.
11317 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11319         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11320         SCALAR_TYPE_MODE instead of TYPE_MODE.
11322 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11323             Alan Hayward  <alan.hayward@arm.com>
11324             David Sherwood  <david.sherwood@arm.com>
11326         * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_TYPE_MODE
11328 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11329             Alan Hayward  <alan.hayward@arm.com>
11330             David Sherwood  <david.sherwood@arm.com>
11332         * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11334 2017-10-23  Richard Biener  <rguenther@suse.de>
11336         PR tree-optimization/82129
11337         * tree-ssa-pre.c (bitmap_set_and): Remove.
11338         (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11339         canonicalizing expressions in the set to those with lowest
11340         ID rather than taking that from the first edge.
11342 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11344         * combine.c (rtx_equal_for_field_assignment_p): Use
11345         byte_lowpart_offset.
11347 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11348             Alan Hayward  <alan.hayward@arm.com>
11349             David Sherwood  <david.sherwood@arm.com>
11351         * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11352         to a SUBREG_PROMOTED_VAR.
11354 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11355             Alan Hayward  <alan.hayward@arm.com>
11356             David Sherwood  <david.sherwood@arm.com>
11358         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11359         (expand_debug_source_expr): Likewise.
11360         * combine.c (combine_simplify_rtx): Likewise.
11361         * cse.c (fold_rtx): Likewise.
11362         * optabs.c (expand_float): Likewise.
11363         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11364         (simplify_binary_operation_1): Likewise.
11366 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11367             Alan Hayward  <alan.hayward@arm.com>
11368             David Sherwood  <david.sherwood@arm.com>
11370         * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11371         (record_promoted_value): Likewise.
11372         * expr.c (expand_expr_real_2): Likewise.
11373         * ree.c (update_reg_equal_equiv_notes): Likewise.
11374         (combine_set_extension): Likewise.
11375         * rtlanal.c (low_bitmask_len): Likewise.
11376         * simplify-rtx.c (neg_const_int): Likewise.
11377         (simplify_binary_operation_1): Likewise.
11379 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11380             Alan Hayward  <alan.hayward@arm.com>
11381             David Sherwood  <david.sherwood@arm.com>
11383         * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11384         * regcprop.c (maybe_mode_change): Likewise.
11385         * reload1.c (alter_reg): Likewise.
11387 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11389         * inchash.h (inchash::hash::add_wide_int): New function.
11390         * lto-streamer-out.c (hash_tree): Use it.
11392 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11394         * inchash.h (inchash::hash::add_wide_int): Rename to...
11395         (inchash::hash::add_hwi): ...this.
11396         * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11397         (polymorphic_call_target_hasher::hash): Likewise.
11398         * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11399         (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11400         (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11401         * lto-streamer-out.c (hash_tree): Likewise.
11402         * optc-save-gen.awk: Likewise.
11403         * tree.c (add_expr): Likewise.
11405 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11407         PR target/52451
11408         * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11409         for ordered inequality comparisons even with TARGET_IEEE_FP.
11411 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11413         PR target/82628
11414         * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11415         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11416         Expand with cmp<dwi>_doubleword.
11418 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11420         * extend.texi: Add x86 specific to 'nocf_check' attribute.
11421         List CET intrinsics.
11422         * invoke.texi: Add -mcet, -mibt, -mshstk options.  Add x86
11423         specific to -fcf-protection option.
11425 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11427         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11428         (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11429         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11430         (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11431         (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11432         * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11433         (extra_objs): Add cet.o for Linux/x86 targets.
11434         (tmake_file): Add i386/t-cet for Linux/x86 targets.
11435         * config/i386/cet.c: New file.
11436         * config/i386/cetintrin.h: Likewise.
11437         * config/i386/t-cet: Likewise.
11438         * config/i386/cpuid.h (bit_SHSTK): New.
11439         (bit_IBT): Likewise.
11440         * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11441         pass IBT and SHSTK bits.
11442         * config/i386/i386-builtin-types.def
11443         (VOID_FTYPE_UNSIGNED_PVOID): New.
11444         (VOID_FTYPE_UINT64_PVOID): Likewise.
11445         * config/i386/i386-builtin.def: Add CET intrinsics.
11446         * config/i386/i386-c.c (ix86_target_macros_internal): Add
11447         OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11448         * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11449         * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11450         prototype.
11451         * config/i386/i386.c (rest_of_insert_endbranch): New.
11452         (pass_data_insert_endbranch): Likewise.
11453         (pass_insert_endbranch): Likewise.
11454         (make_pass_insert_endbranch): Likewise.
11455         (ix86_notrack_prefixed_insn_p): Likewise.
11456         (ix86_target_string): Add -mibt, -mshstk flags.
11457         (ix86_option_override_internal): Add flag_cf_protection
11458         processing.
11459         (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11460         (ix86_print_operand): Add 'notrack' prefix output.
11461         (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11462         (ix86_expand_builtin): Expand CET intrinsics.
11463         (x86_output_mi_thunk): Add 'endbranch' instruction.
11464         * config/i386/i386.h (TARGET_IBT): New.
11465         (TARGET_IBT_P): Likewise.
11466         (TARGET_SHSTK): Likewise.
11467         (TARGET_SHSTK_P): Likewise.
11468         * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11469         UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11470         UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11471         (builtin_setjmp_setup): New pattern.
11472         (builtin_longjmp): Likewise.
11473         (rdssp<mode>): Likewise.
11474         (incssp<mode>): Likewise.
11475         (saveprevssp): Likewise.
11476         (rstorssp): Likewise.
11477         (wrss<mode>): Likewise.
11478         (wruss<mode>): Likewise.
11479         (setssbsy): Likewise.
11480         (clrssbsy): Likewise.
11481         (nop_endbr): Likewise.
11482         * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11483         options.
11484         * config/i386/immintrin.h: Include <cetintrin.h>.
11485         * config/i386/linux-common.h
11486         (file_end_indicate_exec_stack_and_cet): New prototype.
11487         (TARGET_ASM_FILE_END): New.
11489 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11491         * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11492         latencies instead of having separate table; make difference between
11493         integer and float costs.
11494         * i386.h (processor_costs): Remove scalar_stmt_cost,
11495         scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11496         scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11497         vec_store_cost.
11498         * x86-tune-costs.h: Remove entries which has been removed in
11499         procesor_costs from all tables; make cond_taken_branch_cost
11500         and cond_not_taken_branch_cost COST_N_INSNS based.
11502 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11504         * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11506 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11508         * config/i386/i386.md (isa): Remove fma_avx512f.
11509         * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11510         <avx512>_fmadd_<mode>_mask3<round_name>,
11511         <avx512>_fmsub_<mode>_mask<round_name>,
11512         <avx512>_fmsub_<mode>_mask3<round_name>,
11513         <avx512>_fnmadd_<mode>_mask<round_name>,
11514         <avx512>_fnmadd_<mode>_mask3<round_name>,
11515         <avx512>_fnmsub_<mode>_mask<round_name>,
11516         <avx512>_fnmsub_<mode>_mask3<round_name>,
11517         <avx512>_fmaddsub_<mode>_mask<round_name>,
11518         <avx512>_fmaddsub_<mode>_mask3<round_name>,
11519         <avx512>_fmsubadd_<mode>_mask<round_name>,
11520         <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11521         (*vec_widen_umult_even_v16si<mask_name>,
11522         *vec_widen_smult_even_v16si<mask_name>): Likewise.
11523         (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11525 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11527         * extend.texi: Add 'nocf_check' documentation.
11528         * gimple.texi: Add second parameter to
11529         gimple_build_call_from_tree.
11530         * invoke.texi: Add -fcf-protection documentation.
11531         * rtl.texi: Add REG_CALL_NOTRACK documenation.
11533 2017-10-20  Richard Biener  <rguenther@suse.de>
11535         PR tree-optimization/82473
11536         * tree-vect-loop.c (vectorizable_reduction): Properly get at
11537         the largest input type.
11539 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11541         * c-attribs.c (handle_nocf_check_attribute): New function.
11542         (c_common_attribute_table): Add 'nocf_check' handling.
11543         * gimple-parser.c: Add second argument NULL to
11544         gimple_build_call_from_tree.
11545         * attrib.c (comp_type_attributes): Check nocf_check attribute.
11546         * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11547         call insn.
11548         * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11549         * common.opt: Add fcf-protection flag.
11550         * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11551         * flag-types.h: Add enum cf_protection_level.
11552         * gimple.c (gimple_build_call_from_tree): Add second parameter.
11553         Add 'nocf_check' attribute propagation to gimple call.
11554         * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11555         (gimple_build_call_from_tree): Update prototype.
11556         (gimple_call_nocf_check_p): New function.
11557         (gimple_call_set_nocf_check): Likewise.
11558         * gimplify.c: Add second argument to gimple_build_call_from_tree.
11559         * ipa-icf.c: Add nocf_check attribute in statement hash.
11560         * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11561         * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11562         * toplev.c (process_options): Add flag_cf_protection handling.
11564 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11566         * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11568 2017-10-20  Richard Biener  <rguenther@suse.de>
11570         PR tree-optimization/82603
11571         * tree-if-conv.c (predicate_mem_writes): Make sure to only
11572         remove false predicated stores.
11574 2017-10-20  Richard Biener  <rguenther@suse.de>
11576         * graphite-isl-ast-to-gimple.c
11577         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11578         Remove return value and simplify, dump copied stmt after lhs
11579         adjustment.
11580         (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11581         Reduce dump verbosity.
11582         (gsi_insert_earliest): Likewise.
11583         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11584         * graphite.c (print_global_statistics): Adjust dumping.
11585         (print_graphite_scop_statistics): Likewise.
11586         (print_graphite_statistics): Do not dump loops here.
11587         (graphite_transform_loops): But here.
11589 2017-10-20  Nicolas Roche  <roche@adacore.com>
11591         * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11592         * configure: Regenerate.
11594 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11596         PR target/82158
11597         * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11598         functions when optimizing replace GIMPLE_RETURN stmts with
11599         calls to __builtin_unreachable ().
11601         PR sanitizer/82595
11602         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11603         for -fsanitize=thread link of executables.
11604         (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11605         link of executables.
11607         PR target/82370
11608         * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11609         New mode iterators.
11610         (<shift_insn><mode>3<mask_name>): Change the last of the 3
11611         define_insns for logical vector shifts to use VI248_AVX512BW
11612         iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11613         condition, useless isa and prefix attributes.  Change the first
11614         2 of these define_insns to ...
11615         (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11616         define_insn for avx512vl.
11617         (<shift_insn><mode>3): ... and this, new define_insn without
11618         masking for non-avx512vl.
11620         PR target/82370
11621         * config/i386/sse.md (*andnot<mode>3,
11622         <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11623         (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11624         and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11625         not applied use empty suffix even for TARGET_AVX512VL.
11626         * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11627         is applied, supply evex,evex or evex,evex,evex instead of just
11628         evex.
11630 2017-10-20  Julia Koval  <julia.koval@intel.com>
11632         * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11633         (OPTION_MASK_ISA_GFNI_UNSET): New.
11634         (ix86_handle_option): Handle OPT_mgfni.
11635         * config/i386/cpuid.h (bit_GFNI): New.
11636         * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11637         * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11638         * config/i386/i386.c (ix86_target_string): Add -mgfni.
11639         (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11640         * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11641         * config/i386/i386.opt: Add mgfni.
11643 2017-10-20  Orlando Arias  <oarias@knights.ucf.edu>
11645         * config/msp430/msp430.c (msp430_option_override): Disable
11646         -fdelete-null-pointer-checks.
11647         * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11649 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11651         * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11652         of x87 and SSE instructions.
11654 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11656         * asan.c (create_cond_insert_point): Do not update edge count.
11657         * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11658         (afdo_propagate_circuit): Likewise.
11659         (afdo_calculate_branch_prob): Likewise.
11660         (afdo_annotate_cfg): Likewise.
11661         * basic-block.h (struct edge_def): Remove count.
11662         (edge_def::count): New accessor.
11663         * bb-reorder.c (rotate_loop): Update.
11664         (find_traces_1_round): Update.
11665         (connect_traces): Update.
11666         (sanitize_hot_paths): Update.
11667         * cfg.c (unchecked_make_edge): Update.
11668         (make_single_succ_edge): Update.
11669         (check_bb_profile): Update.
11670         (dump_edge_info): Update.
11671         (update_bb_profile_for_threading): Update.
11672         (scale_bbs_frequencies_int): Update.
11673         (scale_bbs_frequencies_gcov_type): Update.
11674         (scale_bbs_frequencies_profile_count): Update.
11675         (scale_bbs_frequencies): Update.
11676         * cfganal.c (connect_infinite_loops_to_exit): Update.
11677         * cfgbuild.c (compute_outgoing_frequencies): Update.
11678         (find_many_sub_basic_blocks): Update.
11679         * cfgcleanup.c (try_forward_edges): Update.
11680         (try_crossjump_to_edge): Update
11681         * cfgexpand.c (expand_gimple_cond): Update
11682         (expand_gimple_tailcall): Update
11683         (construct_exit_block): Update
11684         * cfghooks.c (verify_flow_info): Update
11685         (redirect_edge_succ_nodup): Update
11686         (split_edge): Update
11687         (make_forwarder_block): Update
11688         (duplicate_block): Update
11689         (account_profile_record): Update
11690         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11691         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11692         * cfgloopmanip.c (scale_loop_profile): Update.
11693         (loopify): Update.
11694         (lv_adjust_loop_entry_edge): Update.
11695         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11696         (force_nonfallthru_and_redirect): Update.
11697         (purge_dead_edges): Update.
11698         (rtl_flow_call_edges_add): Update.
11699         * cgraphunit.c (init_lowered_empty_function): Update.
11700         (cgraph_node::expand_thunk): Update.
11701         * gimple-pretty-print.c (dump_probability): Update.
11702         (dump_edge_probability): Update.
11703         * gimple-ssa-isolate-paths.c (isolate_path): Update.
11704         * haifa-sched.c (sched_create_recovery_edges): Update.
11705         * hsa-gen.c (convert_switch_statements): Update.
11706         * ifcvt.c (dead_or_predicable): Update.
11707         * ipa-inline-transform.c (inline_transform): Update.
11708         * ipa-split.c (split_function): Update.
11709         * ipa-utils.c (ipa_merge_profiles): Update.
11710         * loop-doloop.c (add_test): Update.
11711         * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11712         * lto-streamer-in.c (input_cfg): Update.
11713         (input_function): Update.
11714         * lto-streamer-out.c (output_cfg): Update.
11715         * modulo-sched.c (sms_schedule): Update.
11716         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11717         * predict.c (maybe_hot_edge_p): Update.
11718         (unlikely_executed_edge_p): Update.
11719         (probably_never_executed_edge_p): Update.
11720         (dump_prediction): Update.
11721         (drop_profile): Update.
11722         (propagate_unlikely_bbs_forward): Update.
11723         (determine_unlikely_bbs): Update.
11724         (force_edge_cold): Update.
11725         * profile.c (compute_branch_probabilities): Update.
11726         * reg-stack.c (better_edge): Update.
11727         * shrink-wrap.c (handle_simple_exit): Update.
11728         * tracer.c (better_p): Update.
11729         * trans-mem.c (expand_transaction): Update.
11730         (split_bb_make_tm_edge): Update.
11731         * tree-call-cdce.c: Update.
11732         * tree-cfg.c (gimple_find_sub_bbs): Update.
11733         (gimple_split_edge): Update.
11734         (gimple_duplicate_sese_region): Update.
11735         (gimple_duplicate_sese_tail): Update.
11736         (gimple_flow_call_edges_add): Update.
11737         (insert_cond_bb): Update.
11738         (execute_fixup_cfg): Update.
11739         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11740         * tree-complex.c (expand_complex_div_wide): Update.
11741         * tree-eh.c (lower_resx): Update.
11742         (unsplit_eh): Update.
11743         (cleanup_empty_eh_move_lp): Update.
11744         * tree-inline.c (copy_edges_for_bb): Update.
11745         (freqs_to_counts): Update.
11746         (copy_cfg_body): Update.
11747         * tree-ssa-dce.c (remove_dead_stmt): Update.
11748         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11749         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11750         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11751         (unloop_loops): Update.
11752         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11753         * tree-ssa-loop-split.c (connect_loops): Update.
11754         (split_loop): Update.
11755         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11756         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11757         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11758         * tree-ssa-reassoc.c (branch_fixup): Update.
11759         * tree-ssa-tail-merge.c (replace_block_by): Update.
11760         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11761         (compute_path_counts): Update.
11762         (update_profile): Update.
11763         (recompute_probabilities): Update.
11764         (update_joiner_offpath_counts): Update.
11765         (estimated_freqs_path): Update.
11766         (freqs_to_counts_path): Update.
11767         (clear_counts_path): Update.
11768         (ssa_fix_duplicate_block_edges): Update.
11769         (duplicate_thread_path): Update.
11770         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11771         (case_bit_test_cmp): Update.
11772         (collect_switch_conv_info): Update.
11773         (gen_inbound_check): Update.
11774         (do_jump_if_equal): Update.
11775         (emit_cmp_and_jump_insns): Update.
11776         * tree-tailcall.c (decrease_profile): Update.
11777         (eliminate_tail_call): Update.
11778         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11779         (vect_do_peeling): Update.
11780         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11781         * ubsan.c (ubsan_expand_null_ifn): Update.
11782         (ubsan_expand_ptr_ifn): Update.
11783         * value-prof.c (gimple_divmod_fixed_value): Update.
11784         (gimple_mod_pow2): Update.
11785         (gimple_mod_subtract): Update.
11786         (gimple_ic): Update.
11787         (gimple_stringop_fixed_value): Update.
11789 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11791         PR target/82618
11792         * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11794 2017-10-19  Alexander Monakov  <amonakov@ispras.ru>
11796         PR rtl-optimization/82395
11797         * ira-color.c (allocno_priority_compare_func): Fix comparison step
11798         based on non_spilled_static_chain_regno_p.
11800 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11802         * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11803         (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11804         (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11806 2017-10-19  Martin Sebor  <msebor@redhat.com>
11808         PR tree-optimization/82596
11809         * tree.c (array_at_struct_end_p): Handle STRING_CST.
11811 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11813         * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11814         (get_mem_refs_of_builtin_call): Likewise.
11815         * builtins.c (expand_builtin_apply): Adjust call to
11816         allocate_dynamic_stack_space.
11817         (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11818         the third argument to allocate_dynamic_stack_space, otherwise -1.
11819         (expand_builtin): Deal with all alloca variants.
11820         (is_inexpensive_builtin): Likewise.
11821         * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11822         * calls.c (special_function_p): Deal with all alloca variants.
11823         (initialize_argument_information): Adjust call to
11824         allocate_dynamic_stack_space.
11825         (expand_call): Likewise.
11826         * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11827         * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11828         * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11829         use it for the stack usage computation.
11830         * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11831         * function.c (gimplify_parameters): Call build_alloca_call_expr.
11832         * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11833         Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11834         (in_loop_p): Remove first argument and useless check.
11835         (pass_walloca::execute): Remove useless test and adjust call to above.
11836         * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11837         * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11838         (gimplify_call_expr): Deal with all alloca variants.
11839         * hsa-gen.c (gen_hsa_alloca): Likewise.
11840         (gen_hsa_insns_for_call): Likewise.
11841         * ipa-pure-const.c (special_builtin_state): Likewise.
11842         * tree-chkp.c (chkp_build_returned_bound): Likewise.
11843         * tree-object-size.c (alloc_object_size): Likewise.
11844         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11845         (call_may_clobber_ref_p_1): Likewise.
11846         * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11847         (ccp_fold_stmt): Likewise.
11848         (optimize_stack_restore): Likewise.
11849         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11850         (mark_all_reaching_defs_necessary_1): Likewise.
11851         (propagate_necessity): Likewise.
11852         (eliminate_unnecessary_stmts): Likewise.
11853         * tree.c (build_common_builtin_nodes): Build
11854         BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11855         (build_alloca_call_expr): New function.
11856         * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11857         (CASE_BUILT_IN_ALLOCA): Likewise.
11858         (build_alloca_call_expr): Declare.
11859         * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11861 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11863         PR debug/82509
11864         * dwarf2out.c (new_die_raw): New static inline function.
11865         (new_die): Use it to create the DIE.
11866         (add_AT_external_die_ref): Likewise.
11867         (clone_die): Likewise.
11868         (clone_as_declaration): Likewise.
11869         (dwarf2out_vms_debug_main_pointer): Likewise.
11870         (base_type_die): Likewise.  Remove early return for corner cases.
11871         Do not call add_pubtype on the DIE here.
11872         (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11873         (modified_type_die): Adjust the lookup for reverse order DIEs.  Skip
11874         typedefs for base types with DW_AT_endianity.  Make sure a DIE with
11875         native order exists for base types, attach the DIE manually and call
11876         add_pubtype on it.  Do not equate a reverse order DIE to the type.
11878 2017-10-19  Richard Earnshaw  <rearnsha@arm.com>
11880         * config/arm/arm.c (align_ok_ldrd_strd): New function.
11881         (mem_ok_for_ldrd_strd): New parameter align.  Extract the alignment of
11882         the mem into it.
11883         (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11885 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11887         * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN.  Or
11888         SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11889         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11890         BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11891         * opts.c (sanitizer_opts): Add builtin.
11892         * ubsan.c (instrument_builtin): New function.
11893         (pass_ubsan::execute): Call it.
11894         (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11895         * doc/invoke.texi: Document -fsanitize=builtin.
11897         * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
11898         builtins, store max (log2 (align), 0) into uchar field instead of
11899         align into uptr field.
11900         (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
11901         store uchar 0 field instead of uptr 0 field.
11902         (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
11903         instead of passing one address of struct with 2 locations pass
11904         two addresses of structs with 1 location each.
11905         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
11906         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
11907         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
11908         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
11909         (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
11910         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
11911         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
11912         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
11914 2017-10-19  Martin Liska  <mliska@suse.cz>
11916         PR driver/81829
11917         * file-find.c (remove_prefix): Remove.
11918         * file-find.h (remove_prefix): Likewise.
11919         * gcc-ar.c: Remove smartness of lookup.
11921 2017-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
11923         * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
11924         *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
11925         *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
11927 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11929         PR target/82580
11930         * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
11931         (setcc + and to xor + setcc): New peephole2.
11933 2017-10-19  Tom de Vries  <tom@codesourcery.com>
11935         * doc/sourcebuild.texi (Test Directives, Variants of
11936         dg-require-support): Add dg-require-stack-size.
11938 2017-10-19  Martin Liska  <mliska@suse.cz>
11940         PR sanitizer/82517
11941         * gimplify.c (gimplify_decl_expr): Do not instrument variables
11942         that have a large alignment.
11943         (gimplify_target_expr): Likewise.
11945 2017-10-18  Segher Boessenkool  <segher@kernel.crashing.org>
11947         PR rtl-optimization/82602
11948         * ira.c (rtx_moveable_p): Return false for volatile asm.
11950 2017-10-18  Uros Bizjak  <ubizjak@gmail.com>
11952         PR target/82580
11953         * config/i386/i386-modes.def (CCGZ): New CC mode.
11954         * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
11955         * config/i386/predicates.md (ix86_comparison_operator):
11956         Handle CCGZmode.
11957         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11958         Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
11959         with double-word subtraction.
11960         (put_condition_code): Handle CCGZmode.
11962 2017-10-18  Aldy Hernandez  <aldyh@redhat.com>
11964         * wide-int.cc (debug (const wide_int &)): New.
11965         (debug (const wide_int *)): New.
11966         (debug (const widest_int &)): New.
11967         (debug (const widest_int *)): New.
11969 2017-10-18  Vladimir Makarov  <vmakarov@redhat.com>
11971         PR middle-end/82556
11972         * lra-constraints.c (curr_insn_transform): Use non-input operand
11973         instead of output one for matched reload.
11975 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11977         * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
11978         (tree-ssa-loop-ivopts.h): New header file.
11979         (struct builtin_info): New fields.
11980         (classify_builtin_1): Compute and record base and offset parts for
11981         memset builtin partition by calling strip_offset.
11982         (offset_cmp, fuse_memset_builtins): New functions.
11983         (finalize_partitions): Fuse adjacent memset partitions by calling
11984         above function.
11985         * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
11986         Expose the interface.
11987         * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
11989 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11991         PR tree-optimization/82574
11992         * tree-loop-distribution.c (find_single_drs): New parameter.  Check
11993         that data reference must be executed exactly once per iteration
11994         against the outermost loop in nest.
11995         (classify_partition): Update call to above function.
11997 2017-10-18  Richard Biener  <rguenther@suse.de>
11999         PR tree-optimization/82591
12000         * graphite.c (graphite_transform_loops): Move code gen message
12001         printing ...
12002         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12003         Here.  Handle scop_to_isl_ast failing.
12004         (scop_to_isl_ast): Limit the number of ISL operations.
12006 2017-10-18  Richard Biener  <rguenther@suse.de>
12008         * graphite-isl-ast-to-gimple.c
12009         (translate_isl_ast_to_gimple::set_rename): Simplify.
12010         (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
12011         (graphite_copy_stmts_from_block): ... here.
12012         (copy_bb_and_scalar_dependences): Simplify.
12013         (add_parameters_to_ivs_params): Canonicalize.
12014         (generate_entry_out_of_ssa_copies): Simplify.
12015         * graphite-sese-to-poly.c (extract_affine_name): Simplify
12016         by passing in ISL dimension.
12017         (parameter_index_in_region_1): Rename to ...
12018         (parameter_index_in_region): ... this.
12019         (extract_affine): Adjust assert, pass down parameter index.
12020         (add_param_constraints): Use range-info when available.
12021         (build_scop_context): Adjust.
12022         * sese.c (new_sese_info): Adjust.
12023         (free_sese_info): Likewise.
12024         * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
12025         Remove unused typedefs.
12026         (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
12028 2017-10-18  Martin Liska  <mliska@suse.cz>
12030         * combine.c (simplify_compare_const): Add gcc_fallthrough.
12032 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12034         * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
12035         (s390_sched_init): Do not reset s390_sched_state if we entered the
12036         current basic block via a fallthru edge and all others are unlikely.
12038 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
12040         * config/s390/s390.c (NUM_SIDES): New variable.
12041         (LONGRUNNING_THRESHOLD): New variable.
12042         (LATENCY_FACTOR): New variable.
12043         (s390_sched_score): Decrease score for long-running instructions on
12044         wrong side.
12045         (s390_sched_variable_issue): Perform bookkeeping for long-running
12046         instructions.
12048 2017-10-18  Richard Biener  <rguenther@suse.de>
12050         * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
12051         Simplify with removal of the parameter rename map.
12052         (set_rename): Likewise.
12053         (should_copy_to_new_region): Likewise.
12054         (graphite_copy_stmts_from_block): Likewise.
12055         (copy_bb_and_scalar_dependences): Remove initialization of
12056         unused copied_bb_map.
12057         (copy_def): Remove.
12058         (copy_internal_parameters): Likewise.
12059         (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
12060         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
12061         Use INTEGRAL_TYPE_P.
12062         (parameter_index_in_region_1): Rename to ...
12063         (assign_parameter_index_in_region): ... this.  Assert we have
12064         a parameter we handle.
12065         (scan_tree_for_params): Adjust.
12066         * sese.h (parameter_rename_map_t): Remove.
12067         (struct sese_info_t): Remove unused parameter_rename_map and
12068         copied_bb_map members.
12069         * sese.c (new_sese_info): Adjust.
12070         (free_sese_info): Likewise.
12072 2017-10-18  Martin Liska  <mliska@suse.cz>
12074         PR sanitizer/82545
12075         * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
12076         on an abnormal edge.
12078 2017-10-18  Sebastian Huber  <sebastian.huber@embedded-brains.de>
12080         * doc/invoke.texi (ffunction-sections and fdata-sections):
12081         Update.
12083 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12085         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
12086         the use statement can throw internally.
12088 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12090         * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
12091         any RTX present on the RHS of a SET.
12092         * compare-elim.c (try_eliminate_compare): Restore comment.
12094 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12096         * langhooks.h (struct lang_hooks): Document that tree_size langhook
12097         may be also called on tcc_type nodes.
12098         * langhooks.c (lhd_tree_size): Likewise.
12100 2017-10-17  David Malcolm  <dmalcolm@redhat.com>
12102         * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
12103         format_warning_at_substring.
12104         (maybe_warn): Convert source_range * param to a location_t.  Pass
12105         UNKNOWN_LOCATION rather than NULL to fmtwarn.
12106         (format_directive): Remove code to extract source_ranges and
12107         source_range * in favor of just a location_t.
12108         (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
12109         fmtwarn.
12110         * substring-locations.c (format_warning_va): Convert
12111         source_range * param to a location_t.
12112         (format_warning_at_substring): Likewise.
12113         * substring-locations.h (format_warning_va): Likewise.
12114         (format_warning_at_substring): Likewise.
12116 2017-10-17  Jan Hubicka  <hubicka@ucw.cz>
12118         * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
12119         vec_scatter_store
12120         * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
12121         and scatter/gather ops.
12123         * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
12124         vec_gather_load and vec_scatter_store.
12125         * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
12126         * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
12127         * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
12128         * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
12129         * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
12130         * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
12132 2017-10-17  Uros Bizjak  <ubizjak@gmail.com>
12134         * reg-stack.c (compare_for_stack_reg): Add bool argument.
12135         Detect FTST instruction and handle its register pops.  Only pop
12136         second operand if can_pop_second_op is true.
12137         (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
12138         set can_pop_second_op to false in the compare_for_stack_reg call.
12140         * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
12141         output_fp_compare for stack register operands.
12142         * config/i386/i386.c (output_fp_compare): Do not output SSE compare
12143         instructions here.  Do not emit stack register pops here.  Assert
12144         that FCOMPP pops next to top stack register.  Rewrite function.
12146 2017-10-17  Nathan Sidwell  <nathan@acm.org>
12148         PR middle-end/82577
12149         * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
12150         use DECL_ASSEMBLER_NAME_RAW.
12152         PR middle-end/82546
12153         * tree.c (tree_code_size): Reformat.  Punt to lang hook for unknown
12154         TYPE nodes.
12156 2017-10-17  Qing Zhao <qing.zhao@oracle.com>
12157             Wilco Dijkstra <wilco.dijkstra@arm.com>
12159         * builtins.c (expand_builtin_update_setjmp_buf): Add a
12160         converstion to Pmode from the buf_addr.
12162 2017-10-17  Richard Biener  <rguenther@suse.de>
12164         * graphite-dependences.c (scop_get_reads_and_writes): Change
12165         output parameters to references.
12167 2017-10-17  Jackson Woodruff  <jackson.woodruff@arm.com>
12169         PR 71026/tree-optimization
12170         * fold-const.c (distribute_real_division): Removed.
12171         (fold_binary_loc): Remove calls to distribute_real_divison.
12173 2017-10-17  Richard Biener  <rguenther@suse.de>
12175         * graphite-scop-detection.c
12176         (scop_detection::stmt_has_simple_data_refs_p): Always use
12177         the full nest as region.
12178         (try_generate_gimple_bb): Likewise.
12179         * sese.c (scalar_evolution_in_region): Simplify now that
12180         SCEV can handle instantiation in regions.
12181         * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
12182         in the non-loop part of a function if requested.
12184 2017-10-17  Richard Biener  <rguenther@suse.de>
12186         PR tree-optimization/82563
12187         * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
12188         New function.
12189         (graphite_regenerate_ast_isl): Call it.
12190         * graphite-scop-detection.c (build_scops): Remove entry edge split.
12192 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12194         PR tree-optimization/82549
12195         * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
12196         Formatting fixes.  Instead of calling make_bit_field_ref with negative
12197         bitpos return 0.
12199 2017-10-17  Olga Makhotina  <olga.makhotina@intel.com>
12201         * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
12202         _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
12203         _mm_maskz_reduce_ss): New.
12204         * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
12205         __builtin_ia32_reducess_mask): Ditto..
12206         (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
12207         * config/i386/sse.md (reduces<mode>): Renamed to ...
12208         (reduces<mode><mask_scalar_name>): ... this.
12209         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
12210         Changed to ...
12211         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
12212         %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
12214 2017-10-16  David Malcolm  <dmalcolm@redhat.com>
12216         * Makefile.in (OBJS): Add unique-ptr-tests.o.
12217         * selftest-run-tests.c (selftest::run_tests): Call
12218         selftest::unique_ptr_tests_cc_tests.
12219         * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
12220         * unique-ptr-tests.cc: New file.
12222 2017-10-16  Vladimir Makarov  <vmakarov@redhat.com>
12224         PR sanitizer/82353
12225         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12226         locations.
12227         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12228         (make_hard_regno_born, make_hard_regno_dead): Update
12229         bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12231 2017-10-16  Jeff Law  <law@redhat.com>
12233         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12235 2017-10-16  Jan Hubicka  <hubicka@ucw.cz>
12237         * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12239 2017-10-16  Olivier Hainque  <hainque@adacore.com>
12241         * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12242         with_cpu if we were configured for an e500v2 target cpu name.
12244 2017-10-16  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12246         * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12247         * doc/invoke.texi: Document +nodsp as a valid extension for
12248         -mcpu=cortex-m33.
12250 2017-10-16  Martin Liska  <mliska@suse.cz>
12252         * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12253         (test_set_range): Likewise.
12254         (test_range_functions): Rename to ...
12255         (test_bit_in_range): ... this.
12256         (sbitmap_c_tests): Add new test.
12258 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12260         * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12261         New.
12262         (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12263         (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12265 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12267         * config/aarch64/aarch64-builtins.c
12268         (aarch64_types_quadopu_lane_qualifiers): New.
12269         (TYPES_QUADOPU_LANE): New.
12270         * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12271         (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12272         (aarch64_<sur>dot_laneq<vsi2qi>): New.
12273         * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12274         (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12275         * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12276         (Vdottype, DOTPROD): New.
12277         (sur): Add SDOT and UDOT.
12279 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12281         * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12282         (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12283         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12284         Add TARGET_DOTPROD.
12285         * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12286         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12287         Enable TARGET_DOTPROD.
12288         (cortex-a75.cortex-a55): Likewise.
12289         * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12291 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12293         * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12294         (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12295         New.
12296         * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12297         New.
12298         * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12299         (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12300         * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12301         (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12302         * config/arm/types.md (neon_dot, neon_dot_q): New.
12303         * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12305 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12307         * config/arm/arm.h (TARGET_DOTPROD): New.
12308         * config/arm/arm.c (arm_arch_dotprod): New.
12309         (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12310         * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12311         * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12312         (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12313         (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12314         * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12315         * doc/invoke.texi (armv8.2-a): Document dotprod
12317 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12319         * i386.c (ix86_vec_cost): New function.
12320         (ix86_rtx_costs): Handle vector operations better.
12321         * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12322         * x86-tune-costs.h: Add new costs to all tables.
12324 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12326         * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12327         operations.
12328         * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12329         divsd, sqrtss and sqrtsd
12330         * x86-tune-costs.h: Add new entries to all costs.
12331         (znver1_cost): Fix to match real instruction latencies.
12333 2017-10-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12334             Michael Collison <michael.collison@arm.com>
12336         * compare-elim.c: Include emit-rtl.h.
12337         (can_merge_compare_into_arith): New function.
12338         (try_validate_parallel): Likewise.
12339         (try_merge_compare): Likewise.
12340         (try_eliminate_compare): Call the above when no previous clobber
12341         is available.
12342         (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12343         dataflow problems.
12345 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12347         PR middle-end/62263
12348         PR middle-end/82498
12349         * tree-ssa-phiopt.c (value_replacement): Comment fix.  Handle
12350         up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12352         PR middle-end/62263
12353         PR middle-end/82498
12354         * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12355         to be any operand_equal_p operands.  For & (B - 1) require
12356         B to be power of 2.  Recognize
12357         (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12359 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12361         PR bootstrap/82553
12362         * optabs.c (expand_memory_blockage): Fix call of
12363         targetm.have_memory_blockage.
12365 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12367         PR bootstrap/82548
12368         * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12369         x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12370         objects to extra_objs instead of overwriting it.
12372 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12374         * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12375         Use any_fp_register_operand as operand[3] predicate.  Simplify
12376         equality test for operands[2] and operands[4] memory location.
12377         (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12378         (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12379         (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12380         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12381         any_fp_register_operand as operand[1] predicate.  Simplify
12382         equality test for operands[0] and operands[3] memory location.
12383         (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12384         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12385         (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12387 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12389         * target-insns.def: Add memory_blockage.
12390         * optabs.c (expand_memory_blockage): New function.
12391         (expand_asm_memory_barrier): Rename ...
12392         (expand_asm_memory_blockage): ... to this.
12393         (expand_mem_thread_fence): Call expand_memory_blockage
12394         instead of expand_asm_memory_barrier.
12395         (expand_mem_singnal_fence): Ditto.
12396         (expand_atomic_load): Ditto.
12397         (expand_atomic_store): Ditto.
12398         * doc/md.texi (Standard Pattern Names For Generation):
12399         Document memory_blockage instruction pattern.
12401 2017-10-13  Sebastian Perta  <sebastian.perta@renesas.com>
12403         * config/rl78/rl78.c (rl78_emit_libcall): New function.
12404         * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12405         * config/rl78/rl78.md: New define_expand "adddi3".
12407 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12409         * cfghooks.c (verify_flow_info): Disable check that all probabilities
12410         are set correctly.
12412 2017-10-13  Jeff Law  <law@redhat.com>
12414         * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12416 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12418         PR target/82274
12419         * internal-fn.c (expand_mul_overflow): If both operands have
12420         the same highpart of -1 or 0 and the topmost bit of lowpart
12421         is different, overflow is if res <= 0 rather than res < 0.
12423 2017-10-13  Pat Haugen  <pthaugen@us.ibm.com>
12425         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12426         TARGET_P9_VECTOR code for unaligned_load case.
12428 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12430         * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12432 2017-10-13  Nathan Sidwell  <nathan@acm.org>
12434         * tree-core.h (tree_contains_struct): Make bool.
12435         * tree.c (tree_contains_struct): Likewise.
12436         * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12437         (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12438         MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12439         MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12440         MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12442 2017-10-13  Richard Biener  <rguenther@suse.de>
12444         * graphite-isl-ast-to-gimple.c
12445         (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12446         parameters and dominance check.
12447         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12448         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12449         (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12450         Do not update SSA form here or do intermediate IL verification.
12451         * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12452         (graphite_initialize): Remove check on the number of loops in
12453         the function and inline into graphite_transform_loops.
12454         (graphite_finalize): Inline into graphite_transform_loops.
12455         (graphite_transform_loops): Perform SSA update and IL verification
12456         here.
12457         * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12459 2017-10-13  Richard Biener  <rguenther@suse.de>
12461         * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12462         graphite_expression_type_precision): Avoid global constructor
12463         by moving ...
12464         (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12465         (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12466         (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12467         (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12468         (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12469         * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12471 2017-10-13  H.J. Lu  <hongjiu.lu@intel.com>
12473         PR target/82499
12474         * config/i386/i386.h (ix86_red_zone_size): New.
12475         * config/i386/i386.md (push peephole2s): Replace
12476         "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12478 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12479             Alan Hayward  <alan.hayward@arm.com>
12480             David Sherwood  <david.sherwood@arm.com>
12482         * combine.c (can_change_dest_mode): Reject changes in
12483         REGMODE_NATURAL_SIZE.
12485 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12486             Alan Hayward  <alan.hayward@arm.com>
12487             David Sherwood  <david.sherwood@arm.com>
12489         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12490         (expand_debug_source_expr): Likewise.
12491         * combine.c (combine_simplify_rtx): Likewise.
12492         * cse.c (fold_rtx): Likewise.
12493         * fwprop.c (canonicalize_address): Likewise.
12494         * targhooks.c (default_shift_truncation_mask): Likewise.
12496 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12497             Alan Hayward  <alan.hayward@arm.com>
12498             David Sherwood  <david.sherwood@arm.com>
12500         * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12501         (widened_mode): Likewise.
12502         (expand_unop): Likewise.
12503         * ree.c (transform_ifelse): Likewise.
12504         (merge_def_and_ext): Likewise.
12505         (combine_reaching_defs): Likewise.
12506         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12508 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12509             Alan Hayward  <alan.hayward@arm.com>
12510             David Sherwood  <david.sherwood@arm.com>
12512         * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12513         * combine.c (gen_lowpart_for_combine): Likewise.
12514         * dwarf2out.c (rtl_for_decl_location): Likewise.
12515         * final.c (alter_subreg): Likewise.
12516         * rtlhooks.c (gen_lowpart_general): Likewise.
12517         (gen_lowpart_if_possible): Likewise.
12519 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12520             Alan Hayward  <alan.hayward@arm.com>
12521             David Sherwood  <david.sherwood@arm.com>
12523         * calls.c (expand_call): Use subreg_lowpart_offset.
12524         * cse.c (cse_insn): Likewise.
12525         * regcprop.c (copy_value): Likewise.
12526         (copyprop_hardreg_forward_1): Likewise.
12528 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12530         PR target/82524
12531         * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12532         *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12533         =Q constraints to +Q and into insn condition add check
12534         that operands[0] and operands[1] are equal.
12535         (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12536         =Q constraints to +Q and into insn condition add check
12537         that operands[0] is equal to either operands[1] or operands[2].
12539         PR target/82498
12540         * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12541         instead of handling MINUS_EXPR twice (once for each argument),
12542         canonicalize operand order and handle just once, use rtype where
12543         possible.  Handle (A << B) | (A >> (-B & (Z - 1))).
12545         PR target/82498
12546         * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12547         any values of __C while still being pattern recognizable as a simple
12548         rotate instruction.
12550 2017-10-13  Richard Biener  <rguenther@suse.de>
12552         PR tree-optimization/82451
12553         Revert
12554         2017-10-02  Richard Biener  <rguenther@suse.de>
12556         PR tree-optimization/82355
12557         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12558         a mapping for the enclosing loop but avoid generating one for
12559         the loop tree root.
12560         (copy_bb_and_scalar_dependences): Remove premature codegen
12561         error on PHIs in blocks duplicated into multiple places.
12562         * graphite-scop-detection.c
12563         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12564         in the region use it as loop and nest to analyze the DR in.
12565         (try_generate_gimple_bb): Likewise.
12566         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12567         (add_loop_constraints): For blocks in a loop not in the region
12568         create a dimension with a single iteration.
12569         * sese.h (gbb_loop_at_index): Remove assert.
12571         * cfgloop.c (loop_preheader_edge): For the loop tree root
12572         return the single successor of the entry block.
12573         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12574         Reset the SCEV hashtable and niters.
12575         * graphite-scop-detection.c
12576         (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12577         assert that we only have POLYNOMIAL_CHREC that vary in loops
12578         contained in the region.
12579         (scop_detection::graphite_can_represent_expr): Adjust.
12580         (scop_detection::stmt_has_simple_data_refs_p): For loops
12581         not in the region set loop to NULL.  The nest is now the
12582         entry edge to the region.
12583         (try_generate_gimple_bb): Likewise.
12584         * sese.c (scalar_evolution_in_region): Adjust for
12585         instantiate_scev change.
12586         * tree-data-ref.h (graphite_find_data_references_in_stmt):
12587         Make nest parameter the edge into the region.
12588         (create_data_ref): Likewise.
12589         * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12590         entry edge into a region and adjust instantiate_scev calls.
12591         (create_data_ref): Likewise.
12592         (graphite_find_data_references_in_stmt): Likewise.
12593         (find_data_references_in_stmt): Pass the loop preheader edge
12594         from the nest argument.
12595         * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12596         parameter the edge into the region.
12597         (instantiate_parameters): Use the loop preheader edge as entry.
12598         * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12599         NULL loop.
12600         (get_instantiated_value_entry): Make instantiate_below parameter
12601         the edge into the region.
12602         (instantiate_scev_name): Likewise.  Adjust dominance checks,
12603         when we cannot use loop-based instantiation instantiate by
12604         walking use-def chains.
12605         (instantiate_scev_poly): Adjust.
12606         (instantiate_scev_binary): Likewise.
12607         (instantiate_scev_convert): Likewise.
12608         (instantiate_scev_not): Likewise.
12609         (instantiate_array_ref): Remove.
12610         (instantiate_scev_3): Likewise.
12611         (instantiate_scev_2): Likewise.
12612         (instantiate_scev_1): Likewise.
12613         (instantiate_scev_r): Do not blindly handle N-operand trees.
12614         Do not instantiate array-refs.  Handle all constants and invariants.
12615         (instantiate_scev): Make instantiate_below parameter
12616         the edge into the region.
12617         (resolve_mixers): Use the loop preheader edge for the region
12618         parameter to instantiate_scev_r.
12619         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12621 2017-10-13  Richard Biener  <rguenther@suse.de>
12623         PR tree-optimization/82525
12624         * graphite-isl-ast-to-gimple.c
12625         (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12626         out from ...
12627         (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12628         Fail code generation when we cannot represent the isl integer.
12629         (binary_op_to_tree): Elide modulo operations that are no-ops
12630         in the type we code generate.  Remove now superfluous code
12631         generation errors.
12633 2017-10-13  Richard Biener  <rguenther@suse.de>
12635         * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12636         (scop_detection::harmful_loop_in_region): Remove premature
12637         IV type restriction.
12638         (scop_detection::graphite_can_represent_scev): We can handle
12639         pointer IVs just fine.
12641 2017-10-13  Alan Modra  <amodra@gmail.com>
12643         * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12644         "Clobbers and Scratch Registers".  Add paragraph on
12645         alternative to clobbers for scratch registers and OpenBLAS
12646         example.
12648 2017-10-13  Alan Modra  <amodra@gmail.com>
12650         * doc/extend.texi (Clobbers): Correct vax example.  Delete old
12651         example of a memory input for a string of known length.  Move
12652         commentary out of table.  Add a number of new examples
12653         covering array memory inputs.
12655 2017-10-12  Martin Liska  <mliska@suse.cz>
12657         PR tree-optimization/82493
12658         * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12659         (test_range_functions): New function.
12660         (sbitmap_c_tests): Likewise.
12661         * selftest-run-tests.c (selftest::run_tests): Run new tests.
12662         * selftest.h (sbitmap_c_tests): New function.
12664         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12666 2017-10-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
12668         * config/rs6000/amo.h: Fix spacing issue.
12670 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12672         PR target/82498
12673         * config/i386/i386.md (*ashl<mode>3_mask_1,
12674         *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12675         *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12676         patterns.
12678 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12680         * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12681         (profile_probability): Set max_probability
12682         to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12683         in temporaries.
12684         * profile-count.c (profile_probability::differs_from_p): Do not
12685         rely on max_probaiblity == 10000
12687 2017-10-12  Jeff Law  <law@redhat.com>
12689         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12690         negative offsets.
12692 2017-10-12  Martin Sebor  <msebor@redhat.com>
12694         PR other/82301
12695         PR c/82435
12696         * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12697         (handle_alias_pairs): Call it.
12698         * common.opt (-Wattribute-alias): New option.
12699         * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12700         * doc/invoke.texi (-Wattribute-alias): Document.
12702 2017-10-12  Vladimir Makarov  <vmakarov@redhat.com>
12704         Revert
12705         2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12706         PR sanitizer/82353
12707         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12708         locations.
12709         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12710         (make_hard_regno_born, make_hard_regno_dead): Update
12711         bb_killed_pseudos and bb_gen_pseudos.
12713 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12715         * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12717 2017-10-12  Uros Bizjak  <ubizjak@gmail.com>
12719         * config/alpha/alpha.c (alpha_split_conditional_move):
12720         Use std::swap instead of manually swapping.
12721         (alpha_stdarg_optimize_hook): Ditto.
12722         (alpha_canonicalize_comparison): Ditto.
12724 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12726         * tree-loop-distribution.c (struct builtin_info): New struct.
12727         (struct partition): Refactor fields into struct builtin_info.
12728         (partition_free): Free struct builtin_info.
12729         (build_size_arg_loc, build_addr_arg_loc): Delete.
12730         (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12731         information from struct builtin_info.
12732         (find_single_drs): New function refactored from classify_partition.
12733         Also moved builtin validity checks to this function.
12734         (compute_access_range, alloc_builtin): New functions.
12735         (classify_builtin_st, classify_builtin_ldst): New functions.
12736         (classify_partition): Refactor code into functions find_single_drs,
12737         classify_builtin_st and classify_builtin_ldst.
12738         (distribute_loop): Don't do runtime alias check when distributing
12739         loop nest.
12740         (find_seed_stmts_for_distribution): New function.
12741         (pass_loop_distribution::execute): Refactor code finding seed
12742         stmts into above function.  Support distribution for the innermost
12743         two-level loop nest.  Adjust dump information.
12745 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12747         * tree-loop-distribution.c: Adjust the general comment.
12748         (NUM_PARTITION_THRESHOLD): New macro.
12749         (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12750         (classify_partition): Skip builtin pattern of loop nest's inner loop.
12751         (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12752         in call to build_partition_graph.
12753         (finalize_partitions): New parameter.  Make loop distribution more
12754         conservative by fusing more partitions.
12755         (distribute_loop): Don't do runtime alias check in case of loop nest
12756         distribution.
12757         (find_seed_stmts_for_distribution): New function.
12758         (prepare_perfect_loop_nest): New function.
12759         (pass_loop_distribution::execute): Refactor code finding seed stmts
12760         and loop nest into above functions.  Support loop nest distribution.
12761         Adjust dump information accordingly.
12763 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12765         * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12766         and set PTYPE_SEQUENTIAL for merged partition.
12768 2017-10-12  Richard Biener  <rguenther@suse.de>
12770         PR tree-optimization/69728
12771         Revert
12772         2017-09-19  Richard Biener  <rguenther@suse.de>
12774         PR tree-optimization/69728
12775         * graphite-sese-to-poly.c (schedule_error): New global.
12776         (add_loop_schedule): Handle empty domain by failing the
12777         schedule.
12778         (build_original_schedule): Handle schedule_error.
12780         * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12781         domain by returning an unchanged schedule.
12783 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12785         * genrecog.c (validate_pattern): For VEC_SELECT verify that
12786         CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12788 2017-10-12  Aldy Hernandez  <aldyh@redhat.com>
12790         * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12791         Handle params.def.
12793 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12795         PR c++/82159
12796         * expr.c (store_field): Don't optimize away bitsize == 0 store
12797         from CALL_EXPR with addressable return type.
12799 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12801         * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12802         * config/rs6000/rs6000.md (sel): Delete mode attribute.
12803         (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12804         *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12805         TARGET_ISEL instead of TARGET_ISEL<sel>.
12807 2017-10-11  David Edelsohn  <dje.gcc@gmail.com>
12809         * config/rs6000/rs6000.c
12810         (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12812 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12814         * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12815         Move up in file.
12816         (reg_or_cint_operand): Fix comment.
12817         (reg_or_zero_operand): New predicate.
12818         * config/rs6000/rs6000-protos.h (output_isel): Delete.
12819         * config/rs6000/rs6000.c (output_isel): Delete.
12820         * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12821         instead of reg_or_cint_operand.  Output instruction directly (not via
12822         output_isel).
12823         (isel_unsigned_<mode>): Ditto.
12824         (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12825         gpc_reg_operand.  Add an instruction alternative for this.  Output
12826         instruction directly.
12827         (*isel_reversed_unsigned_<mode>): Ditto.
12829 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12831         * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12832         (TARGET_CANONICALIZE_COMPARISON): Define.
12834 2017-10-11  Qing Zhao  <qing.zhao@oracle.com>
12836         PR target/81422
12837         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12838         Check whether the dest is REG before adding REG_EQUIV note.
12840 2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12842         PR sanitizer/82353
12843         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12844         locations.
12845         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12846         (make_hard_regno_born, make_hard_regno_dead): Update
12847         bb_killed_pseudos and bb_gen_pseudos.
12849 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12851         * incpath.h (enum incpath_kind): Name enum, prefix values.
12852         (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12853         * incpath.c (heads, tails): Use INC_MAX.
12854         (add_env_var_paths, add_standard_paths): Use incpath_kind.
12855         (merge_include_chains, split_quote_chain,
12856         register_include_chains): Update incpath_kind names.
12857         (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12858         * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12859         names.
12860         (add_framework_path, darwin_register_objc_includes): Likewise.
12861         * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12863 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12865         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12866         Do not use float_operator operator predicate.
12867         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12868         * config/i386/predicates.md (float_operator): Remove predicate.
12870 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12872         * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12873         (*jccxf_i387): Ditto.
12874         (*jcc<mode>_i387): Ditto.
12875         (*jccu<mode>_i387): Ditto.
12876         (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12877         (*jcc_*_i387 splitters): Remove.
12878         * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12879         * config/i386/i386.c (ix86_split_fp_branch): Remove.
12880         * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12881         Remove predicate.
12883 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12885         * profile-count.h (slow_safe_scale_64bit): New function.
12886         (safe_scale_64bit): New inline.
12887         (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12888         * profile-count.c: Include wide-int.h
12889         (slow_safe_scale_64bit): New.
12891 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12893         * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12894         HAS_DECL_ASSEMBLER_NAME_P.
12895         * gimple-expr.c (gimple_decl_printable_name: Check
12896         HAS_DECL_ASSEMBLER_NAME_P too.
12897         * ipa-utils.h (type_in_anonymous_namespace_p): Check
12898         DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
12899         (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
12900         * passes.c (rest_of_decl_compilation): Check
12901         HAS_DECL_ASSEMBLER_NAME_P too.
12902         * recog.c (verify_changes): Likewise.
12903         * tree-pretty-print.c (dump_decl_name): Likewise.
12904         * tree-ssa-structalias.c (alias_get_name): Likewise.  Reimplement.
12906         * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
12907         (SET_DECL_ASSEMBLER_NAME): Use it.
12908         (DECL_ASSEMBLER_NAME_SET_P): Likewise.
12909         (COPY_DECL_ASSEMBLER_NAME): Likewise.
12910         * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
12912 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12914         * config.gcc (i386, x86_64): Add extra objects.
12915         * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
12916         (ix86_min_insn_size): Declare.
12917         (ix86_issue_rate): Declare.
12918         (ix86_adjust_cost): Declare.
12919         (ia32_multipass_dfa_lookahead): Declare.
12920         (ix86_macro_fusion_p): Declare.
12921         (ix86_macro_fusion_pair_p): Declare.
12922         (ix86_bd_has_dispatch): Declare.
12923         (ix86_bd_do_dispatch): Declare.
12924         (ix86_core2i7_init_hooks): Declare.
12925         (ix86_atom_sched_reorder): Declare.
12926         * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
12927         (COSTS_N_BYTES): Move to x86-tune-costs.h.
12928         (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
12929         (rip_relative_addr_p): Rename to ...
12930         (ix86_rip_relative_addr_p): ... this one; export.
12931         (memory_address_length): Update.
12932         (ix86_issue_rate): Move to x86-tune-sched.c.
12933         (ix86_flags_dependent): Move to x86-tune-sched.c.
12934         (ix86_agi_dependent): Move to x86-tune-sched.c.
12935         (exact_dependency_1): Move to x86-tune-sched.c.
12936         (exact_store_load_dependency): Move to x86-tune-sched.c.
12937         (ix86_adjust_cost): Move to x86-tune-sched.c.
12938         (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
12939         (ix86_macro_fusion_p): Move to x86-tune-sched.c.
12940         (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
12941         (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
12942         (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
12943         (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
12944         (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
12945         (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
12946         (min_insn_size): Rename to ...
12947         (ix86_min_insn_size): ... this one; export.
12948         (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
12949         (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
12950         (core2i7_first_cycle_multipass_backtrack): Move to
12951         x86-tune-sched-core.c.
12952         (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
12953         (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
12954         (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
12955         (ix86_avoid_jump_mispredicts): Update.
12956         (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
12957         (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
12958         (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
12959         (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
12960         (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
12961         (MAX_INSN): Move to ix86-tune-sched-bd.c.
12962         (MAX_IMM): Move to ix86-tune-sched-bd.c.
12963         (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
12964         (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
12965         (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
12966         (MAX_LOAD): Move to ix86-tune-sched-bd.c.
12967         (MAX_STORE): Move to ix86-tune-sched-bd.c.
12968         (BIG): Move to ix86-tune-sched-bd.c.
12969         (enum dispatch_group): Move to ix86-tune-sched-bd.c.
12970         (enum insn_path): Move to ix86-tune-sched-bd.c.
12971         (get_mem_group): Move to ix86-tune-sched-bd.c.
12972         (is_cmp): Move to ix86-tune-sched-bd.c.
12973         (dispatch_violation): Move to ix86-tune-sched-bd.c.
12974         (is_branch): Move to ix86-tune-sched-bd.c.
12975         (is_prefetch): Move to ix86-tune-sched-bd.c.
12976         (init_window): Move to ix86-tune-sched-bd.c.
12977         (allocate_window): Move to ix86-tune-sched-bd.c.
12978         (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
12979         (is_end_basic_block): Move to ix86-tune-sched-bd.c.
12980         (process_end_window): Move to ix86-tune-sched-bd.c.
12981         (allocate_next_window): Move to ix86-tune-sched-bd.c.
12982         (find_constant): Move to ix86-tune-sched-bd.c.
12983         (get_num_immediates): Move to ix86-tune-sched-bd.c.
12984         (has_immediate): Move to ix86-tune-sched-bd.c.
12985         (get_insn_path): Move to ix86-tune-sched-bd.c.
12986         (get_insn_group): Move to ix86-tune-sched-bd.c.
12987         (count_num_restricted): Move to ix86-tune-sched-bd.c.
12988         (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
12989         (add_insn_window): Move to ix86-tune-sched-bd.c.
12990         (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
12991         (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
12992         (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
12993         (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
12994         (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
12995         (do_dispatch): Move to ix86-tune-sched-bd.c.
12996         (has_dispatch): Move to ix86-tune-sched-bd.c.
12997         * config/i386/t-i386: Add new object files.
12998         * config/i386/x86-tune-costs.h: New file.
12999         * config/i386/x86-tune-sched-atom.c: New file.
13000         * config/i386/x86-tune-sched-bd.c: New file.
13001         * config/i386/x86-tune-sched-core.c: New file.
13002         * config/i386/x86-tune-sched.c: New file.
13004 2017-10-11  Liu Hao  <lh_mouse@126.com>
13006         * pretty-print.c [_WIN32] (colorize_init): Remove.  Use
13007         the generic version below instead.
13008         (should_colorize): Recognize Windows consoles as terminals
13009         for MinGW targets.
13010         * pretty-print.c [__MINGW32__] (write_all): New function.
13011         [__MINGW32__] (find_esc_head): Likewise.
13012         [__MINGW32__] (find_esc_terminator): Likewise.
13013         [__MINGW32__] (eat_esc_sequence): Likewise.
13014         [__MINGW32__] (mingw_ansi_fputs): New function that handles
13015         ANSI escape codes.
13016         (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
13017         for MinGW targets.
13019 2017-10-11  Richard Biener  <rguenther@suse.de>
13021         * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
13022         Properly call analyze_scalar_evolution with the loop of the stmt.
13024 2017-10-11  Richard Biener  <rguenther@suse.de>
13026         * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
13027         * tree-core.h (tree_base): Add chrec_var union member.
13028         * tree.h (CHREC_VAR): Remove.
13029         (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
13030         * tree-chrec.h (build_polynomial_chrec): Adjust.
13031         * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
13032         * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
13034 2017-10-11  Marc Glisse  <marc.glisse@inria.fr>
13036         * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
13037         * match.pd: ... here.
13038         ((T) X == (T) Y): Relax condition.
13040 2017-10-11  Bin Cheng  <bin.cheng@arm.com>
13042         PR tree-optimization/82472
13043         * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
13044         comment.
13045         (break_alias_scc_partitions): Update postorder number.
13047 2017-10-11  Martin Liska  <mliska@suse.cz>
13049         PR sanitizer/82490
13050         * opts.c (parse_no_sanitize_attribute): Do not use error_value
13051         variable.
13052         * opts.h (parse_no_sanitize_attribute): Remove last argument.
13054 2017-10-11  Martin Liska  <mliska@suse.cz>
13056         * print-rtl.c (print_insn): Move declaration of idbuf
13057         to same scope as name.
13059 2017-10-11  Martin Liska  <mliska@suse.cz>
13061         Revert r253637:
13063         PR sanitizer/82484
13064         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13065         volatile arguments.
13067 2017-10-11  Martin Liska  <mliska@suse.cz>
13069         PR sanitizer/82484
13070         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
13071         volatile arguments.
13073 2017-10-11  Adhemerval Zanella  <adhemerval.zanella@linaro.org>
13075         * config.gcc (default_gnu_indirect_function): Default to yes for
13076         arm*-*-linux* with glibc.
13078 2017-10-11  Richard Biener  <rguenther@suse.de>
13080         * tree-scalar-evolution.c (get_scalar_evolution): Handle
13081         default-defs and types we do not want to analyze.
13082         (interpret_loop_phi): Replace unreachable code with an assert.
13083         (compute_scalar_evolution_in_loop): Remove and inline ...
13084         (analyze_scalar_evolution_1): ... here, replacing condition with
13085         what makes the intent clearer.  Remove handling of cases
13086         get_scalar_evolution now handles.
13088 2017-10-10  Jim Wilson  <wilson@tuliptree.org>
13090         PR rtl-optimization/81434
13091         * haifa-sched.c (prune_ready_list): Init min_cost_group to 0.  Update
13092         comment for main loop.  In sched_group_found if, also add checks for
13093         pass and min_cost_group.
13095 2017-10-10  Segher Boessenkool  <segher@kernel.crashing.org>
13097         * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
13098         (rs6000_insn_cost): New function.
13099         * config/rs6000/rs6000.md (cost): New attribute.
13101 2017-10-10  Jakub Jelinek  <jakub@redhat.com>
13102             H.J. Lu  <hongjiu.lu@intel.com>
13104         PR target/79565
13105         PR target/82483
13106         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
13107         OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
13108         __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
13109         (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
13110         to OPTION_MASK_ISA_AVX512VL - builtins that have both
13111         OPTION_MASK_ISA_MMX and some other bit set require both
13112         mmx and the ISAs without the mmx bit.
13113         * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
13114         __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
13115         __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
13116         __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
13117         __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
13118         __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
13119         __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
13120         __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
13121         __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
13122         __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
13123         __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
13124         __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
13125         __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
13126         __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
13127         Add OPTION_MASK_ISA_MMX.
13129 2017-10-10  Andreas Tobler  <andreast@gcc.gnu.org>
13131         * config.gcc (armv7*-*-freebsd*): New target.
13132         (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
13134 2017-10-10  Jan Hubicka  <hubicka@ucw.cz>
13136         * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
13137         X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
13138         spot in the file.
13140 2017-10-10  Richard Sandiford  <richard.sandiford@linaro.org>
13142         * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
13143         a template parameter.
13144         (WIDE_INT_REF_FOR): Update accordingly.
13145         * tree.h (wi::int_traits <const_tree>): Delete.
13146         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
13147         (wi::to_widest, wi::to_offset): Use them.  Expand commentary.
13148         (wi::tree_to_wide_ref): New typedef.
13149         (wi::to_wide): New function.
13150         * calls.c (get_size_range): Use wi::to_wide when operating on
13151         trees as wide_ints.
13152         * cgraph.c (cgraph_node::create_thunk): Likewise.
13153         * config/i386/i386.c (ix86_data_alignment): Likewise.
13154         (ix86_local_alignment): Likewise.
13155         * dbxout.c (stabstr_O): Likewise.
13156         * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
13157         * expr.c (const_vector_from_tree): Likewise.
13158         * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
13159         * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
13160         (fold_negate_expr_1, int_const_binop_1, const_binop)
13161         (fold_convert_const_int_from_real, optimize_bit_field_compare)
13162         (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
13163         (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
13164         (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
13165         (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
13166         (fold_not_const, round_up_loc): Likewise.
13167         * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
13168         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
13169         (alloca_call_type): Likewise.
13170         * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
13171         * godump.c (go_output_typedef): Likewise.
13172         * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
13173         * internal-fn.c (get_min_precision): Likewise.
13174         * ipa-cp.c (ipcp_store_vr_results): Likewise.
13175         * ipa-polymorphic-call.c
13176         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
13177         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
13178         (ipa_modify_call_arguments): Likewise.
13179         * match.pd: Likewise.
13180         * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
13181         * print-tree.c (print_node_brief, print_node): Likewise.
13182         * stmt.c (expand_case): Likewise.
13183         * stor-layout.c (layout_type): Likewise.
13184         * tree-affine.c (tree_to_aff_combination): Likewise.
13185         * tree-cfg.c (group_case_labels_stmt): Likewise.
13186         * tree-data-ref.c (dr_analyze_indices): Likewise.
13187         (prune_runtime_alias_test_list): Likewise.
13188         * tree-dump.c (dequeue_and_dump): Likewise.
13189         * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
13190         * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
13191         * tree-pretty-print.c (dump_generic_node): Likewise.
13192         * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
13193         (simple_iv_with_niters): Likewise.
13194         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
13195         * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
13196         * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
13197         * tree-ssa-loop-niter.c (split_to_var_and_offset)
13198         (refine_value_range_using_guard, number_of_iterations_ne_max)
13199         (number_of_iterations_lt_to_ne, number_of_iterations_lt)
13200         (get_cst_init_from_scev, record_nonwrapping_iv)
13201         (scev_var_range_cant_overflow): Likewise.
13202         * tree-ssa-phiopt.c (minmax_replacement): Likewise.
13203         * tree-ssa-pre.c (compute_avail): Likewise.
13204         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
13205         (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
13206         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
13207         * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
13208         * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
13209         * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
13210         (dump_case_nodes, try_switch_expansion): Likewise.
13211         * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
13212         (vect_do_peeling): Likewise.
13213         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13214         * tree-vect-stmts.c (vectorizable_load): Likewise.
13215         * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
13216         (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
13217         (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
13218         (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
13219         (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
13220         (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
13221         (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
13222         (evrp_dom_walker::before_dom_children): Likewise.
13223         * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13224         (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13225         (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13226         (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13227         (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13228         (get_range_pos_neg): Likewise.
13229         * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13230         * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13231         * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13232         * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13233         * config/avr/avr.c (avr_fold_builtin): Likewise.
13234         * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13235         * config/msp430/msp430.c (msp430_attr): Likewise.
13236         * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13237         * config/powerpcspe/powerpcspe-c.c
13238         (altivec_resolve_overloaded_builtin): Likewise.
13239         * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13240         (rs6000_expand_ternop_builtin): Likewise.
13241         * config/rs6000/rs6000-c.c
13242         (altivec_resolve_overloaded_builtin): Likewise.
13243         * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13244         (rs6000_expand_ternop_builtin): Likewise.
13245         * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13247 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13249         * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13250         when copying loop nest with only one inner loop.
13252 2017-10-10  Richard Biener  <rguenther@suse.de>
13254         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13255         blocks if SCEV is active.
13256         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13257         dead code.
13258         (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13259         (scev_initialize): Assert we are not yet initialized.
13261 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13263         * tree-loop-distribution.c (generate_loops_for_partition): Remove
13264         inner loop's exit stmt by making it always exit the loop, otherwise
13265         we would generate an infinite empty loop.
13267 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13269         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13270         renaming variables in new preheader if it's deleted.
13272 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13274         * tree-loop-distribution.c (struct partition): Remove unused field
13275         loops of the structure.
13276         (partition_alloc, partition_free): Ditto.
13277         (build_rdg_partition_for_vertex): Ditto.
13279 2017-10-09  Jeff Law  <law@redhat.com>
13281         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13282         return type to match prototype and documentation.
13284 2010-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13286         * config/rs6000/rs6000.c (processor_costs): Move to ...
13287         * config/rs6000/rs6000.h: ... here.
13288         (rs6000_cost): Declare.
13290 2017-10-09  Eric Botcazou  <ebotcazou@adacore.com>
13292         * except.c (setjmp_fn): New global variable.
13293         (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13294         (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13295         if DONT_USE_BUILTIN_SETJMP is defined.
13297 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13299         * target.def (insn_cost): New hook.
13300         * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13301         * doc/tm.texi: Regenerate.
13302         * rtlanal.c (insn_cost): Use the new hook.
13304 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13306         * combine.c (combine_validate_cost): Compute the new insn_cost,
13307         not just pattern_cost.
13308         (try_combine): Adjust comment.
13310 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13312         * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13313         insn_cost.
13314         * combine.c (uid_insn_cost): Adjust comment.
13315         (combine_validate_cost): Adjust comment.  Use pattern_cost instead
13316         of insn_rtx_cost
13317         (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13318         * dse.c (find_shift_sequence): Ditto.
13319         * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13320         (bb_valid_for_noce_process_p): Use pattern_cost.
13321         * rtl.h (insn_rtx_cost): Delete.
13322         (pattern_cost): New prototype.
13323         (insn_cost): New prototype.
13324         * rtlanal.c (insn_rtx_cost): Rename to...
13325         (pattern_cost): ... this.
13326         (insn_cost): New.
13328 2017-10-09  Uros Bizjak  <ubizjak@gmail.com>
13330         * config/i386/i386.md (*jcc_2): Remove insn pattern.
13331         (*jcc<mode>_0_r_i387): Ditto.
13332         (*jccxf_r_i387): Ditto.
13333         (*jcc<mode>_r_i387): Ditto.
13334         (*jccu<mode>_r_i387): Ditto.
13335         (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13336         (*jcc): Rename from *jcc_1.
13338 2017-10-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13340         * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13341         deferred rescans after the lvx/stvx recombination pre-pass.
13343 2017-10-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
13345         * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13346         memory operation instruction support.
13347         * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13348         (rs6000-ibm-aix[789]*): Likewise.
13349         * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13350         Document new functions.
13352 2017-10-09  Richard Biener  <rguenther@suse.de>
13354         PR tree-optimization/82397
13355         * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13356         equality only for semantically equal trees.
13358 2017-10-09  Richard Biener  <rguenther@suse.de>
13360         PR tree-optimization/82449
13361         * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13362         * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13363         allow constant addresses.
13364         * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13365         are linear.
13367 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13369         * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13370         flags.
13372 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13374         PR target/82463
13375         * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13376         definitions.
13378 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13380         PR target/82465
13381         * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13383 2017-10-09  Jakub Jelinek  <jakub@redhat.com>
13385         PR target/82464
13386         * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13387         s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13389 2017-10-09  Richard Sandiford  <richard.sandiford@linaro.org>
13391         * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13392         (WI_BINARY_PREDICATE_RESULT): Likewise.
13393         (wi::binary_traits::operator_result): New type.
13394         (wi::binary_traits::predicate_result): Likewise.
13395         (generic_wide_int::operator~, unary generic_wide_int::operator-)
13396         (generic_wide_int::operator==, generic_wide_int::operator!=)
13397         (generic_wide_int::operator&, generic_wide_int::and_not)
13398         (generic_wide_int::operator|, generic_wide_int::or_not)
13399         (generic_wide_int::operator^, generic_wide_int::operator+
13400         (binary generic_wide_int::operator-, generic_wide_int::operator*):
13401         Delete.
13402         (operator~, unary operator-, operator==, operator!=, operator&)
13403         (operator|, operator^, operator+, binary operator-, operator*): New
13404         functions.
13405         * expr.c (get_inner_reference): Use wi::bit_and_not.
13406         * fold-const.c (fold_binary_loc): Likewise.
13407         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13408         * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13409         (bit_value_binop): Likewise.
13410         * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13411         * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13412         (extract_range_from_binary_expr_1): Likewise.
13413         (masked_increment): Likewise.
13414         (simplify_bit_ops_using_ranges): Likewise.
13416 2017-10-09  Martin Jambor  <mjambor@suse.cz>
13418         PR hsa/82416
13419         * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13420         * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13421         (hsa_type_for_scalar_tree_type): Use it.  Always force min32int for
13422         COMPLEX types.
13423         (hsa_fixup_mov_insn_type): New function.
13424         (hsa_op_with_type::get_in_type): Use it.
13425         (hsa_build_append_simple_mov): Likewise.  Allow sub-32bit
13426         immediates in an assert.
13427         (hsa_op_with_type::extend_int_to_32bit): New method.
13428         (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13429         types.  Convert to dest type if necessary.
13430         (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13431         (reg_for_gimple_ssa): Pass false as min32int to
13432         hsa_type_for_scalar_tree_type.
13433         (gen_hsa_addr): Fixup type when creating addresable temporary.
13434         (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13435         (gen_hsa_unary_operation): Extend operands and convert to dest type if
13436         necessary.  Call hsa_fixup_mov_insn_type.
13437         (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13438         extend operands and convert to dest type if necessary.
13439         (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13440         to dest type if necessary.
13441         (set_output_in_type): Call hsa_fixup_mov_insn_type.  Just ude dest
13442         if conversion nt necessary and size matches.
13443         (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13444         to dest type if necessary.
13445         (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13446         (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13447         necessary.
13448         (gen_hsa_clrsb): Likewise.
13449         (gen_hsa_ffs): Likewise.
13450         (gen_hsa_divmod): Extend operands and convert to dest type if
13451         necessary.
13452         (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13454 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13456         * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13457         Remove empty default arguments.  Use a brace block as output
13458         statement.
13459         (conditional return): Ditto.
13460         (jump): Ditto.
13461         (indirect_jump): Ditto.  Use b%T0 instead of bctr/blr.
13462         (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13463         Ditto.
13464         (group_ending_nop): Ditto.
13465         (doloop_end): Ditto.
13466         (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13467         (splitters for those): Ditto.
13469 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13471         * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13472         a conditional jump (and the compare for it) so that pc_rtx is the
13473         last operand.
13474         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13475         for the deleted and renamed ctr<mode>_internal[234] patterns.
13476         * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13477         Delete second conditional return pattern.
13478         (ctr<mode>_internal2): Delete this second bdnz pattern.
13479         (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13480         (ctr<mode>_internal4): Delete this second bdz pattern.
13482 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13484         * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13485         (always_initialized_rtx_for_ssa_name_p): New predicate.
13486         * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13487         (finish_out_of_ssa): Free new field of SA.
13488         * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13489         * tree-ssa-coalesce.c: Include tree-ssa.h.
13490         (get_parm_default_def_partitions): Remove extern keyword.
13491         (get_undefined_value_partitions): New function.
13492         * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13493         not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13494         uninitialized bits.
13495         * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13497 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13499         * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13501 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13503         * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13504         for targets that preffer 128bit.
13506 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13508         * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13510 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13512         * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13513         on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13514         prevent DSE.
13515         (thumb_set_return_address): Likewise.
13517 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13519         * common/config/arm/arm-common.c (arm_except_unwind_info):
13520         Handle DWARF2_UNWIND_INFO.
13522 2017-10-07  Michael Collison <michael.collison@arm.com>
13524         * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13525         New pattern.
13527 2017-10-07  Eric Botcazou  <ebotcazou@adacore.com>
13529         * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13530         of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13531         * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13532         defined, force the creation of a new block for a dispatch label.
13534 2017-10-07  Jan Hubicka  <hubicka@ucw.cz>
13536         * invoke.texi (Wsuggest-attribute=cold): Document.
13537         * common.opt (Wsuggest-attribute=cold): New
13538         * ipa-pure-const.c (warn_function_cold): New function.
13539         * predict.c (compute_function_frequency): Use it.
13540         * predict.h (warn_function_cold): Declare.
13542 2017-10-06  Jan Hubicka  <hubicka@ucw.cz>
13544         * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13545         Update profile.
13547 2017-10-06  Martin Liska  <mliska@suse.cz>
13549         * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13550         keyword for member functions.
13551         (struct sanopt_tree_couple): New struct.
13552         (struct sanopt_tree_couple_hash): New function.
13553         (struct sanopt_ctx): Add new hash_map.
13554         (has_dominating_ubsan_ptr_check): New function.
13555         (record_ubsan_ptr_check_stmt): Likewise.
13556         (maybe_optimize_ubsan_ptr_ifn): Likewise.
13557         (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13558         (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13560 2017-10-06  Sudakshina Das  <sudi.das@arm.com>
13562         PR target/82440
13563         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13564         aarch64_simd_valid_immediate on CONST_VECTORs.
13565         (aarch64_reg_or_bic_imm): Likewise.
13567 2017-10-06  Wilco Dijkstra  <wdijkstr@arm.com>
13569         PR rtl-optimization/82396
13570         * haifa-sched.c (ready_sort_real): Disable qsort checking.
13572 2017-10-06  Sebastian Pop  <sebpop@gmail.com>
13574         * graphite-dependences.c (scop_get_reads): Move code to...
13575         (scop_get_must_writes): Move code to...
13576         (scop_get_may_writes): Move code to...
13577         (scop_get_reads_and_writes): ... here.
13578         (scop_get_dependences): Call scop_get_reads_and_writes.
13580 2017-10-06  Jakub Jelinek  <jakub@redhat.com>
13582         PR tree-optimization/82434
13583         * fold-const.h (can_native_encode_type_p,
13584         can_native_encode_string_p): Remove.
13585         * fold-const.c (native_encode_int): Formatting fixes.  If ptr is NULL,
13586         don't encode anything, just return what would be otherwise returned.
13587         (native_encode_fixed, native_encode_complex, native_encode_vector):
13588         Likewise.
13589         (native_encode_string): Likewise.  Inline by hand
13590         can_native_encode_string_p.
13591         (can_native_encode_type_p): Remove.
13592         (can_native_encode_string_p): Remove.
13593         * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13594         STRING_CSTs using can_native_encode_string_p, test all
13595         CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13596         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13597         argument from native_encode_expr.
13598         (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13599         (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13600         but just 2.
13602 2017-10-06  Richard Biener  <rguenther@suse.de>
13604         PR tree-optimization/82397
13605         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13606         operand_equal_p but rely on data_ref_compare_tree for detecting
13607         equalities.
13608         (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13609         to match up with dr_group_sort_cmp.
13611 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13613         PR target/82322
13614         * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13615         builtin.
13616         * config/s390/s390-builtin-types.def: Regenerate.
13618 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13620         PR target/82317
13621         * config/s390/s390-builtin-types.def: Regenerate.
13622         * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13623         Change flag from B_VXE to B_VX.
13624         (s390_vec_min_dbl): Remove B_VXE flag.
13626 2017-10-06  Richard Biener  <rguenther@suse.de>
13628         * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13629         (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13630         translate_isl_ast_to_gimple::is_valid_rename,
13631         translate_isl_ast_to_gimple::get_rename,
13632         translate_isl_ast_to_gimple::get_def_bb_for_const,
13633         translate_isl_ast_to_gimple::get_new_name,
13634         translate_isl_ast_to_gimple::collect_all_ssa_names,
13635         translate_isl_ast_to_gimple::copy_loop_phi_args,
13636         translate_isl_ast_to_gimple::collect_all_ssa_names,
13637         translate_isl_ast_to_gimple::copy_loop_phi_args,
13638         translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13639         translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13640         translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13641         translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13642         translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13643         translate_isl_ast_to_gimple::copy_cond_phi_args,
13644         translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13645         translate_isl_ast_to_gimple::edge_for_new_close_phis,
13646         translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13647         translate_isl_ast_to_gimple::rename_uses,
13648         translate_isl_ast_to_gimple::rename_all_uses): Remove.
13649         (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13650         (set_rename_for_each_def): Likewise.
13651         (graphite_copy_stmts_from_block): Handle debug stmt resetting
13652         here.  Handle rewriting SCEV analyzable uses here.
13653         (copy_bb_and_scalar_dependences): Generate code for PHI
13654         copy-in/outs.
13655         (graphite_regenerate_ast_isl): Adjust.
13656         * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13657         (add_write, add_read): New functions.
13658         (build_cross_bb_scalars_def): Use it and simplify.
13659         (build_cross_bb_scalars_use): Likewise.
13660         (graphite_find_cross_bb_scalar_vars): Inline into...
13661         (try_generate_gimple_bb): ...here.  Add dependences for PHIs,
13662         simulating out-of-SSA.  Compute liveout and add dependencies.
13663         (build_scops): Force an empty entry block.
13664         * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13665         members.
13666         (sese_build_liveouts): Declare.
13667         (sese_trivially_empty_bb_p): Likewise.
13668         * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13669         compute liveout and debug_liveout.
13670         (sese_bad_liveouts_use): Remove.
13671         (sese_reset_debug_liveouts_bb): Likewise.
13672         (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13673         (sese_build_liveouts): Build liveout and debug_liveout and store
13674         it in region.
13675         (new_sese_info): Adjust.
13676         (free_sese_info): Likewise.
13677         (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13678         do not build liveout here.
13679         (move_sese_in_condition): Adjust region entry.
13680         (scev_analyzable_p): Match up with chrec_apply requirements.
13681         (sese_trivially_empty_bb_p): New.
13682         * tree-into-ssa.c (get_reaching_def): Properly support generating
13683         default-defs for incremental rewrite of anonymous names.
13685 2017-10-06  Richard Biener  <rguenther@suse.de>
13687         * graphite-sese-to-poly.c (extract_affine): For casts increasing
13688         precision do not perform modulo reduction.
13690 2017-10-06  Richard Biener  <rguenther@suse.de>
13692         PR tree-optimization/82436
13693         * tree-vect-slp.c (vect_supported_load_permutation_p): More
13694         conservatively choose the vectorization factor when checking
13695         whether we can perform the required load permutation.
13696         (vect_transform_slp_perm_load): Assert when we may not fail.
13698 2017-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
13700         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13701         message for incompatible -msdata=* and -mcall-* options.
13703 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13705         * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13706         rate for post-reload scheduling.
13708 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13710         * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13712 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13714         * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13715         to improve monte carlo in scimark.
13717 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13719         * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13720         pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13721         athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13722         pentium4_cost, nocona_cost): Set reassociation width to 1.
13723         (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13724         width to 2 for fp operations and 1 otherwise.
13725         (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13726         for int and fp.
13727         (atom_cost): Set reassociation width to 2.
13728         (slm_cost, generic_cost): Set fp reassociation width
13729         to 2 and 1 otherwise.
13730         (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13731         (core_cost): Set fp reassociation width to 4 and vector to 2.
13732         (ix86_reassociation_width): Rewrite using cost table; special case
13733         plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13734         and TARGET_AVX128_OPTIMAL.
13735         * config/i386/i386.h (processor_costs): Add
13736         reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13737         (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13738         TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13739         * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13740         (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13741         (X86_TUNE_VECTOR_PARALLEL_EXECUTION):  Remove.
13743 2017-10-05  Nathan Sidwell  <nathan@acm.org>
13745         * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13747 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13749         * config/arm/arm.c (arm_test_fpu_data): New.
13750         (arm_run_selftests): Call arm_test_fpu_data.
13752 2017-10-04  Nathan Sidwell  <nathan@acm.org>
13754         * toplev.c (toplev::main): Remove excess parens on pretty_printer
13755         decl.
13756         * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13758 2017-10-04  Sudakshina Das  <sudi.das@arm.com>
13760         * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13761         check type for aarch64_simd_valid_immediate.
13762         (aarch64_output_simd_mov_immediate): Update prototype.
13763         (aarch64_simd_valid_immediate): Update prototype.
13764         * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13765         support for ORR-immediate.
13766         (and<mode>3): modified pattern to add support for BIC-immediate.
13767         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13768         now checks for valid immediate for BIC and ORR based on new enum
13769         argument.
13770         (aarch64_output_simd_mov_immediate): Function now used to output
13771         BIC/ORR imm as well based on new enum argument.
13772         * config/aarch64/constraints.md (Do): New vector immediate constraint.
13773         (Db) : Likewise.
13774         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13775         (aarch64_reg_or_bic_imm): Likewise.
13777 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13779         * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13780         ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13782 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13784         Revert r253399:
13786         PR rtl-optimization/82396
13787         * haifa-sched.c (autopref_multipass_init): Simplify
13788         initialization.
13789         (autopref_rank_data): Simplify sort order.
13790         * sched-int.h (autopref_multipass_data_): Remove
13791         multi_mem_insn_p, min_offset and max_offset.
13793 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13795         * doc/sourcebuild.texi: Document vect_peeling_profitable.
13797 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13799         * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13800         vect_doubleint_cvt.
13802 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13804         * doc/sourcebuild.texi: Document vect_long_mult.
13806 2017-10-04  Richard Sandiford  <richard.sandiford@linaro.org>
13808         PR tree-optimization/82413
13809         * fold-const.c (build_range_check): Use widest_int when comparing
13810         the maximum ETYPE value with HIGH.
13812 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13814         PR rtl-optimization/82396
13815         * haifa-sched.c (autopref_multipass_init): Simplify
13816         initialization.
13817         (autopref_rank_data): Simplify sort order.
13818         * sched-int.h (autopref_multipass_data_): Remove
13819         multi_mem_insn_p, min_offset and max_offset.
13821 2017-10-04  Jakub Jelinek  <jakub@redhat.com>
13823         PR tree-optimization/82381
13824         * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13825         oeN->rank first.  Return 1 or -1 if one op is SSA_NAME and the other
13826         is not.
13828         PR tree-optimization/82374
13829         * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13830         DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13831         DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13832         current_function_decl to the new decl.
13834 2017-10-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
13836         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13837         helper macro for IEEE float128 hardware built-in functions.
13838         (SQRTF128_ODD): Add built-in functions with the round-to-odd
13839         semantics.
13840         (TRUNCF128_ODD): Likewise.
13841         (ADDF128_ODD): Likewise.
13842         (SUBF128_ODD): Likewise.
13843         (MULF128_ODD): Likewise.
13844         (DIVF128_ODD): Likewise.
13845         (FMAF128_ODD): Likewise.
13846         * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13847         UNSPEC_TRUNC_ROUND_TO_ODD.
13848         (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13849         (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13850         floating point round to odd instructions.
13851         (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13852         (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13853         (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13854         (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13855         (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13856         (trunc<mode>sf2_hw): Change the truncate with round to odd
13857         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13858         (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13859         to odd hardware instructions.
13860         (sub<mode>3_odd): Likewise.
13861         (mul<mode>3_odd): Likewise.
13862         (div<mode>3_odd): Likewise.
13863         (sqrt<mode>2_odd): Likewise.
13864         (fma<mode>4_odd): Likewise.
13865         (fms<mode>4_odd): Likewise.
13866         (nfma<mode>4_odd): Likewise.
13867         (nfms<mode>4_odd): Likewise.
13868         (trunc<mode>df2_odd): Change the truncate with round to odd
13869         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.  Add a generator
13870         function.
13871         * doc/extend.texi (PowerPC built-in functions): Update documentation
13872         for existing IEEE float128-bit built-in functions.  Add built-in
13873         functions that generate the IEEE 128-bit floating point round to
13874         odd instructions.
13876 2017-10-03  Segher Boessenkool  <segher@kernel.crashing.org>
13878         PR rtl-optimization/77729
13879         * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13880         to (X&(C1&~C2))|C2 transformations.
13882 2017-10-03  Martin Jambor  <mjambor@suse.cz>
13884         PR tree-optimization/82363
13885         * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13886         mismatch, mark lacc written regardless of racc.
13888 2017-10-03  Jakub Jelinek  <jakub@redhat.com>
13890         PR tree-optimization/82381
13891         * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13892         stmt_to_insert nor wheather SSA_NAMEs are default defs.
13893         Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13894         fallthrough into reassoc_stmt_dominates_stmt_p.
13896         PR target/82386
13897         * combine.c (combine_instructions): Don't combine in unreachable
13898         basic blocks.
13900 2017-10-02  Peter Bergner  <bergner@vnet.ibm.com>
13902         PR target/80210
13903         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
13904         function to not use the have_cpu variable.  Do not set cpu_index,
13905         rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
13906         or the default cpu.
13907         (rs6000_valid_attribute_p): Remove duplicate initializations of
13908         old_optimize and func_optimize.
13909         (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
13910         (rs6000_activate_target_options): Make global.
13911         * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
13912         prototype.
13914 2017-10-02  Jakub Jelinek  <jakub@redhat.com>
13916         * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
13917         if *poffset + *pmax_size overflows in HOST_WIDE_INT.
13918         Set *poffset to 0 and *psize and *pmax_size to -1 if
13919         *poffset + *psize overflows in HOST_WIDE_INT.
13921         PR tree-optimization/82387
13922         PR tree-optimization/82388
13923         PR tree-optimization/82389
13924         * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
13925         instead of live_bytes non-NULL.
13927 2017-10-02  Georg-Johann Lay  <avr@gjlay.de>
13929         PR target/41076
13930         * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
13931         alternative.
13933 2017-10-02  Richard Biener  <rguenther@suse.de>
13935         * graphite-isl-ast-to-gimple.c (set_codegen_error): With
13936         -fchecking and --param graphite-allow-codegen-errors=0 ICE.
13937         * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
13939 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13941         * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
13942         requested precision matches the type's.
13943         * calls.c (alloc_max_size): Calculate the new candidate size as
13944         a widest_int and use wi::to_widest when comparing it with the
13945         current candidate size.
13946         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
13947         zero rather than integer_zero_node.
13948         * match.pd: Check for a no-op conversion before using wi::add
13949         rather than after.  Use tree_to_uhwi when summing small shift
13950         counts into an unsigned int.
13952 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13953             Alan Hayward  <alan.hayward@arm.com>
13954             David Sherwood  <david.sherwood@arm.com>
13956         PR target/71307
13957         * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
13958         (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
13959         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13960         POINTER_AND_FP_REGS.
13962 2017-10-02  Richard Biener  <rguenther@suse.de>
13964         PR tree-optimization/82355
13965         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
13966         a mapping for the enclosing loop but avoid generating one for
13967         the loop tree root.
13968         (copy_bb_and_scalar_dependences): Remove premature codegen
13969         error on PHIs in blocks duplicated into multiple places.
13970         * graphite-scop-detection.c
13971         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
13972         in the region use it as loop and nest to analyze the DR in.
13973         (try_generate_gimple_bb): Likewise.
13974         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
13975         (add_loop_constraints): For blocks in a loop not in the region
13976         create a dimension with a single iteration.
13977         * sese.h (gbb_loop_at_index): Remove assert.
13979 2017-10-01  Kevin Buettner  <kevinb@redhat.com>
13981         * omp-expand.c (adjust_context_scope): New function.
13982         (expand_parallel_call): Call adjust_context_scope.
13984 2017-10-01  Jeff Law  <law@redhat.com>
13986         * tree-ssa-dom.c (optimize_stmt): Make this a method within the
13987         dom_opt_dom_walker class with direct access to private members.
13988         Add comments.  Call test_for_singularity.
13989         (dom_opt_dom_walker::before_dom_children): Corresponding changes.
13990         (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
13991         m_dummy_cond anymore.
13992         (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
13993         class ctor.
13994         (pass_dominator:execute): Build the dummy_cond here and pass it
13995         to the dom_opt_dom_walker ctor.
13996         (test_for_singularity): New function.
13998 2017-09-30  Krister Walfridsson  <krister.walfridsson@gmail.com>
13999             Maya Rashish  <coypu@sdf.org>
14001         * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
14002         netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
14003         (alpha*-*-netbsd*) Use nbsd_tm_file.
14004         (arm*-*-netbsdelf*) Likewise.
14005         (i[34567]86-*-netbsdelf*) Likewise.
14006         (x86_64-*-netbsd*) Likewise.
14007         (mips*-*-netbsd*) Likewise.
14008         (powerpc-*-netbsd*) Likewise.
14009         (sh*-*-netbsd*) Likewise.
14010         (sparc-*-netbsdelf*) Likewise.
14011         (sparc64-*-netbsd*) Likewise.
14012         (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
14013         to tm_defines.
14014         (vax-*-netbsdelf*) Likewise.
14015         * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
14016         (UINT_FAST8_TYPE) Likewise.
14017         (INT_FAST16_TYPE) Check CHAR_FAST16.
14018         (UINT_FAST16_TYPE) Likewise.
14020 2017-09-30  Jakub Jelinek  <jakub@redhat.com>
14022         PR target/82361
14023         * config/i386/i386.md
14024         (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
14025         (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
14026         *divmodsi4_zext_2): New define_insn_and_split.
14027         (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
14028         (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
14029         (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
14030         *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
14031         New define_insn_and_split.
14032         (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
14033         * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
14034         operands[1] having DImode when mode is SImode.
14036         * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
14037         always SImode for DIV and MOD in REG_EQUAL notes.
14039 2017-09-29  Yury Gribov  <tetra2005@gmail.com>
14041         PR middle-end/82319
14042         * match.pd: Fix handling of NaNs in pattern.
14044 2017-09-29  Jeff Law  <law@redhat.com>
14046         * sbitmap.c (bitmap_bit_in_range_p): New function.
14047         * sbitmap.h (bitmap_bit_in_range_p): Prototype.
14048         * tree-ssa-dse.c (live_bytes_read): New function.
14049         (dse_classify_store): Ignore reads of dead bytes.
14051         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
14052         typos and whitespace errors.
14053         * config/i386/predicates.md (address_no_seg_operand): Likewise.
14054         * config/s390/s390.c (s390_emit_prologue): Likewise.
14056 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14058         PR target/81481
14059         * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
14060         with a symbol for LRA.
14062 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
14064         PR rtl-optimization/82338
14065         * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
14067 2017-09-29  Alexander Monakov  <amonakov@ispras.ru>
14069         * genmodes.c (calc_wider_mode): Suppress qsort macro.
14070         * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
14071         (qsort_chk): Declare.
14072         * vec.c [CHECKING_P] (qsort_chk_error): New static function.
14073         (qsort_chk): New function.
14075 2017-09-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
14077         PR tree-optimization/82337
14078         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
14079         phi definition if the PHI result appears in an abnormal PHI.
14080         (find_basis_for_base_expr): Don't record a basis if the LHS of the
14081         basis appears in an abnormal PHI.
14083 2017-09-29  Richard Biener  <rguenther@suse.de>
14085         * graphite-isl-ast-to-gimple.c
14086         (translate_isl_ast_to_gimple::set_codegen_error): New function.
14087         (binary_op_to_tree): Use it.
14088         (get_rename_from_scev): Likewise.
14089         (copy_loop_phi_nodes): Likewise.
14090         (copy_bb_and_scalar_dependences): Likewise.
14091         (translate_pending_phi_nodes): Likewise.
14093 2017-09-29  Jakub Jelinek  <jakub@redhat.com>
14095         PR target/82339
14096         * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
14097         for movabsq $(i32 << shift), r64.
14099 2017-09-28  Uros Bizjak  <ubizjak@gmail.com>
14101         * config/i386/i386.c (ix86_print_operand_address_as): Do not check
14102         index when encoding %esp as %rsp to avoid 0x67 prefix.
14104 2017-09-28  Sergey Shalnov  <Sergey.Shalnov@intel.com>
14106         * config/i386/i386.md (*movsf_internal, *movdf_internal):
14107         Return 256-bit AVX modes for TARGET_PREFER_AVX256.
14109 2017-09-28  Thomas Preud'homme  <thomas.preudhomme@arm.com>
14111         * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
14112         Extensions with more than 16 double VFP registers.
14113         (cmse_nonsecure_entry_clear_before_return): Remove second entry of
14114         to_clear_mask and all code related to it.  Replace the remaining
14115         entry by a sbitmap and adapt code accordingly.
14117 2017-09-28  Henry Linjamäki  <henry.linjamaki@parmance.com>
14119         * brig-builtins.def: Change pure attributes to const.
14121 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14123         * config.gcc (default_gnu_indirect_function): Default to yes for
14124         sparc*-*-linux* with glibc.
14126 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14128         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
14129         (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
14130         when creating .init_array and .fini_array sections with priority
14131         specified.
14133 2017-09-27  Christophe Lyon  <christophe.lyon@linaro.org>
14135         PR target/71727
14136         * config/aarch64/aarch64.c
14137         (aarch64_builtin_support_vector_misalignment): Always return false
14138         when misalignment is unknown.
14140 2017-09-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14142         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
14143         this function to return false if the definition used by the swap
14144         instruction is artificial, or if the memory address from which the
14145         constant value is loaded is not represented by a base address held
14146         in a register or if the base address register is a frame or stack
14147         pointer.  Additionally, return false if the base address of the
14148         loaded constant is a SYMBOL_REF but is not considered to be a
14149         constant.
14150         (replace_swapped_load_constant): New function.
14151         (rs6000_analyze_swaps): Add a new pass to replace a swap of a
14152         loaded constant vector with a load of a swapped constant vector.
14154 2017-09-27  Carl Love  <cel@us.ibm.com>
14156         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14157         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14158         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14159         fctiw instruction.
14161 2017-09-27  Alexander Monakov  <amonakov@ispras.ru>
14163         * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
14164         first, always call autopref_rank_data otherwise.
14166 2017-09-27  Richard Biener  <rguenther@suse.de>
14168         * graphite-scop-detection.c (find_scop_parameters): Move
14169         loop bound handling ...
14170         (gather_bbs::before_dom_children): ... here, avoiding the need
14171         to build scop_info->loop_nest.
14172         (record_loop_in_sese): Remove.
14173         * sese.h (sese_info_t::loop_nest): Remove.
14174         * sese.c (new_sese_info): Do not allocate loop_nest.
14175         (free_sese_info): Do not free loop_nest.
14177 2017-09-27  Jakub Jelinek  <jakub@redhat.com>
14179         PR c++/82159
14180         * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
14181         lhs from calls if the lhs has addressable type.
14183 2017-09-27  Richard Biener  <rguenther@suse.de>
14185         * graphite.h (scop::max_alias_set): New member.
14186         * graphite-scop-detection.c: Remove references to non-existing
14187         --param in comments.
14188         (build_alias_sets): Record the maximum alias set used for drs.
14189         (build_scops): Support zero as unlimited for
14190         --param graphite-max-arrays-per-scop.
14191         * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
14192         and inline into ...
14193         (build_poly_sr_1): ... here.  Compute alias set based on the
14194         maximum alias set used for drs rather than
14195         PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
14197 2017-09-27  Richard Biener  <rguenther@suse.de>
14199         * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
14200         --param loop-block-tile-size=0 to disable tiling.
14202 2017-09-27  Richard Biener  <rguenther@suse.de>
14204         * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
14205         (graphite-max-nb-scop-params): Document special value zero.
14206         * domwalk.h (dom_walker::STOP): New symbolical constant.
14207         (dom_walker::dom_walker): Add optional parameter for bb to
14208         RPO mapping.
14209         (dom_walker::~dom_walker): Declare.
14210         (dom_walker::before_dom_children): Document STOP return value.
14211         (dom_walker::m_user_bb_to_rpo): New member.
14212         (dom_walker::m_bb_to_rpo): Likewise.
14213         * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
14214         mapping here if not provided by the user.
14215         (dom_walker::~dom_walker): Free bb to RPO mapping if not
14216         provided by the user.
14217         (dom_walker::STOP): Define.
14218         (dom_walker::walk): Do not compute bb to RPO mapping here.
14219         Support STOP return value from before_dom_children to stop
14220         walking.
14221         * graphite-optimize-isl.c (optimize_isl): If the schedule
14222         is the same still generate code if -fgraphite-identity
14223         or -floop-parallelize-all are given.
14224         * graphite-scop-detection.c: Include cfganal.h.
14225         (gather_bbs::gather_bbs): Get and pass through bb to RPO
14226         mapping.
14227         (gather_bbs::before_dom_children): Return STOP for BBs
14228         not in the region.
14229         (build_scops): Compute bb to RPO mapping and pass it to
14230         the domwalk.  Treat --param graphite-max-nb-scop-params=0
14231         as not limiting the number of params.
14232         * graphite.c (graphite_initialize): Remove limit on the
14233         number of basic-blocks in a function.
14234         * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14235         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14236         default value of 10.
14238 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14240         * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14241         Adjust code to eliminate needing to do the shift right 32-bits
14242         operation after XSCVDPSPN.
14244 2017-09-26  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14246         * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14247         ((X / Y) != 0 -> X >= Y): Likewise.
14249 2017-09-26  Carl Love  <cel@us.ibm.com>
14251         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14252         P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14253         vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14254         void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14255         * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14256         * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14257         definitions and overloading.
14258         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14259         statement for P9V_BUILTIN_XST_LEN_R.
14260         (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14261         * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14262         define_expand and define_insn for the instructions and builtins.
14263         * doc/extend.texi: Update the built-in documentation file for the new
14264         built-in functions.
14265         * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14266         define_insn for the instructions
14268 2017-09-26  Krister Walfridsson  <krister.walfridsson@gmail.com>
14270         PR target/39570
14271         * gcc/config/netbsd-protos.h: New file.
14272         * gcc/config/netbsd.c: New file.
14273         * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14274         * gcc/config/t-netbsd: New file.
14275         * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14276         (tmake_file) Add t-netbsd.
14277         (extra_objs) Add netbsd.o.
14279 2017-09-26  Janus Weil  <janus@gcc.gnu.org>
14281         PR fortran/82143
14282         PR fortran/82324
14283         * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14285 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14287         * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14288         sign extension from a vector register to a GPR by doing a 32-bit
14289         direct move and then an EXTSW.
14290         (extendsi<mode>2 splitter): Likewise.
14291         (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14292         right or vector extract after doing XSCVDPSPN.  Use
14293         zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14294         the GPRs.
14295         (movdi_from_sf_zero_ext): Likewise.
14296         (reload_gpr_from_vsxsf): Likewise.
14297         (p8_mfvsrd_4_disf): Delete, no longer used.
14298         (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14299         then needing to move the SFmode to a GPR to use the XSCVDPSP
14300         instruction instead of FRSP and XSCVDPSPN.
14301         * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14302         it is adjacent to the other XSCVSPDP insns.
14303         (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14304         SFmode to be in traditional Altivec registers.
14305         (vsx_xscvdpspn): Eliminate useless alternative constraint.
14306         (vsx_xscvspdpn): Likewise.
14307         (vsx_xscvspdpn_scalar): Likewise.
14309 2017-09-26  Martin Jambor  <mjambor@suse.cz>
14311         * tree-sra.c (compare_access_positions): Put integral types first,
14312         stabilize sorting of integral types, remove conditions putting
14313         non-full-precision integers last.
14314         (sort_and_splice_var_accesses): Disable scalarization if a
14315         non-integert would be represented by a non-full-precision integer.
14317 2017-09-26  Joseph Myers  <joseph@codesourcery.com>
14319         * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14320         * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14321         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14322         * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14323         Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14324         conditionals inside the function instead of around it.  Call
14325         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14326         (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14328 2017-09-26  Richard Biener  <rguenther@suse.de>
14330         * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14331         fold in ...
14332         (scop_detection::build_scop_breadth): ... this.  Removed.
14333         (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14334         (scop_detection::harmful_stmt_in_bb): Likewise.
14335         (scop_detection::graphite_can_represent_stmt): Likewise.
14336         (scop_detection::loop_body_is_valid_scop): Likewise.  Remove recursion.
14337         (scop_detection::can_represent_loop): Remove recursion, fold in ...
14338         (scop_detection::can_represent_loop_1): ... this.  Removed.
14339         (scop_detection::harmful_loop_in_region): Simplify after inlining
14340         the above and remove more quadraticness.
14341         (build_scops): Adjust.
14342         * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14343         quadraticness.
14345 2017-09-26  Jakub Jelinek  <jakub@redhat.com>
14347         PR target/82267
14348         * config/i386/i386.c (ix86_print_operand_address_as): Only test
14349         REGNO (base) == SP_REG if base is a REG.
14351         PR middle-end/35691
14352         * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14353         if it is different SSA_NAME.
14354         (optimize_range_tests_cmp_bitwise): New function.
14355         (optimize_range_tests): Call it.
14357 2017-09-26  Richard Biener  <rguenther@suse.de>
14359         PR tree-optimization/82321
14360         * graphite.c (canonicalize_loop_closed_ssa): Properly check
14361         for the def being inside the loop.
14363 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14365         * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14366         assembler output.
14367         * config/s390/s390-builtins.def: Fix constraint on op4.
14369 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14371         * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14372         independent expanders.
14373         * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14374         ("vec_ordered", "vec_unordered"): New expanders.
14376 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14378         * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14379         for SFmode.
14381 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14383         * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14384         vec_unpacks_lo_v16qi.
14385         ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14387 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14389         * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14390         ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14391         ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14393 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14395         * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14396         predicate.
14397         * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14398         and V16QI.
14399         ("*vec_slb<mode>"): New insn pattern.
14400         ("vec_shr_<mode>"): New expander.
14401         * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14402         and force the shift count operand to V16QImode.
14403         ("vec_srb<mode>"): Set shift count mode to V16QI.
14405 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14407         * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14408         ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14409         ("vec_widen_smult_hi_<mode>"): New expander definitions.
14411 2017-09-26  Richard Earnshaw  <rearnsha@arm.com>
14413         PR target/82175
14414         * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14416 2017-09-26  Richard Biener  <rguenther@suse.de>
14418         PR tree-optimization/82320
14419         * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14420         isn't a change.
14422 2017-09-25  Jeff Law  <law@redhat.com>
14424         * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14425         prototype for new argument.
14426         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14427         mostly extracted from rs6000_emit_allocate_stack.
14428         (rs6000_emit_probe_stack_range_stack_clash): New function.
14429         (rs6000_emit_allocate_stack): Call
14430         rs6000_emit_probe_stack_range_stack_clash as needed.
14431         (rs6000_emit_probe_stack_range): Add additional argument
14432         to call to gen_probe_stack_range{si,di}.
14433         (output_probe_stack_range): New.
14434         (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14435         (output_probe_stack_range_stack_clash): New.
14436         (rs6000_emit_prologue): Emit notes into dump file as requested.
14437         * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14438         (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14439         Add additional operand and pass it to output_probe_stack_range.
14441 2017-09-25  Bin Cheng  <bin.cheng@arm.com>
14443         PR tree-optimization/82163
14444         * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14445         (checking_verify_loop_closed_ssa): New parameter.
14446         * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14447         (check_loop_closed_ssa_stmt): Delete.
14448         (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14449         (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14450         (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14451         changed loops.
14453 2017-09-25  Pekka Jaaskelainen <pekka@parmance.com>
14455         * brig-builtins.def: Treat HSAIL barrier builtins as
14456         setjmp/longjump style functions.
14458 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14460         * target.def (constant_alignment): New hook.
14461         * defaults.h (CONSTANT_ALIGNMENT): Delete.
14462         * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14463         (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14464         * doc/tm.texi: Regenerate.
14465         * targhooks.h (default_constant_alignment): Declare.
14466         (constant_alignment_word_strings): Likewise.
14467         * targhooks.c (default_constant_alignment): New function.
14468         (constant_alignment_word_strings): Likewise.
14469         * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14470         instead of CONSTANT_ALIGNMENT.
14471         * varasm.c (align_variable, get_variable_align, build_constant_desc)
14472         (force_const_mem): Likewise.
14473         * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14474         * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14475         (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14476         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14477         * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14478         definition.
14479         * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14480         * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14481         constant_alignment_word_strings.
14482         * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14483         (CONSTANT_ALIGNMENT): Likewise.
14484         * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14485         (arm_constant_alignment): New function.
14486         * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14487         * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14488         constant_alignment_word_strings.
14489         * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14490         * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14491         constant_alignment_word_strings.
14492         * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14493         * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14494         (cris_constant_alignment): New function.
14495         * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14496         * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14497         (epiphany_constant_alignment): New function.
14498         * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14499         * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14500         constant_alignment_word_strings.
14501         * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14502         * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14503         constant_alignment_word_strings.
14504         * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14505         * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14506         constant_alignment_word_strings.
14507         * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14508         * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14509         * config/i386/i386.c (ix86_constant_alignment): Make static.
14510         Use the same interface as the target hook.
14511         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14512         * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14513         * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14514         constant_alignment_word_strings.
14515         * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14516         * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14517         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14518         * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14519         * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14520         constant_alignment_word_strings.
14521         * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14522         * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14523         constant_alignment_word_strings.
14524         * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14525         * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14526         constant_alignment_word_strings.
14527         * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14528         * config/microblaze/microblaze.c (microblaze_constant_alignment):
14529         New function.
14530         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14531         * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14532         * config/mips/mips.c (mips_constant_alignment): New function.
14533         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14534         * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14535         * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14536         * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14537         (mmix_constant_alignment): Make static.  Use the same interface
14538         as the target hook.
14539         * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14540         * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14541         constant_alignment_word_strings.
14542         * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14543         * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14544         constant_alignment_word_strings.
14545         * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14546         * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14547         constant_alignment_word_strings.
14548         * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14549         * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14550         (rs6000_constant_alignment): New function.
14551         * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14552         * config/riscv/riscv.c (riscv_constant_alignment): New function.
14553         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14554         * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14555         * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14556         (rs6000_constant_alignment): New function.
14557         * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14558         * config/s390/s390.c (s390_constant_alignment): New function.
14559         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14560         * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14561         * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14562         constant_alignment_word_strings.
14563         * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14564         * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14565         (sparc_constant_alignment): New function.
14566         * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14567         * config/spu/spu.c (spu_constant_alignment): New function.
14568         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14569         * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14570         * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14571         constant_alignment_word_strings.
14572         * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14573         * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14574         constant_alignment_word_strings.
14575         * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14576         * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14577         constant_alignment_word_strings.
14578         * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14579         * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14580         (visium_constant_alignment): New function.
14581         * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14582         * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14583         (xtensa_constant_alignment): New function.
14584         * system.h (CONSTANT_ALIGNMENT): Poison.
14586 2017-09-25  Will Schmidt  <will_schmidt@vnet.ibm.com>
14588         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14589         for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14590         (rs6000_builtin_valid_without_lhs): New helper function.
14591         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14592         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14594 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14596         * target.h (vec_perm_indices): Use unsigned short rather than
14597         unsigned char.
14598         (auto_vec_perm_indices): Likewise.
14599         * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14600         Use unsigned int rather than unsigned char.
14601         * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14603 2017-09-25  Richard Biener  <rguenther@suse.de>
14605         * cfgloop.h (sort_sibling_loops): Declare.
14606         * cfgloop.c (sort_sibling_loops_cmp): New helper.
14607         (sort_sibling_loops): New function sorting the sibling loop list
14608         in RPO order.
14609         * graphite.c (graphite_transform_loops): Sort sibling loops.
14611 2017-09-25  Richard Sandiford  <richard.sandifird@linaro.org>
14613         * target.def (vec_perm_const_ok): Change sel parameter to
14614         vec_perm_indices.
14615         * optabs-query.c (can_vec_perm_p): Update accordingly.
14616         * doc/tm.texi: Regenerate.
14617         * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14618         auto_vec_perm_indices and remove separate nelt field.
14619         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14620         (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14621         (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14622         (aarch64_expand_vec_perm_const): Update accordingly.
14623         (aarch64_vectorize_vec_perm_const_ok): Likewise.  Change sel
14624         to vec_perm_indices.
14625         * config/arm/arm.c (expand_vec_perm_d): Change perm to
14626         auto_vec_perm_indices and remove separate nelt field.
14627         (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14628         (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14629         (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14630         accordingly.
14631         (arm_vectorize_vec_perm_const_ok): Likewise.  Change sel
14632         to vec_perm_indices.
14633         * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14634         sel to vec_perm_indices.
14635         * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14636         * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14637         * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14638         Likewise.
14639         * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14640         Likewise.
14642 2017-09-25  Pierre-Marie de Rodat  <derodat@adacore.com>
14644         PR debug/82155
14645         * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14646         on the FUNCTION_DECL function context if it has a DIE that is a
14647         declaration.
14649 2017-09-25  Richard Biener  <rguenther@suse.de>
14651         PR tree-optimization/82285
14652         * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14653         enumeral types.
14655 2017-09-25  Tom de Vries  <tom@codesourcery.com>
14657         PR target/80035
14658         PR target/81069
14659         * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14660         noreturn function.
14662 2017-09-25  Richard Biener  <rguenther@suse.de>
14664         * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14665         ISL errors other than isl_error_quota happen.  Dump if the
14666         schedule is the same.
14667         * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14668         errors instead of aborting inside ISL.
14670 2017-09-25  Iain Sandoe  <iain@codesourcery.com>
14672         PR target/80556
14673         * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14674         of libgcc_eh for m64.
14675         * config/i386/darwin64.h: Likewise.
14677 2017-09-25  Richard Biener  <rguenther@suse.de>
14679         PR middle-end/82144
14680         * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14681         attribute for incomplete types nor twice for complete ones.
14683 2017-09-24  Uros Bizjak  <ubizjak@gmail.com>
14685         PR target/82267
14686         * config/i386/i386.c (ix86_print_operand_address_as): Encode
14687         %esp as %rsp to avoid 0x67 prefix if there is no index or base
14688         register.
14690 2017-09-23  Uros Bizjak  <ubizjak@gmail.com>
14692         PR bootstrap/82306
14693         * config/i386/i386.opt (mprefer-avx256): Use
14694         ix86_target_flags variable.
14695         * config/i386/i386.c (ix86_target_string): Move
14696         -mprefer-avx256 to flag2_opts.
14698 2017-09-22  Jakub Jelinek  <jakub@redhat.com>
14700         PR middle-end/35691
14701         * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14702         and x != -1 | y != -1 into (x & y) != -1.
14704 2017-09-22  Steve Ellcey  <sellcey@cavium.com>
14706         * config.gcc: Add new case statement to set
14707         default_gnu_indirect_function.  Remove it from x86_64-*-linux*,
14708         i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14709         s390x-*-linux* case statements.   Added aarch64 to the list of
14710         supported architectures.
14712 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14714         PR tree-optimization/82289
14715         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14716         STMT_VINFO_RELEVANT_P.
14718 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14719             Alan Hayward  <alan.hayward@arm.com>
14720             David Sherwood  <david.sherwood@arm.com>
14722         * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14723         for VR_RANGE only; don't allow VR_ANTI_RANGE.
14724         (extract_range_from_binary_expr_1): Don't call
14725         extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14727 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14728             Alan Hayward  <alan.hayward@arm.com>
14729             David Sherwood  <david.sherwood@arm.com>
14731         * target.def (preferred_vector_alignment): New hook.
14732         * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14733         hook.
14734         * doc/tm.texi: Regenerate.
14735         * targhooks.h (default_preferred_vector_alignment): Declare.
14736         * targhooks.c (default_preferred_vector_alignment): New function.
14737         * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14738         Expand commentary.
14739         (DR_TARGET_ALIGNMENT): New macro.
14740         (aligned_access_p): Update commentary.
14741         (vect_known_alignment_in_bytes): New function.
14742         * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14743         function.
14744         (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14745         Calculate the misalignment based on the target alignment rather than
14746         the vector size.
14747         (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14748         rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14749         (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14750         the target alignment, rather than masking the element misalignment
14751         with the number of elements in a vector.  Also use the target
14752         alignment when calculating the maximum number of peels.
14753         (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14754         instead of TYPE_ALIGN_UNIT.
14755         (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14756         Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14757         (vect_create_addr_base_for_vector_ref): Update call accordingly.
14758         (vect_create_data_ref_ptr): Likewise.
14759         (vect_setup_realignment): Realign by ANDing with
14760         -DR_TARGET_MISALIGNMENT.
14761         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14762         the number of peels based on DR_TARGET_ALIGNMENT.
14763         * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14764         with the guaranteed alignment boundary when deciding whether
14765         overrun is OK.
14766         (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14767         relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14768         (ensure_base_align): Remove stmt_info parameter.  Get the
14769         target base alignment from DR_TARGET_ALIGNMENT.
14770         (vectorizable_store): Update call accordingly.   Interpret
14771         DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14772         TYPE_ALIGN_UNIT.
14773         (vectorizable_load): Likewise.
14775 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14776             Alan Hayward  <alan.hayward@arm.com>
14777             David Sherwood  <david.sherwood@arm.com>
14779         * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14780         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14781         (vect_enhance_data_refs_alignment): Likewise.
14783 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14785         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14786         error.  Only quit immediately if parsing is complete.
14787         (BEGIN): Initialize fatal_err and parse_done.
14788         (begin fpu, end fpu): Check number of arguments.
14789         (begin arch, end arch): Likewise.
14790         (begin cpu, end cpu): Likewise.
14791         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14792         (optalias): Likewise.
14794 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14796         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14797         * config/arm/arm-isa.h: Delete.  Move definitions to ...
14798         * arm-cpus.in: ... here.  Use new feature and fgroup values.
14799         * config/arm/arm.c (arm_option_override): Use lower case for feature
14800         bit names.
14801         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14802         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14803         * config/arm/parsecpu.awk (END): Add new command 'isa'.
14804         (isa_pfx): Delete.
14805         (print_isa_bits_for): New function.
14806         (gen_isa): New function.
14807         (gen_comm_data): Use print_isa_bits_for.
14808         (define feature): New keyword.
14809         (define fgroup): New keyword.
14810         * config/arm/t-arm (TM_H): Remove.
14811         (GTM_H): Add arm-isa.h.
14812         (arm-isa.h): Add rule to generate file.
14813         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14814         case for feature bit names.
14816 2017-09-22  Richard Biener  <rguenther@suse.de>
14818         * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14819         single caller.
14820         (graphite_regenerate_ast_isl): Do not reset SCEV.  Move debug
14821         print of no dependency loops ...
14822         * graphite.c (graphite_transform_loops): ... here.
14823         (canonicalize_loop_closed_ssa_form): Work from inner to outer
14824         loops.
14825         (same_close_phi_node, remove_duplicate_close_phi,
14826         make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14827         (canonicalize_loop_closed_ssa): ... here and simplify.
14828         * graphite-optimize-isl.c: Include tree-vectorizer.h.
14829         (optimize_isl): Use dump_printf_loc to tell when we stopped
14830         optimizing because of an ISL timeout.
14832 2017-09-22  Richard Biener  <rguenther@suse.de>
14834         PR tree-optimization/82291
14835         * tree-if-conv.c (predicate_mem_writes): Make sure to
14836         remove writes in blocks predicated with false.
14838 2017-09-22  Richard Biener  <rguenther@suse.de>
14840         * sese.c: Include cfganal.h.
14841         (if_region_set_false_region): Remove.
14842         (create_if_region_on_edge): Likewise.
14843         (move_sese_in_condition): Re-implement without destroying
14844         dominators.
14846 2017-09-22  Richard Biener  <rguenther@suse.de>
14848         * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14849         Verify both BBs contain loop PHI nodes before dispatching to
14850         copy_loop_phi_args.
14851         (graphite_regenerate_ast_isl): Do not recompute dominators,
14852         do not verify three times.  Restructure for clarity.
14853         * graphite-scop-detection.c (same_close_phi_node,
14854         remove_duplicate_close_phi, make_close_phi_nodes_unique,
14855         defined_in_loop_p, canonicalize_loop_closed_ssa,
14856         canonicalize_loop_closed_ssa_form): Simplify, remove excess
14857         checking and SSA rewrite, move to ...
14858         * graphite.c: ... here.  Include ssa.h and tree-ssa-loop-manip.h.
14859         (graphite_initialize): Do not pass in ctx, do not reset the
14860         SCEV cache, compute only dominators.
14861         (graphite_transform_loops): Allocate ISL ctx after
14862         graphite_initialize.  Call canonicalize_loop_closed_ssa_form.
14863         Maintain post-dominators only around build_scops.
14864         * sese.c (if_region_set_false_region): Make static.  Free
14865         and recompute dominators.
14866         (move_sese_in_condition): Assert we don't get called with
14867         post-dominators computed.
14868         * sese.h (if_region_set_false_region): Remove.
14870 2017-09-22  Sergey Shalnov  <sergey.shalnov@intel.com>
14872         * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14873         mode attribute for TARGET_AVX512VL.
14875 2017-09-21  Sergey Shalnov  <sergey.shalnov@intel.com>
14877         * config/i386/i386.opt (mprefer-avx256): New option.
14878         * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14879         to flag_opts.
14880         (ix86_preferred_simd_mode): Return 256-bit AVX modes
14881         for TARGET_PREFER_AVX256.
14882         * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14884 2017-09-21  Jeff Law  <law@redhat.com>
14886         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14887         Fix dump output if the only stack space is for pushed registers.
14889 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14891         * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14892         of insn_cost.
14894 2017-09-21  Martin Sebor  <msebor@redhat.com>
14896         PR c/81882
14897         * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
14898         code (in C++) or code that triggers warnings.
14900 2017-09-21  Eric Botcazou  <ebotcazou@adacore.com>
14902         * stor-layout.c (bit_from_pos): Do not distribute the conversion.
14904 2017-09-21  Segher Boessenkool  <segher@kernel.crashing.org>
14906         * haifa-sched.c: Rename insn_cost to insn_sched_cost.
14907         * sched-rgn.c: Ditto.
14908         * sel-sched-ir.c: Ditto.
14910 2017-09-21  Alexander Monakov  <amonakov@ispras.ru>
14912         * toplev.h (set_random_seed): Adjust return type.
14913         * toplev.c (init_local_tick): Move eager initialization of random_seed
14914         to get_random_seed.  Adjust comment.
14915         (init_random_seed): Inline to get_random_seed, delete.
14916         (get_random_seed): Initialize random_seed lazily.
14917         (set_random_seed): Do not return previous value.
14918         (print_switch_value): Do not call get_random_seed.
14920 2017-09-21  Evgeny Kudryashov  <kudryashov@ispras.ru>
14922         * cgraph.c (delete_function_version): New, broken out from...
14923         (cgraph_node::delete_function_version): ...here.  Rename to
14924         cgraph_node::delete_function_version_by_decl.  Update all uses.
14925         (cgraph_node::remove): Call delete_function_version.
14927 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14929         PR sanitizer/81715
14930         * tree-inline.c (expand_call_inline): Emit clobber stmts for
14931         VAR_DECLs to which addressable non-volatile parameters are mapped
14932         and for id->retvar after the return value assignment.  Clear
14933         id->retval and id->retbnd after inlining.
14935 2017-09-21  Richard Biener  <rguenther@suse.de>
14937         PR tree-optimization/82276
14938         PR tree-optimization/82244
14939         * tree-vrp.c (build_assert_expr_for): Set
14940         SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
14941         has it set.
14942         (remove_range_assertions): Revert earlier change.
14944 2017-09-21  Wilco Dijkstra  <wdijkstr@arm.com>
14946         PR target/71951
14947         * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
14949 2017-09-21  Richard Biener  <rguenther@suse.de>
14951         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
14952         Restore valid IL after code generation errors.
14953         * graphite.c (graphite_transform_loops): Diagnose code
14954         generation issues as MSG_MISSED_OPTIMIZATION and continue
14955         with processing SCOPs.
14957 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14958             Alan Hayward  <alan.hayward@arm.com>
14959             David Sherwood  <david.sherwood@arm.com>
14961         * calls.c (compute_argument_addresses): Use simplify_gen_binary
14962         rather than choosing between plus_constant and gen_rtx_<CODE>.
14963         * expr.c (emit_push_insn): Likewise.
14964         (expand_expr_real_2): Likewise.
14966 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14967             Alan Hayward  <alan.hayward@arm.com>
14968             David Sherwood  <david.sherwood@arm.com>
14970         * loop-unroll.c (split_iv): Call copy_rtx on the step.
14972 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14973             Alan Hayward  <alan.hayward@arm.com>
14974             David Sherwood  <david.sherwood@arm.com>
14976         * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
14977         calling tree_to_uhwi.
14979 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14980             Alan Hayward  <alan.hayward@arm.com>
14981             David Sherwood  <david.sherwood@arm.com>
14983         * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
14984         INTEGER_CST rather than a negative test for ADDR_EXPR.
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-vrp.c (extract_range_from_binary_expr_1): Check
14991         int_cst_rangeN before calling value_range_constant_singleton (&vrN).
14993 2017-09-21  Richard Biener  <rguenther@suse.de>
14995         PR tree-optimization/71351
14996         * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
14997         graphite_create_new_loop_guard): Remove, fold remaining parts
14998         into caller ...
14999         (translate_isl_ast_node_for): ... here and simplify.
15001 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
15003         PR target/82260
15004         * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
15005         with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
15006         latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
15007         alternative always use QI mode, for -Os imov (=R,R) alternative
15008         always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
15009         ignore -Os.
15011 2017-09-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
15012             Jeff Law  <law@redhat.com>
15014         * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
15015         (allocate_stack_space): New function, partially extracted from
15016         s390_emit_prologue.
15017         (s390_emit_prologue): Track offset to most recent stack probe.
15018         Code to allocate space moved into allocate_stack_space.
15019         Dump actions when no stack is allocated.
15020         (s390_prologue_plus_offset): New function.
15021         (s390_emit_stack_probe): Likewise.
15023 2017-09-20  Alexandre Oliva  <aoliva@redhat.com>
15025         * common.opt (Wa, Wl, Wp, g, gz=): Add
15026         RejectNegative.
15027         (gno-column-info): Remove.
15028         (gcolumn-info): Drop RejectNegative.
15029         (gno-): New prefix.
15030         (gno-record-gcc-switches): Remove.
15031         (grecord-gcc-switches): Drop RejectNegative.
15032         (gno-split-dwarf): Remove.
15033         (gsplit-dwarf): Drop RejectNegative.
15034         (gno-strict-dwarf): Remove.
15035         (gstrict-dwarf): Drop RejectNegative.
15036         * config/darwin.opt (gfull, gused): Add RejectNegative.
15037         * dwarf2out.c (gen_producer_string): Drop
15038         gno-record-gcc-switches handler.
15039         * optc-gen.awk: Add g to prefixes with negative forms.
15040         * opts-common.c (remapping_prefix_p): New.
15041         (find_opt): Check it.
15042         (generate_canonical_option): Test g prefix.
15043         (option_map): Add -gno- mapping.
15044         (add_misspelling_candidates): Check remapping_prefix_p.
15046 2017-09-20  Jeff Law  <law@redhat.com>
15048         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
15049         thinko in stack clash protection support.
15051         * explow.c (compute_stack_clash_protection_loop_data): Use
15052         CONST_INT_P instead of explicit test.  Verify object is a
15053         CONST_INT_P before looking at INTVAL.
15054         (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
15055         instead of explicit test.
15057 2017-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
15059         PR target/77687
15060         * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
15061         address instead of to r1 and r11.
15063 2017-09-20  Sebastian Peryt  <sebastian.peryt@intel.com>
15065         * config.gcc: Support "knm".
15066         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
15067         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
15068         PROCESSOR_KNM.
15069         * config/i386/i386.c (m_KNM): Define.
15070         (processor_target_table): Add "knm".
15071         (PTA_KNM): Define.
15072         (ix86_option_override_internal): Add "knm".
15073         (ix86_issue_rate): Add PROCESSOR_KNM.
15074         (ix86_adjust_cost): Ditto.
15075         (ia32_multipass_dfa_lookahead): Ditto.
15076         (get_builtin_code_for_version): Handle PROCESSOR_KNM.
15077         (fold_builtin_cpu): Add M_INTEL_KNM.
15078         * config/i386/i386.h (processor_costs): Define TARGET_KNM.
15079         (processor_type): Add PROCESSOR_KNM.
15080         * config/i386/x86-tune.def: Add m_KNM.
15081         * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
15083 2017-09-20  Richard Biener  <rguenther@suse.de>
15085         PR tree-optimization/80213
15086         * graphite-scop-detection.c (trivially_empty_bb_p): Labels
15087         are allowed in empty BBs as well.
15088         (canonicalize_loop_closed_ssa): Also look for other complex
15089         edges.
15090         (scop_detection::get_sese): Include the loop-closed PHI block
15091         in loop SESEs.
15092         (scop_detection::merge_sese): Remove code adding extra blocks.
15093         (scop_detection::region_has_one_loop): Adjust for get_sese changes.
15094         (build_scops): Assert the final returned scop is invalid.
15096 2017-09-20  Richard Biener  <rguenther@suse.de>
15098         PR tree-optimization/82264
15099         * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
15100         for GIMPLE_CONDs.
15101         (vn_phi_lookup): Likewise.
15102         (vn_phi_insert): Likewise.
15104 2017-09-20  Jakub Jelinek  <jakub@redhat.com>
15106         * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
15107         that fits into uhwi or shwi, add DW_AT_const_value regardless
15108         of early_dwarf without going through RTL, using add_AT_unsigned
15109         or add_AT_int.
15111         * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
15112         (DEBUG_LTO_ABBREV_SECTION): Likewise.
15113         (DEBUG_LTO_MACINFO_SECTION): Likewise.
15114         (DEBUG_MACRO_SECTION): Likewise.
15115         (DEBUG_LTO_MACRO_SECTION): Likewise.
15116         (DEBUG_STR_DWO_SECTION): Likewise.
15117         (DEBUG_LTO_STR_DWO_SECTION): Likewise.
15118         (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
15119         (DEBUG_LTO_DWO_LINE_SECTION): Define.
15120         (DEBUG_LTO_LINE_STR_SECTION): Define.
15121         (init_sections_and_labels): Initialize debug_line_str_section
15122         variable.  Initialize debug_loc_section for -gdwarf-5 to
15123         DEBUG_LOCLISTS_SECTION.  Formatting fixes.
15125 2017-09-20  Richard Biener  <rguenther@suse.de>
15127         * graphite-sese-to-poly.c (extract_affine): Properly handle
15128         POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
15130 2017-09-20  Richard Biener  <rguenther@suse.de>
15132         PR tree-optimization/81373
15133         * graphite-scop-detection.c (build_cross_bb_scalars_def):
15134         Force SESE live-out defs to be handled even if they are
15135         scev_analyzable_p.
15137 2017-09-19  Jeff Law  <law@redhat.com>
15139         * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
15140         nothing for stack adjustments with REG_STACK_CHECK.
15141         * sched-deps.c (parse_add_or_inc): Reject insns with
15142         REG_STACK_CHECK from dependency breaking.
15143         * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
15144         (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
15145         * reg-notes.def (STACK_CHECK): New note.
15147         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
15148         (ix86_expand_prologue): Dump stack clash info as needed.
15149         Call ix86_adjust_stack_and_probe_stack_clash as needed.
15151         * function.c (dump_stack_clash_frame_info): New function.
15152         * function.h (dump_stack_clash_frame_info): Prototype.
15153         (enum stack_clash_probes): New enum.
15155         * config/alpha/alpha.c (alpha_expand_prologue): Also check
15156         flag_stack_clash_protection.
15157         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
15158         (arm_expand_prologue, thumb1_expand_prologue): Likewise.
15159         (arm_frame_pointer_required): Likewise.
15160         * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
15161         (ia64_expand_prologue): Likewise.
15162         * config/mips/mips.c (mips_expand_prologue): Likewise.
15163         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
15164         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15165         (sparc_flat_expand_prologue): Likewise.
15166         * config/spu/spu.c (spu_expand_prologue): Likewise.
15168         * explow.c: Include "params.h".
15169         (anti_adjust_stack_and_probe_stack_clash): New function.
15170         (get_stack_check_protect): Likewise.
15171         (compute_stack_clash_protection_loop_data): Likewise.
15172         (emit_stack_clash_protection_loop_start): Likewise.
15173         (emit_stack_clash_protection_loop_end): Likewise.
15174         (allocate_dynamic_stack_space): Use get_stack_check_protect.
15175         Use anti_adjust_stack_and_probe_stack_clash.
15176         * explow.h (compute_stack_clash_protection_loop_data): Prototype.
15177         (emit_stack_clash_protection_loop_start): Likewise.
15178         (emit_stack_clash_protection_loop_end): Likewise.
15179         * rtl.h (get_stack_check_protect): Prototype.
15180         * target.def (stack_clash_protection_final_dynamic_probe): New hook.
15181         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
15182         * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
15183         Prototype.
15184         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
15185         Add @hook.
15186         * doc/tm.texi: Rebuilt.
15187         * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
15188         get_stack_check_protect.
15189         * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
15190         * config/arm/arm.c (arm_expand_prologue): Likewise.
15191         (arm_frame_pointer_required): Likewise.
15192         * config/i386/i386.c (ix86_expand_prologue): Likewise.
15193         * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
15194         * config/mips/mips.c (mips_expand_prologue): Likewise.
15195         * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
15196         * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
15197         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15198         (sparc_flat_expand_prologue): Likewise.
15200         * common.opt (-fstack-clash-protection): New option.
15201         * flag-types.h (enum stack_check_type): Note difference between
15202         -fstack-check= and -fstack-clash-protection.
15203         * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
15204         (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
15205         * toplev.c (process_options): Issue warnings/errors for cases
15206         not handled with -fstack-clash-protection.
15207         * doc/invoke.texi (-fstack-clash-protection): Document new option.
15208         (-fstack-check): Note additional problem with -fstack-check=generic.
15209         Note that -fstack-check is primarily for Ada and refer users
15210         to -fstack-clash-protection for stack-clash-protection.
15211         Document new params for stack clash protection.
15213 2017-09-19  Uros Bizjak  <ubizjak@gmail.com>
15215         * config/i386/i386.c (ix86_split_long_move): Do not handle
15216         address used for LEA in a special way.
15218 2017-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
15220         * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
15222 2017-09-19  Martin Sebor  <msebor@redhat.com>
15224         PR c/81854
15225         * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15226         of incompatible types.
15228 2017-09-19  Will Schmidt  <will_schmidt@vnet.ibm.com>
15230         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15231         for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15232         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15233         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15235 2017-09-19  Richard Biener  <rguenther@suse.de>
15237         PR tree-optimization/82244
15238         * tree-vrp.c (remove_range_assertions): Do not propagate
15239         a constant to abnormals but replace the assert with a copy.
15241 2017-09-19  Alexander Monakov  <amonakov@ispras.ru>
15243         PR rtl-optimization/57878
15244         PR rtl-optimization/68988
15245         * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15246         avoidance test involving non_reload_pseudos.  Move frequency test
15247         below the general fragmentation avoidance test.
15249 2017-09-19  Richard Biener  <rguenther@suse.de>
15251         PR tree-optimization/69728
15252         * graphite-sese-to-poly.c (schedule_error): New global.
15253         (add_loop_schedule): Handle empty domain by failing the
15254         schedule.
15255         (build_original_schedule): Handle schedule_error.
15257 2017-09-19  Richard Biener  <rguenther@suse.de>
15259         * graphite-scop-detection.c (scop_detection::can_represent_loop):
15260         Do not iterate to sibling loops but only to siblings of inner
15261         loops.
15263 2017-09-18  Andreas Schwab  <schwab@linux-m68k.org>
15265         PR target/81613
15266         * config/m68k/m68k.md (moveq feeding equality comparison): Check
15267         that the registers are different.
15269 2017-09-18  Uros Bizjak  <ubizjak@gmail.com>
15271         * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15272         to processor_model and "amdfam17h" to arch_names_table.
15273         * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15275 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15277         PR c/82234
15278         * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15280 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15281             Alan Hayward  <alan.hayward@arm.com>
15282             David Sherwood  <david.sherwood@arm.com>
15284         * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15285         with a vec_info *.
15286         * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15287         accordingly.
15288         * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15289         parameter.  Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15290         vect_schedule_slp_instance.
15291         (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15292         Update call to vect_slp_analyze_node_operations.  Simplify return
15293         value.
15294         (vect_slp_analyze_bb_1): Update call accordingly.
15295         (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15296         Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15297         (vect_schedule_slp): Update call accordingly.
15299 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15300             Alan Hayward  <alan.hayward@arm.com>
15301             David Sherwood  <david.sherwood@arm.com>
15303         * tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
15304         with types that aren't in fact scalar.
15306 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15308         * tree-vect-slp.c (vect_record_max_nunits): New function,
15309         split out from...
15310         (vect_build_slp_tree_1): ...here.
15311         (vect_build_slp_tree_2): Call it for phis too.
15313 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15315         * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15316         to vect_get_vec_def_for_operand when getting the mask operand.
15318 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15319             Alan Hayward  <alan.hayward@arm.com>
15320             David Sherwood  <david.sherwood@arm.com>
15322         * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15323         bitstart.
15325 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15326             Alan Hayward  <alan.hayward@arm.com>
15327             David Sherwood  <david.sherwood@arm.com>
15329         * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15330         calculation for vector booleans.
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-stmts.c (can_vectorize_live_stmts): New function,
15337         split out from...
15338         (vect_transform_stmt): ...here.
15339         (vect_analyze_stmt): Use it instead of calling
15340         vectorizable_live_operation directly.
15342 2017-09-18  Cesar Philippidis  <cesar@codesourcery.com>
15344         * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15345         non-SIMT targets in acc vector loops.
15347 2017-09-18  Claudiu Zissulescu  <claziss@synopsys.com>
15349         * configure.ac: Add arc and check if assembler supports gdwarf2.
15350         * configure: Regenerate.
15352 2017-09-18  Richard Biener  <rguenther@suse.de>
15354         PR tree-optimization/82220
15355         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15356         epilogue niters from the min_profitable_iters compute.
15358 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15360         PR target/82145
15361         * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15362         changes.  Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15363         (ix86_init_pic_reg): Revert 2017-09-01 changes.
15365 2017-09-18  Eric Botcazou  <ebotcazou@adacore.com>
15367         PR target/81361
15368         * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15369         switching to a new text section.
15371 2017-09-18  Richard Biener  <rguenther@suse.de>
15373         * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15374         Simplify.
15375         (build_alias_set): Reject aliases with no access function.
15377 2017-09-18  Richard Biener  <rguenther@suse.de>
15379         PR tree-optimization/79622
15380         * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15381         handle PHIs.
15382         (build_cross_bb_scalars_use): Likewise.
15384 2017-09-18  Pierre-Marie de Rodat  <derodat@adacore.com>
15386         * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15388 2017-09-18  Alan Modra  <amodra@gmail.com>
15390         PR target/81996
15391         * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15392         stack_pointer_rtx for count 0.  Update comments.  Break up
15393         large rtl expression.
15395 2017-09-17  Daniel Santos  <daniel.santos@pobox.com>
15397         * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15398         Increase to 20 bytes.
15399         (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15400         (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15401         or avx version of the stub.
15403 2017-09-17  H.J. Lu  <hongjiu.lu@intel.com>
15405         PR target/82166
15406         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15407         compute the minimum stack alignment.  Also update preferred stack
15408         boundary for leaf functions.
15410 2017-09-16  Richard Sandiford  <richard.sandiford@linaro.org>
15412         PR tree-optimization/82228
15413         * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15414         of ncopies.
15416 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15418         * common/config/nds32/nds32-common.c
15419         (nds32_option_optimization_table): Refine formatting.
15420         (nds32_option_optimization_table): Use -fsched-pressure and
15421         -fomit-frame-pointer for specific optimization level.
15423 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15425         * config/nds32/nds32.c: Refine formatting and comments.
15426         * config/nds32/nds32.h: Likewise.
15427         * config/nds32/nds32.md: Likewise.
15428         * config/nds32/nds32-cost.c: Likewise.
15429         * config/nds32/nds32-isr.c: Likewise.
15430         * config/nds32/nds32-md-auxiliary.c: Likewise.
15431         * config/nds32/nds32-multiple.md: Likewise.
15432         * config/nds32/nds32-predicates.c: Likewise.
15434 2017-09-15  Andrew Sutton  <andrew.n.sutton@gmail.com>
15435             Jakub Jelinek  <jakub@redhat.com>
15437         Add support for -std=c++2a.
15438         * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15439         or -std=gnu+2a.
15440         * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15442 2017-09-15  Steve Ellcey  <sellcey@cavium.com>
15444         PR target/82066
15445         * doc/extend.texi (Common Function Attributes): Add 
15446         references to ARM, AArch64, and S/390 specific attributes.
15447         (Function Specific Option Pragmas): Add AArch64 and S/390
15448         to list of back ends that support the target pragma.
15450 2017-09-15  Nathan Sidwell  <nathan@acm.org>
15452         * doc/standards.texi: Fix C++17 description.  Update URLs for
15453         C++11 & 14.
15455 2017-09-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15457         * common.opt (Wcast-align=strict): New warning option.
15458         * doc/invoke.texi: Document -Wcast-align=strict. 
15460 2017-09-15  Pierre-Marie de Rodat  <derodat@adacore.com>
15462         * cgraph.h (cgraph_thunk_info): Add comments.
15463         * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15464         assert for VIRTUAL_* arguments stricter.
15466 2017-09-15  Jackson Woodruff  <jackson.woodruff@arm.com>
15468         PR tree-optimization/71026
15469         * match.pd: Move RDIV patterns from fold-const.c
15470         * fold-const.c (distribute_real_division): Removed.
15471         (fold_binary_loc): Remove calls to distribute_real_divison.
15473 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15475         * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15476         c++1z and gnu++1z as deprecated.  Change other references to
15477         -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15478         Change -Wc++1z-compat to -Wc++17-compat.
15479         * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15480         * dwarf2out.c (highest_c_language): Handle C++17.
15481         (gen_compile_unit_die): Likewise.
15483 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15485         PR rtl-optimization/82192
15486         * combine.c (make_extraction): Don't look through non-paradoxical
15487         SUBREGs or TRUNCATE if pos + len is or might be bigger than
15488         inner's mode.
15490 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15491             Alan Hayward  <alan.hayward@arm.com>
15492             David Sherwood  <david.sherwood@arm.com>
15494         * target.def (function_arg_offset): New hook.
15495         * targhooks.h (default_function_arg_offset): Declare.
15496         * targhooks.c (default_function_arg_offset): New function.
15497         * function.c (locate_and_pad_parm): Use
15498         targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15499         * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15500         (TARGET_FUNCTION_ARG_OFFSET): ...this.
15501         * doc/tm.texi: Regenerate.
15502         * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15503         * config/spu/spu.c (spu_function_arg_offset): New function.
15504         (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15505         * system.h (FUNCTION_ARG_OFFSET): Poison.
15507 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15508             Alan Hayard  <alan.hayward@arm.com>
15509             David Sherwood  <david.sherwood@arm.com>
15511         * target.def (truly_noop_truncation): New hook.
15512         (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15513         than TRULY_NOOP_TRUNCATION.
15514         * hooks.h (hook_bool_uint_uint_true): Declare.
15515         * hooks.c (hook_bool_uint_uint_true): New function.
15516         * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15517         (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15518         * doc/tm.texi: Regenerate.
15519         * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15520         rather than TRULY_NOOP_TRUNCATION in comments.
15521         (simplify_comparison): Likewise.
15522         (record_truncated_value): Likewise.
15523         * expmed.c (extract_bit_field_1): Likewise.
15524         (extract_split_bit_field): Likewise.
15525         * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15526         instead of TRULY_NOOP_TRUNCATION.
15527         * function.c (assign_parm_setup_block): Likewise.
15528         * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15529         * rtlhooks.c: Include target.h.
15530         * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15531         * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15532         * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15533         * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15534         * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15535         * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15536         * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15537         * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15538         * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15539         * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15540         * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15541         * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15542         * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15543         * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15544         * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15545         * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15546         * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15547         * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15548         * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15549         * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15550         * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15551         * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15552         * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15553         * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15554         * config/mips/mips.c (mips_truly_noop_truncation): New function.
15555         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15556         * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15557         rather than TRULY_NOOP_TRUNCATION in comments.
15558         * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15559         * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15560         * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15561         * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15562         * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15563         * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15564         * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15565         * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15566         * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15567         * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15568         * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15569         * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15570         rather than TRULY_NOOP_TRUNCATION in comments.
15571         * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15572         * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15573         * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15574         * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15575         * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15576         TRULY_NOOP_TRUNCATION condition.
15577         (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15578         (TRULY_NOOP_TRUNCATION): Delete.
15579         * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15580         * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15581         * config/spu/spu.c (spu_truly_noop_truncation): New function.
15582         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15583         * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15584         * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15585         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15586         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15587         * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15588         rather than TRULY_NOOP_TRUNCATION in comments.
15589         * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15590         * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15591         * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15592         * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15593         * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15594         * system.h (TRULY_NOOP_TRUNCATION): Poison.
15596 2017-09-15  Christophe Lyon  <christophe.lyon@linaro.org>
15598         PR target/67591
15599         * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15600         (*cmp_ior): Likewise.
15601         (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15602         (*ior_scc_scc_cmp): Likewise.
15603         (*and_scc_scc): Likewise.
15604         (*and_scc_scc_cmp): Likewise.
15606 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15607             Alan Hayard  <alan.hayward@arm.com>
15608             David Sherwood  <david.sherwood@arm.com>
15610         * target.def (can_change_mode_class): New hook.
15611         (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15612         (hard_regno_nregs): Likewise.
15613         * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15614         * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15615         * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15616         (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15617         (LOAD_EXTEND_OP): Update accordingly.
15618         * doc/tm.texi: Regenerate.
15619         * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15620         CANNOT_CHANGE_MODE_CLASS.
15621         * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15622         (REG_CAN_CHANGE_MODE_P): ...this new macro.
15623         * combine.c (simplify_set): Update accordingly.
15624         * emit-rtl.c (validate_subreg): Likewise.
15625         * recog.c (general_operand): Likewise.
15626         * regcprop.c (mode_change_ok): Likewise.
15627         * reload1.c (choose_reload_regs): Likewise.
15628         (inherit_piecemeal_p): Likewise.
15629         * rtlanal.c (simplify_subreg_regno): Likewise.
15630         * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15631         instead of CANNOT_CHANGE_MODE_CLASS.
15632         (reload_cse_simplify_operands): Likewise.
15633         * reload.c (push_reload): Use targetm.can_change_mode_class
15634         instead of CANNOT_CHANGE_MODE_CLASS.
15635         (push_reload): Likewise.  Also use REG_CAN_CHANGE_MODE_P instead of
15636         REG_CANNOT_CHANGE_MODE_P.
15637         * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15638         * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15639         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15640         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15641         * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15642         (arm_can_change_mode_class): New function.
15643         * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15644         than CANNOT_CHANGE_MODE_CLASS in comments.
15645         * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15646         * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15647         * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15648         (ix86_can_change_mode_class): ...this new function, inverting the
15649         sense of the return value.
15650         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15651         * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15652         * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15653         (ia64_can_change_mode_class): New function.
15654         * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15655         * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15656         * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15657         (m32c_can_change_mode_class): ...this new function, inverting the
15658         sense of the return value.
15659         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15660         * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15661         * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15662         * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15663         (mips_can_change_mode_class): ...this new function, inverting the
15664         sense of the return value.
15665         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15666         * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15667         * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15668         (msp430_can_change_mode_class): New function.
15669         * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15670         * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15671         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15672         * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15673         * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15674         * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15675         * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15676         (pa_cannot_change_mode_class): Replace with...
15677         (pa_can_change_mode_class): ...this new function, inverting the
15678         sense of the return value.
15679         (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15680         than CANNOT_CHANGE_MODE_CLASS in comments.
15681         * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15682         * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15683         * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15684         (pdp11_cannot_change_mode_class): Replace with...
15685         (pdp11_can_change_mode_class): ...this new function, inverting the
15686         sense of the return value.
15687         * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15688         * config/powerpcspe/powerpcspe-protos.h
15689         (rs6000_cannot_change_mode_class_ptr): Delete.
15690         * config/powerpcspe/powerpcspe.c
15691         (rs6000_cannot_change_mode_class_ptr): Delete.
15692         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15693         (rs6000_option_override_internal): Assign to
15694         targetm.can_change_mode_class instead of
15695         rs6000_cannot_change_mode_class_ptr.
15696         (rs6000_cannot_change_mode_class): Replace with...
15697         (rs6000_can_change_mode_class): ...this new function, inverting the
15698         sense of the return value.
15699         (rs6000_debug_cannot_change_mode_class): Replace with...
15700         (rs6000_debug_can_change_mode_class): ...this new function.
15701         * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15702         * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15703         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15704         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15705         * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15706         Delete.
15707         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15708         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15709         (rs6000_option_override_internal): Assign to
15710         targetm.can_change_mode_class instead of
15711         rs6000_cannot_change_mode_class_ptr.
15712         (rs6000_cannot_change_mode_class): Replace with...
15713         (rs6000_can_change_mode_class): ...this new function, inverting the
15714         sense of the return value.
15715         (rs6000_debug_cannot_change_mode_class): Replace with...
15716         (rs6000_debug_can_change_mode_class): ...this new function.
15717         * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15718         * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15719         * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15720         (s390_can_change_mode_class): ...this new function, inverting the
15721         sense of the return value.
15722         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15723         * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15724         * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15725         * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15726         (sh_cannot_change_mode_class): Replace with...
15727         (sh_can_change_mode_class): ...this new function, inverting the
15728         sense of the return value.
15729         * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15730         * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15731         (sparc_can_change_mode_class): New function.
15732         * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15733         * config/spu/spu.c (spu_can_change_mode_class): New function.
15734         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15735         * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15736         * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15737         (visium_can_change_mode_class): New function.
15738         * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15740 2017-09-15  Richard Biener  <rguenther@suse.de>
15742         PR tree-optimization/82217
15743         * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15744         but not undefined case.
15746 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15748         PR target/82145
15749         * postreload.c (reload_cse_simplify_operands): Skip
15750         NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15752 2017-09-15  Richard Biener  <rguenther@suse.de>
15754         PR tree-optimization/68823
15755         * graphite-scop-detection.c (build_alias_set): If we have a
15756         possible dependence check whether we can handle them by just
15757         looking at the DRs DR_ACCESS_FNs.
15758         (build_scops): If build_alias_set fails, fail the SCOP.
15760 2017-09-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
15762         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15763         to support float128 built-in functions that require the ISA 3.0
15764         hardware.
15765         (BU_FLOAT128_3_HW): Likewise.
15766         (SQRTF128): Add support for the IEEE 128-bit square root and fma
15767         built-in functions.
15768         (FMAF128): Likewise.
15769         (FMAQ): Likewise.
15770         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15771         support for built-in functions that need the ISA 3.0 IEEE 128-bit
15772         floating point instructions.
15773         (rs6000_invalid_builtin): Likewise.
15774         (rs6000_builtin_mask_names): Likewise.
15775         * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15776         (RS6000_BTM_FLOAT128_HW): Likewise.
15777         (RS6000_BTM_COMMON): Likewise.
15778         * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15779         function.
15780         * doc/extend.texi (RS/6000 built-in functions): Document the
15781         IEEE 128-bit floating point square root and fused multiply-add
15782         built-in functions.
15784 2017-09-14  Pat Haugen  <pthaugen@us.ibm.com>
15786         * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15787         reg (r2) isn't in the set of registers defined in the prologue.
15789 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15790             Alan Hayward  <alan.hayward@arm.com>
15791             David Sherwood  <david.sherwood@arm.com>
15793         * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15794         (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15795         (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15796         (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15797         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15798         accordingly.
15799         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15800         max_vectorization_factor.
15801         (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
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 (vect_min_worthwhile_factor): Delete.
15808         (vect_worthwhile_without_simd_p): Declare.
15809         * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15810         (vectorizable_reduction): Use it.
15811         * tree-vect-stmts.c (vectorizable_shift): Likewise.
15812         (vectorizable_operation): Likewise.
15814 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15815             Alan Hayward  <alan.hayward@arm.com>
15816             David Sherwood  <david.sherwood@arm.com>
15818         * tree-vectorizer.h (vect_get_num_copies): New function.
15819         * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15820         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15821         (vectorizable_induction): Likewise.
15822         (vectorizable_live_operation): Likewise.
15823         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15824         (vectorizable_bswap): Likewise.
15825         (vectorizable_call): Likewise.
15826         (vectorizable_conversion): Likewise.
15827         (vectorizable_assignment): Likewise.
15828         (vectorizable_shift): Likewise.
15829         (vectorizable_operation): Likewise.
15830         (vectorizable_store): Likewise.
15831         (vectorizable_load): Likewise.
15832         (vectorizable_condition): Likewise.
15833         (vectorizable_comparison): Likewise.
15834         (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15836 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15837             Alan Hayward  <alan.hayward@arm.com>
15838             David Sherwood  <david.sherwood@arm.com>
15840         * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15841         of vect_init_vector.
15843 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15844             Alan Hayward  <alan.hayward@arm.com>
15845             David Sherwood  <david.sherwood@arm.com>
15847         * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15848         an inline wrapper that provides a location.
15849         (gimple_build_vector): Likewise.
15850         * gimple-fold.c (gimple_build_vector_from_val): New function.
15851         (gimple_build_vector): Likewise.
15852         * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15853         functions to build the initial value.  Always return a gimple value.
15854         (get_initial_defs_for_reduction): Likewise.  Only compute
15855         neutral_vec once.
15856         (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15857         vect_init_vector on the results from get_initial_def(s)_for_reduction.
15858         (vectorizable_induction): Use gimple_build_vector rather than
15859         vect_init_vector.
15861 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15862             Alan Hayward  <alan.hayward@arm.com>
15863             David Sherwood  <david.sherwood@arm.com>
15865         * target.h (vec_perm_indices): New typedef.
15866         (auto_vec_perm_indices): Likewise.
15867         * optabs-query.h: Include target.h
15868         (can_vec_perm_p): Take a vec_perm_indices *.
15869         * optabs-query.c (can_vec_perm_p): Likewise.
15870         (can_mult_highpart_p): Update accordingly.  Use auto_vec_perm_indices.
15871         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15872         * tree-vect-generic.c (lower_vec_perm): Likewise.
15873         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15874         (vect_grouped_load_supported): Likewise.
15875         (vect_shift_permute_load_chain): Likewise.
15876         (vect_permute_store_chain): Use auto_vec_perm_indices.
15877         (vect_permute_load_chain): Likewise.
15878         * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15879         (fold_ternary_loc): Update accordingly.  Use auto_vec_perm_indices.
15880         Update uses of can_vec_perm_p.
15881         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15882         mode with a number of elements.  Take a vec_perm_indices *.
15883         (vect_create_epilog_for_reduction): Update accordingly.
15884         Use auto_vec_perm_indices.
15885         (have_whole_vector_shift): Likewise.  Update call to can_vec_perm_p.
15886         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15887         (vect_transform_slp_perm_load): Likewise.
15888         (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15889         * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15890         (vect_gen_perm_mask_checked): Likewise.
15891         * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15892         (vect_gen_perm_mask_checked): Likewise.
15893         (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15894         (vectorizable_store): Likewise.
15895         (vectorizable_load): Likewise.
15896         (perm_mask_for_reverse): Likewise.  Update call to can_vec_perm_p.
15897         (vectorizable_bswap): Likewise.
15899 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15900             Alan Hayward  <alan.hayward@arm.com>
15901             David Sherwood  <david.sherwood@arm.com>
15903         * tree.h (build_vector): Take a vec<tree> instead of a tree *.
15904         * tree.c (build_vector): Likewise.
15905         (build_vector_from_ctor): Update accordingly.
15906         (build_vector_from_val): Likewise.
15907         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
15908         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15909         * tree-vect-generic.c (add_rshift): Likewise.
15910         (expand_vector_divmod): Likewise.
15911         (optimize_vector_constructor): Likewise.
15912         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
15913         (vect_transform_slp_perm_load): Likewise.
15914         (vect_schedule_slp_instance): Likewise.
15915         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15916         (vectorizable_call): Likewise.
15917         (vect_gen_perm_mask_any): Likewise.  Add elements in order.
15918         * expmed.c (make_tree): Likewise.
15919         * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
15920         a vector passed to build_vector.
15921         (fold_convert_const): Likewise.
15922         (exact_inverse): Likewise.
15923         (fold_ternary_loc): Likewise.
15924         (fold_relational_const): Likewise.
15925         (const_binop): Likewise.  Use VECTOR_CST_ELT directly when operating
15926         on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
15927         (const_unop): Likewise.  Store the reduction accumulator in a
15928         variable rather than an array.
15929         (vec_cst_ctor_to_array): Take the number of elements as a parameter.
15930         (fold_vec_perm): Update calls accordingly.  Use auto_vec<tree> for
15931         the new vector, rather than constructing it after the input arrays.
15932         (native_interpret_vector): Use auto_vec<tree> when building
15933         a vector passed to build_vector.  Add elements in order.
15934         * tree-vect-loop.c (get_initial_defs_for_reduction): Use
15935         auto_vec<tree> when building a vector passed to build_vector.
15936         (vect_create_epilog_for_reduction): Likewise.
15937         (vectorizable_induction): Likewise.
15938         (get_initial_def_for_reduction): Likewise.  Fix indentation of
15939         case statements.
15940         * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
15941         to a vec<tree> *.
15942         (sparc_fold_builtin): Use auto_vec<tree> when building a vector
15943         passed to build_vector.
15945 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15946             Alan Hayward  <alan.hayward@arm.com>
15947             David Sherwood  <david.sherwood@arm.com>
15949         * tree-core.h (tree_base::u): Add an "nelts" field.
15950         (tree_vector): Use VECTOR_CST_NELTS as the length.
15951         * tree.c (tree_size): Likewise.
15952         (make_vector): Initialize VECTOR_CST_NELTS.
15953         * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
15954         * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
15955         TYPE_VECTOR_SUBPARTS.
15956         * expr.c (const_vector_mask_from_tree): Consistently use "units"
15957         as the number of units, setting it from VECTOR_CST_NELTS.
15958         (const_vector_from_tree): Likewise.
15959         * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
15960         TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
15961         (fold_negate_expr_1): Likewise.
15962         (fold_convert_const): Likewise.
15963         (const_binop): Likewise.  Differentiate the number of output and
15964         input elements.
15965         (const_unop): Likewise.
15966         (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
15967         in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
15968         in cases that did the opposite.
15970 2017-09-14  Richard Biener  <rguenther@suse.de>
15972         * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
15973         to VN_TOP.
15975 2017-09-14  Eric Botcazou  <ebotcazou@adacore.com>
15977         * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
15979 2017-09-14  Jakub Jelinek  <jakub@redhat.com>
15981         PR target/81325
15982         * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
15983         if and where to split a bb, except for splitting before debug insn
15984         sequences followed by non-label real insn.  Delete debug insns
15985         in between basic blocks.
15987         * combine.c (make_compound_operation_int): Formatting fixes.
15989         * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
15990         * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15991         * config/netbsd.h (LINK_EH_SPEC): Likewise.
15992         * config/sol2.h (LINK_EH_SPEC): Likewise.
15993         * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15994         * config/s390/linux.h (LINK_SPEC): Likewise.
15995         * config/freebsd.h (LINK_EH_SPEC): Likewise.
15996         * config/openbsd.h (LINK_EH_SPEC): Likewise.
15997         * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15998         * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
15999         * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
16000         * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16001         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
16002         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
16004 2017-09-13  Jakub Jelinek  <jakub@redhat.com>
16006         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
16007         support.
16008         (ENDFILE_LINUX_SPEC): Likewise.
16009         (LINK_EH_SPEC): Likewise.
16010         * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
16011         (LINK_OS_LINUX_SPEC32): Likewise.
16012         (LINK_OS_LINUX_SPEC64): Likewise.
16013         * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
16014         (LINK_OS_LINUX_SPEC): Likewise.
16016 2017-09-13  Martin Liska  <mliska@suse.cz>
16018         PR middle-end/82154
16019         * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
16020         CASE_HIGH is NULL_TREE.
16022 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16023             Alan Hayward  <alan.hayward@arm.com>
16024             David Sherwood  <david.sherwood@arm.com>
16026         * target.def (secondary_memory_needed): New hook.
16027         (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
16028         instead of SECONDARY_MEMORY_NEEDED.
16029         (secondary_memory_needed_mode): Likewise.
16030         * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
16031         * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
16032         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
16033         (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
16034         (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
16035         * doc/tm.texi: Regenerate.
16036         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
16037         * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
16038         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16039         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
16040         * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
16041         * config/i386/i386.c (inline_secondary_memory_needed): Put the
16042         mode argument first and change the reg_class arguments to reg_class_t.
16043         (ix86_secondary_memory_needed): Likewise.  Remove the strict parameter.
16044         Make static.  Update the call to inline_secondary_memory_needed.
16045         (ix86_register_move_cost): Update the call to
16046         inline_secondary_memory_needed.
16047         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16048         * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
16049         definition.
16050         * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
16051         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16052         in comment.
16053         * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
16054         * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
16055         * config/mips/mips.c (mips_secondary_memory_needed): Make static
16056         and match hook interface.  Add comment from mips.h.
16057         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16058         * config/mmix/mmix.md (truncdfsf2): Refer to
16059         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16060         in comment.
16061         * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
16062         (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
16063         * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16064         (pa_secondary_memory_needed): New function.
16065         * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
16066         * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
16067         * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16068         (pdp11_secondary_memory_needed): Make static and match hook interface.
16069         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
16070         * config/powerpcspe/powerpcspe-protos.h
16071         (rs6000_secondary_memory_needed_ptr): Delete.
16072         * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
16073         Delete.
16074         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16075         (rs6000_option_override_internal): Assign to
16076         targetm.secondary_memory_needed rather than
16077         rs6000_secondary_memory_needed_ptr.
16078         (rs6000_secondary_memory_needed): Match hook interface.
16079         (rs6000_debug_secondary_memory_needed): Likewise.
16080         * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
16081         * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
16082         (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
16083         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16084         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
16085         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
16086         Delete.
16087         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): 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/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
16095         * config/s390/s390.c (s390_secondary_memory_needed): New function.
16096         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16097         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
16098         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16099         (sparc_secondary_memory_needed): New function.
16100         * lra-constraints.c (check_and_process_move): Refer to
16101         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16102         in comment.
16103         (curr_insn_transform): Likewise.
16104         (process_alt_operands): Use targetm.secondary_memory_needed
16105         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16106         (check_secondary_memory_needed_p): Likewise.
16107         (choose_split_class): Likewise.
16108         * reload.c: Unconditionally include code that was previously
16109         conditional on SECONDARY_MEMORY_NEEDED.
16110         (push_secondary_reload): Use targetm.secondary_memory_needed
16111         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16112         (push_reload): Likewise.
16113         * reload1.c: Unconditionally include code that was previously
16114         conditional on SECONDARY_MEMORY_NEEDED.
16115         (choose_reload_regs): Use targetm.secondary_memory_needed
16116         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16117         (gen_reload): Likewise.
16118         * system.h (SECONDARY_MEMORY_NEEDED): Poison.
16120 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16121             Alan Hayward  <alan.hayward@arm.com>
16122             David Sherwood  <david.sherwood@arm.com>
16124         * target.def (secondary_memory_needed_mode): New hook:
16125         * targhooks.c (default_secondary_memory_needed_mode): Declare.
16126         * targhooks.h (default_secondary_memory_needed_mode): New function.
16127         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
16128         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
16129         * doc/tm.texi: Regenerate.
16130         * lra-constraints.c (check_and_process_move): Use
16131         targetm.secondary_memory_needed_mode instead of
16132         TARGET_SECONDARY_MEMORY_NEEDED_MODE.
16133         (curr_insn_transform): Likewise.
16134         * reload.c (get_secondary_mem): Likewise.
16135         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16136         * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
16137         function.
16138         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16139         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16140         * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
16141         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16142         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
16143         Delete.
16144         * config/powerpcspe/powerpcspe-protos.h
16145         (rs6000_secondary_memory_needed_mode): Delete.
16146         * config/powerpcspe/powerpcspe.c
16147         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16148         (rs6000_secondary_memory_needed_mode): Make static.
16149         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16150         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
16151         Delete.
16152         * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16153         Redefine.
16154         (rs6000_secondary_memory_needed_mode): Make static.
16155         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16156         * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
16157         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16158         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16159         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16160         Redefine.
16161         (sparc_secondary_memory_needed_mode): New function.
16162         * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
16164 2017-09-13  Jackson Woodruff  <jackson.woodruff@arm.com>
16166         * config/aarch64/constraints.md (Umq): New constraint.
16167         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
16168         Change to use Umq.
16169         (mov<mode>): Update condition.
16171 2017-09-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16173         * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
16174         when bitposition is the same.
16176 2017-09-13  Richard Biener  <rguenther@suse.de>
16178         * dwarf2out.c (output_die_symbol): Remove.
16179         (output_die): Do not output a DIEs symbol.
16181 2017-09-13  Richard Biener  <rguenther@suse.de>
16183         PR middle-end/82128
16184         * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
16185         default-def to avoid breaking iterator update with the weird
16186         interaction with cgraph_update_edges_for_call_stmt_node.
16188 2017-09-13  Richard Biener  <rguenther@suse.de>
16190         * tree-cfg.c (verify_gimple_assign_binary): Add verification
16191         for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
16192         VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
16193         (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
16195 2017-09-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
16197         * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
16198         Disable pc relative literal load irrespective of
16199         TARGET_FIX_ERR_A53_84341 for default.
16201 2017-09-12  Eric Botcazou  <ebotcazou@adacore.com>
16203         * config/sparc/sparc.c (output_return): Output the source location of
16204         the insn in the delay slot, if any.
16205         (output_sibcall): Likewise.
16207 2017-09-12  H.J. Lu  <hongjiu.lu@intel.com>
16209         PR driver/81498
16210         * common.opt (-static-pie): New alias.
16211         (shared): Negate static-pie.
16212         (-no-pie): Update help text.
16213         (-pie): Likewise.
16214         (static-pie): New option.
16215         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
16216         -static-pie support.
16217         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
16218         (LINK_EH_SPEC): Likewise.
16219         (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16220         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16221         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16222         * gcc.c (LINK_COMMAND_SPEC): Likewise.
16223         (init_gcc_specs): Likewise.
16224         (init_spec): Likewise.
16225         (display_help): Update help message for -pie.
16226         * doc/invoke.texi: Update -pie, -no-pie and -static.  Document
16227         -static-pie.
16229 2017-09-12  Wilco Dijkstra  <wdijkstr@arm.com>
16231         * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16232         (movdi_aarch64): Likewise.
16233         (movti_aarch64): Likewise.
16235 2017-09-12 Simon Wright <simon@pushface.org>
16237         PR target/80204
16238         * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16239         calculation of the minor version, always output as 0.
16241 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16243         PR target/82112
16244         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16245         ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16246         on it early, rather than manual conversion late.  For
16247         ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16248         instead of performing manual conversion.
16250 2017-09-12  Carl Love  <cel@us.ibm.com>
16252         * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16253         vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16254         vmulouw, vmulosw.
16255         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16256         VMULOSW): Add definitions.
16257         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16258         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16259         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16260         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16261         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16263 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16265         * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16266         types correctly.
16267         (movti_aarch64): Likewise.
16268         (movdf_aarch64): Likewise.
16269         (movtf_aarch64): Likewise.
16270         (load_pairdi): Likewise.
16271         (store_pairdi): Likewise.
16272         (load_pairdf): Likewise.
16273         (store_pairdf): Likewise.
16274         (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16275         (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16276         (ldr_got_small_<mode>): Likewise.
16277         (ldr_got_small_28k_<mode>): Likewise.
16278         (ldr_got_tiny): Likewise.
16279         * config/aarch64/iterators.md (ldst_sz): New.
16280         (ldpstp_sz): Likewise.
16281         * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16282         to store_16.
16283         (thunderx_load): Split load_8 to load_16.
16284         * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16285         load_8 to load_16.
16286         (thunderx2t99_storepair_basic): Split store_8 to store_16.
16287         * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16288         (xgene1_store_pair): Split store_8 to store_16.
16289         * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16290         (falkor_st_0_st_sd): Split store_8 to store_16.
16292 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16294         * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16295         and store1/2/3/4 to store_4/8/12/16.
16296         * config/aarch64/aarch64.md: Update for rename.
16297         * config/arm/arm.md: Likewise.: Likewise.
16298         * config/arm/arm.c: Likewise.
16299         * config/arm/thumb1.md: Likewise.
16300         * config/arm/thumb2.md: Likewise.
16301         * config/arm/vfp.md: Likewise.
16302         * config/arm/arm-generic.md: Likewise.
16303         * config/arm/arm1020e.md: Likewise.
16304         * config/arm/arm1026ejs.md: Likewise.
16305         * config/arm/arm1136jfs.md: Likewise.
16306         * config/arm/arm926ejs.md: Likewise.
16307         * config/arm/cortex-a15.md: Likewise.
16308         * config/arm/cortex-a17.md: Likewise.
16309         * config/arm/cortex-a5.md: Likewise.
16310         * config/arm/cortex-a53.md: Likewise.
16311         * config/arm/cortex-a57.md: Likewise.
16312         * config/arm/cortex-a7.md: Likewise.
16313         * config/arm/cortex-a8.md: Likewise.
16314         * config/arm/cortex-a9.md: Likewise.
16315         * config/arm/cortex-m4.md: Likewise.
16316         * config/arm/cortex-m7.md: Likewise.
16317         * config/arm/cortex-r4.md: Likewise.
16318         * config/arm/exynos-m1.md: Likewise.
16319         * config/arm/fa526.md: Likewise.
16320         * config/arm/fa606te.md: Likewise.
16321         * config/arm/fa626te.md: Likewise.
16322         * config/arm/fa726te.md: Likewise.
16323         * config/arm/fmp626.md: Likewise.
16324         * config/arm/iwmmxt.md: Likewise.
16325         * config/arm/ldmstm.md: Likewise.
16326         * config/arm/marvell-pj4.md: Likewise.
16327         * config/arm/xgene1.md: Likewise.
16328         * config/aarch64/thunderx.md: Likewise.
16329         * config/aarch64/thunderx2t99.md: Likewise.
16330         * config/aarch64/falkor.md: Likewise.
16332 2017-09-12  Martin Liska  <mliska@suse.cz>
16334         * attribs.c (private_lookup_attribute): New function.
16335         * attribs.h (private_lookup_attribute): Declared here.
16336         (lookup_attribute): Called from this place.
16338 2017-09-12  Richard Biener  <rguenther@suse.de>
16340         PR tree-optimization/82157
16341         * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16342         stmts with side-effects.
16344 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16345             Alan Hayward  <alan.hayward@arm.com>
16346             David Sherwood <david.sherwood@arm.com>
16348         * target.def (hard_regno_nregs): New hook.
16349         (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16350         * targhooks.h (default_hard_regno_nregs): Declare.
16351         * targhooks.c (default_hard_regno_nregs): New function.
16352         * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16353         (TARGET_HARD_REGNO_NREGS): ...this hook.
16354         (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16355         (CLASS_MAX_NREGS): Likewise.
16356         * doc/tm.texi: Regenerate.
16357         * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16358         instead of HARD_REGNO_NREGS.
16359         * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16360         HARD_REGNO_NREGS in the comment.
16361         * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16362         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16363         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16364         Return an unsigned int.
16365         (TARGET_HARD_REGNO_NREGS): Redefine.
16366         * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16367         * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16368         * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16369         (arc_hard_regno_nregs): New function.
16370         * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16371         * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16372         (arm_hard_regno_nregs): New function.
16373         * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16374         * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16375         * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16376         (TARGET_HARD_REGNO_NREGS): Redefine.
16377         * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16378         * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16379         (HARD_REGNO_NREGS): Delete.
16380         * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16381         (cr16_hard_regno_nregs): New function.
16382         (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16383         * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16384         * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16385         (cris_hard_regno_nregs): New function.
16386         * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16387         * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16388         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16389         * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16390         (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16391         * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16392         * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16393         (frv_hard_regno_nregs): Make static.  Take and return an
16394         unsigned int.
16395         (frv_class_max_nregs): Remove outdated copy of documentation.
16396         * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16397         * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16398         * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16399         * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16400         * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16401         * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16402         (TARGET_HARD_REGNO_NREGS): Redefine.
16403         * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16404         (CLASS_MAX_NREGS): Update comment.
16405         * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16406         (ia64_hard_regno_nregs): New function.
16407         * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16408         * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16409         * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16410         * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16411         * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16412         an unsigned int.
16413         (m32c_hard_regno_nregs): Likewise.  Make static.
16414         (TARGET_HARD_REGNO_NREGS): Redefine.
16415         * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16416         * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16417         * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16418         (m68k_hard_regno_nregs): New function.
16419         * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16420         * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16421         * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16422         * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16423         * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16424         Take and return an unsigned int.
16425         (TARGET_HARD_REGNO_NREGS): Redefine.
16426         * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16427         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16428         * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16429         * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16430         * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16431         * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16432         * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16433         (msp430_hard_regno_nregs): Make static.  Take and return an
16434         unsigned int.
16435         * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16436         * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16437         * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16438         (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16439         * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16440         * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16441         * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16442         (TARGET_HARD_REGNO_NREGS): Redefine.
16443         * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16444         (PA_HARD_REGNO_NREGS): ...this.
16445         * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16446         (PA_HARD_REGNO_NREGS): ...this.
16447         * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16448         (pa_hard_regno_nregs): New function.
16449         * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16450         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16451         (pdp11_hard_regno_nregs): New function.
16452         * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16453         * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16454         (rs6000_hard_regno_nregs_hook): New function.
16455         * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16456         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16457         * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16458         Take and return an unsigned int.  Move earlier in file.
16459         (TARGET_HARD_REGNO_NREGS): Redefine.
16460         * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16461         * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16462         * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16463         (rl78_hard_regno_nregs): Make static.  Take and return an
16464         unsigned int.
16465         * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16466         * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16467         (rs6000_hard_regno_nregs_hook): New function.
16468         * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16469         * config/rx/rx.c (rx_hard_regno_nregs): New function.
16470         (TARGET_HARD_REGNO_NREGS): Redefine.
16471         * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16472         * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16473         instead of HARD_REGNO_NREGS.
16474         (s390_hard_regno_nregs): New function.
16475         (s390_hard_regno_mode_ok): Add comment from s390.h.
16476         (TARGET_HARD_REGNO_NREGS): Redefine.
16477         * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16478         * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16479         (sh_hard_regno_nregs): New function.
16480         (sh_pass_in_reg_p): Use it.
16481         * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16482         * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16483         (sparc_hard_regno_nregs): New function.
16484         * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16485         * config/spu/spu.c (spu_hard_regno_nregs): New function.
16486         (spu_function_arg_advance): Use it, supplying a valid register number.
16487         (TARGET_HARD_REGNO_NREGS): Redefine.
16488         * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16489         * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16490         * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16491         * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16492         * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16493         * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16494         (CLASS_MAX_NREGS): Remove copy of old documentation.
16495         * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16496         (visium_hard_regno_nregs): New function.
16497         (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16498         * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16499         * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16500         xtensa_hard_regno_nregs): New function.
16501         * system.h (HARD_REGNO_NREGS): Poison.
16503 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16505         * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16506         hard_regno_nregs instead of HARD_REGNO_NREGS.
16507         (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16508         * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16509         (c6x_expand_epilogue): Likewise.
16510         * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16511         (frv_read_iacc_argument): Likewise.
16512         * config/sh/sh.c: Include regs.h.
16513         (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16514         (regs_used): Likewise.
16515         (output_stack_adjust): Likewise.
16516         * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16517         * expmed.c: Include regs.h.
16518         (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16519         * ree.c: Include regs.h.
16520         (combine_reaching_defs): Use hard_regno_nregs instead of
16521         HARD_REGNO_NREGS.
16522         (add_removable_extension): Likewise.
16524 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16526         * regs.h (hard_regno_nregs): Turn into a function.
16527         (end_hard_regno): Update accordingly.
16528         * caller-save.c (setup_save_areas): Likewise.
16529         (save_call_clobbered_regs): Likewise.
16530         (replace_reg_with_saved_mem): Likewise.
16531         (insert_restore): Likewise.
16532         (insert_save): Likewise.
16533         * combine.c (can_change_dest_mode): Likewise.
16534         (move_deaths): Likewise.
16535         (distribute_notes): Likewise.
16536         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16537         * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16538         (rs6000_split_multireg_move): Likewise.
16539         (rs6000_register_move_cost): Likewise.
16540         (rs6000_memory_move_cost): Likewise.
16541         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16542         (rs6000_split_multireg_move): Likewise.
16543         (rs6000_register_move_cost): Likewise.
16544         (rs6000_memory_move_cost): Likewise.
16545         * cselib.c (cselib_reset_table): Likewise.
16546         (cselib_lookup_1): Likewise.
16547         * emit-rtl.c (set_mode_and_regno): Likewise.
16548         * function.c (aggregate_value_p): Likewise.
16549         * ira-color.c (setup_profitable_hard_regs): Likewise.
16550         (check_hard_reg_p): Likewise.
16551         (calculate_saved_nregs): Likewise.
16552         (assign_hard_reg): Likewise.
16553         (improve_allocation): Likewise.
16554         (calculate_spill_cost): Likewise.
16555         * ira-emit.c (modify_move_list): Likewise.
16556         * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16557         (ira_hard_reg_in_set_p): Likewise.
16558         * ira.c (setup_reg_mode_hard_regset): Likewise.
16559         (clarify_prohibited_class_mode_regs): Likewise.
16560         (check_allocation): Likewise.
16561         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16562         (lra_setup_reg_renumber): Likewise.
16563         (setup_try_hard_regno_pseudos): Likewise.
16564         (spill_for): Likewise.
16565         (assign_hard_regno): Likewise.
16566         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16567         * lra-constraints.c (in_class_p): Likewise.
16568         (lra_constraint_offset): Likewise.
16569         (simplify_operand_subreg): Likewise.
16570         (lra_constraints): Likewise.
16571         (split_reg): Likewise.
16572         (split_if_necessary): Likewise.
16573         (invariant_p): Likewise.
16574         (inherit_in_ebb): Likewise.
16575         * lra-lives.c (process_bb_lives): Likewise.
16576         * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16577         (get_hard_regs): Likewise.
16578         (do_remat): Likewise.
16579         * lra-spills.c (assign_spill_hard_regs): Likewise.
16580         * mode-switching.c (create_pre_exit): Likewise.
16581         * postreload.c (reload_combine_recognize_pattern): Likewise.
16582         * recog.c (peep2_find_free_register): Likewise.
16583         * regcprop.c (kill_value_regno): Likewise.
16584         (set_value_regno): Likewise.
16585         (copy_value): Likewise.
16586         (maybe_mode_change): Likewise.
16587         (find_oldest_value_reg): Likewise.
16588         (copyprop_hardreg_forward_1): Likewise.
16589         * regrename.c (check_new_reg_p): Likewise.
16590         (regrename_do_replace): Likewise.
16591         * reload.c (push_reload): Likewise.
16592         (combine_reloads): Likewise.
16593         (find_dummy_reload): Likewise.
16594         (operands_match_p): Likewise.
16595         (find_reloads): Likewise.
16596         (find_equiv_reg): Likewise.
16597         (reload_adjust_reg_for_mode): Likewise.
16598         * reload1.c (count_pseudo): Likewise.
16599         (count_spilled_pseudo): Likewise.
16600         (find_reg): Likewise.
16601         (clear_reload_reg_in_use): Likewise.
16602         (free_for_value_p): Likewise.
16603         (allocate_reload_reg): Likewise.
16604         (choose_reload_regs): Likewise.
16605         (reload_adjust_reg_for_temp): Likewise.
16606         (emit_reload_insns): Likewise.
16607         (delete_output_reload): Likewise.
16608         * rtlanal.c (subreg_get_info): Likewise.
16609         * sched-deps.c (sched_analyze_reg): Likewise.
16610         * sel-sched.c (init_regs_for_mode): Likewise.
16611         (mark_unavailable_hard_regs): Likewise.
16612         (choose_best_reg_1): Likewise.
16613         (verify_target_availability): Likewise.
16614         * valtrack.c (dead_debug_insert_temp): Likewise.
16615         * var-tracking.c (track_loc_p): Likewise.
16616         (emit_note_insn_var_location): Likewise.
16617         * varasm.c (make_decl_rtl): Likewise.
16618         * reginfo.c (choose_hard_reg_mode): Likewise.
16619         (init_reg_modes_target): Refer directly to
16620         this_target_regs->x_hard_regno_nregs.
16622 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16624         * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16625         instead of hard_regno_nregs.
16627 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16629         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16630         end_hard_regno instead of hard_regno_nregs.
16631         * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16632         * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16633         * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16634         * ira-color.c (improve_allocation): Likewise.
16635         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16636         * lra-lives.c (mark_regno_live): Likewise.
16637         (mark_regno_dead): Likewise.
16638         * lra-remat.c (operand_to_remat): Likewise.
16639         * lra.c (collect_non_operand_hard_regs): Likewise.
16640         * postreload.c (reload_combine_note_store): Likewise.
16641         (move2add_valid_value_p): Likewise.
16642         * reload.c (regno_clobbered_p): Likewise.
16644 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16646         * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16647         hard_regno_nregs.
16648         * config/v850/v850.c (v850_reorg): Likewise.
16649         * reload.c (refers_to_regno_for_reload_p): Likewise.
16650         (find_equiv_reg): Likewise.
16651         * reload1.c (reload_reg_reaches_end_p): Likewise.
16653 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16655         * caller-save.c (add_used_regs): Use REG_NREGS instead of
16656         hard_regno_nregs.
16657         * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16658         * config/arm/arm.c (output_move_neon): Likewise.
16659         (arm_attr_length_move_neon): Likewise.
16660         (neon_split_vcombine): Likewise.
16661         * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16662         (c6x_mark_reg_written): Likewise.
16663         (c6x_dwarf_register_span): Likewise.
16664         * config/i386/i386.c (ix86_save_reg): Likewise.
16665         * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16666         (rws_access_reg): Likewise.
16667         * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16668         * mode-switching.c (create_pre_exit): Likewise.
16669         * ree.c (combine_reaching_defs): Likewise.
16670         (add_removable_extension): Likewise.
16671         * regcprop.c (find_oldest_value_reg): Likewise.
16672         (copyprop_hardreg_forward_1): Likewise.
16673         * reload.c (reload_inner_reg_of_subreg): Likewise.
16674         (push_reload): Likewise.
16675         (combine_reloads): Likewise.
16676         (find_dummy_reload): Likewise.
16677         (reload_adjust_reg_for_mode): Likewise.
16678         * reload1.c (find_reload_regs): Likewise.
16679         (forget_old_reloads_1): Likewise.
16680         (reload_reg_free_for_value_p): Likewise.
16681         (reload_adjust_reg_for_temp): Likewise.
16682         (emit_reload_insns): Likewise.
16683         (delete_output_reload): Likewise.
16684         * sel-sched.c (choose_best_reg_1): Likewise.
16685         (choose_best_pseudo_reg): Likewise.
16687 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16688             Alan Hayward  <alan.hayward@arm.com>
16689             David Sherwood <david.sherwood@arm.com>
16691         * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16692         * target.def (slow_unaligned_access): New hook.
16693         * targhooks.h (default_slow_unaligned_access): Declare.
16694         * targhooks.c (default_slow_unaligned_access): New function.
16695         * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16696         (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16697         * doc/tm.texi: Regenerate.
16698         * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16699         * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16700         * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16701         definition.
16702         * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16703         * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16704         Redefine.
16705         (rs6000_slow_unaligned_access): New function.
16706         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16707         (expand_block_compare): Likewise.
16708         (expand_strn_compare): Likewise.
16709         (rs6000_rtx_costs): Likewise.
16710         * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16711         (riscv_slow_unaligned_access): Likewise.
16712         * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16713         (riscv_slow_unaligned_access_p): ...this and make static.
16714         (riscv_option_override): Update accordingly.
16715         (riscv_slow_unaligned_access): New function.
16716         (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16717         * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16718         * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16719         (rs6000_slow_unaligned_access): New function.
16720         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16721         (rs6000_rtx_costs): Likewise.
16722         * config/rs6000/rs6000-string.c (expand_block_compare)
16723         (expand_strn_compare): Use targetm.slow_unaligned_access instead
16724         of SLOW_UNALIGNED_ACCESS.
16725         * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16726         * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16727         * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16728         of SLOW_UNALIGNED_ACCESS.
16729         * expmed.c (simple_mem_bitfield_p): Likewise.
16730         * expr.c (alignment_for_piecewise_move): Likewise.
16731         (emit_group_load_1): Likewise.
16732         (emit_group_store): Likewise.
16733         (copy_blkmode_from_reg): Likewise.
16734         (emit_push_insn): Likewise.
16735         (expand_assignment): Likewise.
16736         (store_field): Likewise.
16737         (expand_expr_real_1): Likewise.
16738         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16739         * lra-constraints.c (simplify_operand_subreg): Likewise.
16740         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16741         * gimple-ssa-store-merging.c: Likewise in block comment at start
16742         of file.
16743         * tree-ssa-strlen.c: Include target.h.
16744         (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16745         of SLOW_UNALIGNED_ACCESS.
16746         * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16748 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16750         PR rtl-optimization/82185
16751         * expmed.c (emit_store_flag_int): Only test tem if it has been
16752         initialized.
16754 2017-09-12  Richard Biener  <rguenther@suse.de>
16756         PR middle-end/82149
16757         * match.pd ((FTYPE) N CMP CST): Fix typo.
16759 2017-09-12  Simon Atanasyan  <simon.atanasyan@imgtec.com>
16761         * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16762         attribute.
16763         (mips_near_type_p): Add 'short_call' attribute as a synonym
16764         for 'near'.
16765         * doc/extend.texi (short_call): Document new function attribute.
16767 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16769         PR target/82112
16770         * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16771         assertion check that in the condition.
16772         (get_atomic_generic_size): Likewise.  Before testing if parameter
16773         has pointer type, if it has array type, call for C++
16774         default_conversion to perform array-to-pointer conversion.
16776 2017-09-12  Richard Biener  <rguenther@suse.de>
16778         * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16779         for operations we cannot scalarize.
16781 2017-09-12  Aldy Hernandez  <aldyh@redhat.com>
16783         * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16784         vectors heap vectors.  Clean up comments.
16785         Make visited_bbs a reference.
16786         (profitable_jump_thread_path): Make GC
16787         vectors heap vectors.  Clean up comments.
16788         Misc cleanups.
16789         (convert_and_register_jump_thread_path): Make GC vectors heap
16790         vectors.
16791         (check_subpath_and_update_thread_path): Same.  Clean up comments.
16792         Make visited_bbs a reference.
16793         (handle_phi): Abstract common code to to
16794         register_jump_thread_path_if_profitable.
16795         Rename VAR_BB to DEF_BB.
16796         Update comments.
16797         Make GC vectors heap vectors.
16798         Make visited_bbs a reference.
16799         (handle_assignment): Same.
16800         (register_jump_thread_path_if_profitable): New.
16801         (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16802         DEF_BB.
16803         Make GC vectors heap vectors.  Clean up comments.
16804         Make visited_bbs a reference.
16805         (find_jump_threads_backwards): Make visited_bbs live in the stack.
16806         * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16807         comment.
16809 2017-09-11  Max Filippov  <jcmvbkbc@gmail.com>
16811         PR target/82181
16812         * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16813         words of E_DImode object are reachable by xtensa_uimm8x4 access.
16815 2017-09-11  Vidya Praveen  <vidyapraveen@arm.com>
16817         Revert r251800 and r251799.
16819 2017-09-11  Martin Jambor  <mjambor@suse.cz>
16821         PR hsa/82119
16822         * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16823         arguments in advance.
16824         * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16825         use it to find predecessor edges.
16826         (naive_outof_ssa): Collect vector of predecessors.
16828 2017-09-08  Jason Merrill  <jason@redhat.com>
16830         PR c++/70029 - ICE with ref-qualifier and -flto
16831         * langhooks.h (struct lang_hooks_for_types): Add
16832         copy_lang_qualifiers.
16833         * attribs.c (build_type_attribute_qual_variant): Use it.
16834         * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16835         NULL.
16836         (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16837         * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16839 2017-09-08  Eric Botcazou  <ebotcazou@adacore.com>
16841         PR target/81988
16842         * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16843         (*mulsi3_sp64): New instruction.
16844         (mulsi3): New expander.
16846 2017-09-08  Uros Bizjak  <ubizjak@gmail.com>
16848         * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16850 2017-09-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16852         * sancov.c: Include memmodel.h.
16854 2017-09-07  Eric Botcazou  <ebotcazou@adacore.com>
16856         PR target/80897
16857         * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16858         large offsets.
16860 2017-09-07  Carl Love  <cel@us.ibm.com>
16862         * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16863         the sldi instruction.
16865 2017-09-07  David Edelsohn  <dje.gcc@gmail.com>
16867         * sancov.c: Include tm_p.h.
16869 2017-09-07  Jakub Jelinek  <jakub@redhat.com>
16871         PR target/81979
16872         * output.h (switch_to_other_text_partition): New declaration.
16873         * varasm.c (switch_to_other_text_partition): New function.
16874         * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16875         NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16876         (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16877         to the other text partition before emitting LCL label and switch back
16878         after emitting the word after it.
16880 2017-09-07  Richard Biener  <rguenther@suse.de>
16882         * passes.def (pass_split_crit_edges): Remove instance before PRE.
16883         * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16884         critical edges here, after loop init.
16885         (pass_data_pre): Remove PROP_no_crit_edges flags.
16886         * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16887         for valueization of call args to avoid leaking VN_TOP.
16888         (visit_use): Assert we do not visit default defs.
16889         (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16890         Use error_mark_node to more easily detect leaking VN_TOP.
16891         All default-defs are varying, not VN_TOP.  Mark them visited.
16892         (run_scc_vn): Make code match comment.
16894 2017-09-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
16896         * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16897         OPTION_MASK_FLOAT128_KEYWORD.
16898         (POWERPC_MASKS): Likewise.
16899         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
16900         support for the -mfloat128-type option, and make -mfloat128
16901         default on PowerPC Linux systems.  Define or undefine
16902         __FLOAT128__ and  __FLOAT128_HARDWARE__ for the current options.
16903         Define __float128 to be __ieee128 if IEEE 128-bit support is
16904         enabled, or undefine it.
16905         (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
16906         Delete defining __FLOAT128_TYPE__.
16907         * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
16908         -mfloat128-type option and make -mfloat128 default on PowerPC
16909         Linux systems.
16910         (TARGET_FLOAT128_TYPE): Likewise.
16911         (-mfloat128-type): Likewise.
16912         * config/rs6000/rs6000.c (rs6000_option_override_internal):
16913         Delete the -mfloat128-type option and make -mfloat128 default on
16914         PowerPC Linux systems.  Always use __ieee128 to be the keyword for
16915         the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
16916         128-bit floating point is enabled.  Change tests from using
16917         -mfloat128-type to -mfloat128.
16918         (rs6000_mangle_type): Use the correct mangling for the __float128
16919         type even if normal long double is restricted to 64-bits.
16920         (floatn_mode): Enable the _Float128 type by default on VSX Linux
16921         systems.
16922         * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
16923         (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
16924         (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
16925         -mfloat128-type.
16926         * doc/invoke.texi (RS/6000 and PowerPC Options): Update
16927         documentation for -mfloat128.
16929 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16931         * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
16933 2017-09-06  Wish Wu  <wishwu007@gmail.com>
16934             Jakub Jelinek  <jakub@redhat.com>
16936         * asan.c (initialize_sanitizer_builtins): Add
16937         BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
16938         BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
16939         BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
16940         BT_FN_VOID_UINT64_PTR variables.
16941         * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
16942         (BT_FN_VOID_UINT16_UINT16): Likewise.
16943         (BT_FN_VOID_UINT32_UINT32): Likewise.
16944         (BT_FN_VOID_FLOAT_FLOAT): Likewise.
16945         (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
16946         (BT_FN_VOID_UINT64_PTR): Likewise.
16947         * common.opt (flag_sanitize_coverage): New variable.
16948         (fsanitize-coverage=trace-pc): Remove.
16949         (fsanitize-coverage=): Add.
16950         * flag-types.h (enum sanitize_coverage_code): New enum.
16951         * fold-const.c (fold_range_test): Disable non-short-circuit
16952         optimization if flag_sanitize_coverage.
16953         (fold_truth_andor): Likewise.
16954         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16955         * opts.c (COVERAGE_SANITIZER_OPT): Define.
16956         (coverage_sanitizer_opts): New array.
16957         (get_closest_sanitizer_option): Add OPTS argument, handle also
16958         OPT_fsanitize_coverage_.
16959         (parse_sanitizer_options): Adjusted to also handle
16960         OPT_fsanitize_coverage_.
16961         (common_handle_option): Add OPT_fsanitize_coverage_.
16962         * sancov.c (instrument_comparison, instrument_switch): New function.
16963         (sancov_pass): Add trace-cmp support.
16964         * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
16965         BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
16966         BUILT_IN_SANITIZER_COV_TRACE_CMP8,
16967         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
16968         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
16969         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
16970         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
16971         BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
16972         BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
16973         * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
16975 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16977         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
16978         error.  Only quit immediately if parsing is complete.
16979         (BEGIN): Initialize fatal_err and parse_done.
16980         (begin fpu, end fpu): Check number of arguments.
16981         (begin arch, end arch): Likewise.
16982         (begin cpu, end cpu): Likewise.
16983         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
16984         (optalias): Likewise.
16986 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16988         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
16989         * config/arm/arm-isa.h: Delete.  Move definitions to ...
16990         * arm-cpus.in: ... here.  Use new feature and fgroup values.
16991         * config/arm/arm.c (arm_option_override): Use lower case for feature
16992         bit names.
16993         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
16994         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
16995         * config/arm/parsecpu.awk (END): Add new command 'isa'.
16996         (isa_pfx): Delete.
16997         (print_isa_bits_for): New function.
16998         (gen_isa): New function.
16999         (gen_comm_data): Use print_isa_bits_for.
17000         (define feature): New keyword.
17001         (define fgroup): New keyword.
17002         * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
17003         (arm-isa.h): Add rule to generate file.
17004         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
17005         case for feature bit names.
17007 2017-09-06  Richard Biener  <rguenther@suse.de>
17009         * tree-ssa-pre.c (NECESSARY): Remove.
17010         (create_expression_by_pieces): Do not touch pass-local flags.
17011         (insert_into_preds_of_block): Likewise.
17012         (do_pre_regular_insertion): Likewise.
17013         (eliminate_insert): Likewise.
17014         (eliminate_dom_walker::before_dom_children): Likewise.
17015         (fini_eliminate): Do not look at inserted_exprs.
17016         (mark_operand_necessary): Remove.
17017         (remove_dead_inserted_code): Replace with simple work-list
17018         algorithm based on inserted_exprs and SSA uses.
17019         (pass_pre::execute): Re-order fini_eliminate and
17020         remove_dead_inserted_code.
17022 2017-09-06  Olivier Hainque  <hainque@adacore.com>
17024         * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
17025         for VxWorks 7.  Adjust surrounding comments.
17027 2017-09-06  Richard Biener  <rguenther@suse.de>
17029         * gimple-ssa-strength-reduction.c
17030         (find_candidates_dom_walker::before_dom_children): Also allow
17031         pointer types.
17033 2017-09-06  Richard Biener  <rguenther@suse.de>
17035         PR tree-optimization/82108
17036         * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
17037         for gap in the non-permutation SLP case.
17039 2017-09-06  Martin Jambor  <mjambor@suse.cz>
17041         PR tree-optimization/82078
17042         * tree-sra.c (sort_and_splice_var_accesses): Move call to
17043         add_access_to_work_queue...
17044         (build_accesses_from_assign): ...here.
17045         (propagate_all_subaccesses): Make sure racc is the group
17046         representative, if there is one.
17048 2017-09-06  Jakub Jelinek  <jakub@redhat.com>
17050         PR middle-end/82095
17051         * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
17052         NULL DECL_INITIAL.
17054 2017-09-06  Richard Biener  <rguenther@suse.de>
17056         * gimple-ssa-strength-reduction.c
17057         (find_candidates_dom_walker::before_doom_children): Use a
17058         type and not a mode check.
17060 2017-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17062         PR target/77308
17063         * config/arm/predicates.md (arm_general_adddi_operand): Create new
17064         non-vfp predicate.
17065         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
17067 2017-09-05  Jeff Law  <law@redhat.com>
17069         PR tree-optimization/64910
17070         * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
17071         cases where we have 3 or more operands.
17073 2017-09-05  Jakub Jelinek  <jakub@redhat.com>
17075         PR middle-end/81768
17076         * omp-low.c (lower_omp_for): Recompute tree invariant if
17077         gimple_omp_for_initial/final is ADDR_EXPR.
17079         PR middle-end/81768
17080         * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
17081         into gimple val before gimplification fo the COND_EXPR.
17083 2017-09-05  Aldy Hernandez  <aldyh@redhat.com>
17085         * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
17086         REGION_COPY argument.
17087         (thread_through_all_blocks): Remove unused argument to
17088         duplicate_thread_path.
17090 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17091             Alan Hayward  <alan.hayward@arm.com>
17092             David Sherwood  <david.sherwood@arm.com>
17094         * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
17095         Take a scalar_mode rather than a machine_mode.
17096         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17097         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
17098         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17099         (aarch64_gen_adjusted_ldpstp): Likewise.
17100         (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
17102 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17103             Alan Hayward  <alan.hayward@arm.com>
17104             David Sherwood  <david.sherwood@arm.com>
17106         * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
17107         Take a scalar_int_mode instead of a machine_mode.
17108         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17109         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17110         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17111         (aarch64_simd_attr_length_rglist): Delete.
17112         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
17113         a scalar_int_mode instead of a machine_mode.
17114         (aarch64_add_offset): Likewise.
17115         (aarch64_internal_mov_immediate): Likewise
17116         (aarch64_add_constant_internal): Likewise.
17117         (aarch64_add_constant): Likewise.
17118         (aarch64_movw_imm): Likewise.
17119         (aarch64_rtx_arith_op_extract_p): Likewise.
17120         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17121         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17122         Remove assert that the mode isn't a vector.
17123         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17124         (aarch64_expand_mov_immediate): Update calls after above changes.
17125         (aarch64_output_casesi): Use as_a <scalar_int_mode>.
17126         (aarch64_and_bitmask_imm): Check for scalar integer modes.
17127         (aarch64_move_imm): Likewise.
17128         (aarch64_can_const_movi_rtx_p): Likewise.
17129         (aarch64_strip_extend): Likewise.
17130         (aarch64_extr_rtx_p): Likewise.
17131         (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
17132         a CONST_INT when the mode parameter is VOIDmode.
17133         (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
17135 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17137         * machmode.h (bitwise_mode_for_mode): Return opt_mode.
17138         * stor-layout.c (bitwise_mode_for_mode): Likewise.
17139         (bitwise_type_for_mode): Update accordingly.
17141 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17143         * stor-layout.h (mode_for_size_tree): Return an opt_mode.
17144         * stor-layout.c (mode_for_size_tree): Likewise.
17145         (mode_for_array): Update accordingly.
17146         (layout_decl): Likewise.
17147         (compute_record_mode): Likewise.  Only set the mode once.
17149 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17151         * target.def (get_mask_mode): Change return type to opt_mode.
17152         Expand commentary.
17153         * doc/tm.texi: Regenerate.
17154         * targhooks.h (default_get_mask_mode): Return an opt_mode.
17155         * targhooks.c (default_get_mask_mode): Likewise.
17156         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17157         * optabs-query.c (can_vec_mask_load_store_p): Update use of
17158         targetm.get_mask_mode.
17159         * tree.c (build_truth_vector_type): Likewise.
17161 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17163         * machmode.h (mode_for_vector): Return an opt_mode.
17164         * stor-layout.c (mode_for_vector): Likewise.
17165         (mode_for_int_vector): Update accordingly.
17166         (layout_type): Likewise.
17167         * config/i386/i386.c (emit_memmov): Likewise.
17168         (ix86_expand_set_or_movmem): Likewise.
17169         (ix86_expand_vector_init): Likewise.
17170         (ix86_get_mask_mode): Likewise.
17171         * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
17172         Likewise.
17173         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
17174         * expmed.c (extract_bit_field_1): Likewise.
17175         * expr.c (expand_expr_real_2): Likewise.
17176         * optabs-query.c (can_vec_perm_p): Likewise.
17177         (can_vec_mask_load_store_p): Likewise.
17178         * optabs.c (expand_vec_perm): Likewise.
17179         * targhooks.c (default_get_mask_mode): Likewise.
17180         * tree-vect-stmts.c (vectorizable_store): Likewise.
17181         (vectorizable_load): Likewise.
17182         (get_vectype_for_scalar_type_and_size): Likewise.
17184 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17186         * machmode.h (mode_for_int_vector): New function.
17187         * stor-layout.c (mode_for_int_vector): Likewise.
17188         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
17189         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
17190         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
17191         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
17192         (s390_expand_vcond): Likewise.
17194 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17196         * machmode.h (opt_machine_mode): New type.
17197         (opt_mode<T>): Allow construction from anything that can be
17198         converted to a T.
17199         (is_a, as_a, dyn_cast): Add overloads for opt_mode.
17200         (mode_for_size): Return an opt_machine_mode.
17201         * stor-layout.c (mode_for_size): Likewise.
17202         (mode_for_size_tree): Update call accordingly.
17203         (bitwise_mode_for_mode): Likewise.
17204         (make_fract_type): Likewise.
17205         (make_accum_type): Likewise.
17206         * caller-save.c (replace_reg_with_saved_mem): Update call
17207         accordingly.
17208         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17209         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17210         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17211         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17212         * expmed.c (extract_bit_field_1): Likewise.
17213         * reload.c (get_secondary_mem): Likewise.
17214         * varasm.c (assemble_integer): Likewise.
17215         * lower-subreg.c (simplify_subreg_concatn): Likewise.  Move
17216         early-out.
17218 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17220         * machmode.h (decimal_float_mode_for_size): New function.
17221         * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
17222         (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
17223         (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17224         (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17225         (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17226         (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17228 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17230         * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17231         (get_builtin_sync_mode): Likewise.
17232         (expand_ifn_atomic_compare_exchange): Likewise.
17233         (expand_builtin_atomic_clear): Likewise.
17234         (expand_builtin_atomic_test_and_set): Likewise.
17235         (fold_builtin_atomic_always_lock_free): Likewise.
17236         * calls.c (compute_argument_addresses): Likewise.
17237         (emit_library_call_value_1): Likewise.
17238         (store_one_arg): Likewise.
17239         * combine.c (combine_instructions): Likewise.
17240         * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17241         * config/arm/arm.c (arm_function_value): Likewise.
17242         (aapcs_allocate_return_reg): Likewise.
17243         * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17244         * config/i386/i386.c (construct_container): Likewise.
17245         (ix86_gimplify_va_arg): Likewise.
17246         (ix86_expand_sse_cmp): Likewise.
17247         (emit_memmov): Likewise.
17248         (emit_memset): Likewise.
17249         (expand_small_movmem_or_setmem): Likewise.
17250         (ix86_expand_pextr): Likewise.
17251         (ix86_expand_pinsr): Likewise.
17252         * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17253         * config/microblaze/microblaze.c (microblaze_block_move_straight):
17254         Likewise.
17255         * config/mips/mips.c (mips_function_value_1) Likewise.
17256         (mips_block_move_straight): Likewise.
17257         (mips_expand_ins_as_unaligned_store): Likewise.
17258         * config/powerpcspe/powerpcspe.c
17259         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17260         (rs6000_darwin64_record_arg_flush): Likewise.
17261         * config/rs6000/rs6000.c
17262         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17263         (rs6000_darwin64_record_arg_flush): Likewise.
17264         * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17265         (sparc_function_value_1): Likewise.
17266         * config/spu/spu.c (adjust_operand): Likewise.
17267         (spu_emit_branch_or_set): Likewise.
17268         (arith_immediate_p): Likewise.
17269         * emit-rtl.c (gen_lowpart_common): Likewise.
17270         * expr.c (expand_expr_real_1): Likewise.
17271         * function.c (assign_parm_setup_block): Likewise.
17272         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17273         * reload1.c (alter_reg): Likewise.
17274         * stor-layout.c (mode_for_vector): Likewise.
17275         (layout_type): Likewise.
17277 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17279         * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17280         (spu_convert_move): Likewise.
17281         * lower-subreg.c (resolve_simple_move): Likewise.
17283 2017-09-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17285         PR target/81833
17286         * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17287         define_insn to a define_expand.
17288         (altivec_vsum2sws_direct): New define_insn.
17289         (altivec_vsumsws): Convert from a define_insn to a define_expand.
17291 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17293         * config/arm/arm.c (arm_option_params_internal): Improve setting of
17294         max_insns_skipped.
17296 2017-09-05  H.J. Lu  <hongjiu.lu@intel.com>
17298         PR target/59501
17299         PR target/81624
17300         PR target/81769
17301         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17302         realign stack if stack alignment needed is less than incoming
17303         stack boundary.
17305 2017-09-05  Marek Polacek  <polacek@redhat.com>
17307         PR sanitizer/82072
17308         * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17309         check earlier.
17311 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17313         * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17315 2017-09-05  Richard Biener  <rguenther@suse.de>
17317         PR tree-optimization/82084
17318         * fold-const.c (can_native_encode_string_p): Handle wide characters.
17320 2017-09-05  Richard Biener  <rguenther@suse.de>
17322         PR tree-optimization/82102
17323         * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17325 2017-09-05  Martin Liska  <mliska@suse.cz>
17327         PR tree-optimization/82032
17328         * tree-cfg.c (generate_range_test): New function.
17329         * tree-cfg.h (generate_range_test): Declared here.
17330         * tree-cfgcleanup.c (convert_single_case_switch): New function.
17331         (cleanup_control_expr_graph): Use it.
17332         * tree-switch-conversion.c (try_switch_expansion): Remove
17333         assert.
17334         (emit_case_nodes): Use generate_range_test.
17336 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17338         PR target/82098
17339         * config/i386/i386.md (*<btsc><mode>_mask): Add
17340         TARGET_USE_BT to insn constraint.
17341         (*btr<mode>_mask): Ditto.
17343 2017-09-04  Wilco Dijkstra  <wdijkstr@arm.com>
17345         * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17346         (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17348 2017-09-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17350         PR target/77308
17351         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17352         TARGET_NEON and TARGET_IWMMXT.
17353         (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17354         TARGET_NEON and TARGET_IWMMXT.
17355         (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17357 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17359         * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17360         (ix86_rewrite_tls_address): Ditto.
17361         * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17362         (ix86_rewrite_tls_address_1): Ditto.
17363         (ix86_rewrite_tls_address): Ditto.
17364         * config/i386/predicates.md (tls_address_pattern): New predicate.
17365         * config/i386/i386.md (TLS address splitter): New splitter.
17367 2017-09-04  Richard Biener  <rguenther@suse.de>
17369         PR tree-optimization/82084
17370         * fold-const.h (can_native_encode_string_p): Declare.
17371         * fold-const.c (can_native_encode_string_p): Factor out from ...
17372         (native_encode_string): ... here.
17373         * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17374         vectorizing stores from constants we later cannot handle.
17376 2017-09-04  Marek Polacek  <polacek@redhat.com>
17378         PR c/81783
17379         * doc/invoke.texi: Update -Wtautological-compare documentation.
17381 2017-09-04  Jeff Law  <law@redhat.com>
17383         PR tree-optimization/64910
17384         * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17385         swap the first and last operand if the last is a constant.
17387 2017-09-04  Marek Polacek  <polacek@redhat.com>
17389         PR sanitizer/82072
17390         * convert.c (do_narrow): When sanitizing signed integer overflows,
17391         bail out for signed types.
17392         (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17394 2017-09-04  Richard Biener  <rguenther@suse.de>
17396         PR tree-optimization/82060
17397         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17398         Move devirtualization after stmt folding and before EH/AB/noreturn
17399         cleanup to get the stmt refs canonicalized.  Use a bool instead
17400         of gimple_modified_p since that doesn't work for NOPs.  Schedule
17401         NOPs generated by folding for removal.
17403 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17404             Alan Hayward  <alan.hayward@arm.com>
17405             David Sherwood  <david.sherwood@arm.com>
17407         * coretypes.h (pad_direction): New enum.
17408         * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17409         (FUNCTION_ARG_PADDING): Likewise.
17410         * target.def (function_arg_padding): New hook.
17411         * targhooks.h (default_function_arg_padding): Declare.
17412         * targhooks.c (default_function_arg_padding): New function.
17413         * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17414         (TARGET_FUNCTION_ARG_PADDING): ...this.
17415         * doc/tm.texi: Regenerate.
17416         * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17417         instead of direction.
17418         (compute_argument_addresses): Likewise.
17419         (load_register_parameters): Likewise.
17420         (emit_library_call_value_1): Likewise.
17421         (store_one_arg): Use targetm.calls.function_arg_padding instead
17422         of FUNCTION_ARG_PADDING.
17423         (must_pass_in_stack_var_size_or_pad): Likewise.
17424         * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17425         (emit_group_store): Likewise.
17426         (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17427         instead of FUNCTION_ARG_PADDING.
17428         (emit_push_insn): Likewise, and propagate enum change throughout
17429         function.
17430         * function.h (direction): Delete.
17431         (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17432         of direction.
17433         * function.c (assign_parm_find_stack_rtl): Likewise.
17434         (assign_parm_setup_block_p): Likewise.
17435         (assign_parm_setup_block): Likewise.
17436         (gimplify_parameters): Likewise.
17437         (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17438         instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17439         function.
17440         * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17441         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17442         * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17443         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17444         (aarch64_function_arg_padding): ...this new function.
17445         (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17446         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17447         * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17448         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17449         * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17450         * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17451         (arm_pad_arg_upward): Replace with...
17452         (arm_function_arg_padding): ...this new function.
17453         * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17454         of direction.
17455         * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17456         * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17457         * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17458         (ia64_hpux_function_arg_padding): Replace with...
17459         (ia64_function_arg_padding): ...this new function.  Use pad_direction
17460         instead of direction.  Check for TARGET_HPUX.
17461         * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17462         * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17463         (iq2000_function_arg_padding): New function.
17464         * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17465         * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17466         (mips_function_arg_padding): ...this new function.
17467         (mips_pad_reg_upward): Update accordingly.
17468         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17469         * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17470         targetm.calls.function_arg_padding.
17471         (FUNCTION_ARG_PADDING): Delete.
17472         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17473         * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17474         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17475         * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17476         (nios2_block_reg_padding): Return pad_direction instead of direction.
17477         * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17478         instead of direction.
17479         (nios2_function_arg_padding): Likewise.  Make static.
17480         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17481         * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17482         (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17483         * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17484         * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17485         (pa_function_arg_padding): Make static.  Return pad_direction instead
17486         of direction.
17487         * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17488         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17489         * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17490         instead of direction.  Use targetm.calls.function_arg_padding.
17491         * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17492         * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17493         * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17494         * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17495         * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17496         Redefine.
17497         (function_arg_padding): Rename to...
17498         (rs6000_function_arg_padding): ...this.  Make static.  Return
17499         pad_direction instead of direction.
17500         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17501         * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17502         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17503         * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17504         instead of direction.  Use targetm.calls.function_arg_padding.
17505         * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17506         * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17507         * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17508         * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17509         * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17510         (function_arg_padding): Rename to...
17511         (rs6000_function_arg_padding): ...this.  Make static.  Return
17512         pad_direction instead of direction.
17513         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17514         * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17515         * config/s390/s390.c (s390_function_arg_padding): New function.
17516         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17517         * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17518         * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17519         * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17520         (function_arg_padding): Rename to...
17521         (sparc_function_arg_padding): ...this.  Make static.  Return
17522         pad_direction instead of direction.
17523         * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17524         * config/spu/spu.c (spu_function_arg_padding): New function.
17525         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17526         * system.h (FUNCTION_ARG_PADDING): Poison.
17528 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17529             Alan Hayward  <alan.hayward@arm.com>
17530             David Sherwood  <david.sherwood@arm.com>
17532         * target.def (modes_tieable_p): New hook.
17533         * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17534         (TARGET_MODES_TIEABLE_P): ...this.
17535         * doc/tm.texi.in: Regenerate.
17536         * hooks.h (hook_bool_mode_mode_true): Declare.
17537         * hooks.c (hook_bool_mode_mode_true): New function.
17538         * combine.c (subst): Use targetm.modes_tieable_p instead of
17539         MODES_TIEABLE_P.
17540         * dse.c (find_shift_sequence): Likewise.
17541         * expmed.c (extract_low_bits): Likewise.
17542         * lower-subreg.c: Include target.h.
17543         (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17544         MODES_TIEABLE_P.
17545         * rtlanal.c (rtx_cost): Likewise.
17546         * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17547         * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17548         * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17549         (TARGET_MODES_TIEABLE_P): Redefine.
17550         * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17551         * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17552         (TARGET_MODES_TIEABLE_P): Redefine.
17553         * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17554         * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17555         (arc_modes_tieable_p): New function.
17556         * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17557         * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17558         * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17559         (arm_modes_tieable_p): Make static.
17560         * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17561         * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17562         * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17563         (TARGET_MODES_TIEABLE_P): Redefine.
17564         * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17565         * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17566         (TARGET_MODES_TIEABLE_P): Redefine.
17567         * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17568         * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17569         (cr16_modes_tieable_p): New function.
17570         * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17571         * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17572         * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17573         (TRULY_NOOP_TRUNCATION): Update comment.
17574         * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17575         (TRULY_NOOP_TRUNCATION): Update comment.
17576         * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17577         (frv_modes_tieable_p): New function.
17578         * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17579         * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17580         * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17581         (TARGET_MODES_TIEABLE_P): Redefine.
17582         * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17583         * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17584         * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17585         (TARGET_MODES_TIEABLE_P): Redefine.
17586         * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17587         * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17588         (ia64_modes_tieable_p): New function.
17589         * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17590         * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17591         (iq2000_modes_tieable_p): New function.
17592         * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17593         * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17594         (lm32_modes_tieable_p): New function.
17595         * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17596         * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17597         * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17598         (TARGET_MODES_TIEABLE_P): Redefine.
17599         * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17600         * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17601         (m32r_modes_tieable_p): New function.
17602         * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17603         * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17604         (m68k_modes_tieable_p): New function.
17605         * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17606         * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17607         (mcore_modes_tieable_p): New function.
17608         * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17609         * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17610         function.
17611         (TARGET_MODES_TIEABLE_P): Redefine.
17612         * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17613         * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17614         * config/mips/mips.c (mips_modes_tieable_p): Make static.
17615         (TARGET_MODES_TIEABLE_P): Redefine.
17616         * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17617         * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17618         * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17619         * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17620         (mn10300_modes_tieable_p): ...this and make static.
17621         (TARGET_MODES_TIEABLE_P): Redefine.
17622         * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17623         * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17624         * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17625         * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17626         (msp430_modes_tieable_p): Make static.
17627         * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17628         * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17629         (TARGET_MODES_TIEABLE_P): Redefine.
17630         * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17631         * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17632         * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17633         (TARGET_MODES_TIEABLE_P): Redefine.
17634         * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17635         * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17636         * config/pa/pa.c (pa_modes_tieable_p): Make static.
17637         (TARGET_MODES_TIEABLE_P): Redefine.
17638         * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17639         * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17640         (pdp11_modes_tieable_p): New function.
17641         * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17642         * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17643         (rs6000_modes_tieable_p): New function.
17644         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17645         * config/powerpcspe/powerpcspe.md: Update comment.
17646         * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17647         * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17648         (TARGET_MODES_TIEABLE_P): Redefine.
17649         * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17650         * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17651         (rl78_modes_tieable_p): New function.
17652         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17653         * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17654         (rs6000_modes_tieable_p): New function.
17655         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17656         * config/rs6000/rs6000.md: Update comment.
17657         * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17658         * config/rx/rx.c (rx_modes_tieable_p): New function.
17659         (TARGET_MODES_TIEABLE_P): Redefine.
17660         * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17661         * config/s390/s390.c (s390_modes_tieable_p): New function.
17662         (TARGET_MODES_TIEABLE_P): Redefine.
17663         * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17664         * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17665         (sh_modes_tieable_p): New function.
17666         * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17667         * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17668         * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17669         (sparc_modes_tieable_p): Make static.
17670         * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17671         * config/spu/spu.c (spu_modes_tieable_p): New function.
17672         (TARGET_MODES_TIEABLE_P): Redefine.
17673         * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17674         * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17675         (TARGET_MODES_TIEABLE_P): Redefine.
17676         * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17677         * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17678         * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17679         * config/v850/v850.c (v850_modes_tieable_p): New function.
17680         (TARGET_MODES_TIEABLE_P): Redefine.
17681         * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17682         * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17683         * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17684         (visium_modes_tieable_p): New function.
17685         * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17686         * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17687         (xtensa_modes_tieable_p): New function.
17688         * system.h (MODES_TIEABLE_P): Poison.
17690 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17691             Alan Hayward  <alan.hayward@arm.com>
17692             David Sherwood  <david.sherwood@arm.com>
17694         * target.def (hard_regno_mode_ok): New hook.
17695         * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17696         (TARGET_HARD_REGNO_MODE_OK): ...this.
17697         * doc/tm.texi.in: Regenerate.
17698         * hooks.h (hook_bool_uint_mode_true): Declare.
17699         * hooks.c (hook_bool_uint_mode_true): New function.
17700         * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17701         HARD_REGNO_MODE_OK.
17702         * genpreds.c (write_insn_preds_c): Add an include of target.h.
17703         * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17704         instead of HARD_REGNO_MODE_OK.
17705         * caller-save.c: Include target.h.
17706         (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17707         HARD_REGNO_MODE_OK.
17708         * combine.c (can_combine_p): Likewise.
17709         (combinable_i3pat): Likewise.
17710         (can_change_dest_mode): Likewise.
17711         * expr.c (init_expr_target): Likewise.
17712         (convert_move): Likewise.
17713         (convert_modes): Likewise.
17714         * ira.c (setup_prohibited_class_mode_regs): Likewise.
17715         (setup_prohibited_mode_move_regs): Likewise.
17716         * ira.h (target_ira): Likewise.
17717         * lra-assigns.c (find_hard_regno_for_1): Likewise.
17718         * lra-constraints.c (process_alt_operands): Likewise.
17719         (split_reg): Likewise.
17720         * recog.c (peep2_find_free_register): Likewise.
17721         * ree.c (combine_reaching_defs): Likewise.
17722         * regcprop.c (maybe_mode_change): Likewise.
17723         * reginfo.c (init_reg_sets_1): Likewise.
17724         (choose_hard_reg_mode): Likewise.
17725         (simplifiable_subregs): Likewise.
17726         * regrename.c (check_new_reg_p): Likewise.
17727         * reload.c (find_valid_class): Likewise.
17728         (find_valid_class_1): Likewise.
17729         (reload_inner_reg_of_subreg): Likewise.
17730         (push_reload): Likewise.
17731         (combine_reloads): Likewise.
17732         (find_dummy_reload): Likewise.
17733         (find_reloads): Likewise.
17734         * reload1.c (find_reg): Likewise.
17735         (set_reload_reg): Likewise.
17736         (allocate_reload_reg): Likewise.
17737         (choose_reload_regs): Likewise.
17738         (reload_adjust_reg_for_temp): Likewise.
17739         * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17740         (simplify_subreg_regno): Likewise.
17741         * sel-sched.c (init_regs_for_mode): Likewise.
17742         * varasm.c (make_decl_rtl): Likewise.
17743         * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17744         (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17745         HARD_REGNO_MODE_OK.
17746         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17747         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17748         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17749         * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17750         * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17751         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17752         * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17753         (arc_mode_class): Delete.
17754         (HARD_REGNO_MODE_OK): Delete.
17755         * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17756         (arc_hard_regno_mode_ok): Rename old array to...
17757         (arc_hard_regno_mode_ok_modes): ...this.
17758         (arc_conditional_register_usage): Update accordingly.
17759         (arc_mode_class): Make static.
17760         (arc_hard_regno_mode_ok): New function.
17761         * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17762         * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17763         * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17764         (arm_hard_regno_mode_ok): Make static.
17765         * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17766         HARD_REGNO_MODE_OK.
17767         * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17768         * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17769         * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17770         return a bool.
17771         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17772         * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17773         * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17774         * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17775         (bfin_hard_regno_mode_ok): ...this.  Make static and return a bool.
17776         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17777         * config/bfin/predicates.md (valid_reg_operand): Use
17778         targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17779         * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17780         * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17781         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17782         * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17783         * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17784         * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17785         (cr16_hard_regno_mode_ok): Make static and return a bool.
17786         * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17787         * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17788         (cris_hard_regno_mode_ok): New function.
17789         * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17790         (epiphany_mode_class): Delete.
17791         (HARD_REGNO_MODE_OK): Delete.
17792         * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17793         * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17794         (hard_regno_mode_ok): Rename to...
17795         (epiphany_hard_regno_mode_ok): ...this.  Make static and return a bool.
17796         * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17797         * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17798         HARD_REGNO_MODE_OK.
17799         * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17800         * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17801         * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17802         (frv_hard_regno_mode_ok): Make static and return a bool.
17803         * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17804         HARD_REGNO_MODE_OK.
17805         * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17806         * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17807         * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17808         * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17809         and return a bool.
17810         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17811         * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17812         * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17813         * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17814         return a bool.
17815         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17816         * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17817         * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17818         (ia64_hard_regno_mode_ok): New function.
17819         * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17820         * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17821         (iq2000_hard_regno_mode_ok): New function.
17822         * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17823         * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17824         (lm32_hard_regno_mode_ok): New function.
17825         * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17826         * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17827         * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17828         instead of HARD_REGNO_MODE_OK.
17829         (m32c_hard_regno_ok): Rename to...
17830         (m32c_hard_regno_mode_ok): ...this.  Make static and return a bool.
17831         (m32c_cannot_change_mode_class): Update accordingly.
17832         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17833         * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17834         (m32r_mode_class): Delete.
17835         (HARD_REGNO_MODE_OK): Delete.
17836         * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17837         (m32r_hard_regno_mode_ok): Rename to...
17838         (m32r_hard_regno_modes): ...this.
17839         (m32r_mode_class): Make static.
17840         (m32r_hard_regno_mode_ok): New function.
17841         * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17842         * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17843         * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17844         (m68k_hard_regno_mode_ok): Make static.
17845         * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17846         * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17847         (mcore_hard_regno_mode_ok): New function.
17848         * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17849         (HARD_REGNO_MODE_OK): Delete.
17850         * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17851         Rename to...
17852         (microblaze_hard_regno_mode_ok_p): ...this and make static.
17853         (microblaze_hard_regno_mode_ok): New function.
17854         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17855         * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17856         (mips_hard_regno_mode_ok): Delete.
17857         * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17858         (mips_hard_regno_mode_ok_p): ...this and make static.
17859         (mips_hard_regno_mode_ok_p): Rename to...
17860         (mips_hard_regno_mode_ok_uncached): ...this.
17861         (mips_hard_regno_mode_ok): New function.
17862         (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17863         of HARD_REGNO_MODE_OK.
17864         (mips_option_override): Update after above name changes.
17865         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17866         * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17867         * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17868         * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17869         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17870         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17871         * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17872         * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17873         * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17874         * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17875         (msp430_hard_regno_mode_ok): Make static and return a bool.
17876         * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17877         * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17878         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17879         and return a bool.
17880         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17881         * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17882         * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17883         * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17884         * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17885         (PA_HARD_REGNO_MODE_OK): ...this
17886         * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17887         (PA_HARD_REGNO_MODE_OK): ...this.
17888         * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17889         (pa_hard_regno_mode_ok): New function.
17890         * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17891         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17892         (pdp11_hard_regno_mode_ok): New function.
17893         * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17894         * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17895         Delete.
17896         * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17897         Make static.
17898         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17899         (rs6000_hard_regno_mode_ok): Rename to...
17900         (rs6000_hard_regno_mode_ok_uncached): ...this.
17901         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17902         (rs6000_hard_regno_mode_ok): New function.
17903         * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
17904         * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
17905         * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
17906         (riscv_hard_regno_mode_ok): ...this and make static.
17907         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17908         * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
17909         * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
17910         * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17911         (rl78_hard_regno_mode_ok): Make static and return bool.
17912         * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
17913         * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
17914         Delete.
17915         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
17916         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17917         (rs6000_hard_regno_mode_ok): Rename to...
17918         (rs6000_hard_regno_mode_ok_uncached): ...this.
17919         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17920         (rs6000_hard_regno_mode_ok): New function.
17921         * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
17922         * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
17923         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17924         * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
17925         * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
17926         * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
17927         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17928         * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
17929         * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
17930         * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17931         (sh_hard_regno_mode_ok): Make static.
17932         * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
17933         instead of HARD_REGNO_MODE_OK.
17934         * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
17935         (sparc_mode_class): Delete.
17936         (HARD_REGNO_MODE_OK): Delete.
17937         * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17938         (hard_regno_mode_classes): Make static.
17939         (sparc_mode_class): Likewise.
17940         (sparc_hard_regno_mode_ok): New function.
17941         * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
17942         * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
17943         * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
17944         function.
17945         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17946         * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
17947         * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
17948         * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
17949         * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
17950         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17951         * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
17952         * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
17953         * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17954         (visium_hard_regno_mode_ok): New function.
17955         * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
17956         instead of HARD_REGNO_MODE_OK.
17957         * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
17958         (HARD_REGNO_MODE_OK): Delete.
17959         * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
17960         (xtensa_hard_regno_mode_ok_p): ...this and make static.
17961         (xtensa_option_override): Update accordingly.
17962         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17963         (xtensa_hard_regno_mode_ok): New function.
17964         * system.h (HARD_REGNO_MODE_OK): Poison.
17966 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17967             Alan Hayward  <alan.hayward@arm.com>
17968             David Sherwood  <david.sherwood@arm.com>
17970         * target.def (hard_regno_call_part_clobbered): New hook.
17971         * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
17972         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
17973         * doc/tm.texi: Regenerate.
17974         * hooks.h (hook_bool_uint_mode_false): Declare.
17975         * hooks.c (hook_bool_uint_mode_false): New function.
17976         * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17977         * cselib.c (cselib_process_insn): Use
17978         targetm.hard_regno_call_part_clobbered instead of
17979         HARD_REGNO_CALL_PART_CLOBBERED.
17980         * ira-conflicts.c (ira_build_conflicts): Likewise.
17981         * ira-costs.c (ira_tune_allocno_costs): Likewise.
17982         * lra-constraints.c (need_for_call_save_p): Likewise.
17983         * lra-lives.c: Include target.h.
17984         (check_pseudos_live_through_calls): Use
17985         targetm.hard_regno_call_part_clobbered instead of
17986         HARD_REGNO_CALL_PART_CLOBBERED.
17987         * regcprop.c: Include target.h.
17988         (copyprop_hardreg_forward_1): Use
17989         targetm.hard_regno_call_part_clobbered instead of
17990         HARD_REGNO_CALL_PART_CLOBBERED.
17991         * reginfo.c (choose_hard_reg_mode): Likewise.
17992         * regrename.c (check_new_reg_p): Likewise.
17993         * reload.c (find_equiv_reg): Likewise.
17994         * reload1.c (emit_reload_insns): Likewise.
17995         * sched-deps.c (deps_analyze_insn): Likewise.
17996         * sel-sched.c (init_regs_for_mode): Likewise.
17997         (mark_unavailable_hard_regs): Likewise.
17998         * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
17999         * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18000         * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
18001         New function.
18002         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18003         * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18004         * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
18005         Delete.
18006         * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
18007         and return a bool.
18008         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18009         * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18010         * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
18011         function.
18012         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18013         * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18014         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
18015         function.
18016         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18017         * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
18018         Delete.
18019         * config/powerpcspe/powerpcspe.c
18020         (rs6000_hard_regno_call_part_clobbered): New function.
18021         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18022         * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18023         * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
18024         New function.
18025         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18026         * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18027         * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
18028         function.
18029         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
18030         * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
18031         * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
18033 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18034             Alan Hayward  <alan.hayward@arm.com>
18035             David Sherwood  <david.sherwood@arm.com>
18037         * rtl.h (subreg_memory_offset): Declare.
18038         * emit-rtl.c (subreg_memory_offset): New function.
18039         * expmed.c (store_bit_field_1): Use it.
18040         * expr.c (undefined_operand_subword_p): Likewise.
18041         * simplify-rtx.c (simplify_subreg): Likewise.
18043 2017-09-04  Alexander Monakov  <amonakov@ispras.ru>
18045         PR rtl-optimization/57448
18046         PR target/67458
18047         PR target/81316
18048         * optabs.c (expand_atomic_load): Place compiler memory barriers if
18049         using atomic_load pattern.
18050         (expand_atomic_store): Likewise.
18052 2017-09-04  Jakub Jelinek  <jakub@redhat.com>
18054         PR sanitizer/81981
18055         * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
18056         and UBSAN_BOUNDS internal calls.  Clean up IFN_UBSAN_OBJECT_SIZE
18057         handling.  Use replace_call_with_value with NULL instead of
18058         gsi_replace, unlink_stmt_vdef and release_defs.
18060         * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
18061         instead of tab.
18063         * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
18065 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
18067         PR bootstrap/82045
18068         * rtl.h (emit_library_call_value_1): Declare.
18069         (emit_library_call): Replace declaration with a series of overloads.
18070         Remove the parameter count argument.
18071         (emit_library_call_value): Likewise.
18072         * calls.c (emit_library_call_value_1): Make global.  Replace varargs
18073         with an "rtx_mode_t *".
18074         (emit_library_call_value): Delete.
18075         (emit_library_call): Likewise.
18076         * asan.c (asan_emit_stack_protection): Update calls accordingly.
18077         (asan_emit_allocas_unpoison): Likewise.
18078         * builtins.c (expand_builtin_powi): Likewise.
18079         (expand_asan_emit_allocas_unpoison): Likewise.
18080         * cfgexpand.c (expand_main_function): Likewise.
18081         * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
18082         * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
18083         * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
18084         * config/arm/arm.c (arm_trampoline_init): Likewise.
18085         (arm_call_tls_get_addr): Likewise.
18086         (arm_expand_divmod_libfunc): Likewise.
18087         * config/bfin/bfin.md (umulsi3_highpart): Likewise.
18088         (smulsi3_highpart): Likewise.
18089         * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
18090         (c6x_expand_compare): Likewise.
18091         (c6x_expand_movmem): Likewise.
18092         * config/frv/frv.c (frv_trampoline_init): Likewise.
18093         * config/i386/i386.c (ix86_trampoline_init): Likewise.
18094         (ix86_expand_divmod_libfunc): Likewise.
18095         * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
18096         (ia64_expand_compare): Likewise.
18097         (ia64_profile_hook): Likewise.
18098         * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
18099         (nonlocal_goto): Likewise.
18100         (restore_stack_nonlocal): Likewise.
18101         * config/m32r/m32r.c (block_move_call): Likewise.
18102         (m32r_trampoline_init): Likewise.
18103         * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
18104         * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
18105         (m68k_call_m68k_read_tp): Likewise.
18106         * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
18107         (microblaze_expand_divide): Likewise.
18108         * config/mips/mips.h (mips_args): Likewise.
18109         * config/mips/sdemtk.h (mips_sync_icache): Likewise.
18110         (MIPS_ICACHE_SYNC): Likewise.
18111         * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
18112         (nios2_trampoline_init): Likewise.
18113         * config/pa/pa.c (hppa_tls_call): Likewise.
18114         (pa_trampoline_init): Likewise.
18115         * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
18116         * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
18117         (expand_strn_compare): Likewise.
18118         (rs6000_generate_compare): Likewise.
18119         (rs6000_expand_float128_convert): Likewise.
18120         (output_profile_hook): Likewise.
18121         (rs6000_trampoline_init): Likewise.
18122         * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
18123         * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
18124         * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
18125         * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
18126         (rs6000_generate_compare): Likewise.
18127         (rs6000_expand_float128_convert): Likewise.
18128         (output_profile_hook): Likewise.
18129         (rs6000_trampoline_init): Likewise.
18130         * config/rs6000/rs6000.md (neg<mode>2): Likewise.
18131         * config/sh/sh.c (sh_trampoline_init): Likewise.
18132         * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
18133         (sparc_emit_float_lib_cmp): Likewise.
18134         (sparc32_initialize_trampoline): Likewise.
18135         (sparc64_initialize_trampoline): Likewise.
18136         (sparc_profile_hook): Likewise.
18137         * config/spu/spu.c (ea_load_store): Likewise.
18138         * config/spu/spu.md (floatunssidf2): Likewise.
18139         * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
18140         * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
18141         * config/visium/visium.c (expand_block_move_4): Likewise.
18142         (expand_block_move_2): Likewise.
18143         (expand_block_move_1): Likewise.
18144         (expand_block_set_4): Likewise.
18145         (expand_block_set_2): Likewise.
18146         (expand_block_set_1): Likewise.
18147         (visium_trampoline_init): Likewise.
18148         (visium_profile_hook): Likewise.
18149         * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
18150         (xtensa_setup_frame_addresses): Likewise.
18151         (xtensa_trampoline_init): Likewise.
18152         * except.c (sjlj_emit_function_enter): Likewise.
18153         (sjlj_emit_function_exit): Likewise.
18154         * explow.c (allocate_dynamic_stack_space): Likewise.
18155         (probe_stack_range): Likewise.
18156         * expr.c (convert_mode_scalar): Likewise.
18157         * optabs.c (expand_binop): Likewise.
18158         (expand_twoval_binop_libfunc): Likewise.
18159         (expand_unop): Likewise.
18160         (prepare_cmp_insn): Likewise.
18161         (prepare_float_lib_cmp): Likewise.
18162         (expand_float): Likewise.
18163         (expand_fix): Likewise.
18164         (expand_fixed_convert): Likewise.
18165         (maybe_emit_sync_lock_test_and_set): Likewise.
18166         (expand_atomic_compare_and_swap): Likewise.
18167         (expand_mem_thread_fence): Likewise.
18168         (expand_atomic_fetch_op): Likewise.
18170 2017-09-03  Gerald Pfeifer  <gerald@pfeifer.com>
18172         * doc/generic.texi (OpenACC): Adjust URL.
18173         * doc/invoke.texi (C Dialect Options): Ditto.
18175 2017-09-03  Uros Bizjak  <ubizjak@gmail.com>
18177         * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
18178         predicate for operand 1.  Add (m,<S>) constraint.
18179         (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
18180         Prevent memory operand 1 with register operand 2.
18182 2017-09-01  Segher Boessenkool  <segher@kernel.crashing.org>
18184         PR rtl-optimization/82024
18185         * combine.c (try_combine): If the combination result is a PARALLEL,
18186         and we only need to retain the SET in there that would be placed
18187         at I2, check that we can place that at I3 instead, before doing so.
18189 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18191         PR target/81766
18192         * config/i386/i386.c (ix86_init_large_pic_reg): Return label
18193         instead of void.
18194         (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
18195         if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
18196         and label.
18198 2017-09-01  Joerg Sonnenberger  <joerg@bec.de>
18199             Jeff Law  <law@redhat.com>
18201         * varasm.c (bss_initializer_p): Do not put constants into .bss
18202         (categorize_decl_for_section): Handle bss_initializer_p returning
18203         false when DECL_INITIAL is NULL.
18205 2017-09-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18207         PR target/82012
18208         * config/s390/s390.c (s390_can_inline_p): New function.
18210 2017-09-01  Jeff Law  <law@redhat.com>
18212         PR tree-optimization/82052
18213         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
18214         Always initialize the returned slot after a hash table miss
18215         when INSERT is true.
18217 2017-09-01  Alexander Monakov  <amonakov@ispras.ru>
18219         * config/s390/s390.md (mem_signal_fence): Remove.
18220         * doc/md.texi (mem_signal_fence): Remove.
18221         * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
18222         Update comments.
18223         * target-insns.def (mem_signal_fence): Remove.
18225 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18227         PR sanitizer/81902
18228         * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18230         PR sanitizer/81923
18231         * asan.c (create_odr_indicator): Strip name encoding from assembler
18232         name before appending it after __odr_asan_.
18234 2017-09-01  Martin Liska  <mliska@suse.cz>
18236         PR tree-optimization/82059
18237         * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18238         frequency only when an edge is redirected.
18240 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18242         * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18243         * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18244         (arc_conditional_register_usage): Remove ARC600 lp_count
18245         exception.
18246         (arc_file_start): Emit Tag_ARC_CPU_variation.
18247         (arc_can_use_doloop_p): New conditions to use ZOLs.
18248         (hwloop_fail): New function.
18249         (hwloop_optimize): Likewise.
18250         (hwloop_pattern_reg): Likewise.
18251         (arc_doloop_hooks): New struct, to be used with reorg_loops.
18252         (arc_reorg_loops): New function, calls reorg_loops.
18253         (arc_reorg): Call arc_reorg_loops.  Remove old ZOL handling.
18254         (arc600_corereg_hazard): Remove ZOL checking, case handled by
18255         hwloop_optimize.
18256         (arc_loop_hazard): Remove function, functionality moved into
18257         hwloop_optimize.
18258         (arc_hazard): Remove arc_loop_hazard call.
18259         (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18260         into hwloop_optimize.
18261         (arc_label_align): Remove ZOL handling.
18262         * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18263         * config/arc/arc.md (doloop_begin): Remove pattern.
18264         (doloop_begin_i): Likewise.
18265         (doloop_end_i): Likewise.
18266         (doloop_fallback): Likewise.
18267         (doloop_fallback_m): Likewise.
18268         (doloop_end): Reimplement expand.
18269         (arc_lp): New pattern for LP instruction.
18270         (loop_end): New pattern.
18271         (loop_fail): Likewise.
18272         (decrement_and_branch_until_zero): Likewise.
18273         * config/arc/arc.opt (mlpc-width): New option.
18274         * doc/invoke.texi (mlpc-width): Document option.
18276 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18278         * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18279         (arc_ccfsm_advance): Fix checking for delay slots.
18280         (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18282 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18284         * config/arc/arc.md (movqi_insn): Add stores to save constant long
18285         immediates.
18286         (movhi_insn): Update store instruction constraint which are saving
18287         6-bit short immediates.
18288         (movsi_insn): Consider also short scaled load operations.
18289         (zero_extendhisi2_i): Use Usd constraint instead of T.
18290         (extendhisi2_i): Add q constraint.
18291         (arc_clzsi2): Add type and length attributes.
18292         (arc_ctzsi2): Likewise.
18293         * config/arc/constraints.md (Usc): Update constraint, the
18294         assembler can parse two relocations for a single instruction.
18296 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18298         * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18299         (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18301 2017-08-31  Olivier Hainque  <hainque@adacore.com>
18303         * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18304         match as powerpc-wrs-vxworks*.
18306 2017-08-31  James Greenhalgh  <james.greenhalgh@arm.com>
18308         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18309         register constraint for by-element operand.
18310         (aarch64_mls_elt_merge<mode>): Likewise.
18312 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18314         * config/arc/arc.c (arc_can_follow_jump): Check for short
18315         branches.
18317 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18319         * config.gcc: Use g.opt for arc.
18320         * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18321         functionality moved to ...
18322         (legitimate_scaled_address_p): New function, ...here.
18323         (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18324         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18325         (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18326         condition.
18327         (arc_override_options): Handle G option.
18328         (arc_output_pic_addr_const): Correct function definition.
18329         (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18330         (arc_decl_anon_ns_mem_p): Delete.
18331         (arc_in_small_data_p): Overhaul this function to take into
18332         consideration the value given via G option.
18333         (arc_rewrite_small_data_1): Renamed and corrected old
18334         arc_rewrite_small_data function.
18335         (arc_rewrite_small_data): New function.
18336         (small_data_pattern): Don't use pic_offset_table_rtx.
18337         * config/arc/arc.h (CC1_SPEC): Recognize G option.
18338         * config/arc/simdext.md (movmisalignv2hi): Use
18339         prepare_move_operands function.
18340         (mov*): Likewise.
18341         (movmisalign*): Likewise.
18342         * doc/invoke.texi (ARC options): Document -G option.
18344 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18346         * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18347         prototype.
18348         * config/arc/arc.c (arc_print_operand): Output scalled address for
18349         sdata whenever is possible.
18350         (arc_in_small_data_p): Allow sdata for 64bit datum when double
18351         load/stores are available.
18352         (compact_sda_memory_operand): Check for the alignment required by
18353         code density instructions.
18354         * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18355         constraint.
18356         * config/arc/constraints.md (Usd): Update constraint.
18357         (Us0): New constraint.
18358         (Usc): Update constraint.
18360 2017-08-31  Richard Biener  <rguenther@suse.de>
18362         PR middle-end/82054
18363         * dwarf2out.c (dwarf2out_early_global_decl): Process each
18364         function only once.
18366 2017-08-31  Tamar Christina  <tamar.christina@arm.com>
18368         * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18369         Resize type_signature.
18371 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18372             Alan Hayward  <alan.hayward@arm.com>
18373             David Sherwood  <david.sherwood@arm.com>
18375         * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18376         subregs whose inner modes can be stored in GPRs.
18377         (aarch64_classify_index): Likewise.
18379 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18380             Alan Hayward  <alan.hayward@arm.com>
18381             David Sherwood  <david.sherwood@arm.com>
18383         * config/aarch64/iterators.md (V_cmp_result): Rename to...
18384         (V_INT_EQUIV): ...this.
18385         (v_cmp_result): Rename to...
18386         (v_int_equiv): ...this.
18387         * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18388         * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18389         (copysign<mode>3): Likewise.
18390         (aarch64_simd_bsl<mode>_internal): Likewise.
18391         (aarch64_simd_bsl<mode>): Likewise.
18392         (vec_cmp<mode><mode>): Likewise.
18393         (vcond<mode><mode>): Likewise.
18394         (vcond<v_cmp_mixed><mode>): Likewise.
18395         (vcondu<mode><v_cmp_mixed>): Likewise.
18396         (aarch64_cm<optab><mode>): Likewise.
18397         (aarch64_cmtst<mode>): Likewise.
18398         (aarch64_fac<optab><mode>): Likewise.
18399         (vec_perm_const<mode>): Likewise.
18400         (vcond_mask_<mode><v_cmp_result>): Rename to...
18401         (vcond_mask_<mode><v_int_equiv>): ...this.
18402         (vec_cmp<mode><v_cmp_result>): Rename to...
18403         (vec_cmp<mode><v_int_equiv>): ...this.
18405 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18406             Alan Hayward  <alan.hayward@arm.com>
18407             David Sherwood  <david.sherwood@arm.com>
18409         * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18410         vector modes.
18411         * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18412         * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18413         (UNSPEC_LD4_DREG): New unspecs.
18414         * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18415         (aarch64_ld2<mode>_dreg_be): Replace with...
18416         (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18417         unspec.
18418         (aarch64_ld3<mode>_dreg_le)
18419         (aarch64_ld3<mode>_dreg_be): Replace with...
18420         (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18421         unspec.
18422         (aarch64_ld4<mode>_dreg_le)
18423         (aarch64_ld4<mode>_dreg_be): Replace with...
18424         (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18425         unspec.
18427 2017-08-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18429         PR tree-optimization/81987
18430         * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18431         insert an initializer in a location not dominated by the stride
18432         definition.
18434 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18436         * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18437         on the entire header of the finally block in the fallthru case.
18439 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18441         * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18443 2017-08-30  Pat Haugen  <pthaugen@us.ibm.com>
18445         * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18446         rs6000_emit_move_from_cr and call renamed function.
18447         (rs6000_emit_prologue): Call renamed functions.
18448         * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18449         movesi_from_cr, remove volatile CRs.
18451 2017-08-30  Jon Beniston  <jon@beniston.com>
18452             Richard Biener  <rguenther@suse.de>
18454         * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18455         of VECTOR_MODE_P check.
18456         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18457         element vector types.
18459 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18461         * df.h (df_read_modify_subreg_p): Remove in favor of...
18462         * rtl.h (read_modify_subreg_p): ...this new function.  Take a
18463         const_rtx instead of an rtx.
18464         * cprop.c (local_cprop_find_used_regs): Update accordingly.
18465         * df-problems.c (df_word_lr_mark_ref): Likewise.
18466         * ira-lives.c (mark_pseudo_reg_live): Likewise.
18467         (mark_pseudo_reg_dead): Likewise.
18468         (mark_ref_dead): Likewise.
18469         * reginfo.c (init_subregs_of_mode): Likewise.
18470         * sched-deps.c (sched_analyze_1): Likewise.
18471         * df-scan.c (df_def_record_1): Likewise.
18472         (df_uses_record): Likewise.
18473         (df_read_modify_subreg_p): Remove in favor of...
18474         * rtlanal.c (read_modify_subreg_p): ...this new function.  Take a
18475         const_rtx instead of an rtx.
18477 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18478             Alan Hayward  <alan.hayward@arm.com>
18479             David Sherwood  <david.sherwood@arm.com>
18481         * rtl.h (partial_subreg_p): New function.
18482         * caller-save.c (save_call_clobbered_regs): Use it.
18483         * calls.c (expand_call): Likewise.
18484         * combine.c (combinable_i3pat): Likewise.
18485         (simplify_set): Likewise.
18486         (make_extraction): Likewise.
18487         (make_compound_operation_int): Likewise.
18488         (gen_lowpart_or_truncate): Likewise.
18489         (force_to_mode): Likewise.
18490         (make_field_assignment): Likewise.
18491         (reg_truncated_to_mode): Likewise.
18492         (record_truncated_value): Likewise.
18493         (move_deaths): Likewise.
18494         * cse.c (record_jump_cond): Likewise.
18495         (cse_insn): Likewise.
18496         * cselib.c (cselib_lookup_1): Likewise.
18497         * expmed.c (extract_bit_field_using_extv): Likewise.
18498         * function.c (assign_parm_setup_reg): Likewise.
18499         * ifcvt.c (noce_convert_multiple_sets): Likewise.
18500         * ira-build.c (create_insn_allocnos): Likewise.
18501         * lra-coalesce.c (merge_pseudos): Likewise.
18502         * lra-constraints.c (match_reload): Likewise.
18503         (simplify_operand_subreg): Likewise.
18504         (curr_insn_transform): Likewise.
18505         * lra-lives.c (process_bb_lives): Likewise.
18506         * lra.c (new_insn_reg): Likewise.
18507         (lra_substitute_pseudo): Likewise.
18508         * regcprop.c (mode_change_ok): Likewise.
18509         (maybe_mode_change): Likewise.
18510         (copyprop_hardreg_forward_1): Likewise.
18511         * reload.c (push_reload): Likewise.
18512         (find_reloads): Likewise.
18513         (find_reloads_subreg_address): Likewise.
18514         * reload1.c (alter_reg): Likewise.
18515         (eliminate_regs_1): Likewise.
18516         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18518 2017-08-30  David Edelsohn  <dje.gcc@gmail.com>
18520         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18521         back to if statements, including unpack.
18523 2017-08-30  Martin Liska  <mliska@suse.cz>
18525         PR inline-asm/82001
18526         * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18527         Rename to ...
18528         (func_checker::compare_asm_inputs_outputs): ... this function.
18529         (func_checker::compare_gimple_asm): Use the function to compare
18530         also ASM constrains.
18531         * ipa-icf-gimple.h: Rename the function.
18533 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18534             Alan Hayward  <alan.hayward@arm.com>
18535             David Sherwood  <david.sherwood@arm.com>
18537         * coretypes.h (complex_mode): New type.
18538         * gdbhooks.py (build_pretty_printer): Handle it.
18539         * machmode.h (complex_mode): New class.
18540         (complex_mode::includes_p): New function.
18541         (is_complex_int_mode): Likewise.
18542         (is_complex_float_mode): Likewise.
18543         * genmodes.c (get_mode_class): Handle complex mode classes.
18544         * function.c (expand_function_end): Use is_complex_int_mode.
18546 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18547             Alan Hayward  <alan.hayward@arm.com>
18548             David Sherwood  <david.sherwood@arm.com>
18550         * coretypes.h (scalar_mode_pod): New typedef.
18551         * gdbhooks.py (build_pretty_printer): Handle it.
18552         * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18553         * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18554         * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18555         * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18556         as_a <scalar_mode>.
18558 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18559             Alan Hayward  <alan.hayward@arm.com>
18560             David Sherwood  <david.sherwood@arm.com>
18562         * machmode.h (mode_for_vector): Take a scalar_mode instead
18563         of a machine_mode.
18564         * stor-layout.c (mode_for_vector): Likewise.
18565         * explow.c (promote_mode): Use as_a <scalar_mode>.
18566         * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
18568 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18569             Alan Hayward  <alan.hayward@arm.com>
18570             David Sherwood  <david.sherwood@arm.com>
18572         * target.def (preferred_simd_mode): Take a scalar_mode
18573         instead of a machine_mode.
18574         * targhooks.h (default_preferred_simd_mode): Likewise.
18575         * targhooks.c (default_preferred_simd_mode): Likewise.
18576         * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18577         * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18578         * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18579         * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18580         * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18581         * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18582         * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18583         * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18584         Likewise.
18585         * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18586         * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18587         * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18588         * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18589         (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18590         * doc/tm.texi: Regenerate.
18591         * optabs-query.c (can_vec_mask_load_store_p): Return false for
18592         non-scalar modes.
18594 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18595             Alan Hayward  <alan.hayward@arm.com>
18596             David Sherwood  <david.sherwood@arm.com>
18598         * target.def (scalar_mode_supported_p): Take a scalar_mode
18599         instead of a machine_mode.
18600         * targhooks.h (default_scalar_mode_supported_p): Likewise.
18601         * targhooks.c (default_scalar_mode_supported_p): Likewise.
18602         * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18603         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18604         * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18605         * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18606         * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18607         * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18608         * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18609         * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18610         * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18611         * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18612         * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18613         * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18614         Likewise.
18615         * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18616         * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18617         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18618         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18619         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18620         Likewise.
18621         * doc/tm.texi: Regenerate.
18623 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18624             Alan Hayward  <alan.hayward@arm.com>
18625             David Sherwood  <david.sherwood@arm.com>
18627         * coretypes.h (opt_scalar_mode): New typedef.
18628         * gdbhooks.py (build_pretty_printers): Handle it.
18629         * machmode.h (mode_iterator::get_2xwider): Add overload for
18630         opt_mode<T>.
18631         * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18632         over scalar modes.
18633         * expr.c (convert_mode_scalar): Likewise.
18634         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18635         * optabs.c (expand_float): Likewise.
18636         (expand_fix): Likewise.
18637         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18639 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18640             Alan Hayward  <alan.hayward@arm.com>
18641             David Sherwood  <david.sherwood@arm.com>
18643         * optabs.c (expand_float): Explicitly check for scalars before
18644         using a branching expansion.
18645         (expand_fix): Likewise.
18647 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18648             Alan Hayward  <alan.hayward@arm.com>
18649             David Sherwood  <david.sherwood@arm.com>
18651         * expr.c (convert_mode): Split scalar handling out into...
18652         (convert_mode_scalar): ...this new function.  Treat the modes
18653         as scalar_modes.
18655 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18656             Alan Hayward  <alan.hayward@arm.com>
18657             David Sherwood  <david.sherwood@arm.com>
18659         * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18660         and scalar_mode.
18661         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18663 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18664             Alan Hayward  <alan.hayward@arm.com>
18665             David Sherwood  <david.sherwood@arm.com>
18667         * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18668         rather than a machine_mode.
18669         (fixed_from_string): Likewise.
18670         (fixed_convert): Likewise.
18671         (fixed_convert_from_int): Likewise.
18672         (fixed_convert_from_real): Likewise.
18673         (real_convert_from_fixed): Likewise.
18674         * fixed-value.c (fixed_from_double_int): Likewise.
18675         (fixed_from_string): Likewise.
18676         (fixed_convert): Likewise.
18677         (fixed_convert_from_int): Likewise.
18678         (fixed_convert_from_real): Likewise.
18679         (real_convert_from_fixed): Likewise.
18680         * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
18682 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18683             Alan Hayward  <alan.hayward@arm.com>
18684             David Sherwood  <david.sherwood@arm.com>
18686         * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18687         of separate mode class checks.  Do not allow vector modes here.
18688         (immed_wide_int_const): Use as_a <scalar_mode>.
18689         * explow.c (trunc_int_for_mode): Likewise.
18690         * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18691         (wi::shwi): Likewise.
18692         (wi::min_value): Likewise.
18693         (wi::max_value): Likewise.
18694         * dwarf2out.c (loc_descriptor): Likewise.
18695         * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18696         for CONST_WIDE_INT.
18698 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18699             Alan Hayward  <alan.hayward@arm.com>
18700             David Sherwood  <david.sherwood@arm.com>
18702         * tree.h (SCALAR_TYPE_MODE): New macro.
18703         * expr.c (expand_expr_addr_expr_1): Use it.
18704         (expand_expr_real_2): Likewise.
18705         * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18706         (fold_convert_const_fixed_from_int): Likewise.
18707         (fold_convert_const_fixed_from_real): Likewise.
18708         (native_encode_fixed): Likewise
18709         (native_encode_complex): Likewise
18710         (native_encode_vector): Likewise.
18711         (native_interpret_fixed): Likewise.
18712         (native_interpret_real): Likewise.
18713         (native_interpret_complex): Likewise.
18714         (native_interpret_vector): Likewise.
18715         * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18716         (simd_clone_adjust_argument_types): Likewise.
18717         (simd_clone_init_simd_arrays): Likewise.
18718         (simd_clone_adjust): Likewise.
18719         * stor-layout.c (layout_type): Likewise.
18720         * tree.c (build_minus_one_cst): Likewise.
18721         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18722         * tree-inline.c (estimate_move_cost): Likewise.
18723         * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18724         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18725         (vectorizable_reduction): Likewise.
18726         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18727         (vect_recog_mixed_size_cond_pattern): Likewise.
18728         (check_bool_pattern): Likewise.
18729         (adjust_bool_pattern): Likewise.
18730         (search_type_for_mask_1): Likewise.
18731         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18732         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18733         (vectorizable_load): Likewise.
18734         (vectorizable_store): Likewise.
18735         * ubsan.c (ubsan_encode_value): Likewise.
18736         * varasm.c (output_constant): Likewise.
18738 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18739             Alan Hayward  <alan.hayward@arm.com>
18740             David Sherwood  <david.sherwood@arm.com>
18742         * coretypes.h (scalar_mode): New class.
18743         * machmode.h (scalar_mode): Likewise.
18744         (scalar_mode::includes_p): New function.
18745         (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18746         * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18747         * genmodes.c (get_mode_class): Handle remaining scalar modes.
18748         * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18749         * expmed.c (store_bit_field_1): Likewise.
18750         (extract_bit_field_1): Likewise.
18751         * expr.c (write_complex_part): Likewise.
18752         (read_complex_part): Likewise.
18753         (emit_move_complex_push): Likewise.
18754         (expand_expr_real_2): Likewise.
18755         * function.c (assign_parm_setup_reg): Likewise.
18756         (assign_parms_unsplit_complex): Likewise.
18757         * optabs.c (expand_binop): Likewise.
18758         * rtlanal.c (subreg_get_info): Likewise.
18759         * simplify-rtx.c (simplify_immed_subreg): Likewise.
18760         * varasm.c (output_constant_pool_2): Likewise.
18762 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18763             Alan Hayward  <alan.hayward@arm.com>
18764             David Sherwood  <david.sherwood@arm.com>
18766         * expmed.c (extract_high_half): Use scalar_int_mode and remove
18767         assertion.
18768         (expmed_mult_highpart_optab): Likewise.
18769         (expmed_mult_highpart): Likewise.
18771 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18772             Alan Hayward  <alan.hayward@arm.com>
18773             David Sherwood  <david.sherwood@arm.com>
18775         * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18776         instead of a machine_mode.
18777         (builtin_memset_read_str): Likewise.
18778         * builtins.c (c_readstr): Likewise.
18779         (builtin_memcpy_read_str): Likewise.
18780         (builtin_strncpy_read_str): Likewise.
18781         (builtin_memset_read_str): Likewise.
18782         (builtin_memset_gen_str): Likewise.
18783         (expand_builtin_signbit): Use scalar_int_mode for local variables.
18784         * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18785         instead of a machine_mode.
18786         * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18787         variables.
18788         (make_extraction): Likewise.
18789         (try_widen_shift_mode): Take and return scalar_int_modes instead
18790         of machine_modes.
18791         * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18792         a scalar_int_mode instead of a machine_mode.
18793         * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18794         (avr_addr_space_pointer_mode): Likewise.
18795         * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18796         * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18797         (msp430_unwind_word_mode): Likewise.
18798         * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18799         (spu_addr_space_pointer_mode): Likewise.
18800         (spu_addr_space_address_mode): Likewise.
18801         (spu_libgcc_cmp_return_mode): Likewise.
18802         (spu_libgcc_shift_count_mode): Likewise.
18803         * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18804         (rl78_addr_space_pointer_mode): Likewise.
18805         (fl78_unwind_word_mode): Likewise.
18806         (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18807         machine_mode.
18808         * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18809         * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18810         * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18811         (mips_valid_pointer_mode): Likewise.
18812         * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18813         * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18814         (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18815         of a machine_mode.
18816         (ft32_addr_space_address_mode): Likewise.
18817         * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18818         scalar_int_mode instead of a machine_mode.
18819         (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18820         of a machine_mode.
18821         (m32c_addr_space_address_mode): Likewise.
18822         * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18823         (rs6000_eh_return_filter_mode): Likewise.
18824         * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18825         (rs6000_eh_return_filter_mode): Likewise.
18826         * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18827         (s390_libgcc_shift_count_mode): Likewise.
18828         (s390_unwind_word_mode): Likewise.
18829         (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18830         machine_mode.
18831         * target.def (mode_rep_extended): Likewise.
18832         (valid_pointer_mode): Likewise.
18833         (addr_space.valid_pointer_mode): Likewise.
18834         (eh_return_filter_mode): Return a scalar_int_mode rather than
18835         a machine_mode.
18836         (libgcc_cmp_return_mode): Likewise.
18837         (libgcc_shift_count_mode): Likewise.
18838         (unwind_word_mode): Likewise.
18839         (addr_space.pointer_mode): Likewise.
18840         (addr_space.address_mode): Likewise.
18841         * doc/tm.texi: Regenerate.
18842         * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18843         a machine_mode.
18844         (do_jump): Use scalar_int_mode for local variables.
18845         * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18846         rather than a machine_mode.
18847         * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18848         (scompare_loc_descriptor_wide): Likewise.
18849         (scompare_loc_descriptor_narrow): Likewise.
18850         * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18851         variables.
18852         * except.c (sjlj_emit_dispatch_table): Likewise.
18853         (expand_builtin_eh_copy_values): Likewise.
18854         * explow.c (convert_memory_address_addr_space_1): Likewise.
18855         Take a scalar_int_mode rather than a machine_mode.
18856         (convert_memory_address_addr_space): Take a scalar_int_mode rather
18857         than a machine_mode.
18858         (memory_address_addr_space): Use scalar_int_mode for local variables.
18859         * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18860         rather than a machine_mode.
18861         * expmed.c (mask_rtx): Likewise.
18862         (init_expmed_one_conv): Likewise.
18863         (expand_mult_highpart_adjust): Likewise.
18864         (extract_high_half): Likewise.
18865         (expmed_mult_highpart_optab): Likewise.
18866         (expmed_mult_highpart): Likewise.
18867         (expand_smod_pow2): Likewise.
18868         (expand_sdiv_pow2): Likewise.
18869         (emit_store_flag_int): Likewise.
18870         (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18871         variables.
18872         (extract_low_bits): Likewise.
18873         * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18874         a machine_mode.
18875         * expr.c (pieces_addr::adjust):  Likewise.
18876         (can_store_by_pieces): Likewise.
18877         (store_by_pieces): Likewise.
18878         (clear_by_pieces_1): Likewise.
18879         (expand_expr_addr_expr_1): Likewise.
18880         (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18881         (expand_expr_real_1): Likewise.
18882         (try_casesi): Likewise.
18883         * final.c (shorten_branches): Likewise.
18884         * fold-const.c (fold_convert_const_int_from_fixed): Change the
18885         type of "mode" to machine_mode.
18886         * internal-fn.c (expand_arith_overflow_result_store): Take a
18887         scalar_int_mode rather than a machine_mode.
18888         (expand_mul_overflow): Use scalar_int_mode for local variables.
18889         * loop-doloop.c (doloop_modify): Likewise.
18890         (doloop_optimize): Likewise.
18891         * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18892         than a machine_mode.
18893         (expand_doubleword_shift_condmove): Likewise.
18894         (expand_doubleword_shift): Likewise.
18895         (expand_doubleword_clz): Likewise.
18896         (expand_doubleword_popcount): Likewise.
18897         (expand_doubleword_parity): Likewise.
18898         (expand_absneg_bit): Use scalar_int_mode for local variables.
18899         (prepare_float_lib_cmp): Likewise.
18900         * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
18901         rather than a machine_mode.
18902         (convert_memory_address_addr_space): Likewise.
18903         (get_mode_bounds): Likewise.
18904         (get_address_mode): Return a scalar_int_mode rather than a
18905         machine_mode.
18906         * rtlanal.c (get_address_mode): Likewise.
18907         * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
18908         than a machine_mode.
18909         * targhooks.c (default_mode_rep_extended): Likewise.
18910         (default_valid_pointer_mode): Likewise.
18911         (default_addr_space_valid_pointer_mode): Likewise.
18912         (default_eh_return_filter_mode): Return a scalar_int_mode rather
18913         than a machine_mode.
18914         (default_libgcc_cmp_return_mode): Likewise.
18915         (default_libgcc_shift_count_mode): Likewise.
18916         (default_unwind_word_mode): Likewise.
18917         (default_addr_space_pointer_mode): Likewise.
18918         (default_addr_space_address_mode): Likewise.
18919         * targhooks.h (default_eh_return_filter_mode): Likewise.
18920         (default_libgcc_cmp_return_mode): Likewise.
18921         (default_libgcc_shift_count_mode): Likewise.
18922         (default_unwind_word_mode): Likewise.
18923         (default_addr_space_pointer_mode): Likewise.
18924         (default_addr_space_address_mode): Likewise.
18925         (default_mode_rep_extended): Take a scalar_int_mode rather than
18926         a machine_mode.
18927         (default_valid_pointer_mode): Likewise.
18928         (default_addr_space_valid_pointer_mode): Likewise.
18929         * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
18930         local variables.
18931         * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
18932         rather than a machine_mode.
18933         * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
18934         for local variables.
18935         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
18936         * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
18937         than a machine_mode.
18939 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18940             Alan Hayward  <alan.hayward@arm.com>
18941             David Sherwood  <david.sherwood@arm.com>
18943         * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
18944         the mode argument to scalar_int_mode.
18945         (do_jump_by_parts_zero_rtx): Likewise.
18946         (do_jump_by_parts_equality_rtx): Likewise.
18947         (do_jump_by_parts_greater): Take a mode argument.
18948         (do_jump_by_parts_equality): Likewise.
18949         (do_jump_1): Update calls accordingly.
18951 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18952             Alan Hayward  <alan.hayward@arm.com>
18953             David Sherwood  <david.sherwood@arm.com>
18955         * is-a.h (safe_dyn_cast): New function.
18956         * rtl.h (rtx_jump_table_data::get_data_mode): New function.
18957         (jump_table_for_label): Likewise.
18958         * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
18959         instead of an rtx_insn *.
18960         (shorten_branches): Use dyn_cast instead of LABEL_P and
18961         JUMP_TABLE_DATA_P.  Use jump_table_for_label and
18962         rtx_jump_table_data::get_data_mode.
18963         (final_scan_insn): Likewise.
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         * combine.c (try_combine): Use is_a <scalar_int_mode> when
18970         trying to combine a full-register integer set with a subreg
18971         integer set.
18973 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18974             Alan Hayward  <alan.hayward@arm.com>
18975             David Sherwood  <david.sherwood@arm.com>
18977         * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
18978         that is always either address_mode or pointer_mode.
18980 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18981             Alan Hayward  <alan.hayward@arm.com>
18982             David Sherwood  <david.sherwood@arm.com>
18984         * expr.c (expand_expr_real_2): Use word_mode instead of innermode
18985         when the two are known to be equal.
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         * simplify-rtx.c (simplify_const_unary_operation): Use
18992         is_a <scalar_int_mode> instead of checking for a nonzero
18993         precision.  Forcibly convert op_mode to a scalar_int_mode
18994         in that case.  More clearly differentiate the operand and
18995         result modes and use the former when deciding what the value
18996         of a count-bits operation should be.  Use is_int_mode instead
18997         of checking for a MODE_INT.  Remove redundant check for whether
18998         this mode has a zero precision.
19000 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19001             Alan Hayward  <alan.hayward@arm.com>
19002             David Sherwood  <david.sherwood@arm.com>
19004         * optabs.c (widen_leading): Change the type of the mode argument
19005         to scalar_int_mode.  Use opt_scalar_int_mode for the mode iterator.
19006         (widen_bswap): Likewise.
19007         (expand_parity): Likewise.
19008         (expand_ctz): Change the type of the mode argument to scalar_int_mode.
19009         (expand_ffs): Likewise.
19010         (epand_unop): Check for scalar integer modes before calling the
19011         above routines.
19013 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19014             Alan Hayward  <alan.hayward@arm.com>
19015             David Sherwood  <david.sherwood@arm.com>
19017         * expr.c (const_scalar_mask_from_tree): Add a mode argument.
19018         Expand commentary.
19019         (expand_expr_real_1): Update call accordingly.
19021 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19022             Alan Hayward  <alan.hayward@arm.com>
19023             David Sherwood  <david.sherwood@arm.com>
19025         * expmed.c (store_bit_field_using_insv): Add op0_mode and
19026         value_mode arguments.  Use scalar_int_mode internally.
19027         (store_bit_field_1): Rename the new integer mode from imode
19028         to op0_mode and use it instead of GET_MODE (op0).  Update calls
19029         to store_split_bit_field, store_bit_field_using_insv and
19030         store_fixed_bit_field.
19031         (store_fixed_bit_field): Add op0_mode and value_mode arguments.
19032         Use scalar_int_mode internally.  Use a bit count rather than a mode
19033         when calculating the largest bit size for get_best_mode.
19034         Update calls to store_split_bit_field and store_fixed_bit_field_1.
19035         (store_fixed_bit_field_1): Add mode and value_mode arguments.
19036         Remove assertion that OP0 has a scalar integer mode.
19037         (store_split_bit_field): Add op0_mode and value_mode arguments.
19038         Update calls to extract_fixed_bit_field.
19039         (extract_bit_field_using_extv): Add an op0_mode argument.
19040         Use scalar_int_mode internally.
19041         (extract_bit_field_1): Rename the new integer mode from imode to
19042         op0_mode and use it instead of GET_MODE (op0).  Update calls to
19043         extract_split_bit_field, extract_bit_field_using_extv and
19044         extract_fixed_bit_field.
19045         (extract_fixed_bit_field): Add an op0_mode argument.  Update calls
19046         to extract_split_bit_field and extract_fixed_bit_field_1.
19047         (extract_fixed_bit_field_1): Add a mode argument.  Remove assertion
19048         that OP0 has a scalar integer mode.  Use as_a <scalar_int_mode>
19049         on the target mode.
19050         (extract_split_bit_field): Add an op0_mode argument.  Update call
19051         to extract_fixed_bit_field.
19053 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19054             Alan Hayward  <alan.hayward@arm.com>
19055             David Sherwood  <david.sherwood@arm.com>
19057         * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
19058         * explow.c (hard_function_value): Likewise.
19059         * expmed.c (extract_fixed_bit_field_1): Likewise.  Move the
19060         convert_to_mode call outside the loop.
19061         * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
19062         for the mode iterator.  Require the mode specified by max_pieces
19063         to exist.
19064         (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
19065         mode iterator.
19066         (copy_blkmode_to_reg): Likewise.
19067         (set_storage_via_setmem): Likewise.
19068         * optabs.c (prepare_cmp_insn): Likewise.
19069         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19070         * stor-layout.c (finish_bitfield_representative): Likewise.
19072 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19073             Alan Hayward  <alan.hayward@arm.com>
19074             David Sherwood  <david.sherwood@arm.com>
19076         * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
19077         * expr.c (convert_move): Use them.
19078         (convert_modes): Likewise.
19079         (store_expr_with_bounds): Likewise.
19081 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19082             Alan Hayward  <alan.hayward@arm.com>
19083             David Sherwood  <david.sherwood@arm.com>
19085         * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
19086         parameter for the mode of "x".  Remove the "known_x", "known_mode"
19087         and "known_ret" arguments.  Change the type of the mode argument
19088         to scalar_int_mode.
19089         (rtl_hooks:reg_num_sign_bit_copies): Likewise.
19090         * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
19091         (reg_num_sign_bit_copies_for_combine): Likewise.
19092         * rtlanal.c (nonzero_bits1): Likewise.
19093         (num_sign_bit_copies1): Likewise.
19094         * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
19095         (reg_num_sign_bit_copies_general): Likewise.
19096         * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
19097         (reg_nonzero_bits_general): Likewise.
19099 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19100             Alan Hayward  <alan.hayward@arm.com>
19101             David Sherwood  <david.sherwood@arm.com>
19103         * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
19104         than in subroutines.  Return 1 for non-integer modes.
19105         (cached_num_sign_bit_copies): Change the type of the mode parameter
19106         to scalar_int_mode.
19107         (num_sign_bit_copies1): Likewise.  Remove early exit for other mode
19108         classes.  Handle CONST_INT_P first and then check whether X also
19109         has a scalar integer mode.  Check the same thing for inner registers
19110         of a SUBREG and for values that are being extended or truncated.
19112 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19113             Alan Hayward  <alan.hayward@arm.com>
19114             David Sherwood  <david.sherwood@arm.com>
19116         * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
19117         in subroutines.  Return the mode mask for non-integer modes.
19118         (cached_nonzero_bits): Change the type of the mode parameter
19119         to scalar_int_mode.
19120         (nonzero_bits1): Likewise.  Remove early exit for other mode
19121         classes.  Handle CONST_INT_P first and then check whether X
19122         also has a scalar integer mode.
19124 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19125             Alan Hayward  <alan.hayward@arm.com>
19126             David Sherwood  <david.sherwood@arm.com>
19128         * expr.c (widest_int_mode_for_size): Make the comment match the code.
19129         Return a scalar_int_mode and assert that the size is greater than
19130         one byte.
19131         (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
19132         (op_by_pieces_d::op_by_pieces_d): Likewise.
19133         (op_by_pieces_d::run): Likewise.
19134         (can_store_by_pieces): Likewise.
19136 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19137             Alan Hayward  <alan.hayward@arm.com>
19138             David Sherwood  <david.sherwood@arm.com>
19140         * combine.c (extract_left_shift): Add a mode argument and update
19141         recursive calls.
19142         (make_compound_operation_int): Change the type of the mode parameter
19143         to scalar_int_mode and update the call to extract_left_shift.
19145 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19146             Alan Hayward  <alan.hayward@arm.com>
19147             David Sherwood  <david.sherwood@arm.com>
19149         * combine.c (simplify_and_const_int): Change the type of the mode
19150         parameter to scalar_int_mode.
19151         (simplify_and_const_int_1): Likewise.  Update recursive call.
19153 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19154             Alan Hayward  <alan.hayward@arm.com>
19155             David Sherwood  <david.sherwood@arm.com>
19157         * combine.c (simplify_compare_const): Check that the mode is a
19158         scalar_int_mode (rather than VOIDmode) before testing its
19159         precision.
19160         (simplify_comparison): Move COMPARISON_P handling out of the
19161         loop and restrict the latter part of the loop to scalar_int_modes.
19162         Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
19163         and when considering SUBREG_REGs.  Use is_int_mode instead of
19164         checking GET_MODE_CLASS against MODE_INT.
19166 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19167             Alan Hayward  <alan.hayward@arm.com>
19168             David Sherwood  <david.sherwood@arm.com>
19170         * combine.c (try_widen_shift_mode): Move check for equal modes to...
19171         (simplify_shift_const_1): ...here.  Use scalar_int_mode for
19172         shift_unit_mode and for modes involved in scalar shifts.
19174 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19175             Alan Hayward  <alan.hayward@arm.com>
19176             David Sherwood  <david.sherwood@arm.com>
19178         * combine.c (force_int_to_mode): New function, split out from...
19179         (force_to_mode): ...here.  Keep xmode up-to-date and use it
19180         instead of GET_MODE (x).
19182 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19183             Alan Hayward  <alan.hayward@arm.com>
19184             David Sherwood  <david.sherwood@arm.com>
19186         * optabs-query.h (extraction_insn::struct_mode): Change type to
19187         opt_scalar_int_mode and update comment.
19188         (extraction_insn::field_mode): Change type to scalar_int_mode.
19189         (extraction_insn::pos_mode): Likewise.
19190         * combine.c (make_extraction): Update accordingly.
19191         * optabs-query.c (get_traditional_extraction_insn): Likewise.
19192         (get_optab_extraction_insn): Likewise.
19193         * recog.c (simplify_while_replacing): Likewise.
19194         * expmed.c (narrow_bit_field_mem): Change the type of the mode
19195         parameter to opt_scalar_int_mode.
19197 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19198             Alan Hayward  <alan.hayward@arm.com>
19199             David Sherwood  <david.sherwood@arm.com>
19201         * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
19202         to a scalar_int_mode instead of a machine_mode.
19203         (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
19204         (get_best_mode): Return a boolean and use a pointer argument to store
19205         the selected mode.  Replace the limit mode parameter with a bit limit.
19206         * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
19207         for the values returned by bit_field_mode_iterator::next_mode.
19208         (store_bit_field): Update call to get_best_mode.
19209         (store_fixed_bit_field): Likewise.
19210         (extract_fixed_bit_field): Likewise.
19211         * expr.c (optimize_bitfield_assignment_op): Likewise.
19212         * fold-const.c (optimize_bit_field_compare): Likewise.
19213         (fold_truth_andor_1): Likewise.
19214         * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
19215         Update for new type of m_mode.
19216         (get_best_mode): As above.
19218 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19219             Alan Hayward  <alan.hayward@arm.com>
19220             David Sherwood  <david.sherwood@arm.com>
19222         * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
19223         to scalar_int_mode.  Remove check for SCALAR_INT_MODE_P.
19224         (store_bit_field): Check is_a <scalar_int_mode> before calling
19225         strict_volatile_bitfield_p.
19226         (extract_bit_field): Likewise.
19228 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19229             Alan Hayward  <alan.hayward@arm.com>
19230             David Sherwood  <david.sherwood@arm.com>
19232         * target.def (cstore_mode): Return a scalar_int_mode.
19233         * doc/tm.texi: Regenerate.
19234         * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19235         * targhooks.h (default_cstore_mode): Likewise.
19236         * targhooks.c (default_cstore_mode): Likewise, using a forced
19237         conversion.
19238         * expmed.c (emit_cstore): Expect the target of the cstore to be
19239         a scalar_int_mode.
19241 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19242             Alan Hayward  <alan.hayward@arm.com>
19243             David Sherwood  <david.sherwood@arm.com>
19245         * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19246         scalar_int_mode.
19247         (niter_desc): Likewise mode.
19248         (iv_analyze): Add a mode parameter.
19249         (biv_p): Likewise.
19250         (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19251         and change its type to scalar_int_mode.
19252         * loop-iv.c: Update commentary at head of file.
19253         (iv_constant): Pass the mode paraeter before the rtx it describes
19254         and change its type to scalar_int_mode.  Remove VOIDmode handling.
19255         (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19256         (iv_extend): Likewise.
19257         (shorten_into_mode): Likewise.
19258         (iv_add): Use scalar_int_mode.
19259         (iv_mult): Likewise.
19260         (iv_shift): Likewise.
19261         (canonicalize_iv_subregs): Likewise.
19262         (get_biv_step_1): Pass the outer_mode parameter before the rtx
19263         it describes and change its mode to scalar_int_mode.   Also change
19264         the type of the returned inner_mode to scalar_int_mode.
19265         (get_biv_step): Likewise, turning outer_mode from a pointer
19266         into a direct parameter.  Update call to get_biv_step_1.
19267         (iv_analyze_biv): Add an outer_mode parameter.  Update calls to
19268         iv_constant and get_biv_step.
19269         (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19270         and change its type to scalar_int_mode.  Don't initialise iv->mode
19271         to VOIDmode and remove later checks for its still being VOIDmode.
19272         Update calls to iv_analyze_op and iv_analyze_expr.  Check
19273         is_a <scalar_int_mode> when changing the mode under consideration.
19274         (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19275         Update call to iv_analyze_expr.
19276         (iv_analyze_op): Add a mode parameter.  Reject subregs whose
19277         inner register is not also a scalar_int_mode.  Update call to
19278         iv_analyze_biv.
19279         (iv_analyze): Add a mode parameter.  Update call to iv_analyze_op.
19280         (biv_p): Add a mode parameter.  Update call to iv_analyze_biv.
19281         (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19282         separate mode class checks.  Update calls to iv_analyze.  Remove
19283         fix-up of VOIDmodes after iv_analyze_biv.
19284         * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19285         don't have a scalar_int_mode.  Update call to biv_p.
19287 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19288             Alan Hayward  <alan.hayward@arm.com>
19289             David Sherwood  <david.sherwood@arm.com>
19291         * cfgexpand.c (convert_debug_memory_address): Use
19292         as_a <scalar_int_mode>.
19293         * combine.c (expand_compound_operation): Likewise.
19294         (make_extraction): Likewise.
19295         (change_zero_ext): Likewise.
19296         (simplify_comparison): Likewise.
19297         * cse.c (cse_insn): Likewise.
19298         * dwarf2out.c (minmax_loc_descriptor): Likewise.
19299         (mem_loc_descriptor): Likewise.
19300         (loc_descriptor): Likewise.
19301         * expmed.c (init_expmed_one_mode): Likewise.
19302         (synth_mult): Likewise.
19303         (emit_store_flag_1): Likewise.
19304         (expand_divmod): Likewise.  Use HWI_COMPUTABLE_MODE_P instead
19305         of a comparison with size.
19306         * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19307         (reduce_to_bit_field_precision): Likewise.
19308         * function.c (expand_function_end): Likewise.
19309         * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19310         * loop-doloop.c (doloop_modify): Likewise.
19311         * optabs.c (expand_binop): Likewise.
19312         (expand_unop): Likewise.
19313         (expand_copysign_absneg): Likewise.
19314         (prepare_cmp_insn): Likewise.
19315         (maybe_legitimize_operand): Likewise.
19316         * recog.c (const_scalar_int_operand): Likewise.
19317         * rtlanal.c (get_address_mode): Likewise.
19318         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19319         (simplify_cond_clz_ctz): Likewise.
19320         * tree-nested.c (get_nl_goto_field): Likewise.
19321         * tree.c (build_vector_type_for_mode): Likewise.
19322         * var-tracking.c (use_narrower_mode): Likewise.
19324 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19325             Alan Hayward  <alan.hayward@arm.com>
19326             David Sherwood  <david.sherwood@arm.com>
19328         * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19329         * builtins.c (expand_builtin_signbit): Use it.
19330         * cfgexpand.c (expand_debug_expr): Likewise.
19331         * dojump.c (do_jump): Likewise.
19332         (do_compare_and_jump): Likewise.
19333         * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19334         * expmed.c (make_tree): Likewise.
19335         * expr.c (expand_expr_real_2): Likewise.
19336         (expand_expr_real_1): Likewise.
19337         (try_casesi): Likewise.
19338         * fold-const-call.c (fold_const_call_ss): Likewise.
19339         * fold-const.c (unextend): Likewise.
19340         (extract_muldiv_1): Likewise.
19341         (fold_single_bit_test): Likewise.
19342         (native_encode_int): Likewise.
19343         (native_encode_string): Likewise.
19344         (native_interpret_int): Likewise.
19345         * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19346         * internal-fn.c (expand_addsub_overflow): Likewise.
19347         (expand_neg_overflow): Likewise.
19348         (expand_mul_overflow): Likewise.
19349         (expand_arith_overflow): Likewise.
19350         * match.pd: Likewise.
19351         * stor-layout.c (layout_type): Likewise.
19352         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19353         * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19354         * tree-ssanames.c (get_range_info): Likewise.
19355         * tree-switch-conversion.c (array_value_type) Likewise.
19356         * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19357         (vect_recog_divmod_pattern): Likewise.
19358         (vect_recog_mixed_size_cond_pattern): Likewise.
19359         * tree-vrp.c (extract_range_basic): Likewise.
19360         (simplify_float_conversion_using_ranges): Likewise.
19361         * tree.c (int_fits_type_p): Likewise.
19362         * ubsan.c (instrument_bool_enum_load): Likewise.
19363         * varasm.c (mergeable_string_section): Likewise.
19364         (narrowing_initializer_constant_valid_p): Likewise.
19365         (output_constant): Likewise.
19367 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19368             Alan Hayward  <alan.hayward@arm.com>
19369             David Sherwood  <david.sherwood@arm.com>
19371         * machmode.h (NARROWEST_INT_MODE): New macro.
19372         * expr.c (alignment_for_piecewise_move): Use it instead of
19373         GET_CLASS_NARROWEST_MODE (MODE_INT).
19374         (push_block): Likewise.
19375         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19376         Likewise.
19377         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19379 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19380             Alan Hayward  <alan.hayward@arm.com>
19381             David Sherwood  <david.sherwood@arm.com>
19383         * postreload.c (move2add_valid_value_p): Change the type of the
19384         mode parameter to scalar_int_mode.
19385         (move2add_use_add2_insn): Add a mode parameter and use it instead
19386         of GET_MODE (reg).
19387         (move2add_use_add3_insn): Likewise.
19388         (reload_cse_move2add): Update accordingly.
19390 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19391             Alan Hayward  <alan.hayward@arm.com>
19392             David Sherwood  <david.sherwood@arm.com>
19394         * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19395         double-word mode.
19396         * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19397         * optabs.c (expand_unop): Likewise.
19399 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19400             Alan Hayward  <alan.hayward@arm.com>
19401             David Sherwood  <david.sherwood@arm.com>
19403         * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19404         (clz_loc_descriptor): Likewise.  Remove SCALAR_INT_MODE_P check.
19405         (popcount_loc_descriptor): Likewise.
19406         (bswap_loc_descriptor): Likewise.
19407         (rotate_loc_descriptor): Likewise.
19408         (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19409         calling the functions above.
19411 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19412             Alan Hayward  <alan.hayward@arm.com>
19413             David Sherwood  <david.sherwood@arm.com>
19415         * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19416         checks.
19417         (try_combine): Likewise.
19418         (simplify_if_then_else): Likewise.
19419         * cse.c (cse_insn): Likewise.
19420         * dwarf2out.c (mem_loc_descriptor): Likewise.
19421         * emit-rtl.c (gen_lowpart_common): Likewise.
19422         * simplify-rtx.c (simplify_truncation): Likewise.
19423         (simplify_binary_operation_1): Likewise.
19424         (simplify_const_relational_operation): Likewise.
19425         (simplify_ternary_operation): Likewise.
19426         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
19428 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19429             Alan Hayward  <alan.hayward@arm.com>
19430             David Sherwood  <david.sherwood@arm.com>
19432         * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19433         * reload.c (push_reload): Likewise.
19434         (find_reloads): Likewise.
19436 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19437             Alan Hayward  <alan.hayward@arm.com>
19438             David Sherwood  <david.sherwood@arm.com>
19440         * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19441         (make_compound_operation_int): Likewise.
19442         (change_zero_ext): Likewise.
19443         * expr.c (convert_move): Likewise.
19444         (convert_modes): Likewise.
19445         * fwprop.c (forward_propagate_subreg): Likewise.
19446         * loop-iv.c (get_biv_step_1): Likewise.
19447         * optabs.c (widen_operand): Likewise.
19448         * postreload.c (move2add_valid_value_p): Likewise.
19449         * recog.c (simplify_while_replacing): Likewise.
19450         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19451         (simplify_binary_operation_1): Likewise.  Remove redundant
19452         mode equality check.
19454 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19455             Alan Hayward  <alan.hayward@arm.com>
19456             David Sherwood  <david.sherwood@arm.com>
19458         * combine.c (combine_simplify_rtx): Add checks for
19459         is_a <scalar_int_mode>.
19460         (simplify_if_then_else): Likewise.
19461         (make_field_assignment): Likewise.
19462         (simplify_comparison): Likewise.
19463         * ifcvt.c (noce_try_bitop): Likewise.
19464         * loop-invariant.c (canonicalize_address_mult): Likewise.
19465         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19467 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19468             Alan Hayward  <alan.hayward@arm.com>
19469             David Sherwood  <david.sherwood@arm.com>
19471         * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19472         is_a <scalar_int_mode> instead of != BLKmode.
19474 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19475             Alan Hayward  <alan.hayward@arm.com>
19476             David Sherwood  <david.sherwood@arm.com>
19478         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19479         instead of != VOIDmode.
19480         * combine.c (if_then_else_cond): Likewise.
19481         (change_zero_ext): Likewise.
19482         * dwarf2out.c (mem_loc_descriptor): Likewise.
19483         (loc_descriptor): Likewise.
19484         * rtlanal.c (canonicalize_condition): Likewise.
19485         * simplify-rtx.c (simplify_relational_operation_1): Likewise.
19487 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19488             Alan Hayward  <alan.hayward@arm.com>
19489             David Sherwood  <david.sherwood@arm.com>
19491         * simplify-rtx.c (simplify_binary_operation_1): Use
19492         is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
19494 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19495             Alan Hayward  <alan.hayward@arm.com>
19496             David Sherwood  <david.sherwood@arm.com>
19498         * wide-int.h (int_traits<unsigned char>) New class.
19499         (int_traits<unsigned short>) Likewise.
19500         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19501         Use GET_MODE_UNIT_PRECISION and remove redundant test for
19502         SCALAR_INT_MODE_P.
19503         * combine.c (set_nonzero_bits_and_sign_copies): Use
19504         is_a <scalar_int_mode>.
19505         (find_split_point): Likewise.
19506         (combine_simplify_rtx): Likewise.
19507         (simplify_logical): Likewise.
19508         (expand_compound_operation): Likewise.
19509         (expand_field_assignment): Likewise.
19510         (make_compound_operation): Likewise.
19511         (extended_count): Likewise.
19512         (change_zero_ext): Likewise.
19513         (simplify_comparison): Likewise.
19514         * dwarf2out.c (scompare_loc_descriptor): Likewise.
19515         (ucompare_loc_descriptor): Likewise.
19516         (minmax_loc_descriptor): Likewise.
19517         (mem_loc_descriptor): Likewise.
19518         (loc_descriptor): Likewise.
19519         * expmed.c (init_expmed_one_mode): Likewise.
19520         * lra-constraints.c (lra_constraint_offset): Likewise.
19521         * optabs.c (prepare_libcall_arg): Likewise.
19522         * postreload.c (move2add_note_store): Likewise.
19523         * reload.c (operands_match_p): Likewise.
19524         * rtl.h (load_extend_op): Likewise.
19525         * rtlhooks.c (gen_lowpart_general): Likewise.
19526         * simplify-rtx.c (simplify_truncation): Likewise.
19527         (simplify_unary_operation_1): Likewise.
19528         (simplify_binary_operation_1): Likewise.
19529         (simplify_const_binary_operation): Likewise.
19530         (simplify_const_relational_operation): Likewise.
19531         (simplify_subreg): Likewise.
19532         * stor-layout.c (bitwise_mode_for_mode): Likewise.
19533         * var-tracking.c (adjust_mems): Likewise.
19534         (prepare_call_arguments): Likewise.
19536 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19537             Alan Hayward  <alan.hayward@arm.com>
19538             David Sherwood  <david.sherwood@arm.com>
19540         * machmode.h (is_int_mode): New fuction.
19541         * combine.c (find_split_point): Use it.
19542         (combine_simplify_rtx): Likewise.
19543         (simplify_if_then_else): Likewise.
19544         (simplify_set): Likewise.
19545         (simplify_shift_const_1): Likewise.
19546         (simplify_comparison): Likewise.
19547         * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19548         * cse.c (notreg_cost): Likewise.
19549         (cse_insn): Likewise.
19550         * cselib.c (cselib_lookup_1): Likewise.
19551         * dojump.c (do_jump_1): Likewise.
19552         (do_compare_rtx_and_jump): Likewise.
19553         * dse.c (get_call_args): Likewise.
19554         * dwarf2out.c (rtl_for_decl_init): Likewise.
19555         (native_encode_initializer): Likewise.
19556         * expmed.c (emit_store_flag_1): Likewise.
19557         (emit_store_flag): Likewise.
19558         * expr.c (convert_modes): Likewise.
19559         (store_field): Likewise.
19560         (expand_expr_real_1): Likewise.
19561         * fold-const.c (fold_read_from_constant_string): Likewise.
19562         * gimple-ssa-sprintf.c (get_format_string): Likewise.
19563         * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19564         * optabs.c (expand_binop): Likewise.
19565         (expand_unop): Likewise.
19566         (expand_abs_nojump): Likewise.
19567         (expand_one_cmpl_abs_nojump): Likewise.
19568         * simplify-rtx.c (mode_signbit_p): Likewise.
19569         (val_signbit_p): Likewise.
19570         (val_signbit_known_set_p): Likewise.
19571         (val_signbit_known_clear_p): Likewise.
19572         (simplify_relational_operation_1): Likewise.
19573         * tree.c (vector_type_mode): Likewise.
19575 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19576             Alan Hayward  <alan.hayward@arm.com>
19577             David Sherwood  <david.sherwood@arm.com>
19579         * machmode.h (smallest_mode_for_size): Fix formatting.
19580         (smallest_int_mode_for_size): New function.
19581         * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19582         instead of smallest_mode_for_size.
19583         * combine.c (make_extraction): Likewise.
19584         * config/arc/arc.c (arc_expand_movmem): Likewise.
19585         * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19586         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19587         * config/s390/s390.c (s390_expand_insv): Likewise.
19588         * config/sparc/sparc.c (assign_int_registers): Likewise.
19589         * config/spu/spu.c (spu_function_value): Likewise.
19590         (spu_function_arg): Likewise.
19591         * coverage.c (get_gcov_type): Likewise.
19592         (get_gcov_unsigned_t): Likewise.
19593         * dse.c (find_shift_sequence): Likewise.
19594         * expmed.c (store_bit_field_1): Likewise.
19595         * expr.c (convert_move): Likewise.
19596         (store_field): Likewise.
19597         * internal-fn.c (expand_arith_overflow): Likewise.
19598         * optabs-query.c (get_best_extraction_insn): Likewise.
19599         * optabs.c (expand_twoval_binop_libfunc): Likewise.
19600         * stor-layout.c (layout_type): Likewise.
19601         (initialize_sizetypes): Likewise.
19602         * targhooks.c (default_get_mask_mode): Likewise.
19603         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19605 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19606             Alan Hayward  <alan.hayward@arm.com>
19607             David Sherwood  <david.sherwood@arm.com>
19609         * machmode.h (opt_mode::else_blk): New function.
19610         (int_mode_for_mode): Declare.
19611         * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19612         * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19613         return type.
19614         * cfgexpand.c (expand_debug_expr): Likewise.
19615         * combine.c (gen_lowpart_or_truncate): Likewise.
19616         (gen_lowpart_for_combine): Likewise.
19617         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19618         * config/avr/avr.c (avr_to_int_mode): Likewise.
19619         (avr_out_plus_1): Likewise.
19620         (avr_out_plus): Likewise.
19621         (avr_out_round): Likewise.
19622         * config/i386/i386.c (ix86_split_to_parts): Likewise.
19623         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19624         (s390_expand_vcond): Likewise.
19625         * config/spu/spu.c (spu_split_immediate): Likewise.
19626         (spu_expand_mov): Likewise.
19627         * dse.c (get_stored_val): Likewise.
19628         * expmed.c (store_bit_field_1): Likewise.
19629         (convert_extracted_bit_field): Use int_mode_for_mode instead of
19630         int_mode_for_size.
19631         (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19632         (extract_low_bits): Likewise.
19633         * expr.c (emit_group_load_1): Likewise.  Separate out the BLKmode
19634         handling rather than repeating the check.
19635         (emit_group_store): Likewise.
19636         (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19637         * optabs.c (expand_absneg_bit): Likewise.
19638         (expand_copysign_absneg): Likewise.
19639         (expand_copysign_bit): Likewise.
19640         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19641         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19642         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19643         * var-tracking.c (prepare_call_arguments):  Likewise.
19644         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19645         int_mode_for_mode instead of mode_for_size.
19646         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19648 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19649             Alan Hayward  <alan.hayward@arm.com>
19650             David Sherwood  <david.sherwood@arm.com>
19652         * machmode.h (int_mode_for_size): New function.
19653         * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19654         instead of mode_for_size.
19655         * calls.c (save_fixed_argument_area): Likewise.  Make use of BLKmode
19656         explicit.
19657         * combine.c (expand_field_assignment): Use int_mode_for_size
19658         instead of mode_for_size.
19659         (make_extraction): Likewise.
19660         (simplify_shift_const_1): Likewise.
19661         (simplify_comparison): Likewise.
19662         * dojump.c (do_jump): Likewise.
19663         * dwarf2out.c (mem_loc_descriptor): Likewise.
19664         * emit-rtl.c (init_derived_machine_modes): Likewise.
19665         * expmed.c (flip_storage_order): Likewise.
19666         (convert_extracted_bit_field): Likewise.
19667         * expr.c (copy_blkmode_from_reg): Likewise.
19668         * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19669         * internal-fn.c (expand_mul_overflow): Likewise.
19670         * lower-subreg.c (simple_move): Likewise.
19671         * optabs-libfuncs.c (init_optabs): Likewise.
19672         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19673         * tree.c (vector_type_mode): Likewise.
19674         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19675         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19676         * tree-vect-generic.c (expand_vector_parallel): Likewise.
19677         * tree-vect-stmts.c (vectorizable_load): Likewise.
19678         (vectorizable_store): Likewise.
19680 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19681             Alan Hayward  <alan.hayward@arm.com>
19682             David Sherwood  <david.sherwood@arm.com>
19684         * coretypes.h (pod_mode): New type.
19685         (scalar_int_mode_pod): New typedef.
19686         * machmode.h (pod_mode): New class.
19687         (int_n_data_t::m): Change type to scalar_int_mode_pod.
19688         * genmodes.c (emit_mode_int_n): Update accordingly.
19689         * lower-subreg.h (target_lower_subreg): Change type to
19690         scalar_int_mode_pod.
19691         * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19692         scalar_int_mode_pod.
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         * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19699         machine_mode to scalar_int_mode.
19700         * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19701         (rs6000_option_override_internal): Remove cast to int.
19702         * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19703         machine_mode to scalar_int_mode.
19704         * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19705         (rs6000_option_override_internal): Remove cast to int.
19706         * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19707         * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19708         to machine_mode.
19709         * config/s390/s390.c (s390_expand_builtin): Likewise.
19710         * coretypes.h (scalar_int_mode): New type.
19711         (opt_scalar_int_mode): New typedef.
19712         * machmode.h (scalar_int_mode): New class.
19713         (scalar_int_mode::includes_p): New function.
19714         (byte_mode): Change type to scalar_int_mode.
19715         (word_mode): Likewise.
19716         (ptr_mode): Likewise.
19717         * emit-rtl.c (byte_mode): Likewise.
19718         (word_mode): Likewise.
19719         (ptr_mode): Likewise.
19720         (init_derived_machine_modes): Update accordingly.
19721         * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19722         and MODE_PARTIAL_INT.
19723         * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19724         opt_scalar_int_mode.
19726 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19727             Alan Hayward  <alan.hayward@arm.com>
19728             David Sherwood  <david.sherwood@arm.com>
19730         * target.def (libgcc_floating_mode_supported_p): Take a
19731         scalar_float_mode.
19732         * doc/tm.texi: Regenerate.
19733         * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19734         scalar_float_mode.
19735         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19736         * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19737         Likewise.
19739 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19740             Alan Hayward  <alan.hayward@arm.com>
19741             David Sherwood  <david.sherwood@arm.com>
19743         * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19744         * doc/tm.texi: Regenerate.
19745         * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19746         * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19747         * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19748         * targhooks.h (default_floatn_mode): Likewise.
19749         * targhooks.c (default_floatn_mode): Likewise.
19750         * tree.c (build_common_tree_nodes): Update accordingly.
19752 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19753             Alan Hayward  <alan.hayward@arm.com>
19754             David Sherwood  <david.sherwood@arm.com>
19756         * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19757         (mode_iterator::iterate_p): Likewise.
19758         (mode_iterator::get_wider): Likewise.
19759         * expr.c (init_expr_target): Use opt_scalar_float_mode.
19761 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19762             Alan Hayward  <alan.hayward@arm.com>
19763             David Sherwood  <david.sherwood@arm.com>
19765         * coretypes.h (opt_scalar_float_mode): New typedef.
19766         * machmode.h (float_mode_for_size): New function.
19767         * emit-rtl.c (double_mode): Delete.
19768         (init_emit_once): Use float_mode_for_size.
19769         * stor-layout.c (layout_type): Likewise.
19770         * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
19772 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19773             Alan Hayward  <alan.hayward@arm.com>
19774             David Sherwood  <david.sherwood@arm.com>
19776         * output.h (assemble_real): Take a scalar_float_mode.
19777         * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19778         * config/arm/arm.md (consttable_4): Likewise.
19779         (consttable_8): Likewise.
19780         (consttable_16): Likewise.
19781         * config/mips/mips.md (consttable_float): Likewise.
19782         * config/s390/s390.c (s390_output_pool_entry): Likewise.
19783         * varasm.c (assemble_real): Take a scalar_float_mode.
19784         (output_constant_pool_2): Update accordingly.
19785         (output_constant): Likewise.
19787 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19788             Alan Hayward  <alan.hayward@arm.com>
19789             David Sherwood  <david.sherwood@arm.com>
19791         * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19792         * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19793         * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19794         (native_encode_real): Likewise.
19795         (native_interpret_real): Likewise.
19796         * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19797         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19799 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19800             Alan Hayward  <alan.hayward@arm.com>
19801             David Sherwood  <david.sherwood@arm.com>
19803         * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19804         <scalar_float_mode>.  Simplify.
19805         (gen_extend_conv_libfunc): Likewise.
19807 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19808             Alan Hayward  <alan.hayward@arm.com>
19809             David Sherwood  <david.sherwood@arm.com>
19811         * coretypes.h (scalar_float_mode): New type.
19812         * machmode.h (mode_traits::from_int): Use machine_mode if
19813         USE_ENUM_MODES is defined.
19814         (is_a): New function.
19815         (as_a): Likewise.
19816         (dyn_cast): Likewise.
19817         (scalar_float_mode): New class.
19818         (scalar_float_mode::includes_p): New function.
19819         (is_float_mode): Likewise.
19820         * gdbhooks.py (MachineModePrinter): New class.
19821         (build_pretty_printer): Use it for scalar_float_mode.
19822         * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19823         (format_helper::format_helper): Turn into a template.
19824         * genmodes.c (get_mode_class): New function.
19825         (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19826         or machine_mode if none.
19827         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19828         as_a <scalar_float_mode>.
19829         * dwarf2out.c (mem_loc_descriptor): Likewise.
19830         (insert_float): Likewise.
19831         (add_const_value_attribute): Likewise.
19832         * simplify-rtx.c (simplify_immed_subreg): Likewise.
19833         * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19834         (expand_unop): Update accordingly.
19835         (expand_abs_nojump): Likewise.
19836         (expand_copysign_absneg): Take a scalar_float_mode.
19837         (expand_copysign_bit): Likewise.
19838         (expand_copysign): Update accordingly.
19840 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19841             Alan Hayward  <alan.hayward@arm.com>
19842             David Sherwood  <david.sherwood@arm.com>
19844         * coretypes.h (opt_mode): New class.
19845         * machmode.h (opt_mode): Likewise.
19846         (opt_mode::else_void): New function.
19847         (opt_mode::require): Likewise.
19848         (opt_mode::exists): Likewise.
19849         (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19850         (GET_MODE_2XWIDER_MODE): Likewise.
19851         (mode_iterator::get_wider): Update accordingly.
19852         (mode_iterator::get_2xwider): Likewise.
19853         (mode_iterator::get_known_wider): Likewise, turning into a template.
19854         * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19855         forcing a wider mode to exist.
19856         * config/cr16/cr16.h (LONG_REG_P): Likewise.
19857         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19858         * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19859         GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19860         * lower-subreg.c (init_lower_subreg): Likewise.
19861         * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19862         on the final iteration.
19863         * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19864         a wider mode exists before asking for a move pattern.
19865         (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19866         forcing a wider mode to exist.
19867         (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19868         returning false if no such mode exists.
19869         * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19870         * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19871         * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19872         Avoid checking for a MODE_INT if we already know the mode is not a
19873         SCALAR_INT_MODE_P.
19874         (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19875         forcing a wider mode to exist.
19876         (expmed_mult_highpart_optab): Likewise.
19877         (expmed_mult_highpart): Likewise.
19878         * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19879         using else_void.
19880         * lto-streamer-in.c (lto_input_mode_table): Likewise.
19881         * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19882         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19883         * internal-fn.c (expand_mul_overflow): Update use of
19884         GET_MODE_2XWIDER_MODE.
19885         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19886         * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19887         GET_MODE_WIDER_MODE.
19888         (convert_plusminus_to_widen): Likewise.
19889         * tree-switch-conversion.c (array_value_type): Likewise.
19890         * var-tracking.c (emit_note_insn_var_location): Likewise.
19891         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19892         Return false inside rather than outside the loop if no wider mode
19893         exists
19894         * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19895         and GET_MODE_2XWIDER_MODE
19896         (can_compare_p): Use else_void.
19897         * gdbhooks.py (OptMachineModePrinter): New class.
19898         (build_pretty_printer): Use it for opt_mode.
19900 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19901             Alan Hayward  <alan.hayward@arm.com>
19902             David Sherwood  <david.sherwood@arm.com>
19904         * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
19905         once.  Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
19907 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19908             Alan Hayward  <alan.hayward@arm.com>
19909             David Sherwood  <david.sherwood@arm.com>
19911         * machmode.h (mode_traits): New structure.
19912         (get_narrowest_mode): New function.
19913         (mode_iterator::start): Likewise.
19914         (mode_iterator::iterate_p): Likewise.
19915         (mode_iterator::get_wider): Likewise.
19916         (mode_iterator::get_known_wider): Likewise.
19917         (mode_iterator::get_2xwider): Likewise.
19918         (FOR_EACH_MODE_IN_CLASS): New mode iterator.
19919         (FOR_EACH_MODE): Likewise.
19920         (FOR_EACH_MODE_FROM): Likewise.
19921         (FOR_EACH_MODE_UNTIL): Likewise.
19922         (FOR_EACH_WIDER_MODE): Likewise.
19923         (FOR_EACH_2XWIDER_MODE): Likewise.
19924         * builtins.c (expand_builtin_strlen): Use new mode iterators.
19925         * combine.c (simplify_comparison): Likewise
19926         * config/i386/i386.c (type_natural_mode): Likewise.
19927         * cse.c (cse_insn): Likewise.
19928         * dse.c (find_shift_sequence): Likewise.
19929         * emit-rtl.c (init_derived_machine_modes): Likewise.
19930         (init_emit_once): Likewise.
19931         * explow.c (hard_function_value): Likewise.
19932         * expmed.c (extract_fixed_bit_field_1): Likewise.
19933         (extract_bit_field_1): Likewise.
19934         (expand_divmod): Likewise.
19935         (emit_store_flag_1): Likewise.
19936         * expr.c (init_expr_target): Likewise.
19937         (convert_move): Likewise.
19938         (alignment_for_piecewise_move): Likewise.
19939         (widest_int_mode_for_size): Likewise.
19940         (emit_block_move_via_movmem): Likewise.
19941         (copy_blkmode_to_reg): Likewise.
19942         (set_storage_via_setmem): Likewise.
19943         (compress_float_constant): Likewise.
19944         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19945         * optabs-query.c (get_best_extraction_insn): Likewise.
19946         * optabs.c (expand_binop): Likewise.
19947         (expand_twoval_unop): Likewise.
19948         (expand_twoval_binop): Likewise.
19949         (widen_leading): Likewise.
19950         (widen_bswap): Likewise.
19951         (expand_parity): Likewise.
19952         (expand_unop): Likewise.
19953         (prepare_cmp_insn): Likewise.
19954         (prepare_float_lib_cmp): Likewise.
19955         (expand_float): Likewise.
19956         (expand_fix): Likewise.
19957         (expand_sfix_optab): Likewise.
19958         * postreload.c (move2add_use_add2_insn): Likewise.
19959         * reg-stack.c (reg_to_stack): Likewise.
19960         * reginfo.c (choose_hard_reg_mode): Likewise.
19961         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19962         * stor-layout.c (mode_for_size): Likewise.
19963         (smallest_mode_for_size): Likewise.
19964         (mode_for_vector): Likewise.
19965         (finish_bitfield_representative): Likewise.
19966         * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
19967         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19968         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
19969         * var-tracking.c (prepare_call_arguments): Likewise.
19971 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19972             Alan Hayward  <alan.hayward@arm.com>
19973             David Sherwood  <david.sherwood@arm.com>
19975         * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
19976         * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
19977         USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
19978         * machmode.h (mode_size): Move earlier in file.
19979         (mode_precision): Likewise.
19980         (mode_inner): Likewise.
19981         (mode_nunits): Likewise.
19982         (mode_unit_size): Likewise.
19983         (unit_unit_precision): Likewise.
19984         (mode_wider): Likewise.
19985         (mode_2xwider): Likewise.
19986         (machine_mode): New class.
19987         (mode_to_bytes): New function.
19988         (mode_to_bits): Likewise.
19989         (mode_to_precision): Likewise.
19990         (mode_to_inner): Likewise.
19991         (mode_to_unit_size): Likewise.
19992         (mode_to_unit_precision): Likewise.
19993         (mode_to_nunits): Likewise.
19994         (GET_MODE_SIZE): Use mode_to_bytes.
19995         (GET_MODE_BITSIZE): Use mode_to_bits.
19996         (GET_MODE_PRECISION): Use mode_to_precision.
19997         (GET_MODE_INNER): Use mode_to_inner.
19998         (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
19999         (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
20000         (GET_MODE_NUNITS): Use mode_to_nunits.
20001         * system.h (ALWAYS_INLINE): New macro.
20002         * config/powerpcspe/powerpcspe-c.c
20003         (altivec_resolve_overloaded_builtin): Use machine_mode instead of
20004         int for arg1_mode and arg2_mode.
20006 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20007             Alan Hayward  <alan.hayward@arm.com>
20008             David Sherwood  <david.sherwood@arm.com>
20010         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
20011         Prefix mode names with E_ in case statements.
20012         * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20013         * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
20014         (aarch64_split_simd_move): Likewise.
20015         (aarch64_gen_storewb_pair): Likewise.
20016         (aarch64_gen_loadwb_pair): Likewise.
20017         (aarch64_gen_store_pair): Likewise.
20018         (aarch64_gen_load_pair): Likewise.
20019         (aarch64_get_condition_code_1): Likewise.
20020         (aarch64_constant_pool_reload_icode): Likewise.
20021         (get_rsqrte_type): Likewise.
20022         (get_rsqrts_type): Likewise.
20023         (get_recpe_type): Likewise.
20024         (get_recps_type): Likewise.
20025         (aarch64_gimplify_va_arg_expr): Likewise.
20026         (aarch64_simd_container_mode): Likewise.
20027         (aarch64_emit_load_exclusive): Likewise.
20028         (aarch64_emit_store_exclusive): Likewise.
20029         (aarch64_expand_compare_and_swap): Likewise.
20030         (aarch64_gen_atomic_cas): Likewise.
20031         (aarch64_emit_bic): Likewise.
20032         (aarch64_emit_atomic_swap): Likewise.
20033         (aarch64_emit_atomic_load_op): Likewise.
20034         (aarch64_evpc_trn): Likewise.
20035         (aarch64_evpc_uzp): Likewise.
20036         (aarch64_evpc_zip): Likewise.
20037         (aarch64_evpc_ext): Likewise.
20038         (aarch64_evpc_rev): Likewise.
20039         (aarch64_evpc_dup): Likewise.
20040         (aarch64_gen_ccmp_first): Likewise.
20041         (aarch64_gen_ccmp_next): Likewise.
20042         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
20043         (alpha_emit_xfloating_libcall): Likewise.
20044         (emit_insxl): Likewise.
20045         (alpha_arg_type): Likewise.
20046         * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
20047         (arc_preferred_simd_mode): Likewise.
20048         (arc_secondary_reload): Likewise.
20049         (get_arc_condition_code): Likewise.
20050         (arc_print_operand): Likewise.
20051         (arc_legitimate_constant_p): Likewise.
20052         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20053         * config/arc/arc.md (casesi_load): Likewise.
20054         (casesi_compact_jump): Likewise.
20055         * config/arc/predicates.md (proper_comparison_operator): Likewise.
20056         (cc_use_register): Likewise.
20057         * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20058         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
20059         (arm_init_iwmmxt_builtins): Likewise.
20060         * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
20061         (neon_expand_vector_init): Likewise.
20062         (arm_attr_length_move_neon): Likewise.
20063         (maybe_get_arm_condition_code): Likewise.
20064         (arm_emit_vector_const): Likewise.
20065         (arm_preferred_simd_mode): Likewise.
20066         (arm_output_iwmmxt_tinsr): Likewise.
20067         (thumb1_output_casesi): Likewise.
20068         (thumb2_output_casesi): Likewise.
20069         (arm_emit_load_exclusive): Likewise.
20070         (arm_emit_store_exclusive): Likewise.
20071         (arm_expand_compare_and_swap): Likewise.
20072         (arm_evpc_neon_vuzp): Likewise.
20073         (arm_evpc_neon_vzip): Likewise.
20074         (arm_evpc_neon_vrev): Likewise.
20075         (arm_evpc_neon_vtrn): Likewise.
20076         (arm_evpc_neon_vext): Likewise.
20077         (arm_validize_comparison): Likewise.
20078         * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
20079         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
20080         * config/avr/avr.c (avr_rtx_costs_1): Likewise.
20081         * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
20082         (c6x_preferred_simd_mode): Likewise.
20083         * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
20084         (epiphany_rtx_costs): Likewise.
20085         * config/epiphany/predicates.md (proper_comparison_operator):
20086         Likewise.
20087         * config/frv/frv.c (condexec_memory_operand): Likewise.
20088         (frv_emit_move): Likewise.
20089         (output_move_single): Likewise.
20090         (output_condmove_single): Likewise.
20091         (frv_hard_regno_mode_ok): Likewise.
20092         (frv_matching_accg_mode): Likewise.
20093         * config/h8300/h8300.c (split_adds_subs): Likewise.
20094         (h8300_rtx_costs): Likewise.
20095         (h8300_print_operand): Likewise.
20096         (compute_mov_length): Likewise.
20097         (output_logical_op): Likewise.
20098         (compute_logical_op_length): Likewise.
20099         (compute_logical_op_cc): Likewise.
20100         (h8300_shift_needs_scratch_p): Likewise.
20101         (output_a_shift): Likewise.
20102         (compute_a_shift_length): Likewise.
20103         (compute_a_shift_cc): Likewise.
20104         (expand_a_rotate): Likewise.
20105         (output_a_rotate): Likewise.
20106         * config/i386/i386.c (classify_argument): Likewise.
20107         (function_arg_advance_32): Likewise.
20108         (function_arg_32): Likewise.
20109         (function_arg_64): Likewise.
20110         (function_value_64): Likewise.
20111         (ix86_gimplify_va_arg): Likewise.
20112         (ix86_legitimate_constant_p): Likewise.
20113         (put_condition_code): Likewise.
20114         (split_double_mode): Likewise.
20115         (ix86_avx256_split_vector_move_misalign): Likewise.
20116         (ix86_expand_vector_logical_operator): Likewise.
20117         (ix86_split_idivmod): Likewise.
20118         (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
20119         (ix86_build_const_vector): Likewise.
20120         (ix86_build_signbit_mask): Likewise.
20121         (ix86_match_ccmode): Likewise.
20122         (ix86_cc_modes_compatible): Likewise.
20123         (ix86_expand_branch): Likewise.
20124         (ix86_expand_sse_cmp): Likewise.
20125         (ix86_expand_sse_movcc): Likewise.
20126         (ix86_expand_int_sse_cmp): Likewise.
20127         (ix86_expand_vec_perm_vpermi2): Likewise.
20128         (ix86_expand_vec_perm): Likewise.
20129         (ix86_expand_sse_unpack): Likewise.
20130         (ix86_expand_int_addcc): Likewise.
20131         (ix86_split_to_parts): Likewise.
20132         (ix86_vectorize_builtin_gather): Likewise.
20133         (ix86_vectorize_builtin_scatter): Likewise.
20134         (avx_vpermilp_parallel): Likewise.
20135         (inline_memory_move_cost): Likewise.
20136         (ix86_tieable_integer_mode_p): Likewise.
20137         (x86_maybe_negate_const_int): Likewise.
20138         (ix86_expand_vector_init_duplicate): Likewise.
20139         (ix86_expand_vector_init_one_nonzero): Likewise.
20140         (ix86_expand_vector_init_one_var): Likewise.
20141         (ix86_expand_vector_init_concat): Likewise.
20142         (ix86_expand_vector_init_interleave): Likewise.
20143         (ix86_expand_vector_init_general): Likewise.
20144         (ix86_expand_vector_set): Likewise.
20145         (ix86_expand_vector_extract): Likewise.
20146         (emit_reduc_half): Likewise.
20147         (ix86_emit_i387_round): Likewise.
20148         (ix86_mangle_type): Likewise.
20149         (ix86_expand_round_sse4): Likewise.
20150         (expand_vec_perm_blend): Likewise.
20151         (canonicalize_vector_int_perm): Likewise.
20152         (ix86_expand_vec_one_operand_perm_avx512): Likewise.
20153         (expand_vec_perm_1): Likewise.
20154         (expand_vec_perm_interleave3): Likewise.
20155         (expand_vec_perm_even_odd_pack): Likewise.
20156         (expand_vec_perm_even_odd_1): Likewise.
20157         (expand_vec_perm_broadcast_1): Likewise.
20158         (ix86_vectorize_vec_perm_const_ok): Likewise.
20159         (ix86_expand_vecop_qihi): Likewise.
20160         (ix86_expand_mul_widen_hilo): Likewise.
20161         (ix86_expand_sse2_abs): Likewise.
20162         (ix86_expand_pextr): Likewise.
20163         (ix86_expand_pinsr): Likewise.
20164         (ix86_preferred_simd_mode): Likewise.
20165         (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
20166         * config/i386/sse.md (*andnot<mode>3): Likewise.
20167         (<mask_codefor><code><mode>3<mask_name>): Likewise.
20168         (*<code><mode>3): Likewise.
20169         * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
20170         (ia64_expand_atomic_op): Likewise.
20171         (ia64_arg_type): Likewise.
20172         (ia64_mode_to_int): Likewise.
20173         (ia64_scalar_mode_supported_p): Likewise.
20174         (ia64_vector_mode_supported_p): Likewise.
20175         (expand_vec_perm_broadcast): Likewise.
20176         * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
20177         (iq2000_function_arg_advance): Likewise.
20178         (iq2000_function_arg): Likewise.
20179         * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
20180         * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
20181         (m68k_libcall_value): Likewise.
20182         (m68k_function_value): Likewise.
20183         (sched_attr_op_type): Likewise.
20184         * config/mcore/mcore.c (mcore_output_move): Likewise.
20185         * config/microblaze/microblaze.c (microblaze_function_arg_advance):
20186         Likewise.
20187         (microblaze_function_arg): Likewise.
20188         * config/mips/mips.c (mips16_build_call_stub): Likewise.
20189         (mips_print_operand): Likewise.
20190         (mips_mode_ok_for_mov_fmt_p): Likewise.
20191         (mips_vector_mode_supported_p): Likewise.
20192         (mips_preferred_simd_mode): Likewise.
20193         (mips_expand_vpc_loongson_even_odd): Likewise.
20194         (mips_expand_vec_unpack): Likewise.
20195         (mips_expand_vi_broadcast): Likewise.
20196         (mips_expand_vector_init): Likewise.
20197         (mips_expand_vec_reduc): Likewise.
20198         (mips_expand_msa_cmp): Likewise.
20199         * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
20200         * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
20201         (cc_flags_for_mode): Likewise.
20202         * config/msp430/msp430.c (msp430_print_operand): Likewise.
20203         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
20204         (nds32_output_casesi_pc_relative): Likewise.
20205         * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20206         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
20207         (nvptx_gen_unpack): Likewise.
20208         (nvptx_gen_pack): Likewise.
20209         (nvptx_gen_shuffle): Likewise.
20210         (nvptx_gen_wcast): Likewise.
20211         (nvptx_preferred_simd_mode): Likewise.
20212         * config/pa/pa.c (pa_secondary_reload): Likewise.
20213         * config/pa/predicates.md (base14_operand): Likewise.
20214         * config/powerpcspe/powerpcspe-c.c
20215         (altivec_resolve_overloaded_builtin): Likewise.
20216         * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
20217         Likewise.
20218         (rs6000_preferred_simd_mode): Likewise.
20219         (output_vec_const_move): Likewise.
20220         (rs6000_expand_vector_extract): Likewise.
20221         (rs6000_split_vec_extract_var): Likewise.
20222         (reg_offset_addressing_ok_p): Likewise.
20223         (rs6000_legitimate_offset_address_p): Likewise.
20224         (rs6000_legitimize_address): Likewise.
20225         (rs6000_emit_set_const): Likewise.
20226         (rs6000_const_vec): Likewise.
20227         (rs6000_emit_move): Likewise.
20228         (spe_build_register_parallel): Likewise.
20229         (rs6000_darwin64_record_arg_recurse): Likewise.
20230         (swap_selector_for_mode): Likewise.
20231         (spe_init_builtins): Likewise.
20232         (paired_init_builtins): Likewise.
20233         (altivec_init_builtins): Likewise.
20234         (do_load_for_compare): Likewise.
20235         (rs6000_generate_compare): Likewise.
20236         (rs6000_expand_float128_convert): Likewise.
20237         (emit_load_locked): Likewise.
20238         (emit_store_conditional): Likewise.
20239         (rs6000_output_function_epilogue): Likewise.
20240         (rs6000_handle_altivec_attribute): Likewise.
20241         (rs6000_function_value): Likewise.
20242         (emit_fusion_gpr_load): Likewise.
20243         (emit_fusion_p9_load): Likewise.
20244         (emit_fusion_p9_store): Likewise.
20245         * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20246         (fusion_gpr_mem_load): Likewise.
20247         (fusion_addis_mem_combo_load): Likewise.
20248         (fusion_addis_mem_combo_store): Likewise.
20249         * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20250         (fusion_gpr_mem_load): Likewise.
20251         (fusion_addis_mem_combo_load): Likewise.
20252         (fusion_addis_mem_combo_store): Likewise.
20253         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20254         Likewise.
20255         * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20256         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20257         (rs6000_preferred_simd_mode): Likewise.
20258         (output_vec_const_move): Likewise.
20259         (rs6000_expand_vector_extract): Likewise.
20260         (rs6000_split_vec_extract_var): Likewise.
20261         (reg_offset_addressing_ok_p): Likewise.
20262         (rs6000_legitimate_offset_address_p): Likewise.
20263         (rs6000_legitimize_address): Likewise.
20264         (rs6000_emit_set_const): Likewise.
20265         (rs6000_const_vec): Likewise.
20266         (rs6000_emit_move): Likewise.
20267         (rs6000_darwin64_record_arg_recurse): Likewise.
20268         (swap_selector_for_mode): Likewise.
20269         (paired_init_builtins): Likewise.
20270         (altivec_init_builtins): Likewise.
20271         (rs6000_expand_float128_convert): Likewise.
20272         (emit_load_locked): Likewise.
20273         (emit_store_conditional): Likewise.
20274         (rs6000_output_function_epilogue): Likewise.
20275         (rs6000_handle_altivec_attribute): Likewise.
20276         (rs6000_function_value): Likewise.
20277         (emit_fusion_gpr_load): Likewise.
20278         (emit_fusion_p9_load): Likewise.
20279         (emit_fusion_p9_store): Likewise.
20280         * config/rx/rx.c (rx_gen_move_template): Likewise.
20281         (flags_from_mode): Likewise.
20282         * config/s390/predicates.md (s390_alc_comparison): Likewise.
20283         (s390_slb_comparison): Likewise.
20284         * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20285         (s390_vector_mode_supported_p): Likewise.
20286         (s390_cc_modes_compatible): Likewise.
20287         (s390_match_ccmode_set): Likewise.
20288         (s390_canonicalize_comparison): Likewise.
20289         (s390_emit_compare_and_swap): Likewise.
20290         (s390_branch_condition_mask): Likewise.
20291         (s390_rtx_costs): Likewise.
20292         (s390_secondary_reload): Likewise.
20293         (__SECONDARY_RELOAD_CASE): Likewise.
20294         (s390_expand_cs): Likewise.
20295         (s390_preferred_simd_mode): Likewise.
20296         * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20297         * config/sh/sh.c (sh_print_operand): Likewise.
20298         (dump_table): Likewise.
20299         (sh_secondary_reload): Likewise.
20300         * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20301         * config/sh/sh.md (casesi_worker_1): Likewise.
20302         (casesi_worker_2): Likewise.
20303         * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20304         (fcc_comparison_operator): Likewise.
20305         * config/sparc/sparc.c (sparc_expand_move): Likewise.
20306         (emit_soft_tfmode_cvt): Likewise.
20307         (sparc_preferred_simd_mode): Likewise.
20308         (output_cbranch): Likewise.
20309         (sparc_print_operand): Likewise.
20310         (sparc_expand_vec_perm_bmask): Likewise.
20311         (vector_init_bshuffle): Likewise.
20312         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20313         (spu_vector_mode_supported_p): Likewise.
20314         (spu_expand_insv): Likewise.
20315         (spu_emit_branch_or_set): Likewise.
20316         (spu_handle_vector_attribute): Likewise.
20317         (spu_builtin_splats): Likewise.
20318         (spu_builtin_extract): Likewise.
20319         (spu_builtin_promote): Likewise.
20320         (spu_expand_sign_extend): Likewise.
20321         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20322         (tilegx_simd_int): Likewise.
20323         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20324         (tilepro_simd_int): Likewise.
20325         * config/v850/v850.c (const_double_split): Likewise.
20326         (v850_print_operand): Likewise.
20327         (ep_memory_offset): Likewise.
20328         * config/vax/vax.c (vax_rtx_costs): Likewise.
20329         (vax_output_int_move): Likewise.
20330         (vax_output_int_add): Likewise.
20331         (vax_output_int_subtract): Likewise.
20332         * config/visium/predicates.md (visium_branch_operator): Likewise.
20333         * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20334         (visium_print_operand_address): Likewise.
20335         * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20336         * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20337         (xtensa_expand_conditional_branch): Likewise.
20338         (xtensa_copy_incoming_a7): Likewise.
20339         (xtensa_output_literal): Likewise.
20340         * dfp.c (decimal_real_maxval): Likewise.
20341         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20343 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20344             Alan Hayward  <alan.hayward@arm.com>
20345             David Sherwood  <david.sherwood@arm.com>
20347         * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20348         (mode_nunits_inline): Likewise.
20349         (mode_inner_inline): Likewise.
20350         (mode_unit_size_inline): Likewise.
20351         (mode_unit_precision_inline): Likewise.
20352         (emit_insn_modes_h): Likewise.  Also emit a #define of the
20353         unprefixed name.
20354         (emit_mode_wider): Add an E_ prefix to mode names.
20355         (emit_mode_complex): Likewise.
20356         (emit_mode_inner): Likewise.
20357         (emit_mode_adjustments): Likewise.
20358         (emit_mode_int_n): Likewise.
20359         * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20360         (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20361         (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20362         (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20363         (CRC32_BUILTIN, ENTRY): Likewise.
20364         * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20365         (aarch64_pop_regs): Likewise.
20366         (aarch64_process_components): Likewise.
20367         * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20368         * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20369         (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20370         (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20371         * config/arm/arm.c (arm_init_libfuncs): Likewise.
20372         * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20373         Likewise.
20374         * config/i386/i386-builtin.def (pcmpestr): Likewise.
20375         (pcmpistr): Likewise.
20376         * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20377         * config/mmix/mmix.c (mmix_output_condition): Likewise.
20378         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20379         Likewise.
20380         * config/rl78/rl78.c (mduc_regs): Likewise.
20381         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20382         (htm_expand_builtin): Likewise.
20383         * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20384         * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20385         * config/xtensa/xtensa.c (print_operand): Likewise.
20386         * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20387         (NUM_MODE_VECTOR_INT): Likewise.
20388         * genoutput.c (null_operand): Likewise.
20389         (output_operand_data): Likewise.
20390         * genrecog.c (print_parameter_value): Likewise.
20391         * lra.c (debug_operand_data): Likewise.
20393 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20395         * dwarf2out.c (scompare_loc_descriptor_wide)
20396         (scompare_loc_descriptor_narrow): New functions, split out from...
20397         (scompare_loc_descriptor): ...here.
20398         * expmed.c (emit_store_flag_int): New function, split out from...
20399         (emit_store_flag): ...here.
20401 2017-08-30  Richard Biener  <rguenther@suse.de>
20403         * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20404         (dwarf2out_early_finish): Move setting of AT_pubnames from
20405         early debug output to early finish.
20407 2017-08-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20409         * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20410         and -mdata-region to the assembler.
20412 2017-08-30  Richard Biener  <rguenther@suse.de>
20414         * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20415         attributes.
20416         (gen_subprogram_die): Add DW_AT_object_pointer only early.
20417         (dwarf2out_early_global_decl): Only generate a DIE for the
20418         abstract origin if it doesn't already exist or is a declaration DIE.
20419         (resolve_addr): Do not add the linkage name twice when
20420         generating a stub DIE for the DW_TAG_GNU_call_site target.
20422 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20424         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20425         Use machine_mode rather than int for arg1_mode.
20427 2017-08-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20429         PR target/82015
20430         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20431         that the second argument of the built-in functions to unpack
20432         128-bit scalar types to 64-bit values is 0 or 1.  Change to use a
20433         switch statement instead a lot of if statements.
20434         * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20435         Allow 64-bit values to be in Altivec registers as well as
20436         traditional floating point registers.
20437         (pack<mode>, FMOVE128_VSX iterator): Likewise.
20439 2017-08-29  Alexander Monakov  <amonakov@ispras.ru>
20441         * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20442         MAX_REGS_PER_ADDRESS == 1.
20444 2017-08-29  Uros Bizjak  <ubizjak@gmail.com>
20446         * config/i386/i386.opt (flag_fentry): Do not init to -1.
20447         * config/i386/i386.c (ix86_option_override_internal): Simplify
20448         setting of opts->x_flag_entry.
20450 2017-08-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20451             Jakub Jelinek  <jakub@redhat.com>
20452             Richard Biener  <rguenther@suse.de>
20454         PR tree-optimization/81503
20455         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20456         folded constant fits in the target type; reorder tests for clarity.
20458 2017-08-29  Martin Liska  <mliska@suse.cz>
20460         * passes.def: Include pass_lower_switch.
20461         * stmt.c (dump_case_nodes): Remove and move to
20462         tree-switch-conversion.
20463         (case_values_threshold): Likewise.
20464         (expand_switch_as_decision_tree_p): Likewise.
20465         (emit_case_decision_tree): Likewise.
20466         (expand_case): Likewise.
20467         (balance_case_nodes): Likewise.
20468         (node_has_low_bound): Likewise.
20469         (node_has_high_bound): Likewise.
20470         (node_is_bounded): Likewise.
20471         (emit_case_nodes): Likewise.
20472         (struct simple_case_node): New struct.
20473         (add_case_node): Remove.
20474         (emit_case_dispatch_table): Use vector instead of case_list.
20475         (reset_out_edges_aux): Remove.
20476         (compute_cases_per_edge): Likewise.
20477         (expand_case): Build list of simple_case_node.
20478         (expand_sjlj_dispatch_table): Use it.
20479         * tree-switch-conversion.c (struct case_node): Moved from
20480         stmt.c and adjusted.
20481         (emit_case_nodes): Likewise.
20482         (node_has_low_bound): Likewise.
20483         (node_has_high_bound): Likewise.
20484         (node_is_bounded): Likewise.
20485         (case_values_threshold): Likewise.
20486         (reset_out_edges_aux): Likewise.
20487         (compute_cases_per_edge): Likewise.
20488         (add_case_node): Likewise.
20489         (dump_case_nodes): Likewise.
20490         (balance_case_nodes): Likewise.
20491         (expand_switch_as_decision_tree_p): Likewise.
20492         (emit_jump): Likewise.
20493         (emit_case_decision_tree): Likewise.
20494         (try_switch_expansion): Likewise.
20495         (do_jump_if_equal): Likewise.
20496         (emit_cmp_and_jump_insns): Likewise.
20497         (fix_phi_operands_for_edge): New function.
20498         (record_phi_operand_mapping): Likewise.
20499         (class pass_lower_switch): New pass.
20500         (pass_lower_switch::execute): New function.
20501         (make_pass_lower_switch): Likewise.
20502         (conditional_probability):
20503         * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20504         * tree-pass.h: Add make_pass_lower_switch.
20506 2017-08-29  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20508         PR target/80993
20509         * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20510         handlers as used.
20512 2017-08-29  Richard Biener  <rguenther@suse.de>
20514         * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20515         we do not add a DW_AT_inline attribute twice.
20516         (gen_subprogram_die): Remove code setting DW_AT_inline on
20517         DECL_ABSTRACT_P nodes.
20519 2017-08-29  Richard Sandiford  <richard.sandiford@linaro.org>
20521         * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20522         calls to internal functions.
20523         (gimplify_modify_expr): Likewise.
20524         * tree-call-cdce.c (use_internal_fn): Likewise.
20525         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20526         (convert_to_divmod): Set the nothrow flag.
20527         * tree-if-conv.c (predicate_mem_writes):  Likewise.
20528         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20529         (vectorizable_call): Likewise.
20530         (vectorizable_store): Likewise.
20531         (vectorizable_load): Likewise.
20532         * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20533         (vect_recog_mask_conversion_pattern): Likewise.
20535 2017-08-29  Martin Liska  <mliska@suse.cz>
20537         PR other/39851
20538         * gcc.c (driver_handle_option): Add new argument.
20539         * opts-common.c (handle_option): Pass
20540         target_option_override_hook.
20541         * opts-global.c (lang_handle_option): Add new option.
20542         (set_default_handlers):  Add new argument.
20543         (decode_options): Likewise.
20544         * opts.c (target_handle_option): Likewise.
20545         (common_handle_option): Call target_option_override_hook.
20546         * opts.h (struct cl_option_handler_func): Add hook for
20547         target option override.
20548         (struct cl_option_handlers): Likewise.
20549         (set_default_handlers): Add new argument.
20550         (decode_options): Likewise.
20551         (common_handle_option): Likewise.
20552         (target_handle_option): Likewise.
20553         * toplev.c (toplev::main): Pass targetm.target_option.override
20554         hook.
20556 2017-08-29  Richard Biener  <rguenther@suse.de>
20557         Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20559         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20560         life to the active subtree.
20562 2017-08-28  Jeff Law  <law@redhat.com>
20564         * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20565         derive_equivalences.
20566         (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20567         Code moved into....
20568         (edge_info::derive_equivalences): New private member function
20570         * tree-ssa-dom.c (class edge_info): Changed from a struct
20571         to a class.  Add ctor/dtor, methods and data members.
20572         (edge_info::edge_info): Renamed from allocate_edge_info.
20573         Initialize additional members.
20574         (edge_info::~edge_info): New.
20575         (free_dom_edge_info): Delete the edge info.
20576         (record_edge_info): Use new class & associated member functions.
20577         Tighten forms for testing for edge equivalences.
20578         (record_temporary_equivalences): Iterate over the simple
20579         equivalences rather than assuming there's only one per edge.
20580         (cprop_into_successor_phis): Iterate over the simple
20581         equivalences rather than assuming there's only one per edge.
20582         (optimize_stmt): Use operand_equal_p rather than pointer
20583         equality for mini-DSE code.
20585 2017-08-28  Nathan Sidwell  <nathan@acm.org>
20587         * gcc.c (execute): Fold SIGPIPE handling into switch
20588         statement.  Adjust internal error message.
20590 2017-08-28  Richard Biener  <rguenther@suse.de>
20592         PR debug/81993
20593         * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20594         Do nothing for removed DIEs.
20596 2017-08-28  Richard Biener  <rguenther@suse.de>
20598         PR tree-optimization/81977
20599         * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20600         memcpy.
20602 2017-08-28  Alexander Monakov  <amonakov@ispras.ru>
20604         PR target/80640
20605         * doc/md.texi (mem_thread_fence): Remove mention of mode.  Rewrite.
20606         * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20607         using targetm.gen_mem_thread_fence.
20609 2017-08-27  Uros Bizjak  <ubizjak@gmail.com>
20611         PR target/81995
20612         * config/i386/i386.md (*<btsc><mode>): Change operand 2
20613         predicate to register_operand.  Reorder operands.
20614         (*btr<mode>): Ditto.
20615         (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20616         (*btr<mode>_mask): Ditto.
20618 2017-08-25  Steven Munroe  <munroesj@gcc.gnu.org>
20620         * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20621         * config/rs6000/xmmintrin.h: New file.
20622         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20624 2017-08-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20626         PR target/81504
20627         * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20628         parameter and_insn and return it.
20629         (recombine_lvx_pattern): Insert a copy to ensure availability of
20630         the base register of the copied masking operation at the point of
20631         the instruction replacement.
20632         (recombine_stvx_pattern): Likewise.
20634 2017-08-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
20636         * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20637         undocumented switches.
20638         (-mpower9-dform-vector): Likewise.
20639         (-mpower9-dform): Likewise.
20640         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20641         comments to delete references to -mpower9-dform* switches.
20642         * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20643         Delete reference to -mpower9-dform* switches, test for
20644         -mpower9-vector instead.
20645         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20646         (OTHER_P9_VECTOR_MASKS): Likewise.
20647         (POWERPC_MASKS): Likewise.
20648         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20649         tests against -mpower9-dform* to -mpower9-vector.  Delete code
20650         that checked for -mpower9-dform* consistancy with other options.
20651         Add test for -mpower9-misc to enable other power9 switches.
20652         (rs6000_init_hard_regno_mode_ok): Likewise.
20653         (rs6000_option_override_internal): Likewise.
20654         (rs6000_emit_prologue): Likewise.
20655         (rs6000_emit_epilogue): Likewise.
20656         (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20657         (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20658         (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20659         -mpower9-vector.
20660         (emit_fusion_p9_store): Likewise.
20661         * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20662         resetting these macros if the assembler does not support ISA 3.0
20663         instructions.
20664         (TARGET_P9_DFORM_VECTOR): Likewise.
20665         * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20666         Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20668 2017-08-25  Alan Modra  <amodra@gmail.com>
20670         PR rtl-optimization/81747
20671         * cse.c (cse_extended_basic_block): Don't attempt to record
20672         equivalences for degenerate conditional branches that branch
20673         to their fall-through.
20675 2017-08-24  Martin Sebor  <msebor@redhat.com>
20677         PR middle-end/81908
20678         * gimple-fold.c (size_must_be_zero_p): New function.
20679         (gimple_fold_builtin_memory_op): Call it.
20681 2017-08-24  Steven Munroe  <munroesj@gcc.gnu.org>
20683         * config/rs6000/mm_malloc.h: New file.
20685 2017-08-24  Bin Cheng  <bin.cheng@arm.com>
20687         PR tree-optimization/81913
20688         * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20689         analysis when either IVs in condition can wrap.
20691 2017-08-24  Uros Bizjak  <ubizjak@gmail.com>
20693         * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20694         * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20696 2017-08-24  Richard Biener  <rguenther@suse.de>
20698         PR target/81921
20699         * targhooks.c (default_target_can_inline_p): Properly
20700         use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20701         is present and always compare.
20702         * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20703         infer -mfpmath=sse from TARGET_SSE_P.
20704         (ix86_can_inline_p): Properly use target_option_default_node when
20705         no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20707 2017-08-24  Richard Biener  <rguenther@suse.de>
20709         PR debug/81936
20710         * dwarf2out.c (output_die): Handle flag_generate_offload like
20711         flag_generate_lto.
20712         (output_comp_unit): Likewise.
20713         (gen_array_type_die): Likewise.
20714         (dwarf2out_early_finish): Likewise.
20715         (note_variable_value_in_expr): Likewise.
20716         (dwarf2out_finish): Likewise.  Adjust assert.
20717         * cgraphunit.c (symbol_table::compile): Move setting of
20718         flag_generate_offload earlier ...
20719         (symbol_table::finalize_compilation_unit): ... here, before
20720         early debug finalization.
20722 2017-08-24  Richard Biener  <rguenther@suse.de>
20724         * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20725         and ipa-fnsummary.h.
20726         (ix86_can_inline_p): When ix86_fpmath flags do not match
20727         check whether the callee uses FP math at all.
20729 2017-08-24  Aldy Hernandez  <aldyh@redhat.com>
20731         PR middle-end/81931
20732         * tree-ssanames.c (get_nonzero_bits): Use element_precision
20733         instead of TYPE_PRECISION.
20735 2017-08-24  Richard Sandiford  <richard.sandiford@linaro.org>
20736             Alan Hayward  <alan.hayward@arm.com>
20737             David Sherwood  <david.sherwood@arm.com>
20739         * combine.c (make_extraction): Use subreg_offset_from_lsb.
20741 2017-08-23  Daniel Santos  <daniel.santos@pobox.com>
20743         * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20744         Remove field.
20745         (ix86_frame::stack_realign_allocate): New field.
20746         (struct machine_frame_state): Modify comments.
20747         (machine_frame_state::sp_realigned_fp_end): New field.
20748         * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20749         layout calculation.
20750         (sp_valid_at): Add assertion to assure no attempt to access invalid
20751         offset of a realigned stack.
20752         (fp_valid_at): Likewise.
20753         (choose_baseaddr): Modify comments.
20754         (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20755         ix86_expand_prologue.
20756         (ix86_expand_prologue): Modify stack realignment and allocation.
20757         (ix86_expand_epilogue): Modify comments.
20758         * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20759         avx2_runtime, avx512f, and avx512f_runtime.
20761 2017-08-23  Uros Bizjak  <ubizjak@gmail.com>
20763         * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20764         (mstackrealign): Do not init to -1.
20765         * config/i386/i386.c (ix86_option_override_internal):
20766         Check opts_set, not opts when setting default value of
20767         opts->x_ix86_force_align_arg_pointer.
20769 2017-08-23  Richard Biener  <rguenther@suse.de>
20771         * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20772         lang_hooks.decl_printable_name.
20773         * print-rtl-function.c (print_rtx_function): Likewise.
20774         * tree-pretty-print.c (dump_function_header): Likewise.
20776 2017-08-23  Richard Biener  <rguenther@suse.de>
20778         PR lto/81940
20779         * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20780         -g0 at compile-time.
20782 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20784         PR middle-end/19706
20785         * doc/sourcebuild.texi (Other hardware attributes):
20786         Document xorsign.
20788 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20790         PR middle-end/19706
20791         * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20792         Fix single-use check.
20794 2017-08-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20796         * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20798 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20800         * doc/install.texi: Modify to add more details on running selected
20801         tests.
20803 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20805         * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20806         is combined with -mabi=ms.
20807         (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20808         ms_abi.
20810 2017-08-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20812         PR tree-optimization/81488
20813         * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20814         and cached_basis fields.
20815         (MAX_SPREAD): New constant.
20816         (alloc_cand_and_find_basis): Initialize new fields.
20817         (clear_visited): New function.
20818         (create_phi_basis_1): Rename from create_phi_basis, set visited
20819         and cached_basis fields.
20820         (create_phi_basis): New wrapper function.
20821         (phi_add_costs_1): Rename from phi_add_costs, add spread
20822         parameter, set visited field, short-circuit when limits reached.
20823         (phi_add_costs): New wrapper function.
20824         (record_phi_increments_1): Rename from record_phi_increments, set
20825         visited field.
20826         (record_phi_increments): New wrapper function.
20827         (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20828         (phi_incr_cost): New wrapper function.
20829         (all_phi_incrs_profitable_1): Rename from
20830         all_phi_incrs_profitable, set visited field.
20831         (all_phi_incrs_profitable): New wrapper function.
20833 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20834             Alan Hayward  <alan.hayward@arm.com>
20835             David Sherwood  <david.sherwood@arm.com>
20837         * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20838         that takes the outer and inner modes.
20839         * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20840         comparison as the canonical test for a paradoxical subreg.
20841         * combine.c (simplify_set): Use paradoxical_subreg_p.
20842         (make_extraction): Likewise.
20843         (force_to_mode): Likewise.
20844         (rtx_equal_for_field_assignment_p): Likewise.
20845         (gen_lowpart_for_combine): Likewise.
20846         (simplify_comparison): Likewise.
20847         * cse.c (equiv_constant): Likewise.
20848         * expmed.c (store_bit_field_1): Likewise.
20849         * final.c (alter_subreg): Likewise.
20850         * fwprop.c (propagate_rtx): Likewise.
20851         (forward_propagate_subreg): Likewise.
20852         * ira-conflicts.c (ira_build_conflicts): Likewise.
20853         * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20854         * lra-constraints.c (curr_insn_transform): Likewise.
20855         (split_reg): Likewise.
20856         * lra-eliminations.c (move_plus_up): Likewise.
20857         (lra_eliminate_regs_1): Likewise.
20858         * recog.c (general_operand): Likewise.
20859         * ree.c (combine_reaching_defs): Likewise.
20860         * reload.c (push_reload): Likewise.
20861         (find_reloads): Likewise.
20862         * reload1.c (elimination_effects): Likewise.
20863         (compute_reload_subreg_offset): Likewise.
20864         (choose_reload_regs): Likewise.
20865         * rtlanal.c (subreg_lsb_1): Likewise.
20866         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20867         (simplify_subreg): Likewise.
20868         * var-tracking.c (track_loc_p): Likewise.
20869         * emit-rtl.c (byte_lowpart_offset): Likewise.
20870         (paradoxical_subreg_p): Delete out-of-line definition.
20872 2017-08-22  Jeff Law  <law@redhat.com>
20874         PR tree-optimization/81741
20875         PR tree-optimization/71947
20876         * tree-ssa-dom.c: Include tree-inline.h.
20877         (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20878         equivalences if one is more expensive to compute than the other.
20879         * tree-ssa-scopedtables.h (class const_or_copies): Make
20880         record_const_or_copy_raw method private.
20881         (class avail_exprs_stack): New method simplify_binary_operation.
20882         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20883         avail_exprs_stack::simplify_binary_operation as needed.
20884         (avail_exprs_stack::simplify_binary_operation): New function.
20886 2017-08-22  Sebastian Huber  <sebastian.huber@embedded-brains.de>
20888         * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20889         * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20890         (DOT_SYMBOLS): Likewise.
20891         (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20892         (RELOCATABLE_NEEDS_FIXUP): Likewise.
20893         (RS6000_ABI_NAME): Likewise.
20894         (TARGET_CMODEL): Likewise.
20895         (TOC_SECTION_ASM_OP): Likewise.
20896         (SET_CMODEL): New macro.
20897         (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
20899 2017-08-22  Richard Biener  <rguenther@suse.de>
20901         * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
20902         to allow for free-lang-data replacements similar to verify_type_variant.
20904 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20905             Alan Hayward  <alan.hayward@arm.com>
20906             David Sherwood  <david.sherwood@arm.com>
20908         * config/aarch64/aarch64.md (casesi): Use DImode rather than
20909         VOIDmode for the LABEL_REF.
20911 2017-08-22  Richard Biener  <rguenther@suse.de>
20913         * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
20915 2017-08-22  Richard Biener  <rguenther@suse.de>
20917         * common.opt (feliminate-dwarf2-dups): Ignore.
20918         * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
20919         * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
20920         same_die_p_wrap, compute_section_prefix,
20921         is_symbol_die, assign_symbol_names, break_out_includes): Remove.
20922         (comdat_symbol_id, comdat_symbol_number): Likewise.
20923         (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
20924         Likewise.
20925         (check_duplicate_cu, record_comdat_symbol_number): Likewise.
20926         (output_die): Mark unreachable path unreachable.
20927         (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
20928         (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
20929         (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
20930         (dwarf2out_early_finish): Likewise.
20932 2017-08-22  Aldy Hernandez  <aldyh@redhat.com>
20934         * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
20936 2017-08-22  Georg-Johann Lay  <avr@gjlay.de>
20938         PR target/81910
20939         * config/avr/avr.c (avr_handle_addr_attribute): Early return if
20940         not VAR_P. Filter attribute warnings with OPT_Wattributes.
20941         (avr_attribute_table) <io, io_low, address>: Initialize
20942         .decl_required with true.
20944 2017-08-21  Michael Meissner  <meissner@linux.vnet.ibm.com>
20946         * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
20947         undocumented debugging options.
20948         (-mvsx-scalar-double): Likewise.
20949         (-mallow-df-permute): Likewise.
20950         (-mvectorize-builtins): Likewise.
20951         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20952         (rs6000_builtin_vectorized_function): Likewise.
20953         (rs6000_builtin_md_vectorized_function): Likewise.
20954         (rs6000_opt_vars): Likewise.
20956 2017-08-21  Uros Bizjak  <ubizjak@gmail.com>
20958         PR target/46091
20959         * config/i386/i386.md (*btsq_imm): Rename from *btsq.
20960         (*btrq_imm): Rename from *btrq.
20961         (*btcq_imm): Rename from *btcq.
20962         (btsc): New code attribute.
20963         (*<btsc><mode>): New insn pattern.
20964         (*btr<mode>): Ditto.
20965         (*<btsc><mode>_mask): New insn_and_split pattern.
20966         (*btr<mode>_mask): Ditto.
20968 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20969             Alan Hayward  <alan.hayward@arm.com>
20970             David Sherwood  <david.sherwood@arm.com>
20972         * function.c (pad_below): Simplify padding calculation.
20974 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20975             Alan Hayward  <alan.hayward@arm.com>
20976             David Sherwood  <david.sherwood@arm.com>
20978         * target.def (function_prologue): Remove frame size argument.
20979         (function_epilogue): Likewise.
20980         * doc/tm.texi: Regenerate.
20981         * final.c (final_start_function): Update call to function_prologue.
20982         (final_end_function): Update call to function_epilogue.
20983         (default_function_pro_epilogue): Remove frame size argument.
20984         * output.h (default_function_pro_epilogue): Likewise.
20985         * config/arm/arm.c (arm_output_function_epilogue): Likewise.
20986         (arm_output_function_prologue): Likewise.
20987         * config/frv/frv.c (frv_function_prologue): Likewise.
20988         (frv_function_epilogue): Likewise.
20989         * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
20990         * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
20991         (ia64_output_function_epilogue): Likewise.
20992         * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
20993         (m32r_output_function_epilogue): Likewise.
20994         * config/microblaze/microblaze.c (microblaze_function_prologue)
20995         (microblaze_function_epilogue): Likewise.
20996         * config/mips/mips.c (mips_output_function_prologue): Likewise.
20997         (mips_output_function_epilogue): Likewise.
20998         * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
20999         (mmix_target_asm_function_epilogue): Likewise.
21000         * config/msp430/msp430.c (msp430_start_function): Likewise.
21001         * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
21002         (nds32_asm_function_epilogue): Likewise.
21003         * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
21004         * config/pa/pa.c (pa_output_function_prologue): Likewise.
21005         (pa_output_function_epilogue): Likewise.
21006         * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
21007         (rs6000_output_function_epilogue): Likewise.
21008         * config/rl78/rl78.c (rl78_start_function): Likewise.
21009         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
21010         (rs6000_output_function_epilogue): Likewise.
21011         * config/rx/rx.c (rx_output_function_prologue): Likewise.
21012         * config/sh/sh.c (sh_output_function_epilogue): Likewise.
21013         * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
21014         (sparc_asm_function_epilogue): Likewise.
21016 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21018         * tree.h (type_has_mode_precision_p): New function.
21019         * convert.c (convert_to_integer_1): Use it.
21020         * expr.c (expand_expr_real_2): Likewise.
21021         (expand_expr_real_1): Likewise.
21022         * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
21023         * match.pd: Likewise.
21024         * tree-ssa-forwprop.c (simplify_rotate): Likewise.
21025         * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
21026         * tree-tailcall.c (process_assignment): Likewise.
21027         * tree-vect-loop.c (vectorizable_reduction): Likewise.
21028         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
21029         (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
21030         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
21031         (vectorizable_assignment): Likewise.
21032         (vectorizable_shift): Likewise.
21033         (vectorizable_operation): Likewise.
21034         * tree-vrp.c (register_edge_assert_for_2): Likewise.
21036 2017-08-21  Wilco Dijkstra  <wdijkstr@arm.com>
21038         * match.pd: Add pow (C, x) simplification.
21040 2017-08-21  Richard Biener  <rguenther@suse.de>
21042         PR tree-optimization/81900
21043         * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
21044         for blocks with abnormal predecessors.
21045         (compute_antic): Do not set visited flag prematurely.
21047 2017-08-21  Georg-Johann Lay  <avr@gjlay.de>
21049         PR target/79883
21050         * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
21052 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21054         * stor-layout.h (vector_type_mode): Move to...
21055         * tree.h (vector_type_mode): ...here.
21056         * stor-layout.c (vector_type_mode): Move to...
21057         * tree.c (vector_type_mode): ...here.  Include rtl.h and regs.h.
21059 2017-08-21  Richard Biener  <rguenther@suse.de>
21061         * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
21062         register_external_die hooks.
21063         (debug_false_tree_charstarstar_uhwistar): Declare.
21064         (debug_nothing_tree_charstar_uhwi): Likewise.
21065         * debug.c (do_nothing_debug_hooks): Adjust.
21066         (debug_false_tree_charstarstar_uhwistar): New do nothing.
21067         (debug_nothing_tree_charstar_uhwi): Likewise.
21068         * dbxout.c (dbx_debug_hooks): Adjust.
21069         (xcoff_debug_hooks): Likewise.
21070         * sdbout.c (sdb_debug_hooks): Likewise.
21071         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
21072         * dwarf2out.c (macinfo_label_base): New global.
21073         (dwarf2out_register_external_die): New function for the
21074         register_external_die hook.
21075         (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
21076         (dwarf2_debug_hooks): Use them.
21077         (dwarf2_lineno_debug_hooks): Adjust.
21078         (struct die_struct): Add with_offset flag.
21079         (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
21080         DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
21081         DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
21082         DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
21083         DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
21084         DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
21085         defining section names for the early LTO debug variants.
21086         (reset_indirect_string): New helper.
21087         (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
21088         (print_dw_val): Add support for offsetted symbol references.
21089         (get_ultimate_context): Split out from is_cxx.
21090         (is_cxx): Use get_ultimate_context.
21091         (is_fortran): Add decl overload.
21092         (compute_comp_unit_symbol): Split out worker from
21093         compute_section_prefix.
21094         (compute_section_prefix): Call compute_comp_unit_symbol and
21095         set comdat_type_p here.
21096         (output_die): Skip DIE symbol output for the LTO added one.
21097         Handle DIE symbol references with offset.
21098         (output_comp_unit): Guard section name mangling properly.
21099         For LTO debug sections emit a symbol at the section beginning
21100         which we use to refer to its DIEs.
21101         (add_abstract_origin_attribute): For DIEs registered via
21102         dwarf2out_register_external_die directly refer to the early
21103         DIE rather than indirectly through the shadow one we created.
21104         Remove obsolete call to dwarf2out_abstract_function for
21105         non-function/block origins.
21106         (gen_array_type_die): When generating early LTO debug do
21107         not emit DW_AT_string_length.
21108         (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
21109         late when in LTO.  As suggested place a gcc_unreachable for
21110         the DECL_ABSTRACT_P case.
21111         (gen_subprogram_die): Avoid another specification DIE
21112         for early built declarations/definitions for the late LTO case.
21113         (gen_variable_die): Add type references for late duplicated VLA dies
21114         when in late LTO.
21115         (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
21116         we have the abstract instance already.
21117         (process_scope_var): Adjust decl DIE contexts in LTO which
21118         first puts them in limbo.
21119         (gen_decl_die): Do not generate type DIEs late apart from
21120         types for VLAs or for decls we do not yet have a DIE.  Do not
21121         call dwarf2out_abstract_function late.
21122         (dwarf2out_early_global_decl): Make sure to create DIEs
21123         for abstract instances of a decl first.
21124         (dwarf2out_late_global_decl): Adjust comment.
21125         (output_macinfo_op): With multiple macro sections use
21126         macinfo_label_base to distinguish labels.
21127         (output_macinfo): Likewise.  Update macinfo_label_base.
21128         Pass in the line info label.
21129         (note_variable_value_in_expr): When generating LTO resolve
21130         all variable values here by generating DIEs as needed.
21131         (init_sections_and_labels): Add early LTO debug flag parameter
21132         and generate different sections and names if set.  Add generation
21133         counter for the labels so we can have multiple of them.
21134         (reset_dies): Helper to allow DIEs to be output multiple times.
21135         (dwarf2out_finish): When outputting DIEs to the fat part of an
21136         LTO object first reset DIEs.
21137         (dwarf2out_early_finish): Output early DIEs when generating LTO.
21138         (modified_type_die): Check for decl_ultimate_origin being self
21139         before recursing.
21140         (gen_type_die_with_usage): Likewise.
21141         (gen_typedef_die): Allow decl_ultimate_origin being self.
21142         (set_decl_abstract_flags): Remove.
21143         (set_block_abstract_flags): Likewise.
21144         (dwarf2out_abstract_function): Treat the early generated DIEs
21145         as the abstract copy and only add DW_AT_inline and
21146         DW_AT_artificial here and call set_decl_origin_self.
21147         If the DIE has an abstract origin don't do anything.
21148         * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
21149         if we have none yet (Go fails to build one, PR78628).
21150         (variably_modified_type_p): Prevent endless recursion for Ada
21151         cyclic pointer types.
21152         * lto-streamer-in.c: Include debug.h.
21153         (dref_queue): New global.
21154         (lto_read_tree_1): Stream in DIE references.
21155         (lto_input_tree): Register DIE references.
21156         (input_function): Stream DECL_DEBUG_ARGS.
21157         * lto-streamer-out.c: Include debug.h.
21158         (lto_write_tree_1): Output DIE references.
21159         (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
21160         Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
21161         (output_function): Stream DECL_DEBUG_ARGS.
21162         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21163         Stream DECL_ABSTRACT_ORIGIN.
21164         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21165         (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
21166         DECL_CONTEXT for file-scope decls.
21167         * lto-streamer.h (struct dref_entry): Declare.
21168         (dref_queue): Likewise.
21169         * cfgexpand.c (pass_expand::execute): Do not call the
21170         outlining_inline_function hook here.
21171         * lto-wrapper.c (debug_obj): New global.
21172         (tool_cleanup): Unlink it if required.
21173         (debug_objcopy): New function.
21174         (run_gcc): Handle early debug sections in the IL files by
21175         extracting them to separate files, partially linkin them and
21176         feeding the result back as result to the linker.
21177         * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
21178         DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
21179         DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
21180         sections into a separate segment.
21181         * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
21182         segments.
21183         (darwin_asm_dwarf_section): Likewise.
21184         (darwin_asm_output_dwarf_offset): Likewise.
21185         * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
21187 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21188             Alan Hayward  <alan.hayward@arm.com>
21189             David Sherwood  <david.sherwood@arm.com>
21191         * read-md.h (md_reader::record_potential_iterator_use): Replace
21192         pointer argument with an rtx and an index.
21193         * read-rtl.c (iterator_group::apply_iterator): Likewise.
21194         (apply_mode_iterator): Likewise.
21195         (apply_code_iterator): Likewise.
21196         (apply_int_iterator): Likewise.
21197         (apply_subst_iterator): Likewise.
21198         (record_iterator_use): Likewise.
21199         (record_attribute_use): Likewise.
21200         (md_reader::record_potential_iterator_use): Likewise.  Update calls
21201         to record_iterator_use and apply_iterator.
21202         (iterator_use): Replace ptr with x and index.
21203         (attribute_use): Likewise.
21204         (apply_attribute_uses): Update calls to apply_iterator.
21205         (apply_iterators): Likewise.  Update initialization of iterator_use.
21206         (rtx_reader::read_rtx_code): Update calls to record_iterator_use
21207         and record_potential_iterator_use.
21208         (rtx_reader::read_rtx_operand): Likewise.
21210 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21211             Alan Hayward  <alan.hayward@arm.com>
21212             David Sherwood  <david.sherwood@arm.com>
21214         * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
21215         CONST_WIDE_INT.
21217 2017-08-21  Richard Biener  <rguenther@suse.de>
21219         PR middle-end/81884
21220         * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
21221         at struct end conservatively when comparing common bases.
21223 2017-08-21  Richard Biener  <rguenther@suse.de>
21225         * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21226         (mem_ref_in_stmt): Remove.
21227         (determine_max_movement): Use ref index to get at the reference.
21228         (invariantness_dom_walker::before_dom_children): Deal with
21229         lim data already initialized.
21230         (gather_mem_refs_stmt): Initialize lim data and record ref index.
21232 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21234         * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21235         (TARGET_ISA_ROUND): Ditto.
21236         (TARGET_ROUND): Ditto.
21237         * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21238         * config/i386/i386.md: Ditto.
21239         * config/i386/sse.md: Ditto.
21240         * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21241         with OPTION_MASK_ISA_SSE4_1.
21243 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21245         PR target/81894
21246         * doc/extend.texi (x86 Built-in Functions): Correct the name of
21247         __builtin_ia32_lzcnt_u16.
21249 2017-08-18  Peter Bergner  <bergner@vnet.ibm.com>
21251         PR target/80210
21252         * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21253         (rs6000_set_current_function): Rewrite function to use it.
21255 2017-08-18  H.J. Lu  <hongjiu.lu@intel.com>
21257         PR c/53037
21258         * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21259         and TYPE_WARN_IF_NOT_ALIGN.
21260         * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21261         (handle_warn_if_not_align): New.
21262         (place_union_field): Call handle_warn_if_not_align.
21263         (place_field): Call handle_warn_if_not_align.
21264         Copy TYPE_WARN_IF_NOT_ALIGN.
21265         (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21266         (layout_type): Likewise.
21267         * tree-core.h (tree_type_common): Add warn_if_not_align.  Set
21268         spare to 18.
21269         (tree_decl_common): Add warn_if_not_align.
21270         * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21271         * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21272         (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21273         (DECL_WARN_IF_NOT_ALIGN): Likewise.
21274         (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21275         * doc/extend.texi: Document warn_if_not_aligned attribute.
21276         * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21278 2017-08-17  Martin Liska  <mliska@suse.cz>
21280         PR bootstrap/81864
21281         * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21282         (get_data_dependence): Use it as pointer type.
21283         (distribute_loop): Likewise.
21285 2017-08-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21287         * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21288         (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21289         (p8_vmrgow_<mode>_direct): New define_insn.
21290         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21291         handle endianness for vmrgew and vmrgow permute patterns.
21293 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21295         * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21296         * config/rs6000/rs6000-cpus.def: Remove comment.
21297         (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21298         (POWERPC_MASKS): Likewise.
21299         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21300         use of TARGET_VSX_TIMODE.
21301         (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21302         (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21303         TARGET_VSX_TIMODE.  Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21304         (rs6000_option_override_internal): Remove dead code.
21305         (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21306         (rs6000_legitimize_reload_address): Likewise.
21307         (rs6000_legitimate_address_p): Likewise.
21308         (rs6000_opt_masks): Delete "vsx-timode".
21309         (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21310         from function comment.
21311         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21312         * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21313         (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21314         condition.
21315         * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21316         * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21317         * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21318         (VSX_TI): Likewise.
21319         (VSX_M): Likewise.
21320         (define_peephole2): Likewise.
21322 2017-08-17  Martin Sebor  <msebor@redhat.com>
21324         PR c/81859
21325         * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21326         past the end of an array.
21327         (test_pp_format): Add test cases.
21329 2017-08-17  Richard Sandiford  <richard.sandiford@linaro.org>
21331         * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21332         missing ECF_NOTHROW flags.
21334 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21336         PR target/72804
21337         * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21338         operands residing in integer registers.
21339         (*vsx_le_perm_load_<mode>): Likewise.
21340         (*vsx_le_perm_store_<mode>): Likewise.
21341         (define_peephole2): Add peepholes to optimize the above.
21343 2017-08-17  Marek Polacek  <polacek@redhat.com>
21345         PR middle-end/81814
21346         * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21347         to mimic what shorten_compare did.  Change the return type to bool.
21348         (fold_cond_expr_with_comparison): Update call to
21349         operand_equal_for_comparison_p.
21350         (fold_ternary_loc): Likewise.
21352 2017-08-17  Jackson Woodruff  <jackson.woodruff@arm.com>
21354         * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21355         register.
21356         (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21358 2017-08-17  Richard Biener  <rguenther@suse.de>
21360         * tree-ssa-structalias.c (solve_graph): When propagating
21361         to successors update the graphs succ edges and avoid duplicate work.
21363 2017-08-17  Maxim Ostapenko  <m.ostapenko@samsung.com>
21365         PR target/81861
21366         * config/i386/i386.c (ix86_option_override_internal): Save target
21367         specific options after ix86_stack_protector_guard_reg was changed.
21369 2017-08-17  Richard Biener  <rguenther@suse.de>
21371         PR tree-optimization/81827
21372         * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21373         (new_var_info): Initialize it conservatively.
21374         (get_call_vi): Mark register vars.
21375         (new_scalar_tmp_constraint_exp): Likewise.
21376         (handle_rhs_call): Likewise.
21377         (handle_const_call): Likewise.
21378         (create_function_info_for): Likewise.
21379         (solve_constraints): Sort varinfos to separate register from
21380         non-register vars to pack points-to solution bitmaps during
21381         iteration.
21383 2017-08-17  Marek Polacek  <polacek@redhat.com>
21385         * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21387 2017-08-17  Richard Biener  <rguenther@suse.de>
21389         * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21390         to true when overflow is undefined and we saturated the result.
21392 2017-08-17  Alan Modra  <amodra@gmail.com>
21394         PR target/80938
21395         * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21396         Don't use store multiple if only one reg needs saving.
21397         (interesting_frame_related_regno): New function.
21398         (rs6000_frame_related): Don't emit frame info for regs that
21399         don't need saving.
21400         (rs6000_emit_epilogue): Likewise.
21402 2017-08-16  Nathan Sidwell  <nathan@acm.org>
21404         * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21405         * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21406         (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21407         * tree.c (free_lang_data_in_type): Use else-if chain.  Always
21408         clear TYPE_LANG_1.  Remove obsolete member-function stripping.
21409         (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21410         (verify_type): Adjust for TYPE_BINFO move.
21411         * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21412         process TYPE_BINFO directly.
21413         (hash_tree): Likewise.
21414         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21415         Likewise.
21416         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21417         Likewise.
21419 2017-08-16  David Malcolm  <dmalcolm@redhat.com>
21421         * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21423 2017-08-16  Uros Bizjak  <ubizjak@gmail.com>
21425         PR target/46091
21426         * config/i386/i386.md (*anddi_1_btr): Change predicates of
21427         operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21428         Add ix86_binary_operator_ok to insn constraint.
21429         (*iordi_1_bts): Ditto.
21430         (*xordi_1_btc): Ditto.
21431         (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21432         Update corresponding peephole2 pattern.
21433         (*btrq): Ditto.
21434         (*btcq): Ditto.
21436 2017-08-16  Bin Cheng  <bin.cheng@arm.com>
21438         PR tree-optimization/81832
21439         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21440         copy loop header which has IFN_LOOP_DIST_ALIAS call.
21442 2017-08-16  Marek Polacek  <polacek@redhat.com>
21444         PR middle/81695
21445         * fold-const.c (fold_indirect_ref_1): Restore original behavior
21446         regarding size_zero_node.
21448 2017-08-16  Martin Liska  <mliska@suse.cz>
21450         PR target/81753
21451         * config.gcc: Respect previously set extra_objs in case
21452         of darwin target.
21454 2017-08-16  Richard Sandiford  <richard.sandiford@linaro.org>
21456         PR tree-optimization/81835
21457         * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21458         the phi SSA_NAME.  Check that the condition in a COND_EXPR does
21459         not depend on the phi.
21461 2017-08-16  Alan Modra  <amodra@gmail.com>
21463         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21464         dead code.
21466 2017-08-16  Alan Modra  <amodra@gmail.com>
21468         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21469         (save_reg_p): ..into this.  Update all callers.
21470         (first_reg_to_save): Simplify.
21472 2017-08-16  Alan Modra  <amodra@gmail.com>
21474         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21475         fixed regs.
21477 2017-08-15  Joseph Myers  <joseph@codesourcery.com>
21479         PR target/78460
21480         PR target/67712
21481         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21482         constant count if that count is less than 32.
21484 2017-08-15  Nathan Sidwell  <nathan@acm.org>
21486         * gcc.c (execute): Emit friendlier message if inferior is killed
21487         by an external cause.
21489 2017-08-15  Richard Biener  <rguenther@suse.de>
21491         PR tree-optimization/81790
21492         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21493         CONSTRUCTORs from simplifying and VN.
21495 2017-08-14  Martin Sebor  <msebor@redhat.com>
21497         * builtin-attrs.def: Add comments.
21499 2017-08-14  Martin Sebor  <msebor@redhat.com>
21501         PR c/81117
21502         * doc/extend.texi (attribute nonstring): Document new attribute.
21504 2017-08-14  Martin Sebor  <msebor@redhat.com>
21506         PR c/81117
21507         * tree-diagnostic.c (default_tree_printer): Handle %G.
21508         * gimple-pretty-print.h (percent_G_format): Declare new function.
21509         * gimple-pretty-print.c (percent_G_format): Define.
21510         * tree-pretty-print.c (percent_K_format): Add argument.
21512 2017-08-14  Martin Sebor  <msebor@redhat.com>
21514         PR translation/79998
21515         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21516         Remove a stray space.
21518 2017-08-14  Uros Bizjak  <ubizjak@gmail.com>
21520         PR target/46091
21521         * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21522         (*iordi_1_bts): Ditto.
21523         (*xordi_1_btc): Ditto.
21525 2017-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21527         PR target/79845
21528         * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21529         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21530         Likewise.
21531         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21532         quoted strings, and make more translator-friendly.
21533         (darwin_rs6000_override_options): Likewise.
21534         (rs6000_option_override_internal): Likewise.
21535         (rs6000_return_in_memory): Fix overlong line.
21536         (init_cmulative_args): Use quoted strings, and make more
21537         translator-friendly.
21538         (rs6000_pass_by_reference): Fix overlong line.
21539         (def_builtin): Use quoted strings.
21540         (altivec_expand_predicate_builtin): Use quoted strings, and make
21541         more translator-friendly.
21542         (htm_expand_builtin): Use quoted strings.
21543         (cpu_expand_builtin): Use quoted strings, and make more
21544         translator-friendly.
21545         (altivec_expand_builtin): Likewise.
21546         (paired_expand_predicate_builtin): Likewise.
21547         (rs6000_invalid_builtin): Likewise.
21548         (builtin_function_type): Use quoted strings.
21549         (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21550         more translator-friendly.
21551         (rs6000_trampoline_init): Likewise.
21552         (rs6000_handle_altivec_attribute): Likewise.
21553         (rs6000_inner_target_options): Use quoted strings.
21554         (rs6000_disable_incompatible_switches): Likewise.
21555         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21556         strings, and make more translator-friendly.
21557         (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21559 2017-08-14  Bin Cheng  <bin.cheng@arm.com>
21561         PR tree-optimization/81799
21562         * tree-loop-distribution.c (version_loop_by_alias_check): Force
21563         cond_expr to simple gimple operand.
21565 2017-08-14  Wilco Dijkstra  <wdijkstr@arm.com>
21567         PR middle-end/46932
21568         * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21570 2017-08-14  Georg-Johann Lay  <avr@gjlay.de>
21572         PR target/81754
21573         PR target/81268
21574         * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21575         * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21576         TARGET_GASISR_PROLOGUES.
21577         * config/avr/avr.c (avr_option_override): Same.
21578         (avr_pass_pre_proep::execute): Same.
21580 2017-08-13  H.J. Lu  <hongjiu.lu@intel.com>
21582         PR target/81820
21583         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21584         frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21586 2017-08-13  Uros Bizjak  <ubizjak@gmail.com>
21588         * config/i386/i386.md (*load_tp_<mode>): Redefine as
21589         define_insn_and_split.  Split to a memory load from 0 in
21590         DEFAULT_TLS_SEG_REG address space.  Merge with *load_tp_x32
21591         using PTR mode iterator.
21592         (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21593         Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21594         (*add_tp_<mode>): Redefine as define_insn_and_split.
21595         Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21596         address space.  Merge with *add_tp_x32 using PTR mode iterator.
21597         (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21598         Split to an add with a  memory load from 0 in
21599         DEFAULT_TLS_SEG_REG address space.
21601 2017-08-12  Andrew Pinski  <apinski@cavium.com>
21603         * config/aarch64/aarch64-option-extensions.def (rdma):
21604         Fix feature string to what Linux prints out in /proc/cpuinfo.
21606 2017-08-12  Pierre-Marie de Rodat  <derodat@adacore.com>
21608         PR ada/79542
21609         * dwarf2out.c (modified_type_die): For C typedef types that have
21610         an ultimate origin, process the ultimate origin instead of the
21611         input type.
21612         (gen_typedef_die): Assert that input DECLs have no ultimate
21613         origin.
21614         (gen_type_die_with_usage): For typedef variants that have an
21615         ultimate origin, just call gen_decl_die on the original DECL.
21616         (process_scope_var): Avoid creating DIEs for local typedefs and
21617         concrete static variables.
21619 2017-08-12  Alan Modra  <amodra@gmail.com>
21621         PR target/81170
21622         PR target/81295
21623         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21624         match gnu-user.h startfile.
21625         (ENDFILE_LINUX_SPEC): Similarly.
21627 2017-08-11  Thomas Schwinge  <thomas@codesourcery.com>
21629         PR lto/81430
21630         * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21631         Remove function.
21632         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21634 2017-08-11  Tamar Christina  <tamar.christina@arm.com>
21635         * config/aarch64/aarch64.md (mov<mode>): Change.
21636         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21637         aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21638         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21640 2017-08-11  Eric Botcazou  <ebotcazou@adacore.com>
21642         * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21643         for storage order barriers.
21645 2017-08-11  Martin Liska  <mliska@suse.cz>
21647         PR tree-opt/79987
21648         * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21649         variables of void type.
21651 2017-08-11  Martin Liska  <mliska@suse.cz>
21653         * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21654         TARGET_SUPPORTS_ALIASES.
21655         * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21656         * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21657         (optimize_weakref): Likewise.
21658         * symtab.c (symtab_node::noninterposable_alias): Likewise.
21659         * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21660         * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21662 2017-08-11  Martin Liska  <mliska@suse.cz>
21664         PR ipa/81213
21665         * config/i386/i386.c (make_resolver_func): Do complete
21666         refactoring of the function.
21668 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21670         PR target/81708
21671         * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21672         * config/i386/i386.c (ix86_stack_protect_guard): Use
21673         ix86_stack_protect_guard_symbol_str to generate varible declaration.
21674         * doc/invoke.texi (x86 Options): Document
21675         -mstack-protector-guard-symbol= option.
21677 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21679         * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21680         * config/i386/i386.c (ix86_split_stack_guard): New function.
21681         (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21682         (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21683         (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21684         (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21685         * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21686         (split_stack_space_check): Call ix86_split_stack_guard.
21688 2017-08-10  Martin Sebor  <msebor@redhat.com>
21690         * print-tree.c (print_node): Print location using the established
21691         format %s:%i%i.
21692         Replace spaces with colons.
21693         (debug_raw, debug): Ditto.
21695 2017-08-10  Martin Sebor  <msebor@redhat.com>
21697         PR c++/81586
21698         * pretty-print.c (pp_format): Correct the handling of %s precision.
21700 2017-08-10  H.J. Lu  <hongjiu.lu@intel.com>
21702         PR target/81736
21703         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21704         to ...
21705         (ix86_finalize_stack_frame_flags): This.  Also clear
21706         frame_pointer_needed if -fno-omit-frame-pointer is used without
21707         stack access.
21708         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21709         with ix86_finalize_stack_frame_flags.
21710         (ix86_expand_epilogue): Likewise.
21711         (ix86_expand_split_stack_prologue): Likewise.
21712         * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21714 2017-08-10  Martin Liska  <mliska@suse.cz>
21716         PR c++/81355
21717         * c-attribs.c (handle_target_attribute):
21718         Report warning for an empty string argument of target attribute.
21720 2017-08-09  Jakub Jelinek  <jakub@redhat.com>
21722         PR c/81687
21723         * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21724         LABEL_DECLs.
21725         * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21726         or DECL_NONLOCAL labels.
21727         (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21728         or DECL_NONLOCAL labels here.
21730 2017-08-09  Will Schmidt  <will_schmidt@vnet.ibm.com>
21732         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21733         to indicate when early gimple folding has been disabled.
21734         (rs6000_gimple_fold_builtin): Add debug content.
21735         (rs6000_invalid_builtin): Fix whitespace.
21736         (rs6000_expand_builtin): Fix whitespace.
21737         * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21739 2017-08-09  Segher Boessenkool  <segher@kernel.crashing.org>
21741         PR target/80938
21742         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21743         SAVE_MULTIPLE if not all the registers that saves, should be saved.
21745 2017-08-09  Jim Wilson  <jim.wilson@linaro.org>
21747         * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21748         (qdf24xx): Likewise.
21749         * config/aarch64/aarch64.md: Include falkor.md.
21750         * config/aarch64/falkor.md: New.
21752 2017-08-09  Marek Polacek  <polacek@redhat.com>
21754         PR c/81233
21755         * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21756         * diagnostic.c (emit_diagnostic): Add a comment.
21757         (emit_diagnostic_valist): New function.
21759 2017-08-09  Marek Polacek  <polacek@redhat.com>
21761         PR c/81417
21762         * input.c (make_location): New overload.
21763         * input.h (make_location): Declare.
21765 2017-08-08  Alan Modra  <amodra@gmail.com>
21766             H.J. Lu  <hongjiu.lu@intel.com>
21768         PR driver/81523
21769         * gcc.c (NO_PIE_SPEC): Delete.
21770         (PIE_SPEC): Define as !no-pie/pie.  Move static|shared|r
21771         exclusion..
21772         (LINK_PIE_SPEC): ..to here.
21773         (LINK_COMMAND_SPEC): Support -no-pie.
21774         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21775         chain of crtbegin*.o selection, update for PIE_SPEC changes and
21776         format.
21777         (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21778         * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21779         (ENDFILE_CRTEND_SPEC): Similarly.
21781 2017-08-08  Uros Bizjak  <ubizjak@gmail.com>
21783         PR target/81708
21784         * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21785         (mstack-protector-guard-offset=): Ditto.
21786         * config/i386/i386.c (ix86_option_override): Handle
21787         -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21788         options.
21789         (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21790         ix86_stack_protect_guard_offset variables.
21791         (TARGET_STACK_PROTECT_GUARD): Always define.
21792         * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21793         and -mstack-protector-guard-offset= options.
21795 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21797         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21798         boundary case for the last candidate.
21800 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21802         * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21803         * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21805 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21807         PR middle-end/19706
21808         * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21809         * config/aarch64/aarch64-builtins.c
21810         (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21811         * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21812         * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21814 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21815             Andrew Pinski <pinskia@gmail.com>
21817         PR middle-end/19706
21818         * internal-fn.def (XORSIGN): New.
21819         * optabs.def (xorsign_optab): New.
21820         * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21821         (convert_expand_mult_copysign): New.
21822         (pass_optimize_widening_mul::execute): Call
21823         convert_expand_mult_copysign.
21825 2017-08-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21827         PR tree-optimization/81354
21828         * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21829         Insert on edges rather than explicitly creating landing pads.
21830         (analyze_candidates_and_replace): Commit edge inserts.
21832 2017-08-08  Richard Biener  <rguenther@suse.de>
21834         PR middle-end/81719
21835         * tree-ssa-loop-niter.c: Include tree-dfa.h.
21836         (expand_simple_operations): Also look through ADDR_EXPRs with
21837         MEM_REF bases treating them as POINTER_PLUS_EXPR.
21839 2017-08-08  Richard Biener  <rguenther@suse.de>
21841         PR tree-optimization/81723
21842         * tree-vect-slp.c (struct bst_traits): New hash traits.
21843         (bst_fail): New global.
21844         (vect_build_slp_tree_2): New worker, split out from ...
21845         (vect_build_slp_tree): ... this now wrapping it with using
21846         bst_fail set to cache SLP tree build fails.  Properly handle
21847         max_tree_size.
21848         (vect_analyze_slp_instance): Allocate and free bst_fail.
21850 2017-08-08  Martin Liska  <mliska@suse.cz>
21852         PR tree-opt/81696
21853         * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21854         LABEL_DECLs that can be from a different function.
21856 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21858         PR tree-optimization/81744
21859         * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21860         loop's number of iterations.
21862 2017-08-08  Martin Liska  <mliska@suse.cz>
21864         * asan.c: Include header files.
21865         * attribs.c (build_decl_attribute_variant): New function moved
21866         from tree.[ch].
21867         (build_type_attribute_qual_variant): Likewise.
21868         (cmp_attrib_identifiers): Likewise.
21869         (simple_cst_list_equal): Likewise.
21870         (omp_declare_simd_clauses_equal): Likewise.
21871         (attribute_value_equal): Likewise.
21872         (comp_type_attributes): Likewise.
21873         (build_type_attribute_variant): Likewise.
21874         (lookup_ident_attribute): Likewise.
21875         (remove_attribute): Likewise.
21876         (merge_attributes): Likewise.
21877         (merge_type_attributes): Likewise.
21878         (merge_decl_attributes): Likewise.
21879         (merge_dllimport_decl_attributes): Likewise.
21880         (handle_dll_attribute): Likewise.
21881         (attribute_list_equal): Likewise.
21882         (attribute_list_contained): Likewise.
21883         * attribs.h (lookup_attribute): New function moved from tree.[ch].
21884         (lookup_attribute_by_prefix): Likewise.
21885         * bb-reorder.c: Include header files.
21886         * builtins.c: Likewise.
21887         * calls.c: Likewise.
21888         * cfgexpand.c: Likewise.
21889         * cgraph.c: Likewise.
21890         * cgraphunit.c: Likewise.
21891         * convert.c: Likewise.
21892         * dwarf2out.c: Likewise.
21893         * final.c: Likewise.
21894         * fold-const.c: Likewise.
21895         * function.c: Likewise.
21896         * gimple-expr.c: Likewise.
21897         * gimple-fold.c: Likewise.
21898         * gimple-pretty-print.c: Likewise.
21899         * gimple.c: Likewise.
21900         * gimplify.c: Likewise.
21901         * hsa-common.c: Likewise.
21902         * hsa-gen.c: Likewise.
21903         * internal-fn.c: Likewise.
21904         * ipa-chkp.c: Likewise.
21905         * ipa-cp.c: Likewise.
21906         * ipa-devirt.c: Likewise.
21907         * ipa-fnsummary.c: Likewise.
21908         * ipa-inline.c: Likewise.
21909         * ipa-visibility.c: Likewise.
21910         * ipa.c: Likewise.
21911         * lto-cgraph.c: Likewise.
21912         * omp-expand.c: Likewise.
21913         * omp-general.c: Likewise.
21914         * omp-low.c: Likewise.
21915         * omp-offload.c: Likewise.
21916         * omp-simd-clone.c: Likewise.
21917         * opts-global.c: Likewise.
21918         * passes.c: Likewise.
21919         * predict.c: Likewise.
21920         * sancov.c: Likewise.
21921         * sanopt.c: Likewise.
21922         * symtab.c: Likewise.
21923         * toplev.c: Likewise.
21924         * trans-mem.c: Likewise.
21925         * tree-chkp.c: Likewise.
21926         * tree-eh.c: Likewise.
21927         * tree-into-ssa.c: Likewise.
21928         * tree-object-size.c: Likewise.
21929         * tree-parloops.c: Likewise.
21930         * tree-profile.c: Likewise.
21931         * tree-ssa-ccp.c: Likewise.
21932         * tree-ssa-live.c: Likewise.
21933         * tree-ssa-loop.c: Likewise.
21934         * tree-ssa-sccvn.c: Likewise.
21935         * tree-ssa-structalias.c: Likewise.
21936         * tree-ssa.c: Likewise.
21937         * tree-streamer-in.c: Likewise.
21938         * tree-vectorizer.c: Likewise.
21939         * tree-vrp.c: Likewise.
21940         * tsan.c: Likewise.
21941         * ubsan.c: Likewise.
21942         * varasm.c: Likewise.
21943         * varpool.c: Likewise.
21944         * tree.c: Remove functions moved to attribs.[ch].
21945         * tree.h: Likewise.
21946         * config/aarch64/aarch64.c: Add attrs.h header file.
21947         * config/alpha/alpha.c: Likewise.
21948         * config/arc/arc.c: Likewise.
21949         * config/arm/arm.c: Likewise.
21950         * config/avr/avr.c: Likewise.
21951         * config/bfin/bfin.c: Likewise.
21952         * config/c6x/c6x.c: Likewise.
21953         * config/cr16/cr16.c: Likewise.
21954         * config/cris/cris.c: Likewise.
21955         * config/darwin.c: Likewise.
21956         * config/epiphany/epiphany.c: Likewise.
21957         * config/fr30/fr30.c: Likewise.
21958         * config/frv/frv.c: Likewise.
21959         * config/ft32/ft32.c: Likewise.
21960         * config/h8300/h8300.c: Likewise.
21961         * config/i386/winnt.c: Likewise.
21962         * config/ia64/ia64.c: Likewise.
21963         * config/iq2000/iq2000.c: Likewise.
21964         * config/lm32/lm32.c: Likewise.
21965         * config/m32c/m32c.c: Likewise.
21966         * config/m32r/m32r.c: Likewise.
21967         * config/m68k/m68k.c: Likewise.
21968         * config/mcore/mcore.c: Likewise.
21969         * config/microblaze/microblaze.c: Likewise.
21970         * config/mips/mips.c: Likewise.
21971         * config/mmix/mmix.c: Likewise.
21972         * config/mn10300/mn10300.c: Likewise.
21973         * config/moxie/moxie.c: Likewise.
21974         * config/msp430/msp430.c: Likewise.
21975         * config/nds32/nds32-isr.c: Likewise.
21976         * config/nds32/nds32.c: Likewise.
21977         * config/nios2/nios2.c: Likewise.
21978         * config/nvptx/nvptx.c: Likewise.
21979         * config/pa/pa.c: Likewise.
21980         * config/pdp11/pdp11.c: Likewise.
21981         * config/powerpcspe/powerpcspe.c: Likewise.
21982         * config/riscv/riscv.c: Likewise.
21983         * config/rl78/rl78.c: Likewise.
21984         * config/rx/rx.c: Likewise.
21985         * config/s390/s390.c: Likewise.
21986         * config/sh/sh.c: Likewise.
21987         * config/sol2.c: Likewise.
21988         * config/sparc/sparc.c: Likewise.
21989         * config/spu/spu.c: Likewise.
21990         * config/stormy16/stormy16.c: Likewise.
21991         * config/tilegx/tilegx.c: Likewise.
21992         * config/tilepro/tilepro.c: Likewise.
21993         * config/v850/v850.c: Likewise.
21994         * config/vax/vax.c: Likewise.
21995         * config/visium/visium.c: Likewise.
21996         * config/xtensa/xtensa.c: Likewise.
21998 2017-08-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
22000         PR target/81593
22001         * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
22002         constraints since the -mupper-regs-* switches have been
22003         eliminated.
22004         (vsx_concat_<mode>_1): New combiner insns to recognize inserting
22005         into a vector from a double word element that was extracted from
22006         another vector, and eliminate extra XXPERMDI instructions.
22007         (vsx_concat_<mode>_2): Likewise.
22008         (vsx_concat_<mode>_3): Likewise.
22009         (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
22010         concat to allow optimizing inserts from previous extracts.
22012 2017-08-07  Uros Bizjak  <ubizjak@gmail.com>
22014         * config/i386/i386.c (ix86_stack_protect_guard): Generate
22015         memory reference to a SSP offset in TLS address space.
22016         (ix86_print_operand) <case '@'>: Remove.
22017         (ix86_print_operand_punct_valid_p): Remove '@' code.
22018         * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
22019         UNSPEC_SP_TLS_TEST.
22020         (stack_tls_protect_set_<mode>): Remove.
22021         (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
22022         (stack_tls_protect_test_<mode>): Remove.
22023         (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
22025 2017-08-07  Olivier Hainque  <hainque@adacore.com>
22027         PR target/81755
22028         * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
22030 2017-08-07  Douglas Rupp  <rupp@adacore.com>
22032         * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
22033         variable was referenced as multidir in command.
22035 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22037         PR c/69389
22038         * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
22039         BIT_FIELD_REF.
22041 2017-08-07  Martin Liska  <mliska@suse.cz>
22043         * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
22044         * config/rl78/rl78.c: Add include of attribs.h.
22045         * config/sh/sh.c: Likewise.
22046         * config/v850/v850.c: Likewise.
22048 2017-08-07  Tom de Vries  <tom@codesourcery.com>
22050         PR middle-end/78266
22051         * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
22053 2017-08-07  Martin Liska  <mliska@suse.cz>
22055         * config/mips/mips.c: Include attribs.h.
22057 2017-08-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
22059         PR fortran/68829
22060         * doc/invoke.texi: Document change in behvaior for -Ofast for
22061         Fortran.
22063 2017-08-07  Wilco Dijkstra  <wdijkstr@arm.com>
22065         * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
22066         Use gen_frame_mem.
22067         (aarch64_pop_regs): Likewise.
22068         (aarch64_gen_load_pair): Likewise.
22069         (aarch64_save_callee_saves): Likewise.
22070         (aarch64_restore_callee_saves): Likewise.
22072 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22074         * config/i386/i386.c: Revert the last change.
22076 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22078         PR target/81736
22079         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
22080         to ...
22081         (ix86_finalize_stack_frame_flags): This.  Also clear
22082         frame_pointer_needed if -fno-omit-frame-pointer is used without
22083         stack access.
22084         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
22085         with ix86_finalize_stack_frame_flags.
22086         (ix86_expand_epilogue): Likewise.
22087         (ix86_expand_split_stack_prologue): Likewise.
22089 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22091         PR target/81743
22092         * config/i386/i386.c (get_builtin_code_for_version): Set priority
22093         to P_AES for Westmere.
22095 2017-08-07  Jonathan Yong  <10walls@gmail.com>
22097         * config/i386/mingw.opt (fset-stack-executable): Removed.
22098         * config/i386/cygming.opt (fset-stack-executable): Moved
22099         from mingw.opt.
22100         * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
22102 2017-08-07  Segher Boessenkool  <segher@kernel.crashing.org>
22104         * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
22106 2017-08-07  Marek Polacek  <polacek@redhat.com>
22108         PR middle-end/81737
22109         * fold-const.c (fold_indirect_ref_1): Check type_domain.
22111 2017-08-07  Martin Liska  <mliska@suse.cz>
22113         * attribs.h (canonicalize_attr_name): New function.
22114         (cmp_attribs): Move from c-format.c and adjusted.
22115         (is_attribute_p): Moved from tree.h.
22116         * tree-inline.c: Add new includes.
22117         * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
22118         (private_is_attribute_p): Remove.
22119         (private_lookup_attribute): Likewise.
22120         (private_lookup_attribute_by_prefix): Simplify.
22121         (remove_attribute): Use is_attribute_p.
22122         * tree.h: Remove removed declarations.
22124 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22126         PR middle-end/81698
22127         * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
22128         instead of computing it in the function.  Formatting fix.
22129         (expand_case): Don't rely on default_edge being the first edge,
22130         clear it if removing it, pass default_edge to
22131         emit_case_dispatch_table.
22132         (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
22133         fix.
22135 2017-08-06  Uros Bizjak  <ubizjak@gmail.com>
22137         * config/alpha/alpha.c (alpha_reorg): If trap is the last active
22138         insn in the function, emit NOP after the insn.
22140 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22142         * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
22143         and element loops.
22145 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22147         * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
22148         loop.
22150 2017-08-04  Yury Gribov  <tetra2005@gmail.com>
22152         PR tree-optimization/57371
22153         * match.pd: New pattern.
22155 2017-08-04  Marek Polacek  <polacek@redhat.com>
22157         PR middle-end/81695
22158         * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
22159         perform the computation in offset_int.
22161 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22163         PR tree-optimization/81136
22164         * tree-vectorizer.h: Include tree-hash-traits.h.
22165         (vec_base_alignments): New typedef.
22166         (vec_info): Add a base_alignments field.
22167         (vect_record_base_alignments): Declare.
22168         * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
22169         field.
22170         (DR_IS_CONDITIONAL_IN_STMT): New macro.
22171         (create_data_ref): Add an is_conditional_in_stmt argument.
22172         * tree-data-ref.c (create_data_ref): Likewise.  Use it to initialize
22173         the is_conditional_in_stmt field.
22174         (data_ref_loc): Add an is_conditional_in_stmt field.
22175         (get_references_in_stmt): Set the is_conditional_in_stmt field.
22176         (find_data_references_in_stmt): Update call to create_data_ref.
22177         (graphite_find_data_references_in_stmt): Likewise.
22178         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
22179         * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
22180         (vect_record_base_alignment): New function.
22181         (vect_record_base_alignments): Likewise.
22182         (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
22183         for nested statements even if we fail to compute a misalignment.
22184         Use pooled base alignments for unconditional references.
22185         (vect_find_same_alignment_drs): Compare base addresses instead
22186         of base objects.
22187         (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
22188         * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
22190 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22192         * tree-vectorizer.h (vec_info): Add a constructor and destructor.
22193         Add an explicit name for the enum.  Use auto_vec for slp_instances
22194         and grouped_stores.
22195         (_loop_vec_info): Add a constructor and destructor.  Use auto_vec
22196         for all vectors.
22197         (_bb_vec_info): Add a constructor and destructor.
22198         (vinfo_for_stmt): Return NULL for uids of -1 as well.
22199         (destroy_loop_vec_info): Delete.
22200         (vect_destroy_datarefs): Likewise.
22201         * tree-vectorizer.c (vect_destroy_datarefs): Delete.
22202         (vec_info::vec_info): New function.
22203         (vec_info::~vec_info): Likewise.
22204         (vectorize_loops): Use delete instead of destroy_loop_vec_info.
22205         * tree-parloops.c (gather_scalar_reductions): Use delete instead of
22206         destroy_loop_vec_info.
22207         * tree-vect-loop.c (new_loop_vec_info): Replace with...
22208         (_loop_vec_info::_loop_vec_info): ...this.
22209         (destroy_loop_vec_info): Replace with...
22210         (_loop_vec_info::~_loop_vec_info): ...this.  Unconditionally delete
22211         the stmt_vec_infos.  Leave handling of vec_info information to its
22212         destructor.  Remove explicit vector releases.
22213         (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
22214         (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
22215         * tree-vect-slp.c (new_bb_vec_info): Replace with...
22216         (_bb_vec_info::_bb_vec_info): ...this.  Don't reserve space in
22217         BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
22218         (destroy_bb_vec_info): Replace with...
22219         (_bb_vec_info::~_bb_vec_info): ...this.  Leave handling of vec_info
22220         information to its destructor.
22221         (vect_slp_analyze_bb_1): Use new and delete instead of
22222         new_bb_vec_info and destroy_bb_vec_info.
22223         (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22224         single delete.
22226 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22228         * tree-data-ref.h (subscript): Add access_fn field.
22229         (data_dependence_relation): Add could_be_independent_p.
22230         (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22231         (same_access_functions): Move to tree-data-ref.c.
22232         * tree-data-ref.c (ref_contains_union_access_p): New function.
22233         (access_fn_component_p): Likewise.
22234         (access_fn_components_comparable_p): Likewise.
22235         (dr_analyze_indices): Add a reference to access_fn_component_p.
22236         (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22237         DR_ACCESS_FN.
22238         (constant_access_functions): Likewise.
22239         (add_other_self_distances): Likewise.
22240         (same_access_functions): Likewise.  (Moved from tree-data-ref.h.)
22241         (initialize_data_dependence_relation): Use XCNEW and remove
22242         explicit zeroing of DDR_REVERSED_P.  Look for a subsequence
22243         of access functions that have the same type.  Allow the
22244         subsequence to end with different bases in some circumstances.
22245         Record the chosen access functions in SUB_ACCESS_FN.
22246         (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22247         a_index and b_index.  Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22248         (subscript_dependence_tester_1): Likewise dra and drb.
22249         (build_classic_dist_vector): Update calls accordingly.
22250         (subscript_dependence_tester): Likewise.
22251         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22252         DDR_COULD_BE_INDEPENDENT_P.
22253         * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22254         comp_alias_ddrs instead of may_alias_ddrs.
22255         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22256         New function.
22257         (vect_analyze_data_ref_dependence): Use it if
22258         DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22259         distance vectors if that fails.
22260         (dependence_distance_ge_vf): New function.
22261         (vect_prune_runtime_alias_test_list): Use it.  Don't clear
22262         LOOP_VINFO_MAY_ALIAS_DDRS.
22264 2017-08-04  Richard Biener  <rguenther@suse.de>
22266         PR middle-end/81705
22267         * fold-const.c (fold_binary_loc): Properly restrict
22268         minus_var0 && minus_var1 case when associating undefined overflow
22269         entities.
22271 2017-08-04  Tom de Vries  <tom@codesourcery.com>
22273         * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22275 2017-08-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22277         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22278         Don't start diagnostic messages with a capital letter.
22279         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22280         Likewise.
22281         (rs6000_invalid_builtin): Likewise.
22282         (rs6000_trampoline_init): Likewise.
22284 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22286         PR target/81621
22287         * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22288         after setting changeable df flags.
22290 2017-08-03  Richard Biener  <rguenther@suse.de>
22292         * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22293         up if the use is in USE - X.
22295 2017-08-03  Alexander Monakov  <amonakov@ispras.ru>
22297         * toplev.c (dumpfile.h): New include.
22298         (internal_error_reentered): New static function.  Use it...
22299         (internal_error_function): ...here to handle reentered internal_error.
22301 2017-08-03  Richard Biener  <rguenther@suse.de>
22303         PR middle-end/81148
22304         * fold-const.c (split_tree): Add minus_var and minus_con
22305         arguments, remove unused loc arg.  Never generate NEGATE_EXPRs
22306         here but always use minus_*.
22307         (associate_trees): Assert we never associate with MINUS_EXPR
22308         and NULL first operand.  Do not recurse for PLUS_EXPR operands
22309         when associating as MINUS_EXPR either.
22310         (fold_binary_loc): Track minus_var and minus_con.
22312 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22314         PR lto/81430
22315         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22316         ACCEL_COMPILER, apply finish_options on
22317         DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22319 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22321         PR target/81662
22322         * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22323         function_entry_patch_area_size > 0.
22325 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22327         PR driver/81650
22328         * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22329         instead of 10??LU, perform unit multiplication in wide_int,
22330         don't change alloc_object_size_limit if the limit is larger
22331         than SSIZE_MAX.
22333         PR tree-optimization/81655
22334         PR tree-optimization/81588
22335         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22336         the case when ranges[i].low and high are 1 for unsigned type with
22337         precision 1.
22339         PR middle-end/81052
22340         * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22341         (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22343 2017-08-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22345         * tree-vrp.h: Add include guard.
22347 2017-08-02  Uros Bizjak  <ubizjak@gmail.com>
22349         PR target/81644
22350         * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22351         (ud2): New insn pattern.
22352         * config/i386/i386.c (ix86_expand_epilogue):
22353         For naked functions, generate ud2 instead of trap insn.
22355 2017-08-02  Marek Polacek  <polacek@redhat.com>
22357         PR other/81667
22358         * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22360 2017-08-02  Tom de Vries  <tom@codesourcery.com>
22361             Cesar Philippidis  <cesar@codesourcery.com>
22363         * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22364         Add missing edge probabilities.
22366 2017-08-02  Tamar Christina  <tamar.christina@arm.com>
22368         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22369         Correct endianness.
22371 2017-08-02  Jakub Jelinek  <jakub@redhat.com>
22373         PR middle-end/79499
22374         * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22375         for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22376         of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22377         prologue_seq sequences - if any.
22379 2017-08-02  Richard Biener  <rguenther@suse.de>
22381         * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22382         via vectors if supported, integer extracts via punning if supported
22383         or otherwise vector extracts.
22385 2017-08-02  Richard Biener  <rguenther@suse.de>
22387         * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22388         into ...
22389         (bitmap_insert_into_set): ... this.
22391 2017-08-02  Richard Biener  <rguenther@suse.de>
22393         PR tree-optimization/81633
22394         Revert
22395         2015-08-17  Alan Hayward  <alan.hayward@arm.com>
22397         PR tree-optimization/71752
22398         * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22400 2017-08-01  Daniel Santos  <daniel.santos@pobox.com>
22402         * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22403         (machine_function::call_ms2sysv_pad_out): Remove field.
22404         * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22405         (ix86_compute_frame_layout): Likewise.
22407 2017-08-01  H.J. Lu  <hongjiu.lu@intel.com>
22409         PR target/81654
22410         * config/i386/i386.c (ix86_set_func_type): Disallow naked
22411         attribute with interrupt attribute.
22413 2017-08-01  Andrew Pinski  <apinski@cavium.com>
22415         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22416         BIT_INSERT_EXPR's operand 1
22417         to see if the types precision matches.
22419 2017-08-01  Martin Liska  <mliska@suse.cz>
22421         PR middle-end/70140
22422         * builtins.c (expand_builtin_memcpy_args): Remove.
22423         (expand_builtin_memcpy): Call newly added function
22424         expand_builtin_memory_copy_args.
22425         (expand_builtin_memcpy_with_bounds): Likewise.
22426         (expand_builtin_mempcpy): Remove last argument.
22427         (expand_builtin_mempcpy_with_bounds): Likewise.
22428         (expand_builtin_memory_copy_args): New function created from
22429         expand_builtin_mempcpy_args with small modifications.
22430         (expand_builtin_mempcpy_args): Remove.
22431         (expand_builtin_stpcpy): Remove unused argument.
22432         (expand_builtin): Likewise.
22433         (expand_builtin_with_bounds): Likewise.
22435 2017-08-01  Martin Liska  <mliska@suse.cz>
22437         Revert r250771
22438         Make mempcpy more optimal (PR middle-end/70140).
22440 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22442         PR target/81622
22443         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22444         __builtin_vec_cmpne verify both arguments are compatible vectors
22445         before looking at TYPE_MODE on the element type.  For __builtin_vec_ld
22446         verify arg1_type is a pointer or array type.  For __builtin_vec_st,
22447         move computation of aligned to after checking the argument types.
22448         Formatting fixes.
22450         PR target/80846
22451         * config/rs6000/vsx.md (vextract_fp_from_shorth,
22452         vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22453         calls.
22455 2017-08-01  Jerome Lambourg  <lambourg@adacore.com>
22456             Doug Rupp  <rupp@adacore.com>
22457             Olivier Hainque  <hainque@adacore.com>
22459         * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22460         well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22461         arm8 (arch v4).
22462         * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22463         for TARGET_OS_CPP_BUILTIN.
22464         (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22465         refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22466         arm_arch7.
22467         (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22468         passing required abi options to the assembler for EABI configurations.
22469         (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22470         of .text.hot and .text.unlikely sections for kernel modules when
22471         using ARM style exceptions.
22472         (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22473         options. Add EXTRA_CC1_SPEC.
22474         (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22475         toolchain options.
22476         (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22477         transition.
22478         (ARM_TARGET2_DWARF_FORMAT): Define.
22479         * config/arm/t-vxworks: Adjust multilib control to removal of the
22480         Diab command line options.
22482 2017-08-01  Martin Liska  <mliska@suse.cz>
22484         PR gcov-profile/81561
22485         * gcov.c (unblock): Make unblocking safe as we need to preserve
22486         index correspondence of blocks and block_lists.
22488 2017-08-01  Richard Biener  <rguenther@suse.de>
22490         PR tree-optimization/81181
22491         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22492         (compute_antic): ... end of iteration here.
22494 2017-08-01  James Greenhalgh  <james.greenhalgh@arm.com>
22496         * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22497         (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22498         (ftree-slp-vectorize): Likewise.
22499         * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22500         can no longer be set independent of flag_tree_loop_vectorize.
22501         * omp-general.c (emp_max_vf): Likewise.
22502         * opts.c (enable_fdo_optimizations): Remove references to
22503         flag_tree_vectorize, these are now implicit.
22504         (common_handle_option): Remove handling for OPT_ftree_vectorize,
22505         and leave it for the options machinery.
22507 2017-08-01  Martin Liska  <mliska@suse.cz>
22509         PR middle-end/70140
22510         * builtins.c (expand_builtin_memcpy_args): Remove.
22511         (expand_builtin_memcpy): Call newly added function
22512         expand_builtin_memory_copy_args.
22513         (expand_builtin_memcpy_with_bounds): Likewise.
22514         (expand_builtin_mempcpy): Remove last argument.
22515         (expand_builtin_mempcpy_with_bounds): Likewise.
22516         (expand_builtin_memory_copy_args): New function created from
22517         expand_builtin_mempcpy_args with small modifications.
22518         (expand_builtin_mempcpy_args): Remove.
22519         (expand_builtin_stpcpy): Remove unused argument.
22520         (expand_builtin): Likewise.
22521         (expand_builtin_with_bounds): Likewise.
22523 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22525         PR target/81641
22526         * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22527         print "ds:" only for immediates in generic address space.
22529 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22531         PR target/81639
22532         * config/i386/i386.c (ix86_funciton_naked): New prototype.
22533         (ix86_function_ok_for_sibcall): Return false for naked functions.
22535 2017-08-01  Richard Biener  <rguenther@suse.de>
22537         * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22538         (compute_antic): Seed worklist with exit block predecessors.
22539         * cfganal.c (dfs_find_deadend): For a cycle return the source
22540         of the edge closing it.
22542 2017-08-01  Tamar Christina  <tamar.christina@arm.com>
22544         * config/aarch64/aarch64.c
22545         (aarch64_can_const_movi_rtx_p): Move 0 check.
22547 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22549         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22550         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22551         above macro.
22552         * match.pd: Ditto in address comparison pattern.
22554 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22556         PR tree-optimization/81627
22557         * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22558         closed ssa form for store-store chain.
22560 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22562         PR tree-optimization/81620
22563         * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22564         for store-store chain.
22566 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22568         PR tree-optimization/81588
22569         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22570         ranges[i].in_p, invert comparison code ccode.  For >/>=,
22571         swap rhs1 and rhs2 and comparison code unconditionally,
22572         for </<= don't do that.  Don't swap rhs1/rhs2 again if
22573         ranges[i].in_p, instead invert comparison code ccode if
22574         opcode or oe->rank is BIT_IOR_EXPR.
22576         PR target/80846
22577         * optabs.def (vec_extract_optab, vec_init_optab): Change from
22578         a direct optab to conversion optab.
22579         * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22580         with GET_MODE_INNER as last argument instead of optab_handler.
22581         * expmed.c (extract_bit_field_1): Likewise.  Use vector from
22582         vector extraction if possible and optab is available.
22583         * expr.c (store_constructor): Use convert_optab_handler instead
22584         of optab_handler.  Use vector initialization from smaller
22585         vectors if possible and optab is available.
22586         * tree-vect-stmts.c (vectorizable_load): Likewise.
22587         * doc/md.texi (vec_extract, vec_init): Document that the optabs
22588         now have two modes.
22589         * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22590         of vec_init from half-sized vectors with the same element mode.
22591         * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22592         (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22593         (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22594         reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22595         reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22596         after mode in gen_vec_extract* calls.
22597         (vec_extract<mode>): Renamed to ...
22598         (vec_extract<mode><ssescalarmodelower>): ... this.
22599         (vec_extract<mode><ssehalfvecmodelower>): New expander.
22600         (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22601         element mode after mode in gen_vec_init* calls.
22602         (VEC_INIT_HALF_MODE): New mode iterator.
22603         (vec_init<mode>): Renamed to ...
22604         (vec_init<mode><ssescalarmodelower>): ... this.
22605         (vec_init<mode><ssehalfvecmodelower>): New expander.
22606         * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22607         (vec_extractv2sfsf): ... this.
22608         (vec_initv2sf): Renamed to ...
22609         (vec_initv2sfsf): ... this.
22610         (vec_extractv2si): Renamed to ...
22611         (vec_extractv2sisi): ... this.
22612         (vec_initv2si): Renamed to ...
22613         (vec_initv2sisi): ... this.
22614         (vec_extractv4hi): Renamed to ...
22615         (vec_extractv4hihi): ... this.
22616         (vec_initv4hi): Renamed to ...
22617         (vec_initv4hihi): ... this.
22618         (vec_extractv8qi): Renamed to ...
22619         (vec_extractv8qiqi): ... this.
22620         (vec_initv8qi): Renamed to ...
22621         (vec_initv8qiqi): ... this.
22622         * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22623         (vec_init<mode>): Renamed to ...
22624         (vec_init<mode><VEC_base_l>): ... this.
22625         (vec_extract<mode>): Renamed to ...
22626         (vec_extract<mode><VEC_base_l>): ... this.
22627         * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22628         (vec_initv2sfsf): ... this.
22629         * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22630         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22631         vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22632         element mode after mode in gen_vec_init* calls.
22633         * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22634         (vec_init<mode><Vel>): ... this.
22635         (vec_extract<mode>): Renamed to ...
22636         (vec_extract<mode><Vel>): ... this.
22637         * config/aarch64/iterators.md (Vel): New mode attribute.
22638         * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22639         Add element mode after mode in gen_vec_extract* calls.
22640         * config/s390/vector.md (non_vec_l): New mode attribute.
22641         (vec_extract<mode>): Renamed to ...
22642         (vec_extract<mode><non_vec_l>): ... this.
22643         (vec_init<mode>): Renamed to ...
22644         (vec_init<mode><non_vec_l>): ... this.
22645         * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22646         s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22647         vec_extract mode.
22648         * config/arm/iterators.md (V_elem_l): New mode attribute.
22649         * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22650         (vec_extract<mode><V_elem_l>): ... this.
22651         (vec_extractv2di): Renamed to ...
22652         (vec_extractv2didi): ... this.
22653         (vec_init<mode>): Renamed to ...
22654         (vec_init<mode><V_elem_l>): ... this.
22655         (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22656         reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22657         reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22658         Add element mode after gen_vec_extract* calls.
22659         * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22660         (vec_init<mode><unitmode>): ... this.
22661         (vec_extract<mode>): Renamed to ...
22662         (vec_extract<mode><unitmode>): ... this.
22663         * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22664         (vec_init<mode><unitmode>): ... this.
22665         * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22666         (vec_initv2sfsf): ... this.
22667         (vec_extractv2sf): Renamed to ...
22668         (vec_extractv2sfsf): ... this.
22669         (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22670         Add element mode after gen_vec_extract* calls.
22671         * config/mips/mips.md (unitmode): New mode iterator.
22672         * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22673         spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22674         * config/spu/spu.md (inner_l): New mode attribute.
22675         (vec_init<mode>): Renamed to ...
22676         (vec_init<mode><inner_l>): ... this.
22677         (vec_extract<mode>): Renamed to ...
22678         (vec_extract<mode><inner_l>): ... this.
22679         * config/sparc/sparc.md (veltmode): New mode iterator.
22680         (vec_init<VMALL:mode>): Renamed to ...
22681         (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22682         * config/ia64/vect.md (vec_initv2si): Renamed to ...
22683         (vec_initv2sisi): ... this.
22684         (vec_initv2sf): Renamed to ...
22685         (vec_initv2sfsf): ... this.
22686         (vec_extractv2sf): Renamed to ...
22687         (vec_extractv2sfsf): ... this.
22688         * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22689         (vec_init<mode>): Renamed to ...
22690         (vec_init<mode><VEC_base_l>): ... this.
22691         (vec_extract<mode>): Renamed to ...
22692         (vec_extract<mode><VEC_base_l>): ... this.
22693         * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22694         (vec_initv2sfsf): ... this.
22695         * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22696         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22697         vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22698         gen_vec_init* calls.
22700 2017-08-01  Richard Biener  <rguenther@suse.de>
22702         PR tree-optimization/81297
22703         * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22704         TREE_OVERFLOW from INTEGER_CSTs.
22706 2017-07-31  Segher Boessenkool  <segher@kernel.crashing.org>
22708         * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22710 2017-07-31  Carl Love  <cel@us.ibm.com>
22712         * config/rs6000/rs6000-c: Add support for built-in functions
22713         vector signed char vec_xl_be (signed long long, signed char *);
22714         vector unsigned char vec_xl_be (signed long long, unsigned char *);
22715         vector signed int vec_xl_be (signed long long, signed int *);
22716         vector unsigned int vec_xl_be (signed long long, unsigned int *);
22717         vector signed long long vec_xl_be (signed long long, signed long long *);
22718         vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22719         vector signed short vec_xl_be (signed long long, signed short *);
22720         vector unsigned short vec_xl_be (signed long long, unsigned short *);
22721         vector double vec_xl_be (signed long long, double *);
22722         vector float vec_xl_be (signed long long, float *);
22723         * config/rs6000/altivec.h (vec_xl_be): Add #define.
22724         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22725         XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22726         for the builtins.
22727         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22728         (altivec_expand_builtin): Add switch statement to call
22729         altivec_expand_xl_be for each builtin.
22730         (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22731         __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22732         __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22733         __builtin_vsx_le_be_v16qi.
22734         * doc/extend.texi: Update the built-in documentation file for the
22735         new built-in functions.
22737 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22739         PR target/25967
22740         * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22741         New function.
22742         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22744 2017-07-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22746         * config.gcc: Add z14.
22747         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22748         CPU model numbers for z13s and z14.
22749         * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22750         arch12 with z14.
22751         * config/s390/s390-opts.h (enum processor_type): Rename
22752         PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22753         * config/s390/s390.c (processor_table): Add field for CPU name to
22754         be passed to Binutils.
22755         (s390_asm_output_machine_for_arch): Use the new field in
22756         processor_table for Binutils.
22757         (s390_expand_builtin): Replace arch12 with z14.
22758         (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22759         (s390_get_sched_attrmask): Likewise.
22760         (s390_get_unit_mask): Likewise.
22761         * config/s390/s390.opt: Add z14 to processor_type enum.
22763 2017-07-31  Martin Jambor  <mjambor@suse.cz>
22765         PR hsa/81477
22766         * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22767         regardless of optimization level.
22769 2017-07-31  Jan Hubicka <hubicka@ucw.cz>
22770             Martin Liska  <mliska@suse.cz>
22772         * predict.def: Remove old comment and adjust probability.
22773         * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22774         PREDICT statements.
22776 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22778         PR target/25967
22779         * config/i386/i386.c (ix86_function_naked): New function.
22780         (ix86_can_use_return_insn_p): Return false for naked functions.
22781         (ix86_expand_prologue): Skip prologue for naked functions.
22782         (ix86_expand_epilogue): Skip epilogue for naked functions
22783         and emit trap instruction.
22784         (ix86_warn_func_return): New function.
22785         (ix86_attribute_table): Add "naked" attribute specification.
22786         (TARGET_WARN_FUNC_RETURN): Define.
22787         * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22789 2017-07-31  Martin Liska  <mliska@suse.cz>
22791         * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22792         (dump_gimple_bb_header): Always dump BB info.
22793         (pp_cfg_jump): Do not append info about BB when dumping a jump.
22795 2017-07-31  Martin Liska  <mliska@suse.cz>
22797         PR sanitize/81530
22798         * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22799         also with current_function_decl non-null equality.
22801 2017-07-31  Jakub Jelinek  <jakub@redhat.com>
22803         PR sanitizer/81604
22804         * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22805         change type to the element type, instead add eltype variable and
22806         use it where we are interested in the element type.
22808         PR tree-optimization/81603
22809         * ipa-polymorphic-call.c
22810         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22811         offset arithmetic in offset_int, bail out if the resulting bit offset
22812         doesn't fit into shwi.
22814 2017-07-31  Martin Liska  <mliska@suse.cz>
22816         * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22817         (gimplify_save_expr): Fix comment.
22819 2017-07-30  H.J. Lu  <hongjiu.lu@intel.com>
22821         PR target/79793
22822         * config/i386/i386.c (ix86_function_arg): Update arguments for
22823         exception handler.
22824         (ix86_compute_frame_layout): Set the initial stack offset to
22825         INCOMING_FRAME_SP_OFFSET.  Update red-zone offset with
22826         INCOMING_FRAME_SP_OFFSET.
22827         (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22828         stack before exception handler returns.
22829         * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22830         the 'ERROR_CODE' for exception handler.
22832 2017-07-30  Uros Bizjak  <ubizjak@gmail.com>
22834         * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22835         (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22836         (ASM_OUTPUT_REG_POP): Ditto.
22837         * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22838         instead of asm_fprintf to output pure string.
22840 2017-07-29  Jakub Jelinek  <jakub@redhat.com>
22842         * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22843         to imported_module_or_decl hook.
22844         (debug_nothing_tree_tree_tree_bool): Remove.
22845         (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22846         * debug.c (do_nothing_debug_hooks): Use
22847         debug_nothing_tree_tree_tree_bool_bool instead of
22848         debug_nothing_tree_tree_tree_bool.
22849         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22850         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22851         * sdbout.c (sdb_debug_hooks): Likewise.
22852         * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22853         (gen_namespace_die): Add DW_AT_export_symbols attribute if
22854         langhook wants it.
22855         (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22856         if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22857         attribute, don't add anything.
22859 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22861         * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22862         (fold_build2_stat_loc): Likewise.
22863         (fold_build3_stat_loc): Likewise.
22864         * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22865         (fold_build1_loc): Remove macro.
22866         (fold_build2_loc): Likewise.
22867         (fold_build3_loc): Likewise.
22869 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22871         * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22872         (gimple_build_debug_bind_source_stat): Likewise.
22873         * gimple.h (gimple_build_debug_bind): Remove macro.
22874         (gimple_build_debug_bind_source): Likewise.
22876 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22878         * bitmap.c (bitmap_alloc): Adjust.
22879         (bitmap_gc_alloc): Likewise.
22880         * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22882 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22884         * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22885         (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22886         * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22887         (bitmap_gc_alloc_stat): Likewise.
22888         (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22890 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22892         * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22893         * rtl.h (shallow_copy_rtx): Remove macro.
22895 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22897         * emit-rtl.c (gen_raw_REG): Adjust.
22898         * gengenrtl.c (gendef): Likewise.
22899         * rtl.c (rtx_alloc_stat): Remove _stat from name.
22900         * rtl.h (rtx_alloc): Remove macro.
22902 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22904         * tree.c (build_tree_list_vec_stat): Remove _stat from name.
22905         (build_tree_list_stat): Likewise.
22906         * tree.h (build_tree_list): Remove macro.
22907         (build_tree_list_vec): Likewise.
22909 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22911         * tree.c (make_vector_stat): Remove _stat from name.
22912         (build_vector_stat): Likewise.
22913         * tree.h (make_vector_stat): Remove macro.
22914         (build_vector_stat): Likewise.
22916 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22918         * tree.h (build_var_debug_value): Remove prototype.
22920 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22922         * tree.c (tree_cons_stat): Remove _stat from name.
22923         * tree.h (tree_cons): Remove macro.
22925 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22927         * tree.c (build_vl_exp_stat): Remove _stat from name.
22928         * tree.h (build_vl_exp): Remove macro.
22930 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22932         * tree.c (build_decl_stat): Remove _stat from name.
22933         * tree.h (build_decl): Remove macro.
22935 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22937         * gimple.c (gimple_build_with_ops_stat): Adjust.
22938         (gimple_alloc_stat): Remove _stat from name.
22939         * gimple.h (gimple_alloc): Remove macro.
22941 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22943         * tree.c (make_tree_vec_stat): Remove _stat from name.
22944         (grow_tree_vec_stat): Likewise.
22945         * tree.h (make_tree_vec_stat): Adjust prototype.
22946         (grow_tree_vec_stat): Likewise.
22947         (make_tree_vec): Remove macro.
22948         (grow_tree_vec): Likewise.
22950 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22952         * fold-const.c (fold_build1_stat_loc): Adjust.
22953         (fold_build2_stat_loc): Likewise.
22954         (fold_build3_stat_loc): Likewise.
22955         * tree.c (build0_stat): Remove _stat from name.
22956         (build1_stat): Likewise.
22957         (build2_stat): Likewise.
22958         (build3_stat): Likewise.
22959         (build4_stat): Likewise.
22960         (build5_stat): Likewise.
22961         * tree.h (build1_loc): Remove macro, and rename _stat function
22962         to this.
22963         (build2_loc): Likewise.
22964         (build3_loc): Likewise.
22965         (build4_loc): Likewise.
22966         (build5_loc): Likewise.
22968 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22970         * tree.c (make_int_cst_stat): Remove _stat from name.
22971         * tree.h (make_int_cst_stat): Adjust prototype.
22972         (make_int_cst): Remove macro.
22974 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22976         * tree.c (make_tre_binfo_stat): Remove _stat from name.
22977         * tree.h (make_tree_binfo_stat): Adjust prototype.
22978         (make_tree_binfo): Remove.
22980 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22982         * tree.c (copy_node_stat): Rename to copy_node.
22983         (build_distinct_type_copy): Adjust.
22984         * tree.h (copy_node_stat): Adjust prototype.
22985         (copy_node): Remove macro.
22987 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22989         * tree.c (make_node_stat): rename to make_node.
22990         (build_tree_list_stat): Adjust.
22991         (build0_stat): Likewise.
22992         (build2_stat): Likewise.
22993         (build3_stat): Likewise.
22994         (build4_stat): Likewise.
22995         (build5_stat): Likewise.
22996         (build_decl_stat): Likewise.
22997         * tree.h (make_node_stat): Adjust prototype.
22998         (make_node): remove macro.
23000 2017-07-28  Peter Bergner  <bergner@vnet.ibm.com>
23002         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
23003         (PPC_FEATURE2_SCV): Likewise.
23004         * config/rs6000/rs6000.c (cpu_supports_info): Use them.
23006 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23008         * config/aarch64/aarch64.c
23009         (aarch64_internal_mov_immediate): Add new special pattern.
23010         * config/aarch64/aarch64.md (*movdi_aarch64):
23011         Add reg/32bit const mov case.
23013 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23014             Richard Sandiford <richard.sandiford@linaro.org>
23016         * config/aarch64/aarch64.md (mov<mode>): Generalize.
23017         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
23018         Add integer and movi cases.
23019         (movi-split-hf-df-sf split, fp16): New.
23020         (enabled): Added TARGET_FP_F16INST.
23021         * config/aarch64/iterators.md (GPF_HF): New.
23022         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
23024 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
23026         * config/aarch64/aarch64.c
23027         (aarch64_simd_container_mode): Add prototype.
23028         (aarch64_expand_mov_immediate): Add HI support.
23029         (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
23030         (aarch64_can_const_movi_rtx_p): New.
23031         (aarch64_preferred_reload_class):
23032         Remove restrictions of using FP registers for certain SIMD operations.
23033         (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
23034         (aarch64_valid_floating_const): Add integer move validation.
23035         (aarch64_simd_imm_scalar_p): Remove.
23036         (aarch64_output_scalar_simd_mov_immediate): Generalize function.
23037         (aarch64_legitimate_constant_p): Expand list of supported cases.
23038         * config/aarch64/aarch64-protos.h
23039         (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
23040         (aarch64_reinterpret_float_as_int): New.
23041         (aarch64_simd_imm_scalar_p): Remove.
23042         * config/aarch64/constraints.md (Uvi): New.
23043         (Dd): Split into Ds and new Dd.
23044         * config/aarch64/aarch64.md (*movsi_aarch64):
23045         Add SIMD mov case.
23046         (*movdi_aarch64): Add SIMD mov case.
23048 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23050         * tree-predcom.c: (struct chain): Handle store-store chain in which
23051         stores for elimination only store loop invariant values.
23052         (execute_pred_commoning_chain): Ditto.
23053         (prepare_initializers_chain_store_elim): Ditto.
23054         (prepare_finalizers): Ditto.
23055         (is_inv_store_elimination_chain): New function.
23056         (initialize_root_vars_store_elim_1): New function.
23058 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23060         * tree-predcom.c: Revise general description of the pass.
23061         (enum chain_type): New enum type for store elimination.
23062         (struct chain): New field supporting store elimination.
23063         (struct component): Ditto.
23064         (dump_chain): Dump store-stores chain.
23065         (release_chain): Release resources.
23066         (split_data_refs_to_components): Compute and create component
23067         contains only stores for elimination.
23068         (get_chain_last_ref_at): New function.
23069         (make_invariant_chain): Initialization.
23070         (make_rooted_chain): Specify chain type in parameter and record it.
23071         (add_looparound_copies): Skip for store-stores chain.
23072         (determine_roots_comp): Compute type of chain and pass it to
23073         make_rooted_chain.
23074         (initialize_root_vars_store_elim_2): New function.
23075         (finalize_eliminated_stores): New function.
23076         (remove_stmt): Handle store for elimination.
23077         (execute_pred_commoning_chain): Execute predictive commoning on
23078         store-store chains.
23079         (determine_unroll_factor): Skip unroll for store-stores chain.
23080         (prepare_initializers_chain_store_elim): New function.
23081         (prepare_initializers_chain): Hanlde store-store chain.
23082         (prepare_finalizers_chain, prepare_finalizers): New function.
23083         (tree_predictive_commoning_loop): Return integer value indicating
23084         if loop is unrolled or lcssa form is corrupted.
23085         (tree_predictive_commoning): Rewrite for lcssa form if necessary.
23087 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23089         * tree-predcom.c (initialize_root): Delete.
23090         (execute_pred_commoning_chain): Initialize root vars and replace
23091         reference of non-combined chain directly, rather than call above
23092         function.
23094 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23096         * tree-predcom.c (ref_at_iteration): Add parameter NITERS.  Compute
23097         memory reference to DR at (NITERS + ITERS)-th iteration of loop.
23099 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23101         * tree-predcom.c (struct chain): New field init_seq.
23102         (release_chain): Release init_seq.
23103         (prepare_initializers_chain): Record intialization stmts in above
23104         field.
23105         (insert_init_seqs): New function.
23106         (tree_predictive_commoning_loop): Call insert_init_seqs.
23108 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23110         * tree-predcom.c (determine_roots_comp): Skip trivial components.
23112 2017-07-28  Richard Biener  <rguenther@suse.de>
23114         * match.pd: Remove superfluous :c.
23115         * genmatch.c (simplify::id): Add member.
23116         (lower_commutative, lower_opt_convert, lower_cond, lower_for):
23117         Copy id.
23118         (current_id): New global.
23119         (dt_node::parent): Move from ...
23120         (dt_operand::parent): ... here.  Add for_id member.
23121         (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
23122         (decision_tree::find_node): Relax order requirement when
23123         merging DT_TRUE nodes to ones inbetween the current simplify
23124         and the one we try to merge with.  Add diagnostic whenever
23125         we need to enforce pattern order by not merging.
23126         (decision_tree::insert): Set current_id.
23127         (decision_tree::print_node): Dump parent node and for_id.
23128         (parser::last_id): Add member.
23129         (parser::push_simplify): Assign unique id.
23130         (parser::parser): Initialize last_id.
23132 2017-07-28  Martin Liska  <mliska@suse.cz>
23134         PR sanitizer/81340
23135         * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
23136         gimple_build_debug_bind.
23138 2017-07-28  Richard Biener  <rguenther@suse.de>
23140         PR tree-optimization/81502
23141         * match.pd: Add pattern combining BIT_INSERT_EXPR with
23142         BIT_FIELD_REF.
23143         * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
23144         size/pos operands.
23145         (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
23146         * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
23147         for BIT_FIELD_REF args.
23148         * fold-const.c (make_bit_field_ref): Likewise.
23149         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23151 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23153         PR sanitizer/80998
23154         * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
23155         * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
23156         * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
23157         Or it into SANITIZER_UNDEFINED.
23158         * ubsan.c: Include gimple-fold.h and varasm.h.
23159         (ubsan_expand_ptr_ifn): New function.
23160         (instrument_pointer_overflow): New function.
23161         (maybe_instrument_pointer_overflow): New function.
23162         (instrument_object_size): Formatting fix.
23163         (pass_ubsan::execute): Call instrument_pointer_overflow
23164         and maybe_instrument_pointer_overflow.
23165         * internal-fn.c (expand_UBSAN_PTR): New function.
23166         * ubsan.h (ubsan_expand_ptr_ifn): Declare.
23167         * sanitizer.def (__ubsan_handle_pointer_overflow,
23168         __ubsan_handle_pointer_overflow_abort): New builtins.
23169         * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
23170         * internal-fn.def (UBSAN_PTR): New internal function.
23171         * opts.c (sanitizer_opts): Add pointer-overflow.
23172         * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
23173         * fold-const.c (build_range_check): Compute pointer range check in
23174         integral type if pointer arithmetics would be needed.  Formatting
23175         fixes.
23177 2017-07-28  Martin Liska  <mliska@suse.cz>
23179         PR sanitizer/81460
23180         * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
23181         parameters that are of a variable-length.
23183 2017-07-28  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23185         * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h.  Add
23186         rs6000/biarch64.h.
23187         * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
23188         (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
23189         (CRT_CALL_STATIC_FUNCTION): Likewise.
23190         (ASM_DEFAULT_SPEC): New define.
23191         (ASM_SPEC32): Likewise.
23192         (ASM_SPEC64): Likewise.
23193         (ASM_SPEC_COMMON): Likewise.
23194         (ASM_SPEC): Likewise.
23195         (INVALID_64BIT): Likewise.
23196         (LINK_OS_DEFAULT_SPEC): Likewise.
23197         (LINK_OS_SPEC32): Likewise.
23198         (LINK_OS_SPEC64): Likewise.
23199         (POWERPC_LINUX): Likewise.
23200         (PTRDIFF_TYPE): Likewise.
23201         (RESTORE_FP_PREFIX): Likewise.
23202         (RESTORE_FP_SUFFIX): Likewise.
23203         (SAVE_FP_PREFIX): Likewise.
23204         (SAVE_FP_SUFFIX): Likewise.
23205         (SIZE_TYPE): Likewise.
23206         (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
23207         (TARGET_64BIT): Likewise.
23208         (TARGET_64BIT): Likewise.
23209         (TARGET_AIX): Likewise.
23210         (WCHAR_TYPE_SIZE): Likewise.
23211         (WCHAR_TYPE): Undefine.
23212         (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
23213         (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
23214         (CPP_OS_RTEMS_SPEC): Delete.
23215         (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems.  Add
23216         asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
23217         link_os_spec64.
23218         * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
23220 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23222         PR tree-optimization/81578
23223         * tree-parloops.c (build_new_reduction): Bail out if
23224         reduction_code isn't one of the standard OpenMP reductions.
23225         Move the details printing after that decision.
23227 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23229         * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23230         related to reload_in_progress.
23231         (splat_input_operand): Likewise.
23232         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23233         Delete prototype.
23234         * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23235         field.
23236         (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23237         (TARGET_INSTANTIATE_DECLS): Likewise.
23238         (legitimate_indexed_address_p): Delete reload_in_progress code.
23239         (rs6000_debug_legitimate_address_p): Likewise.
23240         (rs6000_eliminate_indexed_memrefs): Likewise.
23241         (rs6000_emit_le_vsx_store): Likewise.
23242         (rs6000_emit_move_si_sf_subreg): Likewise.
23243         (rs6000_emit_move): Likewise.
23244         (register_to_reg_type): Likewise.
23245         (rs6000_pre_atomic_barrier): Likewise.
23246         (rs6000_machopic_legitimize_pic_address): Likewise.
23247         (rs6000_allocate_stack_temp): Likewise.
23248         (rs6000_address_for_fpconvert): Likewise.
23249         (rs6000_address_for_altivec): Likewise.
23250         (rs6000_secondary_memory_needed_rtx): Delete function.
23251         (rs6000_check_sdmode): Likewise.
23252         (rs6000_alloc_sdmode_stack_slot): Likewise.
23253         (rs6000_instantiate_decls): Likewise.
23254         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23255         * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23256         Delete reload_in_progress.
23257         (*vec_reload_and_plus_<mptrsize>): Likewise.
23258         * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23259         (vsx_div_v2di): Likewise.
23260         (vsx_udiv_v2di): Likewise.
23262 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23264         * config/rs6000/rs6000.opt (mlra): Replace with stub.
23265         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23266         * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23267         (rs6000_debug_reg_global): Delete print of LRA status.
23268         (rs6000_option_override_internal): Delete dead LRA related code.
23269         (rs6000_lra_p): Delete function.
23270         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23272 2017-07-27  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23274         * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23275         * config/riscv/rtems.h: New file.
23277 2017-07-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23278             Sudakshina Das  <sudi.das@arm.com>
23280         * config/aarch64/aarch64.md
23281         (define_split for and<mode>3nr_compare): Move
23282         non aarch64_logical_operand to a register.
23283         (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23284         register immediate operand to a register.
23285         * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23287 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23289         PR middle-end/81564
23290         * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23292 2017-07-27  Richard Biener  <rguenther@suse.de>
23294         PR tree-optimization/81573
23295         PR tree-optimization/81494
23296         * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23297         multi defuse cycle case.
23299 2017-07-27  Richard Biener  <rguenther@suse.de>
23301         PR tree-optimization/81571
23302         * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23303         PHIs.
23305 2017-07-27  Eric Botcazou  <ebotcazou@adacore.com>
23307         * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23308         earlier and only if MASK_FPU is set.  Adjust formatting.
23310 2017-07-27  Martin Liska  <mliska@suse.cz>
23312         * opt-functions.awk: Add validation of value of Init.
23313         * optc-gen.awk: Pass new argument.
23315 2017-07-27  Martin Liska  <mliska@suse.cz>
23317         * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23318         Fix wrong condition.
23320 2017-07-27  Martin Liska  <mliska@suse.cz>
23322         * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23323         BBs and edges seen by autoFDO.
23325 2017-07-27  Richard Biener  <rguenther@suse.de>
23327         PR tree-optimization/81502
23328         * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23329         with incompatible but same sized type.
23330         (execute_update_addresses_taken): Likewise.
23332 2017-07-27  James Greenhalgh  <james.greenhalgh@arm.com>
23334         * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23335         flag_tree_loop_vectorize rather than flag_tree_vectorize.
23337 2017-07-27  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23339         PR target/81534
23340         * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23341         ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23342         Change s_operand to memory_operand.
23344 2017-07-27  Richard Sandiford  <richard.sandiford@linaro.org>
23346         * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23347         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23348         (rs6000_emit_le_vsx_permute): ...this.  Take the destination as input.
23349         Emit instructions rather than returning an expression.  Handle TFmode
23350         and KFmode by casting to TImode.
23351         (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23352         (rs6000_emit_le_vsx_store): Likewise.
23353         * config/rs6000/vsx.md (VSX_TI): New iterator.
23354         (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23355         (*vsx_le_undo_permute_<mode>): Likewise.
23356         (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23357         emit the split sequence.
23358         (*vsx_le_perm_store_<mode>): Likewise.
23360 2017-07-27  Jakub Jelinek  <jakub@redhat.com>
23362         PR tree-optimization/81555
23363         PR tree-optimization/81556
23364         * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23365         if true, force CHANGED for the recursive invocation.
23366         (reassociate_bb): Remember original length of ops array, pass
23367         len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23369         * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23370         attributes for noipa attribute.  For naked attribute use
23371         lookup_attribute first before lookup_attribute_spec.
23372         * final.c (rest_of_handle_final): Disable IPA RA for functions with
23373         noipa attribute.
23374         * ipa-visibility.c (non_local_p): Fix comment typos.  Return true
23375         for functions with noipa attribute.
23376         (cgraph_externally_visible_p): Return true for functions with noipa
23377         attribute.
23378         * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23379         for functions with noipa attribute.
23380         * doc/extend.texi: Document noipa function attribute.
23381         * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23382         also for functions with noipa attribute.
23383         (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23385 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23387         * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23388         vec_unalign_load_cost and vec_unalign_store_cost.
23390 2017-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
23392         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23393         -mvsx-small-integer option.
23394         (ISA_3_0_MASKS_IEEE): Likewise.
23395         (OTHER_VSX_VECTOR_MASKS): Likewise.
23396         (POWERPC_MASKS): Likewise.
23397         * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23398         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23399         code, only testing for DImode being allowed in non-VSX floating
23400         point registers.
23401         (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23402         to TARGET_P8_VECTOR test.  Remove redundant VSX test inside of
23403         another VSX test.
23404         (rs6000_option_override_internal): Delete -mvsx-small-integer.
23405         (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23406         TARGET_P8_VECTOR test.
23407         (rs6000_secondary_reload_simple_move): Likewise.
23408         (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23409         since TARGET_P9_VECTOR was already tested.
23410         (rs6000_opt_masks): Remove -mvsx-small-integer.
23411         * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23412         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23413         used.
23414         (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23415         test for TARGET_VEXTRACTUB was used, and that uses
23416         TARGET_P9_VECTOR.
23417         (p9 extract splitter): Likewise.
23418         (vsx_extract_<mode>_di_p9): Likewise.
23419         (vsx_extract_<mode>_store_p9): Likewise.
23420         (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23421         for TARGET_P9_VECTOR was used.  Delete code that is now dead with
23422         the elimination of TARGET_VSX_SMALL_INTEGER.
23423         (vsx_extract_<mode>_p8): Likewise.
23424         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23425         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23426         (vsx_set_<mode>_p9): Likewise.
23427         (vsx_set_v4sf_p9): Likewise.
23428         (vsx_set_v4sf_p9_zero): Likewise.
23429         (vsx_insert_extract_v4sf_p9): Likewise.
23430         (vsx_insert_extract_v4sf_p9_2): Likewise.
23431         * config/rs6000/rs6000.md (sign extend splitter): Change
23432         TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23433         (floatsi<mode>2_lfiwax_mem): Likewise.
23434         (floatunssi<mode>2_lfiwzx_mem): Likewise.
23435         (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23436         since a test for TARGET_P9_VECTOR was used.
23437         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23438         (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23439         (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23440         (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23441         TARGET_P8_VECTOR test.
23442         (fix_trunc<mode>si2_stfiwx): Likewise.
23443         (fix_trunc<mode>si2_internal): Likewise.
23444         (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23445         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23446         used.
23447         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23448         (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23449         TARGET_P8_VECTOR test.
23450         (fixuns_trunc<mode>si2_stfiwx): Likewise.
23451         (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23452         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23453         used.
23454         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23455         (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23456         since a test for TARGET_P9_VECTOR was used.
23457         (splitter for loading small constants): Likewise.
23459 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23461         * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23462         vec_fp_stmt_cost.
23464 2017-07-26  H.J. Lu  <hongjiu.lu@intel.com>
23466         PR target/81563
23467         * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23468         (fp_valid_at): Likewise.
23470 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23472         * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23473         (qdf24xx_addrcost_table): Likewise.
23474         (cortexa57_tunings): Update to use generic_branch_cost.
23475         (cortexa72_tunings): Likewise.
23476         (cortexa73_tunings): Likewise.
23477         (qdf24xx_tunings): Likewise.
23479 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23481         * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23482         (thunderx2t99_branch_cost): Likewise.
23483         (cortexa35_tunings): Update to use generic_branch_cost.
23484         (cortexa53_tunings): Likewise.
23485         (cortexa57_tunings): Likewise.
23486         (cortexa72_tunings): Likewise.
23487         (cortexa73_tunings): Likewise.
23488         (thunderx2t99_tunings): Likewise.
23490 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23492         * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23493         (sparc_option_override): Honour MASK_FSMULD.
23494         * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23495         * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23496         * config/sparc/sparc.opt (mfsmuld): New option.
23497         * doc/invoke.texi (mfsmuld): Document option.
23499 2017-07-26  Marek Polacek  <polacek@redhat.com>
23501         PR middle-end/70992
23502         * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23504 2017-07-26  Richard Biener  <rguenther@suse.de>
23506         * gimple-match-head.c (do_valueize): Return OP if valueize
23507         returns NULL_TREE.
23508         (get_def): New helper to get at the def stmt of a SSA name
23509         if valueize allows.
23510         * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23511         do_valueize to get at the def stmt.
23512         (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23514 2017-07-26  Wilco Dijkstra  <wdijkstr@arm.com>
23516         PR middle-end/46932
23517         * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23519 2017-07-26  Martin Liska  <mliska@suse.cz>
23521         PR sanitize/81186
23522         * function.c (expand_function_start): Make expansion of
23523         nonlocal_goto_save_area after parm_birth_insn.
23525 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23527         * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23528         from all CPU target flags enable members.
23530 2017-07-26  Richard Biener  <rguenther@suse.de>
23532         * genmatch.c (dt_simplify::gen): Make iterator vars const.
23533         (decision_tree::gen): Make 'type' const.
23534         (write_predicate): Likewise.
23536 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23538         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23539         Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23540         (rs6000_option_override_internal): Likewise.
23541         (rs6000_expand_vector_set): Likewise.
23542         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23543         (TARGET_UPPER_REGS_SF): Likewise.
23544         (TARGET_UPPER_REGS_DI): Likewise.
23545         (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23546         (TARGET_DIRECT_MOVE_64BIT): Likewise.
23547         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23548         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23549         (Splitters for DI constants in Altivec registers): Likewise.
23550         * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23551         (vsx_set_v4sf_p9): Likewise.
23552         (vsx_set_v4sf_p9_zero): Likewise.
23553         (vsx_insert_extract_v4sf_p9): Likewise.
23554         (vsx_insert_extract_v4sf_p9_2): Likewise.
23556 2017-07-25  Carl Love  <cel@us.ibm.com>
23558         * doc/extend.texi: Update the built-in documentation file for the
23559         existing built-in functions
23560         vector signed char vec_cnttz (vector signed char);
23561         vector unsigned char vec_cnttz (vector unsigned char);
23562         vector signed short vec_cnttz (vector signed short);
23563         vector unsigned short vec_cnttz (vector unsigned short);
23564         vector signed int vec_cnttz (vector signed int);
23565         vector unsigned int vec_cnttz (vector unsigned int);
23566         vector signed long long vec_cnttz (vector signed long long);
23567         vector unsigned long long vec_cnttz (vector unsigned long long);
23569 2017-07-25  Andrew Pinski  <apinski@cavium.com>
23571         * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23572         accesses where the use is for the first operand of a BIT_INSERT.
23574 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23576         PR bootstrap/81521
23577         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23578         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23580 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23582         * config/i386/gstabs.h: Delete.
23583         * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23585 2017-07-25  Uros Bizjak  <ubizjak@gmail.com>
23587         * config/i386/i386.c (ix86_decompose_address): Do not check for
23588         register RTX when looking at index_reg or base_reg.
23589         * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23591 2017-07-25  Eric Botcazou  <ebotcazou@adacore.com>
23593         * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23594         to update EH info here.
23596 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23598         * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23600 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23602         * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23604 2017-07-25  Torsten Duwe  <duwe@suse.de>
23606         * common.opt: Introduce -fpatchable-function-entry
23607         command line option, and its variables function_entry_patch_area_size
23608         and function_entry_patch_area_start.
23609         * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23610         including a two-value parser.
23611         * target.def (print_patchable_function_entry): New target hook.
23612         * targhooks.h (default_print_patchable_function_entry): New function.
23613         * targhooks.c (default_print_patchable_function_entry): Likewise.
23614         * toplev.c (process_options): Switch off IPA-RA if
23615         patchable function entries are being generated.
23616         * varasm.c (assemble_start_function): Look at the
23617         patchable-function-entry command line switch and current
23618         function attributes and maybe generate NOP instructions by
23619         calling the print_patchable_function_entry hook.
23620         * doc/extend.texi: Document patchable_function_entry attribute.
23621         * doc/invoke.texi: Document -fpatchable_function_entry
23622         command line option.
23623         * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23624         New target hook.
23625         * doc/tm.texi: Re-generate.
23627 2017-07-25  Jakub Jelinek  <jakub@redhat.com>
23629         PR target/81532
23630         * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23631         TARGET_AVX512DQ rather than TARGET_AVX512BW.
23633 2017-07-25  Tamar Christina  <tamar.christina@arm.com>
23635         * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23637 2017-07-25  Richard Biener  <rguenther@suse.de>
23639         PR tree-optimization/81455
23640         * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23641         not walk in cycles when looking for guards.
23643 2017-07-25  Richard Biener  <rguenther@suse.de>
23645         PR tree-optimization/81529
23646         * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23647         when optimizing backedge uses.
23649 2017-07-25  David Edelsohn  <dje.gcc@gmail.com>
23651         * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23652         character for AIX.
23653         * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23654         to dl_section_ref.  On AIX, append an expression to subtract
23655         the size of the section length to dl_section_ref.
23657 2017-07-25  Segher Boessenkool  <segher@kernel.crashing.org>
23659         * configure.ac: If any of the config.* scripts fail, exit 1.
23660         * configure: Regenerate.
23662 2017-07-25  Richard Biener  <rguenther@suse.de>
23664         PR middle-end/81546
23665         * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23666         of immediate uses, be more verbose on errors.
23668 2017-07-25  Richard Biener  <rguenther@suse.de>
23670         PR tree-optimization/81510
23671         * tree-vect-loop.c (vect_is_simple_reduction): When the
23672         reduction stmt is not inside the loop bail out.
23674 2017-07-25  Richard Biener  <rguenther@suse.de>
23676         PR tree-optimization/81303
23677         * tree-vect-loop-manip.c (vect_loop_versioning): Build
23678         profitability check against LOOP_VINFO_NITERSM1.
23680 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23682         * domwalk.c (cmp_bb_postorder): Simplify.
23683         (sort_bbs_postorder): New function.  Use it...
23684         (dom_walker::walk): ...here to optimize common cases.
23686 2017-07-25  Martin Liska  <mliska@suse.cz>
23688         PR ipa/81520
23689         * ipa-visibility.c (function_and_variable_visibility): Make the
23690         redirection just on target that supports aliasing.
23691         Fix GNU coding style.
23693 2017-07-25  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23695         PR libgcc/61152
23696         * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23697         Format changes.
23698         * config/arm/rtems.h: Likewise.
23699         * config/bfin/rtems.h: Likewise.
23700         * config/i386/rtemself.h: Likewise.
23701         * config/lm32/rtems.h: Likewise.
23702         * config/m32c/rtems.h: Likewise.
23703         * config/m68k/rtemself.h: Likewise.
23704         * config/microblaze/rtems.h: Likewise.
23705         * config/mips/rtems.h: Likewise.
23706         * config/moxie/rtems.h: Likewise.
23707         * config/nios2/rtems.h: Likewise.
23708         * config/powerpcspe/rtems.h: Likewise.
23709         * config/rs6000/rtems.h: Likewise.
23710         * config/rtems.h: Likewise.
23711         * config/sh/rtems.h: Likewise.
23712         * config/sh/rtemself.h: Likewise.
23713         * config/sparc/rtemself.h: Likewise.
23715 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23717         PR 81487
23718         * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23719         * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23720         * tree-ssa-structalias.c (alias_get_name): Same.
23722 2017-07-25  Bin Cheng  <bin.cheng@arm.com>
23724         PR target/81414
23725         * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23726         instructions if no du chain is found.
23728 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23730         * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23732 2017-07-25  Richard Biener  <rguenther@suse.de>
23734         PR middle-end/81505
23735         * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23736         sticky.
23738 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23740         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23741         upper-regs options.
23742         (ISA_2_7_MASKS_SERVER): Likewise.
23743         (ISA_3_0_MASKS_IEEE): Likewise.
23744         (OTHER_P8_VECTOR_MASKS): Likewise.
23745         (OTHER_VSX_VECTOR_MASKS): Likewise.
23746         (POWERPC_MASKS): Likewise.
23747         (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23748         duplicate list of options.
23749         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23750         explicit -mupper-regs options.
23751         * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23752         -mupper-regs* options.  Delete -mvsx-scalar-memory, which was an
23753         alias for -mupper-regs-df.
23754         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23755         (rs6000_init_hard_regno_mode_ok): Likewise.
23756         (rs6000_option_override_internal): Likewise.
23757         (rs6000_opt_masks): Likewise.
23758         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23759         options in terms of whether -mvsx or -mpower8-vector was used.
23760         (TARGET_UPPER_REGS_DI): Likewise.
23761         (TARGET_UPPER_REGS_SF): Likewise.
23762         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23763         -mupper-regs-* options.
23765 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23767         * passes.c (emergency_dump_function): Print some empty lines and a
23768         header before the RTL dump.
23770 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23772         * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23774 2017-07-24  Wilco Dijkstra  <wdijkstr@arm.com>
23776         PR target/79041
23777         * config/aarch64/aarch64.c (aarch64_classify_symbol):
23778         Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23780 2017-07-24  Carl Love  <cel@us.ibm.com>
23782         * config/rs6000/rs6000-c.c: Add support for built-in functions
23783         vector float vec_extract_fp32_from_shorth (vector unsigned short);
23784         vector float vec_extract_fp32_from_shortl (vector unsigned short);
23785         * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23786         vec_extract_fp_from_shortl): Add defines for the two builtins.
23787         * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23788         VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23789         new builtins.
23790         * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23791         (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23792         * doc/extend.texi: Update the built-in documentation file for the
23793         new built-in function.
23795 2017-07-24  Jakub Jelinek  <jakub@redhat.com>
23797         PR bootstrap/81521
23798         * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23799         documentation.
23800         * doc/generic.texi: Likewise.
23801         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23802         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23804 2017-07-24  Jackson Woodruff  <jackson.woodruff@arm.com>
23806         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23807         (aarch64_mls_elt_merge<mode>): Likewise.
23809 2017-07-23  Krister Walfridsson  <krister.walfridsson@gmail.com>
23811         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23812         having __cxa_atexit.
23814 2017-07-23  Michael Collison  <michael.collison@arm.com>
23816         * config/arm/arm.c (arm_option_override): Deprecate
23817         use of -mstructure-size-boundary.
23818         * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23819         * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23821 2017-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23823         PR target/80695
23824         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23825         Reduce cost estimate for direct moves.
23827 2017-07-23  Uros Bizjak  <ubizjak@gmail.com>
23829         PR target/80569
23830         * config/i386/i386.c (ix86_option_override_internal): Disable
23831         BMI, BMI2 and TBM instructions for -m16.
23833 2017-07-21  Carl Love  <cel@us.ibm.com>
23835         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23836         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23837         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23838         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23839         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23840         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23841         VMULOSW): New enum "unspec" values.
23842         (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23843         altivec_vmulosw): New patterns.
23844         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23845         VMULOSW): Add definitions.
23847 2017-07-21  Jim Wilson  <jim.wilson@linaro.org>
23849         * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23850         (qdf24xx): Likewise.
23851         * config/aarch64/aarch64-options-extensions.def (rdma); New.
23852         * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23853         (AARCH64_FL_V8_1): Renumber.
23854         (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23855         (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23856         * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23857         * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs.  Add
23858         rdma to feature modifiers list.
23860 2017-07-21  Yury Gribov  <tetra2005@gmail.com>
23862         PR middle-end/56727
23863         * ipa-visibility (function_and_variable_visibility): Convert
23864         recursive PLT call to direct call if appropriate.
23866 2017-07-21  Andrew Pinski  <apinski@cavium.com>
23868         * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23869         operand 1 to see if the types precision matches.
23870         * fold-const.c (operand_equal_p): Likewise.
23872 2017-07-21  Richard Biener  <rguenther@suse.de>
23874         PR tree-optimization/81303
23875         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23876         in datarefs vector.  Allow NULL dr0 for no peeling cost estimate.
23877         (vect_peeling_hash_get_lowest_cost): Adjust.
23878         (vect_enhance_data_refs_alignment): Likewise.  Use
23879         vect_get_peeling_costs_all_drs to compute the penalty for no
23880         peeling to match up costs.
23882 2017-07-21  Richard Biener  <rguenther@suse.de>
23884         PR tree-optimization/81500
23885         * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23886         we didn't identify a reduction path.
23888 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23889             Cesar Philippidis  <cesar@codesourcery.com>
23891         PR gcov-profile/81442
23892         * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23893         probabilities.
23895 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23897         PR lto/81430
23898         * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
23899         function.
23900         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
23901         nvptx_override_options_after_change.
23903 2017-07-21  Ulrich Drepper  <drepper@redhat.com>
23905         * dwarf2out.c (output_file_names): Avoid double testing for
23906         dwarf_version >= 5.
23908 2017-07-21  Georg-Johann Lay  <avr@gjlay.de>
23910         * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
23912 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23914         * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
23915         hot/cold regions.
23916         (try_crossjump_to_edge): Do not punt on partitioned functions.
23918 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23920         * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
23921         Put all BBs reachable only via paths crossing cold region to cold
23922         region.
23923         * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
23925 2016-07-21  Richard Biener  <rguenther@suse.de>
23927         PR tree-optimization/81303
23928         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
23929         into account prologue and epilogue iterations when raising
23930         min_profitable_iters to sth at least covering one vector iteration.
23932 2017-07-21  Tamar Christina  <tamar.christina@arm.com>
23934         * config/arm/arm.c (arm_test_cpu_arch_dat):
23935         Check for overlap.
23937 2017-07-20  Nathan Sidwell  <nathan@acm.org>
23939         Remove TYPE_METHODS.
23940         * tree.h (TYPE_METHODS): Delete.
23941         * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
23942         * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
23943         (dbxout_type_methods): Scan TYPE_FIELDS.
23944         (dbxout_type): Don't check TYPE_METHODS here.
23945         * function.c (use_register_for_decl): Always ignore register for
23946         class types when not optimizing.
23947         * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
23948         * tree.c (free_lang_data_in_type): Stitch out member functions and
23949         templates from TYPE_FIELDS.
23950         (build_distinct_type_copy, verify_type_variant,
23951         verify_type): Member fns are on TYPE_FIELDS.
23952         * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
23953         * tree-pretty-print.c (dump_generic_node): Likewise.
23955 2017-07-20  Jakub Jelinek  <jakub@redhat.com>
23957         PR target/80846
23958         * config/i386/i386.c (ix86_expand_vector_init_general): Handle
23959         V2TImode and V4TImode.
23960         (ix86_expand_vector_extract): Likewise.
23961         * config/i386/sse.md (VMOVE): Enable V4TImode even for just
23962         TARGET_AVX512F, instead of only for TARGET_AVX512BW.
23963         (ssescalarmode): Handle V4TImode and V2TImode.
23964         (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
23965         (*vec_extractv2ti, *vec_extractv4ti): New insns.
23966         (VEXTRACTI128_MODE): New mode iterator.
23967         (splitter for *vec_extractv?ti first element): New.
23968         (VEC_INIT_MODE): New mode iterator.
23969         (vec_init<mode>): Consolidate 3 expanders into one using
23970         VEC_INIT_MODE mode iterator.
23972 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23974         * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
23975         non_spilled_static_chain_regno_p.
23977 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23979         * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
23981 2017-07-20  Jan Hubicka  <hubicka@ucw.cz>
23983         * bb-reorder.c (connect_traces): Allow copying of blocks within
23984         single partition.
23986 2017-07-20  Richard Biener  <rguenther@suse.de>
23988         * gimple.h (gimple_phi_result): Add gphi * overload.
23989         (gimple_phi_result_ptr): Likewise.
23990         (gimple_phi_arg): Likewise.  Adjust index assert to only
23991         allow actual argument accesses rather than all slots available
23992         by capacity.
23993         (gimple_phi_arg_def): Add gphi * overload.
23994         * tree-phinodes.c (make_phi_node): Initialize only actual
23995         arguments.
23996         (resize_phi_node): Clear memory not covered by old node,
23997         do not initialize excess argument slots.
23998         (reserve_phi_args_for_new_edge): Initialize new argument slot
23999         completely.
24001 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
24003         PR tree-optimization/81388
24004         Revert r238585:
24005         2016-07-21  Bin Cheng  <bin.cheng@arm.com>
24007         * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
24008         by removing computation of may_be_zero.
24010 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24011             Tom de Vries  <tom@codesourcery.com>
24013         PR middle-end/81030
24014         * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
24015         when gimple level profile disagrees with what RTL expander did.
24017 2017-07-20  Richard Biener  <rguenther@suse.de>
24019         PR tree-optimization/61171
24020         * tree-vectorizer.h (slp_instance): Add reduc_phis member.
24021         (vect_analyze_stmt): Add slp instance parameter.
24022         (vectorizable_reduction): Likewise.
24023         * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
24024         (vect_is_simple_reduction): Deal with chains not detected
24025         as SLP reduction chain, specifically not properly associated
24026         chains containing a mix of plus/minus.
24027         (get_reduction_op): Remove.
24028         (get_initial_defs_for_reduction): Simplify, pass in whether
24029         this is a reduction chain, pass in the SLP node for the PHIs.
24030         (vect_create_epilog_for_reduction): Get the SLP instance as
24031         arg and adjust.
24032         (vectorizable_reduction): Get the SLP instance as arg.
24033         During analysis remember the SLP node with the PHIs in the
24034         instance.  Simplify getting at the vectorized reduction PHIs.
24035         * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
24036         through SLP instance.
24037         (vect_slp_analyze_operations): Likewise.
24038         * tree-vect-stms.c (vect_analyze_stmt): Likewise.
24039         (vect_transform_stmt): Likewise.
24041 2017-07-20  Tom de Vries  <tom@codesourcery.com>
24043         PR tree-optimization/81489
24044         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
24045         read of phi arg location to before loop that modifies phi.
24047 2017-07-20  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
24049         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
24050         New pattern.
24052 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24054         PR middle-end/81331
24055         * except.c (execute): Fix ordering issue.
24057 2018-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24059         PR rtl-optimization/81423
24060         * combine.c (make_compound_operation_int): Don't try to optimize
24061         the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
24063 2017-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
24065         PR rtl-optimization/81423
24066         * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
24067         with a constant that is -1 in the truncated to mode.
24069 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24071         * predict.c (propagate_unlikely_bbs_forward): Break out from ...
24072         (determine_unlikely_bbs): ... here.
24073         * predict.h (propagate_unlikely_bbs_forward): Declare.
24074         * cfgexpand.c (pass_expand::execute): Use it.
24075         * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
24076         unlikely edges.
24077         (find_rarely_executed_basic_blocks_and_crossing_edges): Use
24078         propagate_unlikely_bbs_forward.
24080 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
24082         PR middle-end/81331
24083         * except.c (maybe_add_nop_after_section_switch): New function.
24084         (execute): Use it.
24086 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24088         * gimple.h (gimple_phi_set_arg): Make assert more strict.
24090 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24092         * gimple.h (gimple_phi_arg): Make assert more strict.
24094 2017-07-19  Steven Munroe  <munroesj@gcc.gnu.org>
24096         * config.gcc (powerpc*-*-*): Add mmintrin.h.
24097         * config/rs6000/mmintrin.h: New file.
24098         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
24100 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24102         PR tree-optimization/81346
24103         * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
24105 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24107         * config/nvptx/nvptx.md (VECIM): Add V2DI.
24109 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24111         * config/nvptx/nvptx-modes.def: Add V2DImode.
24112         * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
24113         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
24114         (nvptx_output_mov_insn): Handle lack of mov.b128.
24115         (nvptx_print_operand): Handle 'H' and 'L' codes.
24116         (nvptx_vector_mode_supported): Allow V2DImode.
24117         (nvptx_preferred_simd_mode): New function.
24118         (nvptx_data_alignment): New function.
24119         (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
24120         nvptx_preferred_simd_mode.
24121         * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
24122         64 to 128 bits.
24123         (DATA_ALIGNMENT): Define.  Set to nvptx_data_alignment.
24125 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24127         * config/nvptx/nvptx-modes.def: New file.  Add V2SImode.
24128         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
24129         (nvptx_vector_mode_supported): New function.  Allow V2SImode.
24130         (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
24131         * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
24132         (mov<VECIM>_insn): New define_insn.
24133         (define_expand "mov<VECIM>): New define_expand.
24135 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24137         * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
24139 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24141         PR tree-optimization/81346
24142         * fold-const.h (fold_div_compare, range_check_type): Declare.
24143         * fold-const.c (range_check_type): New function.
24144         (build_range_check): Use range_check_type.
24145         (fold_div_compare): No longer static, rewritten into
24146         a match.pd helper function.
24147         (fold_comparison): Don't call fold_div_compare here.
24148         * match.pd (X / C1 op C2): New optimization using fold_div_compare
24149         as helper function.
24151 2017-07-19  Nathan Sidwell  <nathan@acm.org>
24153         * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
24154         (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
24155         * tree.c (find_decls_types_r, verify_type): Use
24156         TYPE_{MIN,MAX}_VALUE_RAW.
24157         * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
24158         (hash_tree): Likewise.
24159         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
24160         Likewise.
24161         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
24162         Likewise.
24164 2017-07-18  Tom de Vries  <tom@codesourcery.com>
24166         PR middle-end/81464
24167         * omp-expand.c (expand_omp_for_static_chunk): Handle
24168         equal-argument loop exit phi.
24170 2017-07-18  Uros Bizjak  <ubizjak@gmail.com>
24172         PR target/81471
24173         * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
24174         (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
24175         operand 2 predicate.
24176         (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
24177         operand 2 predicate.
24178         (ror,rol -> rorx splitters): Use const_int_operand as
24179         operand 2 predicate.
24181 2017-06-18  Richard Biener  <rguenther@suse.de>
24183         PR tree-optimization/81410
24184         * tree-vect-stmts.c (vectorizable_load): Properly adjust for
24185         the gap in the ! slp_perm SLP case after each group.
24187 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24189         PR middle-end/81463
24190         * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
24191         again.
24193 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24195         PR middle-end/81462
24196         * predict.c (set_even_probabilities): Cleanup; do not affect
24197         probabilities that are already known.
24198         (combine_predictions_for_bb): Call even when count is set.
24200 2017-07-18  Nathan Sidwell  <nathan@acm.org>
24202         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
24203         TYPE_MAX_VALUE.
24205 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24207         PR target/81408
24208         * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
24209         optimization for loop niter analysis.
24211 2017-07-18  Georg-Johann Lay  <avr@gjlay.de>
24213         PR target/81473
24214         * config/avr/avr.c (avr_optimize_casesi): Don't use
24215         INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
24217 2017-07-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
24219         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
24220         body_cost_vec from _vect_peel_extended_info.
24221         (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
24222         (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24223         npeel.
24225 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24227         * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24229 2017-07-18  Richard Biener  <rguenther@suse.de>
24231         PR tree-optimization/80620
24232         PR tree-optimization/81403
24233         * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24234         info when re-using a VN table entry.
24236 2017-07-18  Richard Biener  <rguenther@suse.de>
24238         PR tree-optimization/81418
24239         * tree-vect-loop.c (vectorizable_reduction): Properly compute
24240         vectype_in.  Verify that with lane-reducing reduction operations
24241         we have a single def-use cycle.
24243 2017-07-17  Carl Love  <cel@us.ibm.com>
24245         Revert commit r249424   2017-06-20  Carl Love  <cel@us.ibm.com>
24247         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24248         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24249         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24250         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24251         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24252         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24253         VMULOSW): New enum "unspec" values.
24254         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24255         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24256         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24257         altivec_vmulosw): New patterns.
24258         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24259         VMULOSW): Add definitions.
24261 2017-07-17  Uros Bizjak  <ubizjak@gmail.com>
24263         * config/alpha/alpha.c: Include predict.h.
24265 2017-07-17  Yury Gribov  <tetra2005@gmail.com>
24267         * tree-vrp.c (compare_assert_loc): Fix comparison function
24268         to return predictable results.
24270 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24272         * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24273         option.
24274         (subdi3): Likewise.
24275         * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24276         * doc/invoke.texi (mexpand-adddi): Update text.
24278 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24280         * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24281         that also clobbers the CC register. The old expand code is moved
24282         to ...
24283         (*arc_clzsi2): ... here.
24284         (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24285         the CC register. The old expand code is moved to ...
24286         (arc_ctzsi2): ... here.
24288 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24290         * config/arc/arc.opt (mindexed-loads): Use initial value
24291         TARGET_INDEXED_LOADS_DEFAULT.
24292         (mauto-modify-reg): Use initial value
24293         TARGET_AUTO_MODIFY_REG_DEFAULT.
24294         * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24295         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24296         * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24297         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24299 2017-07-17  Martin Liska  <mliska@suse.cz>
24301         PR sanitizer/81302
24302         * opts.c (finish_options): Do not allow -fgnu-tm
24303         w/ -fsanitize={kernel-,}address.  Say sorry.
24305 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24307         PR target/81369
24308         * tree-loop-distribution.c (classify_partition): Only assert on
24309         numer of iterations.
24310         (merge_dep_scc_partitions): Delete prameter.  Update function call.
24311         (distribute_loop): Remove code handling loop with unknown niters.
24312         (pass_loop_distribution::execute): Skip loop with unknown niters.
24314 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24316         PR target/81369
24317         * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24318         function sort_partitions_by_post_order.
24320 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24322         PR tree-optimization/81374
24323         * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24324         the max index of basic blocks, rather than number of basic blocks.
24326 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24328         * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24329         proto.
24330         (arc_legitimate_pic_operand_p): Likewise.
24331         * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24332         function.
24333         (arc_needs_pcl_p): Likewise.
24334         (arc_legitimate_pc_offset_p): Likewise.
24335         (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24336         function is also used in constrains.md.
24337         (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24338         validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24339         PLUS.  Only return true/false in known cases, otherwise assert.
24340         (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24341         is already called in arc_legitimate_constant_p.
24342         * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24343         pic addresses.
24344         (LEGITIMATE_PIC_OPERAND_P): Use
24345         arc_raw_symbolic_reference_mentioned_p function.
24346         * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24347         function.
24348         (Cal): Likewise.
24349         (C32): Likewise.
24351 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24352         Andrew Burgess  <andrew.burgess@embecosm.com>
24354         * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24355         (arc_return_address_register): New function.
24356         * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24357         (arc_handle_fndecl_attribute): Add naked attribute.
24358         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24359         (TARGET_WARN_FUNC_RETURN): Likewise.
24360         (arc_allocate_stack_slots_for_args): New function.
24361         (arc_warn_func_return): Likewise.
24362         (machine_function): Change type fn_type.
24363         (arc_compute_function_type): Consider new naked function type,
24364         change function return type.
24365         (arc_must_save_register): Adapt to handle new
24366         arc_compute_function_type's return type.
24367         (arc_expand_prologue): Likewise.
24368         (arc_expand_epilogue): Likewise.
24369         (arc_return_address_regs): Delete.
24370         (arc_return_address_register): New function.
24371         (arc_epilogue_uses): Use above function.
24372         * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24373         (arc_function_type): Change encoding, add naked type.
24374         (ARC_INTERRUPT_P): Change to handle the new encoding.
24375         (ARC_FAST_INTERRUPT_P): Likewise.
24376         (ARC_NORMAL_P): Define.
24377         (ARC_NAKED_P): Likewise.
24378         (arc_compute_function_type): Delete prototype.
24379         * config/arc/arc.md (in_ret_delay_slot): Use
24380         arc_return_address_register function.
24381         (simple_return): Likewise.
24382         (p_return_i): Likewise.
24384 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24386         PR tree-optimization/81428
24387         * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24388         can't be built for those types.
24390 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24392         Remove stuff dead since r239246.
24394         * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24395         * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24396         (avr_inform_devices): Remove dead stuff.
24398 2017-07-17  Tamar Christina  <tamar.christina@arm.com>
24400         * config/arm/arm_neon.h: Fix softp typo.
24402 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24404         PR tree-optimization/81365
24405         * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24406         aggregate moves onto bb predecessor edges, make sure there are no
24407         loads that could alias the lhs in between the start of bb and the
24408         loads from *phi.
24410 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24412         PR 80929
24413         * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24414         (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24415         [LSHIFTRT, outer_code = TRUNCATE]: Same.
24417 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24419         PR tree-optimization/81396
24420         * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24421         (init_symbolic_number): Initialize it to 1.
24422         (perform_symbolic_merge): Add n_ops from both operands into the new
24423         n_ops.
24424         (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24425         without base_addr as useless if they need more than one operation.
24426         (bswap_replace): Handle !bswap case for NULL base_addr.
24428 2017-07-17  Tom de Vries  <tom@codesourcery.com>
24430         PR target/81069
24431         * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24432         as possible.
24434 2017-07-17  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24436         * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24437         conditional builtin define __FIX_LEON3FT_B2BST.
24439 2017-07-17  Daniel Cederman  <cederman@gaisler.com>
24441         * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24442         MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24443         with -mfix-ut700.
24445 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24447         PR rtl-optimization/81424
24448         * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24449         to remove potential trapping from operands if -fnon-call-exceptions.
24451 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24453         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24454         profile_proability for scalling.
24455         * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24457 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24459         * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24461 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24463         * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24464         fixpoint arithmetics.
24466 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24468         * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24469         fixpoint arithmetics.
24471 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24473         * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24474         fixpoint arithmetics.
24476 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24478         * profile-count.h (profile_probability::from_reg_br_prob_note,
24479         profile_probability::to_reg_br_prob_note): New functions.
24480         * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24481         * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24482         * predict.c (probability_reliable_p): Update.
24483         (edge_probability_reliable_p): Update.
24484         (br_prob_note_reliable_p): Update.
24485         (invert_br_probabilities): Update.
24486         (add_reg_br_prob_note): New function.
24487         (combine_predictions_for_insn): Update.
24488         * asan.c (asan_clear_shadow): Update.
24489         * cfgbuild.c (compute_outgoing_frequencies): Update.
24490         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24491         (update_br_prob_note): Update.
24492         (rtl_verify_edges): Update.
24493         (purge_dead_edges): Update.
24494         (fixup_reorder_chain): Update.
24495         * emit-rtl.c (try_split): Update.
24496         * ifcvt.c (cond_exec_process_insns): Update.
24497         (cond_exec_process_if_block): Update.
24498         (dead_or_predicable): Update.
24499         * internal-fn.c (expand_addsub_overflow): Update.
24500         (expand_neg_overflow): Update.
24501         (expand_mul_overflow): Update.
24502         * loop-doloop.c (doloop_modify): Update.
24503         * loop-unroll.c (compare_and_jump_seq): Update.
24504         * optabs.c (emit_cmp_and_jump_insn_1): Update.
24505         * predict.h: Update.
24506         * reorg.c (mostly_true_jump): Update.
24507         * rtl.h: Update.
24508         * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24509         * config/alpha/alpha.c (emit_unlikely_jump): Update.
24510         * config/arc/arc.c: (emit_unlikely_jump): Update.
24511         * config/arm/arm.c: (emit_unlikely_jump): Update.
24512         * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24513         * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24514         * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24515         (ix86_print_operand): Update.
24516         (ix86_split_fp_branch): Update.
24517         (predict_jump): Update.
24518         * config/ia64/ia64.c (ia64_print_operand): Update.
24519         * config/mmix/mmix.c (mmix_print_operand): Update.
24520         * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24521         (rs6000_expand_split_stack_prologue): Update.
24522         * config/rs6000/rs6000.c: Update.
24523         * config/s390/s390.c (s390_expand_vec_strlen): Update.
24524         (s390_expand_vec_movstr): Update.
24525         (s390_expand_cs_tdsi): Update.
24526         (s390_expand_split_stack_prologue): Update.
24527         * config/sh/sh.c (sh_print_operand): Update.
24528         (expand_cbranchsi4): Update.
24529         (expand_cbranchdi4): Update.
24530         * config/sparc/sparc.c (output_v9branch): Update.
24531         * config/spu/spu.c (get_branch_target): Update.
24532         (ea_load_store_inline): Update.
24533         * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24534         * config/tilepro/tilepro.c: Update.
24536 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24538         * gimplify.c (mostly_copy_tree_r): Revert latest change.
24539         (gimplify_save_expr): Likewise.
24541 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24543         * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24545 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24547         * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24548         TV_IPA_FNSUMMARY.
24549         * timevar.def (TV_IPA_FNSUMMARY): Define.
24551 2017-07-16  Daniel Cederman  <cederman@gaisler.com>
24553         * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24554         to back store errata sensitive sequence from being generated.
24555         (sqrtdf2_fix): Likewise.
24557 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24559         * tree-ssa-threadupdate.c (compute_path_counts,
24560         update_joiner_offpath_counts): Use profile_probability.
24562 2017-07-15  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24564         Revert:
24565         2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24567         * config/arm/arm-c.c (arm_cpu_builtins): Define
24568         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24570 2017-07-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24572         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24573         array entries to represent __ieee128 versions of the
24574         scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24575         scalar_extract_sig, and scalar_insert_exp built-in functions.
24576         (altivec_resolve_overloaded_builtin): Add special case handling
24577         for the __builtin_scalar_insert_exp function, as represented by
24578         the P9V_BUILTIN_VEC_VSIEDP constant.
24579         * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24580         exponent support for __ieee128 argument.
24581         (VSESQP): Add scalar extract signature support for __ieee128
24582         argument.
24583         (VSTDCNQP): Add scalar test negative support for __ieee128
24584         argument.
24585         (VSIEQP): Add scalar insert exponent support for __int128 argument
24586         with __ieee128 result.
24587         (VSIEQPF): Add scalar insert exponent support for __ieee128
24588         argument with __ieee128 result.
24589         (VSTDCQP): Add scalar test data class support for __ieee128
24590         argument.
24591         (VSTDCNQP): Add overload support for scalar test negative with
24592         __ieee128 argument.
24593         (VSTDCQP): Add overload support for scalar test data class
24594         __ieee128 argument.
24595         * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24596         UNSPEC_VSX_SXSIGDP.
24597         (UNSPEC_VSX_SIEXPQP): New constant.
24598         (xsxexpqp): New insn for VSX scalar extract exponent quad
24599         precision.
24600         (xsxsigqp): New insn for VSX scalar extract significand quad
24601         precision.
24602         (xsiexpqpf): New insn for VSX scalar insert exponent quad
24603         precision with floating point argument.
24604         (xststdcqp): New expand for VSX scalar test data class quad
24605         precision.
24606         (xststdcnegqp): New expand for VSX scalar test negative quad
24607         precision.
24608         (xststdcqp): New insn to match expansions for VSX scalar test data
24609         class quad precision and VSX scalar test negative quad precision.
24610         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24611         special case operand checking to enforce that second operand of
24612         VSX scalar test data class with quad precision argument is a 7-bit
24613         unsigned literal.
24614         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24615         prototypes and descriptions of __ieee128 versions of
24616         scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24617         scalar_test_data_class, and scalar_test_neg built-in functions.
24619 2016-07-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24621         PR tree-optimization/81162
24622         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24623         replace a negate with an add.
24625 2017-07-14  James Greenhalgh  <james.greenhalgh@arm.com>
24627         * doc/invoke.texi (arm/-mcpu): Document +crypto.
24629 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24631         * config/arm/arm-c.c (arm_cpu_builtins): Define
24632         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24634 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24636         * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24637         (armv8-r): Set ARM Cortex-R52 as default CPU.
24638         * config/arm/arm-tables.opt: Regenerate.
24639         * config/arm/arm-tune.md: Regenerate.
24640         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24641         Cortex-R52.
24642         * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24643         extension for -mcpu=cortex-r52.
24645 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24647         * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24648         (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24649         * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24650         (fp-armv8): Define it as FP_ARMv8 only.
24651         config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24652         (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24653         TARGET_FPU_ARMV8.
24654         config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24655         TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24656         * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24657         first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24658         than TARGET_FPU_ARMV8.
24659         * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24660         __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24661         * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24662         TARGET_FPU_ARMV8.
24663         * config/arm/neon.md (neon_vrint): Likewise.
24664         (neon_vcvt): Likewise.
24665         (neon_<fmaxmin_op><mode>): Likewise.
24666         (<fmaxmin><mode>3): Likewise.
24667         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24668         * config/arm/predicates.md (arm_cond_move_operator): Check against
24669         TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24671 2017-07-14  Jackson Woodruff  <jackson.woodruff@arm.com>
24673         * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24674         to top of function.
24676 2017-07-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24678         * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24679         loop in comment with memset.
24681 2017-07-14  Martin Liska  <mliska@suse.cz>
24683         * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24684         * dwarf2out.c (is_java): Remove the function.
24685         (output_pubname): Remove usage of the function.
24686         (lower_bound_default): Remove usage of DW_LANG_Java.
24687         (gen_compile_unit_die): Likewise.
24688         * gcc.c: Remove compiler defaults for .java and .zip files.
24689         * gimple-expr.c (remove_suffix): Change as there's no longer
24690         extension than 4-letter one.
24691         * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24692         (gimplify_save_expr): Likewise.
24693         * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24694         as it's possible even for other languages than Java.
24695         * langhooks.h (struct lang_hooks): Remove Java from a comment.
24696         * lto-opts.c (lto_write_options): Remove reference to Java.
24697         * opts.c (strip_off_ending): Update file extension handling.
24698         * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24699         * tree-eh.c (lower_resx): Likewise.
24700         * tree.c (free_lang_data_in_type): Remove dead code.
24701         (find_decls_types_r): Likewise.
24702         (build_common_builtin_nodes): Remove Java from a comment.
24703         (verify_type): Remove dead code.
24704         * varasm.c (assemble_external): Remove Java from a comment.
24706 2017-07-14  Martin Liska  <mliska@suse.cz>
24708         * opts.c (finish_options): Add quotes.
24709         (common_handle_option): Likewise.
24711 2017-07-14  Martin Liska  <mliska@suse.cz>
24713         * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24714         * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24715         Remove N_SO_PASCAL.
24716         * dwarf2out.c (lower_bound_default): Do not handle
24717         DW_LANG_Pascal83.
24718         (gen_compile_unit_die): Likewise.
24719         * gcc.c: Remove default extension binding for GNU Pascal.
24720         * stmt.c: Remove Pascal language from a comment.
24721         * xcoffout.c: Likewise.
24723 2017-07-13  David Malcolm  <dmalcolm@redhat.com>
24725         PR c/81405
24726         * diagnostic-show-locus.c (fixit_cmp): New function.
24727         (layout::layout): Sort m_fixit_hints.
24728         (column_range::column_range): Assert that the values are valid.
24729         (struct char_span): New struct.
24730         (correction::overwrite): New method.
24731         (struct source_line): New struct.
24732         (line_corrections::add_hint): Add assertions.  Reimplement memcpy
24733         calls in terms of classes source_line and char_span, and
24734         correction::overwrite.
24735         (selftest::test_overlapped_fixit_printing_2): New function.
24736         (selftest::diagnostic_show_locus_c_tests): Call it.
24738 2017-07-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
24740         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24741         early if there is no lhs.
24743 2017-07-13  Martin Liska  <mliska@suse.cz>
24745         * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24746         (gen_reference_type_die): Likewise.
24747         * stor-layout.c: Remove Pascal-related comment.
24749 2017-07-13  Martin Liska  <mliska@suse.cz>
24751         * opts.c (finish_options): Add quotes to error messages.
24752         (parse_sanitizer_options): Likewise.
24754 2017-07-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24756         * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24758 2017-07-13  Richard Earnshaw  <rearnsha@arm.com>
24760         * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24762 2017-07-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
24764         * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24765         during expansion.
24766         * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24768 2017-07-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
24770         PR target/81193
24771         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24772         provides the hardware capability bits, define the macro
24773         __BUILTIN_CPU_SUPPORTS__.
24774         * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24775         if GLIBC does not provide the hardware capability bits.  Add a
24776         gcc_unreachable call if the built-in cpu function is neither
24777         __builtin_cpu_is nor __builtin_cpu_supports.
24778         (rs6000_get_function_versions_dispatcher): Change the warning
24779         that an old GLIBC is used which does not export the capability
24780         bits to be an error.
24781         * doc/extend.texi (target_clones attribute): Document the
24782         restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24783         (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24784         needed by __builtin_cpu_is and __builtin_cpu_supports.  Document
24785         the macros defined by GCC if the newer GLIBC is available.
24787 2017-07-12  Jeff Law  <law@redhat.com>
24789         * config/riscv/riscv.c: Remove unnecessary includes.  Reorder
24790         remaining includes slightly.
24791         * config/riscv/riscv-builtins.c: Include profile-count.h.
24793 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24795         PR target/79883
24796         * config/avr/avr.c (avr_set_current_function): In diagnostic
24797         messages: Quote keywords and (parts of) identifiers.
24798         [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24799         "INTERUPT".
24801 2017-07-12  Carl Love  <cel@us.ibm.com>
24803         * config/rs6000/rs6000-c.c: Add support for built-in functions
24804         vector bool char vec_revb (vector bool char);
24805         vector bool short vec_revb (vector short char);
24806         vector bool int vec_revb (vector bool int);
24807         vector bool long long vec_revb (vector bool long long);
24808         * doc/extend.texi: Update the built-in documentation file for the
24809         new built-in functions.
24811 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24813         * config/s390/s390.md: Remove movcc splitter.
24815 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24817         * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24818         load/store on condition.
24820 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24822         PR target/81407
24823         * config/avr/avr.c (avr_encode_section_info)
24824         [progmem && !TREE_READONLY]: Error if progmem object needs
24825         constructing.
24827 2017-07-11  Michael Collison  <michael.collison@arm.com>
24829         * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24830         New pattern.
24832 2017-07-11  Carl Love  <cel@us.ibm.com>
24834         * config/rs6000/rs6000-c.c: Add support for builtins
24835         vector unsigned int vec_parity_lsbb (vector signed int);
24836         vector unsigned int vec_parity_lsbb (vector unsigned int);
24837         vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24838         vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24839         vector unsigned long long vec_parity_lsbb (vector signed long long);
24840         vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24841         * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24842         * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24843         * doc/extend.texi: Update the built-in documentation file for the
24844         new built-in functions.
24846 2017-07-11  David Malcolm  <dmalcolm@redhat.com>
24848         * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24849         (layout::m_primary_loc): New field.
24850         (layout::layout): Initialize new field.  Move location filtering
24851         logic from here to...
24852         (layout::maybe_add_location_range): ...this new method.  Add
24853         support for filtering to just the lines already specified by other
24854         locations.
24855         (layout::will_show_line_p): New method.
24856         (gcc_rich_location::add_location_if_nearby): New method.
24857         (selftest::test_add_location_if_nearby): New test function.
24858         (selftest::diagnostic_show_locus_c_tests): Call it.
24859         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24860         New method.
24862 2017-07-11  Tom de Vries  <tom@codesourcery.com>
24864         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24865         (bb_first_real_insn): New function.
24866         (nvptx_single): Add extra initialization of broadcasted condition
24867         variables.
24869 2017-07-11  Nathan Sidwell  <nathan@acm.org>
24871         * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24873 2017-07-11  Georg-Johann Lay  <avr@gjlay.de>
24875         * doc/extend.texi (AVR Function Attributes): Remove weblink to
24876         Binutils doc as TEXI will mess them up.
24877         * doc/invoke.texi (AVR Options): Same here.
24879 2017-07-11  Daniel Cederman  <cederman@gaisler.com>
24881         * config/sparc/sparc.opt (mfix-ut700): New option.
24882         (mfix-gr712rc): Likewise.
24883         (sparc_fix_b2bst): New variable.
24884         * doc/invoke.texi (SPARC options): Document them.
24885         (ARM options): Fix warnings.
24886         * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24887         instructions to prevent sequences that can trigger the store-store
24888         errata for certain LEON3FT processors.
24889         (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24890         (sparc_option_override): Set sparc_fix_b2bst appropriately.
24891         * config/sparc/sparc.md (fix_b2bst): New attribute.
24892         (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24894 2017-07-10  Uros Bizjak  <ubizjak@gmail.com>
24896         PR target/81375
24897         * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
24898         (rcpps): Ditto.
24899         (*rsqrtsf2_sse): Ditto.
24900         (rsqrtsf2): Ditto.
24901         (div<mode>3): Macroize insn from divdf3 and divsf3
24902         using MODEF mode iterator.
24904 2017-07-10  Martin Sebor  <msebor@redhat.com>
24906         PR tree-optimization/80397
24907         * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
24908         instead of testing for equality to INTEGER_TYPE.
24910 2017-07-10  Vineet Gupta <vgupta@synopsys.com>
24912         * config.gcc: Remove uclibc from arc target spec.
24914 2017-07-10  Claudiu Zissulescu  <claziss@synopsys.com>
24916         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
24918 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24920         PR lto/80838
24921         * lto-wrapper.c (remove_option): New function.
24922         (merge_and_complain): Merge PIC/PIE options more realistically.
24924 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24926         Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
24928         PR target/20296
24929         PR target/81268
24930         * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
24931         (HAVE_AS_AVR_MGCCISR_OPTION):  If so, AC_DEFINE it.
24932         * config.in: Regenerate.
24933         * configure: Regenerate.
24934         * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
24935         * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
24936         * config/avr/avr.opt (-mgas-isr-prologues): New option and...
24937         (TARGET_GASISR_PROLOGUES): ...target mask.
24938         * common/config/avr/avr-common.c
24939         (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
24940         Set -mgas-isr-prologues.
24941         * config/avr/avr-passes.def (avr_pass_pre_proep): Add
24942         INSERT_PASS_BEFORE for it.
24943         * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
24944         * config/avr/avr.c (avr_option_override)
24945         [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
24946         (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
24947         (avr_attribute_table) <no_gccisr>: Add new function attribute.
24948         (avr_set_current_function) <is_no_gccisr>: Init machine field.
24949         (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
24950         and rtl_opt_pass.
24951         (make_avr_pass_pre_proep): New function.
24952         (emit_push_sfr) <treg>: Add argument to function and use it
24953         instead of TMP_REG.
24954         (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
24955         and set machine->gasisr.yes.
24956         (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
24957         (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
24958         __gcc_isr.n_pushed to .L__stack_usage.
24959         (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
24960         (avr_asm_final_postscan_insn): ...this new static function.
24961         * config/avr/avr.h (machine_function)
24962         <is_no_gccisr, use_L__stack_usage>: New fields.
24963         <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
24964         * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
24965         (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
24966         (gasisr, *gasisr): New expander and insn.
24967         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
24968         [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
24969         * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
24971 2017-07-10  Richard Earnshaw  <rearnsha@arm.com>
24973         * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
24974         in quoted strings.
24976 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24978         Move jump-tables out of .text again.
24980         PR target/81075
24981         * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
24982         (ASM_OUTPUT_ADDR_VEC): New function.
24983         (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
24984         (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
24985         INSN_ADDRESSes as asm comment.
24986         * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
24987         (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
24988         (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
24989         * config/avr/avr.md (*tablejump): Adjust comment.
24990         * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
24991         * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
24992         New detail.
24993         * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
24994         (avr_output_addr_vec): New proto.
24995         (avr_log_t) <insn_addresses>: New field.
24997 2017-07-09  H.J. Lu  <hongjiu.lu@intel.com>
24999         PR target/81313
25000         * config/i386/i386.c (ix86_function_arg_advance): Set
25001         outgoing_args_on_stack to true if there are outgoing arguments
25002         on stack.
25003         (ix86_function_arg): Likewise.
25004         (ix86_get_drap_rtx): Use DRAP only if there are outgoing
25005         arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
25006         * config/i386/i386.h (machine_function): Add
25007         outgoing_args_on_stack.
25009 2017-07-09  Krister Walfridsson  <krister.walfridsson@gmail.com>
25011         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
25012         supporting pthreds.
25013         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
25015 2017-07-08  Richard Sandiford  <richard.sandiford@linaro.org>
25017         * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
25018         (REAL_H): Remove $(MACHMODE_H).
25019         (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
25020         double-int.h.
25021         (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
25022         $(MACHMODE_H) and double-int.h.
25023         (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
25024         $(MACHMODE_H).
25025         (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
25026         double-int.h.
25028 2017-07-07  Andrew Pinski  <apinski@cavium.com>
25030         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
25031         prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
25033 2017-07-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
25035         * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
25036         Add warning if GCC was not configured to link against a GLIBC that
25037         exports the hardware capability bits.
25038         (make_resolver_func): Make resolver function private and not a
25039         COMDAT function.  Create the name with clone_function_name instead
25040         of make_unique_name.
25042         PR target/81348
25043         * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
25044         correct operand in doing the split.
25046 2017-07-07 Carl Love  <cel@us.ibm.com>
25048         * config/rs6000/rs6000-c: Add support for built-in function
25049         vector unsigned short vec_pack_to_short_fp32 (vector float,
25050                                                       vector float).
25051         * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
25052         BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
25053         * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
25054         * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
25055         (convert_4f32_8i16): Add define_expand.
25056         * doc/extend.texi: Update the built-in documentation file for the
25057         new built-in function.
25059 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25061         * config/sparc/m8.md: New file.
25062         * config/sparc/sparc.md: Include m8.md.
25064 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25066         * config/sparc/sparc.opt: New option -mvis4b.
25067         * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
25068         (sparc_option_override): Handle VIS4B.
25069         (enum sparc_builtins): Define
25070         SPARC_BUILTIN_DICTUNPACK{8,16,32},
25071         SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
25072         SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
25073         SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
25074         SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
25075         (check_constant_argument): New function.
25076         (sparc_vis_init_builtins): Define builtins
25077         __builtin_vis_dictunpack{8,16,32},
25078         __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
25079         __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
25080         __builtin_vis_fpcmpde{8,16,32}shl and
25081         __builtin_vis_fpcmpur{8,16,32}shl.
25082         (sparc_expand_builtin): Check that the constant operands to
25083         __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
25084         constant and in range.
25085         * config/sparc/sparc-c.c (sparc_target_macros): Handle
25086         TARGET_VIS4B.
25087         * config/sparc/sparc.h (SPARC_IMM2_P): Define.
25088         (SPARC_IMM5_P): Likewise.
25089         * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
25090         (enabled): Handle vis4b.
25091         (UNSPEC_DICTUNPACK): New unspec.
25092         (UNSPEC_FPCMPSHL): Likewise.
25093         (UNSPEC_FPUCMPSHL): Likewise.
25094         (UNSPEC_FPCMPDESHL): Likewise.
25095         (UNSPEC_FPCMPURSHL): Likewise.
25096         (cpu_feature): New CPU feature `vis4b'.
25097         (dictunpack{8,16,32}): New insns.
25098         (FPCSMODE): New mode iterator.
25099         (fpcscond): New code iterator.
25100         (fpcsucond): Likewise.
25101         (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
25102         (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
25103         (fpcmpde{8,16,32}{si,di}shl): Likewise.
25104         (fpcmpur{8,16,32}{si,di}shl): Likewise.
25105         * config/sparc/constraints.md: Define constraints `q' for unsigned
25106         2-bit integer constants and `t' for unsigned 5-bit integer
25107         constants.
25108         * config/sparc/predicates.md (imm5_operand_dictunpack8): New
25109         predicate.
25110         (imm5_operand_dictunpack16): Likewise.
25111         (imm5_operand_dictunpack32): Likewise.
25112         (imm2_operand): Likewise.
25113         * doc/invoke.texi (SPARC Options): Document -mvis4b.
25114         * doc/extend.texi (SPARC VIS Built-in Functions): Document the
25115         ditunpack* and fpcmp*shl builtins.
25117 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25119         * config.gcc: Handle m8 in --with-{cpu,tune} options.
25120         * config.in: Add HAVE_AS_SPARC6 define.
25121         * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
25122         M8.
25123         * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
25124         TARGET_CPU_m8.
25125         (ASM_CPU32_DEFAUILT_SPEC): Likewise.
25126         (CPP_CPU_SPEC): Handle m8.
25127         (ASM_CPU_SPEC): Likewise.
25128         * config/sparc/sparc-opts.h (enum processor_type): Add
25129         PROCESSOR_M8.
25130         * config/sparc/sparc.c (m8_costs): New struct.
25131         (sparc_option_override): Handle TARGET_CPU_m8.
25132         (sparc32_initialize_trampoline): Likewise.
25133         (sparc64_initialize_trampoline): Likewise.
25134         (sparc_issue_rate): Likewise.
25135         (sparc_register_move_cost): Likewise.
25136         * config/sparc/sparc.h (TARGET_CPU_m8): Define.
25137         (CPP_CPU64_DEFAULT_SPEC): Define for M8.
25138         (ASM_CPU64_DEFAULT_SPEC): Likewise.
25139         (CPP_CPU_SPEC): Handle M8.
25140         (ASM_CPU_SPEC): Likewise.
25141         (AS_M8_FLAG): Define.
25142         * config/sparc/sparc.md: Add m8 to the cpu attribute.
25143         * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
25144         * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
25145         M8 instructions.
25146         * configure: Regenerate.
25147         * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
25148         -mtune=m8.
25150 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25152         * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
25153         subtypes.
25154         * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
25155         ("*movdi_insn_sp32"): Do not set v3pipe.
25156         ("*movsi_insn"): Likewise.
25157         ("*movdi_insn_sp64"): Likewise.
25158         ("*movsf_insn"): Likewise.
25159         ("*movdf_insn_sp32"): Likewise.
25160         ("*movdf_insn_sp64"): Likewise.
25161         ("*zero_extendsidi2_insn_sp64"): Likewise.
25162         ("*sign_extendsidi2_insn"): Likewise.
25163         ("*mov<VM32:mode>_insn"): Likewise.
25164         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25165         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25166         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25167         ("<vlop:code><VL:mode>3"): Likewise.
25168         ("*not_<vlop:code><VL:mode>3"): Likewise.
25169         ("*nand<VL:mode>_vis"): Likewise.
25170         ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
25171         ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
25172         ("one_cmpl<VL:mode>2"): Likewise.
25173         ("faligndata<VM64:mode>_vis"): Likewise.
25174         ("alignaddrsi_vis"): Likewise.
25175         ("alignaddrdi_vis"): Likweise.
25176         ("alignaddrlsi_vis"): Likewise.
25177         ("alignaddrldi_vis"): Likewise.
25178         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25179         ("bmaskdi_vis"): Likewise.
25180         ("bmasksi_vis"): Likewise.
25181         ("bshuffle<VM64:mode>_vis"): Likewise.
25182         ("cmask8<P:mode>_vis"): Likewise.
25183         ("cmask16<P:mode>_vis"): Likewise.
25184         ("cmask32<P:mode>_vis"): Likewise.
25185         ("pdistn<P:mode>_vis"): Likewise.
25186         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25188 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25190         * config/sparc/sparc.md ("subtype"): New insn attribute.
25191         ("*wrgsr_sp64"): Set insn subtype.
25192         ("*rdgsr_sp64"): Likewise.
25193         ("alignaddrsi_vis"): Likewise.
25194         ("alignaddrdi_vis"): Likewise.
25195         ("alignaddrlsi_vis"): Likewise.
25196         ("alignaddrldi_vis"): Likewise.
25197         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25198         ("fexpand_vis"): Likewise.
25199         ("fpmerge_vis"): Likewise.
25200         ("faligndata<VM64:mode>_vis"): Likewise.
25201         ("bshuffle<VM64:mode>_vis"): Likewise.
25202         ("cmask8<P:mode>_vis"): Likewise.
25203         ("cmask16<P:mode>_vis"): Likewise.
25204         ("cmask32<P:mode>_vis"): Likewise.
25205         ("fchksm16_vis"): Likewise.
25206         ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
25207         ("fmean16_vis"): Likewise.
25208         ("fp<plusminus_insn>64_vis"): Likewise.
25209         ("<plusminus_insn>v8qi3"): Likewise.
25210         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25211         ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
25212         ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
25213         ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
25214         ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
25215         ("*movqi_insn"): Likewise.
25216         ("*movhi_insn"): Likewise.
25217         ("*movsi_insn"): Likewise.
25218         ("movsi_pic_gotdata_op"): Likewise.
25219         ("*movdi_insn_sp32"): Likewise.
25220         ("*movdi_insn_sp64"): Likewise.
25221         ("movdi_pic_gotdata_op"): Likewise.
25222         ("*movsf_insn"): Likewise.
25223         ("*movdf_insn_sp32"): Likewise.
25224         ("*movdf_insn_sp64"): Likewise.
25225         ("*zero_extendhisi2_insn"): Likewise.
25226         ("*zero_extendqihi2_insn"): Likewise.
25227         ("*zero_extendqisi2_insn"): Likewise.
25228         ("*zero_extendqidi2_insn"): Likewise.
25229         ("*zero_extendhidi2_insn"): Likewise.
25230         ("*zero_extendsidi2_insn_sp64"): Likewise.
25231         ("ldfsr"): Likewise.
25232         ("prefetch_64"): Likewise.
25233         ("prefetch_32"): Likewise.
25234         ("tie_ld32"): Likewise.
25235         ("tie_ld64"): Likewise.
25236         ("*tldo_ldub_sp32"): Likewise.
25237         ("*tldo_ldub1_sp32"): Likewise.
25238         ("*tldo_ldub2_sp32"): Likewise.
25239         ("*tldo_ldub_sp64"): Likewise.
25240         ("*tldo_ldub1_sp64"): Likewise.
25241         ("*tldo_ldub2_sp64"): Likewise.
25242         ("*tldo_ldub3_sp64"): Likewise.
25243         ("*tldo_lduh_sp32"): Likewise.
25244         ("*tldo_lduh1_sp32"): Likewise.
25245         ("*tldo_lduh_sp64"): Likewise.
25246         ("*tldo_lduh1_sp64"): Likewise.
25247         ("*tldo_lduh2_sp64"): Likewise.
25248         ("*tldo_lduw_sp32"): Likewise.
25249         ("*tldo_lduw_sp64"): Likewise.
25250         ("*tldo_lduw1_sp64"): Likewise.
25251         ("*tldo_ldx_sp64"): Likewise.
25252         ("*mov<VM32:mode>_insn"): Likewise.
25253         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25254         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25256 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25258         * config/sparc/sparc.md ("type"): New insn type viscmp.
25259         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25260         viscmp.
25261         ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25262         ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25263         ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25264         * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25265         viscmp.
25266         ("n7_vis_logical_11cycle"): Likewise.
25267         * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25268         * config/sparc/niagara2.md ("niag3_vis": Likewise.
25269         * config/sparc/niagara.md ("niag_vis"): Likewise.
25270         * config/sparc/ultra3.md ("us3_fga"): Likewise.
25271         * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25273 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25275         * config/sparc/sparc.md: New instruction type `bmask'.
25276         (bmaskdi_vis): Use the `bmask' type.
25277         (bmasksi_vis): Likewise.
25278         * config/sparc/ultra3.md (us3_array): Likewise.
25279         * config/sparc/niagara7.md (n7_array): Likewise.
25280         * config/sparc/niagara4.md (n4_array): Likewise.
25281         * config/sparc/niagara2.md (niag2_vis): Likewise.
25282         (niag3_vis): Likewise.
25283         * config/sparc/niagara.md (niag_vis): Likewise.
25285 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25287         * ipa-comdats.c: Remove optimize check from gate.
25288         * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25289         for functions not optimized.
25290         (ipa_fn_summary_read): Skip optimize check.
25291         (ipa_fn_summary_write): Likewise.
25292         * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25293         is optimized.
25294         * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25295         uninlinable.
25296         (can_inline_edge_p): Check flag_pcc_struct_return for match.
25297         (check_callers): Give up on caller which is not optimized.
25298         (inline_small_functions): Likewise.
25299         (ipa_inline): Do not give up when not optimizing.
25300         * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25301         away unoptimizes cdtors.
25302         (whole_program_function_and_variable_visibility): Do
25303         ipa_discover_readonly_nonaddressable_vars in LTO mode.
25304         * ipa.c (process_references): Do not check optimize.
25305         (symbol_table::remove_unreachable_nodes): Update optimize check.
25306         (set_writeonly_bit): Update optimize check.
25307         (pass_ipa_cdtor_merge::gate): Do not check optimize.
25308         (pass_ipa_single_use::gate): Remove.
25310 2017-07-06  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25312         * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25313         insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25314         rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25315         mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25316         permute_load, permute_store, adjust_extract, adjust_splat,
25317         adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25318         replace_swap_with_copy, dump_swap_insn_table,
25319         alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25320         recombine_lvx_pattern, recombine_stvx_pattern,
25321         recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25322         make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25323         to file rs6000-p8swap.c.
25324         * config/rs6000/rs6000-p8swap.c: New file.
25325         * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25326         * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25327         and rs6000*-*-* targets.
25329 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25331         * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25333 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25335         * lto-wrapper.c (merge_and_complain): Do not merge
25336         fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25337         fsigned_zeros, ftrapping_math, fwrapv.
25338         (append_compiler_options): Do not track these options.
25339         (append_linker_options): Likewie
25341 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25343         * cgraphunit.c (cgraph_node::finalize_function): When
25344         !flag_toplevel_reorde set no_reorder flag.
25345         (varpool_node::finalize_decl): Likewise.
25346         (symbol_table::compile): Drop no toplevel reorder path.
25348 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25350         * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25351         edges; zero probability is not better than uninitialized.
25353 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25355         * asan.h (asan_sanitize_allocas_p): Declare.
25356         * asan.c (asan_sanitize_allocas_p): New function.
25357         (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25358         (handle_builtin_alloca): Likewise.
25359         * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25360         if !asan_sanitize_allocas_p.
25361         * params.def (asan-instrument-allocas): Add new option.
25362         * params.h (ASAN_PROTECT_ALLOCAS): Define.
25363         * opts.c (common_handle_option): Disable allocas sanitization for
25364         KASan by default.
25366 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25368         * asan.c: Include gimple-fold.h.
25369         (get_last_alloca_addr): New function.
25370         (handle_builtin_stackrestore): Likewise.
25371         (handle_builtin_alloca): Likewise.
25372         (asan_emit_allocas_unpoison): Likewise.
25373         (get_mem_refs_of_builtin_call): Add new parameter, remove const
25374         quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25375         BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25376         (instrument_builtin_call): Pass gimple iterator to
25377         get_mem_refs_of_builtin_call.
25378         (last_alloca_addr): New global.
25379         * asan.h (asan_emit_allocas_unpoison): Declare.
25380         * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25381         (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25382         * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25383         if function calls alloca.
25384         * gimple-fold.c (replace_call_with_value): Remove static keyword.
25385         * gimple-fold.h (replace_call_with_value): Declare.
25386         * internal-fn.c: Include asan.h.
25387         * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25388         BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25390 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25392         * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25393         (C_SELFTEST_FLAGS): New.
25394         (CPP_SELFTEST_FLAGS): New.
25395         (SELFTEST_DEPS): New, from deps of s-selftest.
25396         (C_SELFTEST_DEPS): New, from deps of s-selftest.
25397         (CPP_SELFTEST_DEPS): New.
25398         (selftest): Add dependency on s-selftest-c++.
25399         (s-selftest): Rename to...
25400         (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25401         and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25402         than SELFTEST_FLAGS.
25403         (selftest-gdb): Rename to...
25404         (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25405         C_SELFTEST_FLAGS.
25406         (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25407         (selftest-valgrind): Rename to...
25408         (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25409         C_SELFTEST_FLAGS.
25410         (selftest-valgrind): Reintroduce as an alias for
25411         selftest-c-valgrind.
25412         (s-selftest-c++): New.
25413         (selftest-c++-gdb): New.
25414         (selftest-c++-valgrind): New.
25416 2017-07-06  Olivier Hainque  <hainque@adacore.com>
25418         * gcc.c (process_command): When deciding if undefined variables
25419         should be ignored when processing specs, accept "gcc -v" as well.
25421 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25423         * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25424         afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25426 2017-07-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25428         * config/arm/arm-cpus.in (armv8-r): Add new entry.
25429         * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25430         * config/arm/arm-tables.opt: Regenerate.
25431         * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25432         enumerator.
25433         * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25435 2017-07-06  Carl Love  <cel@us.ibm.com>
25437         * ChangeLog: Clean up from mid air collision
25439 2017-07-06  Carl Love  <cel@us.ibm.com>
25441         * config/rs6000/rs6000-c.c: Add support for built-in functions
25442         vector signed int vec_subc (vector signed int, vector signed int);
25443         vector signed __int128 vec_subc (vector signed __int128,
25444                                          vector signed __int128);
25445         vector unsigned __int128 vec_subc (vector unsigned __int128,
25446                                            vector unsigned __int128);
25447         vector signed int vec_sube (vector signed int, vector signed int,
25448                                     vector signed int);
25449         vector unsigned int vec_sube (vector unsigned int,
25450                                       vector unsigned int,
25451                                       vector unsigned int);
25452         vector signed __int128 vec_sube (vector signed __int128,
25453                                          vector signed __int128,
25454                                          vector signed__int128);
25455         vector unsigned __int128 vec_sube (vector unsigned __int128,
25456                                            vector unsigned __int128,
25457                                            vector unsigned __int128);
25458         vector signed int vec_subec (vector signed int, vector signed int,
25459                                      vector signed int);
25460         vector unsigned int vec_subec (vector unsigned int,
25461                                        vector unsigned int,
25462                                        vector unsigned int);
25463         vector signed __int128 vec_subec (vector signed __int128,
25464                                           vector signed __int128,
25465                                           vector signed__int128);
25466         vector unsigned __int128 vec_subec (vector unsigned __int128,
25467                                             vector unsigned __int128,
25468                                             vector unsigned __int128);
25469         * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25470         ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25471         * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25472         BU_ALTIVEC_OVERLOAD_X definitions.
25473         * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25474         * doc/extend.texi: Update the built-in documentation file for the new
25475         built-in functions.
25477 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25479         PR c++/79300
25480         * diagnostic-show-locus.c (layout::layout): Use start and finish
25481         spelling location for the start and finish of each range.
25482         * genmatch.c (linemap_client_expand_location_to_spelling_point):
25483         Add unused aspect param.
25484         * input.c (expand_location_1): Add "aspect" param, and use it
25485         to access the correct part of the location.
25486         (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25487         expand_location_1.
25488         (expand_location_to_spelling_point): Likewise.
25489         (linemap_client_expand_location_to_spelling_point): Add "aspect"
25490         param, and pass it to expand_location_1.
25492 2017-07-06  Sebastian Peryt  <sebastian.peryt@intel.com>
25494         * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25495         _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25496         _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25497         _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25498         _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25499         _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25500         _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25501         _mm_maskz_getmant_ss): New intrinsics.
25502         (__builtin_ia32_getexpss128_mask): Changed to ...
25503         __builtin_ia32_getexpss128_round ... this.
25504         (__builtin_ia32_getexpsd128_mask): Changed to ...
25505         __builtin_ia32_getexpsd128_round ... this.
25506         * config/i386/i386-builtin-types.def
25507         ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25508         (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25509         * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25510         __builtin_ia32_getexpss_mask_round,     __builtin_ia32_getmantsd_mask_round,
25511         __builtin_ia32_getmantss_mask_round): New builtins.
25512         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25513         V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25514         (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25515         CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25516         * config/i386/sse.md
25517         (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25518         avx512f_sgetexp<mode><mask_scalar_name>
25519         <round_saeonly_scalar_name> ... this.
25520         (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25521         %0, %1, %2<round_saeonly_op3>}): Changed to ...
25522         vgetexp<ssescalarmodesuffix>
25523         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25524         %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25525         (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25526         avx512f_vgetmant<mode><mask_scalar_name>
25527         <round_saeonly_scalar_name> ... this.
25528         (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25529         %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25530         vgetmant<ssescalarmodesuffix>
25531         \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25532         %0<mask_scalar_operand4>, %1, %2
25533         <round_saeonly_scalar_mask_op4>, %3} ... this.
25534         * config/i386/subst.md (mask_scalar_operand4,
25535         round_saeonly_scalar_mask_operand4,     round_saeonly_scalar_mask_op4,
25536         round_saeonly_scalar_nimm_predicate): New subst attributes.
25538 2017-07-06  Julia Koval  <julia.koval@intel.com>
25540         * config/i386/i386.c (ix86_erase_embedded_rounding):
25541         Remove code for old rounding pattern.
25543 2017-07-06  Richard Earnshaw  <rearnsha@arm.com>
25545         * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25547 2017-07-06  Christophe Lyon  <christophe.lyon@linaro.org>
25549         * doc/sourcebuild.texi (Test Directives, Variants of
25550         dg-require-support): Add documentation for dg-require-stack-check.
25552 2017-07-05  Sebastian Peryt  <sebastian.peryt@intel.com>
25554         * config/i386/subst.md (mask_scalar, round_scalar,
25555         round_saeonly_scalar): New meta-templates.
25556         (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25557         round_scalar_mask_operand3, round_scalar_mask_op3,
25558         round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25559         round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25560         round_saeonly_scalar_constraint,
25561         round_saeonly_scalar_prefix): New subst attribute.
25562         * config/i386/sse.md
25563         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25564         <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25565         <round_scalar_name> ... this.
25566         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25567         <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25568         <round_scalar_name> ... this.
25569         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25570         <sse>_vm<code><mode>3<mask_scalar_name>
25571         <round_saeonly_scalar_name> ... this.
25572         (v<plusminus_mnemonic><ssescalarmodesuffix>
25573         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25574         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25575         v<plusminus_mnemonic><ssescalarmodesuffix>
25576         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25577         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25578         (v<multdiv_mnemonic><ssescalarmodesuffix>
25579         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25580         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25581         v<multdiv_mnemonic><ssescalarmodesuffix>
25582         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25583         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25584         (v<maxmin_float><ssescalarmodesuffix>
25585         \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25586         %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25587         v<maxmin_float><ssescalarmodesuffix>
25588         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25589         %0<mask_scalar_operand3>, %1, %<iptr>2
25590         <round_saeonly_scalar_mask_op3>} ... this.
25592 2017-07-05  Richard Earnshaw  <rearnsha@arm.com>
25594         * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25595         (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25597 2017-07-05  Richard Sandiford  <richard.sandiford@linaro.org>
25598             Alan Hayward  <alan.hayward@arm.com>
25599             David Sherwood  <david.sherwood@arm.com>
25601         * combine.c (simplify_if_then_else): Remove "enum" before
25602         "machine_mode".
25603         * compare-elim.c (can_eliminate_compare): Likewise.
25604         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25605         Likewise.
25606         (aarch64_lookup_simd_builtin_type): Likewise.
25607         (aarch64_simd_builtin_type): Likewise.
25608         (aarch64_init_simd_builtin_types): Likewise.
25609         (aarch64_simd_expand_args): Likewise.
25610         * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25611         Likewise.
25612         (aarch64_reverse_mask): Likewise.
25613         (aarch64_simd_emit_reg_reg_move): Likewise.
25614         (aarch64_gen_adjusted_ldpstp): Likewise.
25615         (aarch64_ccmp_mode_to_code): Likewise.
25616         (aarch64_operands_ok_for_ldpstp): Likewise.
25617         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25618         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25619         Likewise.
25620         (aarch64_min_divisions_for_recip_mul): Likewise.
25621         (aarch64_reassociation_width): Likewise.
25622         (aarch64_get_condition_code_1): Likewise.
25623         (aarch64_simd_emit_reg_reg_move): Likewise.
25624         (aarch64_simd_attr_length_rglist): Likewise.
25625         (aarch64_reverse_mask): Likewise.
25626         (aarch64_operands_ok_for_ldpstp): Likewise.
25627         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25628         (aarch64_gen_adjusted_ldpstp): Likewise.
25629         * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25630         Likewise.
25631         * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25632         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25633         (arm_lookup_simd_builtin_type): Likewise.
25634         (arm_simd_builtin_type): Likewise.
25635         (arm_init_simd_builtin_types): Likewise.
25636         (arm_expand_builtin_args): Likewise.
25637         * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25638         * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25639         (ft32_setup_incoming_varargs): Likewise.
25640         (ft32_function_arg): Likewise.
25641         (ft32_function_arg_advance): Likewise.
25642         (ft32_pass_by_reference): Likewise.
25643         (ft32_arg_partial_bytes): Likewise.
25644         (ft32_valid_pointer_mode): Likewise.
25645         (ft32_addr_space_pointer_mode): Likewise.
25646         (ft32_addr_space_legitimate_address_p): Likewise.
25647         * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25648         Likewise.
25649         * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25650         (ix86_emit_outlined_ms2sysv_restore): Likewise.
25651         (iamcu_alignment): Likewise.
25652         (canonicalize_vector_int_perm): Likewise.
25653         (ix86_noce_conversion_profitable_p): Likewise.
25654         (ix86_mpx_bound_mode): Likewise.
25655         (ix86_operands_ok_for_move_multiple): Likewise.
25656         * config/microblaze/microblaze-protos.h
25657         (microblaze_expand_conditional_branch_reg): Likewise.
25658         * config/microblaze/microblaze.c
25659         (microblaze_expand_conditional_branch_reg): Likewise.
25660         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25661         Likewise.
25662         (rs6000_reassociation_width): Likewise.
25663         (rs6000_invalid_binary_op): Likewise.
25664         (fusion_p9_p): Likewise.
25665         (emit_fusion_p9_load): Likewise.
25666         (emit_fusion_p9_store): Likewise.
25667         * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25668         Likewise.
25669         (riscv_hard_regno_mode_ok_p): Likewise.
25670         (riscv_address_insns): Likewise.
25671         (riscv_split_symbol): Likewise.
25672         (riscv_legitimize_move): Likewise.
25673         (riscv_function_value): Likewise.
25674         (riscv_hard_regno_nregs): Likewise.
25675         (riscv_expand_builtin): Likewise.
25676         * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25677         (riscv_build_integer): Likewise.
25678         (riscv_split_integer): Likewise.
25679         (riscv_legitimate_constant_p): Likewise.
25680         (riscv_cannot_force_const_mem): Likewise.
25681         (riscv_regno_mode_ok_for_base_p): Likewise.
25682         (riscv_valid_base_register_p): Likewise.
25683         (riscv_valid_offset_p): Likewise.
25684         (riscv_valid_lo_sum_p): Likewise.
25685         (riscv_classify_address): Likewise.
25686         (riscv_legitimate_address_p): Likewise.
25687         (riscv_address_insns): Likewise.
25688         (riscv_load_store_insns): Likewise.
25689         (riscv_force_binary): Likewise.
25690         (riscv_split_symbol): Likewise.
25691         (riscv_force_address): Likewise.
25692         (riscv_legitimize_address): Likewise.
25693         (riscv_move_integer): Likewise.
25694         (riscv_legitimize_const_move): Likewise.
25695         (riscv_legitimize_move): Likewise.
25696         (riscv_address_cost): Likewise.
25697         (riscv_subword): Likewise.
25698         (riscv_output_move): Likewise.
25699         (riscv_canonicalize_int_order_test): Likewise.
25700         (riscv_emit_int_order_test): Likewise.
25701         (riscv_function_arg_boundary): Likewise.
25702         (riscv_pass_mode_in_fpr_p): Likewise.
25703         (riscv_pass_fpr_single): Likewise.
25704         (riscv_pass_fpr_pair): Likewise.
25705         (riscv_get_arg_info): Likewise.
25706         (riscv_function_arg): Likewise.
25707         (riscv_function_arg_advance): Likewise.
25708         (riscv_arg_partial_bytes): Likewise.
25709         (riscv_function_value): Likewise.
25710         (riscv_pass_by_reference): Likewise.
25711         (riscv_setup_incoming_varargs): Likewise.
25712         (riscv_print_operand): Likewise.
25713         (riscv_elf_select_rtx_section): Likewise.
25714         (riscv_save_restore_reg): Likewise.
25715         (riscv_for_each_saved_reg): Likewise.
25716         (riscv_register_move_cost): Likewise.
25717         (riscv_hard_regno_mode_ok_p): Likewise.
25718         (riscv_hard_regno_nregs): Likewise.
25719         (riscv_class_max_nregs): Likewise.
25720         (riscv_memory_move_cost): Likewise.
25721         * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25722         * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25723         (rl78_addr_space_address_mode): Likewise.
25724         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25725         Likewise.
25726         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25727         (rs6000_reassociation_width): Likewise.
25728         (rs6000_invalid_binary_op): Likewise.
25729         (fusion_p9_p): Likewise.
25730         (emit_fusion_p9_load): Likewise.
25731         (emit_fusion_p9_store): Likewise.
25732         * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25733         (ok_for_simple_move_operands): Likewise.
25734         (ok_for_simple_move_strict_operands): Likewise.
25735         (ok_for_simple_arith_logic_operands): Likewise.
25736         (visium_legitimize_reload_address): Likewise.
25737         (visium_select_cc_mode): Likewise.
25738         (output_cbranch): Likewise.
25739         (visium_split_double_move): Likewise.
25740         (visium_expand_copysign): Likewise.
25741         (visium_expand_int_cstore): Likewise.
25742         (visium_expand_fp_cstore): Likewise.
25743         * config/visium/visium.c (visium_pass_by_reference): Likewise.
25744         (visium_function_arg): Likewise.
25745         (visium_function_arg_advance): Likewise.
25746         (visium_libcall_value): Likewise.
25747         (visium_setup_incoming_varargs): Likewise.
25748         (visium_legitimate_constant_p): Likewise.
25749         (visium_legitimate_address_p): Likewise.
25750         (visium_legitimize_address): Likewise.
25751         (visium_secondary_reload): Likewise.
25752         (visium_register_move_cost): Likewise.
25753         (visium_memory_move_cost): Likewise.
25754         (prepare_move_operands): Likewise.
25755         (ok_for_simple_move_operands): Likewise.
25756         (ok_for_simple_move_strict_operands): Likewise.
25757         (ok_for_simple_arith_logic_operands): Likewise.
25758         (visium_function_value_1): Likewise.
25759         (rtx_ok_for_offset_p): Likewise.
25760         (visium_legitimize_reload_address): Likewise.
25761         (visium_split_double_move): Likewise.
25762         (visium_expand_copysign): Likewise.
25763         (visium_expand_int_cstore): Likewise.
25764         (visium_expand_fp_cstore): Likewise.
25765         (visium_split_cstore): Likewise.
25766         (visium_select_cc_mode): Likewise.
25767         (visium_split_cbranch): Likewise.
25768         (output_cbranch): Likewise.
25769         (visium_print_operand_address): Likewise.
25770         * expmed.c (flip_storage_order): Likewise.
25771         * expmed.h (emit_cstore): Likewise.
25772         (flip_storage_order): Likewise.
25773         * genrecog.c (validate_pattern): Likewise.
25774         * hsa-gen.c (gen_hsa_addr): Likewise.
25775         * internal-fn.c (expand_arith_overflow): Likewise.
25776         * ira-color.c (allocno_copy_cost_saving): Likewise.
25777         * lra-assigns.c (find_hard_regno_for_1): Likewise.
25778         * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25779         (process_invariant_for_inheritance): Likewise.
25780         * lra-eliminations.c (move_plus_up): Likewise.
25781         * omp-low.c (lower_oacc_reductions): Likewise.
25782         * simplify-rtx.c (simplify_subreg): Likewise.
25783         * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25784         (TARGET_CHKP_BOUND_MODE): Likewise..
25785         * targhooks.c (default_chkp_bound_mode): Likewise.
25786         (default_setup_incoming_vararg_bounds): Likewise.
25787         * targhooks.h (default_chkp_bound_mode): Likewise.
25788         (default_setup_incoming_vararg_bounds): Likewise.
25789         * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25790         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25791         (have_whole_vector_shift): Likewise.
25792         * tree-vect-stmts.c (vectorizable_load): Likewise.
25793         * doc/tm.texi: Regenerate.
25795 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25797         Graceful degrade if Binutils PR21472 is not available.
25799         PR target/81072
25800         * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25801         .rodata in flash test fails.
25802         (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25803         * confgure: Regenerate.
25804         * config.in: Regenerate.
25805         * config/avr/avr.c (avr_asm_named_section)
25806         [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25807         __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25808         (avr_asm_init_sections): Same.
25810 2017-07-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25812         * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25813         (fma<VH:mode>4_intrinsic): Likewise.
25814         (*fmsub<VCVTF:mode>4): Likewise.
25815         (*fmsub<VH:mode>4_intrinsic): Likewise.
25817 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25819         PR target/81305
25820         * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25821         Don't depend on "optimize > 0".
25822         (out_movhi_r_mr, out_movqi_mr_r): Same.
25823         (out_movhi_mr_r, out_movqi_r_mr): Same.
25824         (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25825         io_address_operand on "optimize > 0".
25827 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25829         * tree-loop-distribution.c: Add general explanantion on the pass.
25830         (generate_loops_for_partition): Mark distributed loop.
25831         (pg_add_dependence_edges): New parameter.  Handle alias data
25832         dependence specially and record it in the parameter if asked.
25833         (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25834         (init_partition_graph_vertices, add_partition_graph_edge): New.
25835         (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25836         (free_partition_graph_vdata, build_partition_graph): New.
25837         (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25838         (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25839         (data_ref_segment_size, latch_dominated_by_data_ref): New.
25840         (compute_alias_check_pairs, version_loop_by_alias_check): New.
25841         (version_for_distribution_p, finalize_partitions): New.
25842         (distribute_loop): Handle alias data dependence specially.  Factor
25843         out loop fusion code as functions and call these functions.
25845 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25847         * tree-loop-distribution.c (classify_partition): New parameter and
25848         better handle reduction statement.
25849         (rdg_build_partitions): Revise comment.
25850         (distribute_loop): Compute statements in all partitions and pass it
25851         to classify_partition.
25853 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25855         * tree-loop-distribution.c (enum partition_type): New.
25856         (struct partition): New field type.
25857         (partition_merge_into): Add parameter.  Update partition type.
25858         (data_dep_in_cycle_p, update_type_for_merge): New functions.
25859         (build_rdg_partition_for_vertex): Compute partition type.
25860         (rdg_build_partitions): Dump partition type.
25861         (distribute_loop): Update calls to partition_merge_into.
25863 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25865         * tree-loop-distribution.c (struct ddr_hasher): New.
25866         (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25867         (ddrs_table): New.
25868         (classify_partition): Call get_data_dependence.
25869         (pg_add_dependence_edges): Ditto.
25870         (distribute_loop): Release data dependence hash table.
25872 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25874         * tree-loop-distribution.c (ref_base_address): Delete.
25875         (similar_memory_accesses): Rename ...
25876         (share_memory_accesses): ... to this.  Check if partitions access
25877         the same memory reference.
25878         (distribute_loop): Call share_memory_accesses.
25880 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25882         * tree-loop-distribution.c (struct partition): New field recording
25883         its data reference.
25884         (partition_alloc, partition_free): Init and release data refs.
25885         (partition_merge_into): Merge data refs.
25886         (build_rdg_partition_for_vertex): Collect data refs for partition.
25887         (pg_add_dependence_edges): Change parameters from vector to bitmap.
25888         Update uses.
25889         (distribute_loop): Remve data refs from vertice data of partition
25890         graph.
25892 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25894         * tree-loop-distribution.c (params.h): Include header file.
25895         (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25896         (datarefs_vec): New global var.
25897         (create_rdg_vertices): Use datarefs_vec directly.
25898         (free_rdg): Don't free data references.
25899         (build_rdg): Update use.  Don't free data references.
25900         (distribute_loop): Compute global variable for data references.
25901         Bail out if there are too many data references.
25903 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25905         * tree-loop-distribution.c (loop_nest): New global var.
25906         (build_rdg): Use loop directly, rather than loop nest.
25907         (pg_add_dependence_edges): Remove loop nest parameter.  Use global
25908         variable directly.
25909         (distribute_loop): Compute global variable loop nest.  Update use.
25911 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25913         * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
25914         (partition_merge_into): New parameter.  Dump reason for fusion.
25915         (distribute_loop): Update use of partition_merge_into.
25917 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25919         * tree-loop-distribution.c (bb_top_order_index): New.
25920         (bb_top_order_index_size, bb_top_order_cmp): New.
25921         (stmts_from_loop): Use topological order.
25922         (pass_loop_distribution::execute): Compute and release topological
25923         order for basic blocks.
25925 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25927         * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
25928         if no loops.
25930 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25932         * cfgloop.h (struct loop): Add comment.  New field orig_loop_num.
25933         * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
25934         * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
25935         * internal-fn.def (LOOP_DIST_ALIAS): New.
25936         * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
25937         (fold_loop_internal_call): ... this.
25938         (vect_loop_dist_alias_call): New function.
25939         (set_uid_loop_bbs): Call fold_loop_internal_call.
25940         (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
25941         internal calls.
25943 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25945         PR target/81300
25946         * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
25947         Require dead FLAGS_REG at the beginning of a peephole.
25949 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25951         PR target/81294
25952         * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
25953         arguments in the call to __builtin_ia32_sbb_u32.
25954         (_subborrow_u64): Swap _X and _Y arguments in the call to
25955         __builtin_ia32_sbb_u64.
25957 2017-07-04  Jakub Jelinek  <jakub@redhat.com>
25959         PR debug/81278
25960         * tree-vrp.c (compare_assert_loc): Turn into a function template
25961         with stable template parameter.  Only test if a->e is NULL,
25962         !a->e == !b->e has been verified already.  Use e == NULL or
25963         e != NULL instead of e or ! e tests.  If stable is true, don't use
25964         iterative_hash_expr, on the other side allow a or b or both NULL
25965         and sort the NULLs last.
25966         (process_assert_insertions): Sort using compare_assert_loc<false>
25967         instead of compare_assert_loc, later sort using
25968         compare_assert_loc<true> before calling process_assert_insertions_for
25969         in a loop.  Use break instead of continue once seen NULL pointer.
25971 2017-07-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25973         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
25974         Cortex-R7 and Cortex-R8 processors.
25976 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25978         * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
25979         uninitialized while src is not.
25981 2017-07-04  Richard Earnshaw  <rearnsha@arm.com>
25983         * common/config/arm/arm-common.c: Adjust include path for
25984         arm-cpu-cdata.h
25985         * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
25986         (arm-cpu.h): Create in build directory.  Adjust dependency rules.
25987         (arm-cpu-data.h): Likewise.
25988         (arm-cpu-cdata.h): Likewise.
25989         * config/arm/arm-cpu.h: Delete.
25990         * config/arm/arm-cpu-cdata.h: Delete.
25991         * config/arm/arm-cpu-data.h: Delete.
25993 2017-07-04  James Greenhalgh  <james.greenhalgh@arm.com>
25995         * config/arm/arm-cpus.in (cortex-a55): New.
25996         (cortex-a75): Likewise.
25997         (cortex-a75.cortex-a55): Likewise.
25998         * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
25999         cortex-a75.
26000         * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
26001         * config/arm/arm-cpu-cdata.h: Regenerate.
26002         * config/arm/arm-cpu-data.h: Regenerate.
26003         * config/arm/arm-cpu.h: Regenerate.
26004         * config/arm/arm-tables.opt: Regenerate.
26005         * config/arm/arm-tune.md: Regenerate.
26007 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26009         * haifa-sched.c (sched_create_recovery_edges): Update profile.
26011 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
26013         * bb-reorder.c (better_edge_p): Fix handling of uninitialized
26014         probability.
26016 2017-07-04  Richard Sandiford  <richard.sandiford@linaro.org>
26018         PR tree-optimization/81292
26019         * tree-ssa-strlen.c (handle_builtin_strlen): When setting
26020         full_string_p, also call adjust_related_strinfos if the adjustment
26021         is simple, otherwise invalidate related strinfos.
26023 2017-07-04  Martin Liska  <mliska@suse.cz>
26025         PR sanitizer/81040
26026         * sanopt.c (sanitize_rewrite_addressable_params): Mark the
26027         newly created variable as DECL_IGNORED_P.
26029 2017-07-04  Martin Liska  <mliska@suse.cz>
26031         PR ipa/81293
26032         * ipa-inline.c (inline_small_functions):
26033         Use xstrdup_for_dump.
26035 2017-07-04  Tom de Vries  <tom@codesourcery.com>
26037         * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
26039 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
26041         PR target/81033
26042         * config/darwin.c (darwin_function_switched_text_sections):
26043         Fix spaces.
26045 2017-07-03  Jan Hubicka  <hubicka@ucw.cz>
26047         * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
26049 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26051         * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
26053 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26055         * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
26056         min_profitable_iters, and th as inclusive lower bounds.
26057         Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
26058         (vect_estimate_min_profitable_iters): Return inclusive lower bounds
26059         for min_profitable_iters and min_profitable_estimate.
26060         (vect_transform_loop): Treat th as an inclusive lower bound.
26061         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
26063 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
26065         PR target/81033
26066         * config/darwin.c (darwin_function_switched_text_sections):
26067         Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
26068         in two pieces, and suppress the use of buf.
26070 2017-07-03  Nathan Sidwell  <nathan@acm.org>
26072         * hash-table.h (hash_table_mod1): Fix indentation.
26074 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26076         PR middle-end/81290
26077         * predict.c (force_edge_cold): Be more careful about propagation
26078         backward.
26079         * profile-count.h (profile_probability::guessed,
26080         profile_probability::fdo, profile_count::guessed, profile_count::fdo):
26081         New.
26082         * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
26084 2017-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
26086         * doc/invoke.texi (rcpc architecture extension): Document it.
26088 2017-07-03  Richard Biener  <rguenther@suse.de>
26090         PR tree-optimization/60510
26091         * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
26092         the scalar reduction PHI and use it.
26093         (vectorizable_reduction): Properly guard the single_defuse_cycle
26094         path for non-SLP reduction chains where we cannot use it.
26095         Rework reduc_def/index and vector type deduction.  Rework
26096         vector operand gathering during reduction op code-gen.
26097         * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
26098         chains dissolve the chain and leave it to non-SLP reduction
26099         handling.
26101 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26103         * tree-data-ref.h (dr_alignment): Declare.
26104         * tree-data-ref.c (dr_alignment): New function.
26105         * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
26106         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26107         set it.
26108         * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
26110 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26112         * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
26113         and base_misalignment fields.
26114         (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
26115         * tree-data-ref.c: Include builtins.h.
26116         (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
26117         * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
26118         (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
26119         * tree-vect-data-refs.c: Include tree-cfg.h.
26120         (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
26121         fields instead of calculating an alignment here.
26122         (vect_analyze_data_refs): Use dr_analyze_innermost.  Dump the new
26123         innermost_loop_behavior fields.
26125 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26127         * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
26128         field.
26129         (DR_STEP_ALIGNMENT): New macro.
26130         * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
26131         * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
26132         (create_data_ref): Print it.
26133         * tree-vect-stmts.c (vectorizable_load): Use the step alignment
26134         to tell whether the step preserves vector (mis)alignment.
26135         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26136         Move the check for an integer step and generalise to all INTEGER_CST.
26137         (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
26138         Print the outer step alignment.
26140 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26142         * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
26143         with offset_alignment.
26144         (DR_ALIGNED_TO): Delete.
26145         (DR_OFFSET_ALIGNMENT): New macro.
26146         * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
26147         (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
26148         * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
26149         (create_data_ref): Likewise.
26150         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26151         (vect_analyze_data_refs): Likewise.
26152         * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
26153         creating dummy innermost behavior.
26155 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26157         * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
26158         with a "innermost_loop_behavior *" and refeence tree.
26159         * tree-data-ref.c (dr_analyze_innermost): Likewise.
26160         (create_data_ref): Update call accordingly.
26161         * tree-predcom.c (find_looparound_phi): Likewise.
26163 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26165         * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
26166         fields with dr_wrt_vec_loop.
26167         (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
26168         (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
26169         (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
26170         (vect_dr_behavior): New function.
26171         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26172         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26173         vect_dr_behavior.  Use a step_preserves_misalignment_p boolean to
26174         track whether the step preserves the misalignment.
26175         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26176         Use vect_dr_behavior.
26177         (vect_setup_realignment): Update call accordingly.
26178         (vect_create_data_ref_ptr): Likewise.  Use vect_dr_behavior.
26179         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
26180         call to vect_create_addr_base_for_vector_ref.
26181         (vect_create_cond_for_align_checks): Likewise.
26182         * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
26183         STMT_VINFO_DR_WRT_VEC_LOOP as a block.
26184         (vect_recog_mask_conversion_pattern): Likewise.
26185         * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
26186         (new_stmt_vec_info): Remove redundant zeroing.
26188 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26190         * common/config/arm/arm-common.c (arm_be8_option): New function.
26191         * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
26192         (ISA_ARMv6): Add isa_bit_be8.
26193         * config/arm/arm.h (arm_be8_option): Add prototype.
26194         (BE8_SPEC_FUNCTION): New define.
26195         (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
26196         * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
26197         (mlittle-endian): Similarly.
26198         (mbe8, mbe32): New options.
26199         * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
26200         * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
26202 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26204         * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
26206 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26208         * tree-cfgcleanup.c (want_merge_blocks_p): New function.
26209         (cleanup_tree_cfg_bb): Use it.
26210         * profile-count.h (profile_count::of_for_merging, profile_count::merge):
26211         New functions.
26212         * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
26214 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26216         PR bootstrap/81285
26217         * loop-doloop.c (add_test): Update profile.
26219 2017-07-03  Martin Liska  <mliska@suse.cz>
26221         PR sanitize/81040
26222         * sanopt.c (rewrite_usage_of_param): New function.
26223         (sanitize_rewrite_addressable_params): Likewise.
26224         (pass_sanopt::execute): Call rewrite_usage_of_param.
26226 2017-07-03  Richard Biener  <rguenther@suse.de>
26228         * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26229         back to using VIEW_CONVERT_EXPR.
26231 2017-07-03  Martin Liska  <mliska@suse.cz>
26233         PR other/78366
26234         * doc/extend.texi: Document when a resolver function is
26235         generated for target_clones.
26237 2017-07-03  Martin Liska  <mliska@suse.cz>
26239         * asan.c (asan_emit_stack_protection): Unpoison just red zones
26240         and shadow memory of auto variables which are subject of
26241         use-after-scope sanitization.
26242         (asan_expand_mark_ifn): Add do set only when is_poison.
26244 2016-07-03  Richard Biener  <rguenther@suse.de>
26246         * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26247         reduction PHIs.
26248         (vect_force_simple_reduction): Record reduction def -> phi mapping.
26249         (vectorizable_reduction): Perform reduction PHI creation when
26250         visiting a reduction PHI and adjust and simplify code generation
26251         phase of the reduction op.  Cache dts, use fold_binary, not fold_build2.
26252         (vect_transform_loop): Visit reduction PHIs.
26253         * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26254         defs into the SLP tree.
26255         (vect_build_slp_tree): Reduction defs terminate the recursion.
26256         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26257         of reduction defs.
26258         (vect_get_vec_defs_for_stmt_copy): Export.
26259         (vect_get_vec_defs): Likewise.
26260         * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26261         purpose.
26262         (vect_get_vec_defs_for_stmt_copy): Declare.
26263         (vect_get_vec_defs): Likewise.
26265 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26267         * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26268         parameter with a "loop" parameter and use it instead of the
26269         loop containing DR_STMT.  Don't check simple_iv when doing
26270         BB analysis.  Describe the two analysis modes in the comment.
26272 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26274         PR tree-optimization/69468
26275         * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26276         (find_same_succ_bb): Handle ignore_edge_flags.
26278 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26280         PR tree-optimization/81192
26281         * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26282         hash.
26283         (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26284         differs.
26285         (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26287 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26289         PR tree-optimization/81192
26290         * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26291         BB_SAME_SUCC (bb) == NULL.
26293 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26295         * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26296         consistency.
26298 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26300         * dumpfile.c: Include profile-count.h
26301         * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26302         update profile.
26303         (insert_cond_bb): Update profile.
26304         * tree-cfg.h (insert_cond_bb): Update prototype.
26305         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26306         * tree-dump.c: Do not include tree-cfg.
26308 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26310         * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26312 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26314         * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26315         bb.
26317 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26319         * tree-complex.c (expand_complex_div_wide): update profile.
26321 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26322             Alan Hayward  <alan.hayward@arm.com>
26323             David Sherwood  <david.sherwood@arm.com>
26325         * Makefile.in (MACHMODE_H): Remove insn-modes.h
26326         (CORETYPES_H): New define.
26327         (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26328         (insn-modes-inline.h, s-modes-inline-h): New rules.
26329         (generated_files): Add insn-modes-inline.h.
26330         (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26331         (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26332         (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26333         (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26334         (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26335         (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26336         (build/gencodes.o, build/genconditions.o): Likewise.
26337         (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26338         (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26339         (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26340         (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26341         (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26342         (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26343         * coretypes.h: Include everything up to real.h for generators.
26344         Include insn-modes.h first.  Include wide-int-print.h after
26345         wide-int.h.  Include insn-modes-inline.h and then machmode.h.
26346         * machmode.h: Don't include insn-modes.h here.
26347         * function-tests.c: Remove includes of signop.h, machmode.h,
26348         double-int.h and wide-int.h.
26349         * rtl.h: Likewise.
26350         * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26351         and wide-int.h.
26352         * optc-save-gen.awk: Likewise.
26353         * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26354         * godump.c: Remove include of wide-int-print.h.
26355         * pretty-print.h: Likewise.
26356         * wide-int-print.cc: Likewise.
26357         * wide-int.cc: Likewise.
26358         * hash-map-tests.c: Remove include of signop.h.
26359         * hash-set-tests.c: Likewise.
26360         * rtl-tests.c: Likewise.
26361         * mkconfig.sh: Remove include of machmode.h.
26362         * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26363         into...
26364         (emit_insn_modes_inline_h): ...this new function.  Emit the code
26365         into an insn-modes-inline.h header file, adding appropriate
26366         include guards and end comments.
26367         (emit_insn_modes_c_header): Remove include of machmode.h.
26368         (emit_min_insn_modes_c_header): Include coretypes.h rather than
26369         machmode.h.
26370         (main): Handle -i flag and call emit_insn_modes_inline_h when
26371         it is passed.
26373 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26375         * tree-ssa-strlen.c (strinfo): Rename the length field to
26376         nonzero_chars.  Add a full_string_p field.
26377         (compare_nonzero_chars, zero_length_string_p): New functions.
26378         (get_addr_stridx): Add an offset_out parameter.
26379         Use compare_nonzero_chars.
26380         (get_stridx): Update accordingly.  Use compare_nonzero_chars.
26381         (new_strinfo): Update after above changes to strinfo.
26382         (set_endptr_and_length): Set full_string_p.
26383         (get_string_length): Update after above changes to strinfo.
26384         (unshare_strinfo): Update call to new_strinfo.
26385         (maybe_invalidate): Likewise.
26386         (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26387         Use compare_nonzero_chars and zero_string_p.  Treat nonzero_chars
26388         as a uhwi instead of an shwi.  Update after above changes to
26389         strinfo and new_strinfo.
26390         (zero_length_string): Assert that chainsi contains full strings.
26391         Use zero_length_string_p.  Update call to new_strinfo.
26392         (adjust_related_strinfos): Update after above changes to strinfo.
26393         Copy full_string_p from origsi.
26394         (adjust_last_stmt): Use zero_length_string_p.
26395         (handle_builtin_strlen): Update after above changes to strinfo and
26396         new_strinfo.  Install the lhs as the string length if the previous
26397         entry didn't describe a full string.
26398         (handle_builtin_strchr): Update after above changes to strinfo
26399         and new_strinfo.
26400         (handle_builtin_strcpy): Likewise.
26401         (handle_builtin_strcat): Likewise.
26402         (handle_builtin_malloc): Likewise.
26403         (handle_pointer_plus): Likewise.
26404         (handle_builtin_memcpy): Likewise.  Track nonzero characters
26405         that aren't necessarily followed by a nul terminator.
26406         (handle_char_store): Likewise.
26408 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26410         PR tree-optimization/80769
26411         * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26412         for malloc and calloc.  Document the new invariant that all related
26413         strinfos have delayed lengths or none do.
26414         (verify_related_strinfos): Move earlier in file.
26415         (set_endptr_and_length): New function, split out from...
26416         (get_string_length): ...here.  Also set the lengths of related
26417         strinfos.
26418         (zero_length_string): Assert that chainsi has known (rather than
26419         delayed) lengths.
26420         (adjust_related_strinfos): Likewise.
26422 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26424         PR tree-optimization/81136
26425         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26426         assert that two references with the same misalignment have the same
26427         compile-time misalignment if those compile-time misalignments
26428         are known.
26430 2017-07-01  Andi Kleen  <ak@linux.intel.com>
26432         * print-tree.c (print_node): Print all attributes.
26434 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26436         * cfg.c (scale_bbs_frequencies): New function.
26437         * cfg.h (scale_bbs_frequencies): Declare it.
26438         * cfgloopanal.c (single_likely_exit): Cleanup.
26439         * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26440         as parameter.
26441         (scale_loop_profile): Likewise.
26442         (loop_version): Likewise.
26443         (create_empty_loop_on_edge): Update.
26444         * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26445         scale_loop_frequencies, scale_loop_profile, loopify,
26446         loop_version): Update prototypes.
26447         * modulo-sched.c (sms_schedule): Update.
26448         * predict.c (unlikely_executed_edge_p): Also check probability.
26449         (probably_never_executed_edge_p): Fix typo.
26450         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26451         * tree-parloops.c (gen_parallel_loop): Update.
26452         * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26453         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26454         * tree-ssa-loop-split.c (split_loop): Update.
26455         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26456         * tree-vect-loop-manip.c (vect_do_peeling): Update.
26457         (vect_loop_versioning): Update.
26458         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26460 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26462         * trans-mem.c (split_bb_make_tm_edge): Update profile.
26464 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26466         * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26467         to keep profile consistent.
26469 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26471         * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26472         * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26473         * profile-count.h (max_safe_multiplier): Make unsigned.
26474         (profile_count::guessed_zero): New.
26476 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26478         * bb-reorder.c (fix_up_crossing_landing_pad,
26479         fix_crossing_conditional_branches): Use make_single_succ_edge
26480         to keep profile consistent.
26482 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26484         * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26485         to update profile.
26487 2017-07-01  Jakub Jelinek  <jakub@redhat.com>
26489         PR sanitizer/81262
26490         * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26491         the right scopes, make sure cond_jump isn't preserved between multiple
26492         iterations.  Search for fallthru edge whenever there are 3+ edges and
26493         use find_fallthru_edge for it.
26495 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26497         Patch by Alexander Monakov <amonakov@ispras.ru>
26498         * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26499         probabilities consistently.
26501 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26503         * pa.c (pa_expand_compare_and_swap_loop): Update call of
26504         emit_cmp_and_jump_insns.
26506 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26508         PR ipa/81261
26509         * tree-inline.c (expand_call_inline): Combine profile statuses.
26511 2017-06-30  Andrew Pinski  <apinski@cavium.com>
26513         * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26514         fold_stmt returned true.
26516 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26518         * ggc.h (empty_string): Delete.
26519         * cfgexpand.c (expand_asm_stmt): Use plain "".
26520         * optabs.c (expand_asm_memory_barrier): Likewise.
26521         * stringpool.c (empty_string): Delete.
26522         (digit_vector, digit_string): Delete.
26523         (ggc_alloc_string): Use plain "", don't optimize single digit
26524         strings.  Use ggc_alloc_atomic.
26526 2017-06-30  Richard Earnshaw  <rearnsha@arm.com>
26528         * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26529         comparison set and one other set, use the cost of the non-comparison
26530         set.
26532 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26534         * ggc.h: Replace all 'static inline' with plain 'inline'.  Fix
26535         some formatting.
26537 2017-06-30  Peter Bergner  <bergner@vnet.ibm.com>
26539         * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26540         loops.  Remove now unneeded calls to gimple_switch_set_label() that
26541         just set removed labels to NULL_TREE.
26543 2017-06-30  Aldy Hernandez  <aldyh@redhat.com>
26545         * tree-ssanames.c (set_range_info_raw): Abstract from ...
26546         (set_range_info): ...here.  Only call set_range_info_raw if domain
26547         is useful.
26548         (set_nonzero_bits): Call set_range_info_raw.
26549         * tree-ssanames.h (set_range_info_raw): New.
26551 2017-06-30  Jakub Jelinek  <jakub@redhat.com>
26553         PR target/81225
26554         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26555         V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26556         of nonimmediate_operand and <store_mask_constraint> instead of m
26557         for the input operand.  For V8FI iterator, always split if input
26558         is a MEM.  For V16FI and V8SF_256 iterators, don't test if both
26559         operands are MEM if <mask_applied>.  For VI4F_256 iterator, use
26560         <store_mask_predicate> instead of register_operand and
26561         <store_mask_constraint> instead of v for the input operand.  Make
26562         sure both operands aren't MEMs for if not <mask_applied>.
26564 2017-06-30  Sylvestre Ledru  <sylvestre@debian.org>
26566         * lto-wrapper.c (copy_file) Close both file descriptors before
26567         exiting normally.
26569 2017-06-30  Martin Liska  <mliska@suse.cz>
26571         PR ipa/81214
26572         * multiple_target.c (create_dispatcher_calls): Make ifunc
26573         also for function that don't have calls or are not referenced.
26575 2017-06-30  Richard Biener  <rguenther@suse.de>
26577         * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26578         analyze the first scalar stmt.  Move vector type computation
26579         for the BB case here from ...
26580         * tree-vect-stmts.c (vect_analyze_stmt): ... here.  Guard
26581         live operation processing in the SLP case properly.
26583 2017-06-30  Richard Biener  <rguenther@suse.de>
26585         * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26587 2017-06-30  Martin Liska  <mliska@suse.cz>
26589         PR sanitizer/81021
26590         * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26591         before BUILT_IN_UNWIND_RESUME when ASAN is used.
26593 2017-06-30  Yvan Roux  <yvan.roux@linaro.org>
26595         * doc/invoke.texi (AArch64): Add missing options and remove redundant
26596         ones.
26598 2017-06-30  Richard Biener  <rguenther@suse.de>
26600         PR tree-optimization/81249
26601         * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26602         condition reduction result to original scalar type.
26604 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26606         * profile-count.h (enum profile_quality): Fix typos and whitespace
26607         issues.
26609 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26611         * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26612         type for branch probabilities.
26614 2017-06-29  Julian Brown  <julian@codesourcery.com>
26615             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26617         * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26618         * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26619         (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26620         (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26622 2017-06-29  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26624         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26625         check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26626         * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26627         CC usage from generic code to here.
26628         * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26629         CC usage into the target macros.
26631 2017-06-29  Maya Rashish  <coypu@sdf.org>
26633         * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26634         objects.
26636 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26638         * arm/arm-builtins.c: Include profile-count.h
26639         * except.c (sjlj_emit_function_enter): Use
26640         profile_probability::unlikely.
26642 2017-06-29  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26644         * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26645         and tocrel_offset be pointer args rather than implicitly using
26646         static versions.
26647         (legitimate_constant_pool_address_p, rs6000_emit_move,
26648         const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26649         tocrel_offset and use in toc_relative_expr_p call.
26650         (print_operand, print_operand_address): Use static tocrel_base_oac
26651         and tocrel_offset_oac.
26652         (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26653         tocrel_offset_oac.
26655 2017-06-29  Maya Rashish  <coypu@sdf.org>
26657         * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26659 2017-06-29  Eric Botcazou  <ebotcazou@adacore.com>
26661         * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26662         objects, take into account only the alignment of 'op0' and 'mode1' if
26663         'op0' is a MEM.
26665 2017-06-29  Steve Ellcey  <sellcey@cavium.com>
26667         * ccmp.c (ccmp_tree_comparison_p): New function.
26668         (ccmp_candidate_p): Update to use above function.
26669         (get_compare_parts): New function.
26670         (expand_ccmp_next): Update to use new functions.
26671         (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26672         new functions.
26673         (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26674         take mode as argument.
26675         * ccmp.h (expand_ccmp_expr): Add mode as argument.
26676         * expr.c (expand_expr_real_1): Pass mode as argument.
26678 2017-06-29  Segher Boessenkool  <segher@kernel.crashing.org>
26680         * combine.c (combine_instructions): Print insns to dump_file, together
26681         with their costs.
26683 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26685         * asan.c (asan_emit_stack_protection): Update.
26686         (create_cond_insert_point): Update.
26687         * auto-profile.c (afdo_propagate_circuit): Update.
26688         * basic-block.h (struct edge_def): Turn probability to
26689         profile_probability.
26690         (EDGE_FREQUENCY): Update.
26691         * bb-reorder.c (find_traces_1_round): Update.
26692         (better_edge_p): Update.
26693         (sanitize_hot_paths): Update.
26694         * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26695         (make_single_succ_edge): Update.
26696         (check_bb_profile): Update.
26697         (dump_edge_info): Update.
26698         (update_bb_profile_for_threading): Update.
26699         * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26700         probabilitycount to 0.
26701         * cfgbuild.c (compute_outgoing_frequencies): Update.
26702         * cfgcleanup.c (try_forward_edges): Update.
26703         (outgoing_edges_match): Update.
26704         (try_crossjump_to_edge): Update.
26705         * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26706         (expand_gimple_tailcall): Update.
26707         (construct_init_block): Use make_single_succ_edge.
26708         (construct_exit_block): Use make_single_succ_edge.
26709         * cfghooks.c (verify_flow_info): Update.
26710         (redirect_edge_succ_nodup): Update.
26711         (split_edge): Update.
26712         (account_profile_record): Update.
26713         * cfgloopanal.c (single_likely_exit): Update.
26714         * cfgloopmanip.c (scale_loop_profile): Update.
26715         (set_zero_probability): Remove.
26716         (duplicate_loop_to_header_edge): Update.
26717         * cfgloopmanip.h (loop_version): Update prototype.
26718         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26719         (force_nonfallthru_and_redirect): Update.
26720         (update_br_prob_note): Update.
26721         (rtl_verify_edges): Update.
26722         (purge_dead_edges): Update.
26723         (rtl_lv_add_condition_to_bb): Update.
26724         * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26725         * cgraphunit.c (init_lowered_empty_function): Update.
26726         (cgraph_node::expand_thunk): Update.
26727         * cilk-common.c: Include profile-count.h
26728         * dojump.c (inv): Remove.
26729         (jumpifnot): Update.
26730         (jumpifnot_1): Update.
26731         (do_jump_1): Update.
26732         (do_jump): Update.
26733         (do_jump_by_parts_greater_rtx): Update.
26734         (do_compare_rtx_and_jump): Update.
26735         * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26736         do_jump_1. do_compare_rtx_and_jump): Update prototype.
26737         * dwarf2cfi.c: Include profile-count.h
26738         * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26739         (sjlj_emit_dispatch_table): Likewise.
26740         * explow.c: Include profile-count.h
26741         * expmed.c (emit_store_flag_force): Update.
26742         (do_cmp_and_jump): Update.
26743         * expr.c (compare_by_pieces_d::generate): Update.
26744         (compare_by_pieces_d::finish_mode): Update.
26745         (emit_block_move_via_loop): Update.
26746         (store_expr_with_bounds): Update.
26747         (store_constructor): Update.
26748         (expand_expr_real_2): Update.
26749         (expand_expr_real_1): Update.
26750         * expr.h (try_casesi, try_tablejump): Update prototypes.
26751         * gimple-pretty-print.c (dump_probability): Update.
26752         (dump_profile): New.
26753         (dump_gimple_label): Update.
26754         (dump_gimple_bb_header): Update.
26755         * graph.c (draw_cfg_node_succ_edges): Update.
26756         * hsa-gen.c (convert_switch_statements): Update.
26757         * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26758         (find_if_case_1): Update.
26759         (find_if_case_2): Update.
26760         * internal-fn.c (expand_arith_overflow_result_store): Update.
26761         (expand_addsub_overflow): Update.
26762         (expand_neg_overflow): Update.
26763         (expand_mul_overflow): Update.
26764         (expand_vector_ubsan_overflow): Update.
26765         * ipa-cp.c (good_cloning_opportunity_p): Update.
26766         * ipa-split.c (split_function): Use make_single_succ_edge.
26767         * ipa-utils.c (ipa_merge_profiles): Update.
26768         * loop-doloop.c (add_test): Update.
26769         (doloop_modify): Update.
26770         * loop-unroll.c (compare_and_jump_seq): Update.
26771         (unroll_loop_runtime_iterations): Update.
26772         * lra-constraints.c (lra_inheritance): Update.
26773         * lto-streamer-in.c (input_cfg): Update.
26774         * lto-streamer-out.c (output_cfg): Update.
26775         * mcf.c (adjust_cfg_counts): Update.
26776         * modulo-sched.c (sms_schedule): Update.
26777         * omp-expand.c (expand_omp_for_init_counts): Update.
26778         (extract_omp_for_update_vars): Update.
26779         (expand_omp_ordered_sink): Update.
26780         (expand_omp_for_ordered_loops): Update.
26781         (expand_omp_for_generic): Update.
26782         (expand_omp_for_static_nochunk): Update.
26783         (expand_omp_for_static_chunk): Update.
26784         (expand_cilk_for): Update.
26785         (expand_omp_simd): Update.
26786         (expand_omp_taskloop_for_outer): Update.
26787         (expand_omp_taskloop_for_inner): Update.
26788         * omp-simd-clone.c (simd_clone_adjust): Update.
26789         * optabs.c (expand_doubleword_shift): Update.
26790         (expand_abs): Update.
26791         (emit_cmp_and_jump_insn_1): Update.
26792         (expand_compare_and_swap_loop): Update.
26793         * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26794         * predict.c (predictable_edge_p): Update.
26795         (edge_probability_reliable_p): Update.
26796         (set_even_probabilities): Update.
26797         (combine_predictions_for_insn): Update.
26798         (combine_predictions_for_bb): Update.
26799         (propagate_freq): Update.
26800         (estimate_bb_frequencies): Update.
26801         (force_edge_cold): Update.
26802         * profile-count.c (profile_count::dump): Add missing space into dump.
26803         (profile_count::debug): Add newline.
26804         (profile_count::differs_from_p): Explicitly convert to unsigned.
26805         (profile_count::stream_in): Update.
26806         (profile_probability::dump): New member function.
26807         (profile_probability::debug): New member function.
26808         (profile_probability::differs_from_p): New member function.
26809         (profile_probability::differs_lot_from_p): New member function.
26810         (profile_probability::stream_in): New member function.
26811         (profile_probability::stream_out): New member function.
26812         * profile-count.h (profile_count_quality): Rename to ...
26813         (profile_quality): ... this one.
26814         (profile_probability): New.
26815         (profile_count): Update.
26816         * profile.c (compute_branch_probabilities): Update.
26817         * recog.c (peep2_attempt): Update.
26818         * sched-ebb.c (schedule_ebbs): Update.
26819         * sched-rgn.c (find_single_block_region): Update.
26820         (compute_dom_prob_ps): Update.
26821         (schedule_region): Update.
26822         * sel-sched-ir.c (compute_succs_info): Update.
26823         * stmt.c (struct case_node): Update.
26824         (do_jump_if_equal): Update.
26825         (get_outgoing_edge_probs): Update.
26826         (conditional_probability): Update.
26827         (emit_case_dispatch_table): Update.
26828         (expand_case): Update.
26829         (expand_sjlj_dispatch_table): Update.
26830         (emit_case_nodes): Update.
26831         * targhooks.c: Update.
26832         * tracer.c (better_p): Update.
26833         (find_best_successor): Update.
26834         * trans-mem.c (expand_transaction): Update.
26835         * tree-call-cdce.c: Update.
26836         * tree-cfg.c (gimple_split_edge): Upate.
26837         (move_sese_region_to_fn): Upate.
26838         * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26839         * tree-eh.c (lower_resx): Upate.
26840         (cleanup_empty_eh_move_lp): Upate.
26841         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26842         * tree-inline.c (copy_edges_for_bb): Update.
26843         (copy_cfg_body): Update.
26844         * tree-parloops.c (gen_parallel_loop): Update.
26845         * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26846         (gimple_gen_time_profiler): Update.
26847         * tree-ssa-dce.c (remove_dead_stmt): Update.
26848         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26849         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26850         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26851         (unloop_loops): Update.
26852         (try_peel_loop): Update.
26853         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26854         * tree-ssa-loop-split.c (connect_loops): Update.
26855         (split_loop): Update.
26856         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26857         (hoist_guard): Update.
26858         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26859         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26860         (value_replacement): Update.
26861         * tree-ssa-reassoc.c (branch_fixup): Update.
26862         * tree-ssa-tail-merge.c (replace_block_by): Update.
26863         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26864         (create_edge_and_update_destination_phis): Update.
26865         (compute_path_counts): Update.
26866         (recompute_probabilities): Update.
26867         (update_joiner_offpath_counts): Update.
26868         (freqs_to_counts_path): Update.
26869         (duplicate_thread_path): Update.
26870         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26871         (struct switch_conv_info): Update.
26872         (gen_inbound_check): Update.
26873         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26874         (vect_do_peeling): Update.
26875         (vect_loop_versioning): Update.
26876         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26877         (optimize_mask_stores): Update.
26878         * ubsan.c (ubsan_expand_null_ifn): Update.
26879         * value-prof.c (gimple_divmod_fixed_value): Update.
26880         (gimple_divmod_fixed_value_transform): Update.
26881         (gimple_mod_pow2): Update.
26882         (gimple_mod_pow2_value_transform): Update.
26883         (gimple_mod_subtract): Update.
26884         (gimple_mod_subtract_transform): Update.
26885         (gimple_ic): Update.
26886         (gimple_stringop_fixed_value): Update.
26887         (gimple_stringops_transform): Update.
26888         * value-prof.h: Update.
26890 2017-06-29  Carl Love  <cel@us.ibm.com>
26892         * config/rs6000/rs6000-c.c: Add support for built-in functions
26893         vector signed int vec_signed (vector float);
26894         vector signed long long vec_signed (vector double);
26895         vector signed int vec_signed2 (vector double, vector double);
26896         vector signed int vec_signede (vector double);
26897         vector signed int vec_signedo (vector double);
26898         * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
26899         instruction generator.
26900         * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26901         UNSPEC_VSX_VSIGNED2): Add UNSPECS.
26902         (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
26903         Add define_insn.
26904         (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
26905         vunsignede_v2df): Add define_expands.
26906         * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
26907         VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
26908         VEC_UNSIGNEDO): Add definitions.
26909         * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26910         UNSPEC_VSX_VSIGNED2): Add UNSPECs.
26911         (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
26912         (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
26913         vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
26914         * config/rs6000/altivec.h (vec_signed, vec_signed2,
26915         vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
26916         vec_unsignede, vec_unsignedo): Add builtin defines.
26917         * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
26918         declaration.
26919         * doc/extend.texi: Update the built-in documentation file for the
26920         new built-in functions.
26922 2017-06-29  Richard Biener  <rguenther@suse.de>
26924         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
26925         reduction chains to LOOP_VINFO_REDUCTIONS.
26926         * tree-vect-slp.c (vect_analyze_slp): Continue looking for
26927         SLP reductions after processing reduction chains.
26929 2017-06-29  Nathan Sidwell  <nathan@acm.org>
26931         * builtins.c (fold_builtin_FUNCTION): Use
26932         lang_hooks.decl_printable_name.
26934 2017-06-29  Peter Bergner  <bergner@vnet.ibm.com>
26936         PR middle-end/81194
26937         * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
26938         with only one label.
26939         * stmt.c (expand_case): Assert NCASES is greater than one.
26941 2017-06-29  Richard Biener  <rguenther@suse.de>
26943         * tree-cfg.c (group_case_labels_stmt): Return whether we changed
26944         anything.
26945         (group_case_labels): Likewise.
26946         (find_taken_edge): Push sanity checking on val to workers...
26947         (find_taken_edge_cond_expr): ... here
26948         (find_taken_edge_switch_expr): ... and here, handle cases
26949         with just a default label.
26950         * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
26951         (group_case_labels): Likewise.
26952         * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
26953         group_case_labels does anything cleanup the CFG again.
26955 2017-06-29  Bin Cheng  <bin.cheng@arm.com>
26957         PR tree-optimization/81196
26958         * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
26959         exit condition comparing two IVs.
26961 2017-06-29  Richard Earnshaw  <rearnsha@arm.com>
26963         * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
26964         profile to the dummy entry at the end of the list of architectures.
26965         * config/arm/arm-cpu-cdata.h: Regenerated.
26967 2017-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26968             Michael Collison <michael.collison@arm.com>
26970         PR target/70119
26971         * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
26972         New pattern.
26973         (*aarch64_reg_<mode>3_neg_mask2): New pattern.
26974         (*aarch64_reg_<mode>3_minus_mask): New pattern.
26975         (*aarch64_<optab>_reg_di3_mask2): New pattern.
26976         * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
26977         of shift when the shift amount is masked with constant equal to
26978         the size of the mode.
26979         * config/aarch64/predicates.md (subreg_lowpart_operator): New
26980         predicate.
26982 2017-06-29  Martin Liska  <mliska@suse.cz>
26984         * config/i386/i386.opt: Change range from [1,5] to [0,5].
26986 2017-06-29  Yury Gribov  <tetra2005@gmail.com>
26988         PR bootstrap/80565
26989         * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
26990         code.
26991         * ipa-inline.h
26992         (edge_growth_cache_entry::edge_growth_cache_entry): New
26993         function.
26994         (reset_edge_growth_cache): Update to use constructor.
26996 2017-06-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
26998         * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
26999         (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
27000         (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
27002 2017-06-28  Sebastian Peryt  <sebastian.peryt@intel.com>
27004         * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
27005         (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
27007 2017-06-28  Szabolcs Nagy  <szabolcs.nagy@arm.com>
27009         * config.gcc (*-linux-musl*): Add t-musl tmake_file.
27010         (*-linux-uclibc*): Add t-uclibc tmake_file.
27011         * config/t-musl: New.
27012         * config/t-uclibc: New.
27014 2017-06-28  Richard Earnshaw  <rearnsha@arm.com>
27016         * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
27017         context.
27018         (gen_comm_data): Emit architectural setting of arch_prof.
27019         * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
27020         profile.
27021         (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
27022         (armv8-m.base, armv8-m.main): Likewise.
27023         * arm-protos.h (arm_build_target): Add profile field.
27024         (arch_option): Likewise.
27025         * config/arm/arm.c (arm_configure_build_target): Copy the profile to
27026         the active target.
27027         * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
27028         arm_active_target.profile.
27030 2017-06-28  Richard Biener  <rguenther@suse.de>
27032         PR middle-end/81227
27033         * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
27034         TYPE_OVERFLOW_WRAPS.
27035         * match.pd (negate_expr_p): Likewise.
27036         * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
27037         fold_build2, not fold_binary.
27039 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
27041         * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
27042         Convert memory address to Pmode.
27043         (aarch64_print_operand): Assert MEM operands are always Pmode.
27045 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
27047         PR target/79665
27048         * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
27049         Remove redundant if.
27050         (aarch_forward_to_shift_is_not_shifted_reg): Remove.
27051         * config/arm/aarch-common-protos.h
27052         (aarch_forward_to_shift_is_not_shifted_re): Remove.
27053         * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
27055 2017-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
27057         PR ipa/81238
27058         * multiple_target.c (create_dispatcher_calls): Set the default
27059         clone to be static, not public.
27061 2017-06-28  Richard Biener  <rguenther@suse.de>
27063         * tree-vect-loop.c (vectorizable_reduction): Move special
27064         cond reduction IV var creation ...
27065         (vect_create_epilog_for_reduction): ... here.  Remove induction_index
27066         parameter.  Use STMT_VINFO_VECTYPE.
27067         * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
27068         constant_p.
27070 2017-06-28  Martin Liska  <mliska@suse.cz>
27072         PR ipa/81128
27073         * ipa-visibility.c (non_local_p): Handle visibility.
27075 2017-06-28  Martin Liska  <mliska@suse.cz>
27077         PR driver/79659
27078         * common.opt: Add IntegerRange to various options.
27079         * opt-functions.awk (integer_range_info): New function.
27080         * optc-gen.awk: Add integer_range_info to cl_options struct.
27081         * opts-common.c (decode_cmdline_option): Handle
27082         CL_ERR_INT_RANGE_ARG.
27083         (cmdline_handle_error): Likewise.
27084         * opts.c (print_filtered_help): Show valid interval in
27085         when --help is provided.
27086         * opts.h (struct cl_option): Add range_min and range_max fields.
27087         * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
27089 2017-06-28  Marc Glisse  <marc.glisse@inria.fr>
27091         * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
27092         (x * C EQ/NE y * C): New transformation.
27094 2017-06-28  Christophe Lyon  <christophe.lyon@linaro.org>
27096         * genmultilib (combination_space): Accept '+' in option names.
27098 2017-06-28  Martin Liska  <mliska@suse.cz>
27100         PR sanitizer/81224
27101         * asan.c (instrument_derefs): Bail out inner references
27102         that are hard register variables.
27104 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
27106         PR target/81175
27107         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
27108         rather than def_builtin_pure for __builtin_ia32_gatherpf*.
27110 2017-06-28  Richard Biener  <rguenther@suse.de>
27112         * tree-vectorizer.h (vect_get_vec_defs): Remove.
27113         (vect_get_slp_defs): Adjust.
27114         * tree-vect-loop.c (get_initial_defs_for_reduction): Split
27115         out from ...
27116         * tree-vect-slp.c (vect_get_constant_vectors): ... here and
27117         simplify.
27118         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
27119         get_initial_defs_for_reduction instead of vect_get_vec_defs.
27120         (vectorizable_reduction): Adjust.
27121         * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
27122         handling.
27123         (vect_get_slp_defs): Likewise.
27124         * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
27125         (vectorizable_bswap): Adjust.
27126         (vectorizable_call): Likewise.
27127         (vectorizable_conversion): Likewise.
27128         (vectorizable_assignment): Likewise.
27129         (vectorizable_shift): Likewise.
27130         (vectorizable_operation): Likewise.
27131         (vectorizable_store): Likewise.
27132         (vectorizable_condition): Likewise.
27133         (vectorizable_comparison): Likewise.
27135 2017-06-28  Michael Collison  <michael.collison@arm.com>
27137         PR target/68535
27138         * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
27139         set of base_reg
27140         (arm_gen_movmemqi): Removed unused variable 'i'.
27141         Convert 'for' loop into 'while' loop.
27142         (arm_expand_prologue): Remove last unnecessary set of insn.
27143         (thumb_pop): Remove unused variable 'pushed_words'.
27144         (thumb_exit): Remove last unnecessary set of regs_to_pop.
27146 2017-06-28  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27148         * config/s390/predicates.md: Use s390_rel_address_ok_p.
27149         * config/s390/s390-protos.h: Add prototype of
27150         s390_rel_address_ok_p.
27151         * config/s390/s390.c (s390_got_symbol): New function.
27152         (s390_rel_address_ok_p): New function.
27153         (legitimize_pic_address): Use s390_rel_address_ok_p.
27154         (s390_load_got): Use s390_got_symbol.
27155         (s390_option_override): Issue error if
27156         -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
27157         * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
27158         New macro.
27159         * config/s390/s390.opt: New option mpic-data-is-text-relative.
27161 2017-06-27  Andrew Pinski  <apinski@cavium.com>
27163         * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
27164         (X * copysign (1.0, X)): New pattern.
27165         (X * copysign (1.0, -X)): New pattern.
27166         (copysign (-1.0, CST)): New pattern.
27168 2017-06-27  Joseph Myers  <joseph@codesourcery.com>
27170         * genmultilib (combination_space): Remove variable.
27171         Validate reuse rules against regular expression for any sequence
27172         of multilib options in any order.
27174 2017-06-27  Michael Collison  <michael.collison@arm.com>
27176         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
27177         call aarch64_split_simd_combine.
27178         * (aarch64_combine_internal<mode>): Delete pattern.
27179         * config/aarch64/aarch64.c (aarch64_split_simd_combine):
27180         Allow register and subreg operands.
27182 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27184         * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
27185         specific need, just fallback on defaults.
27186         (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
27188 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27189             Olivier Hainque  <hainque@adacore.com>
27191         * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
27192         map for 64bits.
27193         (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
27194         targets. Pick a default if no particular attempt applied.
27195         (STACK_CHECK_PROTECT): Double for 64bit targets, which have
27196         larger contexts.
27198 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27200         * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
27201         (x86_64-wrs-vxworks7): Likewise.
27203 2017-06-27  Marek Polacek  <polacek@redhat.com>
27205         PR sanitizer/81223
27206         * ubsan.c (instrument_null): Check get_base_address's result for null.
27208 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27210         * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
27212 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27214         * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
27215         BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
27216         (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
27217         BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
27218         New function types.
27219         * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
27220         BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
27221         BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
27222         BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
27223         BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27224         BUILT_IN_FEUPDATEENV): New builtins.
27225         * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27226         TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27227         * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27228         fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27229         macros.
27230         (builtin_structptr_types): Adjust size.
27231         * tree.c (builtin_structptr_types): Add four entries.
27233 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27234             Olivier Hainque  <hainque@adacore.com>
27236         * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27237         (TLS_SYM): New local macro, forcing reference to __tls__ on
27238         link command lines for VxWorks 7 RTPs, triggering initialization
27239         of tlsLib.
27240         (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27241         OS features TLS support, true for RTPs on VxWorks 7.
27242         * config/vxworks.c (vxworks_override_options): Setup emutls
27243         accordingly.
27245 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27247         * predict.c (test_prediction_value_range): Use -1U instead of -1
27248         to avoid narrowing conversion warning.
27249         * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27250         to avoid narrowing conversion warning.
27251         * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27252         -1.
27253         * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27255 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27257         * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27258         64bit configurations.
27259         (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27260         (SIZE_TYPE): Likewise.
27261         * config/vxworks.c (vxworks_emutls_var_fields): Use
27262         long_unsigned_type_node instead of unsigned_type_node as the offset
27263         field type, which is "pointer" mode in emutls.c.
27265 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27267         PR sanitizer/81209
27268         * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27270         PR middle-end/81207
27271         * gimple-fold.c (replace_call_with_call_and_fold): Handle
27272         gimple_vuse copying separately from gimple_vdef copying.
27274 2017-06-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27276         * value-prof.c (free_hist): Remove call to memset and the enclosing if
27277         condition.
27279 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27280             Olivier Hainque  <hainque@adacore.com>
27282         * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27283         for all vxworks7 targets.
27284         * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27285         (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27286         (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27287         variations for VX6/VX7 and 32/64bits later on in ...
27288         (VXWORKS_LIB_SPEC): Leverage new macros.
27289         (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27290         as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27292 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27294         * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27295         _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27297 2017-06-26  Carl Love  <cel@us.ibm.com>
27299         * config/rs6000/rs6000-c.c: Add support for built-in functions
27300         vector bool char vec_reve (vector bool char);
27301         vector signed char vec_reve (vector signed char);
27302         vector unsigned char vec_reve (vector unsigned char);
27303         vector bool int vec_reve (vector bool int);
27304         vector signed int vec_reve (vector signed int);
27305         vector unsigned int vec_reve (vector unsigned int);
27306         vector bool long long vec_reve (vector bool long long);
27307         vector signed long long vec_reve (vector signed long long);
27308         vector unsigned long long vec_reve (vector unsigned long long);
27309         vector bool short vec_reve (vector bool short);
27310         vector signed short vec_reve (vector signed short);
27311         vector double vec_reve (vector double);
27312         vector float vec_reve (vector float);
27313         * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27314         VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27315         * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27316         (altivec_vreve): New pattern.
27317         * config/rs6000/altivec.h (vec_reve): New define.
27318         * doc/extend.texi (vec_rev): Update the built-in documentation file
27319         for the new built-in functions.
27321 2016-06-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27323         PR tree-optimization/71815
27324         * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27325         function.
27326         (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27327         has_single_use.
27328         (slsr_process_phi): Likewise.
27329         (replace_uncond_cands_and_profitable_phis): Don't replace a
27330         multiply candidate with a stride of 1 (copy or cast).
27331         (phi_incr_cost): Call uses_consumed_by_stmt rather than
27332         has_single_use.
27333         (lowest_cost_path): Likewise.
27334         (total_savings): Likewise.
27336 2017-06-26  Richard Biener  <rguenther@suse.de>
27338         PR target/81175
27339         * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27340         Use def_builtin_pure for all gather builtins.
27342 2017-06-26  Richard Biener  <rguenther@suse.de>
27344         PR tree-optimization/81203
27345         * tree-tailcall.c (find_tail_calls): Do not move stmts into
27346         non-dominating BBs.
27348 2017-06-26  Marek Polacek  <polacek@redhat.com>
27350         PR c/80116
27351         * doc/invoke.texi: Document -Wmultistatement-macros.
27353 2017-06-26  Christophe Lyon  <christophe.lyon@linaro.org>
27355         * doc/sourcebuild.texi (ARM-specific attributes): Document new
27356         arm_neon_ok_no_float_abi effective target.
27358 2017-06-26  Richard Biener  <rguenther@suse.de>
27360         PR tree-optimization/80928
27361         * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27362         (copy_bbs): Set BB_DUPLICATED flag early.
27363         (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27364         marked blocks.
27365         (execute_on_shrinking_pred): Likewise.
27366         * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27367         BB_DUPLICATED blocks.
27368         * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27369         iterate over all PHIs considering removal of *gsi.
27371 2017-06-23  Jim Wilson  <jim.wilson@linaro.org>
27373         * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27374         qdf24xx.
27376 2017-06-23  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27378         * config/rs6000/rs6000-string.c: (expand_block_clear,
27379         do_load_for_compare, select_block_compare_mode,
27380         compute_current_alignment, expand_block_compare,
27381         expand_strncmp_align_check, expand_strn_compare,
27382         expand_block_move, rs6000_output_load_multiple)
27383         Move functions related to string/block move/compare
27384         to a separate file.
27385         * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27386         * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27387         for this function which is now used in two files.
27388         * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27389         * config.gcc: Add rs6000-string.o to extra_objs for
27390         targets powerpc*-*-* and rs6000*-*-*.
27392 2017-06-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
27394         PR target/80510
27395         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27396         32-bit, since indexed is not valid for DImode.
27397         (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27398         3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27399         (define_peephole2 for Altivec d-form load): Add 32-bit support.
27400         (define_peephole2 for Altivec d-form store): Likewise.
27402         PR ipa/81185
27403         * multiple_target.c (create_dispatcher_calls): Only create the
27404         dispatcher call if the function is the default clone of a
27405         versioned function.
27407 2017-06-23  Segher Boessenkool  <segher@kernel.crashing.org>
27409         PR middle-end/80902
27410         * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27411         a call, force the call to not be a tail call.
27413 2017-06-23  Jeff Law  <law@redhat.com>
27415         * doc/contrib.texi: Add entry for Steven Pemberton's work on
27416         enquire.
27418 2017-06-23  Will Schmidt  <will_schmidt@vnet.ibm.com>
27420         * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27421         update_call_from_tree().  (rs6000_gimple_fold_builtin): Add
27422         handling for early expansion of vector shifts (sl,sr,sra,rl).
27423         (builtin_function_type): Add vector shift right instructions
27424         to the unsigned argument list.
27426 2017-06-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27428         rtl-optimizatoin/79286
27429         * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27430         * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27431         trap.  PIC register plus a const unspec without offset can never trap.
27433 2017-06-23  Marc Glisse  <marc.glisse@inria.fr>
27435         * tree.h (builtin_structptr_type): New type.
27436         (builtin_structptr_types): Declare new array.
27437         * tree.c (builtin_structptr_types): New array.
27438         (free_lang_data, build_common_tree_nodes): Use it.
27440 2017-06-23  Jonathan Wakely  <jwakely@redhat.com>
27442         PR c++/81187
27443         * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27444         -Wnoexcept.
27446 2017-06-22  Matt Turner  <mattst88@gmail.com>
27448         * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27449         Lake models to skylake case.  Assume skylake for unknown
27450         models with clflushopt.
27452 2017-06-22  Jeff Law  <law@redhat.com>
27454         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27455         frame sizes that do not satisfy aarch64_uimm12_shift.
27457 2017-06-22  Jan Hubicka <hubicka@ucw.cz>
27459         * profile-count.h (apply_probability,
27460         apply_scale, probability_in): Fix checks for zero.
27462 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27464         * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27465         * doc/cppdiropts.texi (-I @var{dir}): Document it.
27467 2016-06-22  Richard Biener  <rguenther@suse.de>
27469         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27470         COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27471         REDUC_MAX_EXPR support.
27472         (vectorizable_reduction): Likewise.
27473         (vect_create_epilog_for_reduction): Likewise.
27475 2017-06-22  James Greenhalgh  <james.greenhalgh@arm.com>
27477         * match.pd (A / (1 << B) -> A >> B): New.
27478         * generic-match-head.c: Include optabs-tree.h.
27479         * gimple-match-head.c: Likewise.
27480         * optabs-tree.h (target_supports_op_p): New.
27481         * optabs-tree.c (target_supports_op_p): New.
27483 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27485         * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27486         $gcc_cv_ld --help output.
27487         (gcc_cv_ld_demangle): Likewise.
27488         (gcc_cv_ld_eh_frame_hdr): Likewise.
27489         (gcc_cv_ld_pie): Likewise.
27490         (gcc_cv_ld_as_needed): Likewise.  Prefer native forms unless $gnu_ld.
27491         (gcc_cv_ld_buildid): Likewise.
27492         (gcc_cv_ld_sysroot): Likewise.
27493         (ld_bndplt_support): Likewise.
27494         (ld_pushpopstate_support): Likewise.
27495         * configure: Regenerate.
27496         * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27498 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27500         PR target/81151
27501         * config/i386/sse.md (round<mode>2): Renumber match_dup and
27502         operands indexes to avoid gap between operands and match_dups.
27504 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27506         * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27507         Increment Arith_shift and Arith_shift_reg by 1.
27508         * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27509         New tuning flag.
27510         * config/aarch64/aarch64.c (thunderx_tunings): Enable
27511         AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27512         (aarch64_strip_extend): Add new argument and test for it.
27513         (aarch64_cheap_mult_shift_p): New function.
27514         (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27515         add a cost if it is true.
27516         Update calls to aarch64_strip_extend.
27517         (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27519 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27521         * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27522         tunings.
27523         (thunderxt88): Likewise.
27524         * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27525         (thunderx_prefetch_tune): New variable.
27526         (thunderx2t99_prefetch_tune): Update for the correct values.
27527         (thunderxt88_tunings): New variable.
27528         (thunderx_tunings): Use thunderx_prefetch_tune instead of
27529         generic_prefetch_tune.
27530         (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27532 2017-06-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27534         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27535         SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27536         (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27537         (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27538         (aarch64_atomic_cas<mode>, GPI): Likewise.
27540 2017-06-21  Martin Liska  <mliska@suse.cz>
27542         * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27543         statements on cold and hot labels.
27544         * predict.c (tree_estimate_probability_bb): Remove the
27545         prediction from this place.
27547 2017-06-21  Martin Liska  <mliska@suse.cz>
27549         PR tree-optimization/79489
27550         * gimplify.c (maybe_add_early_return_predict_stmt): New
27551         function.
27552         (gimplify_return_expr): Call the function.
27553         * predict.c (tree_estimate_probability_bb): Remove handling
27554         of early return.
27555         * predict.def: Update comment about early return predictor.
27556         * gimple-predict.h (is_gimple_predict): New function.
27557         * predict.def: Change default value of early return to 66.
27558         * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27559         statements.
27560         * passes.def: Put pass_strip_predict_hints to the beginning of
27561         IPA passes.
27563 2017-06-21  Pierre-Marie de Rodat  <derodat@adacore.com>
27565         * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27566         FUNCTION_DECL declarations.
27567         (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27568         declarations.
27569         (dwaf2out_decl): Likewise.
27570         * godump.c (go_early_global_decl): Skip call to the real debug hook
27571         for FUNCTION_DECL declarations.
27572         * passes.c (rest_of_decl_compilation): Skip call to the
27573         early_global_decl debug hook for FUNCTION_DECL declarations, unless
27574         -fdump-go-spec is passed.
27576 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27578         * config/i386/i386.c (struct builtin_isa): New field pure_p.
27579         Reorder for compactness.
27580         (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27581         (def_builtin_pure, def_builtin_pure2): New functions.
27582         (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27584 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27586         * match.pd (nop_convert): New predicate.
27587         ((A +- CST1) +- CST2): Allow some NOP conversions.
27589 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27591         PR c++/81130
27592         * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27593         with ctors/dtors if GOVD_SHARED is set.
27595 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27597         * config/aarch64/aarch64.md (movti_aarch64):
27598         Emit mov rather than orr.
27599         (movtf_aarch64): Likewise.
27600         * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27601         Emit mov rather than orr.
27603 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27605         * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27606         Swap alternatives, make integer dup more expensive.
27608 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27610         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27611         Return true for non-tls symbols.
27613 2017-06-21  James Greenhalgh  <james.greenhalgh@arm.com>
27615         * config/aarch64/aarch64-cores.def (cortex-a55): New.
27616         (cortex-a75): Likewise.
27617         (cortex-a75.cortex-a55): Likewise.
27618         * config/aarch64/aarch64-tune.md: Regenerate.
27619         * doc/invoke.texi (-mtune): Document new values for -mtune.
27621 2017-06-21  Tom de Vries  <tom@codesourcery.com>
27623         * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27624         stack_size feature.
27625         (Effective-Target Keywords, Other attributes): Suggest using
27626         dg-add-options stack_size feature to get stack limit in stack_size
27627         effective target documentation.
27629 2017-06-21  Julian Brown  <julian@codesourcery.com>
27630             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
27632         * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27633         (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27634         * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27635         reservation.
27636         * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27637         attribute type list for neon_multiply.
27638         * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27639         attribute type list for neon_multiply.
27640         * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27641         * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27642         attribute type list for neon_multiply.
27643         * config/arm/types.md (crypto_pmull): Add.
27644         * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27645         attribute type list.
27647 2017-06-20  Andreas Tobler  <andreast@gcc.gnu.org>
27649         * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27650         arm1176jzf-s.
27652 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27654         * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27655         to make sure not to dereference a NULL cost_classes_ptr pointer.
27657 2017-06-20  Carl Love  <cel@us.ibm.com>
27659         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27660         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27661         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27662         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27663         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27664         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27665         VMULOSW): New enum "unspec" values.
27666         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27667         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27668         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27669         altivec_vmulosw): New patterns.
27670         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27671         VMULOSW): Add definitions.
27673 2017-06-20  Julia Koval  <julia.koval@intel.com>
27675         * config/i386/i386.c: Fix rounding expand for new pattern.
27676         * config/i386/subst.md: Fix pattern (parallel -> unspec).
27678 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27680         * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27681         * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27683 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27685         * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27686         feature string.
27688 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27690         * config/aarch64/aarch64-cores.def: Rearrange to sort by
27691         architecture, then by implementer ID.
27692         * config/aarch64/aarch64-tune.md: Regenerate.
27694 2017-06-20  Richard Biener  <rguenther@suse.de>
27696         PR middle-end/81097
27697         * fold-const.c (split_tree): Fold to type before negating.
27699 2017-06-20  David Malcolm  <dmalcolm@redhat.com>
27701         * diagnostic-show-locus.c
27702         (selftest::test_fixit_deletion_affecting_newline): New function.
27703         (selftest::diagnostic_show_locus_c_tests): Call it.
27705 2017-06-20  Andreas Schwab  <schwab@suse.de>
27707         PR target/80970
27708         * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27709         instead of "+d".
27711 2017-06-20  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
27713         * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27714         __ARM_FEATURE_COPROC according to support.
27716 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27718         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27719         Rewritten to avoid overflow for > 32-bit pointers.
27721         PR sanitizer/81125
27722         * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27723         by removing enum keyword.
27724         (ubsan_type_descriptor): Likewise.  Formatting fix.
27726         PR target/81121
27727         * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27728         splitter): Require TARGET_SSE2 in the condition.
27730 2017-06-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
27732         PR target/79799
27733         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27734         for doing vector set of SFmode on ISA 3.0.
27735         * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27736         (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27737         element.
27738         (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27739         SFmode value into a V4SF variable that was extracted from another
27740         V4SF variable without converting the element to double precision
27741         and back to single precision vector format.
27742         (vsx_insert_extract_v4sf_p9_2): Likewise.
27744 2017-06-19  Jakub Jelinek  <jakub@redhat.com>
27746         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27747         in UWHI to avoid undefined overflow.
27749         PR sanitizer/81125
27750         * ubsan.h (enum ubsan_encode_value_phase): New.
27751         (ubsan_encode_value): Change second argument to
27752         enum ubsan_encode_value_phase with default value of
27753         UBSAN_ENCODE_VALUE_GENERIC.
27754         * ubsan.c (ubsan_encode_value): Change second argument to
27755         enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27756         adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27757         create_tmp_var_raw instead of create_tmp_var and use a
27758         TARGET_EXPR.
27759         (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27760         instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27761         ubsan_encode_value callers.
27763         PR sanitizer/81111
27764         * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27765         use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27766         just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27768 2017-06-19  Richard Biener  <rguenther@suse.de>
27770         PR middle-end/81118
27771         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27772         estimates if we changed anything.
27774 2017-06-19  Richard Biener  <rguenther@suse.de>
27776         PR tree-optimization/80887
27777         * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27778         (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27779         simplified lookups, then reset mprts_hook.
27780         (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27781         simplifying.
27782         (try_to_simplify): Likewise.
27784 2017-06-19  Martin Liska  <mliska@suse.cz>
27786         PR sanitizer/80879
27787         * gimplify.c (gimplify_switch_expr):
27788         Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27790 2017-06-19  Martin Liska  <mliska@suse.cz>
27792         * doc/install.texi: Document that PGO runs in 4 stages.
27794 2017-06-19  Martin Liska  <mliska@suse.cz>
27796         PR ipa/80732
27797         * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27798         to dispatcher function name.
27799         * multiple_target.c (replace_function_decl): New function.
27800         (create_dispatcher_calls): Redirect both edges and references.
27802 2017-06-19  Jan Hubicka <hubicka@ucw.cz>
27804         * profile-count.c (profile_count::dump): Dump quality.
27805         (profile_count::differs_from_p): Update for unsigned val.
27806         * profile-count.h (profile_count_quality): New enum.
27807         (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27809 2017-06-19  Richard Biener  <rguenther@suse.de>
27811         * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27812         struct function as arg.
27813         (estimate_numbers_of_iterations): Export overload with loop arg.
27814         (free_numbers_of_iterations_estimates_loop): Use an overload of
27815         free_numbers_of_iterations_estimates instead.
27816         * tree-cfg.c (remove_bb): Adjust.
27817         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27818         * tree-parloops.c (gen_parallel_loop): Likewise.
27819         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27820         Likewise.
27821         (tree_unroll_loops_completely): Likewise.
27822         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27823         Use an overload instead and export.
27824         (estimated_loop_iterations): Adjust.
27825         (max_loop_iterations): Likewise.
27826         (likely_max_loop_iterations): Likewise.
27827         (estimate_numbers_of_iterations): Take struct function as arg
27828         and adjust.
27829         (loop_exits_before_overflow): Adjust.
27830         (free_numbers_of_iterations_estimates_loop): Use an overload.
27831         * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27832         * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27834 2017-06-19  Richard Biener  <rguenther@suse.de>
27836         PR ipa/81112
27837         * ipa-prop.c (find_constructor_constant_at_offset): Handle
27838         RANGE_EXPR conservatively.
27840 2017-06-16  Carl Love  <cel@us.ibm.com>
27842         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27843         definitions for vec_float, vec_float2, vec_floato,
27844         vec_floate built-ins.
27845         * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27846         for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27847         floate.
27848         * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27849         FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27850         UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27851         * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27852         define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27853         * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27854         vec_floato): Add builtin defines.
27855         * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27856         Update the built-in documentation file for the new built-in
27857         functions.
27859 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27861         * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27862         (mthumb): Mark as the negative of -marm.
27864 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27866         * doc/invoke.texi (ARM Options, -mcpu): Document supported
27867         extension options.
27868         (ARM Options, -mtune): Document that this accepts the same
27869         extension options as -mcpu.
27870         (ARM Options, -mfpu): Document addition of -mfpu=auto.
27872 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27874         * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27875         permitted extensions.
27877 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27879         * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27880         (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27881         (armv8-m.main): Add option +nodsp.
27882         * config/arm/arm-cpu-cdata.h: Regenerated.
27884 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27886         * config/arm/t-fuchsia: New file.
27887         * config.gcc (arm*-*-fuchsia*): Use it.
27889 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27891         * config/arm/t-symbian: Rewrite for new option infrastructure.
27893 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27895         * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27896         (MULTILIB_REQUIRED): Likewise.
27898 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27900         * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
27901         (MULTILIB_RESUE): Likewise.
27902         (MULTILIB_MATCHES): Likewise.
27903         (MULTLIB_REQUIRED): Likewise.
27905 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27907         * config/arm/t-rtems: Rewrite for new option framework.
27909 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27911         * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
27912         (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
27913         (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
27914         (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
27915         * config/arm/t-multilib: ... here.
27916         (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
27917         (MULTILIB_MATCHES): Use armv7 libraries for armv7-r.  Also use for
27918         armv7-a and armv8*-a when A-profile libraries have not been built.
27919         * config/arm/t-rmprofile: Rewrite.
27921 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27923         * genmultilib (multilib_reuse): Allow an explicit period to be escaped
27924         with a backslash.  Remove the backslash after substituting unescaped
27925         periods.
27926         * doc/fragments.texi (MULTILIB_REUSE): Document it.
27928 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27930         * config.gcc: (arm*-*-*): When building a-profile libraries, force
27931         the driver to pass through the default setting of -mfloat-abi.
27932         * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
27933         rather than NULL.
27934         * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
27935         (all_feat_combs): New rule.
27936         (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories.  Rework
27937         default libraries.
27938         * config/arm/t-aprofile: Rewrite.
27940 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27942         * config/arm/arm.h (FPUTYPE_AUTO): Define.
27943         * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
27944         fpu is not specified by the user/command-line.
27945         * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
27946         * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
27947         * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
27948         * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
27949         * common/config/arm/arm-common.c (arm_canon_arch_option): Use
27950         FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
27952 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27954         * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
27955         * config/arm/t-arm-elf: Rewritten.
27957 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27959         * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
27960         have some floating-point instructions.
27961         (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
27962         (TARGET_MAYBE_HARD_FLOAT): New macro.
27963         * config/arm/arm-builtins.c (arm_init_builtins): Use
27964         TARGET_MAYBE_HARD_FLOAT.
27965         * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
27967 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27969         * common/config/arm/arm-common.c: Define INCLUDE_LIST.
27970         (configargs.h): Include it.
27971         (arm_print_hint_for_fpu_option): New function.
27972         (arm_parse_fpu_option): New function.
27973         (candidate_extension): New class.
27974         (arm_canon_for_multilib): New function.
27975         * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
27976         (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
27977         (ARCH_CANONICAL_SPECS): New macro.
27978         (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
27980 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27982         * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
27983         are set after handling multilib fragments.  Set target_cpu_default2
27984         from with_cpu.
27986 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27988         * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
27989         cpu name.
27990         (arm*-*-*): Set target_cpu_default2 to a quoted string.
27991         * config/arm/parsecpu.awk (check_cpu): Validate any extension
27992         options.
27993         (check_arch): Likewise.
27994         * config/arm/arm.c (arm_configure_build_target): Handle
27995         TARGET_CPU_DEFAULT being a string constant.  Scan any feature
27996         options in the default.
27998 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28000         * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
28001         when an option is an alias of another.
28002         * config/arm/parsecpu.awk (optalias): New parser token.
28003         (gen_comm_data): Mark non-alias options as such.  Emit entries
28004         for extension aliases.
28005         * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
28006         (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
28007         (armv6kz, armv6zk, armv6t2): Likewise.
28008         (armv7): Make vfpv3-d16 an alias.
28009         (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases.  Sort in
28010         canonical order.
28011         (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
28012         Sort in canonical order.
28013         (armv8-a): Sort in canonical order.
28014         (armv8.1-a, armv8.2-a):  Likewise.
28015         (generic-armv7-a): Make neon and neon-vfpv3 aliases.  Sort in
28016         canonical order.
28017         (cortex-a9): Sort in canonical order.
28018         * config/arm/arm.c (selftests.h): Include it.
28019         (arm_test_cpu_arch_data): New function.
28020         (arm_run_self_tests): New function.
28021         (TARGET_RUN_TARGET_SELFTESTS): Redefine.
28022         (targetm): Move declaration to the end of the file.
28023         * arm-cpu-cdata.h: Regenerated.
28025 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28027         * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
28028         call to target_mode_check describing the type of option passed.
28029         * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
28030         (arm_target_thumb_only): Use arm_parse_arch_option_name or
28031         arm_parse_cpu_option_name to match parameters against list of
28032         available targets.
28033         * config/arm/parsecpu.awk (gen_comm_data): Don't generate
28034         arm_arch_core_flags data structure.
28035         * config/arm/arm-cpu_cdata.h: Regenerated.
28037 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28039         * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
28040         config/arm/arm.c.
28041         (arm_print_hint_for_cpu_option): Likewise.
28042         (arm_print_hint_for_arch_option): Likewise.
28043         (arm_parse_cpu_option_name): Likewise.
28044         (arm_parse_arch_option_name): Likewise.
28045         * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
28046         of entries in the all_fpus list.
28047         * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
28048         (arm_parse_cpu_option_name): Declare.
28049         (arm_parse_arch_option_name): Declare.
28050         (arm_parse_option_features): Declare.
28051         (arm_intialize_isa): Declare.
28052         * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
28053         data tables to ...
28054         (gen_comm_data): ... here.  Make definitions non-static.
28055         * config/arm/arm-cpu-data.h: Regenerated.
28056         * config/arm/arm-cpu-cdata.h: Regenerated.
28058 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28060         * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
28061         (cpu_arch_extension): New structure.
28062         (cpu_arch_option, arch_option, cpu_option): New structures.
28063         * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
28064         architecture types.
28065         (gen_data): Generate new format data tables.
28066         * config/arm/arm.c (cpu_tune): New structure.
28067         (cpu_option, processors): Delete.
28068         (arm_print_hint_for_core_or_arch): Delete.  Replace with ...
28069         (arm_print_hint_for_cpu_option): ... this and ...
28070         (arm_print_hint_for_arch_option): ... this.
28071         (arm_parse_arch_cpu_name): Delete.  Replace with ...
28072         (arm_parse_cpu_option_name): ... this and ...
28073         (arm_parse_arch_option_name): ... this.
28074         (arm_unrecognized_feature): Change type of target parameter to
28075         cpu_arch_option.
28076         (arm_parse_arch_cpu_features): Delete.  Replace with ...
28077         (arm_parse_option_features): ... this.
28078         (arm_configure_build_target): Rework to use new configuration data
28079         tables.
28080         (arm_print_tune_info): Rework for new configuration data tables.
28081         * config/arm/arm-cpu-data.h: Regenerated.
28082         * config/arm/arm-cpu.h: Regenerated.
28084 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28086         * Makefile.in (OBJS): Move sbitmap.o from here ...
28087         (OBJS-libcommon): ... to here.
28089 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28091         * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
28092         (ISA_ALL_CRYPTO): New macro.
28093         (ISA_ALL_SIMD): New macro
28094         (ISA_ALL_FP): New macro.
28095         * config/arm/arm.c (fpu_bitlist): Update initializer.
28096         * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
28097         simd or fp.
28098         (arm9e): Add fpu.  Add option for nofp
28099         (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
28100         (arm926ej-s, arm1026ej-s): Likewise.
28101         (generic-armv7-a): Add fpu.  Add options for simd, vfpv3, vfpv3-d16,
28102         vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
28103         neon-fp16, neon-vfpv4, nofp and nosimd.
28104         (cortex-a5, cortex-a7): Add fpu.  Add options for nosimd and nofp.
28105         (cortex-a8): Add fpu.  Add option for nofp.
28106         (cortex-a9): Add fpu.  Add options for nosimd and nofp.
28107         (cortex-a12, cortex-a15, cortex-a17): Add fpu.  Add option for nofp.
28108         (cortex-r4f): Add fpu.
28109         (cortex-r5): Add fpu.  Add options for nofp.dp and nofp.
28110         (cortex-r7): Use idiv option from architecture.  Add fpu.  Add option
28111         for nofp.
28112         (cortex-r8): Likewise.
28113         (cortex-m4): Add fpu.  Add option for nofp.
28114         (cortex-a15.cortex-a7): Add fpu.  Add option for nofp.
28115         (cortex-a17.cortex-a7): Likewise.
28116         (cortex-a32): Add fpu.  Add options for crypto and nofp.
28117         (cortex-a35, cortex-a53): Likewise.
28118         (cortex-a57): Add fpu.  Add option for crypto.
28119         (cortex-a72, cortex-a73): Likewise.
28120         (exynos-m1): Likewise.
28121         (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
28122         (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
28123         (cortex-m33): Add fpu.  Add option for nofp.
28124         * config/arm/arm-cpu-cdata.h: Regenerated
28125         * config/arm/arm-cpu-data.h: Regenerated.
28127 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28129         * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
28130         (armv5te, armv5tej): Likewise.
28131         (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
28132         (armv7): Add options fp and vfpv3-d16.
28133         (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
28134         vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
28135         nofp and nosimd.
28136         (armv7ve): Likewise.
28137         (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
28138         (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
28139         (armv8-a): Add nocrypto option.
28140         (armv8.1-a, armv8.2-a): Likewise.
28141         (armv8-m.main): add options fp, fp.dp and nofp.
28143 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28145         * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
28146         nofp.
28147         (armv8-a+crc): Delete.
28148         (armv8.1-a): Add options simd, crypto and nofp.
28149         (armv8.2-a): Add options fp16, simd, crypto and nofp.
28150         (armv8.2-a+fp16): Delete.
28151         (armv8-m.main): Add option dsp.
28152         (armv8-m.main+dsp): Delete.
28153         (cortex-a8): Add fpu.  Add nofp option.
28154         (cortex-a9): Add fpu.  Add nofp and nosimd options.
28155         * config/arm/parsecpu.awk (gen_data): Generate option tables and
28156         link to main cpu and architecture data structures.
28157         (gen_comm_data): Only put isa attributes from the main architecture
28158         in common tables.
28159         (option): New statement for architecture and CPU entries.
28160         * arm.c (struct cpu_option): New structure.
28161         (struct processors): Add entry for options.
28162         (arm_unrecognized_feature): New function.
28163         (arm_parse_arch_cpu_name): Ignore any characters after the first
28164         '+' character.
28165         (arm_parse_arch_cpu_feature): New function.
28166         (arm_configure_build_target): Separate out any CPU and architecture
28167         features and parse separately.  Don't error out if -mfpu=auto is
28168         used with only an architecture string.
28169         (arm_print_asm_arch_directives): New function.
28170         (arm_file_start): Call it.
28171         * config/arm/arm-cpu-cdata.h: Regenerated.
28172         * config/arm/arm-cpu-data.h: Likewise.
28173         * config/arm/arm-tables.opt: Likewise.
28175 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28177         * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
28178         assembler when it is not -mfpu=auto.
28180 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28182         * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
28183         (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
28184         (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
28185         (ASM_CPU_SPEC): Rewrite.
28186         (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
28187         (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef.  Use
28188         MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS.  Remove
28189         reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
28190         * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
28191         copied string is NUL-terminated.  Also strip any characters prefixed
28192         by '+'.
28193         (arm_rewrite_selected_arch): New function.
28194         (arm_rewrite_march): New function.
28196 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28198         * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
28199         (x_arm_cpu_string, x_arm_tune_string): Likewise.
28200         (march, mcpu, mtune): Convert to string-based options.
28201         * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
28202         (arm_parse_arch_cpu_name): New function.
28203         (arm_configure_build_target): Use arm_parse_arch_cpu_name to
28204         identify selected architecture or CPU.
28205         (arm_option_save): New function.
28206         (TARGET_OPTION_SAVE): Redefine.
28207         (arm_option_restore): Restore string options.
28208         (arm_option_print): Print string options.
28210 2017-06-16  Martin Sebor  <msebor@redhat.com>
28212         PR tree-optimization/80933
28213         PR tree-optimization/80934
28214         * builtins.c (fold_builtin_3): Do not handle bcmp here.
28215         * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
28216         (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
28217         (gimple_fold_builtin): Call them.
28219 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28221         * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
28222         as unlikely; update profile.
28224 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28226         * predict.c (force_edge_cold): Handle declaring edges impossible
28227         more aggresively.
28229 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28231         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28232         profile.
28233         (try_unroll_loop_completely): Fix reporting.
28235 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28237         * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28239 2017-06-16  James Greenhalgh  <james.greenhalgh@arm.com>
28241         PR target/71778
28242         * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28243         if given a non-constant argument for an intrinsic which requires a
28244         constant.
28246 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28248         * profile.c (compare_freqs): New function.
28249         (branch_prob): Sort edge list.
28250         (find_spanning_tree): Assume that the list is priority sorted.
28252 2017-06-16  Richard Biener  <rguenther@suse.de>
28254         PR tree-optimization/81090
28255         * passes.def (pass_record_bounds): Remove.
28256         * tree-pass.h (make_pass_record_bounds): Likewise.
28257         * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28258         make_pass_record_bounds): Likewise.
28259         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28260         not free niter estimates at the beginning but at the end.
28261         * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28263 2017-06-16  Richard Biener  <rguenther@suse.de>
28265         * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28266         initializer to workaround ICE in host GCC 4.8.
28268 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28270         * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28271         counts.
28272         (clone_inlined_nodes): Update.
28274 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28276         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28277         prefetch settings, and enable prefetching by default at -O3.
28279 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28281         * config/aarch64/aarch64.c (aarch64_override_options_internal):
28282         Set flag_prefetch_loop_arrays according to tuning data.
28284 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28286         * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28287         New tune structure.
28288         (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28289         [Unrelated to main purpose of the patch] Place the pointer field last
28290         to enable type checking errors when tune structure are wrongly merged.
28291         * config/aarch64/aarch64.c (generic_prefetch_tune,)
28292         (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28293         (thunderx2t99_prefetch_tune): New tune constants.
28294         (tune_params *_tunings): Update all tunings (no functional change).
28295         (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28296         PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28297         from tunings structures.
28299 2017-06-16  Jakub Jelinek  <jakub@redhat.com>
28301         PR sanitizer/81094
28302         * ubsan.c (instrument_null): Add T argument, use it instead
28303         of computing it based on IS_LHS.
28304         (instrument_object_size): Likewise.
28305         (pass_ubsan::execute): Adjust instrument_null and
28306         instrument_object_size callers to pass gimple_get_lhs or
28307         gimple_assign_rhs1 result to it.  Use instrument_null instead of
28308         calling get_base_address and instrument_mem_ref.  Handle
28309         aggregate call arguments for object-size sanitization.
28311 2017-06-16  Yury Gribov  <tetra2005@gmail.com>
28313         PR tree-optimization/81089
28314         * tree-vrp.c (is_masked_range_test): Validate operands of
28315         subexpression.
28317 2017-06-15  Martin Sebor  <msebor@redhat.com>
28319         PR c++/80560
28320         * dumpfile.c (dump_register): Avoid calling memset to initialize
28321         a class with a default ctor.
28322         * gcc.c (struct compiler): Remove const qualification.
28323         * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28324         * hash-table.h: Ditto.
28325         * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28326           assignment.
28327         * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28328         * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28329         default ctor.
28330         * params.h (struct param_info): Make struct members non-const.
28331         * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28332         with default initialization.
28333         * vec.h (vec_copy_construct, vec_default_construct): New helper
28334         functions.
28335         (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28336         with vec_copy_construct.
28337         (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28338         (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28339         * doc/invoke.texi (-Wclass-memaccess): Document.
28341 2017-06-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28343         * emit-rtl.h (is_leaf): Update comment about local
28344         register allocator.
28346 2017-06-15  Jozef Lawrynowicz   <jozef.l@somniumtech.com>
28348         PR target/78818
28349         * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28350         for a variable to have a section before checking if the section has a
28351         name.
28352         Set section to.persistent if persistent attribute is set.
28353         Warn if .persistent attribute is used on an automatic variable.
28355 2017-06-15  Eric Botcazou  <ebotcazou@adacore.com>
28357         PR rtl-optimization/80474
28358         * reorg.c (update_block): Do not ignore instructions in a delay slot.
28360 2017-06-15  Segher Boessenkool  <segher@kernel.crashing.org>
28362         * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28363         of REGNO.
28365 2017-06-14  Maciej W. Rozycki  <macro@imgtec.com>
28367         * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28368         (casesi): Emit bounds checking as RTL.
28369         (casesi_internal_mips16_<mode>): Remove bounds checking.
28371 2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
28373         * config/xtensa/xtensa.c (xtensa_option_override): Append
28374         MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28375         (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28376          xtensa_doloop_hooks): Define unconditionally.
28377         (xtensa_reorg_loops): Only call reorg_loops in the presence of
28378         TARGET_LOOPS.
28379         * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28380         (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28381         for it in xtensa_option_override.
28382         (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28383          HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28385 2017-06-14  Boris Kolpackov  <boris@codesynthesis.com>
28387         * doc/cppopts.texi: Document '-' special value to -MF.
28389 2017-06-14  Wilco Dijkstra  <wdijkstr@arm.com>
28391         * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28392         (cortex_a53_fconst): Likewise.
28393         (cortex_a53_fpmul): Likewise.
28394         (cortex_a53_f_load_64): Likewise.
28395         (cortex_a53_f_load_many): Likewise.
28396         (cortex_a53_advsimd_alu): Likewise.
28397         (cortex_a53_advsimd_alu_q): Likewise.
28398         (cortex_a53_advsimd_mul): Likewise.
28399         (cortex_a53_advsimd_mul_q): Likewise.
28400         (fpmac bypass): Add new bypass for fpmac-fpmac case.
28401         Add missing fmul, r2f_cvt and fconst cases.
28403 2017-06-14  Richard Biener  <rguenther@suse.de>
28405         PR middle-end/81088
28406         * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28407         literal constants.
28408         (fold_binary_loc): When associating do not treat pre-existing
28409         TREE_OVERFLOW on literal constants as a reason to allow
28410         TREE_OVERFLOW on associated literal constants.
28412 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28414         * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28415         (MASK_FEATURES): New macro.
28416         * config/sparc/sparc.c (sparc_option_override): Remove the special
28417         handling of -mfpu and generalize it to all MASK_FEATURES switches.
28419 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28421         * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28422         a division of 0 if non-call exceptions are enabled.
28424 2017-06-14  Andrew Pinski  <apinski@cavium.com>
28425             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
28427         PR target/71663
28428         * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28429         Improve vector initialization code gen for only variable case.
28431 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28433         * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28435 2017-06-14  Richard Biener  <rguenther@suse.de>
28437         PR tree-optimization/81083
28438         * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28439         as values.
28441 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28443         * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28444         (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28445         * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28446         * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete.  Adjust former use.
28447         * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28448         * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28450 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28452         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28453         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28455 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28457         * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28459 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28461         * config/rs6000/t-rtems: Don't handle SPE.
28463 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28465         * config/rs6000/t-linux: Don't handle SPE.
28467 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28469         * config/rs6000/eabispe.h: Delete file.
28471 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28473         * config/rs6000/t-spe: Delete file.
28475 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28477         * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28478         (rs6000_legitimate_offset_address_p): Return false for anything in
28479         V2SImode or V2SFmode.
28481 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28483         * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28484         except V2SF and V2SI.  Rearrange the vector modes, and add comments.
28485         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28486         and V4HImode.
28487         (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28488         (rs6000_legitimate_offset_address_p): Ditto.
28489         (rs6000_emit_move): Ditto.
28490         (rs6000_init_builtins): Remove V4HI_type_node.
28492 2017-06-13  Martin Liska  <mliska@suse.cz>
28494         PR sanitize/78204
28495         * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28496         (gate_asan): Likewise.
28497         * asan.h (asan_no_sanitize_address_p): Remove the function.
28498         (sanitize_flags_p): New function.
28499         * builtins.def: Fix coding style.
28500         * common.opt: Use renamed enum value.
28501         * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28502         * doc/extend.texi: Document no_sanitize attribute.
28503         * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28504         to SANITIZE_UNDEFINED_NONDEFAULT.
28505         * gcc.c (sanitize_spec_function): Use the renamed enum value.
28506         * gimple-fold.c (optimize_atomic_compare_exchange_p):
28507         Use sanitize_flags_p.
28508         * gimplify.c (gimplify_function_tree): Likewise.
28509         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28510         * opts.c (parse_no_sanitize_attribute): New function.
28511         (common_handle_option): Use renamed enum value.
28512         * opts.h (parse_no_sanitize_attribute): Declare.
28513         * tree.c (sanitize_flags_p): New function.
28514         * tree.h: Declared here.
28515         * tsan.c: Use sanitize_flags_p.
28516         * ubsan.c (ubsan_expand_null_ifn): Likewise.
28517         (instrument_mem_ref): Likewise.
28518         (instrument_bool_enum_load): Likewise.
28519         (do_ubsan_in_current_function): Remove the function.
28520         (pass_ubsan::execute): Use sanitize_flags_p.
28521         * ubsan.h: Remove do_ubsan_in_current_function
28522         * tree-cfg.c (print_no_sanitize_attr_value): New function.
28523         (dump_function_to_file): Use it here.
28525 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28527         PR tree-optimization/80803
28528         PR tree-optimization/81063
28529         * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28530         (propagate_subaccesses_across_link): Enqueue subtree whenever
28531         necessary instead of relying on the caller.
28533 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28535         * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28536         that have a first_link.
28537         (sort_and_splice_var_accesses): Do not check first_link before
28538         enquing.
28539         (subtree_mark_written_and_enqueue): Likewise.
28540         (propagate_all_subaccesses): Likewise and do not stop at first
28541         parent with a first_link.
28543 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28545         * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28546         instead of f.
28548 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28550         * match.pd: New pattern.
28552 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28554         * tree-vrp.c (is_masked_range_test): New function.
28555         (register_edge_assert_for): Determine ranges for
28556         some bit tests.
28558 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28560         PR tree-optimization/67328
28561         * fold-const.c (maskable_range_p): New function.
28562         (build_range_check): Generate bittests if possible.
28564 2017-06-13  Martin Liska  <mliska@suse.cz>
28566         * gimple-pretty-print.c (dump_probability): Add new argument.
28567         (dump_edge_probability): Dump both probability and count.
28568         (dump_gimple_label): Likewise.
28569         (dump_gimple_bb_header): Likewise.
28571 2017-06-13  Georg-Johann Lay  <avr@gjlay.de>
28573         PR target/81072
28574         * config/avr/avr-devices.c: Fix indentation.
28575         * config/avr/gen-avr-mmcu-specs.c: Dito.
28577 2017-06-13  Richard Biener  <rguenther@suse.de>
28579         * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28580         instead get vector type from stmt_info.
28581         (vectorizable_reduction): Adjust.  Remove dead code.
28583 2017-06-13  Richard Biener  <rguenther@suse.de>
28585         PR middle-end/81065
28586         * fold-const.c (extract_muldiv_1): Remove bogus distribution
28587         case of C * (x * C2 + C3).
28588         (fold_addr_of_array_ref_difference): Properly fold index difference.
28590 2017-06-12  David S. Miller  <davem@davemloft.net>
28592         PR target/80968
28593         * config/sparc/sparc.md (return expander): Emit frame blockage if
28594         function uses alloca.
28596 2017-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
28598         * combine.c (make_field_assignment): Check len rather than the mode
28599         precision when calling force_to_mode.
28601 2017-06-12  Georg-Johann Lay  <avr@gjlay.de>
28603         Support multilibs and devices that see flash in RAM address range.
28605         PR target/81072
28606         * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28607         (avr_mcu_t) <flash_pm_offset>: New field.
28608         (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28609         * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28610         (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28611         (AVR_TINY_PM_OFFSET): Remove macro.
28612         * config/avr/avr.opt (-mshort-calls): New option.
28613         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28614         [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28615         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28616         <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28617         <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28618         instead of avr_arch->have_jmp_call.
28619         <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28620         [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28621         avr_arch->flash_pm_offset to define.
28622         * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28623         new field flash_pm_offset.  Add entry for avrxmega3.
28624         (avr_texinfo): Add entry for avrxmega3.
28625         * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28626         attiny212, attiny214,
28627         attiny412, attiny414, attiny416, attiny417,
28628         attiny814, attiny816, attiny817,
28629         attiny1614, attiny1616, attiny1617,
28630         attiny3214, attiny3216, attiny3217.
28631         * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28632         avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28633         (avr_print_operand_address) [AVR_TINY]: Same.
28634         (avr_asm_init_sections) <readonly_data_section>: Only patch
28635         callback if avr_arch->flash_pm_offset = 0.
28636         (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28637         for rodata if avr_arch->flash_pm_offset != 0.
28638         (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28639         * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28640         (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28641         (m_options): Append opt_rcall.
28642         (m_dirnames): Append dir_rcall.
28643         * config/avr/t-multilib: Regenerate.
28645         * configure.ac [target=avr]: Check whether avrxmega3 default
28646         linker description file works as needed.
28647         * configure: Regenerate.
28648         * doc/avr-mmcu.texi: Regenerate.
28649         * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28650         <__AVR_ARCH__>: Document avrxmega3 and 103.
28651         <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28652         <__AVR_SHORT_CALLS__>: Document it.
28653         <__AVR_PM_BASE_ADDRESS__>: Document it.
28654         * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28655         (AVR Variable Attributes) <progmem>: Document this is
28656         not needed for avrxmega3.
28657         (AVR Named Address Spaces) <__flash>: Dito.
28659 2017-06-12  Jan Hubicka  <hubicka@ucw.cz>
28661         * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28663 2017-06-12  Doug Rupp  <rupp@adacore.com>
28665         * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28666         Append vxworks-stdint.h to the tm_file list.
28667         * config/vxworks-stdint.h: New file.
28669 2017-06-12  Martin Liska  <mliska@suse.cz>
28671         PR tree-optimization/81041
28672         * tree-profile.c (gimple_gen_ic_func_profiler):
28673         Create an extra BB in profile-generate
28674         (gimple_gen_time_profiler): Likewise.
28676 2017-06-12  Jakub Jelinek  <jakub@redhat.com>
28678         PR tree-optimization/81003
28679         * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28680         (update_range_test): Use it instead of force_gimple_operand_gsi.
28682 2017-06-12  Richard Biener  <rguenther@suse.de>
28684         PR tree-optimization/81053
28685         * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28686         with backedge value not defined in loop.  Simplify def stmt
28687         compute.
28689 2017-06-11  Tom de Vries  <tom@codesourcery.com>
28691         PR target/79939
28692         * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28693         Return true.
28694         (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28695         nvptx_cannot_force_const_mem.
28697 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28699         * opts.c (finish_options): Move test for flag_split_stack after
28700         it has been initialized.
28702 2017-06-11  Jason Merrill  <jason@redhat.com>
28704         * tree.h (id_equal): New.
28705         * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28706         omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28707         instead of strcmp of IDENTIFIER_POINTER.
28709 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28711         * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28712         (mark_all_inlined_calls_cdtor): Fix formating.
28713         (inline_transform): Rescale profile before inlining.
28715 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28717         * cgraph.h (cgraph_edge::clone): Update prototype.
28718         * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28719         (cgraph_node::create_clone): Update.
28720         (cgraph_node::create_version_clone): Update.
28721         * tree-inline.c (copy_bb): Update.
28722         (expand_call_inline): Update.
28724 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28726         * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28727         factored out from ...
28728         (rs6000_emit_prologue): ... here.
28730 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28732         * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28733         factored out from ...
28734         (rs6000_emit_prologue): ... here.
28736 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28738         * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28739         edge counts.
28740         (handle_missing_profiles): Fix computation of tp_first_run.
28741         (counts_to_freqs): Do not touch freqs when count is 0.
28743 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28745         * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28746         profile.
28748 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28750         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28751         attributes): Document signal effective target.
28753 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28755         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28756         Document effective target stack_size.
28758 2017-06-09  David Malcolm  <dmalcolm@redhat.com>
28760         * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28761         to the edit_context if they can be auto-applied.
28763 2017-06-9  Ian Lance Taylor  <iant@golang.org>
28765         * opts.c (finish_options): If -fsplit-stack, disable implicit
28766         -forder-blocks-and-partition.
28767         * doc/invoke.texi (Optimize Options): Document that when using
28768         -fsplit-stack -forder-blocks-and-partition is not implicitly
28769         enabled.
28771 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28773         * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28774         ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28775         ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28776         * builtins.def (abort, trap, unreachable): Declare cold.
28777         * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28778         * tree-core.h (ECF_COLD): New.
28779         * tree.c (set_call_expr_flags): Handle ECF_COLD.
28780         (build_common_builtin_nodes): Mark unreachable and abort as cold.
28782 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28784         * predict.c (unlikely_executed_stmt_p): Cleanup.
28786 2017-06-09  Richard Biener  <rguenther@suse.de>
28788         * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28789         model if the ref is always written to.
28791 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28793         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28795 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28797         * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28798         than udiv.
28800 2017-06-09  Tom de Vries  <tom@codesourcery.com>
28802         PR target/80855
28803         * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28804         "target cannot support label values" when encountering LABEL_REF.
28806 2017-06-09  Martin Liska  <mliska@suse.cz>
28808         * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28809         (gimple_gen_ic_func_profiler): Emit direct comparison
28810         of __gcov_indirect_call_callee with NULL.
28811         (gimple_gen_time_profiler): Change probability from
28812         PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28814 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28816         * profile.c (edge_gcov_counts): Turn to pointer.
28817         (compute_branch_probabilities, compute_branch_probabilities): Update.
28818         (branch_prob): Do not clear edge_gcov_count.
28819         * profile.h (edge_gcov_counts): Turn to pointer.
28820         (edge_gcov_count): Update.
28822 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28824         * gimple.h (gimple_check_failed): Mark cold.
28826 2017-06-09  Richard Biener  <rguenther@suse.de>
28828         PR tree-optimization/66623
28829         * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28830         refactor check_reduction into two parts, properly computing
28831         whether we have to check reduction validity for outer loop
28832         vectorization.
28834 2017-06-09  Richard Biener  <rguenther@suse.de>
28836         PR tree-optimization/79483
28837         * graphite-scop-detection.c (order): New global.
28838         (get_order): Compute bb to order mapping that satisfies code
28839         generation constraints.
28840         (cmp_pbbs): New helper.
28841         (build_scops): Start domwalk at entry block, sort generated
28842         pbbs.
28844 2017-06-09  Richard Biener  <rguenther@suse.de>
28846         PR middle-end/81007
28847         * ipa-polymorphic-call.c
28848         (ipa_polymorphic_call_context::restrict_to_inner_class):
28849         Skip FIELD_DECLs with error_mark_node type.
28850         * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28851         last again.
28853 2017-06-09  Martin Liska  <mliska@suse.cz>
28855         * predict.c (struct branch_predictor): New struct.
28856         (test_prediction_value_range): New test.
28857         (predict_c_tests): New function.
28858         * selftest-run-tests.c (selftest::run_tests): Run the function.
28859         * selftest.h: Declare new tests.
28861 2017-06-09  Segher Boessenkool  <segher@kernel.crashing.org>
28863         PR target/80966
28864         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28865         gen_add3_insn did not fail.
28866         * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28867         r0, construct that number in a temporary reg and add that reg to r0.
28868         If asked to put the result in r0 as well, fail.
28870 2017-06-08  Will Schmidt  <will_schmidt@vnet.ibm.com>
28872         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28873         for early expansion of vec_eqv.
28875 2017-06-08  Jakub Jelinek  <jakub@redhat.com>
28877         PR middle-end/81005
28878         * ubsan.c (instrument_null): Avoid pointless code temporary.
28879         (pass_ubsan::execute): Instrument aggregate arguments of calls.
28881 2017-06-08  Uros Bizjak  <ubizjak@gmail.com>
28883         PR target/81015
28884         Revert:
28885         2016-12-14  Uros Bizjak  <ubizjak@gmail.com>
28887         PR target/59874
28888         * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28889         (*clzhi2): Ditto.
28891 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28893         * predict.c (unlikely_executed_edge_p): Move ahead.
28894         (probably_never_executed_edge_p): Use it.
28896 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28898         PR middle-end/79988
28899         * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
28900         gimple_call_builtin_p call.
28902 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28904         * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
28905         * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
28906         rtl_check_failed_type2, rtl_check_failed_code1,
28907         rtl_check_failed_code2, rtl_check_failed_code_mode,
28908         rtl_check_failed_block_symbol, cwi_check_failed_bounds,
28909         rtvec_check_failed_bounds, rtl_check_failed_flag,
28910         _fatal_insn_not_found, _fatal_insn): Likewise.
28911         * tree.h (tree_contains_struct_check_failed,
28912         tree_check_failed, tree_not_check_failed,
28913         tree_class_check_failed, tree_range_check_failed,
28914         tree_not_class_check_failed, tree_int_cst_elt_check_failed,
28915         tree_vec_elt_check_failed, phi_node_elt_check_failed,
28916         tree_operand_check_failed, omp_clause_check_failed,
28917         omp_clause_operand_check_failed, omp_clause_range_check_failed):
28918         Likewise.
28920 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28922         * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
28923         flag_branch_probabilities.
28924         * ipa-inline.c (edge_badness): Likewise.
28925         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
28926         * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
28927         * predict.c (maybe_hot_frequency_p): Likewise.
28928         (probably_never_executed): Likewise.
28929         * sched-ebb.c (schedule_ebbs): Likewise.
28930         * sched-rgn.c (find_single_block_region): Likewise.
28931         * tracer.c (tail_duplicate): Likewise.
28933 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28935         * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
28936         longer requires x_flag_profile_use.
28938 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28940         * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
28941         instead of flag_reorder_blocks_and_partition.
28942         * dbxout.c (dbxout_function_end): Likewise.
28943         * dwarf2out.c (gen_subprogram_die): Likewise.
28944         * haifa-sched.c (sched_create_recovery_edges): Likewise.
28945         * hw-doloop.c (reorg_loops): Likewise.
28946         * varasm.c (assemble_start_function,
28947         assemble_end_function): Likewise.
28948         (decide_function_section): Do not check for
28949         flag_reorder_blocks_and_partition.
28951 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28953         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
28954         New function.
28955         (chkp_get_hard_register_fake_addr_expr): Ditto.
28956         (chkp_build_addr_expr): Add check for hard reg case.
28957         (chkp_parse_array_and_component_ref): Ditto.
28958         (chkp_find_bounds_1): Ditto.
28959         (chkp_process_stmt): Don't generate bounds store for
28960         hard reg case.
28962 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28964         * predict.c (maybe_hot_bb_p): Do not check profile status.
28965         (maybe_hot_edge_p): Likewise.
28966         (probably_never_executed): Check for zero counts even if profile
28967         is not read.
28968         (unlikely_executed_edge_p): New function.
28969         (unlikely_executed_stmt_p): New function.
28970         (unlikely_executed_bb_p): New function.
28971         (set_even_probabilities): Use unlikely predicates.
28972         (combine_predictions_for_bb): Likewise.
28973         (predict_paths_for_bb): Likewise.
28974         (predict_paths_leading_to_edge): Likewise.
28975         (determine_unlikely_bbs): New function.
28976         (estimate_bb_frequencies): Use it.
28977         (compute_function_frequency): Use zero counts even if profile is
28978         not read.
28979         * profile-count.h: Fix typo.
28981 2017-08-08  Julia Koval  <julia.koval@intel.com>
28983         * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
28984         _mm512_mask_cvtsepi16_storeu_epi8,
28985         _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
28986         * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
28987         _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
28988         _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
28989         _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
28990         * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
28991         (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
28992         VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
28993         * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
28994         __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
28995         __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
28996         __builtin_ia32_pmovuswb256mem_mask,
28997         __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
28998         __builtin_ia32_pmovwb512mem_mask): New builtins.
29000 2017-08-08  Julia Koval  <julia.koval@intel.com>
29002         PR target/73350,80862
29003         * config/i386/subst.md (round): Fix round pattern.
29004         * config/i386/i386.c (ix86_erase_embedded_rounding):
29005         Fix erasing rounding for the fixed pattern.
29007 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
29009         * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
29011 2017-06-08  Martin Liska  <mliska@suse.cz>
29013         PR gcov-profile/80911
29014         * gcov.c (block_info::block_info): New constructor.
29016 2017-06-07  Carl Love  <cel@us.ibm.com>
29018         * config/rs6000/rs6000-c: The return type of the following
29019         built-in functions was implemented as int not long long.  Fix sign
29020         of return value for the unsigned version of vec_mulo and vec_mule.
29021         vector unsigned long long vec_bperm (vector unsigned long long,
29022                                              vector unsigned char)
29023         vector signed long long vec_mule (vector signed int,
29024                                           vector signed int)
29025         vector unsigned long long vec_mule (vector unsigned int,
29026                                             vector unsigned int)
29027         vector signed long long vec_mulo (vector signed int,
29028                                           vector signed int)
29029         vector unsigned long long vec_mulo (vector unsigned int,
29030                                             vector unsigned int)
29031         * doc/extend.texi: Fix the documentation for the built-in
29032         functions.
29034 2017-06-07  Carl Love  <cel@us.ibm.com>
29036         PR target/80982
29037         * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
29038         for BE.
29040 2017-06-07  Carl Love  <cel@us.ibm.com>
29042         * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
29043         support, Generate       doublehv for signed int/float for BE case only.
29045 2017-06-07  Alexander Monakov  <amonakov@ispras.ru>
29047         * doc/invoke.texi (mcx16): Rewrite.
29049 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29051         * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
29052         * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
29053         *mov<mode>_softfloat, and an anonymous splitter): Use
29054         nonimmediate_operand instead of rs6000_nonimmediate_operand.
29056 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29058         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
29059         SPEFSCR registers.
29060         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29061         (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
29062         (rs6000_debug_reg_global): Adjust.
29063         (rs6000_init_hard_regno_mode_ok): Adjust.
29064         (rs6000_dbx_register_number): Adjust.
29065         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
29066         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29067         Remove SPE_ACC and SPEFSCR.
29068         (REG_ALLOC_ORDER): Ditto.
29069         (FRAME_POINTER_REGNUM): Change to 111.
29070         (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
29071         (REG_CLASS_NAMES): Ditto.
29072         (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
29073         (REGISTER_NAMES): Ditto.
29074         (ADDITIONAL_REG_NAMES): Ditto.
29075         (rs6000_reg_names): Ditto.
29076         * config/rs6000/rs6000.md: Renumber some register number
29077         define_constants.
29079 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29081         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
29082         registers.
29083         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29084         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
29085         to 117.
29086         (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
29087         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29088         Delete the SPE high registers.
29089         (REG_ALLOC_ORDER): Ditto.
29090         (enum reg_class): Remove SPE_HIGH_REGS.
29091         (REG_CLASS_NAMES): Ditto.
29092         (REG_CLASS_CONTENTS): Delete the SPE high registers.
29093         (REGISTER_NAMES): Ditto.
29094         (rs6000_reg_names): Ditto.
29095         * doc/tm.texi.in: Remove SPE as example.
29096         * doc/tm.texi: Regenerate.
29098 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29100         * config/rs6000/8540.md (ppc8540_brinc): Delete.
29101         * config/rs6000/e500mc.md (e500mc_brinc): Delete.
29102         * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
29103         * config/rs6000/rs6000.md (type): Remove "brinc".
29105 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29107         * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
29108         (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
29109         * config/rs6000/linuxspe.h: Delete file.
29110         * config/rs6000/rs6000.md: Don't include spe.md.
29111         * config/rs6000/spe.h: Delete file.
29112         * config/rs6000/spe.md: Delete file.
29113         * config/rs6000/t-rs6000: Remove spe.md.
29115 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29117         * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
29118         (reg_or_none500mem_operand): Delete.
29119         * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
29120         instead of reg_or_none500mem_operand.
29122 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29124         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29125         handling of SPE flags.
29126         * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
29128 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29130         * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
29131         SPE ABI handling.
29132         * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
29133         (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
29134         paired_divv2sf3): Similar.
29135         * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
29136         SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
29137         * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
29138         RS6000_BUILTIN_S.
29139         Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
29140         Rename the paired_* instruction patterns.
29141         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
29142         define __SPE__.
29143         * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
29144         * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
29145         (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
29146         spe_padding_size, and spe_64bit_regs_used.  Replace TARGET_SPE and
29147         TARGET_SPE_ABI with 0, simplify.  Replace SPE_VECTOR_MODE with
29148         PAIRED_VECTOR_MODE.
29149         (struct machine_function): Delete field spe_insn_chain_scanned_p.
29150         (spe_func_has_64bit_regs_p): Delete.
29151         (spe_expand_predicate_builtin): Delete.
29152         (spe_expand_evsel_builtin): Delete.
29153         (TARGET_DWARF_REGISTER_SPAN): Do not define.
29154         (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
29155         (invalid_e500_subreg): Delete.
29156         (rs6000_legitimize_address): Always force_reg op2 as well, for
29157         paired single memory accesses.
29158         (rs6000_member_type_forces_blk): Delete.
29159         (rs6000_spe_function_arg): Delete.
29160         (rs6000_expand_unop_builtin): Delete SPE handling.
29161         (rs6000_expand_binop_builtin): Ditto.
29162         (spe_expand_stv_builtin): Delete.
29163         (bdesc_2arg_spe): Delete.
29164         (spe_expand_builtin): Delete.
29165         (spe_expand_predicate_builtin): Delete.
29166         (spe_expand_evsel_builtin): Delete.
29167         (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
29168         (spe_init_builtins): Delete.
29169         (spe_func_has_64bit_regs_p): Delete.
29170         (savres_routine_name): Delete "info" parameter.  Adjust callers.
29171         (rs6000_emit_stack_reset): Ditto.
29172         (rs6000_dwarf_register_span): Delete.
29173         * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
29174         UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
29175         SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
29176         Delete.
29177         * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
29178         Delete.
29179         * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
29180         * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
29181         * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
29182         mulv2sf3, divv2sf3): Delete expanders.
29184 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29186         config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
29188 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29190         * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
29191         * config/rs6000/rs6000.c: Ditto.
29193 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29195         * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
29196         * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
29197         comparison_operator.
29199 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29201         * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
29202         * config/rs6000/rs6000.opt: Ditto.
29203         * config/rs6000/t-rtems: Ditto.
29205 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29207         * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
29208         TARGET_E500_SINGLE by 0, simplify.
29209         * config/rs6000/rs6000.c: Ditto.
29210         (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
29211         (spe_build_register_parallel): Delete.
29212         * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
29213         TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
29214         * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
29215         TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
29216         (E500_CONVERT): Delete.
29217         * config/rs6000/spe.md: Remove many patterns and all define_constants.
29219 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29221         * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
29222         * config/rs6000/dfp.md: Ditto.
29223         (negdd2, *negdd2_fpr): Merge.
29224         (absdd2, *absdd2_fpr): Merge.
29225         (negtd2, *negtd2_fpr): Merge.
29226         (abstd2, *abstd2_fpr): Merge.
29227         * config/rs6000/e500.h: Delete file.
29228         * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29229         TARGET_FPRS by 1 and simplify.
29230         * config/rs6000/rs6000-c.c: Ditto.
29231         * config/rs6000/rs6000.c: Ditto.  Also replace TARGET_SF_SPE and
29232         TARGET_DF_SPE by 0.
29233         * config/rs6000/rs6000.h: Ditto.  Delete TARGET_SF_SPE and
29234         TARGET_DF_SPE.
29235         * config/rs6000/rs6000.md: Ditto.
29236         (floatdidf2, *floatdidf2_fpr): Merge.
29237         (move_from_CR_gt_bit): Delete.
29238         * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29239         (E500_CR_IOR_COMPARE): Delete.
29240         (All patterns that require !TARGET_FPRS): Delete.
29241         * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29243 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29245         * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29247 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29249         * graphds.c (add_edge): Intitialize edge's attached data.
29250         (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29251         pointer parameter.  Call pointed function on each edge during
29252         graph traversing.  Skip traversing the edge when the function
29253         returns true.
29254         (graphds_dfs, graphds_scc): Ditto.
29255         (for_each_edge): New parameter.  Pass the new parameter to callback
29256         function.
29257         * graphds.h (skip_edge_callback): New function pointer type.
29258         (graphds_dfs, graphds_scc): New function pointer parameter.
29259         (graphds_edge_callback, for_each_edge): New parameter.
29261 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29263         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29264         out code checking if runtime alias check is possible to below ...
29265         Call the new function.
29266         * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29267         * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29269 2017-06-07  Marek Polacek  <polacek@redhat.com>
29271         PR sanitizer/80932
29272         * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29273         TYPE_OVERFLOW_WRAPS check.
29275 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29277         * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29278         if versioning is required.
29279         * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29280         peeling with the check for versioning.
29282 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29284         * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29285         * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29286         Set true to new parameter if new ssa variable is defined.
29287         (vect_gen_vector_loop_niters): Refactor.  Set range information
29288         for the new vector loop bound variable.
29289         (vect_do_peeling): Ditto.
29291 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29293         * tree-affine.c (ssa.h): Include header file.
29294         (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29295         has wrapping overflow behavior.
29297 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29299         * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29301 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29303         (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29304         (tree_to_aff_combination): ... here.
29306 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29308         * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29309         reg_pressure model function.
29310         (ivopts_global_cost_for_size): Delete.
29311         (determine_set_costs, iv_ca_recount_cost): Call new model function
29312         ivopts_estimate_reg_pressure.
29314 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29316         * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29317         expensive than udiv.  Remove floating point cases from mod.
29319 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29321         * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29322         Increase idiv cost.
29324 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29326         * config/aarch64/aarch64.md
29327         (copysignsf3): Fix mask generation.
29329 2017-06-07  Jakub Jelinek  <jakub@redhat.com>
29331         * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29332         TDI_gimple.
29333         (class dump_manager): Add register_dumps method.
29334         * dumpfile.c: Include langhooks.h.
29335         (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29336         (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29337         (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29338         (dump_manager::dump_register): Start with 512 entries instead of 32.
29339         (dump_manager::register_dumps): New method.
29340         * toplev.c (general_init): Instead of invoking register_dumps
29341         langhook, invoke register_dumps method on the dump manager.
29342         * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29343         TDI_generic.
29345 2017-06-07  Richard Sandiford  <richard.sandiford@linaro.org>
29347         * doc/md.texi: Clarify the restrictions on a define_insn condition.
29348         Say that # requires an associated define_split to exist, and that
29349         the define_split must be suitable for use after register allocation.
29351 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29353         * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29354         (compute_outgoing_frequencies): Also initialize zero counts.
29355         (find_many_sub_basic_blocks): Do not produce uninitialized profile
29356         around loops; preserve more of profile when nothing changes.
29358 2017-06-06  Jim Wilson  <jim.wilson@linaro.org>
29360         * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29361         here.
29362         * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29363         * config/arm/arm-cpu-cdata.h: Regenerate.
29364         * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29365         * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29366         * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29367         * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29368         * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29369         support.
29370         * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29371         support.
29372         * config/arm/t-rmprofile: Likewise.
29373         * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29375 2017-06-06  David S. Miller  <davem@davemloft.net>
29377         PR target/80968
29378         * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29379         blockage if function uses alloca.
29381 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29383         * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29384         New "uid" fields to hold pretty-print IDs of group and ref.
29385         Memory references are now identified as <group_id>:<ref_id>
29386         instead of using [random] addresses.
29387         (dump_mem_details): Simplify, no functional change.
29388         (dump_mem_ref): Simplify and make output more concise.
29389         Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29390         (find_or_create_group): Initialize group uid.
29391         (record_ref): Initialize ref uid.  Improve debug output.
29392         (prune_group_by_reuse, should_issue_prefetch_p,)
29393         (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29394         (mark_nontemporal_store, determine_loop_nest_reuse):
29395         Improve debug output.
29397 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29399         * dbgcnt.def (prefetch): New debug counter.
29400         * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29401         (schedule_prefetches): Stop issueing prefetches if debug counter
29402         tripped.
29404 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29406         * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29407         gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29409 2017-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29411         * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29412         Use aarch64_reg_or_zero predicate for operand 4.
29413         (aarch64_compare_and_swap<mode> define_insn_and_split):
29414         Use aarch64_reg_or_zero predicate for operand 3.  Add 'Z' constraint.
29415         (aarch64_store_exclusive<mode>): Likewise for operand 2.
29417 2017-06-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29419         * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29420         (arm_compute_save_core_reg_mask): This.
29421         (thumb1_compute_save_reg_mask): Rename into ...
29422         (thumb1_compute_save_core_reg_mask): This.
29423         (arm_compute_save_reg0_reg12_mask): Adapt comment.
29424         (arm_compute_frame_layout): Likewise.
29426 2017-06-06  Richard Biener  <rguenther@suse.de>
29428         PR tree-optimization/80974
29429         * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29430         keep or clear leaders SSA info.
29432 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29434         * config/nvptx/nvptx.c (split_mode_p): New function.
29435         (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29437 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29439         * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29441 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29443         PR bootstrap/80978
29444         * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29445         profile.
29447 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29449         * shrink-wrap.c (handle_simple_exit): Update profile.
29450         (try_shrink_wrapping): Upate profile.
29452 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29454         * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29455         (tree_guess_outgoing_edge_probabilities): New.
29456         * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29457         * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29459 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29461         * ipa-split.c (split_function): Initialize return bb profile.
29463 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29465         * profile.c (compute_branch_probabilities): Also initialize
29466         EXIT_BLOCK profile.
29468 2017-06-06  Richard Biener  <rguenther@suse.de>
29470         PR tree-optimization/80928
29471         * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29472         (vect_analyze_loop_operations): Properly guard analysis for
29473         pure SLP case.
29474         (vect_transform_loop): Likewise.
29475         (vect_analyze_loop_2): Also reset SLP type on PHIs.
29476         (vect_model_induction_cost): Do not cost for pure SLP.
29477         (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29478         of induction in inner loop vectorization.
29479         * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29480         (vect_get_and_check_slp_defs): Handle vect_induction_def.
29481         (vect_build_slp_tree): Likewise.  Handle PHIs as terminating the
29482         recursion.
29483         (vect_analyze_slp_cost_1): Cost induction.
29484         (vect_detect_hybrid_slp_stmts): Handle PHIs.
29485         (vect_get_slp_vect_defs): Likewise.
29486         * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29487         (vect_transform_stmt): Handle SLP reductions.
29488         * tree-vectorizer.h (vectorizable_induction): Adjust.
29490 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29492         * config/rs6000/rs6000.c (make_resolver_func): Update
29493         init_lowered_empty_function call.
29495 2017-06-05  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29497         * doc/invoke.texi: Document the -fprofile-abs-path option.
29498         * common.opt (fprofile-abs-path): New option.
29499         * gcov-io.h (gcov_write_filename): Declare.
29500         * gcov-io.c (gcov_write_filename): New function.
29501         * coverage.c (coverage_begin_function): Use gcov_write_filename.
29502         * profile.c (output_location): Likewise.
29504 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29506         * shring-wrap.c: Revert accidental commit.
29508 2017-06-05  Volker Reichelt  <v.reichelt@netcologne.de>
29510         * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29512 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29514         * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29515         new edge.
29516         * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29517         profile in callgraph edge.
29518         * profile-count.h (apply_probability): If THIS is 0, then result is 0
29519         (apply_scale): Likewise.
29520         * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29521         Also scale profile when inlining function with zero profile.
29522         (initialize_cfun): Update exit block profile even when it is zero.
29523         * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29524         when profile is read.
29526 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29528         * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29529         (CLONE_*): New constants to define the processors we can generate
29530         code for with the target_clone attribute.
29531         (rs6000_clone_map): New array to identify which clone processors
29532         the current program is running on.
29533         (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29534         target_clone attribute.
29535         (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29536         (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29537         (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29538         (cpu_expand_builtin): Add support for target_clone attribute.
29539         (rs6000_valid_attribute_p): Allow "default" attribute.
29540         (get_decl_name): New debug function to simplify printing the
29541         current function name in debugging statements.
29542         (rs6000_clone_priority): New functions to support the target_clone
29543         attribute, and be able to generate code to switch between ISA 2.05
29544         through ISA 3.0 (power6 through power9).
29545         (rs6000_compare_version_priority): Likewise.
29546         (rs6000_get_function_versions_dispatcher): Likewise.
29547         (make_resolver_func): Likewise.
29548         (add_condition_to_bb): Likewise.
29549         (dispatch_function_versions): Likewise.
29550         (rs6000_generate_version_dispatcher_body): Likewise.
29551         (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29552         (fusion_gpr_load_p): Fix a spacing issue.
29553         * doc/extend.texi (Common Function Attributes): Document that the
29554         PowerPC supports the target_clone attribute.
29556 2017-06-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29558         * config/arm/arm.h: explain F symbol found in description of ARM
29559         register allocation in its legend.
29561 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29563         * config/mips/frame-header-opt.c: Include profile-count.h.
29564         * config/riscv/riscv.c: Include profile-count.h
29566 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29568         * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29569         update profile.
29570         (sm_set_flag_if_changed): Add bbs field.
29571         (execute_sm_if_changed_flag_set): Pass BBS.
29572         (execute_sm): Update.
29574 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29576         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29577         New pattern.
29579 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29581         * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29582         (peephole2): New peephole2 to emit the above.
29583         * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29585 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29587         * config/aarch64/aarch64.c (define_peephole2 above
29588         *sub_<shift>_<mode>): New peephole.
29590 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
29592         * config/i386/i386.c (make_resolver_func): Update.
29593         * Makefile.in: Add profile-count.h and profile-count.o
29594         * auto-profile.c (afdo_indirect_call): Update to new API.
29595         (afdo_set_bb_count): Update.
29596         (afdo_propagate_edge): Update.
29597         (afdo_propagate_circuit): Update.
29598         (afdo_calculate_branch_prob): Update.
29599         (afdo_annotate_cfg): Update.
29600         * basic-block.h: Include profile-count.h
29601         (struct edge_def): Turn count to profile_count.
29602         (struct basic_block_def): Likewie.
29603         (REG_BR_PROB_BASE): Move to profile-count.h
29604         (RDIV): Move to profile-count.h
29605         * bb-reorder.c (max_entry_count): Turn to profile_count.
29606         (find_traces): Update.
29607         (rotate_loop):Update.
29608         (connect_traces):Update.
29609         (sanitize_hot_paths):Update.
29610         * bt-load.c (migrate_btr_defs): Update.
29611         * cfg.c (RDIV): Remove.
29612         (init_flow): Use alloc_block.
29613         (alloc_block): Uninitialize count.
29614         (unchecked_make_edge): Uninitialize count.
29615         (check_bb_profile): Update.
29616         (dump_edge_info): Update.
29617         (dump_bb_info): Update.
29618         (update_bb_profile_for_threading): Update.
29619         (scale_bbs_frequencies_int): Update.
29620         (scale_bbs_frequencies_gcov_type): Update.
29621         (scale_bbs_frequencies_profile_count): New.
29622         * cfg.h (update_bb_profile_for_threading): Update.
29623         (scale_bbs_frequencies_profile_count): Declare.
29624         * cfgbuild.c (compute_outgoing_frequencies): Update.
29625         (find_many_sub_basic_blocks): Update.
29626         * cfgcleanup.c (try_forward_edges): Update.
29627         (try_crossjump_to_edge): Update.
29628         * cfgexpand.c (expand_gimple_tailcall): Update.
29629         (construct_exit_block): Update.
29630         * cfghooks.c (verify_flow_info): Update.
29631         (dump_bb_for_graph): Update.
29632         (split_edge): Update.
29633         (make_forwarder_block): Update.
29634         (duplicate_block): Update.
29635         (account_profile_record): Update.
29636         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29637         (get_estimated_loop_iterations): Update.
29638         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29639         (single_likely_exit): Update.
29640         * cfgloopmanip.c (scale_loop_profile): Update.
29641         (loopify): Update.
29642         (set_zero_probability): Update.
29643         (lv_adjust_loop_entry_edge): Update.
29644         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29645         (purge_dead_edges): Update.
29646         (rtl_account_profile_record): Update.
29647         * cgraph.c (cgraph_node::create): Uninitialize count.
29648         (symbol_table::create_edge): Uninitialize count.
29649         (cgraph_update_edges_for_call_stmt_node): Update.
29650         (cgraph_edge::dump_edge_flags): Update.
29651         (cgraph_node::dump): Update.
29652         (cgraph_edge::maybe_hot_p): Update.
29653         * cgraph.h: Include profile-count.h
29654         (create_clone), create_edge, create_indirect_edge): Update.
29655         (cgraph_node): Turn count to profile_count.
29656         (cgraph_edge0: Likewise.
29657         (make_speculative, clone): Update.
29658         (create_edge): Update.
29659         (init_lowered_empty_function): Update.
29660         * cgraphclones.c (cgraph_edge::clone): Update.
29661         (duplicate_thunk_for_node): Update.
29662         (cgraph_node::create_clone): Update.
29663         * cgraphunit.c (cgraph_node::analyze): Update.
29664         (cgraph_node::expand_thunk): Update.
29665         * final.c (dump_basic_block_info): Update.
29666         * gimple-streamer-in.c (input_bb): Update.
29667         * gimple-streamer-out.c (output_bb): Update.
29668         * graphite.c (print_global_statistics): Update.
29669         (print_graphite_scop_statistics): Update.
29670         * hsa-brig.c: Include basic-block.h.
29671         * hsa-dump.c: Include basic-block.h.
29672         * hsa-gen.c (T sum_slice): Update.
29673         (convert_switch_statements):Update.
29674         * hsa-regalloc.c: Include basic-block.h.
29675         * ipa-chkp.c (chkp_produce_thunks): Update.
29676         * ipa-cp.c (struct caller_statistics): Update.
29677         (init_caller_stats): Update.
29678         (gather_caller_stats): Update.
29679         (ipcp_cloning_candidate_p): Update.
29680         (good_cloning_opportunity_p): Update.
29681         (get_info_about_necessary_edges): Update.
29682         (dump_profile_updates): Update.
29683         (update_profiling_info): Update.
29684         (update_specialized_profile): Update.
29685         (perhaps_add_new_callers): Update.
29686         (decide_about_value): Update.
29687         (ipa_cp_c_finalize): Update.
29688         * ipa-devirt.c (struct odr_type_warn_count): Update.
29689         (struct decl_warn_count): Update.
29690         (struct final_warning_record): Update.
29691         (possible_polymorphic_call_targets): Update.
29692         (ipa_devirt): Update.
29693         * ipa-fnsummary.c (redirect_to_unreachable): Update.
29694         * ipa-icf.c (sem_function::merge): Update.
29695         * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29696         * ipa-inline.c (compute_uninlined_call_time): Update.
29697         (compute_inlined_call_time): Update.
29698         (want_inline_small_function_p): Update.
29699         (want_inline_self_recursive_call_p): Update.
29700         (edge_badness): Update.
29701         (lookup_recursive_calls): Update.
29702         (recursive_inlining): Update.
29703         (inline_small_functions): Update.
29704         (dump_overall_stats): Update.
29705         (dump_inline_stats): Update.
29706         * ipa-profile.c (ipa_profile_generate_summary): Update.
29707         (ipa_propagate_frequency): Update.
29708         (ipa_profile): Update.
29709         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29710         * ipa-utils.c (ipa_merge_profiles): Update.
29711         * loop-doloop.c (doloop_modify): Update.
29712         * loop-unroll.c (report_unroll): Update.
29713         (unroll_loop_runtime_iterations): Update.
29714         * lto-cgraph.c (lto_output_edge): Update.
29715         (lto_output_node): Update.
29716         (input_node): Update.
29717         (input_edge): Update.
29718         (merge_profile_summaries): Update.
29719         * lto-streamer-in.c (input_cfg): Update.
29720         * lto-streamer-out.c (output_cfg): Update.
29721         * mcf.c (create_fixup_graph): Update.
29722         (adjust_cfg_counts): Update.
29723         (sum_edge_counts): Update.
29724         * modulo-sched.c (sms_schedule): Update.
29725         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29726         * predict.c (maybe_hot_count_p): Update.
29727         (probably_never_executed): Update.
29728         (dump_prediction): Update.
29729         (combine_predictions_for_bb): Update.
29730         (propagate_freq): Update.
29731         (handle_missing_profiles): Update.
29732         (counts_to_freqs): Update.
29733         (rebuild_frequencies): Update.
29734         (force_edge_cold): Update.
29735         * predict.h: Include profile-count.h
29736         (maybe_hot_count_p, counts_to_freqs): UPdate.
29737         * print-rtl-function.c: Do not include cfg.h
29738         * print-rtl.c: Include basic-block.h
29739         * profile-count.c: New file.
29740         * profile-count.h: New file.
29741         * profile.c (is_edge_inconsistent): Update.
29742         (correct_negative_edge_counts): Update.
29743         (is_inconsistent): Update.
29744         (set_bb_counts): Update.
29745         (read_profile_edge_counts): Update.
29746         (compute_frequency_overlap): Update.
29747         (compute_branch_probabilities): Update; Initialize and deinitialize
29748         gcov_count tables.
29749         (branch_prob): Update.
29750         * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29751         (edge_gcov_count): New.
29752         (bb_gcov_count): New.
29753         * shrink-wrap.c (try_shrink_wrapping): Update.
29754         * tracer.c (better_p): Update.
29755         * trans-mem.c (expand_transaction): Update.
29756         (ipa_tm_insert_irr_call): Update.
29757         (ipa_tm_insert_gettmclone_call): Update.
29758         * tree-call-cdce.c: Update.
29759         * tree-cfg.c (gimple_duplicate_sese_region): Update.
29760         (gimple_duplicate_sese_tail): Update.
29761         (gimple_account_profile_record): Update.
29762         (execute_fixup_cfg): Update.
29763         * tree-inline.c (copy_bb): Update.
29764         (copy_edges_for_bb): Update.
29765         (initialize_cfun): Update.
29766         (freqs_to_counts): Update.
29767         (copy_cfg_body): Update.
29768         (expand_call_inline): Update.
29769         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29770         * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29771         (try_unroll_loop_completely): Update.
29772         (try_peel_loop): Update.
29773         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29774         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29775         * tree-ssa-loop-split.c (connect_loops): Update.
29776         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29777         * tree-ssa-reassoc.c (branch_fixup): Update.
29778         * tree-ssa-tail-merge.c (replace_block_by): Update.
29779         * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29780         (compute_path_counts): Update.
29781         (update_profile): Update.
29782         (recompute_probabilities): Update.
29783         (update_joiner_offpath_counts): Update.
29784         (estimated_freqs_path): Update.
29785         (freqs_to_counts_path): Update.
29786         (clear_counts_path): Update.
29787         (ssa_fix_duplicate_block_edges): Update.
29788         (duplicate_thread_path): Update.
29789         * tree-switch-conversion.c (case_bit_test_cmp): Update.
29790         (struct switch_conv_info): Update.
29791         * tree-tailcall.c (decrease_profile): Update.
29792         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29793         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29794         * value-prof.c (check_counter): Update.
29795         (gimple_divmod_fixed_value): Update.
29796         (gimple_mod_pow2): Update.
29797         (gimple_mod_subtract): Update.
29798         (gimple_ic_transform): Update.
29799         (gimple_stringop_fixed_value): Update.
29800         * value-prof.h (gimple_ic): Update.
29802 2017-06-02  Carl Love  <cel@us.ibm.com>
29804         * config/rs6000/rs6000-c: Add support for built-in functions
29805         vector double vec_doublee (vector signed int);
29806         vector double vec_doublee (vector unsigned int);
29807         vector double vec_doublee (vector float);
29808         vector double vec_doubleh (vector signed int);
29809         vector double vec_doubleh (vector unsigned int);
29810         vector double vec_doubleh (vector float);
29811         vector double vec_doublel (vector signed int);
29812         vector double vec_doublel (vector unsigned int);
29813         vector double vec_doublel (vector float);
29814         vector double vec_doubleo (vector signed int);
29815         vector double vec_doubleo (vector unsigned int);
29816         vector double vec_doubleo (vector float);.
29817         * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29818         DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29819         UNS_DOUBLEL.
29820         * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29821         unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29822         unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29823         VS_sxwsp.
29824         * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29825         vec_doublel, vec_doubleh.
29826         * doc/extend.texi: Update the built-in documentation file for the
29827         new built-in functions.
29829 2017-06-02  David Malcolm  <dmalcolm@redhat.com>
29831         PR jit/80954
29832         * ipa-inline-analysis.c (free_growth_caches): Set
29833         edge_removal_hook_holder to NULL after removing it.
29835 2017-06-02  Sudakshina Das  <sudi.das@arm.com>
29837         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29838         comparision with zero.
29840 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29841         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29842         for early expansion of vec_min and vec_max builtins.
29843         (builtin_function_type): Add min/max unsigned variants to those
29844         identified as having unsigned arguments.
29846 2017-06-02  Olivier Hainque  <hainque@adacore.com>
29848         * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29850 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29852         * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29853         Use VALL_F16 iterator rather than VALL.
29855 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29857         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29858         Emit CBNZ inside loop when doing a strong exchange and comparing
29859         against zero.  Generate the CC flags after the loop.
29861 2017-06-02  David Edelsohn  <dje.gcc@gmail.com>
29863         * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29864         (dl_section_ref): New.
29865         (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29866         On AIX, append an expression to subtract the size of the
29867         section length to dl_section_ref.
29869 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29871         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29872         for early expansion of vector absolute builtins.
29874 2017-06-02  Richard Biener  <rguenther@suse.de>
29876         * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29877         what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29879 2017-06-02  Richard Biener  <rguenther@suse.de>
29881         PR tree-optimization/80948
29882         * tree-tailcall.c (find_tail_calls): Track stmts to move in
29883         stmt order as well.
29885 2017-06-02  Richard Biener  <rguenther@suse.de>
29887         * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29888         PHIs are ok.
29889         * tree-vect-stmts.c (process_use): Do not mark backedge defs
29890         for inductions as relevant.
29892 2017-06-02  Richard Biener  <rguenther@suse.de>
29894         * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29895         (vectorizable_induction): ... this.  Remove dead code.
29897 2017-06-02  Eric Botcazou  <ebotcazou@adacore.com>
29899         * builtins. (expand_builtin_alloca): Remove second parameter and
29900         infer its value from the first parameter instead.
29901         (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
29903 2017-06-02  Jakub Jelinek  <jakub@redhat.com>
29905         PR rtl-optimization/80903
29906         * loop-doloop.c (add_test): Unshare sequence.
29908 2017-06-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29910         * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
29912 2017-06-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29914         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
29915         static.
29916         (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
29917         xlogue_layout::get_instance, logue_layout::xlogue_layout,
29918         sp_valid_at, fp_valid_at, choose_basereg): Formatting.
29919         (xlogue_layout::get_stub_rtx): Make static.
29920         (xlogue_layout::get_stub_name): Avoid const-cast, make static.
29921         (xlogue_layout::compute_stub_managed_regs): Rename to...
29922         (xlogue_layout::count_stub_managed_regs): ...this.
29923         (xlogue_layout::is_stub_managed_reg): New function.
29924         (xlogue_layout::m_stub_names): Rename to...
29925         (xlogue_layout::s_stub_names): ...this, make static.
29926         (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
29927         xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
29928         xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
29929         xlogue_layout::s_stub_names): Instantiate statics.
29930         (stub_managed_regs): Remove.
29931         (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
29932         (disable_call_ms2sysv_xlogues): Rename to...
29933         (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
29934         (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
29935         warning logic.
29936         (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
29937         change after reload_completed.
29938         (ix86_can_use_return_insn_p): Use the ix86_frame data structure
29939         directly.
29940         (ix86_expand_prologue): Likewise.
29941         (ix86_expand_epilogue): Likewise.
29942         (ix86_expand_split_stack_prologue): Likewise.
29943         (ix86_compute_frame_layout): Remove frame parameter ...
29944         (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
29945         (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
29946         only if necessary.
29947         (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
29948         (ix86_frame): Move from here ...
29949         * config/i386/i386.h (ix86_frame): ... to here.
29950         (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
29951         complete ix86_frame data structure instead.  Remove some_ld_name.
29953 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29955         * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
29956         symbols that hold a DECL_VALUE_EXPR.
29958 2017-06-01  Martin Jambor  <mjambor@suse.cz>
29960         PR tree-optimization/80898
29961         * tree-sra.c (process_subtree_disqualification): Removed.
29962         (disqualify_candidate): Do not acll
29963         process_subtree_disqualification.
29964         (subtree_mark_written_and_enqueue): New function.
29965         (propagate_all_subaccesses): Set grp_write of LHS subtree if the
29966         RHS has been disqualified and re-queue LHS if necessary.  Apart
29967         from that, ignore disqualified RHS.
29969 2017-06-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29971         * config/s390/s390.c (s390_emit_epilogue): Disable early return
29972         address fetch for z10 or later.
29974 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29976         * config/arc/arc.md (tst_movb): Add guard when splitting.
29978 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29980         * config/arc/arc.c (arc_can_eliminate): Test against
29981         arc_frame_pointer_needed.
29983 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29985         * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
29986         to prevent store reordering.
29987         * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
29988         (type): Add block type.
29989         (stack_tie): Define special instruction to be used in
29990         expand_prologue.
29992 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29994         * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
29995         constraint. It is not valid for the pattern.
29996         (noncommutative_binary_comparison): Likewise.
29998 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30000         * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
30001         scaled addresses.
30003 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30005         * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
30006         be used by the reg-alloc.
30008 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30010         * config/arc/arc.md (mulsi3): Avoid use of hard registers before
30011         reg-alloc when having mul64 or mul32x16 instructions.
30012         (mulsidi3): Likewise.
30013         (umulsidi3): Likewise.
30014         (mulsi32x16): New pattern.
30015         (mulsi64): Likewise.
30016         (mulsidi64): Likewise.
30017         (umulsidi64): Likewise.
30018         (MUL32x16_REG): Define.
30019         (mul64_600): Use MUL32x16_REG.
30020         (mac64_600): Likewise.
30021         (umul64_600): Likewise.
30022         (umac64_600): Likewise.
30024 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
30026         * config/arc/arc.md (mulsi3_700): Make it commutative.
30028 2017-06-01  Jose E. Marchesi  <jose.marchesi@oracle.com>
30030         * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
30031         type for movstouw.
30032         (*sign_extendsidi2_insn): Likewise for movstosw.
30034 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
30036         * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
30037         the type of the input discriminant value.  Convert the
30038         discriminant value of signedness vary.
30040 2017-06-01  Volker Reichelt  <v.reichelt@netcologne.de>
30042         * doc/invoke.texi (-Wcatch-value): Document new shortcut.
30043         Add to -Wall section.
30045 2017-06-01  Richard Biener  <rguenther@suse.de>
30047         PR middle-end/66313
30048         * fold-const.c (fold_plusminus_mult_expr): If the factored
30049         factor may be zero use a wrapping type for the inner operation.
30050         * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
30051         and handle moved defs.
30052         (process_assignment): Properly guard the unary op case.  Return a
30053         tri-state indicating that moving the stmt before the call may allow
30054         to continue.  Pass through to_move.
30055         (find_tail_calls): Handle moving unrelated defs before
30056         the call.
30058 2017-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
30060         PR target/80618
30061         * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
30062         splitter result in the canonical way.
30064 2017-05-31  Uros Bizjak  <ubizjak@gmail.com>
30066         * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
30067         also for 32bit target.  Update insn attributes.
30068         (zero-extendsidi2 splitter): Allow all registers for operand 1.
30070 2017-05-31  Sebastian Peryt  <sebastian.peryt@intel.com>
30072         * config/i386/avx512fintrin.h (_mm_mask_max_sd)
30073         (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
30074         (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
30075         (_mm_maskz_min_ss): New intrinsics.
30077 2017-05-31  Martin Liska  <mliska@suse.cz>
30079         * tree-vect-loop.c (vect_create_epilog_for_reduction):
30080         Change comment style to one we normally use.
30081         (vectorizable_reduction): Likewise.
30082         (vectorizable_induction): Likewise.
30083         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
30084         (vectorizable_call): Likewise.
30085         (vectorizable_simd_clone_call): Likewise.
30086         (vectorizable_conversion): Likewise.
30087         (vectorizable_assignment): Likewise.
30088         (vectorizable_shift): Likewise.
30089         (vectorizable_operation): Likewise.
30090         (vectorizable_store): Likewise.
30091         (vectorizable_load): Likewise.
30092         * tree-vectorizer.h: Likewise.
30094 2017-05-31  Alexander Monakov  <amonakov@ispras.ru>
30096         * passes.c (emergency_dump_function): New.
30097         * tree-pass.h (emergency_dump_function): Declare.
30098         * plugin.c (plugins_internal_error_function): Remove.
30099         * plugin.h (plugins_internal_error_function): Remove declaration.
30100         * toplev.c (internal_error_function): New static function.  Use it...
30101         (general_init): ...here.
30103 2017-05-31  Graham Markall  <graham.markall@embecosm.com>
30105         * config/arc/arc.c (arc_print_operand): Handle constant operands.
30106         (arc_rtx_costs): Add costs for new patterns.
30107         * config/arc/arc.md: Additional *add_n and *sub_n patterns.
30108         * config/arc/predicates.md: Add _1_2_3_operand predicate.
30110 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30112         * tree-ssa-strlen.c (get_next_strinfo): New function.
30113         (get_stridx_plus_constant): Use it.
30114         (zero_length_string): Likewise.
30115         (adjust_related_strinfos): Likewise.
30116         (adjust_last_stmt): Likewise.
30118 2017-05-31  Richard Biener  <rguenther@suse.de>
30120         PR target/80880
30121         * config/i386/i386.c (ix86_expand_builtin): Remove assert
30122         for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
30124 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30126         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
30127         loop_vinfo argument and use of dependence distance vectors.
30128         Check instead whether the two references differ only in their
30129         initial value and assume that they have the same alignment if the
30130         difference is a multiple of the vector alignment.
30131         (vect_analyze_data_refs_alignment): Update call accordingly.
30133 2017-05-31  Martin Liska  <mliska@suse.cz>
30135         PR target/79155
30136         * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
30138 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30140         * tree-vect-loop-manip.c (create_intersect_range_checks_index)
30141         (create_intersect_range_checks): Move from ...
30142         * tree-data-ref.c (create_intersect_range_checks_index)
30143         (create_intersect_range_checks): ... to here.
30144         (create_runtime_alias_checks): New function factored from ...
30145         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
30146         here.  Call above function.
30147         * tree-data-ref.h (create_runtime_alias_checks): New function.
30149 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30151         * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
30152         segment length for dr_b and compute it in wide_int.
30154 2017-05-31  Richard Biener  <rguenther@suse.de>
30156         PR tree-optimization/80906
30157         * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
30158         and pass through iv_map.
30159         (copy_bb_and_scalar_dependences): Adjust.
30160         (translate_pending_phi_nodes): Likewise.
30161         (copy_loop_close_phi_args): Handle code-generating IVs instead
30162         of ICEing.
30164 2017-05-30  David Malcolm  <dmalcolm@redhat.com>
30166         * diagnostic-color.c (color_dict): Add "type-diff".
30167         (parse_gcc_colors): Update comment.
30168         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
30169         -fdiagnostics-show-template-tree and -fno-elide-type.
30170         (GCC_COLORS): Add type-diff to example.
30171         (type-diff=): New.
30172         (-fdiagnostics-show-template-tree): New.
30173         (-fno-elide-type): New.
30174         * pretty-print.c (pp_format): Pass quote and formatters[argno] to
30175         the pp_format_decoder callback.  Call any m_format_postprocessor's
30176         "handle" method.
30177         (pretty_printer::pretty_printer): Initialize
30178         m_format_postprocessor.
30179         (pretty_printer::~pretty_printer): Delete any
30180         m_format_postprocessor.
30181         * pretty-print.h (printer_fn): Add bool and const char ** parameters.
30182         (class format_postprocessor): New class.
30183         (struct pretty_printer::format_decoder): Document the new parameters.
30184         (struct pretty_printer::m_format_postprocessor): New field.
30185         * tree-diagnostic.c (default_tree_printer): Update for new
30186         bool and const char ** params.
30187         * tree-diagnostic.h (default_tree_printer): Likewise.
30189 2017-05-30  Segher Boessenkool  <segher@kernel.crashing.org>
30191         * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
30192         (lwa_operand): Delete rs6000_gen_cell_microcode test.
30193         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30194         rs6000_gen_cell_microcode code.
30195         (rs6000_final_prescan_insn): Delete.
30196         (rs6000_opt_vars): Delete the "gen-cell-microcode" and
30197         "warn-cell-microcode" entries.
30198         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
30199         * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
30200         throughout.  Change cc_reg_not_micro_cr0_operand to
30201         cc_reg_not_cr0_operand throughout.
30202         (*extendhi<mode>2_noload): Delete.
30203         * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
30204         (mwarn-cell-microcode): Delete.
30205         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
30206         -mgen-cell-microcode and -mwarn-cell-microcode.
30208 2017-05-30  Uros Bizjak  <ubizjak@gmail.com>
30210         PR target/80833
30211         * config/i386/constraints.md (Yd): New constraint.
30212         (Ye): Ditto.
30213         * config/i386/i386.md (*movti_internal): Add (?r, Ye)
30214         and (?Yd, r) alternatives.  Update insn attributes.
30215         * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
30216         and (?*Yd, r) alternatives.  Update insn attributes.
30217         (double-mode inter-unit splitters): Add new GR<->XMM splitters.
30219 2017-05-30  Pierre-Marie de Rodat  <derodat@adacore.com>
30221         * gimplify.c (gimplify_modify_expr): Don't create a
30222         DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30223         function.
30225 2017-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
30227         * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30229 2017-05-30  Richard Biener  <rguenther@suse.de>
30231         * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30232         and reduc_def fields.
30233         (STMT_VINFO_REDUC_TYPE): New define.
30234         (STMT_VINFO_REDUC_DEF): Likewise.
30235         (vect_force_simple_reduction): Adjust prototype.
30236         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30237         (vect_is_simple_reduction): Remove check_reduction argument.
30238         (vect_force_simple_reduction): Adjust and set
30239         STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30240         (vectorizable_reduction): Do not re-do reduction analysis
30241         but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30242         * tree-parloops.c (gather_scalar_reductions): Adjust.
30244 2017-05-30  Richard Biener  <rguenther@suse.de>
30246         PR middle-end/80901
30247         * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30248         split_edge code.
30250 2017-05-24  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30252         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30253         Introduce unknown_misalignment parameter and remove vf.
30254         (vect_peeling_hash_get_lowest_cost):
30255         Pass unknown_misalignment parameter.
30256         (vect_enhance_data_refs_alignment):
30257         Fix unsupportable data ref treatment.
30259 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30261         * tree-vect-data-refs.c (vect_get_data_access_cost):
30262         Workaround for SLP handling.
30263         (vect_enhance_data_refs_alignment):
30264         Compute costs for doing no peeling at all, compare to the best
30265         peeling costs so far and avoid peeling if cheaper.
30267 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30269         * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30270         Return peeling info and set costs to zero for unlimited cost
30271         model.
30272         (vect_enhance_data_refs_alignment): Also inspect all datarefs
30273         with unknown misalignment. Compute and costs for unknown
30274         misalignment, compare them to the costs for known misalignment
30275         and choose the cheapest for peeling.
30277 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30279         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30280         (vect_get_peeling_costs_all_drs): Create function.
30281         (vect_peeling_hash_get_lowest_cost):
30282         Use vect_get_peeling_costs_all_drs.
30283         (vect_peeling_supportable): Create function.
30284         (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30286 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30288         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30289         DR_HAS_NEGATIVE_STEP.
30290         (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30291         (vect_enhance_data_refs_alignment): Use.
30292         (vect_duplicate_ssa_name_ptr_info): Use.
30293         * tree-vectorizer.h (dr_misalignment): Use.
30294         (known_alignment_for_access_p): Use.
30296 2017-05-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
30298         PR target/78838
30299         * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30300         .lowtext.
30301         (has_section_name): New function.
30303 2017-05-30  Martin Liska  <mliska@suse.cz>
30305         PR other/80909
30306         * auto-profile.c (get_function_decl_from_block): Fix
30307         parenthesis.
30309 2017-05-30  Richard Biener  <rguenther@suse.de>
30311         PR middle-end/80876
30312         * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30314 2017-05-30  Martin Liska  <mliska@suse.cz>
30316         * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30317         * dumpfile.h (struct dump_file_info): Remove ctors.
30319 2017-05-30  Martin Liska  <mliska@suse.cz>
30321         * predict.def: Fix GNU coding style.
30323 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30325         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30326         Mark 'to' argument with ATTRIBUTE_UNUSED.
30328 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30330         * config/xtensa/xtensa.c (xtensa_emit_call): Use
30331         HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30332         (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30333         format string.
30335 2017-05-29  Eric Botcazou  <ebotcazou@adacore.com>
30337         * doc/install.texi (Options specification): Restore entry of
30338         --enable-sjlj-exceptions.
30340 2017-05-27  Michael Eager  <eager@eagercon.com>
30342         Revert:
30343         2016-01-21  Ajit Agarwal  <ajitkum@xilinx.com>
30345         See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30347         * config/microblaze/microblaze.h
30348         (FIXED_REGISTERS): Update in macro.
30349         (CALL_USED_REGISTERS): Update in macro.
30351 2017-05-27  François-Xavier Coudett  <fxcoudert@gcc.gnu.org>
30353         * doc/install.texi: Add links to macOS binary distributions.
30355 2017-05-27  Jakub Jelinek  <jakub@redhat.com>
30357         PR bootstrap/80887
30358         Revert:
30359         2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30361         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30363 2017-05-26  Martin Liska  <mliska@suse.cz>
30365         * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30367 2017-05-26  Martin Liska  <mliska@suse.cz>
30369         * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30370         always leading ';; '.
30371         (dump_bb_info): Likewise.
30372         (brief_dump_cfg): Likewise.
30373         * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30374         * dumpfile.c: Remove usage of TDF_VERBOSE.
30375         * dumpfile.h (enum dump_kind): Likewise.
30376         (dump_gimple_bb_header): Do not use TDF_COMMENT.
30377         * print-tree.c (debug_verbose): Remove.
30378         * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30379         (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30380         * tree-diagnostic.c (default_tree_printer): Replace
30381         TDF_DIAGNOSTIC with TDF_SLIM.
30383 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30385         * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30386         in parameter loop, rather than loop_vinfo.
30387         (create_intersect_range_checks): Ditto.
30388         (vect_create_cond_for_alias_checks): Update call to above functions.
30390 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30392         PR tree-optimization/80815
30393         * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30394         for merging runtime alias checks.  Handle negative DR_STEPs.
30396 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30398         * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30399         Move from ...
30400         * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30401         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30402         out code pruning runtime alias checks.
30403         * tree-data-ref.c (prune_runtime_alias_test_list): New function
30404         factored out from above.
30405         * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30406         Move from ...
30407         * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30408         ... to here.
30409         (prune_runtime_alias_test_list): New decalaration.
30411 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30413         * tree-vect-data-refs.c (compare_tree): Rename and move ...
30414         * tree-data-ref.c (data_ref_compare_tree): ... to here.
30415         * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30416         * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30417         (operator==, comp_dr_with_seg_len_pair): Ditto.
30418         (vect_prune_runtime_alias_test_list): Ditto.
30420 2017-05-26  Martin Liska  <mliska@suse.cz>
30422         PR ipa/80663
30423         * params.def: Bound partial-inlining-entry-probability param.
30425 2017-05-26  Marek Polacek  <polacek@redhat.com>
30427         PR sanitizer/80875
30428         * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30429         can be negated.
30431 2017-05-26  Richard Biener  <rguenther@suse.de>
30433         PR tree-optimization/80842
30434         * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30435         value.
30437 2017-05-26  Richard Biener  <rguenther@suse.de>
30439         PR tree-optimization/80844
30440         * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30442 2017-05-25  Sebastian Peryt  <sebastian.peryt@intel.com>
30444         * doc/md.texi (Machine Constraints): Update x86 family
30445         machine constraints section to match 'config/i386/constraints.md'.
30447 2017-05-25  Volker Reichelt  <v.reichelt@netcologne.de>
30449         * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30451 2017-05-25  Nathan Sidwell  <nathan@acm.org>
30453         * doc/invoke.texi (--enable-languages): Update documentation.
30455 2017-05-25  Martin Liska  <mliska@suse.cz>
30457         * dumpfile.c: Add TDF_FOLDING.
30458         * dumpfile.h (enum dump_kind): Likewise.
30459         * genmatch.c (dt_simplify::gen_1): Use it.
30461 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30463         * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30465 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30467         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30468         * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30470 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30472         * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30473         * match.pd (X == C): Rewrite it here.
30474         (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30475         with_certain_nonzero_bits2): New predicates.
30476         * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30478 2017-05-24  Nathan Sidwell  <nathan@acm.org>
30480         * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30481         avoid warning.
30483         * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30484         warning.
30486 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30488         * config/powerpcspe: New port.  Files are copied from the rs6000
30489         port, with "rs6000" in filenames replaced by "powerpcspe".
30491 2017-05-24  Wilco Dijkstra  <wdijkstr@arm.com>
30493         PR rtl-optimization/80754
30494         * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30496 2017-05-24  Sheldon Lobo  <smlobo@sheldon.us.oracle.com>
30498         * config/sparc/sparc.md (length): Return the correct value for -mflat
30499         sibcalls to match output_sibcall.
30501 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30503         PR bootstrap/80860
30504         PR bootstrap/80843
30505         * config/rs6000/rs6000.c (struct machine_function): Add new field
30506         n_components.
30507         (rs6000_get_separate_components): Init that field, use it.
30508         (rs6000_components_for_bb): Use the field.
30510 2017-05-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30512         * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30514 2017-05-24  Peter Bergner  <bergner@vnet.ibm.com>
30516         PR middle-end/80823
30517         * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30519 2017-05-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30521         PR target/80725
30522         * config/s390/s390.c (s390_check_qrst_address): Check incoming
30523         address against address_operand predicate.
30524         * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30526 2017-05-24  Eric Botcazou  <ebotcazou@adacore.com>
30528         * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30529         parameters passed indirectly.
30531 2017-05-23  Uros Bizjak  <ubizjak@gmail.com>
30533         * config/i386/i386.md (*movdi_internal): Remove SSE4
30534         alternative 18 (?r, *v).  Update insn attributes.
30535         (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30536         Update insn attributes.
30537         (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30538         Update insn attributes.
30539         * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30540         alternative 1 (r, v). Remove isa attribute.
30541         * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30542         Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30543         and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30545 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30547         * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30548         dg-line directive.
30550 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30552         * cgraphunit.c (symbol_table::process_new_functions): Update.
30553         * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30554         (inline_generate_summary): Rename to ...
30555         (ipa_fn_summary_generate): ... this one.
30556         (inline_read_summary): Rename to ...
30557         (ipa_fn_summary_read): ... this one.
30558         (inline_write_summary): Rename to ...
30559         (ipa_fn_summary_write): ... this one.
30560         (inline_free_summary): Rename to ...
30561         (ipa_free_fn_summary): ... this one.
30562         (pass_data_local_fn_summary, pass_local_fn_summary,
30563         make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30564         pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30565         pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30566         make_pass_ipa_fn_summary): New.
30567         * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30568         inline_write_summary, inline_free_summary): Remove.
30569         (ipa_free_fn_summary) : New.
30570         * ipa-inline.c (ipa_inline): Update.
30571         (pass_ipa_inline): Do not generate summaries.
30572         * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30573         Remove.
30574         * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30575         and add pass_ipa_fn_summary.
30576         * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30577         New.
30578         (make_pass_inline_parameters): Remove.
30580 2017-05-23  Thomas Schwinge  <thomas@codesourcery.com>
30582         * omp-low.c (struct omp_context): Remove "default_kind" member.
30583         Adjust all users.
30585         * omp-offload.c (execute_oacc_device_lower): Remove the
30586         parallelism dimensions function attributes for unparallelized
30587         OpenACC kernels constructs.
30589 2017-05-23  Martin Liska  <mliska@suse.cz>
30591         * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30592         functions.
30593         (cgraph_edge::make_speculative): Likewise.
30594         (cgraph_edge::resolve_speculation): Likewise.
30595         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30596         (cgraph_node::dump): Likewise.
30597         * cgraph.h: Likewise.
30598         * cgraphunit.c (analyze_functions): Likewise.
30599         (symbol_table::compile): Likewise.
30600         * ipa-cp.c (print_all_lattices): Likewise.
30601         (determine_versionability): Likewise.
30602         (initialize_node_lattices): Likewise.
30603         (ipcp_verify_propagated_values): Likewise.
30604         (estimate_local_effects): Likewise.
30605         (update_profiling_info): Likewise.
30606         (create_specialized_node): Likewise.
30607         (perhaps_add_new_callers): Likewise.
30608         (decide_about_value): Likewise.
30609         (decide_whether_version_node): Likewise.
30610         (identify_dead_nodes): Likewise.
30611         (ipcp_store_bits_results): Likewise.
30612         * ipa-devirt.c (dump_targets): Likewise.
30613         (ipa_devirt): Likewise.
30614         * ipa-icf.c (sem_item::dump): Likewise.
30615         (sem_function::equals): Likewise.
30616         (sem_variable::equals): Likewise.
30617         (sem_item_optimizer::read_section): Likewise.
30618         (sem_item_optimizer::execute): Likewise.
30619         (congruence_class::dump): Likewise.
30620         * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30621         (dump_inline_summary): Likewise.
30622         (estimate_node_size_and_time): Likewise.
30623         (inline_analyze_function): Likewise.
30624         * ipa-inline-transform.c (inline_call): Likewise.
30625         * ipa-inline.c (report_inline_failed_reason): Likewise.
30626         (want_early_inline_function_p): Likewise.
30627         (edge_badness): Likewise.
30628         (update_edge_key): Likewise.
30629         (inline_small_functions): Likewise.
30630         * ipa-profile.c (ipa_profile): Likewise.
30631         * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30632         (ipa_make_edge_direct_to_target): Likewise.
30633         (remove_described_reference): Likewise.
30634         (ipa_impossible_devirt_target): Likewise.
30635         (propagate_controlled_uses): Likewise.
30636         (ipa_print_node_params): Likewise.
30637         (ipcp_transform_function): Likewise.
30638         * ipa-pure-const.c (pure_const_read_summary): Likewise.
30639         (propagate_pure_const): Likewise.
30640         * ipa-reference.c (generate_summary): Likewise.
30641         (read_write_all_from_decl): Likewise.
30642         (propagate): Likewise.
30643         (ipa_reference_read_optimization_summary): Likewise.
30644         * ipa-utils.c (ipa_merge_profiles): Likewise.
30645         * ipa.c (walk_polymorphic_call_targets): Likewise.
30646         (symbol_table::remove_unreachable_nodes): Likewise.
30647         (ipa_single_use): Likewise.
30648         * passes.c (execute_todo): Likewise.
30649         * predict.c (drop_profile): Likewise.
30650         * symtab.c (symtab_node::get_dump_name): New function.
30651         (symtab_node::dump_name): Likewise.
30652         (symtab_node::dump_asm_name): Likewise.
30653         (symtab_node::dump_references): Likewise.
30654         (symtab_node::dump_referring): Likewise.
30655         (symtab_node::dump_base): Likewise.
30656         (symtab_node::debug_symtab): Likewise.
30657         * tree-sra.c (convert_callers_for_node): Likewise.
30658         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30659         * value-prof.c (init_node_map): Likewise.
30661 2017-05-23  Martin Liska  <mliska@suse.cz>
30663         * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30664         and symtab_node::debug_symtab to symbol_table::debug.
30665         * cgraphunit.c (analyze_functions): Use the renamed function.
30666         (symbol_table::compile): Likewise.
30667         * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30668         * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30669         * passes.c (execute_todo): Likewise.
30670         * symtab.c (symbol_table::dump): New function.
30671         * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30673 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30675         * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30676         that nonconst implies exec.
30678 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30680         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30681         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30682         (inline_edge_summary_vec): Turn into ...
30683         (ipa_call_summaries): ... this one.
30684         (redirect_to_unreachable, edge_set_predicate,
30685         evaluate_properties_for_edge, inline_summary_alloc,
30686         reset_ipa_call_summary, reset_inline_summary,
30687         inline_summary_t::duplicate): Update.
30688         (inline_edge_duplication_hook): Turn to ...
30689         (ipa_call_summary_t::duplicate): ... this one.
30690         (inline_edge_removal_hook): Turn to ...
30691         (ipa_call_summary_t::remove): ... this one.
30692         (dump_inline_edge_summary): Turn to ...
30693         (dump_ipa_call_summary): ... this one.
30694         (estimate_function_body_sizes): Update.
30695         (inline_update_callee_summaries): Update.
30696         (remap_edge_change_prob): Update.
30697         (remap_edge_summaries): Update.
30698         (inline_merge_summary): Update.
30699         (do_estimate_edge_time): Update.
30700         (inline_generate_summary): Update.
30701         (inline_read_section): Update.
30702         (inline_read_summary): Update.
30703         (inline_free_summary): Update.
30704         * ipa-inline.c (can_inline_edge_p): Update.
30705         (compute_inlined_call_time): Update.
30706         (want_inline_small_function_p): Update.
30707         (edge_badness): Update.
30708         (early_inliner): Update.
30709         * ipa-inline.h (inline_edge_summary): Turn to ...
30710         (ipa_call_summary): ... this one.
30711         (ipa_call_summary_t): New class.
30712         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30713         (ipa_call_summaries): New.
30714         (inline_edge_summary): Remove.
30715         (estimate_edge_growth): Update.
30716         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30717         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30718         * ipa-split.c (execute_split_functions): Update.
30719         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30721 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30723         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30724         attributes): Document rdrand effective target.
30726 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30728         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30729         attributes): Sort alphabetically.
30731 2017-05-23  Georg-Johann Lay  <avr@gjlay.de>
30733         * config/avr/genmultilib.awk: Use gsub instead of gensub.
30735 2017-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30737         PR target/80718
30738         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30739         V2DF/V2DI splat into two separate patterns, one that handles
30740         registers, and the other that only handles memory.  Drop support
30741         for splatting from a GPR on ISA 2.07 and then splitting the
30742         splat into direct move and splat.
30743         (vsx_splat_<mode>_reg): Likewise.
30744         (vsx_splat_<mode>_mem): Likewise.
30746 2017-05-22  Segher Boessenkool  <segher@kernel.crashing.org>
30748         * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30750 2017-05-22  Jakub Jelinek  <jakub@redhat.com>
30752         PR middle-end/80809
30753         * omp-low.c (finish_taskreg_remap): New function.
30754         (finish_taskreg_scan): If unit size of ctx->record_type
30755         is non-constant, unshare the size expression and replace
30756         decls in it with possible outer var refs.
30758         PR middle-end/80809
30759         * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30760         GOVD_SHARED rather than GOVD_PRIVATE with it.
30761         (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30762         GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30764         PR middle-end/80853
30765         * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30766         as last argument to build_outer_var_ref for pointer bases of array
30767         section reductions.
30769 2017-05-19  Martin Sebor  <msebor@redhat.com>
30771         * print-tree.c (print_node): Print DECL_READ_P flag.
30773 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30775         * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30776         * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30777         * cgraph.c: Likewise.
30778         * cgraphunit.c: Likewise.
30779         * gengtype.c: Likewise.
30780         * ipa-cp.c: Likewise.
30781         * ipa-devirt.c: Likewise.
30782         * ipa-icf.c: Likewise.
30783         * ipa-predicate.c: Likewise.
30784         * ipa-profile.c: Likewise.
30785         * ipa-prop.c: Likewise.
30786         * ipa-split.c: Likewise.
30787         * ipa.c: Likewise.
30788         * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30789         edge_predicate_pool, dump_inline_hints,
30790         inline_summary::account_size_time, redirect_to_unreachable,
30791         edge_set_predicate, set_hint_predicate,
30792         evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30793         inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30794         inline_summary_t::remove, remap_hint_predicate_after_duplication,
30795         inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30796         ipa_call_summary_t::remove, initialize_growth_caches,
30797         free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30798         debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30799         mark_modified, unmodified_parm_1, unmodified_parm,
30800         unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30801         set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30802         compute_bb_predicates, will_be_nonconstant_expr_predicate,
30803         will_be_nonconstant_predicate, record_modified_bb_info,
30804         get_minimal_bb, record_modified, param_change_prob,
30805         phi_result_unknown_predicate, predicate_for_phi_result,
30806         array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30807         estimate_function_body_sizes, compute_inline_parameters,
30808         compute_inline_parameters_for_curren, pass_data_inline_parameters,
30809         estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30810         inline_update_callee_summaries, remap_edge_change_prob,
30811         remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30812         inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30813         inline_analyze_function, inline_summary_t::insert,
30814         inline_generate_summary, read_ipa_call_summary, inline_read_section,
30815         inline_read_summary, write_ipa_call_summary, inline_write_summary,
30816         inline_free_summary): Move to ipa-fnsummary.h
30817         (predicate_t): Remove.
30818         * ipa-fnsummary.c: New file.
30819         * ipa-inline.h:  Do not include sreal.h and ipa-predicate.h
30820         (enum inline_hints_vals, inline_hints, agg_position_info,
30821         INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30822         inline_summaries, ipa_call_summary, ipa_call_summary_t,
30823         ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30824         dump_inline_summary, dump_inline_hints, inline_generate_summary,
30825         inline_read_summary, inline_write_summary, inline_free_summary,
30826         inline_analyze_function, initialize_inline_failed,
30827         inline_merge_summary, inline_update_overall_summary,
30828         compute_inline_parameters): Move to ipa-fnsummary.h
30829         * ipa-fnsummary.h: New file.
30830         * ipa-inline-transform.h: Include ipa-inline.h.
30831         * ipa-inline.c: LIkewise.
30833 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30835         * ipa-inline.c (edge_badness): Use inlined_time instead of
30836         inline_summaries->get.
30838 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30840         * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30842 2017-05-22  Nathan Sidwell  <nathan@acm.org>
30844         * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30845         (fdump-lang): Document 'raw' option.
30846         * dumpfile.h (TDI_tu): Delete.
30847         * dumpfile.c (dump_files): Remove translation-unit.
30848         (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30850 2017-05-22  Georg-Johann Lay  <avr@gjlay.de>
30852         * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30853         command option from $(AWK) call.
30854         * config/avr/genmultilib.awk: Simplify and rewrite so that it
30855         generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30856         [FORMAT]: Remove handling of variable.
30857         * config/avr/t-multilib: Regenerate.
30859 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30861         * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30862         self_time.
30863         (dump_inline_summary): Do not print self_time.
30864         (estimate_function_body_sizes): Do not set self_time.
30865         (compute_inline_parameters): Likewise.
30866         (inline_read_section, inline_write_summary): Do not stream self_time.
30867         * ipa-inline.h (inline_summary): Drop self_time.
30869 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30871         * ipa-inline-analysis.c (account_size_time): Rename to ...
30872         (inline_summary::account_size_time): ... this one.
30873         (reset_ipa_call_summary): Turn to ...
30874         (ipa_call_summary::reset): ... this one.
30875         (reset_inline_summary): Turn to ...
30876         (inline_summary::reset): ... this one.
30877         (inline_summary_t::remove): Update.
30878         (inline_summary_t::duplicate): Update.
30879         (ipa_call_summary_t::remove): Update.
30880         (dump_inline_summary): Update.
30881         (estimate_function_body_sizes): Update.
30882         (compute_inline_parameters): Update.
30883         (estimate_node_size_and_time): Update.
30884         (inline_merge_summary): Update.
30885         (inline_update_overall_summary): Update.
30886         (inline_read_section): Update.
30887         (inline_write_summary): Update.
30888         * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30889         add account_size_time and reset member functions.
30890         (ipa_call_summary): Add reset function.
30891         * ipa-predicate.h (predicate::operator &): Constify.
30893 2017-05-22  Richard Biener  <rguenther@suse.de>
30895         * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30897 2017-05-19  Jason Merrill  <jason@redhat.com>
30899         * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
30901 2017-05-19  Marek Polacek  <polacek@redhat.com>
30903         PR sanitizer/80800
30904         * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
30905         TYPE_OVERFLOW_WRAPS checks.
30907 2017-05-19  Thomas Schwinge  <thomas@codesourcery.com>
30909         * tree-core.h (enum omp_clause_default_kind): Add
30910         "OMP_CLAUSE_DEFAULT_PRESENT".
30911         * tree-pretty-print.c (dump_omp_clause): Handle it.
30912         * gimplify.c (enum gimplify_omp_var_data): Add
30913         "GOVD_MAP_FORCE_PRESENT".
30914         (gimplify_adjust_omp_clauses_1): Map it to
30915         "GOMP_MAP_FORCE_PRESENT".
30916         (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
30918         * gimplify.c (oacc_default_clause): Clarify.
30920 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30922         LANG_HOOK_REGISTER_DUMPS
30923         * toplev.c (general_init): Call register dump lang hook.
30924         * doc/invoke.texi: Document -fdump-lang option family.
30925         * dumpfile.c (dump_files): Remove class dump here.
30926         (FIRST_AUTO_NUMBERED_DUMP): Adjust.
30927         * dumpfile.h (tree_dump_index): Remove TDI_class.
30928         * langhooks-def.h (lhd_register_dumps): Declare.
30929         (LANG_HOOKS_REGISTER_DUMPS): Define.
30930         (LANG_HOOKS_INITIALIZER): Add it.
30931         * langhooks.c (lhd_register_dumps): Define.
30932         * langhooks.h (struct lang_hooks): Add register_dumps.
30934 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30936         * context.h (context::set_passes): New.
30937         * context.c (context::context): Do not create pass manager.
30938         * toplev.c (general_init): Create pass manager here.
30940 2017-05-19  Segher Boessenkool  <segher@kernel.crashing.org>
30942         * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
30943         use this splitter if two add or or instructions would also work for
30944         the constant we want to generate.
30946 2017-05-19  Richard Biener  <rguenther@suse.de>
30948         PR build/80821
30949         * genmatch.c (dt_node::gen_kids_1): Add missing scope around
30950         predicate evaluation.
30952 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30954         * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
30955         add ctor.
30956         * ipa-inline.c (want_inline_small_function_p): Do not cast to
30957         unsigned.
30959 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30961         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30962         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30963         (inline_edge_summary_vec): Turn into ...
30964         (ipa_call_summaries): ... this one.
30965         (redirect_to_unreachable, edge_set_predicate,
30966         evaluate_properties_for_edge, inline_summary_alloc,
30967         reset_ipa_call_summary, reset_inline_summary,
30968         inline_summary_t::duplicate): Update.
30969         (inline_edge_duplication_hook): Turn to ...
30970         (ipa_call_summary_t::duplicate): ... this one.
30971         (inline_edge_removal_hook): Turn to ...
30972         (ipa_call_summary_t::remove): ... this one.
30973         (dump_inline_edge_summary): Turn to ...
30974         (dump_ipa_call_summary): ... this one.
30975         (estimate_function_body_sizes): Update.
30976         (inline_update_callee_summaries): Update.
30977         (remap_edge_change_prob): Update.
30978         (remap_edge_summaries): Update.
30979         (inline_merge_summary): Update.
30980         (do_estimate_edge_time): Update.
30981         (inline_generate_summary): Update.
30982         (inline_read_section): Update.
30983         (inline_read_summary): Update.
30984         (inline_free_summary): Update.
30985         * ipa-inline.c (can_inline_edge_p): Update.
30986         (compute_inlined_call_time): Update.
30987         (want_inline_small_function_p): Update.
30988         (edge_badness): Update.
30989         (early_inliner): Update.
30990         * ipa-inline.h (inline_edge_summary): Turn to ...
30991         (ipa_call_summary): ... this one.
30992         (ipa_call_summary_t): New class.
30993         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30994         (ipa_call_summaries): New.
30995         (inline_edge_summary): Remove.
30996         (estimate_edge_growth): Update.
30997         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30998         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30999         * ipa-split.c (execute_split_functions): Update.
31000         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
31002 2017-05-19  Richard Biener  <rguenther@suse.de>
31004         PR middle-end/80764
31005         * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
31007 2017-05-18  Segher Boessenkool  <segher@kernel.crashing.org>
31009         * config/rs6000/rs6000.c (struct machine_function): Add field
31010         fpr_is_wrapped_separately.
31011         (rs6000_get_separate_components): Use 64 components.  Handle the
31012         new FPR components.
31013         (rs6000_components_for_bb): Handle the FPR components.
31014         (rs6000_emit_prologue_components): Handle the FPR components.
31015         (rs6000_emit_epilogue_components): Handle the FPR components.
31016         (rs6000_set_handled_components): Handle the FPR components.
31017         (rs6000_emit_prologue): Don't output prologue code for those FPRs
31018         that are already separately shrink-wrapped.
31019         (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
31020         that are already separately shrink-wrapped.
31022 2017-05-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
31024         PR target/80510
31025         * config/rs6000/predicates.md (simple_offsettable_mem_operand):
31026         New predicate.
31028         * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
31029         (define_peephole2 for Altivec d-form load): Add peepholes to catch
31030         cases where the register allocator uses a move and an offsettable
31031         memory operation to/from a FPR register on ISA 2.06/2.07.
31032         (define_peephole2 for Altivec d-form store): Likewise.
31034 2017-05-18  Uros Bizjak  <ubizjak@gmail.com>
31036         PR target/80799
31037         * config/i386/mmx.md (*mov<mode>_internal): Enable
31038         alternatives 11, 12, 13 and 14 also for 32bit targets.
31039         Remove alternatives 15, 16, 17 and 18.
31040         * config/i386/sse.md (vec_concatv2di): Change
31041         alternative (!x, *y) to (x, ?!*Yn).
31043 2017-05-18  Paolo Carlini  <paolo.carlini@oracle.com>
31045         * dumpfile.h (enum dump_kind): Remove stray comma.
31047 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31049         * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
31050         * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
31051         predicate::num_conditions
31052         (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
31053         (CHANGED): turn into predicate::changed.
31054         (agg_position_info): Move to ipa-predicate.h
31055         (add_condition, predicate::add_clause, predicate::operator &=,
31056         predicate::or_with, predicate::evaluate, predicate::probability,
31057         dump_condition, dump_clause, predicate::dump,
31058         predicate::remap_after_duplication, predicate::remap_after_inlining,
31059         predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
31060         (evaluate_conditions_for_known_args): Update.
31061         (set_cond_stmt_execution_predicate): Update.
31062         * ipa-inline.h: Include ipa-predicate.h
31063         (condition, inline_param_summary, conditions, agg_position_info,
31064         predicate): Move to ipa-predicate.h
31065         * ipa-predicate.c: New file.
31066         * ipa-predicate.h: New file.
31068 2017-05-18  Wilco Dijkstra  <wdijkstr@arm.com>
31070         * final.c (leaf_function_p): Check we are not in a sequence.
31072 2017-05-18  Martin Liska  <mliska@suse.cz>
31074         * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
31075         * dumpfile.c (dump_register): Use new enum dump_kind.
31076         (get_dump_file_name): Likewise.
31077         (dump_enable_all): Likewise.
31078         (dump_switch_p_1): Likewise.
31079         (enable_rtl_dump_file): Remove usage of TDF_RTL.
31080         * dumpfile.h (enum dump_kind): New enum type.
31081         (struct dump_file_info): Create constructor and
31082         format fields and comments.
31083         * passes.c (pass_manager::register_one_dump_file):
31084         Use num dump_kind.
31085         * statistics.c (statistics_early_init): Likewise.
31086         * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
31087         TDF_TREE with TDF_SLIM.
31088         (gather_memory_references_ref): Likewise.
31090 2017-05-18  Martin Liska  <mliska@suse.cz>
31092         * vec.h (struct vnull): Use it.
31094 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31096         * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
31097         (true_predicate, false_predicate, true_predicate_p,
31098         false_predicate_p): Remove.
31099         (single_cond_predicate, not_inlined_predicate): Turn to member function
31100         in ipa-inline.h
31101         (add_condition): Update.
31102         (add_clause): Turn to...
31103         (predicate::add_clause): ... this one; update; allow passing NULL
31104         as parameter.
31105         (and_predicates): Turn to ...
31106         (predicate::operator &=): ... this one.
31107         (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
31108         (or_predicates): Turn to ...
31109         (predicate::or_with): ... this one.
31110         (evaluate_predicate): Turn to ...
31111         (predicate::evaluate): ... this one.
31112         (predicate_probability): Turn to ...
31113         (predicate::probability): ... this one.
31114         (dump_condition): Update.
31115         (dump_predicate): Turn to ...
31116         (predicate::dump): ... this one.
31117         (account_size_time): Update.
31118         (edge_set_predicate): Update.
31119         (set_hint_predicate): UPdate.
31120         (evaluate_conditions_for_known_args): Update.
31121         (evaluate_properties_for_edge): Update.
31122         (remap_predicate_after_duplication): Turn to...
31123         (predicate::remap_after_duplication): ... this one.
31124         (remap_hint_predicate_after_duplication): Update.
31125         (inline_summary_t::duplicate): UPdate.
31126         (dump_inline_edge_summary): Update.
31127         (dump_inline_summary): Update.
31128         (set_cond_stmt_execution_predicate): Update.
31129         (set_switch_stmt_execution_predicate): Update.
31130         (compute_bb_predicates): Update.
31131         (will_be_nonconstant_expr_predicate): Update.
31132         (will_be_nonconstant_predicate): Update.
31133         (phi_result_unknown_predicate): Update.
31134         (predicate_for_phi_result): Update.
31135         (array_index_predicate): Update.
31136         (estimate_function_body_sizes): Update.
31137         (estimate_node_size_and_time): Update.
31138         (estimate_ipcp_clone_size_and_time): Update.
31139         (remap_predicate): Rename to ...
31140         (predicate::remap_after_inlining): ... this one.
31141         (remap_hint_predicate): Update.
31142         (inline_merge_summary): Update.
31143         (inline_update_overall_summary): Update.
31144         (estimate_size_after_inlining): Update.
31145         (read_predicate): Rename to ...
31146         (predicate::stream_in): ... this one.
31147         (read_inline_edge_summary): Update.
31148         (write_predicate): Rename to ...
31149         (predicate::stream_out): ... this one.
31150         (write_inline_edge_summary): Update.
31151         * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
31152         (clause_t): Turn to uint32_t
31153         (predicate): Turn to class; implement constructor and operators
31154         ==, !=, &
31155         (size_time_entry): Update.
31156         (inline_summary): Update.
31157         (inline_edge_summary): Update.
31159 2017-05-18  Marc Glisse  <marc.glisse@inria.fr>
31161         * fold-const.c (fold_binary_loc): Move transformation...
31162         * match.pd (C - X CMP X): ... here.
31164 2017-05-18  Sheldon Lobo  <sheldon.lobo@oracle.com>
31166         * config/sparc/sparc.c (sparc_option_override): Set function
31167         alignment for -mcpu=niagara7 to 64 to match the I$ line.
31168         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
31169         latency to 1.
31170         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
31171         latency to 2.
31172         * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
31174 2017-05-18  Marek Polacek  <polacek@redhat.com>
31176         PR sanitizer/80797
31177         * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
31178         (pass_ubsan::execute): Call gimple_assign_single_p instead of
31179         gimple_assign_load_p.
31181 2017-05-17  Segher Boessenkool  <segher@kernel.crashing.org>
31183         PR middle-end/80692
31184         * real.c (do_compare): Give decimal_do_compare preference over
31185         comparing just the signs.
31187 2017-05-17  Uros Bizjak  <ubizjak@gmail.com>
31189         * doc/md.texi (Canonicalization of Instructions): Describe the
31190         canonical form of instructions that inherently set a condition
31191         code register.
31193 2017-05-17  Peter Bergner  <bergner@vnet.ibm.com>
31195         PR middle-end/80775
31196         * tree-cfg.c: Move deletion of unreachable case statements to after
31197         the merging of consecutive case labels.
31199 2017-05-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
31201         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
31202         readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
31203         restoring of callee-saved registers.
31205 2017-05-17  Eric Botcazou  <ebotcazou@adacore.com>
31207         * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
31208         * config/visium/visium.c (single_set_and_flags): Likewise.
31209         * config/visium/visium.md (Substitutions): Likewise.
31211 2017-05-17  Martin Liska  <mliska@suse.cz>
31213         * cfg.c: Introduce dump_flags_t type and
31214         use it instead of int type.
31215         * cfg.h: Likewise.
31216         * cfghooks.c: Likewise.
31217         * cfghooks.h (struct cfg_hooks): Likewise.
31218         * cfgrtl.c: Likewise.
31219         * cfgrtl.h: Likewise.
31220         * cgraph.c (cgraph_node::get_body): Likewise.
31221         * coretypes.h: Likewise.
31222         * domwalk.c: Likewise.
31223         * domwalk.h: Likewise.
31224         * dumpfile.c (struct dump_option_value_info): Likewise.
31225         (dump_enable_all): Likewise.
31226         (dump_switch_p_1): Likewise.
31227         (opt_info_switch_p): Likewise.
31228         * dumpfile.h (enum tree_dump_index): Likewise.
31229         (struct dump_file_info): Likewise.
31230         * genemit.c: Likewise.
31231         * generic-match-head.c: Likewise.
31232         * gengtype.c (open_base_files): Likewise.
31233         * gimple-pretty-print.c: Likewise.
31234         * gimple-pretty-print.h: Likewise.
31235         * graph.c (print_graph_cfg): Likewise.
31236         * graphite-scop-detection.c (dot_all_sese): Likewise.
31237         * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31238         * loop-unroll.c (report_unroll): Likewise.
31239         * passes.c (pass_manager::register_one_dump_file): Likewise.
31240         * print-tree.c: Likewise.
31241         * statistics.c: Likewise.
31242         * tree-cfg.c: Likewise.
31243         * tree-cfg.h: Likewise.
31244         * tree-dfa.c: Likewise.
31245         * tree-dfa.h: Likewise.
31246         * tree-dump.c (dump_function): Likewise.
31247         * tree-dump.h (struct dump_info): Likewise.
31248         * tree-pretty-print.c: Likewise.
31249         * tree-pretty-print.h: Likewise.
31250         * tree-ssa-live.c: Likewise.
31251         * tree-ssa-live.h: Likewise.
31252         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31253         * tree-vect-loop.c: Likewise.
31254         * tree-vect-slp.c: Likewise.
31256 2017-05-16  James Greenhalgh  <james.greenhalgh@arm.com>
31257             Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31259         PR tree-optimization/80457
31260         * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31261         of all arguments to a statement as scalar_to_vec operations.
31262         (vectorizable_call): Adjust call to vect_model_simple_cost for
31263         new parameter.
31264         (vectorizable_conversion): Likewise.
31265         (vectorizable_assignment): Likewise.
31266         (vectorizable_shift): Likewise.
31267         (vectorizable_operation): Likewise.
31268         (vectorizable_comparison): Likewise.
31269         (vect_is_simple_cond): Record the def types for operands.
31270         (vectorizable_condition): Likewise, call vect_model_simple_cost.
31271         * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31272         for statement argument count.
31274 2017-05-16  Carl Love  <cel@us.ibm.com>
31276         * config/rs6000/rs6000-c: Add support for built-in functions
31277         vector unsigned long long vec_bperm (vector unsigned long long,
31278                                              vector unsigned char)
31279         vector signed long long vec_mule (vector signed int,
31280                                           vector signed int)
31281         vector unsigned long long vec_mule (vector unsigned int,
31282                                             vector unsigned int)
31283         vector signed long long vec_mulo (vector signed int,
31284                                           vector signed int)
31285         vector unsigned long long vec_mulo (vector unsigned int,
31286                                             vector unsigned int)
31287         vector signed char vec_sldw (vector signed char,
31288                                      vector signed char,
31289                                      const int)
31290         vector unsigned char vec_sldw (vector unsigned char,
31291                                        vector unsigned char,
31292                                        const int)
31293         vector signed short vec_sldw (vector signed short,
31294                                       vector signed short,
31295                                       const int)
31296         vector unsigned short vec_sldw (vector unsigned short,
31297                                         vector unsigned short,
31298                                         const int)
31299         vector signed int vec_sldw (vector signed int,
31300                                     vector signed int,
31301                                     const int)
31302         vector unsigned int vec_sldw (vector unsigned int,
31303                                       vector unsigned int,
31304                                       const int)
31305         vector signed long long vec_sldw (vector signed long long,
31306                                           vector signed long long,
31307                                           const int)
31308         vector unsigned long long vec_sldw (vector unsigned long long,
31309                                             vector unsigned long long,
31310                                             const int)
31311         * config/rs6000/rs6000-c: Add support for built-in functions
31312         * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31313         * config/rs6000/altivec.h: Add defintion for vec_sldw.
31314         * doc/extend.texi: Update the built-in documentation for the
31315         new built-in functions.
31317 2017-05-16  Marek Polacek  <polacek@redhat.com>
31319         PR sanitizer/80536
31320         PR sanitizer/80386
31321         * tree.c (save_expr): Don't fold the expression.
31323 2017-05-16  Uros Bizjak  <ubizjak@gmail.com>
31325         * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31326         to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31327         and (?*y,m).  Update insn attributes.
31329 2017-05-16  Martin Liska  <mliska@suse.cz>
31331         * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31332         flags argument of print_gimple_stmt, print_gimple_expr,
31333         print_generic_stmt and print_generic_expr.
31334         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31335         * coretypes.h: Likewise.
31336         * except.c (dump_eh_tree): Likewise.
31337         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31338         * gimple-pretty-print.h: Likewise.
31339         * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31340         (backprop::push_to_worklist): Likewise.
31341         (backprop::pop_from_worklist): Likewise.
31342         (backprop::process_use): Likewise.
31343         (backprop::intersect_uses): Likewise.
31344         (note_replacement): Likewise.
31345         * gimple-ssa-store-merging.c
31346         (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31347         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31348         (pass_store_merging::execute): Likewise.
31349         * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31350         (ssa_base_cand_dump_callback): Likewise.
31351         (dump_incr_vec): Likewise.
31352         (replace_refs): Likewise.
31353         (replace_mult_candidate): Likewise.
31354         (create_add_on_incoming_edge): Likewise.
31355         (create_phi_basis): Likewise.
31356         (insert_initializers): Likewise.
31357         (all_phi_incrs_profitable): Likewise.
31358         (introduce_cast_before_cand): Likewise.
31359         (replace_one_candidate): Likewise.
31360         * gimplify.c (gimplify_expr): Likewise.
31361         * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31362         (set_rename): Likewise.
31363         (rename_uses): Likewise.
31364         (copy_loop_phi_nodes): Likewise.
31365         (add_close_phis_to_merge_points): Likewise.
31366         (copy_loop_close_phi_args): Likewise.
31367         (copy_cond_phi_args): Likewise.
31368         (graphite_copy_stmts_from_block): Likewise.
31369         (translate_pending_phi_nodes): Likewise.
31370         * graphite-poly.c (print_pdr): Likewise.
31371         (dump_gbb_cases): Likewise.
31372         (dump_gbb_conditions): Likewise.
31373         (print_scop_params): Likewise.
31374         * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31375         (build_cross_bb_scalars_use): Likewise.
31376         (gather_bbs::before_dom_children): Likewise.
31377         * hsa-dump.c (dump_hsa_immed): Likewise.
31378         * ipa-cp.c (print_ipcp_constant_value): Likewise.
31379         (get_replacement_map): Likewise.
31380         * ipa-inline-analysis.c (dump_condition): Likewise.
31381         (estimate_function_body_sizes): Likewise.
31382         * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31383         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31384         * ipa-prop.c (ipa_dump_param): Likewise.
31385         (ipa_print_node_jump_functions_for_edge): Likewise.
31386         (ipa_modify_call_arguments): Likewise.
31387         (ipa_modify_expr): Likewise.
31388         (ipa_dump_param_adjustments): Likewise.
31389         (ipa_dump_agg_replacement_values): Likewise.
31390         (ipcp_modif_dom_walker::before_dom_children): Likewise.
31391         * ipa-pure-const.c (check_stmt): Likewise.
31392         (pass_nothrow::execute): Likewise.
31393         * ipa-split.c (execute_split_functions): Likewise.
31394         * omp-offload.c (dump_oacc_loop_part): Likewise.
31395         (dump_oacc_loop): Likewise.
31396         * trans-mem.c (tm_log_emit): Likewise.
31397         (tm_memopt_accumulate_memops): Likewise.
31398         (dump_tm_memopt_set): Likewise.
31399         (dump_tm_memopt_transform): Likewise.
31400         * tree-cfg.c (gimple_verify_flow_info): Likewise.
31401         (print_loop): Likewise.
31402         * tree-chkp-opt.c (chkp_print_addr): Likewise.
31403         (chkp_gather_checks_info): Likewise.
31404         (chkp_get_check_result): Likewise.
31405         (chkp_remove_check_if_pass): Likewise.
31406         (chkp_use_outer_bounds_if_possible): Likewise.
31407         (chkp_reduce_bounds_lifetime): Likewise.
31408         * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31409         (chkp_mark_completed_bounds): Likewise.
31410         (chkp_register_incomplete_bounds): Likewise.
31411         (chkp_mark_invalid_bounds): Likewise.
31412         (chkp_maybe_copy_and_register_bounds): Likewise.
31413         (chkp_build_returned_bound): Likewise.
31414         (chkp_get_bound_for_parm): Likewise.
31415         (chkp_build_bndldx): Likewise.
31416         (chkp_get_bounds_by_definition): Likewise.
31417         (chkp_generate_extern_var_bounds): Likewise.
31418         (chkp_get_bounds_for_decl_addr): Likewise.
31419         * tree-chrec.c (chrec_apply): Likewise.
31420         * tree-data-ref.c (dump_data_reference): Likewise.
31421         (dump_subscript): Likewise.
31422         (dump_data_dependence_relation): Likewise.
31423         (analyze_overlapping_iterations): Likewise.
31424         * tree-inline.c (expand_call_inline): Likewise.
31425         (tree_function_versioning): Likewise.
31426         * tree-into-ssa.c (dump_defs_stack): Likewise.
31427         (dump_currdefs): Likewise.
31428         (dump_names_replaced_by): Likewise.
31429         (dump_update_ssa): Likewise.
31430         (update_ssa): Likewise.
31431         * tree-object-size.c (pass_object_sizes::execute): Likewise.
31432         * tree-parloops.c (build_new_reduction): Likewise.
31433         (try_create_reduction_list): Likewise.
31434         (ref_conflicts_with_region): Likewise.
31435         (oacc_entry_exit_ok_1): Likewise.
31436         (oacc_entry_exit_single_gang): Likewise.
31437         * tree-pretty-print.h: Likewise.
31438         * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31439         (get_scalar_evolution): Likewise.
31440         (add_to_evolution): Likewise.
31441         (get_loop_exit_condition): Likewise.
31442         (analyze_evolution_in_loop): Likewise.
31443         (analyze_initial_condition): Likewise.
31444         (analyze_scalar_evolution): Likewise.
31445         (instantiate_scev): Likewise.
31446         (number_of_latch_executions): Likewise.
31447         (gather_chrec_stats): Likewise.
31448         (final_value_replacement_loop): Likewise.
31449         (scev_const_prop): Likewise.
31450         * tree-sra.c (dump_access): Likewise.
31451         (disqualify_candidate): Likewise.
31452         (create_access): Likewise.
31453         (reject): Likewise.
31454         (maybe_add_sra_candidate): Likewise.
31455         (create_access_replacement): Likewise.
31456         (analyze_access_subtree): Likewise.
31457         (analyze_all_variable_accesses): Likewise.
31458         (sra_modify_assign): Likewise.
31459         (initialize_constant_pool_replacements): Likewise.
31460         (find_param_candidates): Likewise.
31461         (decide_one_param_reduction): Likewise.
31462         (replace_removed_params_ssa_names): Likewise.
31463         * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31464         * tree-ssa-copy.c (dump_copy_of): Likewise.
31465         (copy_prop_visit_cond_stmt): Likewise.
31466         * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31467         * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31468         (record_equivalences_from_stmt): Likewise.
31469         * tree-ssa-dse.c (compute_trims): Likewise.
31470         (delete_dead_call): Likewise.
31471         (delete_dead_assignment): Likewise.
31472         * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31473         (forward_propagate_into_cond): Likewise.
31474         (pass_forwprop::execute): Likewise.
31475         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31476         * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31477         Likewise.
31478         (move_computations_worker): Likewise.
31479         (execute_sm): Likewise.
31480         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31481         (remove_exits_and_undefined_stmts): Likewise.
31482         (remove_redundant_iv_tests): Likewise.
31483         * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31484         (adjust_iv_update_pos): Likewise.
31485         * tree-ssa-math-opts.c (bswap_replace): Likewise.
31486         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31487         (value_replacement): Likewise.
31488         * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31489         * tree-ssa-pre.c (print_pre_expr): Likewise.
31490         (get_representative_for): Likewise.
31491         (create_expression_by_pieces): Likewise.
31492         (insert_into_preds_of_block): Likewise.
31493         (eliminate_insert): Likewise.
31494         (eliminate_dom_walker::before_dom_children): Likewise.
31495         (eliminate): Likewise.
31496         (remove_dead_inserted_code): Likewise.
31497         * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31498         * tree-ssa-reassoc.c (get_rank): Likewise.
31499         (eliminate_duplicate_pair): Likewise.
31500         (eliminate_plus_minus_pair): Likewise.
31501         (eliminate_not_pairs): Likewise.
31502         (undistribute_ops_list): Likewise.
31503         (eliminate_redundant_comparison): Likewise.
31504         (update_range_test): Likewise.
31505         (optimize_range_tests_var_bound): Likewise.
31506         (optimize_vec_cond_expr): Likewise.
31507         (rewrite_expr_tree): Likewise.
31508         (rewrite_expr_tree_parallel): Likewise.
31509         (linearize_expr): Likewise.
31510         (break_up_subtract): Likewise.
31511         (linearize_expr_tree): Likewise.
31512         (attempt_builtin_powi): Likewise.
31513         (attempt_builtin_copysign): Likewise.
31514         (transform_stmt_to_copy): Likewise.
31515         (transform_stmt_to_multiply): Likewise.
31516         (dump_ops_vector): Likewise.
31517         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31518         (print_scc): Likewise.
31519         (set_ssa_val_to): Likewise.
31520         (visit_reference_op_store): Likewise.
31521         (visit_use): Likewise.
31522         (sccvn_dom_walker::before_dom_children): Likewise.
31523         (run_scc_vn): Likewise.
31524         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31525         Likewise.
31526         (expr_hash_elt::print): Likewise.
31527         (const_and_copies::pop_to_marker): Likewise.
31528         (const_and_copies::record_const_or_copy_raw): Likewise.
31529         * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31530         * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31531         (dump_predicates): Likewise.
31532         (find_uninit_use): Likewise.
31533         (warn_uninitialized_phi): Likewise.
31534         (pass_late_warn_uninitialized::execute): Likewise.
31535         * tree-ssa.c (verify_vssa): Likewise.
31536         (verify_ssa): Likewise.
31537         (maybe_optimize_var): Likewise.
31538         * tree-vrp.c (dump_value_range): Likewise.
31539         (dump_all_value_ranges): Likewise.
31540         (dump_asserts_for): Likewise.
31541         (register_edge_assert_for_2): Likewise.
31542         (vrp_visit_cond_stmt): Likewise.
31543         (vrp_visit_switch_stmt): Likewise.
31544         (vrp_visit_stmt): Likewise.
31545         (vrp_visit_phi_node): Likewise.
31546         (simplify_cond_using_ranges_1): Likewise.
31547         (fold_predicate_in): Likewise.
31548         (evrp_dom_walker::before_dom_children): Likewise.
31549         (evrp_dom_walker::push_value_range): Likewise.
31550         (evrp_dom_walker::pop_value_range): Likewise.
31551         (execute_early_vrp): Likewise.
31553 2017-05-16  Richard Biener  <rguenther@suse.de>
31555         * dwarf2out.c (loc_list_from_tree_1): Do not create
31556         DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31558 2017-05-16  Richard Biener  <rguenther@suse.de>
31560         * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31561         just generated.
31562         (note_variable_value_in_expr): If we resolved the decl ref
31563         do not push to the stack.
31565 2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
31567         * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31568         operations in fast-math mode.
31569         (vaddq_f16): Likewise.
31570         (vmul_f16): Likewise.
31571         (vmulq_f16): Likewise.
31572         (vsub_f16): Likewise.
31573         (vsubq_f16): Likewise.
31574         * config/arm/neon.md (add<mode>3): New.
31575         (sub<mode>3): New.
31576         (fma:<VH:mode>3): New.  Also remove outdated comment.
31577         (mul<mode>3): New.
31579 2017-05-16  Martin Liska  <mliska@suse.cz>
31581         PR ipa/79849.
31582         PR ipa/79850.
31583         * ipa-devirt.c (warn_types_mismatch): Fix typo.
31584         (odr_types_equivalent_p): Likewise.
31586 2017-05-15  Sylvestre Ledru  <sylvestre@debian.org>
31588         * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31590 2017-05-15  Uros Bizjak  <ubizjak@gmail.com>
31592         PR target/80425
31593         * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31594         non-interunit SSE move alternatives with '?'.
31595         (zero-extendsidi peephole2): New peephole to skip intermediate
31596         general register in SSE zero-extend sequence.
31598 2017-05-15  Jeff Law  <law@redhat.com>
31600         * reorg.c (relax_delay_slots): Create a new variable to hold
31601         the temporary target rather than clobbering TARGET_LABEL.
31603         * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31604         missing argument to extract_bit_field call.
31605         * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31607 2017-05-15  Martin Liska  <mliska@suse.cz>
31609         PR driver/31468
31610         * gcc.c (process_command): Do not allow empty argument of -o option.
31612 2017-05-15  Renlin Li  <renlin.li@arm.com>
31614         * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31615         * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31616         * config/aarch64/constraints.md (Usf): Add long call check.
31617         * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31618         (call_value): Likewise.
31619         (sibcall): Likewise.
31620         (sibcall_value): Likewise.
31621         (call_insn): New.
31622         (call_value_insn): New.
31623         (sibcall_insn): Update rtx pattern.
31624         (sibcall_value_insn): Likewise.
31625         (call_internal): Remove.
31626         (call_value_internal): Likewise.
31627         (sibcall_internal): Likewise.
31628         (sibcall_value_internal): Likewise.
31629         (call_reg): Likewise.
31630         (call_symbol): Likewise.
31631         (call_value_reg): Likewise.
31632         (call_value_symbol): Likewise.
31634 2017-05-14  Krister Walfridsson  <krister.walfridsson@gmail.com>
31636         PR target/80600
31637         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31639 2017-05-14  Uros Bizjak  <ubizjak@gmail.com>
31641         * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31642         compatible with CCGOCmode and with CCZmode.
31644 2017-05-14  Martin Sebor  <msebor@redhat.com>
31646         PR middle-end/77671
31647         * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31648         (gimple_fold_builtin_snprintf): Same.
31649         * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31650         (gimple_fold_builtin_snprintf): Same.
31651         * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31652         of character types.
31653         (is_call_safe): New function.
31654         (try_substitute_return_value): Call it.
31655         (try_simplify_call): New function.
31656         (pass_sprintf_length::handle_gimple_call): Call it.
31658 2017-05-14  Martin Sebor  <msebor@redhat.com>
31660         PR middle-end/80669
31661         * builtins.c (expand_builtin_stpncpy): Simplify.
31663 2017-05-14  Daniel Santos  <daniel.santos@pobox.com>
31665         * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31666         * config/i386/i386.h
31667         (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31668         (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31669         (struct machine_function): Add new members call_ms2sysv,
31670         call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31671         (struct machine_frame_state): New fields sp_realigned and
31672         sp_realigned_offset.
31673         * config/i386/i386.c
31674         (enum xlogue_stub): New enum.
31675         (enum xlogue_stub_sets): New enum.
31676         (class xlogue_layout): New class.
31677         (struct ix86_frame): New fields stack_realign_allocate_offset,
31678         stack_realign_offset and outlined_save_offset.  Modify comments to
31679         detail stack layout when using out-of-line stubs.
31680         (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31681         (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31682         -mcall-ms2sysv-xlogues.
31683         (stub_managed_regs): New static variable.
31684         (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31685         registers managed by out-of-line stub.
31686         (disable_call_ms2sysv_xlogues): New function.
31687         (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31688         m->call_ms2sysv when appropriate and compute frame layout for
31689         out-of-line stubs.
31690         (sp_valid_at, fp_valid_at): New inline functions.
31691         (choose_basereg): New function.
31692         (choose_baseaddr): Add align parameter, use choose_basereg and modify
31693         all callers.
31694         (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31695         Use align parameter of choose_baseaddr to generated aligned SSE movs
31696         when possible.
31697         (pro_epilogue_adjust_stack): Modify to track
31698         machine_frame_state::sp_realigned.
31699         (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31700         (ix86_nsaved_sseregs): Likewise.
31701         (ix86_emit_save_regs): Likewise.
31702         (ix86_emit_save_regs_using_mov): Likewise.
31703         (ix86_emit_save_sse_regs_using_mov): Likewise.
31704         (get_scratch_register_on_entry): Likewise.
31705         (gen_frame_set): New function.
31706         (gen_frame_load): Likewise.
31707         (gen_frame_store): Likewise.
31708         (emit_outlined_ms2sysv_save): Likewise.
31709         (emit_outlined_ms2sysv_restore): Likewise.
31710         (ix86_expand_prologue): Modify stack re-alignment code and call
31711         emit_outlined_ms2sysv_save when appropriate.
31712         (ix86_emit_leave): Clear machine_frame_state::sp_realigned.  Add
31713         parameter rtx_insn *insn, which allows the function to be used to only
31714         generate the notes.
31715         (ix86_expand_epilogue): Modify validity checks of frame and stack
31716         pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31717         (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31718         * config/i386/predicates.md
31719         (save_multiple): New predicate.
31720         (restore_multiple): Likewise.
31721         * config/i386/sse.md
31722         (save_multiple<mode>): New pattern.
31723         (save_multiple_realign<mode>): Likewise.
31724         (restore_multiple<mode>): Likewise.
31725         (restore_multiple_and_return<mode>): Likewise.
31726         (restore_multiple_leave_return<mode>): Likewise.
31727         * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31729 2017-05-14  Julia Koval  <julia.koval@intel.com>
31731         * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31732         * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31733         (__builtin_ia32_xsetbv): New builtins.
31734         * config/i386/i386.c (ix86_expand_special_args_builtin):
31735         Process new types.
31736         (ix86_expand_builtin): Special expand for new intrinsics.
31737         * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31738         (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31739         * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31741 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31743         * cfganal.c (inverted_post_order_compute): Change argument type
31744         to vec *.
31745         * cfganal.h (inverted_post_order_compute): Adjust prototype.
31746         * df-core.c (rest_of_handle_df_initialize): Adjust.
31747         (rest_of_handle_df_finish): Likewise.
31748         (df_analyze_1): Likewise.
31749         (df_analyze): Likewise.
31750         (loop_inverted_post_order_compute): Change argument to be a vec *.
31751         (df_analyze_loop): Adjust.
31752         (df_get_n_blocks): Likewise.
31753         (df_get_postorder): Likewise.
31754         * df.h (struct df_d): Change field to be a vec.
31755         * lcm.c (compute_laterin): Adjust.
31756         (compute_available): Likewise.
31757         * lra-lives.c (lra_create_live_ranges_1): Likewise.
31758         * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31759         * tree-ssa-pre.c (compute_antic): Likewise.
31761 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31763         * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31764         (depth_first_search::depth_first_search): Change structure init
31765         function to this constructor.
31766         (depth_first_search::add_bb): Rename function to this member.
31767         (depth_first_search::execute): Likewise.
31768         (flow_dfs_compute_reverse_finish): Adjust.
31770 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31772         * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31773         (longest_simple_path): Likewise.
31774         * shrink-wrap.c (spread_components): Likewise.
31775         (disqualify_problematic_components): Likewise.
31776         (emit_common_heads_for_components): Likewise.
31777         (emit_common_tails_for_components): Likewise.
31778         (insert_prologue_epilogue_for_components): Likewise.
31780 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31782         * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31783         auto_sbitmap.
31785 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31787         * df-core.c (df_set_blocks): Start using auto_bitmap.
31788         (df_compact_blocks): Likewise.
31789         * df-problems.c (df_rd_confluence_n): Likewise.
31790         * df-scan.c (df_insn_rescan_all): Likewise.
31791         (df_process_deferred_rescans): Likewise.
31792         (df_update_entry_block_defs): Likewise.
31793         (df_update_exit_block_uses): Likewise.
31794         (df_entry_block_bitmap_verify): Likewise.
31795         (df_exit_block_bitmap_verify): Likewise.
31796         (df_scan_verify): Likewise.
31797         * lra-constraints.c (lra_constraints): Likewise.
31798         (undo_optional_reloads): Likewise.
31799         (lra_undo_inheritance): Likewise.
31800         * lra-remat.c (calculate_gen_cands): Likewise.
31801         (do_remat): Likewise.
31802         * lra-spills.c (assign_spill_hard_regs): Likewise.
31803         (spill_pseudos): Likewise.
31804         * tree-ssa-pre.c (bitmap_set_and): Likewise.
31805         (bitmap_set_subtract_values): Likewise.
31807 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31809         * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31810         management with auto_bitmap.
31811         (fix_inter_tick): Likewise.
31812         (fix_recovery_deps): Likewise.
31813         * ira.c (add_store_equivs): Likewise.
31814         (find_moveable_pseudos): Likewise.
31815         (split_live_ranges_for_shrink_wrap): Likewise.
31816         * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31817         (rtx_reuse_manager::seen_def_p): Likewise.
31818         (rtx_reuse_manager::set_seen_def): Likewise.
31819         * print-rtl.h (class rtx_reuse_manager): Likewise.
31821 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31823         * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31824         lifetime.
31825         (migrate_btr_def): Likewise.
31826         * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31827         * df-core.c (loop_post_order_compute): Likewise.
31828         (loop_inverted_post_order_compute): Likewise.
31829         * hsa-common.h: Likewise.
31830         * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31831         * init-regs.c (initialize_uninitialized_regs): Likewise.
31832         * ipa-inline.c (resolve_noninline_speculation): Likewise.
31833         (inline_small_functions): Likewise.
31834         * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31835         * ira.c (combine_and_move_insns): Likewise.
31836         (build_insn_chain): Likewise.
31837         * loop-invariant.c (find_invariants): Likewise.
31838         * lower-subreg.c (propagate_pseudo_copies): Likewise.
31839         * predict.c (tree_predict_by_opcode): Likewise.
31840         (predict_paths_leading_to): Likewise.
31841         (predict_paths_leading_to_edge): Likewise.
31842         (estimate_loops_at_level): Likewise.
31843         (estimate_loops): Likewise.
31844         * shrink-wrap.c (try_shrink_wrapping): Likewise.
31845         (spread_components): Likewise.
31846         * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31847         * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31848         * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31849         * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31850         * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31851         * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31852         * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31853         * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31854         (mark_threaded_blocks): Likewise.
31855         (thread_through_all_blocks): Likewise.
31856         * tree-ssa.c (verify_ssa): Likewise.
31857         (execute_update_addresses_taken): Likewise.
31858         * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31860 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31862         * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31863         auto_vec.
31864         (post_order_compute): Likewise.
31865         (inverted_post_order_compute): Likewise.
31866         (pre_and_rev_post_order_compute_fn): Likewise.
31868 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31870         * genrecog.c (int_set::int_set): Explicitly construct our
31871         auto_vec base class.
31872         * vec.h (auto_vec::auto_vec): New constructor.
31874 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31876         * bitmap.h (class auto_bitmap): New constructor taking
31877         bitmap_obstack * argument.
31879 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31881         * bitmap.h (class auto_bitmap): Change type of m_bits to
31882         bitmap_head, and adjust ctor / dtor and member operators.
31884 2017-05-13  Uros Bizjak  <ubizjak@gmail.com>
31886         * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31887         when returned register mode doesn't match original mode.
31889 2017-05-12  Jeff Law  <law@redhat.com>
31890             Jakub Jelinek  <jakub@redhat.com>
31892         * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31893         we look for cc setter after the compare-elim changes.
31894         * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31895         within the vector to match what compare-elim now expects.
31896         (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31897         (xorsi3_flags, one_cmplsi2_flags): Likewise.
31899         * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
31900         after the compare-elim changes.
31901         * config/rx/rx.md (abssi2_flags): Fix order of patterns within
31902         the vector to match what compare-elim now expects.
31903         (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
31904         (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
31905         (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
31906         (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
31907         (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
31909         * config/visium/visium.c (single_set_and_flags): Fix where
31910         we look for cc setter after the compare-elim changes.
31911         * config/visium/visium.md (flags_subst_logic): Fix order of patterns
31912         with the vector to match what compare-elim now expects.
31913         (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
31914         (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
31915         (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
31916         (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
31917         (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
31918         (neg<mode>2_insn_set_overflow): Likewise.
31920 2017-05-12  Jim Wilson  <jim.wilson@linaro.org>
31922         PR middle-end/79794
31923         * expmed.c (extract_bit_field_1): Add alt_rtl argument.  Before
31924         maybe_expand_insn call, set ops[0].target.  If still set after call,
31925         set alt_rtl.  Add extra arg to recursive calls.
31926         (extract_bit_field): Add alt_rtl argument.  Pass to
31927         extract_bit_field.
31928         * expmed.h (extract_bit_field): Fix prototype.
31929         * expr.c (emit_group_load_1, copy_blkmode_from_reg)
31930         (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
31931         to extract_bit_field_calls.
31932         (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
31933         Pass alt_rtl to extract_bit_field calls.
31934         * calls.c (store_unaligned_arguments_into_psuedos)
31935         load_register_parameters): Pass extra NULL to extract_bit_field calls.
31936         * optabs.c (maybe_legitimize_operand): Clear op->target when call
31937         gen_reg_rtx.
31938         * optabs.h (struct expand_operand): Add target bitfield.
31940 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31942         * compare-elim.c (try_eliminate_compare): Canonicalize
31943         operation with embedded compare to
31944         [(set (reg:CCM) (compare:CCM (operation) (immediate)))
31945          (set (reg) (operation)].
31947         * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
31949 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31951         PR target/80723
31952         * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
31953         cost of adding a carry flag for ADC instruction.
31954         [case MINUS]: Ignore the cost of subtracting a carry flag
31955         for SBB instruction.
31957 2017-05-12  Steven Munroe  <munroesj@gcc.gnu.org>
31959         * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
31960         and x86intrin.h
31961         * config/rs6000/bmiintrin.h: New file.
31962         * config/rs6000/bmi2intrin.h: New file.
31963         * config/rs6000/x86intrin.h: New file.
31965 2017-05-12  Jeff Law  <law@redhat.com>
31967         * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
31968         markers.
31970 2017-05-12  Peter Bergner  <bergner@vnet.ibm.com>
31972         PR middle-end/80707
31973         * tree-cfg.c: Remove cfg edges of unreachable case statements.
31975 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31977         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31978         early expansion of vector divide builtins.
31979         (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
31980         builtins identified as having unsigned arguments.
31982 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31984         * config/rs6000/rs6000.c (gimple-fold.h): New #include.
31985         (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
31986         expansion of vector logical operations (and, andc, or, xor,
31987         nor, orc, nand).
31989 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31991         * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
31992         * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
31994 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31996         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31997         early GIMPLE expansion of vector multiplies.
31999 2017-05-12  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32001         * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
32002         TARGET_HAVE_MOVT conditional.
32003         (movt splitter): Likewise.
32005 2017-05-12  Richard Biener  <rguenther@suse.de>
32007         * tree-ssa-sccvn.h (has_VN_INFO): Declare.
32008         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32009         Fold all stmts not inplace.
32011 2017-05-12  Richard Biener  <rguenther@suse.de>
32013         PR tree-optimization/80713
32014         * tree-ssa-pre.c (remove_dead_inserted_code): Clear
32015         inserted_exprs bit for not removed stmts.
32017 2017-05-12  Thomas Schwinge  <thomas@codesourcery.com>
32019         PR middle-end/69921
32020         * tree-parloops.c (create_parallel_loop): Set "oacc kernels
32021         parallelized" attribute for parallelized OpenACC kernels.
32022         * omp-offload.c (execute_oacc_device_lower): Use it.
32024         * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
32025         Set "oacc kernels" attribute.
32026         * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
32027         parameter.  Adjust all users.
32028         (oacc_fn_attrib_kernels_p): Remove function.
32029         * omp-offload.c (execute_oacc_device_lower): Look for "oacc
32030         kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
32031         * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
32032         * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
32033         assert "oacc kernels" attribute is set.
32035 2017-05-11  Carl Love  <cel@us.ibm.com>
32037         * config/rs6000/rs6000-c: Add support for built-in functions
32038         vector unsigned char vec_popcnt (vector signed char)
32039         vector unsigned char vec_popcnt (vector unsigned char)
32040         vector unsigned short vec_popcnt (vector signed short)
32041         vector unsigned short vec_popcnt (vector unsigned short)
32042         vector unsigned int vec_popcnt (vector signed int)
32043         vector unsigned int vec_popcnt (vector unsigned int)
32044         vector unsigned long long vec_popcnt (vector signed long long)
32045         vector unsigned long long vec_popcnt (vector unsigned long long)
32046         vector signed long long vec_slo (vector signed long long,
32047                                          vector signed char)
32048         vector signed long long vec_slo (vector signed long long,
32049                                          vector unsigned char)
32050         vector unsigned long long vec_slo (vector unsigned long long,
32051                                            vector signed char)
32052         vector unsigned long long vec_slo (vector unsigned long long,
32053                                            vector unsigned char)
32054         * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
32055         VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
32056         * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
32057         vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
32058         * doc/extend.texi: Update the built-in documentation file for the
32059         new built-in functions.
32061 2017-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
32063         * attribs.h (sorted_attr_string): Move machine independent
32064         functions for target clone support from the i386 port to common
32065         code.  Rename ix86_function_versions to common_function_versions.
32066         Rename make_name to make_unique_name.
32067         (common_function_versions): Likewise.
32068         (make_unique_name): Likewise.
32069         (make_dispatcher_decl): Likewise.
32070         (is_function_default_version): Likewise.
32071         * attribs.c (attr_strcmp): Likewise.
32072         (sorted_attr_string): Likewise.
32073         (common_function_versions): Likewise.
32074         (make_unique_name): Likewise.
32075         (make_dispatcher_decl): Likewise.
32076         (is_function_default_version): Likewise.
32077         * config/i386/i386.c (attr_strcmp): Likewise.
32078         (sorted_attr_string): Likewise.
32079         (ix86_function_versions): Likewise.
32080         (make_name): Likewise.
32081         (make_dispatcher_decl): Likewise.
32082         (is_function_default_version): Likewise.
32083         (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
32085 2017-05-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32087         PR target/80695
32088         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
32089         Account for direct move costs for vec_construct of integer
32090         vectors.
32092 2017-05-11  Uros Bizjak  <ubizjak@gmail.com>
32094         PR target/80706
32095         * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
32096         (UNSPEC_STX_ATOMIC): Ditto.
32097         (loaddi_via_sse): New insn.
32098         (storedi_via_sse): Ditto.
32099         (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
32100         Update corresponding peephole2 patterns.
32101         (atomic_storedi_fpu): Ditto.
32103 2017-05-11  Julia Koval  <julia.koval@intel.com>
32105         * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
32106         (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
32107         New intrinsics.
32108         * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
32109         (__builtin_ia32_rsqrt14ss_mask): New builtins.
32110         * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
32112 2017-05-11  Nathan Sidwell  <nathan@acm.org>
32114         * graphite-poly.c: Include dumpfile.h.
32116         * dumpfle.h (dump_function): Declare here ...
32117         * tree-dump.h (dump_function): ... not here.
32118         * dumpfile.c: #include tree-cfg.h.
32119         (dump_function): Move here from ...
32120         * tree-dump.c (dump_function): ... here.
32121         * gimplify.c: #include splay-tree.h, not tree-dump.h.
32122         * graphite-poly.c: Don't include tree-dump.h.
32123         * cgraphclones.c: Include dumpfile.h not tree-dump.h.
32124         * print-tree.c: Likewise.
32125         * stor-layout.c: Likewise.
32126         * tree-nested.c: Likewise.
32128         * dumpfile.c (dump_start): Use TDF_FLAGS.
32129         (dump_enable_all): Fix TDF_KIND check thinko.
32131 2017-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32133         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
32134         array entries to represent two legal parameterizations of the
32135         overloaded __builtin_cmpb function, as represented by the
32136         P6_OV_BUILTIN_CMPB constant.
32137         (altivec_resolve_overloaded_builtin): Add special case handling
32138         for the __builtin_cmpb function, as represented by the
32139         P6_OV_BUILTIN_CMPB constant.
32140         * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
32141         (BU_P6_64BIT_2): New macro.
32142         (BU_P6_OVERLOAD_2): New macro
32143         (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
32144         (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
32145         (CMPB): Add overload support to represent both 32-bit and 64-bit
32146         compare-bytes function.
32147         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
32148         support for TARGET_CMPB.
32149         * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
32150         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
32151         documentation of the __builtin_cmpb overloaded built-in function.
32153 2017-05-11  Richard Biener  <rguenther@suse.de>
32155         PR tree-optimization/80705
32156         * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
32157         bases are not vectorizable.
32159 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32161         * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
32162         when counting register pressure.
32164 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32166         * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
32167         (iv_ca_more_deps): Renamed to ...
32168         (iv_ca_compare_deps): ... this.
32169         (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
32171 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32173         * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
32174         to ...
32175         (determine_group_iv_costs): ... here.
32176         (find_inv_vars_cb): Record inv var if it's not recorded before.
32178 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32180         * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
32181         (get_shiftadd_cost): Ditto.
32183 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32185         * tree-ssa-address.c: Include header file.
32186         (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
32187         address.
32188         (add_to_parts): Refactor.
32189         (addr_to_parts): New parameter.  Update use of move_hint_to_base.
32190         (create_mem_ref): Update use of addr_to_parts.  Re-associate addr
32191         in new order.
32193 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32195         PR tree-optimization/53090
32196         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
32197         COMP_IV_EXPR_2.
32198         (extract_cond_operands): Detect condition with IV on both sides
32199         and return COMP_IV_EXPR_2.
32200         (find_interesting_uses_cond): Add iv_use for both IVs in condition.
32201         (rewrite_use_compare): Simplify by removing call to function
32202         extract_cond_operands.
32204 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32206         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
32207         (extract_cond_operands): Detect condition comparing against non-
32208         invariant bound and return appropriate enum value.
32209         (find_interesting_uses_cond): Update use of extract_cond_operands.
32210         Handle its return value accordingly.
32211         (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
32213 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32215         * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
32216         nonlinear iv_use computation in loop invariant sensitive way.
32218 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32220         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
32221         (find_iv_candidates): Call relate_compare_use_with_all_cands.
32223 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32225         * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32226         (dump_cand): Support iv_cand.inv_exprs.
32227         (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32228         for candidates.
32229         (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32230         iv_cand.inv_exprs.
32232 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32234         * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32235         from ...
32236         * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32237         as local function.  Include necessary header files.
32238         * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32240 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32242         * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32244 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32246         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32247         operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32248         RSHIFT_EXPR and BIT_NOT_EXPR.
32250 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32252         * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32253         (adjust_setup_cost): New parameter supporting round up adjustment.
32254         (struct address_cost_data): Delete.
32255         (force_expr_to_var_cost): Don't bound cost with spill_cost.
32256         (split_address_cost, ptr_difference_cost): Delete.
32257         (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32258         (struct ainc_cost_data): New struct.
32259         (get_address_cost_ainc): New function.
32260         (get_address_cost, get_computation_cost): Reimplement.
32261         (determine_group_iv_cost_address): Record inv_expr for all uses of
32262         a group.
32263         (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32264         (iv_ca_has_deps): Reimplemented to ...
32265         (iv_ca_more_deps): ... this.  Check if NEW_CP introduces more deps
32266         than OLD_CP.
32267         (iv_ca_extend): Call iv_ca_more_deps.
32269 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32271         * tree-ssa-address.c (struct mem_address): Move to header file.
32272         (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32273         * tree-ssa-address.h (struct mem_address): Move from C file.
32274         (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32276 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32278         * tree-affine.h (aff_combination_type): New interface.
32279         (aff_combination_zero_p): Remove static.
32280         (aff_combination_const_p): New interface.
32281         (aff_combination_singleton_var_p): New interfaces.
32283 2017-05-11  Richard Biener  <rguenther@suse.de>
32285         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32286         Skip unreachable blocks and destinations.
32287         (eliminate): Move stmt removal and fixup ...
32288         (fini_eliminate): ... here.  Skip inserted exprs.
32289         (pass_pre::execute): Move fini_pre after fini_eliminate.
32290         * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32291         (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32292         PRE to get rid of dead code that has invalid SSA form and
32293         split critical edges again.
32295 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32297         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32299 2017-05-11  Richard Biener  <rguenther@suse.de>
32301         * passes.c (execute_function_todo): Verify loops if they are
32302         said to be up-to-date.
32303         * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32304         * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32306 2017-05-10  John David Anglin  <danglin@gcc.gnu.org>
32308         PR target/80090
32309         * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32310         handle calling assemble_external ourself.
32312         PR target/79027
32313         * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32314         modes with zero size.  Enhance comment.
32316 2017-05-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32318         * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32319         built-ins for vec_xl and vec_xst with short and char pointer
32320         arguments.
32322 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32324         * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32325         (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32326         (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32327         (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32328         (_mm_maskz_min_round_ss): New intrinsics.
32329         * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32330         (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32331         * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32332         (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32333         (__builtin_ia32_minss_mask_round): New builtins.
32334         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32335         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32336         * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32337         Rename to ...
32338         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32339         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32340         Change to ...
32341         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32342         ... this.
32344 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32346         * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32347         (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32348         (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32349         (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32350         (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32351         (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32352         (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32353         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32354         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32355         * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32356         (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32357         (__builtin_ia32_mulss_mask_round): New builtins.
32358         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32359         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32360         * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32361         Rename to ...
32362         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32363         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32364         Change to ...
32365         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32366         ... this.
32368 2017-05-10  Julia Koval  <julia.koval@intel.com>
32370         * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32371         (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32372         (_mm256_setr_m128i): New intrinsics.
32374 2017-05-10  Julia Koval  <julia.koval@intel.com>
32376         * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32377         (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32378         (_mm_maskz_rcp14_ss): New intrinsics.
32379         * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32380         (__builtin_ia32_rcp14ss_mask): New builtins.
32381         * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32383 2017-05-10  Peter Bergner  <bergner@vnet.ibm.com>
32385         PR tree-optimization/51513
32386         * tree-cfg.c (gimple_seq_unreachable_p): New function.
32387         (assert_unreachable_fallthru_edge_p): Use it.
32388         (group_case_labels_stmt): Likewise.
32389         * tree-cfg.h: Prototype it.
32390         * stmt.c: Include cfghooks.h and tree-cfg.h.
32391         (emit_case_dispatch_table) <gap_label>: New local variable.
32392         Use it to fill dispatch table gaps.
32393         Test for default_label before updating probabilities.
32394         (expand_case) <default_label>: Remove unneeded initialization.
32395         Test for unreachable default case statement and remove its edge.
32396         Set default_label accordingly.
32397         * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32399 2017-05-10  Carl Love  <cel@us.ibm.com>
32401         * config/rs6000/rs6000-c: Add support for built-in functions
32402         vector signed char      vec_neg (vector signed char)
32403         vector signed short int vec_neg (vector short int)
32404         vector signed int       vec_neg (vector signed int)
32405         vector signed long long vec_neg (vector signed long long)
32406         vector float            vec_neg (vector float)
32407         vector double           vec_neg (vector double)
32408         * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32409         overload.
32410         * config/rs6000/altivec.h: Add define for vec_neg
32411         * doc/extend.texi: Update the built-in documentation for the
32412         new built-in functions.
32414 2017-05-10  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32416         PR tree-optimization/77644
32417         * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32419 2017-05-10  Nathan Sidwell  <nathan@acm.org>
32421         * dumpfile.h (TDI_lang_all): New.
32422         (TDF_KIND): New. Renumber others
32423         (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32424         than bits.
32425         * dumpfile.c (dump_files): Mark language dumps as TDF_LANG.  add
32426         lang-all.
32427         (get_dump_file_name): Adjust suffix generation.
32428         (dump_enable_all): Use TDF_KIND.
32429         * doc/invoke.texi (-fdump-lang-all): Document.
32431         * dumpfile.h: Tabify.
32433 2017-05-10  Wilco Dijkstra  <wdijkstr@arm.com>
32435         PR target/80671
32436         * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32437         Move member access before delete.
32439 2017-05-10  Alexandre Oliva  <aoliva@redhat.com>
32441         * tree-inline.c (expand_call_inline): Split block at stmt
32442         before the call.
32444 2017-05-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
32446         PR target/68163
32447         * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32448         are now unused after splitting mov{sf,sd}_hardfloat.
32449         (f32_lr2): Likewise.
32450         (f32_lm): Likewise.
32451         (f32_lm2): Likewise.
32452         (f32_li): Likewise.
32453         (f32_li2): Likewise.
32454         (f32_lv): Likewise.
32455         (f32_sr): Likewise.
32456         (f32_sr2): Likewise.
32457         (f32_sm): Likewise.
32458         (f32_sm2): Likewise.
32459         (f32_si): Likewise.
32460         (f32_si2): Likewise.
32461         (f32_sv): Likewise.
32462         (f32_dm): Likewise.
32463         (f32_vsx): Likewise.
32464         (f32_av): Likewise.
32465         (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32466         For movsf, order stores so the VSX stores occur before the GPR
32467         store which encourages the register allocator to use a traditional
32468         FPR instead of a GPR.  For movsd, order the stores so that the GPR
32469         store comes before the VSX stores to allow the power6 to work.
32470         This is due to the power6 not having a 32-bit integer store
32471         instruction from a FPR.
32472         (movsf_hardfloat): Likewise.
32473         (movsd_hardfloat): Likewise.
32475 2017-05-09  Martin Sebor  <msebor@redhat.com>
32477         PR translation/80280
32478         * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32479         added in r247778.
32481         PR translation/80280
32482         * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32483         data member added in r247778.
32484         (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32486 2017-05-09  Nathan Sidwell  <nathan@acm.org>
32488         * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32490         * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32491         typedefs.
32493 2017-05-09  Marek Polacek  <polacek@redhat.com>
32495         * doc/invoke.texi: Fix typo.
32497 2017-05-09  Richard Biener  <rguenther@suse.de>
32499         * tree-vrp.c (vrp_val_is_max): Adjust comment.
32500         (vrp_val_is_min): Likewise.
32501         (set_value_range_to_value): Likewise.
32502         (set_value_range_to_nonnegative): Likewise.
32503         (gimple_assign_nonzero_p): Likewise.
32504         (gimple_stmt_nonzero_p): Likewise.
32505         (vrp_int_const_binop): Likewise.  Remove unreachable case.
32506         (adjust_range_with_scev): Adjust comments.
32507         (compare_range_with_value): Likewise.
32508         (extract_range_from_phi_node): Likewise.
32509         (test_for_singularity): Likewise.
32511 2017-05-09  Richard Biener  <rguenther@suse.de>
32513         * tree-vrp.c (get_single_symbol): Add assert that we don't
32514         get overflowed constants as invariant part.
32515         (compare_values_warnv): Add comment before the TREE_NO_WARNING
32516         checks.  Use wi::cmp instead of recursing for integer constants.
32517         (compare_values): Just ignore whether we assumed undefined
32518         overflow instead of failing the compare.
32519         (extract_range_for_var_from_comparison_expr): Add comment before the
32520         TREE_NO_WARNING sets.
32521         (test_for_singularity): Likewise.
32522         (extract_range_from_comparison): Do not disable optimization
32523         when we assumed undefined overflow.
32524         (extract_range_basic): Remove init of unused var.
32526 2017-05-09  Richard Biener  <rguenther@suse.de>
32528         * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32529         (extract_range_from_multiplicative_op_1): Adjust.
32530         (extract_range_from_binary_expr_1): Use int_const_binop.
32532 2017-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32534         PR target/80101
32535         * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32536         rs6000_store_data_bypass_p in seven define_bypass directives and
32537         in several comments.
32538         * config/rs6000/rs6000-protos.h: Add prototype for
32539         rs6000_store_data_bypass_p function.
32540         * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32541         function implements slightly different (rs6000-specific) semantics
32542         than store_data_bypass_p, returning false rather than aborting
32543         with assertion error when arguments do not satisfy the
32544         requirements of store data bypass.
32545         (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32546         rs6000_store_data_bypass_p.
32548 2017-05-08  Max Filippov  <jcmvbkbc@gmail.com>
32550         * config/xtensa/xtensa-protos.h
32551         (xtensa_initial_elimination_offset): New declaration.
32552         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32553         New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32554         macro definition, add case for FRAME_POINTER_REGNUM when
32555         FRAME_GROWS_DOWNWARD.
32556         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32557         (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32558         xtensa_initial_elimination_offset.
32560 2017-05-08  Nathan Sidwell  <nathan@acm.org>
32562         * doc/invoke.texi: Alphabetize -fdump options.
32564 2017-05-08  Martin Sebor  <msebor@redhat.com>
32566         PR translation/80280
32567         * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32569 2017-05-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
32571         * target.def (compute_frame_layout): New optional target hook.
32572         * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32573         * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32574         * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32575         target hook.
32576         * reload1.c (verify_initial_elim_offsets): Likewise.
32577         * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32578         (use_simple_return_p): Call arm_compute_frame_layout if needed.
32579         (arm_get_frame_offsets): Split up into this ...
32580         (arm_compute_frame_layout): ... and this function.
32582 2017-05-08  Richard Sandiford  <richard.sandiford@arm.com>
32584         * config/aarch64/constraints.md (Usa): New constraint.
32585         * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32587 2017-05-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32589         * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32590         with_multilib_list after it has been checked.
32592 2017-05-08  Richard Biener  <rguenther@suse.de>
32594         * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32595         (bitmap_set_subtract_values): Likewise.
32597 2017-05-08  Richard Biener  <rguenther@suse.de>
32599         * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32600         (gimple_assign_nonzero): ... this and remove strict_overflow_p
32601         argument.
32602         (gimple_stmt_nonzero_warnv_p): Rename to ...
32603         (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32604         argument.
32605         (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32606         (extract_range_basic): Adjust, do not disable propagation on
32607         strict overflow sensitive simplification.
32608         (vrp_visit_cond_stmt): Likewise.
32610 2017-05-05  Jan Hubicka  <hubicka@ucw.cz>
32612         * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32613         body size unconditionally.
32615 2017-05-07  Jeff Law  <law@redhat.com>
32617         Revert:
32618         2017-05-06  Jeff Law  <law@redhat.com>
32619         PR tree-optimization/78496
32620         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32621         code.
32623         PR tree-optimization/78496
32624         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32625         (simplify_stmt_using_ranges): Call it.
32626         (vrp_dom_walker::before_dom_children): Extract equivalences
32627         from an ASSERT_EXPR with an equality comparison against a
32628         constant.
32630 2017-05-06  Jeff Law  <law@redhat.com>
32632         PR tree-optimization/78496
32633         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32634         code.
32636         PR tree-optimization/78496
32637         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32638         (simplify_stmt_using_ranges): Call it.
32639         (vrp_dom_walker::before_dom_children): Extract equivalences
32640         from an ASSERT_EXPR with an equality comparison against a
32641         constant.
32643 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32645         * lra-constraints.c (lra_copy_reg_equiv): New function.
32646         (split_reg): Use it to copy equivalence information from the
32647         original register to the spill register.
32649 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32651         PR rtl-optimization/75964
32652         * simplify-rtx.c (simplify_const_relational_operation): Remove
32653         invalid handling of comparisons of integer ABS.
32655 2017-05-06  Uros Bizjak  <ubizjak@gmail.com>
32657         * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32658         initialize to zero.
32659         (init_regs): Remove declaration.
32660         (function_arg_advance_32): Initialize error_p as boolean variable.
32662 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32664         * store-motion.c (remove_reachable_equiv_notes): Reformat long
32665         lines.  Use for (;;).
32667 2017-05-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32669         * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32670         (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32671         (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32672         (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32673         VF=2 that require versioning.
32675 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32677         * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32678         int.
32680 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32682         * diagnostic.h (diagnostic_override_option_index): Convert from
32683         macro to inline function.
32685 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32687         * diagnostic.c (last_module_changed_p): New function.
32688         (set_last_module): New function.
32689         (diagnostic_report_current_module): Convert macro usage to
32690         the above functions.
32691         * diagnostic.h (diagnostic_context::last_module): Strengthen
32692         from const line_map * to const line_map_ordinary *.
32693         (diagnostic_last_module_changed): Delete macro.
32694         (diagnostic_set_last_module): Delete macro.
32696 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32698         * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32699         with diagnostic_report_diagnostic.
32700         (diagnostic_n_impl_richloc): Likewise.
32701         * diagnostic.h (report_diagnostic): Delete macro.
32702         * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32703         with diagnostic_report_diagnostic.
32704         * substring-locations.c (format_warning_va): Likewise.
32706 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32708         * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32709         save/restor of format_spec.  Move option-printing code to...
32710         (print_option_information): ...this new function, and
32711         reimplement by simply printing to the pretty_printer,
32712         rather than appending to the format string.
32714 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32716         * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32717         handling logic into...
32718         (update_effective_level_from_pragmas): ...this new function.
32720 2017-05-04  Andrew Waterman  <andrew@sifive.com>
32722         * config/riscv/riscv.opt (mstrict-align): New option.
32723         * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it.  Update comment.
32724         (SLOW_UNALIGNED_ACCESS): Define.
32725         (riscv_slow_unaligned_access): Declare.
32726         * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32727         field.
32728         (riscv_slow_unaligned_access): New variable.
32729         (rocket_tune_info): Set slow_unaligned_access to true.
32730         (optimize_size_tune_info): Set slow_unaligned_access to false.
32731         (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32732         (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32733         (riscv_option_override): Set riscv_slow_unaligned_access.
32734         * doc/invoke.texi: Add -mstrict-align to RISC-V.
32736 2017-05-04  Kito Cheng  <kito.cheng@gmail.com>
32738         * config/riscv/riscv.md: Unify indentation.
32740 2017-05-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
32742         PR target/79038
32743         PR target/79202
32744         PR target/79203
32745         * config/rs6000/rs6000.md (u code attribute): Add FIX and
32746         UNSIGNED_FIX.
32747         (extendsi<mode>2): Add support for doing sign extension via
32748         VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32749         don't have ISA 3.0 instructions.
32750         (extendsi<mode>2 splitter): Likewise.
32751         (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32752         generate the normal insns since SImode can now go in vector
32753         registers.  Disallow the special UNSPECs needed for previous
32754         machines to hide SImode being used.  Add new insns
32755         fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32756         (fix_trunc<mode>si2_stfiwx): Likewise.
32757         (fix_trunc<mode>si2_internal): Likewise.
32758         (fixuns_trunc<mode>si2): Likewise.
32759         (fixuns_trunc<mode>si2_stfiwx): Likewise.
32760         (fctiw<u>z_<mode>_smallint): Likewise.
32761         (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32762         of floating point to 32-bit integer from doing a direct move to
32763         the GPR registers to do a store.
32764         (fctiwz_<mode>): Break long line.
32766 2017-05-05  Bin Cheng  <bin.cheng@arm.com>
32768         * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32769         * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32770         (addr_list, addr_offset_valid_p): New.
32771         (split_address_groups): Check offset validity with above function.
32772         (gt-tree-ssa-loop-ivopts.h): Include header file.
32774 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32776         * config.gcc (arm*-*-*): Add missing 'fi'.
32778 2017-05-05  Steve Ellcey  <sellcey@cavium.com>
32780         * doc/invoke.texi (-fopt-info): Explicitly say order of options
32781         included in -fopt-info does not matter.
32782         * doc/optinfo.texi (-fopt-info): Fix description of default
32783         behavour. Explicitly say order of options included in -fopt-info
32784         does not matter.
32786 2017-05-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32788         * config.gcc: Allow combinations of aprofile and rmprofile values for
32789         --with-multilib-list.
32790         * config/arm/t-multilib: New file.
32791         * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32792         variables.  Remove setting of ISA and floating-point ABI in
32793         MULTILIB_OPTIONS and MULTILIB_DIRNAMES.  Set architecture and FPU in
32794         MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32795         and MULTILIB_DIRNAMES respectively.  Add comment to introduce all
32796         matches.  Add architecture matches for marvel-pj4 and generic-armv7-a
32797         CPU options.
32798         * config/arm/t-rmprofile: Likewise except for the matches changes.
32799         * doc/install.texi (--with-multilib-list): Document the combination of
32800         aprofile and rmprofile values and warn about pitfalls in doing that.
32802 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32804         * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32805         (movdi_aarch64): Likewise.
32807 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32809         PR tree-optimization/80632
32810         * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32811         field.
32812         (build_arrays): Initialize it for virtual phis.
32813         (fix_phi_nodes): Use it for virtual phis.
32815         PR tree-optimization/80558
32816         * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32817         [x, y] op z into [x op, y op z] for op & or | if conditions
32818         are met.
32820 2017-05-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
32821             Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32823         PR target/71607
32824         * config/arm/arm.md (use_literal_pool): Remove.
32825         (64-bit immediate split): No longer takes cost into consideration
32826         if arm_disable_literal_pool is enabled.
32827         * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32828         used when arm_disable_literal_pool is enabled.
32829         (arm_max_const_double_inline_cost): Remove use of
32830         arm_disable_literal_pool.
32831         (push_minipool_fix): Add assert.
32832         (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32833         * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32834         (no_literal_pool_sf_immediate): New.
32836 2017-05-05  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32838         PR tree-optimization/80613
32839         * tree-ssa-dce.c (propagate_necessity): Remove cases for
32840         BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32842 2017-05-05  Richard Biener  <rguenther@suse.de>
32844         * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32846 2017-05-05  Georg-Johann Lay  <avr@gjlay.de>
32848         * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32849         of this flag from insn conditions due to removal from r247495.
32851 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32853         * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32854         New function.
32855         (arm_early_store_addr_dep_ptr): Likewise.
32856         * config/arm/aarch-common-protos.h
32857         (arm_early_load_addr_dep_ptr): Add prototype.
32858         (arm_early_store_addr_dep_ptr): Likewise.
32859         * config/arm/cortex-a53.md: Add new bypasses.
32861 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32863         * tree.c (next_type_uid): Change type to unsigned.
32864         (type_hash_canon): Decrement back next_type_uid if
32865         freeing a type node with the highest TYPE_UID.  For INTEGER_TYPEs
32866         also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32867         if possible.
32869 2017-05-04  Martin Sebor  <msebor@redhat.com>
32871         * builtins.c: Fix a trivial typo in a comment.
32873         PR middle-end/79234
32874         * builtins.c (check_sizes): Adjust to handle reading past the end.
32875         Avoid printing excessive upper bound of ranges.  Use %E to print
32876         tree nodes instead of converting them to %wu.
32877         (expand_builtin_memchr): New function.
32878         (compute_dest_size): Rename...
32879         (compute_objsize): ...to this.
32880         (expand_builtin_memcpy): Adjust.
32881         (expand_builtin_mempcpy): Adjust.
32882         (expand_builtin_strcat): Adjust.
32883         (expand_builtin_strcpy): Adjust.
32884         (check_strncat_sizes): Adjust.
32885         (expand_builtin_strncat): Adjust.
32886         (expand_builtin_strncpy): Adjust and simplify.
32887         (expand_builtin_memset): Adjust.
32888         (expand_builtin_bzero): Adjust.
32889         (expand_builtin_memcmp): Adjust.
32890         (expand_builtin): Handle memcmp.
32891         (maybe_emit_chk_warning): Check strncat just once.
32893 2017-05-04  Martin Sebor  <msebor@redhat.com>
32895         PR preprocessor/79214
32896         PR middle-end/79222
32897         PR middle-end/79223
32898         * builtins.c (check_sizes): Add inlining context and issue
32899         warnings even when -Wno-system-headers is set.
32900         (check_strncat_sizes): Same.
32901         (expand_builtin_strncat): Same.
32902         (expand_builtin_memmove): New function.
32903         (expand_builtin_stpncpy): Same.
32904         (expand_builtin): Handle memmove and stpncpy.
32906 2017-05-04  Bin Cheng  <bin.cheng@arm.com>
32908         * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
32909         which is not used any more.
32911 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32913         * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
32915 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32917         * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
32918         (cortexa53_tunings): Likewise.
32919         (cortexa57_tunings): Likewise.
32920         (cortexa72_tunings): Likewise.
32921         (cortexa73_tunings): Likewise.
32923 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32925         * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
32926         Set loop alignment to 8.
32928 2017-05-04  Martin Sebor  <msebor@redhat.com>
32930         PR translation/80280
32931         * builtins.c (expand_builtin_object_size): Add missing quoting to
32932         %D and like directives.
32933         * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
32934         (hsa_type_for_tree_type): Same.
32935         (verify_function_arguments): Same.
32936         * symtab.c (symbol_table::change_decl_assembler_name): Same.
32937         * varasm.c (get_section): Same.
32938         (mark_weak): Same.
32940 2017-05-04  Martin Sebor  <msebor@redhat.com>
32942         PR translation/80280
32943         * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
32945 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32947         * config/aarch64/aarch64.c (generic_addrcost_table):
32948         Change HI/TI mode setting.
32950 2017-05-04  Martin Jambor  <mjambor@suse.cz>
32952         PR tree-optimization/80622
32953         * tree-sra.c (comes_initialized_p): New function.
32954         (build_accesses_from_assign): Only set write lazily when
32955         comes_initialized_p is false.
32956         (analyze_access_subtree): Use comes_initialized_p.
32957         (propagate_subaccesses_across_link): Assert !comes_initialized_p
32958         instead of testing for PARM_DECL.
32960 2017-05-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32962         * config/aarch64/aarch64.md (prefetch); Adjust predicate and
32963         constraint on operand 0 to allow more general addressing modes.
32964         Adjust output template.
32965         * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
32966         New function.
32967         * config/aarch64/aarch64-protos.h
32968         (aarch64_address_valid_for_prefetch_p): Declare prototype.
32969         * config/aarch64/constraints.md (Dp): New address constraint.
32970         * config/aarch64/predicates.md (aarch64_prefetch_operand): New
32971         predicate.
32973 2017-05-04  Jan Hubicka  <hubicka@ucw.cz>
32975         * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
32976         update use of estimate_ipcp_clone_size_and_time.
32977         (estimate_local_effects): Update use of
32978         estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
32979         * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
32980         * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
32981         Return nonspecialized time.
32983 2017-05-04  Richard Biener  <rguenther@suse.de>
32985         * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
32986         for the last VUSE which def dominates the PHI.  Directly call
32987         maybe_skip_until.
32988         (get_continuation_for_phi_1): Remove.
32990 2017-05-04  Richard Sandiford  <richard.sandiford@linaro.org>
32992         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
32993         to explain the use of truncating division.  Cap the number of
32994         iterations to the maximum given by nb_iterations_upper_bound,
32995         if defined.
32997 2017-05-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32999         * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
33000         * configure: Regenerate.
33001         * config.in: Regenerate.
33002         * config/i386/driver-mingw32.c: new file.
33003         * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
33004         * config.host: Link driver-mingw32.o on MinGW host.
33005         * doc/install.texi: Document new --enable-mingw-wildcard configure
33006         option.
33008 2017-05-04  Marek Polacek  <polacek@redhat.com>
33010         PR tree-optimization/80612
33011         * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
33013 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
33014             Andre Simoes Dias Vieira  <andre.simoesdiasvieira@arm.com>
33016         * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
33017         (movt splitter): Likewise.
33018         * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
33019         to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
33020         (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
33021         block for Thumb-1 with MOVT.
33022         (thumb2_legitimate_address_p): Move code block ...
33023         (can_avoid_literal_pool_for_label_p): ... into this new function.
33024         (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
33025         literal pool.
33026         (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
33027         * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
33028         "M-profile targets with the MOVT instruction".
33030 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
33032         * config/arm/arm-builtins.c (arm_init_builtins): Rename
33033         __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
33034         __builtin_arm_stfscr to __builtin_arm_set_fpscr.
33036 2017-05-04  Martin Liska  <mliska@suse.cz>
33038         * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
33039         variable cond_code.
33041 2017-05-04  Richard Biener  <rguenther@suse.de>
33043         * tree.c (array_at_struct_end_p): Handle arrays at struct
33044         end with flexarrays more conservatively.  Refactor and treat
33045         arrays of arrays or aggregates more strict.  Fix
33046         VIEW_CONVERT_EXPR handling.  Remove allow_compref argument.
33047         * tree.c (array_at_struct_end_p): Adjust prototype.
33048         * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
33049         * gimple-fold.c (get_range_strlen): Likewise.
33050         * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
33052 2017-05-04  Richard Biener  <rguenther@suse.de>
33054         PR tree-optimization/31130
33055         * tree-vrp.c (needs_overflow_infinity): Remove as always returning
33056         false.
33057         (supports_overflow_infinity): Likewise.
33058         (is_negative_overflow_infinity): Likewise.
33059         (is_positive_overflow_infinity): Likewise.
33060         (is_overflow_infinity): Likewise.
33061         (stmt_overflow_infinity): Likewise.
33062         (overflow_infinity_range_p): Likewise.
33063         (usable_range_p): Remove as always returning true.
33064         (make_overflow_infinity): Remove.
33065         (negative_overflow_infinity): Likewise.
33066         (positive_overflow_infinity): Likewise.
33067         (avoid_overflow_infinity): Likewise.
33068         (set_value_range): Adjust accordingly.
33069         (set_value_range_to_nonnegative): Likewise, remove now unused
33070         overflow_infinity arg.
33071         (vrp_operand_equal_p): Adjust.
33072         (update_value_range): Likewise.
33073         (range_int_cst_singleton_p): Likewise.
33074         (operand_less_p): Likewise.
33075         (compare_values_warnv): Likewise.
33076         (extract_range_for_var_from_comparison_expr): Likewise.
33077         (vrp_int_const_binop): Likewise.
33078         (zero_nonzero_bits_from_vr): Likewise.
33079         (extract_range_from_multiplicative_op_1): Likewise.
33080         (extract_range_from_binary_expr_1): Likewise.
33081         (extract_range_from_unary_expr): Likewise.
33082         (extract_range_from_comparison): Likewise.
33083         (extract_range_basic): Likewise.
33084         (adjust_range_with_scev): Likewise.
33085         (compare_ranges): Likewise.
33086         (compare_range_with_value): Likewise.
33087         (dump_value_range): Likewise.
33088         (test_for_singularity): Likewise, remove strict_overflow_p parameter
33089         never used.
33090         (simplify_cond_using_ranges): Adjust.
33092 2017-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
33094         * brig-builtins.def: Added a builtin for class_f64.
33095         * builtin-types.def: Added a builtin type needed by class_f64.
33097 2017-05-03  Jason Merrill  <jason@redhat.com>
33099         * timevar.def: Add TV_CONSTEXPR.
33101 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33103         * common.opt (fdiagnostics-parseable-fixits): Fix typo.
33105 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33107         * ipa-prop.c (ipa_update_after_lto_read): Removed.
33108         * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
33109         * ipa-cp.c (ipcp_propagate_stage): Do not call
33110         ipa_update_after_lto_read.
33111         * ipa-inline.c (ipa_inline): Likewise.
33113 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33115         * ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
33116         tag.  Added a default constructor and a destructor.
33117         (ipa_edge_args_sum_t): New class;
33118         (ipa_edge_args_sum): Declare.
33119         (ipa_edge_args_vector): Remove declaration.
33120         (IPA_EDGE_REF): Use ipa_edge_args_sum.
33121         (ipa_free_edge_args_substructures): Remove declaration.
33122         (ipa_check_create_edge_args): Use ipa_edge_args_sum.
33123         (ipa_edge_args_info_available_for_edge_p): Likewise.
33124         * ipa-prop.c (ipa_edge_args_vector): Removed.
33125         (edge_removal_hook_holder): Likewise.
33126         (edge_duplication_hook_holder): Likewise.
33127         (ipa_edge_args_sum): New variable.
33128         (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
33129         ipa_edge_args_vector.
33130         (ipa_free_edge_args_substructures): Likewise.
33131         (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
33132         ipa_edge_args_vector.
33133         (ipa_edge_removal_hook): Turned into method
33134         ipa_edge_args_sum_t::remove.
33135         (ipa_edge_duplication_hook): Turned into method
33136         ipa_edge_args_sum_t::duplicate.
33137         (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
33138         registering edge hooks.
33139         (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
33140         * ipa-inline-analysis.c (estimate_function_body_sizes): Test
33141         ipa_edge_args_sum instead of ipa_edge_args_vector.
33142         * ipa-profile.c (ipa_profile): Likewise.
33144 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33146         * symbol-summary.h (function_summary): New method exists.
33147         (function_summary::symtab_removal): Deallocate through release.
33148         (call_summary): New class.
33149         (gt_ggc_mx): New overload.
33150         (gt_pch_nx): Likewise.
33151         (gt_pch_nx): Likewise.
33153 2017-05-03  Jeff Law  <law@redhat.com>
33155         PR tree-optimization/78496
33156         * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
33157         from simplify_cond_using_ranges.  Split off code to walk
33158         backwards through casts into ...
33159         (simplify_cond_using_ranges_2): New function.
33160         (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
33161         (execute_vrp): After identifying jump threads, call
33162         simplify_cond_using_ranges_2.
33164 2017-05-03  Jan Hubicka  <hubicka@ucw.cz>
33166         PR bootstrap/80609
33167         * ipa-inline.h (inline_summary): Add ctor.
33168         (create_ggc): Do not use ggc_cleared_alloc.
33170 2017-05-03  Jeff Downs  <heydowns@somuchpressure.net>
33171             Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33173         * gcc.c (handle_braces): Support escaping in switch matching
33174         text.
33175         * doc/invoke.texi (Spec Files): Document it.
33176         Remove superfluous @code markup in items.
33178 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33180         * diagnostic-show-locus.c (struct column_range): New struct.
33181         (get_affected_columns): New function.
33182         (get_printed_columns): New function.
33183         (struct correction): New struct.
33184         (correction::ensure_capacity): New function.
33185         (correction::ensure_terminated): New function.
33186         (struct line_corrections): New struct.
33187         (line_corrections::~line_corrections): New dtor.
33188         (line_corrections::add_hint): New function.
33189         (layout::print_trailing_fixits): Reimplement in terms of the new
33190         classes.
33191         (selftest::test_overlapped_fixit_printing): New function.
33192         (selftest::diagnostic_show_locus_c_tests): Call it.
33194 2017-05-03  Nathan Sidwell  <nathan@acm.org>
33196         Canonicalize canonical type hashing
33197         * tree.h (type_hash_canon_hash): Declare.
33198         * tree.c (type_hash_list, attribute_hash_list): Move into
33199         type_hash_canon_hash.
33200         (build_type_attribute_qual_variant): Break out hash code calc into
33201         type_hash_canon_hash.
33202         (type_hash_canon_hash): New.  Generic type hash computation.
33203         (build_range_type_1, build_array_type_1, build_function_type,
33204         build_method_type_directly, build_offset_type, build_complex_type,
33205         make_vector_type): Call it.
33207 2017-05-03  Richard Biener  <rguenther@suse.de>
33209         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
33210         When all DRs have unknown misaligned do not always peel
33211         when there is a store but apply the same costing model as if
33212         there were only loads.
33214 2017-05-03  Richard Biener  <rguenther@suse.de>
33216         Revert
33217         PR tree-optimization/80492
33218         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33219         compare_base_decls returning dont-know properly.
33221 2017-05-03  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33223         * config/arm/iterators.md (CCSI): New mode iterator.
33224         (arch): New mode attribute.
33225         * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33226         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33227         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This.  Use CCSI
33228         code iterator for success result mode.
33229         * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33230         the corresponding new insn generators.
33232 2017-05-03  Bin Cheng  <bin.cheng@arm.com>
33234         Revert r247509
33235         2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33236         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33238 2017-05-03  Richard Sandiford  <richard.sandiford@linaro.org>
33240         * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33241         (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33242         (DDR_A): Wrap DDR argument in brackets.
33243         (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33244         (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33245         (DDR_REVERSED_P): Likewise.
33247 2017-05-03  Jakub Jelinek  <jakub@redhat.com>
33249         PR tree-optimization/79472
33250         * tree-switch-conversion.c (struct switch_conv_info): Add
33251         contiguous_range and default_case_nonstandard fields.
33252         (collect_switch_conv_info): Compute contiguous_range and
33253         default_case_nonstandard fields, don't clear final_bb if
33254         contiguous_range and only the default case doesn't have the required
33255         structure.
33256         (check_all_empty_except_final): Set default_case_nonstandard instead
33257         of failing if contiguous_range and the default case doesn't have empty
33258         block.
33259         (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33260         and only the default case doesn't have the required constants.  Skip
33261         virtual phis.
33262         (gather_default_values): Skip virtual phis.  Allow non-NULL CASE_LOW
33263         if default_case_nonstandard.
33264         (build_constructors): Build constant 1 just once.  Assert that default
33265         values aren't inserted in between cases if contiguous_range.  Skip
33266         virtual phis.
33267         (build_arrays): Skip virtual phis.
33268         (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33269         (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33270         Handle virtual phis.
33271         (gen_inbound_check): Handle default_case_nonstandard case.
33272         (process_switch): Adjust check_final_bb caller.  Call
33273         gather_default_values with the first non-default case instead of
33274         default case if default_case_nonstandard.
33276 2017-05-02  Nathan Sidwell  <nathan@acm.org>
33278         * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33279         check.  Fix formatting.
33281 2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
33283         * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33284         errors when comparing specialized and unspecialized times.
33286 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
33288         * diagnostic-show-locus.c
33289         (layout::should_print_annotation_line_p): Make private.
33290         (layout::print_annotation_line): Make private.
33291         (layout::annotation_line_showed_range_p): Make private.
33292         (layout::show_ruler): Make private.
33293         (layout::print_source_line): Make private.  Pass in line and
33294         line_width, rather than calling location_get_source_line.  Drop
33295         returned value.
33296         (layout::print_leading_fixits): New method.
33297         (layout::print_any_fixits): Rename to...
33298         (layout::print_trailing_fixits): ...this, and make private.
33299         Don't print newline fixits.
33300         (diagnostic_show_locus): Move logic for printing one row into...
33301         (layout::print_line): ...this new function.  Move the
33302         location_get_source_line call and error-handling from
33303         print_source_line to here.  Call print_leading_fixits, and rename
33304         print_any_fixits to print_trailing_fixits.
33305         (selftest::test_fixit_insert_containing_newline): Update now that
33306         newlines are partially supported.
33307         (selftest::test_fixit_insert_containing_newline_2): New test.
33308         (selftest::test_fixit_replace_containing_newline): Update comments.
33309         (selftest::diagnostic_show_locus_c_tests): Call the new test.
33310         * edit-context.c (class added_line): New class.
33311         (class edited_line): Describe newline handling in comment.
33312         (edited_line::actually_edited_p): New method.
33313         (edited_line::print_content): Delete redundant decl.
33314         (edited_line::m_predecessors): New field.
33315         (edited_file::print_content): Call edited_line::print_content.
33316         (edited_file::print_diff): Update to support newlines.
33317         (edited_file::print_diff_hunk): Likewise.
33318         (edited_file::print_run_of_changed_lines): New function.
33319         (edited_file::print_diff_line): Convert to...
33320         (print_diff_line): ...this.
33321         (edited_file::get_effective_line_count): New function.
33322         (edited_line::edited_line): Initialize new field m_predecessors.
33323         (edited_line::~edited_line): Clean up m_predecessors.
33324         (edited_line::apply_fixit): Handle newlines.
33325         (edited_line::get_effective_line_count): New function.
33326         (edited_line::print_content): New function.
33327         (edited_line::print_diff_lines): New function.
33328         (selftest::test_applying_fixits_insert_containing_newline): New
33329         test.
33330         (selftest::test_applying_fixits_replace_containing_newline): New
33331         test.
33332         (selftest::insert_line): New function.
33333         (selftest::test_applying_fixits_multiple_lines): Add example of
33334         inserting a line.
33335         (selftest::edit_context_c_tests): Call the new tests.
33337 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33339         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33340         parameter cand.  Update dump information.
33341         (get_computation_cost): Update uses.
33343 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33345         * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33346         (get_computation_aff): Reorder parameters.  Use get_computation_aff_1.
33347         (get_computation_at, rewrite_use_address): Update use of
33348         get_computation_aff.
33350 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33352         * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33353         (get_computation): Delete.
33354         (get_computation_cost): Implement like get_computation_cost_at.
33355         Use get_computation_at.
33356         (get_computation_cost_at): Delete.
33357         (rewrite_use_nonlinear_expr): Use get_computation_at.
33358         (rewrite_use_compare, remove_unused_ivs): Ditto.
33360 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33362         * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33364 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33366         * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33367         (ivopts_global_cost_for_size): Rename parameter and update uses.
33368         (iv_ca_recount_cost): Update uses.
33369         (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33370         candidates seperately in n_invs and n_cands.
33371         (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33373 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33375         * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33376         (find_inv_vars_cb): New.
33377         (find_depends): Renamed to ...
33378         (find_inv_vars): ... this.
33379         (add_candidate_1, force_var_cost): Call find_inv_vars.
33380         (split_address_cost, determine_group_iv_cost_cond): Ditto.
33382 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33384         * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33385         inv_vars.  Add inv_exprs.
33386         (struct iv_cand): Rename depends_on to inv_vars.
33387         (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33388         max_inv_var_id/n_inv_var_uses.  Move max_inv_expr_id around.
33389         Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33390         (dump_cand): Dump inv_vars.
33391         (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33392         (record_invariant, find_depends, add_candidate_1): Ditto.
33393         (set_group_iv_cost, force_var_cost): Ditto.
33394         (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33395         (get_computation_cost_at, get_computation_cost): Ditto.
33396         (determine_group_iv_cost_generic): Ditto.
33397         (determine_group_iv_cost_address): Ditto.
33398         (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33399         (determine_group_iv_costs): Ditto.
33400         (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33401         (iv_ca_set_remove_invariants): Renamed to ...
33402         (iv_ca_set_remove_invs): ... this.  Support inv_vars and inv_exprs.
33403         (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33404         (iv_ca_set_add_invariants):  Renamed to ...
33405         (iv_ca_set_add_invs): ... this.  Support inv_vars and inv_exprs.
33406         (iv_ca_set_cp): Use iv_ca_set_add_invs.
33407         (iv_ca_has_deps): Support inv_vars and inv_exprs.
33408         (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33409         (create_new_ivs): Remove useless dump.
33411 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33413         * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33414         iv_cand code.
33415         (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33416         (iv_ca_set_no_cp, create_new_iv): Ditto.
33418 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33420         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33422 2017-05-02  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33424         * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33425         function tree_check2.
33427 2017-05-02  Martin Liska  <mliska@suse.cz>
33429         * doc/gcov.texi: Add missing preposition.
33430         * gcov.c (function_info::function_info): Properly fill up
33431         all member variables.
33433 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33435         * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33437 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33439         * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33441 2017-05-02  Martin Liska  <mliska@suse.cz>
33443         PR lto/77954.
33444         * lto-streamer-in.c (lto_read_tree_1): Remove
33445         LTO_STREAMER_DEBUG.
33446         * lto-streamer.c (struct tree_hash_entry): Likewise.
33447         (struct tree_entry_hasher): Likewise.
33448         (tree_entry_hasher::hash): Likewise.
33449         (tree_entry_hasher::equal): Likewise.
33450         (lto_streamer_init): Likewise.
33451         (lto_orig_address_map): Likewise.
33452         (lto_orig_address_get): Likewise.
33453         (lto_orig_address_remove): Likewise.
33454         * lto-streamer.h: Likewise.
33455         * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33456         * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33458 2017-05-02  Sebastian Peryt  <sebastian.peryt@intel.com>
33460         * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33461         (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33462         (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33463         (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33464         (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33465         (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33466         (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33467         (mm_maskz_sub_ss): New intrinsics.
33468         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33469         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33470         * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33471         (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33472         (__builtin_ia32_subss_mask_round): New builtins.
33473         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33474         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33475         * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33476         Renamed to ...
33477         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33478         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33479         Changed to ...
33480         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33481         ... this.
33483 2017-05-02  Martin Jambor  <mjambor@suse.cz>
33485         PR tree-optimization/78687
33486         * tree-sra.c (access): New field parent.
33487         (process_subtree_disqualification): New function.
33488         (disqualify_candidate): Call it.
33489         (build_accesses_from_assign): Reset write flag if creating an
33490         assighnment link.
33491         (build_access_subtree): Fill in parent field and also prpagate
33492         down grp_write flag.
33493         (create_artificial_child_access): New parameter set_grp_write, set
33494         grp_write to its value.
33495         (propagate_subaccesses_across_link): Also propagate grp_write flag
33496         values.
33497         (propagate_all_subaccesses): Push the closest parent back to work
33498         queue if add_access_to_work_queue returned true.
33500 2017-05-02  Richard Biener  <rguenther@suse.de>
33502         * common.opt (fstrict-overflow): Alias negative to fwrapv.
33503         * doc/invoke.texi (fstrict-overflow): Remove all traces of
33504         -fstrict-overflow documentation.
33505         * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33506         (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33507         flag_strict_overflow.
33508         * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33509         * lto-opts.c (lto_write_options): Do not stream it.
33510         * lto-wrapper.c (merge_and_complain): Do not handle it.
33511         * opts.c (default_options_table): Do not set -fstrict-overflow.
33512         (finish_options): Likewise do not clear it when sanitizing.
33513         * simplify-rtx.c (simplify_const_relational_operation): Do not
33514         test flag_strict_overflow.
33516 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33518         * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33519         using enabled attribute.
33520         (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33521         (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33522         (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33523         (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33524         (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33525         (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33526         (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33527         (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33528         (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33529         (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33531 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33533         * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33535 2017-05-02  Richard Biener  <rguenther@suse.de>
33537         PR tree-optimization/80591
33538         Revert
33539         2017-04-10  Richard Biener  <rguenther@suse.de>
33541         * tree-ssa-structalias.c (find_func_aliases): Properly handle
33542         asm inputs.
33544 2017-05-02  Richard Biener  <rguenther@suse.de>
33546         PR tree-optimization/80549
33547         * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33548         (cleanup_tree_cfg_noloop): Create forwarders to known loop
33549         headers if they do not have a preheader.
33551 2017-05-02  Martin Liska  <mliska@suse.cz>
33553         PR other/80589
33554         * common.opt: Fix typo.
33555         * doc/invoke.texi: Likewise.
33557 2017-05-01  Jan Beulich  <jbeulich@suse.com>
33559         * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33560         swapping, add (x,x,m,x,n) alternative.
33562 2017-05-01  Nathan Sidwell  <nathan@acm.org>
33564         * calls.c (combine_pending_stack_adjustment_and_call): Remove
33565         unnecessary unadjusted_alignment check.
33567 2017-05-01  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33569         PR c++/80038
33570         * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33571         operations here.
33572         * gimplify.c (gimplify_cilk_detach): New function.
33573         (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33574         * tree-core.h: Document EXPR_CILK_SPAWN.
33575         * tree.h (EXPR_CILK_SPAWN): Define.
33577 2017-05-01  David Malcolm  <dmalcolm@redhat.com>
33579         * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33580         to use new fixit_hint representation, using the "replace" logic.
33581         (get_line_span_for_fixit_hint): Likewise.
33582         (layout::print_any_fixits): Likewise.
33583         (selftest::test_one_liner_many_fixits): Rename to...
33584         (selftest::test_one_liner_many_fixits_1): ...this, and update
33585         comment and expected output to reflect that the multiple fix-it
33586         hints are now consolidated into one insertion.
33587         (selftest::test_one_liner_many_fixits_2): New test.
33588         (selftest::test_diagnostic_show_locus_one_liner): Update for
33589         above.
33590         (selftest::test_fixit_consolidation): Update for fix-it API
33591         change.
33592         * diagnostic.c (print_parseable_fixits): Likewise.
33593         * edit-context.c (edited_line::m_line_events): Convert from
33594         auto_vec <line_event *> to auto_vec <line_event>.
33595         (class line_event): Convert from abstract base class to a concrete
33596         class, taking over the role of replace_event.
33597         (class insert_event): Delete.
33598         (class replace_event): Rename to class line_event.  Convert to
33599         half-open range.
33600         (edit_context::add_fixits): Reimplement.
33601         (edit_context::apply_insert): Delete.
33602         (edit_context::apply_replace): Rename to...
33603         (edit_context::apply_fixit): ...this.  Convert to half-open range.
33604         (edited_file::apply_insert): Delete.
33605         (edited_file::apply_replace): Rename to...
33606         (edited_file::apply_fixit): ...this.
33607         (edited_line::~edited_line): Drop deletion of events.
33608         (edited_line::apply_insert): Delete.
33609         (edited_line::apply_replace): Rename to...
33610         (edited_line::apply_fixit): ...this.  Convert to half-open range.
33611         Update for change to type of m_line_events.
33612         * edit-context.h (edit_context::apply_insert): Delete.
33613         (edit_context::apply_replace): Rename to...
33614         (edit_context::apply_fixit): ...this.
33616 2017-05-01  Martin Sebor  <msebor@redhat.com>
33618         * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33619         known.
33621 2017-05-01  Uros Bizjak  <ubizjak@gmail.com>
33623         PR target/68491
33624         * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33625         __get_cpuid_max returns 0.
33626         (__get_cpuid_count): Ditto.
33628 2017-05-01  Eric Botcazou  <ebotcazou@adacore.com>
33630         * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33631         replacement expression is another instance of one of its arguments.
33633 2017-05-01  Jakub Jelinek  <jakub@redhat.com>
33635         PR target/79430
33636         * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33637         check for stack push/pop autoinc.
33638         * config/i386/i386.c (ix86_agi_dependent): Return false
33639         if the only reason why modified_in_p returned true is that
33640         addr is SP based and set_insn is a push or pop.
33642 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33644         * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33645         overflow check.
33647 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33649         PR ipa/79224
33650         * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33651         (account_size_time): Use two predicates - exec_pred and
33652         nonconst_pred_ptr.
33653         (evaluate_conditions_for_known_args): Compute both clause and
33654         nonspec_clause.
33655         (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33656         (inline_summary_t::duplicate): Update.
33657         (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33658         separately.
33659         (compute_inline_parameters): Likewise.
33660         (estimate_edge_size_and_time): Update caluclation of time.
33661         (estimate_node_size_and_time): Compute both time and nonspecialized
33662         time.
33663         (estimate_ipcp_clone_size_and_time): Update.
33664         (inline_merge_summary): Update.
33665         (do_estimate_edge_time): Update.
33666         (do_estimate_edge_size): Update.
33667         (do_estimate_edge_hints): Update.
33668         (inline_read_section, inline_write_summary): Stream both new predicates.
33669         * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33670         as argument.
33671         (compute_inlined_call_time): Cleanup.
33672         (big_speedup_p): Update.
33673         (edge_badness): Update.
33674         * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33675         (size_time_entry): Replace predicate by exec_predicate and
33676         nonconst_predicate.
33677         (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33678         (estimate_edge_time): Return also nonspec_time.
33679         (reset_edge_growth_cache): Update.
33681 2017-04-29  Jakub Jelinek  <jakub@redhat.com>
33683         PR rtl-optimization/80491
33684         * ifcvt.c (noce_process_if_block): When looking for x setter
33685         with missing else_bb, don't check only the insn right before
33686         cond_earliest, but look for the last insn that x is modified in
33687         within the same bb.
33689         PR rtl-optimization/80491
33690         * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33692 2017-04-29  Marc Glisse  <marc.glisse@inria.fr>
33694         PR tree-optimization/80487
33695         * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33697 2017-04-29  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33699         PR tree-optimization/79697
33700         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33701         is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33702         (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33703         BUILT_IN_STRNDUP.
33704         * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33705         (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33707 2017-04-28  Martin Sebor  <msebor@redhat.com>
33709         PR tree-optimization/80523
33710         * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33711         (init_target_to_host_charmap, target_to_host, target_strtol10): New
33712         functions.
33713         (maybe_warn, format_directive, parse_directive): Use new functions.
33714         (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33716 2017-04-28  Marc Glisse  <marc.glisse@inria.fr>
33718         * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33720 2017-04-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
33722         * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33723         target_header_dir): Set correctly.
33724         * configure: Regenerated.
33725         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33726         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33727         instead of SYSTEM_HEADER_DIR.
33729 2017-04-28  Jan Hubicka  <hubicka@ucw.cz>
33731         * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33732         (estimate_local_effects): Likewise.
33733         * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33734         edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33735         * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33736         estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33737         do_estimate_edge_time, estimate_edge_time): Likewise.
33738         * ipa-inline-analysis.c (estimate_node_size_and_time,
33739         estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33740         (estimate_time_after_inlining): Remove.
33742 2017-04-28  Martin Liska  <mliska@suse.cz>
33744         * doc/gcov.texi: Enhance documentation of gcov.
33746 2017-04-28  Martin Liska  <mliska@suse.cz>
33748         * doc/gcov.texi: Sort options in alphabetic order.
33749         * doc/gcov-dump.texi: Likewise.
33750         * doc/gcov-tool.texi: Likewise.
33751         * gcov.c (print_usage): Likewise.
33752         * gcov-dump.c (print_usage): Likewise.
33753         * gcov-tool.c (print_merge_usage_message): Likewise.
33754         (print_rewrite_usage_message): Likewise.
33755         (print_overlap_usage_message): Likewise.
33757 2017-04-28  Martin Liska  <mliska@suse.cz>
33759         PR gcov-profile/53915
33760         * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33762 2017-04-28  Martin Liska  <mliska@suse.cz>
33764         PR gcov-profile/79891
33765         * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33766         is marked by compiler as living on a line.
33767         (get_cycles_count): Remove usage of the union.
33768         (output_intermediate_file): Likewise.
33769         (find_source): Fix GNU coding style.
33770         (accumulate_line_counts): Remove old non-all block mode.
33771         (output_lines): Remove usage of the union.
33772         * profile.c (output_location): Include all BBs, even if
33773         belonging to a same line (and file) as a previous BB.
33775 2017-04-28  Martin Liska  <mliska@suse.cz>
33777         * gcov.c (process_args): Handle new argument 'w'.
33778         (read_graph_file): Assign ID to BBs.
33779         (output_branch_count): Display BB # if verbose flag is set.
33780         (output_lines): Likewise for arcs.
33781         (print_usage): Add '--verbose' option help.
33782         * doc/gcov.texi: Document --verbose (-w) option.
33784 2017-04-28  Martin Liska  <mliska@suse.cz>
33786         * gcov.c (struct block_location_info): New struct.
33787         (process_file): Fill up the new structure.
33788         (read_graph_file): Replace usage of encoding by the newly added
33789         struct.
33790         (add_line_counts): Likewise.
33791         (accumulate_line_counts): Remove usage of the union.
33792         (function_info::function_info): New function.
33793         (function_info::~function_info): Likewise.
33794         (process_file): Call delete instead of release_function.
33795         (release_function): Release the function.
33796         (release_structures): Call delete instead of release_function.
33797         (solve_flow_graph): Replace usage of num_blocks.
33798         (find_exception_blocks): Likewise.
33799         (output_lines): Fix GNU coding style.
33801 2017-04-28  Martin Liska  <mliska@suse.cz>
33803         PR driver/56469
33804         * coverage.c (coverage_remove_note_file): New function.
33805         * coverage.h: Declare the function.
33806         * toplev.c (finalize): Clean if an error has been seen.
33808 2017-04-28  Martin Liska  <mliska@suse.cz>
33810         PR gcov-profile/80031
33811         * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33812         * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33813         * gcov.c (read_graph_file): Read just number of blocks.
33814         * profile.c (branch_prob): Do not stream 0 flags per a basic
33815         block.
33817 2017-04-28  Martin Liska  <mliska@suse.cz>
33819         * gcov-dump.c (tag_*): Add new argument to declarations.
33820         (dump_gcov_file): Likewise.
33821         (tag_blocks): Add and use new argument depth.
33822         (tag_arcs): Likewise.
33823         (tag_lines): Likewise.
33824         (tag_counters): Likewise.
33825         (tag_summary): Likewise.
33826         (dump_working_sets): Use depth to do a proper indentation.
33828 2017-04-28  Jakub Jelinek  <jakub@redhat.com>
33830         PR bootstrap/80531
33831         * cgraph.h (symtab_node::debug_symtab): No longer inline.
33832         * symtab.c (symtab_node::debug_symtab): Move definition here.
33834 2017-04-28  Richard Biener  <rguenther@suse.de>
33836         * lto-streamer.h (LTO_major_version): Bump to 7.
33838 2017-04-28  Richard Biener  <rguenther@suse.de>
33840         * tree-vrp.c (assert_info): New struct.
33841         (add_assert_info): New helper.
33842         (register_edge_assert_for_2): Refactor to add asserts to a vector
33843         of assert_info.
33844         (register_edge_assert_for_1): Likewise.
33845         (register_edge_assert_for): Likewise.
33846         (finish_register_edge_assert_for): New helper actually registering
33847         asserts where live on edge.
33848         (find_conditional_asserts): Adjust.
33849         (find_switch_asserts): Likewise.
33850         (evrp_dom_walker::try_find_new_range): Generalize.
33851         (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33853 2017-04-27  Marek Polacek  <polacek@redhat.com>
33855         PR sanitizer/80349
33856         * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33857         arg10 and arg11 to itype.
33859 2017-04-27  Jonathan Wakely  <jwakely@redhat.com>
33861         * doc/extend.texi (Object Size Checking): Improve grammar.
33863 2017-04-27  Richard Earnshaw  <rearnsha@arm.com>
33865         PR target/80530
33866         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33867         that the logic for permitting reciprocal estimates matches that
33868         in use_rsqrt_p.
33870 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33872         PR c++/80534
33873         * tree.c (type_cache_hasher::equal): Only compare
33874         TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33875         (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33876         non-aggregate element types.
33877         * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33878         about the flag on ARRAY_TYPEs in the comment, formatting fix.
33880 2017-04-27  Richard Biener  <rguenther@suse.de>
33882         PR middle-end/80533
33883         * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33884         stripping ARRAY_REFs from MEM_EXPR make sure we're not
33885         keeping a reference to a trailing array.
33887 2017-04-27  Richard Biener  <rguenther@suse.de>
33889         PR middle-end/80539
33890         * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33891         being in loop-closed SSA form conservatively.
33892         (chrec_fold_multiply_poly_poly): Likewise.
33894 2017-04-27  Tamar Christina  <tamar.christina@arm.com>
33896         PR middle-end/79665
33897         * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
33898         CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
33900 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33902         PR target/77728
33903         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
33904         (aarch64_function_arg_alignment): Return unsigned int again, but still
33905         ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
33906         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33907         Don't emit -Wpsabi note.
33908         (aarch64_function_arg_boundary): Likewise.
33909         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33910         caller.
33912 2017-04-26  Nathan Sidwell  <nathan@acm.org>
33914         * tree.h (crc32_unsigned_n): Declare.
33915         (crc32_unsigned, crc32_unsigned): Make inline.
33916         * tree.c (crc32_unsigned_bits): Replace with ...
33917         (crc32_unsigned_n): ... this.
33918         (crc32_unsigned, crc32_byte): Remove.
33919         (crc32_string): Remove unnecessary braces.
33921 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33923         * ipa-cp.c (estimate_local_effects): Convert sreal to int.
33924         * ipa-inline-analysis.c (MAX_TIME): Remove.
33925         (account_size_time): Use sreal for time.
33926         (dump_inline_summary): Update.
33927         (estimate_function_body_sizes): Update.
33928         (estimate_edge_size_and_time): Update.
33929         (estimate_calls_size_and_time): Update.
33930         (estimate_node_size_and_time): Update.
33931         (inline_merge_summary): Update.
33932         (inline_update_overall_summary): Update.
33933         (estimate_time_after_inlining): Update.
33934         (inline_read_section): Update.
33935         (inline_write_summary): Update.
33936         * ipa-inline.c (compute_uninlined_call_time): Update.
33937         (compute_inlined_call_time): Update.
33938         (recursive_inlining): Update.
33939         (inline_small_functions): Update.
33940         (dump_overall_stats): Update.
33941         * ipa-inline.h: Include sreal.h.
33942         (size_time_entry): Turn time to sreal.
33943         (inline_summary): Turn self_time nad time to sreal.
33945 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33947         * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
33948         data-streamer.h
33949         (sreal::stream_out, sreal::stream_in): New.
33950         * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
33952 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
33954         * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
33955         environment.
33957 2017-04-25  Uros Bizjak  <ubizjak@gmail.com>
33959         PR target/70799
33960         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
33961         Handle ASHIFTRT.
33962         (dimode_scalar_chain::compute_convert_gain): Ditto.
33963         (dimode_scalar_chain::make_vector_copies): Ditto.
33964         (dimode_scalar_chain::convert_reg): Ditto.
33965         (dimode_scalar_chain::convert_insn): Ditto.
33966         * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
33967         (VI248_AVX512BW_1): New mode iterator.
33968         (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
33969         <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
33970         mode iterator.
33972 2017-04-25  Martin Sebor  <msebor@redhat.com>
33974         PR tree-optimization/80497
33975         * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
33976         constants are representable in HOST_WIDE_INT.
33977         (parse_directive): Ditto.
33979 2017-04-25  Martin Sebor  <msebor@redhat.com>
33981         PR bootstrap/80486
33982         * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
33983         (new_zero_array): Adjust signature.
33984         (dom_info::dom_init): Used unsigned rather that size_t.
33985         (dom_info::dom_info): Same.
33987 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33988             Jakub Jelinek  <jakub@redhat.com>
33990         PR target/77728
33991         * config/arm/arm.c: Include gimple.h.
33992         (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33993         returns negative, increment ncrn only if it returned positive.
33994         (arm_needs_doubleword_align): Return int instead of bool,
33995         ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
33996         members, but if there is any such non-FIELD_DECL
33997         > PARM_BOUNDARY aligned decl, return -1 instead of false.
33998         (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33999         returns negative, increment nregs only if it returned positive.
34000         (arm_setup_incoming_varargs): Likewise.
34001         (arm_function_arg_boundary): Emit -Wpsabi note if
34002         arm_needs_doubleword_align returns negative, return
34003         DOUBLEWORD_ALIGNMENT only if it returned positive.
34005 2017-04-25  Marek Polacek  <polacek@redhat.com>
34007         PR sanitizer/80349
34008         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34009         first argument to type.
34011 2017-04-25  Bill Seurer  <seurer@linux.vnet.ibm.com>
34013         PR target/80482
34014         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
34015         type checks to test for compatibility instead of equality.
34017 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
34018             Jakub Jelinek  <jakub@redhat.com>
34020         PR target/77728
34021         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
34022         type.
34023         (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
34024         struct.  Ignore DECL_ALIGN of decls other than FIELD_DECL for
34025         the alignment computation, but return their maximum in warn_alignment.
34026         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
34027         Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
34028         is smaller.
34029         (aarch64_function_arg_boundary): Likewise.  Simplify using MIN/MAX.
34030         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
34031         caller.
34033 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34035         * config/arc/simdext.md (dmpyh): Fix typo.
34037 2017-04-25  Richard Biener  <rguenther@suse.de>
34039         PR tree-optimization/80492
34040         * alias.c (compare_base_decls): Handle registers with asm
34041         specification conservatively.
34042         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
34043         compare_base_decls returning dont-know properly.
34045 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34047         * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
34048         (legitimate_offset_address_p): New function.
34049         (arc_legitimate_address_p): Use above function.
34051 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34053         * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
34055 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34057         * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
34058         ACCH registers whenever they are available.
34060 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34062         * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
34063         double regs fix when not used.
34065 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34067         * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
34068         core registers.
34069         (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
34070         (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
34072 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34074         * config/arc/arc.c (arc_output_addsi): Check for h-register class
34075         when emitting short ADD instructions.
34077 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
34079         * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
34080         constraint.
34081         (cmpsi_cc_c_insn): Likewise.
34082         (cbranchsi4_scratch): Compute proper instruction length using
34083         compact_hreg_operand.
34084         * config/arc/predicates.md (compact_hreg_operand): New predicate.
34086 2017-04-25  Richard Biener  <rguenther@suse.de>
34088         PR middle-end/80509
34089         * passes.c (pass_manager::pass_manager): Initialize
34090         m_name_to_pass_map.
34092 2017-04-25  Richard Biener  <rguenther@suse.de>
34094         PR tree-optimization/79201
34095         * tree-ssa-sink.c (statement_sink_location): Handle calls.
34097 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34099         PR target/80464
34100         * config/s390/vector.md: Split MEM->GPR vector moves for
34101         non-s_operand addresses.
34103 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34105         PR target/79895
34106         * config/s390/predicates.md (reload_const_wide_int_operand): New
34107         predicate.
34108         * config/s390/s390.md ("movti"): Remove d/P alternative.
34109         ("movti_bigconst"): New pattern definition.
34111 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34113         PR target/80080
34114         * s390-protos.h (s390_expand_cs_hqi): Removed.
34115         (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
34116         * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
34117         modes as well as CCZ1mode and CCZmode.
34118         (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
34119         signature of s390_emit_compare_and_swap.
34120         (s390_expand_cs_hqi): Likewise, make static.
34121         (s390_expand_cs_tdsi): Generate an explicit compare before trying
34122         compare-and-swap, in some cases.
34123         (s390_expand_cs): Wrapper function.
34124         (s390_expand_atomic_exchange_tdsi): New backend specific expander for
34125         atomic_exchange.
34126         (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
34127         * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
34128         patterns for small and large integers.  Forbid symref memory operands.
34129         Move expander to s390.c.  Require cc register.
34130         ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
34131         ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
34132         ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
34133         ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
34134         symref memory operands.  Remove CC mode and call s390_match_ccmode
34135         instead.
34136         ("atomic_exchange<mode>"): Allow and implement all integer modes.
34138 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34140         * config/s390/s390.md (define_peephole2): New peephole to help
34141         combining the load-and-test pattern with volatile memory.
34143 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34145         * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
34146         with CCZmode for TARGET_Z196.
34148 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34150         PR rtl-optimization/80501
34151         * combine.c (make_compound_operation_int): Set subreg_code to SET
34152         even for AND with mask of the sign bit of mode.
34154         PR rtl-optimization/80500
34155         * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
34156         sum's initial value.
34158 2017-04-25  Julian Brown  <julian@codesourcery.com>
34159             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34161         * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
34163 2017-04-25  Marc Glisse  <marc.glisse@inria.fr>
34165         * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
34167 2017-04-25  Julian Brown  <julian@codesourcery.com>
34168             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34170         * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
34171         (thunderx2t99_sha): New Reservation.
34173 2017-04-25  Julian Brown  <julian@codesourcery.com>
34174             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34176         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
34177         type for 1-element load.
34179 2017-04-24  Marc Glisse  <marc.glisse@inria.fr>
34181         * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
34183 2017-04-24  Martin Jambor  <mjambor@suse.cz>
34185         PR tree-optimization/80293
34186         * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
34187         char arrays not totally scalarizable if it is false.
34188         (analyze_all_variable_accesses): Pass correct value in the new
34189         parameter.  Add a statistics counter.
34191 2017-04-24  Jan Hubicka  <hubicka@ucw.cz>
34193         PR middle-end/79931
34194         * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
34196 2017-04-24  Richard Biener  <rguenther@suse.de>
34198         PR tree-optimization/80494
34199         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
34200         out for complex types.
34202 2017-04-24  Richard Biener  <rguenther@suse.de>
34204         * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
34205         * tree-ssa-sccvn.c (print_scc): Print SCC size.
34206         (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
34207         (DFS): Adjust and never fail.
34208         (sccvn_dom_walker::fail): Remove.
34209         (sccvn_dom_walker::before_dom_children): Adjust.
34210         (run_scc_vn): Likewise and never fail.
34211         * tree-ssa-pre.c (pass_pre::execute): Adjust.
34212         (pass_fre::execute): Likewise.
34214 2017-04-24  Richard Biener  <rguenther@suse.de>
34216         PR tree-optimization/79725
34217         * tree-ssa-sink.c (statement_sink_location): Return whether
34218         failure reason was zero uses.  Move that check later.
34219         (sink_code_in_bb): Deal with zero uses by removing the stmt
34220         if possible.
34222 2017-04-24  Richard Biener  <rguenther@suse.de>
34224         PR c++/2972
34225         * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34226         pointer-based references.
34228 2017-04-24  Richard Biener  <rguenther@suse.de>
34230         PR bootstrap/79814
34231         * pass_manager.h (pass_manager::operator new): Remove.
34232         (pass_manager::operator delete): Likewise.
34233         * passes.c (pass_manager::operator new): Remove.
34234         (pass_manager::operator delete): Likewise.
34235         (pass_manager::pass_manager): Zero individual pass members.
34237 2017-04-23  Uros Bizjak  <ubizjak@gmail.com>
34239         PR target/70799
34240         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34241         <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34242         Check "XEXP (src, 1)" operand here.
34243         <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34244         Check "XEXP (src, 1)" operand here.
34245         (dimode_scalar_chain::make_vector_copies): Detect count register
34246         of a shift instruction.  Zero extend count register from QImode
34247         to DImode to satisfy vector shift pattern count operand predicate.
34248         Substitute vector shift count operand with a DImode copy.
34249         (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34250         vector register.
34252 2017-04-21  Uros Bizjak  <ubizjak@gmail.com>
34254         * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34255         Remove UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34256         (*insvqi_1_mem_rex64): Move above insv<mode>_1.  Remove
34257         UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34258         (UNSPEC_NOREX_MEM): Remove definition.
34260 2017-04-21  Richard Biener  <rguenther@suse.de>
34262         PR tree-optimization/79547
34263         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34264         Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34265         bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34266         without any constraints.
34268 2017-04-21  Richard Biener  <rguenther@suse.de>
34270         PR tree-optimization/78847
34271         * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34273 2017-04-21  Richard Biener  <rguenther@suse.de>
34275         * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34276         (build_distinct_type_copy): Likewise.
34277         (build_variant_type_copy): Likewise.
34278         * tree.c (build_qualified_type): Pass down mem-stat info.
34279         (build_distinct_type_copy): Likewise.
34280         (build_variant_type_copy): Likewise.
34282 2017-04-21  Richard Biener  <rguenther@suse.de>
34284         PR tree-optimization/80237
34285         * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34286         defaulted to NULL.
34287         (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34288         for a simplified result.
34290 2016-04-21  Richard Biener  <rguenther@suse.de>
34292         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34293         sth as strict as a simple_iv but a chrec without symbols and an
34294         operand defined in the loop we are peeling (and not some subloop).
34295         (propagate_constants_for_unrolling): Propagate all constants.
34297 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34299         PR target/79804
34300         * config/i386/i386.c (print_reg): Remove assert for disalowed
34301         regno values, call output_operand_lossage instead.
34303 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34305         PR target/78090
34306         * config/i386/constraints.md (Yc): New register constraint.
34307         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34308         Use Yc constraint for alternative 2 of operand 0.  Remove
34309         preferred_for_speed attribute.
34311 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34313         * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34314         lastprivate clauses in SIMT case.
34316 2017-04-20  Volker Reichelt  <v.reichelt@netcologne.de>
34318         * doc/invoke.texi (-Wextra-semi): Document new warning option.
34320 2017-04-20  Richard Biener  <rguenther@suse.de>
34322         PR tree-optimization/57796
34323         * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34324         as N scalar stores.
34325         (vect_model_load_cost): Cost gathers as N scalar loads.
34327 2017-04-20  Richard Biener  <rguenther@suse.de>
34329         * ggc-page.c (ggc_allocated_p): Rename to ...
34330         (safe_lookup_page_table_entry): ... this and return the lookup
34331         result.
34332         (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34334 2017-04-20  Richard Biener  <rguenther@suse.de>
34336         PR tree-optimization/80453
34337         * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34338         * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34339         from the conditions.
34340         (vn_phi_eq): Pass them down.
34341         (vn_phi_lookup): Record them.
34342         (vn_phi_insert): Likewise.
34344 2017-04-20  Matthew Fortune  <matthew.fortune@imgtec.com>
34346         * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34347         uninitialized variable warning to avoid buffer overrun.
34349 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34351         PR other/71250
34352         * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34353         is suppressed for '{ 0 }' in C.
34355 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34357         * BASE-VER: Set to 8.0.0.
34359 2017-04-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34361         * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34362         priority .init_array and .fini_array section with SECTION_NOTYPE
34363         flag.
34365 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34367         PR middle-end/80423
34368         * tree.h (build_array_type): Add typeless_storage default argument.
34369         * tree.c (type_cache_hasher::equal): Also compare
34370         TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34371         (build_array_type): Add typeless_storage argument, set
34372         TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34373         recursive call.
34374         (build_nonshared_array_type): Adjust build_array_type_1 caller.
34375         (build_array_type): Likewise.  Add typeless_storage argument.
34377 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34378             Jakub Jelinek  <jakub@redhat.com>
34380         PR tree-optimization/80426
34381         * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34382         operation on symbolic operands, also compute the overflow for the
34383         invariant part when the operation degenerates into a negation.
34385 2017-04-19  Jakub Jelinek  <jakub@redhat.com>
34387         PR debug/80461
34388         * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34389         Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34391         PR debug/80436
34392         * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34394 2017-04-19  Georg-Johann Lay  <avr@gjlay.de>
34396         PR target/80462
34397         * config/avr/avr.c (tree.h): Include it.
34398         (cgraph.h): Include it.
34399         (avr_encode_section_info): Don't warn for uninitialized progmem
34400         variable if it's just an alias.
34402 2017-04-19  Richard Biener  <rguenther@suse.de>
34404         PR ipa/65972
34405         * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34406         when needed by AutoPGO.
34408 2017-04-19  Paulo J. Matos  <paulo@matos-sorge.com>
34410         PR lto/50345
34411         * doc/lto.texi: Remove an extra 'that'.
34413 2017-04-19  Segher Boessenkool  <segher@kernel.crashing.org>
34415         PR rtl-optimization/80429
34416         * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34417         are only used in debug insns.
34419 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34420             Vladimir Makarov  <vmakarov@redhat.com>
34422         * config/sparc/predicates.md (input_operand): Add comment.  Return
34423         true for any memory operand when LRA is in progress.
34424         * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34426 2017-04-18  Jeff Law  <law@redhat.com>
34428         PR target/74563
34429         * mips.md ({return,simple_return}_internal): Do not overwrite
34430         operands[0].
34432 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34434         PR tree-optimization/80443
34435         * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34436         instead of adding 1, subtract -1 and similarly instead of subtracting
34437         1 add -1.
34439 2017-04-18  Richard Sandiford  <richard.sandiford@arm.com>
34441         PR rtl-optimization/80357
34442         * haifa-sched.c (tmp_bitmap): New variable.
34443         (model_recompute): Handle duplicate use records.
34444         (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34445         (free_global_sched_pressure_data): Free it.
34447 2017-04-18  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34449         Revert:
34450         2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34451         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34452         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34453         instead of SYSTEM_HEADER_DIR.
34455 2017-04-18  Jeff Law  <law@redhat.com>
34457         PR middle-end/80422
34458         * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34459         predecessors after walking up the insn chain.
34461 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34463         PR debug/80263
34464         * dwarf2out.c (modified_type_die): Try harder not to emit internal
34465         sizetype type into debug info.
34467 2017-04-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
34469         PR target/80099
34470         * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34471         unneeded test for TARGET_UPPER_REGS_SF.
34472         * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34474 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34476         PR sanitizer/80444
34477         * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34478         instead of gsi_after_labels.
34480 2017-04-18  Jeff Law  <law@redhat.com>
34482         * regcprop.c (maybe_mode_change): Avoid creating copies of the
34483         stack pointer.
34485         Revert:
34486         2017-04-13  Jeff Law  <law@redhat.com>
34487         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34488         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34490 2017-04-18  Georg-Johann Lay  <avr@gjlay.de>
34492         PR target/79453
34493         * config/avr/avr.c (intl.h): Include it.
34494         (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34496 2017-04-18  Martin Liska  <mliska@suse.cz>
34498         PR gcov-profile/78783
34499         * gcov-tool.c (gcov_output_files): Validate that destination
34500         file is either removed by the tool or by a user.
34502 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34503             Guy Benyei  <guybe@mellanox.com>
34505         * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34506         block, and do not negate it, the stored id is already negative.
34508 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34510         * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34512 2017-04-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
34514         PR target/80098
34515         * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34516         masks of options that should be turned off if the VSX vector
34517         options are turned off.
34518         (OTHER_P8_VECTOR_MASKS): Likewise.
34519         (OTHER_VSX_VECTOR_MASKS): Likewise.
34520         * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34521         rs6000_disable_incompatible_switches to validate no type switches
34522         like -mvsx.
34523         (rs6000_incompatible_switch): New function to disallow turning on
34524         other vector options if -mno-vsx, -mno-power8-vector, or
34525         -mno-power9-vector are specified.
34527 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34529         * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34531 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34533         * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34534         * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34535         * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34536         (ARG_POINTER_CFA_OFFSET): Likewise.
34538 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34540         * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34541         conditions to take advantage of various optimizations.
34543 2017-04-13  Jeff Law  <law@redhat.com>
34545         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34546         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34547         (zero_extendsidi2_dext): Likewise.
34549 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34551         PR sanitizer/80403
34552         * fold-const.c (fold_ternary_loc): Revert
34553         use op0 instead of fold_convert_loc (loc, type, arg0) part of
34554         2017-04-12 change.
34556 2017-04-13  Vladimir Makarov  <vmakarov@redhat.com>
34558         PR rtl-optimization/80343
34559         * lra-remat.c (update_scratch_ops): Assign original hard reg to
34560         new scratch pseudo.
34562 2017-04-13  Denis Khalikov <d.khalikov@partner.samsung.com>
34564         PR sanitizer/80414
34565         * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34566         to ubsan_encode_value.
34568 2017-04-13  Jeff Law  <law@redhat.com>
34570         * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34571         appearing in DEBUG_INSNs.
34573 2017-04-13  Martin Liska  <mliska@suse.cz>
34575         PR gcov-profile/80413
34576         * gcov-io.c (gcov_write_string): Copy to buffer just when
34577         allocated size is greater than zero.
34579 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34581         PR debug/80321
34582         * dwarf2out.c (decls_for_scope): Ignore declarations of
34583         current_function_decl in BLOCK_NONLOCALIZED_VARS.
34585 2017-04-12  Jan Hubicka  <hubicka@ucw.cz>
34587         PR lto/69953
34588         * ipa-visibility.c (non_local_p): Fix typos.
34589         (localize_node): When localizing symbol in same comdat group,
34590         dissolve the group only when we know external symbols are going
34591         to be privatized.
34592         (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34594 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34596         PR tree-optimization/79390
34597         * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34598         order does not result in usable sequence, retry with reversed operand
34599         order.
34601         PR sanitizer/80403
34602         PR sanitizer/80404
34603         PR sanitizer/80405
34604         * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34605         to fold_build2_loc.  Convert TREE_OPERAND (tem, 0) to type.  Use
34606         op0 instead of fold_convert_loc (loc, type, arg0).
34608 2017-04-12  Jeff Law  <law@redhat.com>
34610         * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34611         has a delay slot in the generated code.
34613         * config/cris/cris.md (cris_preferred_reload_class): Return
34614         GENNONACR_REGS rather than GENERAL_REGS.
34616 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34618         PR c/80163
34619         * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34620         vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34621         signedness of the result type.
34623 2017-04-12  Richard Biener  <rguenther@suse.de>
34624             Jeff Law  <law@redhat.com>
34626         PR tree-optimization/80359
34627         * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34628         trim stores to TARGET_MEM_REFs.
34630 2017-04-12  Richard Biener  <rguenther@suse.de>
34632         PR tree-optimization/79390
34633         * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34634         threading case even more.
34636 2017-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
34638         PR target/80382
34639         * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34640         for quad_address_p for TImode, instead of just not indexed_address.
34642 2017-04-12  Richard Biener  <rguenther@suse.de>
34643             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34645         PR middle-end/79671
34646         * alias.c (component_uses_parent_alias_set_from): Handle
34647         TYPE_TYPELESS_STORAGE.
34648         (get_alias_set): Likewise.
34649         * tree-core.h (tree_type_common): Add typeless_storage flag.
34650         * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34651         * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34652         for types containing members with TYPE_TYPELESS_STORAGE.
34653         (place_field): Likewise.
34654         (layout_type): Likewise for ARRAY_TYPE.
34655         * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34656         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34657         TYPE_TYPELESS_STORAGE.
34658         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34660 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34662         PR sanitizer/80349
34663         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34664         first argument to type.
34666 2017-04-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34668         PR target/80376
34669         PR target/80315
34670         * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34671         CONST0_RTX (mode) rather than const0_rtx where appropriate.
34672         (rs6000_expand_binop_builtin): Likewise.
34673         (rs6000_expand_ternop_builtin): Likewise; also add missing
34674         vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34675         vshasigma built-ins.
34676         * doc/extend.texi: Document that vec_xxpermdi's third argument
34677         must be a constant.
34679 2017-04-11  Uros Bizjak  <ubizjak@gmail.com>
34681         * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34682         Use shift_const cost parameter when calculating gain of STV shifts.
34684 2017-04-11  Vladimir Makarov  <vmakarov@redhat.com>
34686         PR rtl-optimization/70478
34687         * lra-constraints.c (process_alt_operands): Check memory for
34688         disfavoring memory insn operand.
34690 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34692         PR middle-end/80100
34693         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34694         left shift in unsigned HOST_WIDE_INT type.
34696         PR rtl-optimization/80385
34697         * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34698         (not (neg X)) into (plus X -1) for complex or non-integral modes.
34700         PR libgomp/80394
34701         * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34702         if they have any depend clauses.
34704 2017-04-11  Martin Liska  <mliska@suse.cz>
34706         PR ipa/80212
34707         * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34708         * ipa-split.c (split_function): Create a local comdat symbol
34709         if caller is in a comdat group.
34711 2017-04-11  Martin Liska  <mliska@suse.cz>
34713         PR ipa/80212
34714         * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34715         flags.
34717 2017-04-11  Martin Sebor  <msebor@redhat.com>
34719         PR middle-end/80364
34720         * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34721         always use the int type.  Use INTEGRAL_TYPE_P() rather than testing
34722         for INTEGER_TYPE.
34723         (directive::set_width, directive::set_precision, format_character):
34724         Adjust.
34725         (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34726         INTEGER_TYPE.
34728 2017-04-11  Richard Earnshaw  <rearnsha@arm.com>
34730         PR target/80389
34731         * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34732         conflict, set target->arch_name instead of target->cpu_name.
34734 2017-04-11  Richard Biener  <rguenther@suse.de>
34736         PR tree-optimization/80374
34737         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34738         build_zero_cst, remove fold_convertible_p check again.
34740 2017-04-11  Martin Liska  <mliska@suse.cz>
34742         PR sanitizer/70878
34743         * ubsan.c (instrument_object_size): Do not instrument register
34744         variables.
34746 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34748         PR target/80381
34749         * config/i386/i386-builtin-types.def
34750         (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34751         V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34752         V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34753         V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34754         V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34755         V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34756         V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34757         V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34758         V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34759         V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34760         V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34761         V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34762         V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34763         V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34764         V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34765         V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34766         V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34767         V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34768         * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34769         __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34770         __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34771         __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34772         __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34773         __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34774         __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34775         __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34776         __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34777         __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34778         __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34779         __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34780         __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34781         __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34782         __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34783         __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34784         __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34785         __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34786         __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34787         __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34788         __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34789         __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34790         __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34791         __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34792         __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34793         __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34794         __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34795         __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34796         aliases.
34797         * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34798         flag to second_arg_count, handle 4 argument function type _COUNT
34799         aliases, handle second_arg_count on second argument rather than last.
34801 2017-04-10  Jeff Law  <law@redhat.com>
34803         PR tree-optimization/80374
34804         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34805         record anything if we can not convert integer_zero_node to the
34806         desired type.
34808 2017-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
34810         PR target/80108
34811         * config/rs6000/rs6000.c (rs6000_option_override_internal):
34812         Enhance special handling given to the TARGET_P9_MINMAX option in
34813         relation to certain other options.
34815 2017-04-10  Bin Cheng  <bin.cheng@arm.com>
34817         PR tree-optimization/80153
34818         * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34819         remove POINTER_PLUS_EXPR's base part directly, rather than through
34820         aff_tree.
34822 2017-04-10  Richard Biener  <rguenther@suse.de>
34823             Bin Cheng  <bin.cheng@arm.com>
34825         PR tree-optimization/80153
34826         * tree-affine.c (aff_combination_to_tree): Get base pointer from
34827         the first element of pointer type aff_tree.  Build result expr in
34828         aff_tree's type.
34829         (add_elt_to_tree): Convert to type unconditionally.  Remove other
34830         fold_convert calls.
34831         * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34832         (rewrite_use_nonlinear_expr): Check invariant using iv information.
34834 2017-04-10  Richard Biener  <rguenther@suse.de>
34836         * tree-ssa-structalias.c (find_func_aliases): Properly handle
34837         asm inputs.
34839 2017-04-10  Vladimir Makarov  <vmakarov@redhat.com>
34841         PR rtl-optimization/70478
34842         * lra-constraints.c (curr_small_class_check): New.
34843         (update_and_check_small_class_inputs): New.
34844         (process_alt_operands): Update curr_small_class_check.  Disfavor
34845         alternative insn memory operands.  Check available regs for small
34846         class operands.
34848 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
34850         PR target/80057
34851         * config/mips/mips.opt (-mvirt): Update description.
34852         * doc/invoke.texi (-mvirt): Likewise.
34854 2017-04-10  Richard Biener  <rguenther@suse.de>
34856         PR middle-end/80362
34857         * fold-const.c (fold_binary_loc): Look at unstripped ops when
34858         looking for NEGATE_EXPR in -A / -B to A / B folding.
34860 2017-04-10  Martin Liska  <mliska@suse.cz>
34862         PR gcov-profile/80224
34863         * gcov.c (print_usage): Fix usage string.
34864         (get_gcov_intermediate_filename): Remove.
34865         (output_gcov_file): Use both for normal and intermediate format.
34866         (generate_results): Do not initialize special file for
34867         intermediate format.
34869 2017-04-10  Richard Biener  <rguenther@suse.de>
34871         PR tree-optimization/80304
34872         * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34873         for safelen.
34875 2017-04-10  Nathan Sidwell  <nathan@acm.org>
34877         PR target/79905
34878         * config/rs6000/rs6000.c (rs6000_vector_type): New.
34879         (rs6000_init_builtins): Use it.
34881 2016-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34883         * config/arm/arm.md (<mrc>): Add mode to SET source.
34884         (<mrrc>): Likewise.
34886 2017-04-10  Richard Biener  <rguenther@suse.de>
34888         PR middle-end/80344
34889         * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34891 2017-04-10  Jakub Jelinek  <jakub@redhat.com>
34893         PR target/80324
34894         * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34895         _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34896         _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34897         _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
34898         _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
34899         _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
34900         _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
34901         _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
34902         _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
34903         _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
34904         _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
34905         _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
34906         _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
34907         _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
34908         _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
34909         _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
34910         _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
34911         _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
34912         _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
34913         _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
34914         _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
34915         _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
34916         _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
34918 2017-04-08  Vladimir Makarov  <vmakarov@redhat.com>
34920         PR rtl-optimization/70478
34921         * lra-constraints.c: Reverse the last patch.
34923 2017-04-08  Andreas Tobler  <andreast@gcc.gnu.org>
34925         * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
34926         Add comment for WCHAR_T.
34928 2017-04-08  Martin Liska  <mliska@suse.cz>
34930         Revert:
34931         2017-04-07  Martin Liska  <mliska@suse.cz>
34933         PR ipa/80212
34934         * ipa-split.c (split_function): Add function part to a same comdat
34935         group.
34937 2017-04-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34939         PR target/80358
34940         * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
34942 2017-04-07  Pat Haugen  <pthaugen@us.ibm.com>
34944         * rs6000/rs6000.c (vec_load_pendulum): Rename...
34945         (vec_pairing): ...to this.
34946         (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
34947         (rs6000_sched_init): Adjust for name change.
34948         (struct rs6000_sched_context): Likewise.
34949         (rs6000_init_sched_context): Likewise.
34950         (rs6000_set_sched_context): Likewise.
34952 2017-04-07  Jakub Jelinek  <jakub@redhat.com>
34954         PR target/80322
34955         PR target/80323
34956         PR target/80325
34957         PR target/80326
34958         * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
34959         intrinsics.
34960         * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
34961         _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
34962         _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
34964 2017-04-07  Andreas Tobler  <andreast@gcc.gnu.org>
34966         * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
34968 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34970         PR rtl-optimization/70703
34971         * ira-color.c (update_conflict_hard_regno_costs): Use
34972         int64_t instead of HOST_WIDE_INT.
34974 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34976         PR rtl-optimization/70478
34977         * lra-constraints.c (process_alt_operands): Disfavor alternative
34978         insn memory operands.
34980 2017-04-07  Jeff Law  <law@redhat.com>
34982         * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
34983         CALL and NOTE_INSN_CALL_ARG_LOCATION.
34985 2017-04-07  Martin Liska  <mliska@suse.cz>
34987         PR target/79889
34988         * config/aarch64/aarch64.c (aarch64_process_target_attr):
34989         Show error message instead of an ICE.
34991 2017-04-07  Martin Liska  <mliska@suse.cz>
34993         PR ipa/80212
34994         * ipa-split.c (split_function): Add function part to a same comdat
34995         group.
34997 2017-04-07  Richard Biener  <rguenther@suse.de>
34999         PR middle-end/80341
35000         * tree.c (get_unwidened): Also handle ! for_type case for
35001         INTEGER_CSTs.
35002         * convert.c (do_narrow): Split out from ...
35003         (convert_to_integer_1): ... here.  Do not pass final truncation
35004         type to get_unwidened for TRUNC_DIV_EXPR.
35006 2017-04-07  Richard Biener  <rguenther@suse.de>
35008         * tree-affine.c (wide_int_ext_for_comb): Take type rather
35009         than aff_tree.
35010         (aff_combination_const): Adjust.
35011         (aff_combination_scale): Likewise.
35012         (aff_combination_add_elt): Likewise.
35013         (aff_combination_add_cst): Likewise.
35014         (aff_combination_convert): Likewise.
35015         (add_elt_to_tree): Likewise.  Remove unused argument.
35016         (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
35018 2017-04-07  Sebastian Huber  <sebastian.huber@embedded-brains.de>
35020         * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
35021         definition.
35022         * config/arm/arm.c (arm_default_short_enums): Use
35023         ARM_DEFAULT_SHORT_ENUMS.
35024         * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
35026 2017-04-06  Jakub Jelinek  <jakub@redhat.com>
35028         PR debug/80234
35029         * dwarf2out.c (gen_member_die): Handle C++17 inline static data
35030         members with redundant out-of-class redeclaration.
35032 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
35034         PR target/80286
35035         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
35036         * config/i386/i386.md (*zero_extendsidi2):
35037         Add (?*x,*x) and (?*v,*v) alternatives.
35039 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
35041         PR target/79733
35042         * config/i386/i386.c (ix86_expand_builtin)
35043         <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
35044         mode from insn data. Convert operands to insn operand mode.
35045         Copy operands that don't satisfy insn predicate to a register.
35047 2017-04-06  Sam Thursfield  <sam.thursfield@codethink.co.uk>
35049         * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
35050         Update comments.
35052 2017-04-06  Richard Biener  <rguenther@suse.de>
35054         PR tree-optimization/80334
35055         * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
35056         preserve alignment of accesses.
35058 2017-04-06  Richard Biener  <rguenther@suse.de>
35060         PR tree-optimization/80262
35061         * tree-sra.c (build_ref_for_offset): Preserve address-space
35062         information.
35063         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
35064         Drop useless address-space information on MEM_REF offsets.
35066 2017-04-05  Andreas Schwab  <schwab@linux-m68k.org>
35068         * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
35070 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35072         PR rtl-optimization/70703
35073         * ira-color.c (update_conflict_hard_regno_costs): Use
35074         HOST_WIDE_INT instead of long.
35076 2017-04-05  Uros Bizjak  <ubizjak@gmail.com>
35078         PR target/80298
35079         * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
35080         not defined for x86_64 target.  Add -mmmx target option when __SSE2__
35081         is not defined.
35082         * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
35083         for x86_64 target.  Handle -m3dnowa option.
35085 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35087         PR rtl-optimization/70703
35088         * ira-color.c (update_costs_from_allocno): Use the smallest mode.
35089         (update_conflict_hard_regno_costs): Use long instead of unsigned
35090         arithmetic for cost calculation.
35092 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35093             Bernd Edlinger  <bernd.edlinger@hotmail.de>
35095         PR sanitizer/80308
35096         * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
35097         for big endian.
35099 2017-04-05  Eric Botcazou  <ebotcazou@adacore.com>
35101         PR target/78002
35102         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
35103         ptr_mode with Pmode throughout.
35104         * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
35105         into probe_stack_range and use DImode.
35107 2017-04-05  Dominik Vogt  <vogt@linux.vnet.ibm.com>
35109         PR target/79890
35110         * config/s390/s390.c (s390_register_info_gprtofpr): Return if
35111         call_eh_return is true.
35113 2017-04-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35115         * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
35116         Initialize last_match_fntype_index.
35118 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35120         PR target/80310
35121         * tree-nvr.c: Include internal-fn.h.
35122         (pass_return_slot::execute): Ignore internal calls without
35123         direct optab.
35125 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35126             Richard Biener  <rguenther@suse.de>
35128         PR c++/80297
35129         * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
35130         captures used multiple times, except for the last use.
35131         * generic-match-head.c: Include gimplify.h.
35133 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35135         PR tree-optimization/79390
35136         * target.h (struct noce_if_info): Declare.
35137         * targhooks.h (default_noce_conversion_profitable_p): Declare.
35138         * target.def (noce_conversion_profitable_p): New target hook.
35139         * ifcvt.h (struct noce_if_info): New type, moved from ...
35140         * ifcvt.c (struct noce_if_info): ... here.
35141         (noce_conversion_profitable_p): Renamed to ...
35142         (default_noce_conversion_profitable_p): ... this.  No longer
35143         static nor inline.
35144         (noce_try_store_flag_constants, noce_try_addcc,
35145         noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
35146         noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
35147         instead of noce_conversion_profitable_p.
35148         * config/i386/i386.c: Include ifcvt.h.
35149         (ix86_option_override_internal): Don't override
35150         PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
35151         (ix86_noce_conversion_profitable_p): New function.
35152         (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
35153         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
35154         * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
35155         * doc/tm.texi: Regenerated.
35157 2017-04-04  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35159         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
35160         correction.
35162 2017-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35164         PR target/80307
35165         * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
35166         instructions for small multiply cores.
35168 2017-04-04  Jeff Law  <law@redhat.com>
35170         * config/mips/mips.c (mips_multi_add): Zero initialize the newly
35171         added member.
35172         (mips_expand_vec_perm_const): Initialize elements in orig_perm
35173         that are not set by the loop over the elements.
35175 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35177         PR target/80286
35178         * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
35179         int mode, convert_modes it to mode as unsigned, otherwise use
35180         lowpart_subreg to mode rather than SImode.
35181         * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
35182         ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
35183         Use DImode instead of SImode for the shift count operand.
35184         * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
35185         Likewise.
35187 2017-04-04  Richard Biener  <rguenther@suse.de>
35189         PR middle-end/80281
35190         * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
35191         arithmetic done for the negate or the plus.  Simplify.
35192         (A - (-B) -> A + B): Likewise.
35193         * fold-const.c (split_tree): Make sure to not negate pointers.
35195 2017-04-04  Segher Boessenkool  <segher@kernel.crashing.org>
35197         PR rtl-optimization/60818
35198         * simplify-rtx.c (simplify_binary_operation_1): Do not replace
35199         a compare of comparisons with the thing compared if this results
35200         in a different machine mode.
35202 2017-04-03  Jonathan Wakely  <jwakely@redhat.com>
35204         * alias.c (base_alias_check): Fix typo in comment.
35205         * cgraph.h (class ipa_polymorphic_call_context): Likewise.
35206         * cgraphunit.c (symbol_table::compile): Likewise.
35207         * collect2.c (maybe_run_lto_and_relink): Likewise.
35208         * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
35209         * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
35210         * config/avr/avr.c (avr_map_op_t): Likewise.
35211         * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
35212         * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
35213         * config/epiphany/epiphany.md (movcc): Likewise.
35214         * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
35215         * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
35216         Likewise.
35217         * config/mips/mips.c (mips_save_restore_reg): Likewise.
35218         * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
35219         * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
35220         * config/sh/sh.c (sh_rtx_costs): Likewise.
35221         * fold-const.c (fold_truth_andor): Likewise.
35222         * genautomata.c (collapse_flag): Likewise.
35223         * gengtype.h (struct type::u::s): Likewise.
35224         * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35225         * input.c (FORMAT_AMOUNT): Likewise.
35226         * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35227         (known_aggs_to_agg_replacement_list): Likewise.
35228         * ipa-inline-analysis.c: Likewise.
35229         * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35230         * ipa-polymorphic-call.c
35231         (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35232         * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35233         * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35234         Likewise.
35235         * modulo-sched.c (apply_reg_moves): Likewise.
35236         * omp-expand.c (build_omp_regions_1): Likewise.
35237         * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35238         * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35239         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35240         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35241         * value-prof.c: Likewise.
35242         * var-tracking.c (val_reset): Likewise.
35244 2017-04-03  Richard Biener  <rguenther@suse.de>
35246         PR tree-optimization/80275
35247         * fold-const.c (split_address_to_core_and_offset): Handle
35248         POINTER_PLUS_EXPR.
35250 2017-04-03  Eric Botcazou  <ebotcazou@adacore.com>
35252         * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35253         descriptors is at least equal to that of functions.
35255 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35257         * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35259 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35261         PR target/80250
35262         * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35263         (mov<IMOD4:mode>): New expander.
35264         (*mov<IMOD4:mode>_internal): New insn and split pattern.
35266 2017-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
35268         PR rtl-optimization/79405
35269         * fwprop.c (propagations_left): New variable.
35270         (forward_propagate_into): Decrement it.
35271         (fwprop_init): Initialize it.
35272         (fw_prop): If the variable has reached zero, stop propagating.
35273         (fwprop_addr): Ditto.
35275 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35277         PR debug/79255
35278         * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35279         a FUNCTION_DECL, pass it as decl instead of origin to
35280         process_scope_var.
35282 2017-03-31  Alexander Monakov  <amonakov@ispras.ru>
35284         * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35285         string.
35287 2017-03-31  Pat Haugen  <pthaugen@us.ibm.com>
35289         PR target/80107
35290         * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35291         TARGET_VSX_SMALL_INTEGER.
35293 2017-03-31  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35295         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35296         reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35298 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
35300         * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35301         extraction from odd-numbered MSA register.
35303 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35305         PR middle-end/80173
35306         * expmed.c (store_bit_field_1): Don't attempt to create
35307         a word subreg out of hard registers wider than word if they
35308         have HARD_REGNO_NREGS of 1 for their mode.
35310         PR middle-end/80163
35311         * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35312         conversions to integer types wider than word and pointer.
35314         PR debug/80025
35315         * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35316         (rtx_equal_for_cselib_p): Pass 0 to it.
35317         * cselib.c (cselib_hasher::equal): Likewise.
35318         (rtx_equal_for_cselib_1): Add depth argument.  If depth
35319         is 128, don't look up VALUE locs and punt.  Increment
35320         depth in recursive calls when walking VALUE locs.
35322 2017-03-31  Bernd Edlinger  <bernd.edlinger@hotmail.de>
35324         * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35325         (make_gcov_file_name): Use the canonical path name for generating
35326         the MD5 value.
35327         (read_line): Fix handling of files with ascii null bytes.
35329 2017-03-30  Matthew Fortune  <matthew.fortune@imgtec.com>
35331         * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35332         to initialise a vector register instead
35333         of using a const_int.
35335 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35337         PR translation/80189
35338         * gimplify.c (omp_default_clause): Use %qs instead of %s in
35339         diagnostic messages.
35341 2017-03-30  Peter Bergner  <bergner@vnet.ibm.com>
35343         PR target/80246
35344         * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35345         (dfp_diex_<mode>): Update mode of operand 1.
35346         * doc/extend.texi (dxex, dxexq): Document change to return type.
35347         (diex, diexq): Document change to argument type.
35349 2017-03-30  Martin Jambor  <mjambor@suse.cz>
35351         PR ipa/77333
35352         * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35353         * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35354         it reflects the signature changes performed at the callee side.
35355         * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35356         to cgraph_build_function_type_skip_args.
35357         (build_function_decl_skip_args): Adjust call to the above function.
35359 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35361         PR target/80206
35362         * config/i386/sse.md
35363         (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35364         register as dest whenever it is a MEM not rtx_equal_p to the
35365         corresponding dup operand, and when forcing into reg move the
35366         reg into the memory afterwards.
35367         (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35368         Likewise.  Use <ssehalfvecmode> instead of <ssequartermode>
35369         for the force_reg mode.
35370         (avx512vl_vextractf128<mode>): Use register as dest either
35371         always when a MEM, or when it is a MEM not rtx_equal_p to the
35372         corresponding dup operand, or even not when it is a CONST_VECTOR
35373         depending on the mode and lo vs. hi.
35374         (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35375         parens.
35376         (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35377         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35378         Likewise.  Require that operands[2] is even.
35379         (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35380         Remove extraneous parens.  Require that operands[2] is a multiple
35381         of 4.
35382         (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35383         operands[0] is a MEM if <mask_applied>, the predicates/constraints
35384         disallow memory then.
35386 2017-03-30  Richard Biener  <rguenther@suse.de>
35388         PR tree-optimization/77498
35389         * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35390         to non-constants over backedges.
35392 2017-03-29  Segher Boessenkool  <segher@kernel.crashing.org>
35394         PR rtl-optimization/80233
35395         * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35396         as last_combined_insn.  Do not test for BARRIER_P separately.
35398 2017-03-29  Andreas Schwab  <schwab@suse.de>
35400         PR ada/80146
35401         * calls.c (prepare_call_address): Convert funexp to Pmode before
35402         copying to temp reg.
35404 2017-03-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35406         PR tree-optimization/80158
35407         * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35408         Handle possible future case of more than one alternate
35409         interpretation.
35410         (replace_rhs_if_not_dup): Likewise.
35411         (replace_one_candidate): Likewise.
35413 2017-03-28  Vladimir Makarov  <vmakarov@redhat.com>
35415         PR rtl-optimization/80193
35416         * ira.c (ira): Do not check allocation for LRA.
35418 2017-03-28  Alexander Monakov  <amonakov@ispras.ru>
35420         * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35421         (nvptx_output_simt_exit): Declare.
35422         * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35423         cfun->machine->unisimt_location.  Handle NULL unisimt_predicate.
35424         (init_softstack_frame): Move initialization of crtl->is_leaf to...
35425         (nvptx_declare_function_name): ...here.  Emit declaration of local
35426         memory space buffer for omp_simt_enter insn.
35427         (nvptx_output_unisimt_switch): New.
35428         (nvptx_output_softstack_switch): New.
35429         (nvptx_output_simt_enter): New.
35430         (nvptx_output_simt_exit): New.
35431         * config/nvptx/nvptx.h (struct machine_function): New fields
35432         has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35433         * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35434         (UNSPECV_SIMT_EXIT): Ditto.
35435         (omp_simt_enter_insn): New insn.
35436         (omp_simt_enter): New expansion.
35437         (omp_simt_exit): New insn.
35438         * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35440         * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35441         (expand_GOMP_SIMT_ENTER_ALLOC): New.
35442         (expand_GOMP_SIMT_EXIT): New.
35443         * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35444         (GOMP_SIMT_ENTER_ALLOC): Ditto.
35445         (GOMP_SIMT_EXIT): Ditto.
35446         * target-insns.def (omp_simt_enter): New insn.
35447         (omp_simt_exit): Ditto.
35448         * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35449         simt_dlist.
35450         (lower_rec_simd_input_clauses): Implement SIMT privatization.
35451         (lower_rec_input_clauses): Likewise.
35452         (lower_lastprivate_clauses): Handle SIMT privatization.
35454         * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35455         (ompdevlow_adjust_simt_enter): New.
35456         (find_simtpriv_var_op): New.
35457         (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35458         IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35460         * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35461         * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35462         (copy_decl_for_dup_finish): Ditto.
35464         * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35466 2017-03-28  Uros Bizjak  <ubizjak@gmail.com>
35468         PR target/53383
35469         * config/i386/i386.c (ix86_option_override_internal): Always
35470         allow -mpreferred-stack-boundary=3 for 64-bit targets.
35472 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35474         * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35476 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35478         * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35479         mark new edge's irreducible flag accordign to it.
35480         (vect_do_peeling): Check loop preheader edge's irreducible flag
35481         and pass it to function slpeel_add_loop_guard.
35483 2017-03-28  Richard Sandiford  <richard.sandiford@arm.com>
35485         PR tree-optimization/80218
35486         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35487         Update block frequencies and counts.
35489 2017-03-28  Richard Biener  <rguenther@suse.de>
35491         PR tree-optimization/78644
35492         * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35493         of a simplification result we may not use it at all.
35495 2017-03-28  Richard Biener  <rguenther@suse.de>
35497         PR ipa/80205
35498         * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35499         without arguments, generate default definition of a SSA name.
35501 2017-03-28  Richard Biener  <rguenther@suse.de>
35503         PR middle-end/80222
35504         * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35505         TYPE_REF_CAN_ALIAS_ALL references.
35506         * fold-const.c (fold_indirect_ref_1): Likewise.
35508 2017-03-28  Martin Liska  <mliska@suse.cz>
35510         PR ipa/80104
35511         * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35512         thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35514 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35515             Thomas Petazzoni  <thomas.petazzoni@free-electrons.com>
35517         * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35518         (EXTRA_SPECS): Define.
35519         (SUBTARGET_EXTRA_SPECS): Likewise.
35520         (SUBTARGET_CPP_SPEC): Likewise.
35521         * config/arc/elf.h (EXTRA_SPECS): Renamed to
35522         SUBTARGET_EXTRA_SPECS.
35523         * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35525 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35527         * config/arc/simdext.md (vst64_insn): Update pattern.
35528         (vld32wh_insn): Likewise.
35529         (vld32wl_insn): Likewise.
35530         (vld64_insn): Likewise.
35531         (vld32_insn): Likewise.
35533 2017-03-28  Marek Polacek  <polacek@redhat.com>
35535         PR sanitizer/80067
35536         * fold-const.c (fold_comparison): Use protected_set_expr_location
35537         instead of SET_EXPR_LOCATION.
35539 2017-03-28  Markus Trippelsdorf  <markus@trippelsdorf.de>
35541         * tree.c (add_expr): Avoid name lookup warning.
35543 2017-03-27  Jeff Law  <law@redhat.com>
35545         PR tree-optimization/80216
35546         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35547         function name.  Limit recursion depth.
35548         (record_temporary_equivalences): Corresponding changes.
35550 2017-03-27  Jonathan Wakely  <jwakely@redhat.com>
35552         * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35553         covered first.
35555 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35557         PR target/80102
35558         * reg-notes.def (REG_CFA_NOTE): Define.  Use it for CFA related
35559         notes.
35560         * cfgcleanup.c (reg_note_cfa_p): New array.
35561         (insns_have_identical_cfa_notes): New function.
35562         (old_insns_match_p): Don't cross-jump in between /f
35563         and non-/f instructions.  If both i1 and i2 are frame related,
35564         verify all CFA notes, their order and content.
35566 2017-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
35568         PR target/78543
35569         * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35570         HImode and SImode with zero extend to DImode to one insn.
35571         (bswap<mode>2_extenddi): Likewise.
35572         (bswapsi2_extenddi): Likewise.
35573         (bswaphi2_extendsi): Likewise.
35574         (bswaphi2): Combine bswap HImode and SImode into one insn.
35575         Separate memory insns from swapping register.
35576         (bswapsi2): Likewise.
35577         (bswap<mode>2): Likewise.
35578         (bswaphi2_internal): Delete, no longer used.
35579         (bswapsi2_internal): Likewise.
35580         (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35581         store, and gpr<-gpr swap insns.
35582         (bswap<mode>2_store): Likewise.
35583         (bswaphi2_reg): Register only splitter, combine with the splitter.
35584         (bswaphi2 splitter): Likewise.
35585         (bswapsi2_reg): Likewise.
35586         (bswapsi2 splitter): Likewise.
35587         (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35588         the insns into load, store, and register/register insns.
35589         (bswapdi2_ldbrx): Likewise.
35590         (bswapdi2_load): Likewise.
35591         (bswapdi2_store): Likewise.
35592         (bswapdi2_reg): Likewise.
35594 2017-03-27  Gunther Nikl  <gnikl@users.sourceforge.net>
35596         * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35597         (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35599 2017-03-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35601         PR target/80103
35602         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35603         add comments.
35604         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35605         special handling for target option conflicts between dform
35606         options (-mpower9-dform, -mpower9-dform-vector,
35607         -mpower9-dform-scalar) and -mno-direct-move.
35609 2017-03-27  Richard Biener  <rguenther@suse.de>
35611         PR tree-optimization/80181
35612         * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35614 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35616         * config/arc/predicates.md (move_double_src_operand): Replace the
35617         call to move_double_src_operand with a call to address_operand.
35619 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35621         * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35622         * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35623         * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35625 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35627         * config/arc/predicates.md (long_immediate_loadstore_operand):
35628         Consider scaled addresses cases.
35630 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35632         * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35633         restored when in interrupt.
35634         * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35635         doesn't have delay slot.
35637 2017-03-27  Richard Biener  <rguenther@suse.de>
35639         PR ipa/79776
35640         * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35641         inlined thunk clones.
35643 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35645         PR sanitizer/80168
35646         * asan.c (instrument_derefs): Copy over last operand from
35647         original COMPONENT_REF to the new COMPONENT_REF with
35648         DECL_BIT_FIELD_REPRESENTATIVE.
35649         * ubsan.c (instrument_object_size): Likewise.
35651 2017-03-27  Richard Biener  <rguenther@suse.de>
35653         PR tree-optimization/80170
35654         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35655         sure DR/SCEV didnt fold in constants we do not see when looking
35656         at the reference base alignment.
35658 2017-03-27  Richard Biener  <rguenther@suse.de>
35660         PR middle-end/80171
35661         * gimple-fold.c (fold_ctor_reference): Properly guard against
35662         NULL return value from canonicalize_constructor_val.
35664 2017-03-25  Uros Bizjak  <ubizjak@gmail.com>
35666         PR target/80180
35667         * config/i386/i386.c (ix86_expand_builtin)
35668         <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35669         flags reg setting and flags reg using instructions.
35670         <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto.  Use non-flags reg
35671         clobbering instructions to zero extend op2.
35673 2017-03-25  Gerald Pfeifer  <gerald@pfeifer.com>
35675         * doc/install.texi (Configuration) <--with-aix-soname>:
35676         Update link to AIX ld.
35678 2017-03-25  Bernd Schmidt  <bschmidt@redhat.com>
35680         PR rtl-optimization/80160
35681         PR rtl-optimization/80159
35682         * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35683         reg_alternate_class into account.
35685 2017-03-24  Vladimir Makarov  <vmakarov@redhat.com>
35687         PR target/80148
35688         * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35689         to consider in curr_insn_transform.
35691 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35693         * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35694         * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35695         and emit_mode_inner.
35697 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35699         * config/s390/s390-builtins.def: Add VXE builtins.  Add a flags
35700         argument to the overloaded builtin variants.  Use the new flag to
35701         deprecate certain builtin variants.
35702         * config/s390/s390-builtin-types.def: Add new builtin types.
35703         * config/s390/s390-builtins.h: Support new flags field for
35704         overloaded builtins.
35705         * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35706         (s390_macro_to_expand): Enable vector float data type.
35707         (s390_cpu_cpp_builtins_internal): Indicate support of the new
35708         builtins by incrementing the __VEC__ version number.
35709         (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35710         vec_xst.
35711         (s390_resolve_overloaded_builtin): Emit error messages depending
35712         on the builtin flags.
35713         * config/s390/s390.c (s390_expand_builtin): Support additional
35714         flags argument.  Change error message to match the messages
35715         emitted in s390-c.c.
35716         * config/s390/s390.md: New UNSPEC_* constants.
35717         (op_type): Add new instruction types.
35718         * config/s390/vecintrin.h: Add new builtins and test data class
35719         constants.
35720         * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35721         (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35722         (VEC_INEXACT, VEC_NOINEXACT): New constants.
35723         ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35724         ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35725         ("vec_mergel<mode>"): V_HW -> VEC_HW.
35727         ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35728         ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35729         ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35730         ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35732         ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35733         ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35734         ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35735         ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35737         ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35738         ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35739         ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35740         ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35741         ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35742         ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35743         ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35745         ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35746         ("vec_scatter_element<V_HW_4:mode>_DI")
35747         ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35748         ("vec_fpint<mode>", "vflls")
35749         ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35750         ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35751         ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35752         ("*vec_cmphe<mode>_cc"): ... these.
35754         ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35755         mode constant instead of magic value.
35757 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35759         * config/s390/s390.c (s390_expand_vec_compare): Support other
35760         vector floating point modes than just V2DF.
35761         (s390_expand_vcond): Likewise.
35762         (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35763         (s390_cannot_change_mode_class): Prevent mode changes between TF
35764         and V1TF in vector registers.
35765         * config/s390/s390.md (DF, SF): New mode attributes.
35766         ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35767         ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35768         SFmode support for VRs.
35769         * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35770         vector fp modes.
35771         (VFT, VF_HW): New mode iterators.
35772         (vw, sdx): New mode attributes.
35773         ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35774         ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35775         ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35776         ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35777         ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35778         also the new vector floating point modes.  Renaming to ...
35780         ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35781         ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35782         ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35783         ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35784         ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35785         ("vec_unordered<mode>"): ... these.
35787         ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35788         ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35789         ("*vec_extendv2df"): New insn definitions.
35791 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35793         * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35794         ("mulditi3_2", "*muldi3_sign"): New patterns.
35795         ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35796         rename the pattern definition.
35798 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35800         * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35801         expander.
35802         ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35804 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35806         * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35807         instruction if possible.
35808         * config/s390/vector.md (vec_halfnumelts): New mode
35809         attribute.
35810         ("*vec_vllezlf<mode>"): New pattern.
35812 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35814         * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35815         ("popcountv4si2", "popcountv2di2"): Rename to ...
35816         ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35817         ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35818         condition.
35819         ("popcount<mode>2_vxe"): New pattern.
35821 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35823         * common/config/s390/s390-common.c (processor_flags_table): Add
35824         arch12.
35825         * config.gcc: Add arch12.
35826         * config/s390/driver-native.c (s390_host_detect_local_cpu):
35827         Default to arch12 for unknown CPU model numbers.
35828         * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35829         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35830         PROCESSOR_max sanity check.
35831         * config/s390/s390-opts.h (enum processor_type): Add
35832         PROCESSOR_ARCH12.
35833         * config/s390/s390.c (processor_table): Add arch12.
35834         (s390_expand_builtin): Add check for B_VXE flag.
35835         (s390_issue_rate): Add PROCESSOR_ARCH12.
35836         (s390_get_sched_attrmask): Likewise.
35837         (s390_get_unit_mask): Likewise.
35838         (s390_sched_score): Enable z13 scheduling for arch12.
35839         (s390_sched_reorder): Likewise.
35840         (s390_sched_variable_issue): Likewise.
35841         * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35842         PF_VXE.
35843         (s390_tune_attr): Use z13 scheduling also for arch12.
35844         (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35845         (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35846         (TARGET_VXE_P): New macros.
35847         * config/s390/s390.md: Add arch12 to cpu attribute.  Add arch12
35848         and vxe to cpu_facility.  Add arch12 and vxe to enabled attribute.
35849         * config/s390/s390.opt: Add arch12 as processor_type.
35851 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35853         * config/s390/s390.md
35854         ("fixuns_truncdddi2", "fixuns_trunctddi2")
35855         ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35856         ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35858         ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35859         Rename expanders to ...
35861         ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35862         ("fixuns_truncdddi2_emu"): ... these.
35864         ("fixuns_trunc<mode>si2_emu"): New expander.
35866         ("*fixuns_truncdfdi2_z13"): Rename to ...
35867         ("*fixuns_truncdfdi2_vx"): ... this.
35869 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35871         * config/s390/2964.md: Remove the single element vector compare
35872         instructions which are no longer used.
35873         * config/s390/s390.c (s390_select_ccmode): Remove handling of
35874         vector CCmodes.
35875         (s390_canonicalize_comparison): Remove handling of DFmode
35876         compares.
35877         (s390_expand_vec_compare_scalar): Remove function.
35878         (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35879         * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35880         pattern.
35881         ("*cmp<mode>_ccs"): Add wfcdb instruction.
35883 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35885         * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35886         FP zero.
35887         ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35888         will anyway by matched by mov<mode>_64dfp.
35890 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35892         * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35893         vlef/vstef.  Add missing operand to vleif.
35895 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35897         * config/s390/s390.c (s390_expand_vec_init): Enable vector load
35898         pair for all vector types with 64 bit elements.
35899         * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
35900         * config/s390/vector.md (V_HW_64): ... here.
35901         (V_128_NOSINGLE): New mode iterator.
35902         ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
35903         ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
35904         ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
35905         ("*vec_load_pairv2di"): Change to ...
35906         ("*vec_load_pair<mode>"): ... this one.
35908 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35910         * config/s390/constraints.md: Add comments.
35911         (jKK): Reject element sizes > 8 bytes.
35912         * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
35913         s_operands.
35914         * config/s390/s390.md: Add the s_operand checks formerly in
35915         s390_split_ok_p to various splitters where they are still
35916         required.
35917         * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
35918         for 128 bit vectors.  Plus two splitters.
35920 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35922         * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
35923         the file.
35925 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35927         PR target/79893
35928         * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
35929         error if the boundary argument is not constant.
35931 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35933         PR rtl-optimization/80112
35934         * loop-doloop.c (doloop_condition_get): Don't check condition
35935         if cmp isn't SET with IF_THEN_ELSE src.
35937 2017-03-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35939         PR tree-optimization/80158
35940         * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
35941         replacing a candidate statement, also replace it for the
35942         candidate's alternate interpretation.
35943         (replace_rhs_if_not_dup): Likewise.
35944         (replace_one_candidate): Likewise.
35946 2017-03-24  Richard Biener  <rguenther@suse.de>
35948         PR tree-optimization/80167
35949         * graphite-isl-ast-to-gimple.c
35950         (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
35951         properly.
35952         (translate_isl_ast_to_gimple::get_rename): Likewise.
35954 2017-03-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35956         * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
35957         handling of certain combinations of target options, including the
35958         combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
35959         -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
35961 2017-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
35963         PR target/71436
35964         * config/arm/arm.md (*load_multiple): Add reload_completed to
35965         matching condition.
35967 2017-03-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35968             Richard Biener  <rguenth@suse.de>
35970         PR tree-optimization/79908
35971         PR tree-optimization/80136
35972         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
35973         been cast away, gimplify_and_add suffices.
35975 2017-03-23  Markus Trippelsdorf  <markus@trippelsdorf.de>
35977         * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
35979 2017-03-23  Richard Biener  <rguenther@suse.de>
35981         PR tree-optimization/80032
35982         * gimplify.c (gimple_push_cleanup): Forced unconditional
35983         cleanups still have to go to the conditional_cleanups
35984         sequence.
35986 2017-03-22  Jakub Jelinek  <jakub@redhat.com>
35988         PR tree-optimization/80072
35989         * tree-ssa-reassoc.c (struct operand_entry): Change id field type
35990         to unsigned int.
35991         (next_operand_entry_id): Change type to unsigned int.
35992         (sort_by_operand_rank): Make sure to return the right return value
35993         even if unsigned fields are bigger than INT_MAX.
35994         (struct oecount): Change cnt and id type to unsigned int.
35995         (oecount_hasher::equal): Formatting fix.
35996         (oecount_cmp): Make sure to return the right return value
35997         even if unsigned fields are bigger than INT_MAX.
35998         (undistribute_ops_list): Change next_oecount_id type to unsigned int.
36000         PR c++/80129
36001         * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
36002         TREE_READONLY on result if writing it more than once.
36004         PR sanitizer/80110
36005         * doc/invoke.texi (-fsanitize=thread): Document that with
36006         -fnon-call-exceptions atomics are not able to throw
36007         exceptions.
36009         PR sanitizer/80110
36010         * tsan.c: Include tree-eh.h.
36011         (instrument_builtin_call): Call maybe_clean_eh_stmt or
36012         maybe_clean_or_replace_eh_stmt where needed.
36013         (instrument_memory_accesses): Add cfg_changed argument.
36014         Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
36015         if it returned true.
36016         (tsan_pass): Adjust caller.  Return TODO_cleanup_cfg if cfg_changed.
36018         PR rtl-optimization/63191
36019         * config/i386/i386.c (ix86_delegitimize_address): Turn into small
36020         wrapper function, moved the whole old content into ...
36021         (ix86_delegitimize_address_1): ... this.  New inline function.
36022         (ix86_find_base_term): Use ix86_delegitimize_address_1 with
36023         true as last argument instead of ix86_delegitimize_address.
36025 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
36027         * config/aarch64/aarch64.c (generic_branch_cost): Copy
36028         cortexa57_branch_cost.
36030 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
36032         * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
36034 2017-03-21  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36036         PR target/80123
36037         * doc/md.texi (Constraints): Document wA constraint.
36038         * config/rs6000/constraints.md (wA): New.
36039         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
36040         (rs6000_init_hard_regno_mode_ok): Init wA constraint.
36041         * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
36042         * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
36044 2017-03-22  Cesar Philippidis  <cesar@codesourcery.com>
36046         PR c++/80029
36047         * gimplify.c (is_oacc_declared): New function.
36048         (oacc_default_clause): Use it to set default flags for acc declared
36049         variables inside parallel regions.
36050         (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
36051         declared variables.
36052         (gimplify_oacc_declare): Gimplify the declare clauses.  Add the
36053         declare attribute to any decl as necessary.
36055 2017-03-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
36057         PR target/80082
36058         * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
36059         (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
36060         * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
36061         (arm_arch_lpae): This.
36062         * config/arm/arm.c (arm_arch7ve): Rename into ...
36063         (arm_arch_lpae): This.  Define it in term of isa_bit_lpae.
36064         * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
36065         arm_arch_lpae.
36067 2017-03-22  Martin Liska  <mliska@suse.cz>
36069         PR target/79906
36070         * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
36071         error message instead of an ICE.
36073 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36075         * doc/extend.texi (6.11 Additional Floating Types): Revise.
36077 2017-03-21  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36079         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
36080         comments.
36081         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
36082         comments.
36084 2017-03-21  Martin Sebor  <msebor@redhat.com>
36086         * doc/extend.texi: Use "cannot" instead of "can't."
36087         * doc/hostconfig.texi: Same.
36088         * doc/install.texi: Same.
36089         * doc/invoke.texi: Same.
36090         * doc/loop.texi: Same.
36091         * doc/md.texi: Same.
36092         * doc/objc.texi: Same.
36093         * doc/rtl.texi: Same.
36094         * doc/tm.texi: Same.
36095         * doc/tm.texi.in: Same.
36096         * doc/trouble.texi: Same.
36098 2017-03-21  Alexandre Oliva  <aoliva@redhat.com>
36100         PR debug/63238
36101         * dwarf2out.c (struct checksum_attributes): Add at_alignment.
36102         (collect_checksum_attributes): Set it.
36103         (die_checksum_ordered): Use it.
36105 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36107         PR tree-optimization/79908
36108         * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
36109         change: For a VA_ARG whose LHS has been cast away, use
36110         force_gimple_operand to construct the side effects.
36112 2017-03-21  David Malcolm  <dmalcolm@redhat.com>
36114         PR translation/80001
36115         * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
36116         more amenable to translation.
36117         (oacc_loop_auto_partitions): Likewise.
36119 2017-03-21  Marek Polacek  <polacek@redhat.com>
36120             Martin Sebor  <msebor@redhat.com>
36122         PR tree-optimization/80109
36123         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
36124         on INTEGRAL_TYPE_P.
36126 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36127             Segher Boessenkool  <segher@kernel.crashing.org>
36129         PR target/80125
36130         * combine.c (can_combine_p): Revert the 2017-03-20 change, only
36131         check reg_used_between_p between insn and one of succ or succ2
36132         depending on if succ is artificial insn not inserted into insn
36133         stream.
36135 2017-03-21  Martin Liska  <mliska@suse.cz>
36137         PR gcov-profile/80081
36138         * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
36139         * doc/gcc.texi: Include gcov-dump stuff.
36140         * doc/gcov-dump.texi: New file.
36142 2017-03-21  Toma Tabacu  <toma.tabacu@imgtec.com>
36144         PR rtl-optimization/79150
36145         * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
36146         conditional jump, if the jump is the last insn of the loop.
36148 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36149             Richard Biener  <rguenth@suse.de>
36151         PR tree-optimization/79908
36152         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36153         been cast away, use force_gimple_operand to construct the side
36154         effects.
36156 2017-03-21  Martin Liska  <mliska@suse.cz>
36158         PR libfortran/79956
36159         * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
36160         to NULL.
36162 2017-03-21  Brad Spengler <spender@grsecurity.net>
36164         PR plugins/80094
36165         * plugin.c (htab_hash_plugin): New function.
36166         (add_new_plugin): Use it and adjust.
36167         (parse_plugin_arg_opt): Adjust.
36168         (init_one_plugin): Likewise.
36170 2017-03-21  Richard Biener  <rguenther@suse.de>
36172         PR tree-optimization/80032
36173         * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
36174         if set force the cleanup to happen unconditionally.
36175         (gimplify_target_expr): Push inserted clobbers with force_uncond
36176         to avoid them being removed by control-dependent DCE.
36178 2017-03-21  Richard Biener  <rguenther@suse.de>
36180         PR tree-optimization/80122
36181         * tree-inline.c (copy_bb): Do not expans va-arg packs or
36182         va_arg_pack_len when the inlined call stmt requires pack
36183         expansion itself.
36184         * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
36186 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36188         PR sanitizer/78158
36189         * tsan.c (instrument_builtin_call): If the memory model argument
36190         is not a constant, assume it is valid.
36192         PR c/67338
36193         * fold-const.c (round_up_loc): Negate divisor in unsigned type to
36194         avoid UB.
36196 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36198         PR rtl-optimization/79910
36199         * combine.c (can_combine_p): Do not allow combining an I0 or I1
36200         if its dest is used by an insn before I2 (other than the combined
36201         insns themselves, which are properly handled already).
36203 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36205         Revert:
36206         2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36208         * combine.c (record_used_regs): New static function.
36209         (try_combine): Handle situations where there is an additional
36210         instruction between I2 and I3 which needs to have a LOG_LINK
36211         updated.
36213         Revert:
36214         2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36216         * combine.c (try_combine): Delete redundant i1 test.  Call
36217         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36219 2017-03-20  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36221         PR target/80083
36222         * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
36223         alternatives 13/14.
36225 2017-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36227         PR tree-optimization/80054
36228         * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36229         the optimization if a PHI or any of its arguments is not dominated
36230         by the candidate's basis.  Use gphi* rather than gimple* as
36231         appropriate.
36232         (replace_profitable_candidates): Clean up a gimple* variable that
36233         should be a gphi* variable.
36235 2017-03-20  Martin Sebor  <msebor@redhat.com>
36237         PR c++/52477
36238         * doc/extend.texi (attribute constructor): Document present limitation.
36240 2017-03-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36242         PR target/79963
36243         * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36244         __POWER9_VECTOR__ #ifdef control, change template definition to
36245         use Power9-specific built-in function.
36246         (vec_any_eq): Likewise.
36247         * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36248         to control outcomes from this test.
36249         (vector_ae_<mode>p): For VEC_F modes, likewise.
36251 2017-03-20  Ian Lance Taylor  <iant@google.com>
36253         * config/i386/i386.c (ix86_function_regparm): Save an extra
36254         register for -fsplit-stack with DECL_STATIC_CHAIN.
36256 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36258         PR target/79912
36259         * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36260         (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36262 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36264         * config/riscv/riscv.c (riscv_print_operand): Use "fence
36265         iorw,ow".
36266         * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36267         iorw,iorw".
36269 2017-03-20  Marek Polacek  <polacek@redhat.com>
36271         PR sanitizer/80063
36272         * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36274 2017-03-20  Richard Biener  <rguenther@suse.de>
36276         PR tree-optimization/80113
36277         * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36278         allocate extra SSA name for PHI def.
36279         (add_close_phis_to_outer_loops): Likewise.
36280         (add_close_phis_to_merge_points): Likewise.
36281         (copy_loop_close_phi_args): Likewise.
36282         (copy_cond_phi_nodes): Likewise.
36284 2017-03-20  Martin Liska  <mliska@suse.cz>
36286         PR middle-end/79753
36287         * tree-chkp.c (chkp_build_returned_bound): Do not build
36288         returned bounds for a LHS that's not a BOUNDED_P type.
36290 2017-03-20  Martin Liska  <mliska@suse.cz>
36292         PR target/79769
36293         PR target/79770
36294         * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36295         COMPLEX_CST and VECTOR_CST.
36297 2017-03-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36299         PR target/78857
36300         * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36301         target operand.  A new splitter adds the clobber statement in case
36302         the target operand is dead anyway.
36304 2017-03-19  Gerald Pfeifer  <gerald@pfeifer.com>
36306         * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36307         to age-old versions of binutils and glibc.
36309 2017-03-18  Segher Boessenkool  <segher@kernel.crashing.org>
36311         * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36313 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36315         * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36317 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36319         * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36320         requirement for binutils 2.13.
36322 2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36324         * combine.c (try_combine): Delete redundant i1 test.  Call
36325         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36327 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com
36329         * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36330         riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36331         contents.
36332         <riscv64-*-elf>: Re-arrange section
36333         <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36334         <riscv32-*-linux>: Likewise.
36335         <riscv64-*-elf>: Likewise
36336         <riscv64-*-linux>: Likewise.
36338 2017-03-17  Richard Earnshaw  <rearnsha@arm.com>
36340         PR target/80052
36341         * aarch64.opt(verbose-cost-dump): Fix typo.
36343 2017-03-17  Pat Haugen  <pthaugen@us.ibm.com>
36345         PR target/79951
36346         * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36347         for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36349 2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36351         * reload.c (find_reloads): When reloading a nonoffsettable address,
36352         use RELOAD_OTHER for it and its address reloads.
36354         PR rtl-optimization/79910
36355         * combine.c (record_used_regs): New static function.
36356         (try_combine): Handle situations where there is an additional
36357         instruction between I2 and I3 which needs to have a LOG_LINK
36358         updated.
36360 2017-03-17  Jeff Law  <law@redhat.com>
36362         PR tree-optimization/71437
36363         * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36364         conditional in the hash table first.
36365         (vrp_dom_walker::before_dom_children): Extract condition from
36366         ASSERT_EXPR.  Record condition, its inverion and any implied
36367         conditions as well.
36369 2017-03-17  Marek Polacek  <polacek@redhat.com>
36370             Markus Trippelsdorf  <markus@trippelsdorf.de>
36372         PR tree-optimization/80079
36373         * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36374         m_stores_head.
36376 2017-03-17  Richard Biener  <rguenther@suse.de>
36378         PR middle-end/80075
36379         * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36380         Properly verify the LHS before the RHS possibly claims to be
36381         handled.
36382         (stmt_could_throw_p): Hande gimple conds fully here.  Clobbers
36383         do not throw.
36385 2017-03-17  Martin Jambor  <mjambor@suse.cz>
36387         * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36388         (List of -O2 options): Likewise.
36389         (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36390         (-fipa-vrp) New.
36392 2017-03-17  Tom de Vries  <tom@codesourcery.com>
36394         * gcov-dump.c (print_usage): Print bug_report_url.
36396 2017-03-17  Richard Biener  <rguenther@suse.de>
36398         PR middle-end/80050
36399         * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36400         (parser::peek): Likewise.
36402 2017-03-17  Richard Biener  <rguenther@suse.de>
36404         PR tree-optimization/80048
36405         * sese.c (free_sese_info): Properly release rename_map and
36406         copied_bb_map elements.
36408 2017-03-16  Alexandre Oliva  <aoliva@redhat.com>
36410         * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36411         Add linked-list forward and backlinks.  Insert on
36412         construction, remove on destruction.
36413         (class pass_store_merging): Add m_stores_head field.
36414         (pass_store_merging::terminate_and_process_all_chains):
36415         Iterate over m_stores_head list.
36416         (pass_store_merging::terminate_all_aliasing_chains):
36417         Likewise.
36418         (pass_store_merging::execute): Check for debug stmts first.
36419         Push new chains onto the m_stores_head stack.
36421 2017-03-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
36423         PR target/71294
36424         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36425         SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36426         but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36428 2017-03-16  Jeff Law  <law@redhat.com>
36430         PR tree-optimization/71437
36431         * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36432         member function.  Implementation moved into after_dom_children
36433         member function and into the threader's thread_outgoing_edges
36434         function.
36435         (dom_opt_dom_walker::after_dom_children): Simplify by moving
36436         some code into new thread_outgoing_edges.
36437         * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36438         definition.  Simplify marker handling (do it here).   Assume we always
36439         have the available expression and the const/copies tables.
36440         (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36441         and tree-vrp.c
36442         * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36443         * tree-vrp.c (equiv_stack): No longer file scoped.
36444         (vrp_dom_walker): New class.
36445         (vrp_dom_walker::before_dom_children): New member function.
36446         (vrp_dom_walker::after_dom_children): Likewise.
36447         (identify_jump_threads):  Setup domwalker.  Use it rather than
36448         walking edges in a random order by hand.  Simplify setup/finalization.
36449         (finalize_jump_threads): Remove.
36450         (vrp_finalize): Do not call identify_jump_threads here.
36451         (execute_vrp): Do it here instead and call thread_through_all_blocks
36452         here too.
36454         PR tree-optimization/71437
36455         * tree-ssa-dom.c (pfn_simplify): Add basic_block argument.  All
36456         callers changed.
36457         (simplify_stmt_for_jump_threading): Add basic_block argument.  All
36458         callers changed.
36459         (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36460         (dom_opt_dom_walker::thread_across_edge): Remove
36461         handle_dominating_asserts argument.  All callers changed.
36462         (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36463         changes.  Remove calls to lhs_of_dominating_assert.  Other
36464         uses of handle_dominating_asserts turn into unconditional code
36465         (simplify_control_stmt_condition_1): Likewise.
36466         (simplify_control_stmt_condition): Likewise.
36467         (thread_through_normal_block, thread_across_edge): Likewise.
36468         * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36469         * tree-vrp.c (lhs_of_dominating_assert): Move here.  Return original
36470         object if it is not an SSA_NAME.
36471         (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36472         before calling into the VRP specific simplifiers.
36473         (identify_jump_threads): Remove handle_dominating_asserts
36474         argument.
36476 2017-03-16  Jakub Jelinek  <jakub@redhat.com>
36478         PR fortran/79886
36479         * tree-diagnostic.c (default_tree_printer): No longer static.
36480         * tree-diagnostic.h (default_tree_printer): New prototype.
36482 2017-03-16  Tamar Christina  <tamar.christina@arm.com>
36484         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36485         Change ins into fmov.
36487 2017-03-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36489         * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36490         * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36491         Use h_con constraint for operand 1.
36492         (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36493         (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36495 2017-03-15  Jeff Law  <law@redhat.com>
36497         PR tree-optimization/71437
36498         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36499         (record_temporary_equivalences): Use it.
36501         PR tree-optimization/71437
36502         * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36503         tree-ssa-scopedtables.
36504         (lookup_avail_expr, build_and_record_new_cond): Likewise.
36505         (record_conditions, record_cond, vuse_eq): Likewise.
36506         (record_edge_info): Adjust to API tweak of record_conditions.
36507         (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36508         (record_temporary_equivalences, optimize_stmt): Likewise.
36509         (eliminate_redundant_computations): Likewise.
36510         (record_equivalences_from_stmt): Likewise.
36511         * tree-ssa-scopedtables.c: Include options.h and params.h.
36512         (vuse_eq): New function, moved from tree-ssa-dom.c
36513         (build_and_record_new_cond): Likewise.
36514         (record_conditions): Likewise.  Accept vector of conditions rather
36515         than edge_equivalence structure for first argument.
36516         for the first argument.
36517         (avail_exprs_stack::lookup_avail_expr): New member function, moved
36518         from tree-ssa-dom.c.
36519         (avail_exprs_stack::record_cond): Likewise.
36520         * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36521         from tree-ssa-dom.c.
36522         (avail_exprs_stack): Add new member functions lookup_avail_expr
36523         and record_cond.
36524         (record_conditions): Declare.
36526 2017-03-15  Vladimir Makarov  <vmakarov@redhat.com>
36528         PR target/80017
36529         * lra-constraints.c (process_alt_operands): Increase reject for
36530         reloading an input/output operand.
36532 2017-03-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
36534         PR target/79038
36535         * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36536         insns to convert from signed/unsigned char/short to IEEE 128-bit
36537         floating point.
36538         (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36540 2017-03-15  Uros Bizjak  <ubizjak@gmail.com>
36542         PR target/80019
36543         * config/i386/i386.c (ix86_vector_duplicate_value): Create
36544         subreg of inner mode for values already in registers.
36546 2017-03-15  Bernd Schmidt  <bschmidt@redhat.com>
36548         * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36549         iteration reg is used after the loop.
36551 2017-03-14  Martin Sebor  <msebor@redhat.com>
36553         PR tree-optimization/79800
36554         * gimple-ssa-sprintf.c (format_floating: Add argument.  Handle
36555         precision in negative-positive range.
36556         (format_floating): Call non-const overload with adjusted precision.
36558 2017-03-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
36560         PR target/79947
36561         * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36562         -mpowerpc-gfxopt.
36564 2017-03-14  Martin Sebor  <msebor@redhat.com>
36566         PR middle-end/80020
36567         * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36568         * builtins.def (aligned_alloc): Use it.
36570         PR c/79936
36571         * Makefile.in (GTFILES): Add calls.c.
36572         * calls.c: Include "gt-calls.h".
36574 2017-03-14  Bernd Schmidt  <bschmidt@redhat.com>
36576         PR rtl-optimization/79728
36577         * regs.h (struct target_regs): New field
36578         x_contains_allocatable_regs_of_mode.
36579         (contains_allocatable_regs_of_mode): New macro.
36580         * reginfo.c (init_reg_sets_1): Initialize it, and change
36581         contains_reg_of_mode so it includes global regs as well.
36582         * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36583         rather than contains_regs_of_mode.
36585 2017-03-14  Martin Liska  <mliska@suse.cz>
36587         * doc/invoke.texi: Document options that can't be combined with
36588         -fcheck-pointer-bounds.
36590 2017-03-14  Martin Liska  <mliska@suse.cz>
36592         PR middle-end/79831
36593         * doc/invoke.texi (-Wchkp): Document the option.
36595 2017-03-14  Martin Liska  <mliska@suse.cz>
36597         * Makefile.in: Install gcov-dump.
36599 2017-03-14  Martin Liska  <mliska@suse.cz>
36601         * multiple_target.c (expand_target_clones): Bail out for
36602         an invalid attribute.
36604 2017-03-14  Richard Biener  <rguenther@suse.de>
36606         * alias.c (struct alias_set_entry): Pack properly.
36607         * cfgloop.h (struct loop): Likewise.
36608         * cse.c (struct set): Likewise.
36609         * ipa-utils.c (struct searchc_env): Likewise.
36610         * loop-invariant.c (struct invariant): Likewise.
36611         * lra-remat.c (struct cand): Likewise.
36612         * recog.c (struct change_t): Likewise.
36613         * rtl.h (struct address_info): Likewise.
36614         * symbol-summary.h (function_summary): Likewise.
36615         * tree-loop-distribution.c (struct partition): Likewise.
36616         * tree-object-size.c (struct object_size_info): Likewise.
36617         * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36618         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36619         * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36620         * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36621         * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36622         * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36623         (struct _stmt_vec_info): Likewise.
36625 2017-03-14  Martin Liska  <mliska@suse.cz>
36627         PR target/79892
36628         * multiple_target.c (create_dispatcher_calls): Check that
36629         a target can create a function dispatcher.
36631 2017-03-14  Martin Liska  <mliska@suse.cz>
36633         PR lto/66295
36634         * multiple_target.c (expand_target_clones): Drop local.local
36635         flag for default implementation.
36637 2017-03-14  Richard Biener  <rguenther@suse.de>
36639         PR tree-optimization/80030
36640         * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36642 2017-03-13  Kito Cheng  <kito.cheng@gmail.com>
36644         * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36645         gcc_fallthrough() instead of __attribute__((fallthrough));
36647 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36649         * doc/gcc.texi: Remove "up" link to (DIR).
36650         * doc/gccint.texi: Ditto.
36652 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36654         * doc/install.texi (Specific) <avr>: Remove reference to
36655         binutils 2.13.
36657 2017-03-13  Jeff Law  <law@redhat.com>
36659         * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36660         attribute rather than comments.
36662         * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36663         match_scratch operand is highest.
36665 2017-03-13  Martin Liska  <mliska@suse.cz>
36667         PR middle-end/78339
36668         * ipa-pure-const.c (warn_function_noreturn): If the declarations
36669         is a CHKP clone, use original declaration.
36671 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36673         * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36674         (arc_conditional_register_usage): Use a different allocation order
36675         when optimizing for size.
36676         * common/config/arc/arc-common.c (arc_option_optimization_table):
36677         Section anchors default on when optimizing for size.
36679 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36681         * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36683 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36685         * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36686         * config/arc/arc.md (cpu_facility): Add cd variant.
36687         (*movqi_insn): Add code density variant.
36688         (*movhi_insn): Likewise.
36689         (*movqi_insn): Likewise.
36690         (*addsi3_mixed): Likewise.
36691         (subsi3_insn): Likewise.
36693 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36695         * config/arc/arc.md (movsi_cond_exec): Update constraint.
36697 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36699         * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36700         expressions with MINUS and UNARY ops.
36702 2017-03-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36704         PR target/79911
36705         * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36706         Rename to...
36707         (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36708         between vec_select and vector argument.
36709         (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36710         (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36711         (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36712         (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36713         (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36714         (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36716 2017-03-13  Richard Biener  <rguenther@suse.de>
36718         PR other/79991
36719         * params.def (vect-max-peeling-for-alignment): Fix typo.
36721 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36723         * doc/install.texi (Specific) <mips-*-*>: Remove description of
36724         issue that only occurred with binutils below 2.18.
36726 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36728         * doc/install.texi (Specific) <cris-axis-elf>: No longer
36729         refer to binutils 2.11/2.12 minimum.
36731 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36733         * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36734         ftp.kernel.org and simplify binutils requirement.
36736 2017-03-11  Gerald Pfeifer  <gerald@pfeifer.com>
36738         * doc/invoke.texi (Warning Options): Fix spelling of link-time
36739         optimization.
36740         (Optimize Options): Ditto.  Also remove redundancy.
36742 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36744         PR translation/79848
36745         * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36746         "%qs".
36747         * ipa-pure-const.c (suggest_attribute): Likewise.  Convert _
36748         to G_ to avoid double translation.
36750 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36752         PR translation/79923
36753         * auto-profile.c (get_combined_location): Convert leading
36754         character of diagnostics to lower case and remove trailing period.
36755         (read_profile): Likewise for various diagnostics.
36756         * config/arm/arm.c (arm_option_override): Remove trailing period
36757         from various diagnostics.
36758         * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36759         (msp430_expand_delay_cycles): Likewise.
36761 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36763         PR target/79925
36764         * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36765         full command-line argument, rather than just "str".
36766         (aarch64_validate_march): Likewise.
36767         (aarch64_validate_mtune): Likewise.
36769 2017-03-10  Bernd Schmidt  <bschmidt@redhat.com>
36771         PR rtl-optimization/78911
36772         * lra-assigns.c (must_not_spill_p): New function.
36773         (spill_for): Use it.
36775 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36777         PR tree-optimization/79981
36778         * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36779         ATOMIC_COMPARE_EXCHANGE ifn result.
36780         (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36781         IFN_ATOMIC_COMPARE_EXCHANGE.
36783 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36785         PR driver/79875
36786         * opts.c (parse_sanitizer_options): Add missing question mark to
36787         "did you mean" message.
36789 2017-03-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36791         * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36792         built-in.
36793         (VMULEUH_UNS): Likewise.
36794         (VMULOUB_UNS): Likewise.
36795         (VMULOUH_UNS): Likewise.
36796         * config/rs6000/rs6000.c (builtin_function_type): Remove
36797         references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36799 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36801         PR bootstrap/79952
36802         * read-rtl-function.c (function_reader::read_rtx_operand): Update
36803         x with result of extra_parsing_for_operand_code_0.
36804         (function_reader::extra_parsing_for_operand_code_0): Convert
36805         return type from void to rtx, returning x.  When reading
36806         SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36807         larger size containing struct block_symbol.
36809 2017-03-10  Segher Boessenkool  <segher@kernel.crashing.org>
36811         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36812         -mfloat128-hardware without -m64.
36814 2017-03-10  Will Schmidt <will_schmidt@vnet.ibm.com>
36816         PR target/79941
36817         * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36818         entries to the case statement that marks unsigned arguments to
36819         overloaded functions.
36821 2017-03-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36823         * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36824         two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36826 2017-03-10  Pat Haugen  <pthaugen@us.ibm.com>
36828         PR target/79907
36829         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36830         TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36832 2017-03-10  Martin Liska  <mliska@suse.cz>
36834         PR target/65705
36835         PR target/69804
36836         * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36837         * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36838         FIELD != NULL.
36840 2017-03-10  Olivier Hainque  <hainque@adacore.com>
36842         * tree-switch-conversion (array_value_type): Start by resetting
36843         candidate type to it's main variant.
36845 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36847         PR rtl-optimization/79909
36848         * combine.c (try_combine): Use simplify_replace_rtx on individual
36849         CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36850         of the whole CALL_INSN_FUNCTION_USAGE.
36852         PR tree-optimization/79972
36853         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36854         get_range_info on SSA_NAMEs.  Formatting fixes.
36856 2017-03-10  Richard Biener  <rguenther@suse.de>
36857             Jakub Jelinek  <jakub@redhat.com>
36859         PR tree-optimization/77975
36860         * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36861         edge to be constant.
36862         (get_val_for): For constant x return it.  Formatting fix.
36863         (loop_niter_by_eval): Avoid pointless looping if the next iteration
36864         would use the same bases as the current one.
36866 2017-03-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36868         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36869         instead of vec_select for V1TImode.
36870         * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36871         longer needed.
36872         (VSX_LE_128): Add V1TI to this mode iterator.
36873         (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36874         (*vsx_le_perm_store_<mode>): Likewise.
36875         (pre-reload splitter for VSX stores): Likewise.
36876         (post-reload splitter for VSX stores): Likewise.
36877         (*vsx_xxpermdi2_le_<mode>): Likewise.
36878         (*vsx_lxvd2x2_le_<mode>): Likewise.
36879         (*vsx_stxvd2x2_le_<mode>): Likewise.
36881 2017-03-09  Michael Eager  <eager@eagercon.com>
36883         Correct failures with --enable-checking=yes,rtl.
36885         * config/microblaze/microblaze.c (microblaze_expand_shift):
36886         Replace GET_CODE test with CONST_INT_P and INTVAL test with
36887         test for const0_rtx.
36888         * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36889         lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36891 2017-03-09  Richard Biener  <rguenther@suse.de>
36893         PR tree-optimization/79977
36894         * graphite-scop-detection.c (scop_detection::merge_sese):
36895         Handle the case of extra exits to blocks dominating the entry.
36897 2017-03-09  Toma Tabacu  <toma.tabacu@imgtec.com>
36899         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
36900         Document rdynamic.
36902 2017-03-09  Vladimir Makarov  <vmakarov@redhat.com>
36904         PR rtl-optimization/79949
36905         * lra-constraints.c (process_alt_operands): Check memory when
36906         trying to predict a cycle.  Print about the overall increase.
36908 2017-03-09  Richard Biener  <rguenther@suse.de>
36910         PR middle-end/79971
36911         * gimple-expr.c (useless_type_conversion_p): Preserve
36912         TYPE_SATURATING for fixed-point types.
36914 2017-03-09  Richard Biener  <rguenther@suse.de>
36916         PR ipa/79970
36917         * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
36918         alignment of BLKmode params.
36920 2017-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36922         PR target/79913
36923         * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
36924         (VALL_NO_V2Q): Likewise.
36925         (VDQF_DF): Delete.
36926         * config/aarch64/aarch64-simd.md
36927         (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
36928         iterator.
36929         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
36930         VALL_NO_V2Q mode iterator.
36931         (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
36933 2017-03-09  Martin Liska  <mliska@suse.cz>
36935         PR tree-optimization/79631
36936         * tree-chkp-opt.c (chkp_is_constant_addr): Call
36937         tree_int_cst_sign_bit just for INTEGER constants.
36939 2017-03-09  Martin Liska  <mliska@suse.cz>
36941         PR target/65705
36942         PR target/69804
36943         * toplev.c (process_options): Disable -fcheck-pointer-bounds with
36944         sanitizers.
36946 2017-03-09  Marek Polacek  <polacek@redhat.com>
36948         PR c++/79672
36949         * tree.c (inchash::add_expr): Handle TREE_VEC.
36951 2017-03-09  Martin Liska  <mliska@suse.cz>
36953         PR ipa/79764
36954         (chkp_narrow_size_and_offset): New function.
36955         (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
36956         (void chkp_parse_bit_field_ref): New function.
36957         (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
36958         (chkp_process_stmt): Use chkp_parse_bit_field_ref.
36960 2017-03-09  Martin Liska  <mliska@suse.cz>
36962         PR ipa/79761
36963         * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
36964         (chkp_find_bounds_1): Remove gcc_unreachable.
36966 2017-03-09  Jakub Jelinek  <jakub@redhat.com>
36968         PR sanitizer/79944
36969         * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
36970         BUILT_IN_SYNC*, determine the access type from the size suffix and
36971         always build a MEM_REF with that type.  Handle forgotten
36972         BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
36974         PR target/79932
36975         * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
36976         _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
36977         _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
36978         _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
36979         _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
36980         _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
36981         _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
36982         _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
36983         _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
36984         _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
36985         _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
36986         _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
36987         _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
36988         _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
36989         _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
36990         _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
36991         _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
36992         _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
36993         _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
36994         _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
36995         _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
36996         _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
36997         _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
36998         _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
36999         _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
37000         _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
37001         _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
37002         _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
37003         _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
37004         _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
37005         definitions outside of __OPTIMIZE__ guarded section.
37007         PR target/79932
37008         * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
37009         _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
37010         _mm512_packus_epi32, _mm512_maskz_packus_epi32,
37011         _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
37012         guarded section.
37014 2017-03-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
37016         * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
37017         ("vfenez<mode>"): Add missing constraints.
37019 2017-03-08  Martin Sebor  <msebor@redhat.com>
37021         PR target/79928
37022         * config/nds32/nds32.c (nds32_option_override):
37023         Fix misspelled diagnostic.
37025 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37027         PR c/79940
37028         * gimplify.c (gimplify_omp_for): Replace index var in outer
37029         taskloop statement with an artificial variable and add
37030         OMP_CLAUSE_PRIVATE clause for it.
37032 2017-03-08  Richard Biener  <rguenther@suse.de>
37034         PR tree-optimization/79955
37035         * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
37036         for accesses that are completely outside of the variable.
37038 2017-03-08  Andrew Haley  <aph@redhat.com>
37040         PR tree-optimization/79943
37041         * tree-ssa-loop-split.c (compute_new_first_bound): When
37042         calculating the new upper bound, (END-BEG) should be added, not
37043         subtracted.
37045 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37047         * config/avr/avr.md (setmemhi): Make sure match_dup
37048         operand number comes before match_scratch.
37050 2017-03-08  Richard Biener  <rguenther@suse.de>
37052         PR tree-optimization/79920
37053         * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
37054         with ncopies == 1 to ...
37055         (vect_transform_slp_perm_load): ... here.  Properly compute
37056         all element loads by iterating VF times over the group.  Do
37057         not handle ncopies (computed in a broken way) in
37058         vect_create_mask_and_perm.
37060 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
37062         PR sanitizer/79904
37063         * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
37064         is a uniform vector, use uniform_vector_p return value instead of
37065         building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
37067 2017-03-07  Marek Polacek  <polacek@redhat.com>
37069         PR middle-end/79809
37070         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
37071         (alloca_call_type): Likewise.
37073 2017-03-07  Martin Liska  <mliska@suse.cz>
37075         * gcov.c (process_args): Put comment to correct location.
37077 2017-03-07  Martin Liska  <mliska@suse.cz>
37079         PR middle-end/68270
37080         * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
37081         Use array_at_struct_end_p instead of DECL_CHAIN (field).
37082         (chkp_narrow_bounds_for_field): Likewise.
37083         (chkp_parse_array_and_component_ref): Pass one more argument to
37084         call.
37086 2017-03-07  Richard Biener  <rguenther@suse.de>
37088         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
37089         preheaders.
37091 2017-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
37093         * config/i386/i386.c (ix86_local_alignment): Align most aggregates
37094         of 16 bytes and more to 16 bytes, not those of 16 bits and more.
37096 2017-03-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37098         PR c/79855
37099         * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
37100         to end of description.
37101         (PARAM_MAX_STORES_TO_MERGE): Likewise.
37103 2017-03-07  Jakub Jelinek  <jakub@redhat.com>
37105         PR rtl-optimization/79901
37106         * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
37107         ...
37108         (*avx512f_<code><mode>3<mask_name>): ... this.
37109         (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
37110         iterator instead of VI8_AVX2_AVX512BW.
37112         PR rtl-optimization/79901
37113         * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
37114         min/max expander, expand it using expand_vec_cond_expr.
37116         PR sanitizer/79897
37117         * ubsan.c (ubsan_encode_value): Call mark_addressable on the
37118         temporary.
37120 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37122         PR c++/79821
37123         * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
37124         to void * for PCH reasons.
37125         * dwarf2out.c (output_loc_operands, output_die): Cast
37126         v.val_vec.array to unsigned char *.
37128 2017-03-06  John David Anglin  <danglin@gcc.gnu.org>
37130         PR target/77850
37131         * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
37132         vector types.
37134 2017-03-06  Vladimir Makarov  <vmakarov@redhat.com>
37136         PR rtl-optimization/79571
37137         * lra-constraints.c (process_alt_operands): Calculate static
37138         reject and subtract it from overall when only addresses will be
37139         reloaded.
37141 2017-03-06  Julia Koval  <julia.koval@intel.com>
37143         PR target/79793
37144         * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
37145         incoming stack boundary to 128 for 64-bit targets.
37147 2017-03-06  Richard Biener  <rguenther@suse.de>
37149         PR tree-optimization/79894
37150         * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
37151         to NULL after folding it.
37153 2017-03-06  Richard Biener  <rguenther@suse.de>
37155         PR tree-optimization/79824
37156         * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
37157         check disabling peeling for gaps.
37159 2017-03-06  Toma Tabacu  <toma.tabacu@imgtec.com>
37161         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
37162         attributes): Document gettimeofday.
37164 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37166         * config/s390/s390.c (s390_option_override_internal): Set
37167         PARAM_MIN_VECT_LOOP_BOUND
37169 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37171         * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
37172         * config/s390/s390.md: Likewise.
37174 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37176         PR target/79812
37177         * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
37178         (<avx2_avx512>_perm<mode>): Rename to ...
37179         (avx2_perm<mode>): ... this.  Use VI8F_256 iterator instead
37180         of VI8F_256_512.
37181         (<avx512>_perm<mode>_mask): Rename to ...
37182         (avx512vl_perm<mode>_mask): ... this.  Use VI8F_256 iterator instead
37183         of VI8F_256_512.
37184         (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
37185         (avx2_perm<mode>_1<mask_name): ... this.  Use VI8F_256 iterator
37186         instead of VI8F_256_512.
37187         (avx512f_perm<mode>): New define_expand.
37188         (avx512f_perm<mode>_mask): Likewise.
37189         (avx512f_perm<mode>_1<mask_name>): New define_insn.
37190         (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
37192 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37194         * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
37195         msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
37196         if_then_else.
37197         (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
37199 2017-03-06  Martin Liska  <mliska@suse.cz>
37201         PR sanitize/79783
37202         * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
37203         when having a SSA NAME w/o VAR_DECL assigned to it.
37205 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37207         * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
37208         msa_dpsub_<su>_d): Fix MODE for vec_select.
37210 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37212         * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
37213         argument.
37214         * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
37216 2017-03-06  Richard Biener  <rguenther@suse.de>
37218         * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
37219         * plugin.c (register_plugin_info): Likewise.
37220         * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
37222 2017-03-05  Jakub Jelinek  <jakub@redhat.com>
37224         * config/i386/sse.md (sse_storehps, sse_storelps,
37225         avx_<castmode><avxsizesuffix>_<castmode>,
37226         avx512f_<castmode><avxsizesuffix>_<castmode>,
37227         avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37228         in condition that at least one operand is not a MEM.
37230 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37232         PR middle-end/79805
37233         * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37234         ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37235         ECF_NOTHROW.
37236         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37237         gimple_call_nothrow_p flag based on whether original builtin can throw.
37238         If it can, emit following stmts on the fallthrough edge.
37239         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37240         don't create new bb if inserting just debug stmts on the edge, try to
37241         insert them on the fallthru bb or just reset debug stmts.
37243 2017-03-03  Segher Boesssenkool  <segher@kernel.crashing.org>
37245         PR target/43763
37246         * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37247         restore recog_data (including the operand rtxes inside it) around
37248         the call to get_insn_template.
37250 2017-03-03  Martin Sebor  <msebor@redhat.com>
37252         PR tree-optimization/79699
37253         * context.c (context::~context): Free MPFR caches to avoid
37254         a memory leak on program exit.
37256 2017-03-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37258         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37259         Use wide_int::ulow () instead of .elt (0).
37261 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37263         * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37264         (*pushxf): Limit oF constraint to 32bit targets and add oC
37265         constraint for 64bit targets.
37266         (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37267         (*pushdf): Change rmF constraint to rmC.
37269 2017-03-03  Martin Liska  <mliska@suse.cz>
37271         * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37272         Remove unused variable.
37274 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37276         PR target/79807
37277         * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37278         is a memory operand, increase num_memory.
37279         (ix86_expand_args_builtin): Likewise.
37281 2017-03-03  Jan Hubicka  <jh@suse.cz>
37283         PR lto/79760
37284         * ipa-devirt.c (maybe_record_node): Properly handle
37285         __cxa_pure_virtual visibility.
37287 2017-03-03  Martin Liska  <mliska@suse.cz>
37289         PR tree-optimization/79803
37290         * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37291         assert.
37292         (pass_loop_prefetch::execute): Disabled optimization if an
37293         assumption about L1 cache size is not met.
37295 2017-03-03  Martin Liska  <mliska@suse.cz>
37297         PR rtl-optimization/79574
37298         * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37299         (hash_scan_set): Likewise.
37300         (dump_hash_table): Likewise.
37301         (hoist_code): Likewise.
37303 2017-03-03  Richard Biener  <rguenther@suse.de>
37305         * fixed-value.c (fixed_from_string): Restore use of elt (1)
37306         in place of uhigh ().
37307         (fixed_convert_from_real): Likewise.
37309 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37311         PR target/79514
37312         * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37314 2017-03-03  Richard Biener  <rguenther@suse.de>
37316         PR middle-end/79818
37317         * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37318         TYPE_OVERFLOW_UNDEFINED check.
37320 2017-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37322         * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37323         numbers.
37324         (vector_ae_<mode>_p): Likewise.
37325         (vector_nez_<mode>_p): Likewise.
37326         (vector_ne_v2di_p): Likewise.
37327         (vector_ae_v2di_p): Likewise.
37328         (vector_ne_<mode>_p): Likewise.
37329         * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37330         numbers.
37331         (vsx_tsqrt<mode>2_fe): Likewise.
37333 2017-03-02  Uros Bizjak  <ubizjak@gmail.com>
37335         PR target/79514
37336         * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37338 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37340         PR rtl-optimization/79780
37341         * cprop.c (one_cprop_pass): When second and further conditional trap
37342         in a single basic block is turned into an unconditional trap, turn it
37343         into a deleted note to avoid RTL verification failures.
37345 2017-03-02  Richard Biener  <rguenther@suse.de>
37347         * fold-const.c (const_binop): Use ulow () instead of elt (0).
37349 2017-03-02  Richard Biener  <rguenther@suse.de>
37351         PR tree-optimization/79345
37352         PR c++/42000
37353         * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37354         param and abort the walk, returning -1 if it is hit.
37355         (walk_aliased_vdefs): Take a limit param and pass it on.
37356         * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37357         defaulting to 0 and return a signed int.
37358         * tree-ssa-uninit.c (struct check_defs_data): New struct.
37359         (check_defs): New helper.
37360         (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37361         about uninitialized memory.
37362         * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37363         bogus uninitialized warning.
37364         (fixed_convert_from_real): Likewise.
37366 2017-03-02  Bin Cheng  <bin.cheng@arm.com>
37368         PR tree-optimization/66768
37369         * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37370         iv_use if base object can't be determined.
37372 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37374         PR tree-optimization/79345
37375         * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37376         * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37377         (get_pattern_stats): Initialize it.
37378         * genemit.c (gen_expand): Verify match_scratch numbers come after
37379         match_operand/match_dup numbers.
37380         * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37381         match_scratch numbers.
37382         * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37383         Likewise.
37384         * config/s390/s390.md (trunctdsd2): Likewise.
37386 2017-03-02  Richard Biener  <rguenther@suse.de>
37388         * wide-int.h (wide_int_storage::operator=): Implement in terms
37389         of wi::copy.
37391 2017-03-02  Richard Biener  <rguenther@suse.de>
37393         PR tree-optimization/79777
37394         * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37395         the to insert expression to sth existing.
37397 2017-03-01  Martin Sebor  <msebor@redhat.com>
37399         PR middle-end/79692
37400         * gimple-ssa-sprintf.c
37401         (directive::known_width_and_precision): New function.
37402         (format_integer): Use it.
37403         (get_mpfr_format_length): Consider the full range of precision
37404         when computing %g output with the # flag.  Set the likely byte
37405         count to 3 rather than 1 when precision is indeterminate.
37406         (format_floating): Correct the lower bound of precision.
37408 2017-03-01  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37410         * doc/invoke.texi: Document default code model for 64-bit Linux.
37412 2017-03-01  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
37414         PR target/79752
37415         * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37416         udiv rather than div since input pattern is unsigned.
37418 2017-03-01  Uros Bizjak  <ubizjak@gmail.com>
37420         * config/i386/i386.c (print_reg): Warn for values of
37421         unsupported size in integer register.
37423 2017-03-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
37425         PR target/79439
37426         * config/rs6000/predicates.md (current_file_function_operand): Do
37427         not allow self calls to be local if the function is replaceable.
37429 2017-03-01  Kelvin Nilsen  <kelvin@gcc.gnu.org>
37431         PR target/79395
37432         * config/rs6000/altivec.h (vec_ctz and others): Change the
37433         preprocessor macro that controls conditional compilation from
37434         _ARCH_PWR9 to __POWER9_VECTOR__.
37435         (vec_all_ne): Change parameterization of __altivec_scalar_pred
37436         macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37437         control (instead of _ARCH_PWR9 control) so that template
37438         definition uses power9-specific function.
37439         (vec_any_eq): Likewise.
37440         (vec_all_ne): Change macro definition to use a power9-specific
37441         expansion under #ifdef __POWER9_VECTOR__ control (instead of
37442         _ARCH_PWR9 control).
37443         (vec_any_eq) Likewise.
37444         * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37445         expansion for CMPNEF to remove support for xvcmpnesp instruction.
37446         (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37447         support for xvcmpnedp instruction.
37448         (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37449         macro expansion so that Power9 implementation of vec_all_ne does
37450         not use the AltiVec predicate framework.
37451         (VCMPNEH_P): Likewise.
37452         (VCMPNEW_P): Likewise.
37453         (VCMPNED_P): Likewise.
37454         (VCMPNEFP_P): Likewise.
37455         (VCMPNEDP_P): Likewise.
37456         (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37457         implementation of vec_any_eq to not use AltiVec predicate
37458         framework.
37459         (VCMPAEH_P): Likewise.
37460         (VCMPAEW_P): Likewise.
37461         (VCMPAED_P): Likewise.
37462         (VCMPAEFP_P): Likewise.
37463         (VCMPAEDP_P): Likewise.
37464         (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37465         BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37466         not use the AltiVec predicate framework.
37467         (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37468         of vec_any_eq to not use AltiVec predicate framework.
37469         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37470         support for predefined __POWER9_VECTOR__ macro to indicate that
37471         Power9 instruction selection is enabled.
37472         (altivec_overloaded_builtins): Remove extraneous
37473         ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37474         function argument types RS6000_BTI_bool_V16QI and
37475         RS6000_BTI_bool_V16QI.  Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37476         entry for overloaded function argument types RS6000_BTI_bool_V4SI
37477         andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB.  Remove
37478         two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37479         P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37480         Power9 for implementations of vec_cmpne.  Change the signature for
37481         all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37482         (representing vec_all_ne) to remove the previously described first
37483         argument of type RS6000_BTI_INTSI, as this was an artifact of
37484         reliance on the AltiVec predicate framework, which is no longer
37485         used in the implementation of these functions.  Add
37486         P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37487         function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37488         since, unlike the AltiVec predicate framework implementation, we
37489         do not share function descriptors between vec_alle and vec_anyeq.
37490         (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37491         set of modes that receive special treatment even when
37492         TARGET_P9_VECTOR is true.  The special treatment emits code that
37493         does not depend on Power9 instructions.
37494         * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37495         define_expand to not rely on AltiVec predicate framework.
37496         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37497         function.
37498         (vector_ne_v2di_p): Change this define_expand to not rely on
37499         AltiVec predicate framework.
37500         (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37501         function.
37502         (vector_ne_<mode>_p): Change this define_expand to not rely on
37503         AltiVec predicate framework.
37504         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37505         function.
37506         * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37507         (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37508         define_insn pattern.
37509         (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37510         define_insn pattern because the xvcmpne<VSs>. instruction is not
37511         supported.
37512         (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37513         instruction is not supported.
37515 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37517         * config/nvptx/nvptx.c: Include intl.h.
37519 2017-03-01  Martin Jambor  <mjambor@suse.cz>
37521         PR lto/78140
37522         * ipa-prop.h (ipa_bits): Removed field known.
37523         (ipa_jump_func): Removed field vr_known.  Changed fields bits and m_vr
37524         to pointers.  Adjusted their comments to warn about their sharing.
37525         (ipcp_transformation_summary): Change bits to a vector of pointers.
37526         (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37527         (ipa_get_ipa_bits_for_value): Declare.
37528         * tree-vrp.h (value_range): Mark as GTY((for_user)).
37529         * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37530         (ipa_bits_hash_table): Likewise.
37531         (ipa_vr_ggc_hash_traits): Likewise.
37532         (ipa_vr_hash_table): Likewise.
37533         (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37534         being pointers and vr_known being removed.
37535         (ipa_set_jf_unknown): Likewise.
37536         (ipa_get_ipa_bits_for_value): New function.
37537         (ipa_set_jfunc_bits): Likewise.
37538         (ipa_get_value_range): New overloaded functions.
37539         (ipa_set_jfunc_vr): Likewise.
37540         (ipa_compute_jump_functions_for_edge): Use the above functions to
37541         construct bits and vr parts of jump functions.
37542         (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37543         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37544         exist.
37545         (ipcp_grow_transformations_if_necessary): Also allocate
37546         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37547         exist.
37548         (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37549         them.  Fix too long lines.
37550         (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37551         vr_known being removed.
37552         (ipa_read_jump_function): Use new setter functions to construct bits
37553         and vr parts of jump functions or set them to NULL.
37554         (write_ipcp_transformation_info): Adjust for bits being pointers.
37555         (read_ipcp_transformation_info): Likewise.
37556         (ipcp_update_bits): Likewise.  Fix excessively long lines a trailing
37557         space.
37558         Include gt-ipa-prop.h.
37559         * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37560         being pointers.
37561         (ipcp_store_bits_results): Likewise.
37562         (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37563         Do not write to existing jump functions but use a temporary instead.
37565 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37567         PR c++/79681
37568         * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37569         attempt to use its first operand as BIT_FIELD_REF base.
37571 2017-03-01  Richard Biener  <rguenther@suse.de>
37573         PR middle-end/79721
37574         * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37575         interpolating formula in wrapping arithmetic.
37576         (chrec_apply): Convert chrec_evaluate return value to wanted type.
37578 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37580         PR tree-optimization/79734
37581         * tree-vect-generic.c (expand_vector_condition): Optimize
37582         AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37583         Handle VEC_COND_EXPR where comparison has different inner width from
37584         type's inner width.
37586 2017-02-28  Sandra Loosemore  <sandra@codesourcery.com>
37588         * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37589         markup, and similar issues.  Remove @opindex entries for things
37590         that aren't options.  Add missing -mmpy-option entries.
37592 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37594         PR tree-optimization/79737
37595         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37596         a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37597         tmpbuf[byte_size - 1].  Call natice_encode_expr with byte_size - 1
37598         instead of byte_size.  Formatting fix.
37599         (shift_bytes_in_array_right): Formatting fix.
37601 2017-02-28  Eric Botcazou  <ebotcazou@adacore.com>
37603         PR target/79749
37604         * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37605         condition on optimize for the leaf function test.
37607 2017-02-28  Martin Liska  <mliska@suse.cz>
37609         PR lto/79625
37610         * read-rtl-function.c (function_reader::handle_unknown_directive):
37611         Bail out when one uses -flto.
37613 2017-02-28  Martin Liska  <mliska@suse.cz>
37615         * common.opt: Replace space with tabular for options of <number>
37616         type.
37617         * config/i386/i386.opt: Show <number> value for
37618         -mlarge-data-threshold.
37619         * opts.c (print_filtered_help): Do not display number in hexadecimal
37620         format.
37622 2017-02-28  Martin Liska  <mliska@suse.cz>
37624         * common.opt: Fix --help=option -Q for options which are of
37625         an enum type.
37627 2017-02-28  Uros Bizjak  <ubizjak@gmail.com>
37629         * config/i386/i386.c (print_reg): Error out for values
37630         of 8-bit size in invalid integer register.
37632 2017-02-28  Martin Sebor  <msebor@redhat.com>
37634         PR tree-optimization/79691
37635         * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37637 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37639         PR target/79729
37640         * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37641         gcc_unreachable with output_operand_lossage.
37643 2017-02-28  Richard Biener  <rguenther@suse.de>
37645         PR tree-optimization/79740
37646         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37647         inserts.
37648         (visit_nary_op): Insert the nary into the hashtable if we
37649         pattern-matched sth.
37650         * tree-ssa-pre.c (eliminate_insert): Robustify.
37652 2017-02-28  Richard Biener  <rguenther@suse.de>
37654         PR middle-end/79731
37655         * fold-const.c (decode_field_reference): Reject out-of-bound
37656         accesses.
37658 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37660         * config/i386/i386.c: Include intl.h.
37661         (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37662         instead of just cond ? "..." : "...".
37663         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37664         * coverage.c (read_counts_file): Likewise.
37665         * omp-offload.c: Include intl.h.
37666         (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37667         of just cond ? "..." : "...".
37668         * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37669         of just cond ? "..." : "...".
37671 2017-02-28  Richard Earnshaw  <rearnsha@arm.com>
37673         PR target/79742
37674         * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37675         entry, if present.
37676         * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37677         'tune for' CPU name.
37678         * config/arm/arm-cpu-data.h: Regenerated.
37680 2017-02-28  Richard Biener  <rguenther@suse.de>
37682         PR tree-optimization/79732
37683         * tree-inline.c (expand_call_inline): Do not shadow var.
37685 2017-02-28  Richard Biener  <rguenther@suse.de>
37687         PR tree-optimization/79723
37688         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37689         address-space properly.
37691 2017-02-28  Thomas Schwinge  <thomas@codesourcery.com>
37693         * doc/optinfo.texi (Optimization groups): Fix option used for
37694         OPTGROUP_ALL.
37695         * doc/invoke.texi (-fopt-info): Document "omp".
37696         * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37697         (OPTGROUP_ALL): Add OPTGROUP_OMP.
37698         * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37699         * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37700         * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37702         * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP.  Adjust
37703         all users.
37704         * dumpfile.c (optgroup_options): Instead of "openmp", associate
37705         OPTGROUP_OMP with "omp".
37707 2017-02-27  Pat Haugen  <pthaugen@us.ibm.com>
37709         PR target/79544
37710         * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37711         for arithmetic shift of unsigned V2DI.
37713 2017-02-27  Claudiu Zissulescu  <claziss@synopsys.com>
37715         * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37716         arc/linux.h headers.
37717         * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37718         (LINK_SPEC): Likewise.
37719         (ARC_TLS_EXTRA_START_SPEC): Likewise.
37720         (EXTRA_SPECS): Likewise.
37721         (STARTFILE_SPEC): Likewise.
37722         (ENDFILE_SPEC): Likewise.
37723         (LIB_SPEC): Likewise.
37724         (TARGET_SDATA_DEFAULT): Likewise.
37725         (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37726         (MULTILIB_DEFAULTS): Likewise.
37727         (DWARF2_UNWIND_INFO): Likewise.
37728         * config/arc/big.h: New file.
37729         * config/arc/elf.h: Likewise.
37730         * config/arc/linux.h: Likewise.
37731         * config/arc/t-uClibc: Remove.
37733 2017-02-27  Bin Cheng  <bin.cheng@arm.com>
37735         PR tree-optimization/77536
37736         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37737         (tree_transform_and_unroll_loop): Use above function to compute the
37738         estimated niter of unrolled loop and use it when scaling profile.
37739         Also use count info rather than frequency if it's non-zero.
37740         * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37741         * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37742         (vect_transform_loop): Call above function.
37744 2017-02-27  Richard Biener  <rguenther@suse.de>
37746         PR tree-optimization/45397
37747         * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37748         * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37749         (visit_nary_op): Add pattern matching for CSEing sign-changed
37750         or truncated operations with wider ones.
37752 2017-02-27  Richard Biener  <rguenther@suse.de>
37754         PR tree-optimization/79690
37755         * tree-vect-stmts.c (vectorizable_store): Use vector type
37756         built from the DR with address-space.
37758 2017-02-26  Gerald Pfeifer  <gerald@pfeifer.com>
37760         * doc/invoke.texi (Optimize Options): Refine the description
37761         of asan-use-after-return.
37763 2017-02-25  Alan Modra  <amodra@gmail.com>
37765         PR rtl-optimization/79584
37766         * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37767         base, not ad->base_term, the reg within base.  Remove assertion
37768         that ad->base == ad->base_term.  Replace gen_int_mode using
37769         bogus mode with const0_rtx.
37771 2017-02-25  Jakub Jelinek  <jakub@redhat.com>
37773         PR middle-end/79396
37774         * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37775         FMA_EXPR like tcc_binary or tcc_unary.
37777         * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37779         PR debug/77589
37780         * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37781         bitfield.
37782         (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37783         (output_loc_operands): Handle DW_OP_call_ref and
37784         DW_OP_GNU_variable_value.
37785         (struct variable_value_struct): New type.
37786         (struct variable_value_hasher): Likewise.
37787         (variable_value_hash): New variable.
37788         (string_types): Remove.
37789         (copy_loc_descr): New function.
37790         (add_loc_descr_to_each): Clarify comment.  Use copy_loc_descr.
37791         (prepend_loc_descr_to_each): New function.
37792         (add_loc_list): Fix comment typo.  Use prepend_loc_descr_to_each
37793         instead of add_loc_descr_to_each if the first argument is single
37794         location list and the second has multiple.
37795         (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37796         (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37797         when looking for variable value which doesn't have other location info.
37798         (loc_list_from_tree): Formatting fix.
37799         (gen_array_type_die): Simplify DW_AT_string_length handling.
37800         (adjust_string_types): Remove.
37801         (gen_subprogram_die): Don't call adjust_string_types nor test/set
37802         string_types.  Call resolve_variable_values.
37803         (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37804         (resolve_addr_in_expr): Likewise.  Add A argument.
37805         (copy_deref_exprloc): Remove deref argument.  Adjust for the
37806         original expression being DW_OP_GNU_variable_value with optionally
37807         DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37808         optionally after it.
37809         (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37810         (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37811         callers.  Set remove_AT_byte_size if removing DW_AT_string_length.
37812         (variable_value_hasher::hash, variable_value_hasher::equal): New
37813         methods.
37814         (resolve_variable_value_in_expr, resolve_variable_value,
37815         resolve_variable_values, note_variable_value_in_expr,
37816         note_variable_value): New functions.
37817         (dwarf2out_early_finish): Call note_variable_value on all toplevel
37818         DIEs.
37820 2017-02-24  Jakub Jelinek  <jakub@redhat.com>
37822         PR c/79677
37823         * opts.h (handle_generated_option): Add GENERATED_P argument.
37824         * opts-common.c (handle_option): Adjust function comment.
37825         (handle_generated_option): Add GENERATED_P argument, pass it to
37826         handle_option.
37827         (control_warning_option): Pass false to handle_generated_option
37828         GENERATED_P.
37829         * opts.c (maybe_default_option): Pass true to handle_generated_option
37830         GENERATED_P.
37831         * optc-gen.awk: Likewise.
37833 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37835         * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37836         a REG, look at the REG it is a SUBREG of.
37837         (splitter for cmpeqsi_t): Ditto.
37839 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37841         * config/pa/pa.c (pa_combine_instructions): Do not share RTL.  Make
37842         the special USEs with the pattern of the insn, not the insn itself.
37844 2017-02-24  Matthew Fortune  <matthew.fortune@imgtec.com>
37846         PR target/79473
37847         * doc/invoke.texi: Document -mload-store-pairs.
37849 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37850             Sandra Loosemore  <sandra@codesourcery.com>
37852         * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37853         argument isn't a CONST_INT.
37854         (nios2_alternate_compare_const): Assert op is a CONST_INT.
37855         (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37856         (nios2_validate_compare): Bypass alternate compare logic if *op2
37857         is not a CONST_INT.
37858         (ldstwm_operation_p): Return false if first_base is not a REG or
37859         if first_offset is not a CONST_INT.
37861 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37863         * config/cris/cris.md: Use correct operand in a define_peephole2.
37865 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37867         * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37869 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37871         * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37872         this_insn if it is an INSN or JUMP_INSN.
37873         (force_offsettable): Look at base, not at addr.
37874         * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37875         on things that aren't necessarily CONST_INTs.
37877 2017-02-24  Uros Bizjak  <ubizjak@gmail.com>
37879         * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37880         -mfpmath=sse is the default also for x86-32 targets with SSE2
37881         instruction set when @option{-ffast-math} is enabled
37883 2017-02-24  Jeff Law  <law@redhat.com>
37885         PR rtl-optimizatoin/79286
37886         * ira.c (update_equiv_regs): Drop may_trap_p exception to
37887         dominance test.
37889 2017-02-24  Richard Biener  <rguenther@suse.de>
37891         PR tree-optimization/79389
37892         * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37893         debug insns.
37895 2017-02-24  Aldy Hernandez  <aldyh@redhat.com>
37897         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
37898         function comment to reflect reality.
37899         (loop_exits_before_overflow): Fix typo in function description.
37901 2017-02-24  Richard Biener  <rguenther@suse.de>
37903         PR tree-optimization/79389
37904         * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
37905         properly that a threading opportunity exists.  Detect conditional
37906         copy/constant propagation opportunities.
37908 2017-02-23  Eric Botcazou  <ebotcazou@adacore.com>
37910         * config/visium/visium.md (type): Add trap.
37911         (b): New mode attribute.
37912         (*btst): Rename into...
37913         (*btst<mode>): ...this and adjust.
37914         (*cbranchsi4_btst_insn): Rename into...
37915         (*cbranch<mode>4_btst_insn): ...this and adjust.
37916         (trap): New define_insn.
37918 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37920         PR tree-optimization/79389
37921         * ifcvt.c (struct noce_if_info): Add rev_cond field.
37922         (noce_reversed_cond_code): New function.
37923         (noce_emit_store_flag): Use rev_cond if non-NULL instead of
37924         reversed_comparison_code.  Formatting fix.
37925         (noce_try_store_flag): Test rev_cond != NULL in addition to
37926         reversed_comparison_code.
37927         (noce_try_store_flag_constants): Likewise.
37928         (noce_try_store_flag_mask): Likewise.
37929         (noce_try_addcc): Use rev_cond if non-NULL instead of
37930         reversed_comparison_code.
37931         (noce_try_cmove_arith): Likewise.  Formatting fixes.
37932         (noce_try_minmax, noce_try_abs): Clear rev_cond.
37933         (noce_find_if_block): Initialize rev_cond.
37934         (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
37935         instead of false as last argument never attempt to reverse it
37936         afterwards.
37938 2017-02-23  Bin Cheng  <bin.cheng@arm.com>
37940         PR tree-optimization/79663
37941         * tree-predcom.c (combine_chains): Process refs in reverse order
37942         only for ZERO length chains, and add explaining comment.
37944 2017-02-23  Jeff Law  <law@redhat.com>
37946         PR tree-optimization/79578
37947         * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
37948         in call to operand_equal_p.
37950 2017-01-23  Dominique d'Humieres  <dominiq@lps.ens.fr>
37952         PR target/71017
37953         * config/i386/cpuid.h: Fix another undefined behavior.
37955 2017-02-23  Richard Biener  <rguenther@suse.de>
37957         PR tree-optimization/79683
37958         * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
37959         vector types for data-refs.
37961 2017-02-23  Martin Liska  <mliska@suse.cz>
37963         * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
37965 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37967         PR middle-end/79665
37968         * internal-fn.c (get_range_pos_neg): Moved to ...
37969         * tree.c (get_range_pos_neg): ... here.  No longer static.
37970         * tree.h (get_range_pos_neg): New prototype.
37971         * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
37972         are known to be in between 0 and signed maximum inclusive, try to
37973         expand both unsigned and signed divmod and use the cheaper one from
37974         those.
37976 2017-02-22  Jeff Law  <law@redhat.com>
37978         PR tree-optimization/79578
37979         * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
37980         to compare base operands.
37982 2017-02-22  Segher Boessenkool  <segher@kernel.crashing.org>
37984         PR target/79211
37985         * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
37986         gpc_reg_operand instead of fpr_reg_operand.
37988 2017-02-22  Sameera Deshpande  <sameera.deshpande@imgtec.com>
37990         * config/mips/mips.c (mips_return_in_memory): Force FP
37991         vector types to be returned in memory for o32 ABI.
37993 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37995         * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
37996         instead of DW_TAG_member for static data member declarations and don't
37997         set no_linkage_name for static inline data members.
37998         (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
37999         to DW_TAG_member.
38001 2017-02-22  Martin Liska  <mliska@suse.cz>
38003         * doc/invoke.texi: Replace inequality signs with square brackets
38004         for -Wnormalized.
38006 2017-02-22  Matthew Fortune  <matthew.fortune@imgtec.com>
38008         PR target/78660
38009         * lra-constraints.c (simplify_operand_subreg): Handle
38010         WORD_REGISTER_OPERATIONS targets.
38012 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
38014         PR target/70465
38015         * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
38016         and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
38017         elimination by swapping fld*.
38019 2017-02-22  Richard Biener  <rguenther@suse.de>
38021         PR tree-optimization/79673
38022         * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
38023         convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
38024         irrelevant address-space qualifiers and avoiding a
38025         ADDR_SPACE_CONVERT_EXPR from fold_convert.
38027 2017-02-22  Richard Biener  <rguenther@suse.de>
38029         PR tree-optimization/79666
38030         * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
38031         to not symbolically negate if that may introduce undefined
38032         overflow.
38034 2017-02-22  Martin Liska  <mliska@suse.cz>
38036         PR lto/79587
38037         * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
38038         * data-streamer-out.c (streamer_write_gcov_count_stream):
38039         Likewise.
38040         * value-prof.c (stream_out_histogram_value): Make assert more
38041         precise based on type of counter.
38043 2017-02-21  Uros Bizjak  <ubizjak@gmail.com>
38045         PR target/79593
38046         * config/i386/i386.md (standard_x87sse_constant_load splitter):
38047         Use nonimmediate_operand instead of memory_operand for operand 1.
38048         (float-extend standard_x87sse_constant_load splitter): Ditto.
38050 2017-02-21  Jeff Law  <law@redhat.com>
38052         PR tree-optimization/79621
38053         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
38054         blocks with edges to themselves.
38056 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38058         PR target/79633
38059         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
38060         is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
38061         Use gimple_call_builtin_p.
38063         PR target/79570
38064         * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
38065         on temporarily removed DEBUG_INSNs.
38067         PR tree-optimization/79649
38068         * tree-loop-distribution.c (classify_partition): Give up on
38069         non-generic address space loads/stores.
38071 2017-02-21  Aldy Hernandez  <aldyh@redhat.com>
38073         * doc/loop.texi (Loop manipulation): Remove nonexistent
38074         tree_ssa_loop_version from the documentation.
38075         * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
38077 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38079         PR target/79494
38080         * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
38081         make_reg_eh_region_note_nothrow_nononlocal on call_insn.
38082         * config/rs6000/rs6000.c: Include except.h.
38083         (rs6000_expand_split_stack_prologue): Call
38084         make_reg_eh_region_note_nothrow_nononlocal on the call insn.
38086 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38088         PR lto/79579
38089         * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
38090         have been analyzed.
38092 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38094         * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
38095         for backward compatibility only.
38096         * doc/invoke.texi (Option Summary): Remove all references to
38097         -fipa-cp-alignment.
38099 2017-02-21  Matthew Fortune  <matthew.fortune@imgtec.com>
38101         PR target/78660
38102         Revert:
38103         2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38105         * lra-constraints.c (curr_insn_transform): Handle
38106         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38108 2017-02-21  Martin Liska  <mliska@suse.cz>
38110         * config/i386/i386.opt: Replace -masm-dialect with -masm.
38112 2017-02-21  Thomas Schwinge  <thomas@codesourcery.com>
38114         PR translation/79638
38115         * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
38117 2017-02-21  Eric Botcazou  <ebotcazou@adacore.com>
38119         PR ada/67205
38120         * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
38121         (arm_function_ok_for_sibcall): Return false for an indirect call by
38122         descriptor if all the argument registers are used.
38123         (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
38124         alignment of the function.
38126 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38128         PR tree-optimization/61441
38129         * simplify-rtx.c (simplify_const_unary_operation): For
38130         -fsignaling-nans and sNaN operand, return NULL_RTX rather than
38131         the sNaN unmodified.
38133 2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
38135         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
38136         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
38137         instead of SYSTEM_HEADER_DIR.
38139 2017-02-20  Gerald Pfeifer  <gerald@pfeifer.com>
38140             Martin LiÅ¡ka  <mliska@suse.cz>
38142         * doc/invoke.texi (use-after-scope-direct-emission-threshold):
38143         Fix typos and grammar, use active voice, and clarify.
38145 2017-02-20  Marek Polacek  <polacek@redhat.com>
38147         PR middle-end/79537
38148         * gimplify.c (gimplify_expr): Handle unused *&&L;.
38150         PR sanitizer/79558
38151         * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
38153 2017-02-20  Jakub Jelinek  <jakub@redhat.com>
38155         PR target/79568
38156         * config/i386/i386.c (ix86_expand_builtin): Handle
38157         OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
38158         ix86_builtins_isa[fcode].isa as a requirement of those
38159         flags and any other flag in the bitmask.
38160         (ix86_init_mmx_sse_builtins): Use 0 instead of
38161         ~OPTION_MASK_ISA_64BIT as mask.
38162         * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
38163         __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
38164         __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
38165         __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
38167 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38169         PR target/78012
38170         * lra-constraints.c (split_reg): Check requested split mode
38171         is supported by the register.
38173 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38175         * lra-constraints.c (simplify_operand_subreg): Remove early
38176         return false.
38178 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38180         PR target/78660
38181         * lra-constraints.c (curr_insn_transform): Tighten condition
38182         for converting SUBREG reloads from OP_OUT to OP_INOUT.
38184 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38186         PR target/78660
38187         * lra-constraints.c (curr_insn_transform): Handle
38188         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38190 2017-02-19  Uros Bizjak  <ubizjak@gmail.com>
38192         Revert:
38193         2016-05-30  Uros Bizjak  <ubizjak@gmail.com>
38195         * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
38197 2017-02-19  Jonathan Wakely  <jwakely@redhat.com>
38199         PR c++/69523
38200         * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
38201         description.
38203 2017-02-19  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38205         * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
38206         for FMA_EXPR.
38208 2017-02-18  Jakub Jelinek  <jakub@redhat.com>
38210         * final.c (last_columnnum, override_columnnum): New variables.
38211         (final_start_function): Set last_columnnum, pass it to begin_prologue
38212         hook and pass 0 to dwarf2out_begin_prologue.
38213         (final_scan_insn): Update override_columnnum.  Pass last_columnnum
38214         to source_line debug hook.
38215         (notice_source_line): Compute last_columnnum and for debug_column_info
38216         return true on column changes.
38217         * debug.h (struct gcc_debug_hooks): Add column argument to
38218         source_line and begin_prologue hooks.
38219         (debug_nothing_int_charstar_int_bool): Remove prototype.
38220         (debug_nothing_int_int_charstar,
38221         debug_nothing_int_int_charstar_int_bool): New prototypes.
38222         (dwarf2out_begin_prologue): Add column argument.
38223         * debug.c (do_nothing_debug_hooks): Adjust source_line and
38224         begin_prologue hooks.
38225         (debug_nothing_int_charstar_int_bool): Remove.
38226         (debug_nothing_int_int_charstar,
38227         debug_nothing_int_int_charstar_int_bool): New functions.
38228         * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38229         through to dwarf2out_source_line.
38230         (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38231         (dwarf2out_source_line): Add column argument, emit it if requested.
38232         * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38233         arguments.
38234         * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38235         * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38236         * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38237         through to dwarf2out_begin_prologue.
38238         (vmsdbgout_source_line): Add column argument, pass it through to
38239         dwarf2out_source_line.
38240         * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38241         dbxout_source_line caller.
38242         (dbxout_source_line): Add column argument.
38244         * common.opt (gno-column-info, gcolumn-info): New options.
38245         * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38246         (check_die): Also test for multiple DW_AT_decl_column attributes.
38247         (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38248         DW_AT_decl_column if requested.
38249         (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38250         if requested.
38251         (gen_variable_die): Likewise.
38252         (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38253         * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38255         PR target/79569
38256         * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38257         * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38258         (ix86_handle_option): Handle OPT_m3dnowa.
38259         * doc/invoke.texi (-m3dnowa): Document.
38260         * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38261         -m3dnowa instead of -m3dnow -march=athlon.
38263         PR target/79559
38264         * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38265         instead of gcc_assert for K, r and R code checks.  Formatting fixes.
38267 2017-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
38269         PR target/79261
38270         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38271         support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38272         * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38273         generator for vsx_xxpermdi_<mode>_be.
38274         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38275         force big-endian semantics.
38276         (vsx_xxpermdi_<mode>_be): New define_expand with same
38277         implementation as previous version of vsx_xxpermdi_<mode>.
38279 2017-02-17  Jakub Jelinek  <jakub@redhat.com>
38281         PR tree-optimization/79327
38282         * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38283         variable, its initialization and use.
38285 2017-02-17  Julia Koval  <julia.koval@intel.com>
38287         * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38288         (OPTION_MASK_ISA_PKU_UNSET): New.
38289         (ix86_handle_option): Handle -mrdpid.
38290         * config/i386/cpuid.h (bit_RDPID): New.
38291         * config/i386/driver-i386.c (host_detect_local_cpu):
38292         Detect RDPID feature.
38293         * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38294         * config/i386/i386-c.c (ix86_target_macros_internal):
38295         Handle RDPID flag.
38296         * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38297         (ix86_valid_target_attribute_inner_p): Add "rdpid".
38298         (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38299         * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38300         * config/i386/i386.md (define_insn "rdpid"): New.
38301         * config/i386/i386.opt Add -mrdpid.
38302         * config/i386/immintrin.h (_rdpid_u32): New.
38304 2017-02-17  Vladimir Makarov  <vmakarov@redhat.com>
38306         PR rtl-optimization/79541
38307         * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38308         instead of transforming it into USE.
38310 2017-02-17  Segher Boessenkool  <segher@kernel.crashing.org>
38312         * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38313         If HONOR_SNANS (SFmode) force the input to a register.
38314         (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38315         (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38316         an frsp or similar insn.
38318 2017-02-17  Martin Liska  <mliska@suse.cz>
38320         PR rtl-optimization/79577
38321         * params.def (selsched-max-sched-times): Increase minimum to 1.
38323 2017-02-17  Martin Liska  <mliska@suse.cz>
38325         PR rtl-optimization/79574
38326         * gcse.c (want_to_gcse_p): Prevent integer overflow.
38328 2017-02-17  Martin Liska  <mliska@suse.cz>
38330         PR tree-optimization/79529
38331         * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38332         ssa_defined_default_def_p to handle cases which are implicitly
38333         defined.
38334         * tree-ssa.c (ssa_defined_default_def_p): New function.
38335         (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38336         which are implicitly defined.
38337         * tree-ssa.h (ssa_defined_default_def_p): Declare.
38339 2017-02-17  Richard Biener  <rguenther@suse.de>
38341         PR middle-end/79576
38342         * params.def (max-ssa-name-query-depth): Limit to 10.
38344 2017-02-17  Richard Biener  <rguenther@suse.de>
38346         PR tree-optimization/79552
38347         * tree-ssa-structalias.c (visit_loadstore): Properly verify
38348         default defs.
38350 2017-02-17  Richard Biener  <rguenther@suse.de>
38352         PR bootstrap/79567
38353         * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38355 2017-02-17  Marek Polacek  <polacek@redhat.com>
38357         PR middle-end/79536
38358         * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38359         (fold_negate_expr): New wrapper.
38361 2017-02-16  Sandra Loosemore  <sandra@codesourcery.com>
38363         * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]: 
38364         Correct terminology and de-emphasize pre-standard behavior.
38366 2017-02-16  Alan Modra  <amodra@gmail.com>
38368         PR rtl-optimization/79286
38369         * ira.c (def_dominates_uses): New function.
38370         (update_equiv_regs): Don't create an equivalence for insns that
38371         may trap where the register def does not dominate the use.
38373 2017-02-16  Vladimir Makarov  <vmakarov@redhat.com>
38375         PR rtl-optimization/78127
38376         * lra.c (lra): Call lra_eliminate before finish the loop after
38377         lra_constraint.
38379 2017-02-16  Richard Biener  <rguenther@suse.de>
38381         * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38382         isl/isl_val.h.
38383         * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38384         (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38385         * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38386         (isl_val_int_from_wi): New function.
38387         (extract_affine_gmp): Rename to ...
38388         (extract_affine_wi): ... this, take a widest_int.
38389         (extract_affine_int): Just wrap extract_affine_wi.
38390         (add_param_constraints): Use isl_val_int_from_wi.
38391         (add_loop_constraints): Likewise, and extract_affine_wi.
38393 2017-02-15  Jeff Law  <law@redhat.com>
38395         PR middle-end/79521
38396         * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38397         ira_init_register_move_cost_if_necessary.
38399 2017-02-15  Martin Sebor  <msebor@redhat.com>
38401         PR middle-end/32003
38402         * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38403         removed in a prior commit.
38405 2017-02-15  Bin Cheng  <bin.cheng@arm.com>
38407         PR tree-optimization/79347
38408         * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38409         counters during peeling.
38411 2017-02-15  Thomas Schwinge  <thomas@codesourcery.com>
38413         * Makefile.in (site.exp): Remove "set ISLVER".
38415 2017-02-15  Jakub Jelinek  <jakub@redhat.com>
38417         PR target/79487
38418         * real.c (real_from_integer): Call real_convert even for decimal.
38420 2017-02-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38422         PR target/79421
38423         * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38425 2017-02-14  Andrew Pinski  <apinski@cavium.com>
38427         * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38428         cores and change the partno/implementer to be correct.
38429         (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38430         the 'B" as the implementer.
38431         * config/aarch64/aarch64-tune.md: Regenerate.
38433 2017-02-14  Carl Love  <cel@us.ibm.com>
38435         * config/rs6000/rs6000.c: Add case statement entry to make the
38436         xvcvuxdsp built-in argument unsigned.
38437         * config/rs6000/vsx.md: Fix the source and return operand types so they
38438         match the instruction definitions from the ISA document.  Fix typo
38439         in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38440         statement.
38442 2017-02-14  Vladimir Makarov  <vmakarov@redhat.com>
38444         PR target/79282
38445         * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38446         member early_clobber_alts.
38447         * lra-lives.c (reg_early_clobber_p): New.
38448         (process_bb_lives): Use it.
38449         * lra.c (new_insn_reg): New arg early_clobber_alts.  Use it.
38450         (debug_operand_data): Initialize early_clobber_alts.
38451         (setup_operand_alternative): Set up early_clobber_alts.
38452         (collect_non_operand_hard_regs): Ditto.  Pass early clobber
38453         alternatives to new_insn_reg.
38454         (add_regs_to_insn_regno_info): Add arg early_clobber_alts.  Use
38455         it.
38456         (lra_update_insn_regno_info): Pass the new arg.
38458 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38460         PR middle-end/79505
38461         * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38462         (new_oacc_loop_raw): Don't clear already cleared fields.
38464         PR target/79481
38465         * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38466         _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38467         _mm512_prefetch_i64gather_ps): New inline functions and macros.
38469 2017-02-14  Uros Bizjak  <ubizjak@gmail.com>
38471         PR target/79495
38472         * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38474 2017-02-14  H.J. Lu  <hongjiu.lu@intel.com>
38476         PR target/79498
38477         * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38478         the extra instruction to the right place to store 128-bit constant
38479         when needed.
38481 2017-02-14  Martin Sebor  <msebor@redhat.com>
38483         PR middle-end/79448
38484         * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38485           warning for strings of unknown length.
38487 2017-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
38489         * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38491 2017-02-14  Jeff Law  <law@redhat.com>
38493         PR target/79404
38494         * ira-costs.c (scan_one_insn): Initialize register move costs
38495         for pseudos seen in USE/CLOBBER insns.
38497         PR tree-optimization/79095
38498         * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38499         if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38500         (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38501         if the operands are known to be not equal, then the resulting range
38502         is ~[0,0].
38503         (intersect_ranges): If the new range is ~[0,0] and the old range is
38504         wide, then prefer ~[0,0].
38505         * tree-vrp.c (overflow_comparison_p_1): New function.
38506         (overflow_comparison_p): New function.
38507         * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38508         if NAME is used in an overflow test.
38509         (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38510         overflow check that can be expressed as an equality test, then adjust
38511         ops to be that equality test.
38513 2017-02-14  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38515         * config/s390/s390-builtin-types.def: Remove flags argument.
38516         * config/s390/s390.c (s390_init_builtins): Likewise.
38518 2017-02-14  Martin Liska  <mliska@suse.cz>
38520         * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38521         vector.  Fix trailing white spaces.
38523 2017-02-14  James Greenhalgh  <james.greenhalgh@arm.com>
38525         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38526         HFmode.
38528 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38530         PR rtl-optimization/68664
38531         * config/arm/arm.c (arm_sched_can_speculate_insn):
38532         New function.  Declare prototype.
38533         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38535 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38537         PR rtl-optimization/68664
38538         * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38539         New function.
38540         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38542 2017-02-14  Amit Pawar  <amit.pawar@amd.com>
38544         * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38545         max skip bytes for function, loop and jump.
38547 2017-02-14  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38549         * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38550         ABS_EXPR for gimple dump.
38552 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38554         PR target/79462
38555         * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38557         PR tree-optimization/79408
38558         * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38559         case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38560         (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38561         also if rhs1 is INTEGER_CST.
38563 2017-02-14  Richard Biener  <rguenther@suse.de>
38565         PR middle-end/79432
38566         * tree-into-ssa.c (insert_phi_nodes): When the function can
38567         have abnormal edges rewrite SSA names with broken use-def
38568         dominance out of SSA and register them for PHI insertion.
38570 2017-02-13  Martin Sebor  <msebor@redhat.com>
38572         PR middle-end/79496
38573         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38574         clearing info.nowrite flag when snprintf size argument is a range.
38576 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38578         * cprop.c (cprop_jump): Add missing space in string literal.
38579         * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38580         (get_constraint_for_component_ref): Likewise.
38581         * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38582         * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38583         * lra-constraints.c (process_alt_operands): Likewise.
38584         * ipa-inline.c (inline_small_functions): Likewise.
38585         * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38586         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38587         * trans-mem.c (diagnose_tm_1_op): Likewise.
38588         * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38589         (grid_parallel_clauses_gridifiable): Likewise.
38591         * config/nvptx/mkoffload.c (process): Add space in between
38592         , and %d.
38594         * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38595         "MOD4_SSE_REGS" and "ALL_REGS".
38597         * spellcheck.c (test_data): Add , in between "foo" and "food".
38599 2017-02-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38601         PR target/79449
38602         * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38603         boundary crossing check and subsequent code generation agree.
38605 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38607         * config/aarch64/aarch64.c (has_memory_op): Delete.
38608         (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38609         has_memory_op.
38611 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38613         PR rtl-optimization/79388
38614         PR rtl-optimization/79450
38615         * combine.c (distribute_notes): When removing TEM_INSN for which
38616         corresponding dest has last value recorded, invalidate that last
38617         value.
38619 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38621         * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38622         of explicit '@'.  Add missing assembly comment marker on branch costs
38623         printout.
38625 2017-02-13  Nathan Sidwell  <nathan@acm.org>
38627         * gengtype-lex.l (<in_struct>): Add '/'.
38629 2017-02-13  Martin Liska  <mliska@suse.cz>
38631         PR c/79471
38632         * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38634 2017-02-13  Richard Biener  <rguenther@suse.de>
38636         * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38637         Remove.
38638         * configure: Re-generate.
38639         * config.in: Likewise.
38640         * graphite-dependences.c: Simplify as if
38641         HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38642         * graphite-isl-ast-to-gimple.c: Likewise.
38643         * graphite-optimize-isl.c: Likewise.
38644         * graphite-poly.c: Likewise.
38645         * graphite-sese-to-poly.c: Likewise.
38646         * graphite.h: Likewise.
38647         * toplev.c: Include isl/version.h and use isl_version () for
38648         printing the ISL version.
38649         * doc/install.texi: Update ISL requirement.
38651 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38653         * doc/standards.texi (Standards): Update reference to
38654         Objective-C 2.0.
38656 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38658         * doc/extend.texi (Named Address Spaces): sourceware.org now
38659         defaults to https.
38660         * doc/install.texi (Binaries): Ditto.
38661         (Specific): Ditto.
38663 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38665         * doc/cpp.texi: Replace "stringify"/"stringification" with C
38666         standard terminology "stringize"/"stringizing" throughout.
38667         * doc/cppinternals.texi: Likewise.
38669 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38671         * doc/extend.texi: Fix some spelling mistakes and typos.
38672         * doc/invoke.texi: Likewise.
38674 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38676         PR ipa/79224
38677         * params.def (inline-min-speedup) Change from 10 to 8.
38679 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38681         * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38682         4.5.
38684 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38686         PR ipa/79224
38687         * ipa-inline-analysis.c (get_minimal_bb): New function.
38688         (record_modified): Use it.
38689         (remap_edge_change_prob): Handle also ancestor functions.
38691 2017-02-11  Gerald Pfeifer  <gerald@pfeifer.com>
38693         * doc/contrib.texi (Contributors): Remove broken link into
38694         the Mauve CVS repository.
38696 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38698         PR middle-end/79454
38699         * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38700         result computation whenever lhs doesn't have vector mode, not
38701         just when it has BLKmode.
38703 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38705         * doc/makefile.texi (profiledbootstrap): Refer to the
38706         installation instructions only in textual form.
38708 2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38710         PR target/79295
38711         * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38713 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38715         * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38716         (Specific): Update mingw-w64 reference.
38717         (Binaries): Ditto.
38718         (Specific): Remove broken link to Renesas RX processor.
38720 2017-02-10  Richard Biener  <rguenther@suse.de>
38722         * toplev.c (process_options): Do not mention obsolete graphite
38723         options when printing sorry message about missing graphite support.
38724         Mention -floop-nest-optimize.
38726 2017-02-10  Christophe Lyon  <christophe.lyon@linaro.org>
38728         * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38729         (vtst_p16): Likewise.
38730         (vtstq_p8): Likewise.
38731         (vtstq_p16): Likewise.
38732         (vtst_p64): New.
38733         (vtstq_p64): Likewise.
38734         * config/arm/arm_neon.h (vgetq_lane_p64): New.
38735         (vset_lane_p64): New.
38736         (vsetq_lane_p64): New.
38738 2017-02-10  Jakub Jelinek  <jakub@redhat.com>
38740         PR tree-optimization/79411
38741         * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38742         stmt operands are SSA_NAMEs used in abnormal phis.
38743         (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38744         phis.
38746 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38748         PR ipa/70795
38749         * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38750         flag if needed.
38752 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38754         * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38756 2017-02-09  Jakub Jelinek  <jakub@redhat.com>
38758         * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38759         to avoid warning.
38761         PR c/79413
38762         * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38763         not arbitrary TREE_CONSTANT.
38765         PR c/79431
38766         * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38767         "omp declare target link" attribute unless is_global_var.
38768         * omp-offload.c (find_link_var_op): Likewise.
38770 2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
38771             Chung-Lin Tang  <cltang@codesourcery.com>
38773         * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38774         OMP_CLAUSE_TILE.
38775         (gimplify_adjust_omp_clauses): Don't delete TILE.
38776         (gimplify_omp_for): Deal with TILE.
38777         * internal-fn.c (expand_GOACC_TILE): New function.
38778         * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38779         (GOACC_TILE): New.
38780         * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38781         (expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
38782         element fields.
38783         (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38784         avoid DIV for outermost collapse var.
38785         (expand_oacc_for): Insert tile element loop as needed.  Adjust.
38786         Remove out of date comments, fix whitespace.
38787         * omp-general.c (omp_extract_for_data): Deal with tiling.
38788         * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38789         adjust OLF_DIM_BASE value.
38790         (struct omp_for_data): Add tiling field.
38791         * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38792         (lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
38793         for auto loops.  Remove default auto determining, moved to
38794         oacc_loop_fixed_partitions.
38795         * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38796         stmts, add e_mask field.
38797         (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38798         (oacc_thread_numbers): Use oacc_dim_call.
38799         (oacc_xform_tile): New.
38800         (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38801         (finish_oacc_loop): Adjust for ifns vector.
38802         (oacc_loop_discover_walk): Append loop abstraction sites to list,
38803         add case for GOACC_TILE fns.
38804         (oacc_loop_xform_loop): Delete.
38805         (oacc_loop_process): Iterate over call list directly, and add
38806         handling for GOACC_TILE fns.
38807         (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38808         dump partitioning.
38809         (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38810         vector partitioning to outer loops.  Assign 2 partitions to loops
38811         when available. Add TILE handling.
38812         (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38813         (execite_oacc_device_lower): Process GOACC_TILE fns,
38814         ignore unknown specs.
38815         * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38816         * tree.c (omp_clause_num_ops): Adjust TILE ops.
38817         * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38819 2017-02-09  Gerald Pfeifer  <gerald@pfeifer.com>
38821         * configure.ac (ACX_BUGURL): Update.
38822         * configure: Regenerate.
38824 2017-02-09  Richard Biener  <rguenther@suse.de>
38826         PR tree-optimization/69823
38827         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38828         Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.
38830 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38832         * config/arc/arc-c.def: Add __NPS400__ definition.
38833         * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38834         (TARGET_NPS400): Define.
38836 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38838         * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38839         file.
38840         (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38841         pointer, arch_info.
38842         (arc_cpu_types): Fill the arch_info field with a pointer into the
38843         arc_arch_types table.
38844         (arc_selected_cpu): Declare.
38845         * config/arc/arc.c (arc_selected_cpu): Make global.
38846         (arc_selected_arch): Delete.
38847         (arc_base_cpu): Delete.
38848         (arc_override_options): Remove references to deleted variables,
38849         update access to arch information.
38850         (ARC_OPT): Update access to arch information.
38851         (ARC_OPTX): Likewise.
38852         * config/arc/arc.h (arc_base_cpu): Remove declaration.
38853         (TARGET_ARC600): Update access to arch information.
38854         (TARGET_ARC601): Likewise.
38855         (TARGET_ARC700): Likewise.
38856         (TARGET_EM): Likewise.
38857         (TARGET_HS): Likewise.
38858         * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38859         information.
38861 2017-02-08  Pat Haugen  <pthaugen@us.ibm.com>
38863         PR target/78604
38864         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38865         condition/operands for integer GE/LE/GEU/LEU operations.
38867 2017-02-08  Segher Boessenkool  <segher@kernel.crashing.org>
38869         PR translation/79397
38870         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38871         of AltiVec.
38873 2017-02-08  Martin Jambor  <mjambor@suse.cz>
38875         PR ipa/79375
38876         * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38877         whether allocation happened.
38878         (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38879         nothing was allocated.
38881 2017-02-08  Jakub Jelinek  <jakub@redhat.com>
38883         PR tree-optimization/79408
38884         * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38885         constant, but SSA_NAME with a known integer range, use the minimum
38886         of that range instead of op1 to determine if modulo can be replaced
38887         with its first operand.
38889 2016-02-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38891         * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38893 2017-02-08  Richard Biener  <rguenther@suse.de>
38895         PR tree-optimization/71824
38896         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38897         Check all loops contained in the merged region.
38899 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38901         * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
38903 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38905         * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
38906         (thunderxt88): Likewise.
38907         (thunderxt81): Disable LSE and change v8.1 to v8.
38908         (thunderxt83): Likewise.
38910 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38911             Richard Biener  <rguenther@suse.de>
38913         PR middle-end/79399
38914         * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
38915         type from int to size_t.
38916         * ira-costs.c (struct_costs_size): Change type from int to size_t.
38918 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38920         PR rtl-optimization/79386
38921         * cprop.c (bypass_conditional_jumps): Initialize
38922         bypass_last_basic_block already before splitting bbs after
38923         unconditional traps...
38924         (bypass_conditional_jumps): ... rather than here.
38926         PR target/79299
38927         * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
38928         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
38929         *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
38930         fix -masm=intel patterns.
38932 2017-02-07  Richard Biener  <rguenther@suse.de>
38934         PR tree-optimization/79256
38935         PR middle-end/79278
38936         * builtins.c (get_object_alignment_2): Use min_align_of_type
38937         to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
38938         and ADJUST_FIELD_ALIGN.
38940         * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
38941         type parameter.
38942         * doc/tm.texi: Regenerate.
38943         * stor-layout.c (layout_decl): Adjust.
38944         (update_alignment_for_field): Likewise.
38945         (place_field): Likewise.
38946         (min_align_of_type): Likewise.
38947         * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
38948         * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
38949         * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
38950         * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
38951         * config/frv/frv.c (frv_adjust_field_align): Likewise.
38952         * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
38953         * config/i386/i386.c (x86_field_alignment): Likewise.
38954         * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
38955         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
38956         * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
38957         * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
38958         * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
38959         * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
38960          Likewise.
38962         Revert
38963         2017-01-30  Richard Biener  <rguenther@suse.de>
38965         PR tree-optimization/79256
38966         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
38967         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
38968         alignment on TYPE.
38970 2017-02-07  Toma Tabacu  <toma.tabacu@imgtec.com>
38972         * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
38973         argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
38974         builtins to SImode and emit a zero-extend, if necessary.
38976 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38978         * docs/invoke.texi (RISC-V Options): Alphabetize.
38980 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38982         * doc/invoke.texi (RISC-V Options): Use two spaces to separate
38983         options.
38985 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38987         * config/riscv/riscv.c: New file.
38988         * common/config/riscv/riscv-common.c: Likewise.
38989         * config.gcc: Likewise.
38990         * config/riscv/constraints.md: Likewise.
38991         * config/riscv/elf.h: Likewise.
38992         * config/riscv/generic.md: Likewise.
38993         * config/riscv/linux.h: Likewise.
38994         * config/riscv/multilib-generator: Likewise.
38995         * config/riscv/peephole.md: Likewise.
38996         * config/riscv/pic.md: Likewise.
38997         * config/riscv/predicates.md: Likewise.
38998         * config/riscv/riscv-builtins.c: Likewise.
38999         * config/riscv/riscv-c.c: Likewise.
39000         * config/riscv/riscv-ftypes.def: Likewise.
39001         * config/riscv/riscv-modes.def: Likewise.
39002         * config/riscv/riscv-opts.h: Likewise.
39003         * config/riscv/riscv-protos.h: Likewise.
39004         * config/riscv/riscv.h: Likewise.
39005         * config/riscv/riscv.md: Likewise.
39006         * config/riscv/riscv.opt: Likewise.
39007         * config/riscv/sync.md: Likewise.
39008         * config/riscv/t-elf-multilib: Likewise.
39009         * config/riscv/t-linux: Likewise.
39010         * config/riscv/t-linux-multilib: Likewise.
39011         * config/riscv/t-riscv: Likewise.
39012         * configure.ac: Likewise.
39013         * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
39014         Waterman as RISC-V maintainers.
39015         * doc/install.texi: Add RISC-V entries.
39016         * doc/invoke.texi: Add RISC-V options section.
39017         * doc/md.texi: Add RISC-V constraints section.
39018         * configure: Regenerated.
39020 2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
39022         PR target/66144
39023         * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
39024         false values to be constant vectors with all 0 or all 1 bits set.
39025         (vcondu<mode><mode>): Likewise.
39026         * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
39027         predicate.
39028         (fpmask_comparison_operator): Update comment.
39029         (vecint_comparison_operator): New predicate.
39030         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
39031         vector conditionals when the true and false values are constant
39032         vectors with all 0 bits or all 1 bits set.
39034 2017-02-06  Martin Sebor  <msebor@redhat.com>
39036         PR  tree-optimization/79376
39037         * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
39039 2017-02-06  Uros Bizjak  <ubizjak@gmail.com>
39041         * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
39042         explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
39043         to simplify split condition.
39045 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39047         * omp-expand.c (oxpand_omp_atomic_fetch_op,
39048         expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
39049         false.
39051 2017-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
39053         PR rtl-optimization/68664
39054         * target.def (can_speculate_insn): New hook.
39055         * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
39056         * doc/tm.texi: Regenerate.
39057         * sched-rgn.c (can_schedule_ready_p): Use the new hook.
39058         * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
39059         (rs6000_sched_can_speculate_insn): New function.
39061 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
39063         PR tree-optimization/79284
39064         * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
39065         * tree-vect-stmts.c (vect_get_vec_def_for_operand,
39066         vectorizable_mask_load_store, vectorizable_operation,
39067         vect_is_simple_cond, get_same_sized_vectype): Use it instead
39068         of comparing TREE_CODE of a type against BOOLEAN_TYPE.
39069         * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
39070         vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
39071         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
39072         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
39073         Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
39074         is_gimple_assign (stmt).  Replace another such test with
39075         is_gimple_assign (stmt).
39077 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39079         PR target/78883
39080         * config/avr/avr.c (rtl-iter.h): Include it.
39081         (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
39082         (avr_legitimate_combined_insn): ...and implementation.
39084 2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39086         * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
39087         * config/s390/s390.c (s390_const_operand_ok)
39088         (s390_canonicalize_comparison, s390_extract_part)
39089         (s390_single_part, s390_contiguous_bitmask_nowrap_p)
39090         (s390_contiguous_bitmask_p, s390_rtx_costs)
39091         (legitimize_pic_address): Likewise.
39092         * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
39093         * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
39094         ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
39095         ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
39096         * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
39098 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39100         * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
39101         REGNO($0) == REGNO($1).
39103 2017-02-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39105         * config/s390/linux.h(SIZE_TYPE): Add comment.
39107 2017-02-06  Julian Brown  <julian@codesourcery.com>
39108             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39109             Virendra Pathak  <virendra.pathak@broadcom.com>
39111         * config/aarch64/aarch64-cores.def: Change the scheduler
39112         to Thunderx2t99.
39113         * config/aarch64/aarch64.md: Include thunderx2t99.md.
39114         * config/aarch64/thunderx2t99.md: New file.
39116 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39118         * doc/standards.texi (Go Language): Update link to language
39119         standard.
39121 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39123         * tree-eh.c (lower_resx): Sanitize profile.
39124         (cleanup_empty_eh_move_lp): Likewise.
39126 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39128         PR tree-ssa/79347
39129         * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
39130         ELSE_PROB.
39131         * cfgloopmanip.h (loop_version): Update prototype.
39132         * modulo-sched.c (sms_schedule): Update call of loop_version.
39133         * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
39134         * tree-parloops.c (gen_parallel_loop): Likewise.
39135         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
39136         * tree-ssa-loop-split.c (split_loop): Likewise.
39137         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
39138         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
39140 2017-02-05  Martin Liska  <mliska@suse.cz>
39142         PR bootstrap/78985
39143         * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
39144         variable to NULL.
39145         (print_operand_address): Initialize a struct to zero.
39147 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39149         * doc/contrib.texi (Contributors): Refer to Hans Boehm's
39150         garbage collector only in textual form.
39152 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39154         * doc/extend.texi (x86 specific memory model extensions for
39155         transactional memory): Simplify a phrase.
39157 2017-02-05  Eric Botcazou  <ebotcazou@adacore.com>
39159         PR target/79353
39160         * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
39161         'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
39162         (atomic_storedi_1): Likewise.
39164 2017-02-04  Jakub Jelinek  <jakub@redhat.com>
39166         PR tree-optimization/79338
39167         * tree-parloops.c (gather_scalar_reductions): Don't call
39168         vect_analyze_loop_form for loop->inner before destroying loop's
39169         loop_vinfo.
39171 2017-02-03  Martin Sebor  <msebor@redhat.com>
39173         PR tree-optimization/79327
39174         * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
39175         when precision has resulted in leading zeros.
39176         (format_integer): Adjust the likely counter to assume an unknown
39177         argument that may be zero is non-zero.
39179 2017-02-03  Jason Merrill  <jason@redhat.com>
39181         PR c++/78689
39182         * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
39183         avoid copying non-taken branch.
39185 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39187         PR tree-optimization/79340
39188         * tree-vect-loop.c (vectorizable_reduction): Release
39189         vec_defs elements after safe_splicing them into other vectors.
39190         Formatting fixes.
39192         PR tree-optimization/79327
39193         * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
39194         true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
39195         dirtype.
39196         (format_integer): Use wide_int_to_tree instead of build_int_cst
39197         + to_?hwi.  If argmin is NULL, just set argmin and argmax to
39198         TYPE_{MIN,MAX}_VALUE of argtype.  Simplify and fix computation
39199         of shortest and longest sequence.
39201 2017-02-03  Uros Bizjak  <ubizjak@gmail.com>
39203         * config/i386/i386.c (dimode_scalar_chain::convert_reg):
39204         Use pextrd for TARGET_SSE4_1 when creating scalar copy.
39206 2017-02-03  Walter Lee  <walt@tilera.com>
39208         PR target/78862
39209         * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
39210         after initial stackframe link reg save.
39211         * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
39213 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39215         PR target/79354
39216         * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
39217         wu for stxssp alternative.
39219 2017-02-03  Martin Sebor  <msebor@redhat.com>
39221         PR tree-optimization/79352
39222         * gimple-fold.c (get_range_strlen): Add argument.
39223         (get_range_strlen): Change return type to bool.
39224         (get_maxval_strlen): Pass in a dummy argument.
39225         * gimple-fold.h (get_range_strlen): Change return type to bool.
39226         * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39227         * tree.h (array_at_struct_end_p): Add argument.
39228         * tree.c (array_at_struct_end_p): Handle it.
39230 2017-02-03  Martin Liska  <mliska@suse.cz>
39232         PR lto/66295
39233         * multiple_target.c (create_dispatcher_calls): Redirect edge
39234         from a caller of a dispatcher.
39235         (expand_target_clones): Make the clones local.
39236         (ipa_target_clone): Do both target clones and resolvers.
39237         (ipa_dispatcher_calls): Remove the pass.
39238         (pass_dispatcher_calls::gate): Likewise.
39239         (make_pass_dispatcher_calls): Likewise.
39240         * passes.def (pass_target_clone): Put as very first IPA early
39241         pass.
39243 2017-02-03  Martin Liska  <mliska@suse.cz>
39245         * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39246         in case of a function with ifunc attribute.
39248 2017-02-03  Martin Liska  <mliska@suse.cz>
39250         * cgraph.c (cgraph_node::dump): Dump function version info.
39251         * symtab.c (symtab_node::dump_base): Add missing new line.
39253 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39255         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39256         (ifcombine_ifandif): Use it.
39258 2017-02-03  Martin Liska  <mliska@suse.cz>
39260         * doc/invoke.texi: Document default value for
39261         use-after-scope-direct-emission-threshold.
39263 2017-02-03  Martin Liska  <mliska@suse.cz>
39265         PR tree-optimization/79339
39266         * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39267         (format_floating): Likewise.
39269 2017-02-03  Martin Liska  <mliska@suse.cz>
39271         PR ipa/79337
39272         * ipa-prop.c (ipa_node_params_t::insert): Remove current
39273         implementation.
39274         (ipa_node_params_t::remove): Likewise.
39275         * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39276         initialization from removed ipa_node_params_t::insert.
39277         (ipa_node_params::~ipa_node_params): Move from removed
39278         ipa_node_params_t::release.
39279         * symbol-summary.h (symbol_summary::m_released): New member.
39280         Do not release a summary twice.  Do not allow to call finalizer
39281         for types of a summary that live in GGC memory.
39283 2017-02-02  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39285         * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39286         cmp_branch fusion.
39288 2017-02-02  Martin Sebor  <msebor@redhat.com>
39290         PR middle-end/79275
39291         * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39292         (format_string): Tighten up the range of output for non-constant
39293         strings and correct the expected range for wide non-constant strings.
39295 2017-02-02  Martin Sebor  <msebor@redhat.com>
39297         * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39299         PR middle-end/32003
39300         * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39301         index.
39302         (-fdump-tree-@var): Add to index and document how to come up
39303         with pass-specific option and dump file names.
39304         (-fdump-passes): Clarify where to look for output.
39306 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39308         PR middle-end/77445
39309         * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39310         statistics of the analyzed path; allow threading for speed when
39311         any of BBs along the path are optimized for speed.
39313 2017-02-02  Eric Botcazou  <ebotcazou@adacore.com>
39315         PR middle-end/78468
39316         * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39317         settings of the virtual registers.
39319         Revert again
39320         2016-08-23  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39322         * explow.c (get_dynamic_stack_size): Take known alignment of stack
39323         pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39324         needed.
39326 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39328         * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39329         ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39331 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39333         * config/s390/s390.md: Add missing comments with the expanded
39334         mnemonics.
39335         * config/s390/vector.md: Likewise.
39336         * config/s390/vx-builtins.md: Likewise.
39338 2017-02-02  Jakub Jelinek  <jakub@redhat.com>
39340         PR target/79197
39341         * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39342         (fixuns_trunc<mode>di2): ... this, remove previous expander.  Put all
39343         conditions on a single line.
39345 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39347         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39348         __S390_VX__ to __VX__.
39350 2017-02-01  Andrew Pinski  <apinski@cavium.com>
39352         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39353         stmt_info to record_stmt_cost.
39354         (vect_get_known_peeling_cost): Pass stmt_info if known to
39355         record_stmt_cost.
39356         * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39357         cpu_vector_cost field into
39358         scalar_int_stmt_cost and scalar_fp_stmt_cost.  Split vec_stmt_cost
39359         field into vec_int_stmt_cost and vec_fp_stmt_cost.
39360         * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39361         splitting of scalar_stmt_cost and vec_stmt_cost.
39362         (thunderx_vector_cost): Likewise.
39363         (cortexa57_vector_cost): LIkewise.
39364         (exynosm1_vector_cost): Likewise.
39365         (xgene1_vector_cost): Likewise.
39366         (thunderx2t99_vector_cost): Improve after the splitting of the two
39367         fields.
39368         (aarch64_builtin_vectorization_cost): Update for the splitting of
39369         scalar_stmt_cost and vec_stmt_cost.
39371 2017-02-01  Torvald Riegel  <triegel@redhat.com>
39372             Richard Henderson  <rth@redhat.com>
39374         * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39375         conditional on existance of a fast atomic load.
39376         * optabs-query.c (can_atomic_load_p): New function.
39377         * optabs-query.h (can_atomic_load_p): Declare it.
39378         * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39379         no fast atomic load is available for the particular size of access.
39380         (expand_atomic_compare_and_swap): Likewise.
39381         (expand_atomic_load): Likewise.
39382         (expand_atomic_store): Likewise.
39383         (expand_atomic_fetch_op): Likewise.
39384         * testsuite/lib/target-supports.exp
39385         (check_effective_target_sync_int_128): Remove x86 because it provides
39386         no fast atomic load.
39387         (check_effective_target_sync_int_128_runtime): Likewise.
39389 2017-02-01  Richard Biener  <rguenther@suse.de>
39391         * graphite.c: Include tree-vectorizer.h for find_loop_location.
39392         (graphite_transform_loops): Provide opt-info for optimized nests.
39393         * tree-parloop.c (parallelize_loops): Provide opt-info for
39394         parallelized loops.
39396 2017-02-01  Richard Biener  <rguenther@suse.de>
39398         PR middle-end/79315
39399         * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39400         was not set before.
39402 2017-02-01  Richard Biener  <rguenther@suse.de>
39404         PR tree-optimization/71824
39405         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39406         Verify the loops are valid in the merged SESE region.
39407         (scop_detection::can_represent_loop_1): Check analyzing the
39408         evolution of the number of iterations in the region succeeds.
39410 2017-01-31  Ian Lance Taylor  <iant@golang.org>
39412         * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39413         REG_ARGS_SIZE note to 32-bit push insns and call insn.
39415 2017-01-31  David Malcolm  <dmalcolm@redhat.com>
39417         PR preprocessor/79210
39418         * input.c (get_substring_ranges_for_loc): Replace line_width
39419         assertion with error-handling.
39421 2017-01-31  Richard Biener  <rguenther@suse.de>
39423         PR tree-optimization/77318
39424         * graphite-sese-to-poly.c (extract_affine): Fix assert.
39425         (create_pw_aff_from_tree): Take loop parameter.
39426         (add_condition_to_pbb): Pass loop of the condition to
39427         create_pw_aff_from_tree.
39429 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39431         * config/s390/s390.c (s390_asan_shadow_offset): New function.
39432         (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39434 2017-01-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
39436         PR target/78597
39437         PR target/79038
39438         * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39439         no longer used.
39440         (convert_int_to_float128): Likewise.
39441         * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39442         (convert_int_to_float128): Likewise.
39443         * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39444         (UNSPEC_IEEE128_CONVERT): Likewise.
39445         (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39446         rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39447         Use local variables for IBM extended format.
39448         (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39449         (fix_trunc<mode>si2_fprs): Likewise.
39450         (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39451         (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39452         (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39453         to know that we can now have integers of all sizes in vector
39454         registers.
39455         (fix<uns>_<mode>di2_hw): Likewise.
39456         (float<uns>_<mode>si2_hw): Likewise.
39457         (fix_<mode>si2_hw): Likewise.
39458         (fixuns_<mode>si2_hw): Likewise.
39459         (float<uns>_<mode>di2_hw): Likewise.
39460         (float_<mode>di2_hw): Likewise.
39461         (float_<mode>si2_hw): Likewise.
39462         (floatuns_<mode>di2_hw): Likewise.
39463         (floatuns_<mode>si2_hw): Likewise.
39464         (xscvqp<su>wz_<mode>): Delete, no longer used.
39465         (xscvqp<su>dz_<mode>): Likewise.
39466         (xscv<su>dqp_<mode>): Likewise.
39467         (ieee128_mfvsrd_64bit): Likewise.
39468         (ieee128_mfvsrd_32bit): Likewise.
39469         (ieee128_mfvsrwz): Likewise.
39470         (ieee128_mtvsrw): Likewise.
39471         (ieee128_mtvsrd_64bit): Likewise.
39472         (ieee128_mtvsrd_32bit): Likewise.
39474 2017-01-31  Martin Liska  <mliska@suse.cz>
39476         PR ipa/79285
39477         * ipa-prop.c (ipa_free_all_node_params): Call release method
39478         instead of ~sumbol_summary to not to trigger double times
39479         dtor of hash_map.
39481 2017-01-31  Aldy Hernandez  <aldyh@redhat.com>
39483         PR tree-optimization/71691
39484         * bitmap.h (class auto_bitmap): New.
39485         * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39486         is_maybe_undefined instead of ssa_undefined_value_p.
39488 2017-01-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39490         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39491         __S390_ARCH_LEVEL__ to __ARCH__.
39493 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39495         PR tree-optimization/79267
39496         * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39497         if should_remove_lhs_p is true.
39499 2017-01-30  Alexandre Oliva  <aoliva@redhat.com>
39501         PR debug/63238
39502         * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39503         (add_alignment_attribute): New.
39504         (base_type_die): Add alignment attribute.
39505         (subrange_type_die): Likewise.
39506         (modified_type_die): Likewise.
39507         (gen_array_type_die): Likewise.
39508         (gen_descr_array_type_die: Likewise.
39509         (gen_enumeration_type_die): Likewise.
39510         (gen_subprogram_die): Likewise.
39511         (gen_variable_die): Likewise.
39512         (gen_field_die): Likewise.
39513         (gen_ptr_to_mbr_type_die): Likewise.
39514         (gen_struct_or_union_type_die): Likewise.
39515         (gen_subroutine_type_die): Likewise.
39516         (gen_typedef_die): Likewise.
39517         (base_type_cmp): Compare alignment attribute.
39519 2017-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
39521         PR target/79170
39522         * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39523         (setb_unsigned) New pattern for setb with CCUNS.
39524         * config/rs6000/rs6000.c (expand_block_compare): Use a different
39525         subfc./subfe sequence to avoid overflow problems.  Generate a
39526         shorter sequence with cmpld/setb for power9.
39527         * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39528         for generating subfc. instruction.
39529         (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39530         now uses this instruction.
39532 2017-01-30  Ian Lance Taylor  <iant@google.com>
39534         PR debug/79289
39535         * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39536         for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39538 2017-01-30  Martin Sebor  <msebor@redhat.com>
39540         * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39541         Move constant to the right of a relational operator.
39542         (get_mpfr_format_length, format_character, format_string): Ditto.
39543         (should_warn_p, maybe_warn): Same.
39545         * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39547 2017-01-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
39549         PR lto/79061
39550         * asan.c (get_translation_unit_decl): Remove function.
39551         (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39553 2017-01-30  Martin Liska  <mliska@suse.cz>
39555         PR gcov-profile/79259
39556         * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39557         -fprofile-generate.
39559 2017-01-30  Martin Liska  <mliska@suse.cz>
39561         PR bootstrap/78985
39562         * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39563         Initialize variables with NULL value.
39565 2017-01-30  Richard Earnshaw  <rearnsh@arm.com>
39567         PR target/79260
39568         * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39569         tm_p_file.
39570         * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39572 2017-01-30  Richard Biener  <rguenther@suse.de>
39574         PR tree-optimization/79276
39575         * tree-vrp.c (process_assert_insertions): Properly adjust common
39576         when removing a duplicate.
39578 2017-01-30  Richard Biener  <rguenther@suse.de>
39580         PR tree-optimization/79256
39581         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39582         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39583         alignment on TYPE.
39584         * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39586 2017-01-30  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39588         PR target/79240
39589         * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39590         ("*r<noxa>sbg_<mode>_sll_bitmask")
39591         ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39592         ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39593         Use contiguous_bitmask_nowrap_operand.
39595 2017-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
39597         PR target/79268
39598         * config/rs6000/altivec.h (vec_xl): Revise #define.
39599         (vec_xst): Likewise.
39601 2017-01-27  Uros Bizjak  <ubizjak@gmail.com>
39603         * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39605 2017-01-27  Bernd Schmidt  <bschmidt@redhat.com>
39607         PR rtl-optimization/79194
39608         * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39609         traps before call to bypass_conditional_jumps.
39611 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39613         PR tree-optimization/71374
39614         * lra-constraints.c (check_conflict_input_operands): New.
39615         (match_reload): Use it.
39617 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39619         PR target/79131
39620         * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39621         account to calculate conflict_set.
39623 2017-01-27  Bin Cheng  <bin.cheng@arm.com>
39625         PR rtl-optimization/78559
39626         * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39627         other_insn in combine.
39629 2017-01-27  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
39631         * builtin-types.def: Use unsigned_char_type_node for BT_UINT8.  Use
39632         uint16_type_node for BT_UINT16.
39634 2017-01-27  David Malcolm  <dmalcolm@redhat.com>
39636         * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39637         "RTL Tests" to menu.
39638         (GIMPLE Tests): New node.
39639         (RTL Tests): New node.
39641 2017-01-27  Richard Biener  <rguenther@suse.de>
39643         PR tree-optimization/79245
39644         * tree-loop-distribution.c (distribute_loop): Apply cost
39645         modeling also to detected patterns.
39647 2017-01-27  Richard Biener  <rguenther@suse.de>
39649         PR tree-optimization/71433
39650         * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39651         (compare_assert_loc): New function.
39652         (process_assert_insertions): Sort and optimize assert locations
39653         to remove duplicates and push down identical assertions on
39654         edges to their destination block.
39656 2017-01-27  Richard Biener  <rguenther@suse.de>
39658         PR tree-optimization/79244
39659         * tree-vrp.c (remove_range_assertions): Forcefully propagate
39660         out SSA names even if abnormal.
39662 2017-01-27  Jakub Jelinek  <jakub@redhat.com>
39664         * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39665         * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39666         instead of MPFR_RNDN.
39668 2017-01-27  Richard Earnshaw  <rearnsha@arm.com>
39670         PR target/79239
39671         * arm.c (arm_option_override): Don't call build_target_option_node
39672         until after doing all option overrides.
39673         (arm_valid_target_attribute_tree): Likewise.
39675 2017-01-27  Martin Liska  <mliska@suse.cz>
39677         * doc/invoke.texi (-fprofile-arcs): Document profiling support
39678         for {cd}tors and C++ {cd}tors.
39680 2017-01-27  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39682         * config/s390/s390.md ("*setmem_long_and")
39683         ("*setmem_long_and_31z"): Use zero_extend instead of and.
39685 2017-01-26  Martin Sebor  <msebor@redhat.com>
39687         * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39688         of precision.
39690 2017-01-26  Martin Sebor  <msebor@redhat.com>
39692         * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39693         HAVE_DFmode before using XFmode or DFmode.
39694         (parse_directive): Avoid using the z length modifier to avoid
39695         the ISO C++98 does not support the â€˜z’ gnu_printf length modifier.
39697         PR middle-end/78703
39698         * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39699         to accept adjustment as an array.
39700         (get_int_range): New function.
39701         (struct directive): Make width and prec arrays.
39702         (directive::set_width, directive::set_precision): Call get_int_range.
39703         (format_integer, format_floating): Handle width and precision ranges.
39704         (format_string, parse_directive): Same.
39706 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39708         PR debug/79129
39709         * dwarf2out.c (generate_skeleton_bottom_up): For children with
39710         comdat_type_p set, just clone them, but keep the children in the
39711         original DIE.
39713         PR debug/78835
39714         * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39715         which have direct callers with -fvar-tracking-assignments enabled
39716         in the current TU.
39717         (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39718         inside of type units.
39720 2017-01-26  Martin Sebor  <msebor@redhat.com>
39722         PR middle-end/78703
39723         * gimple-ssa-sprintf.c (struct result_range): Add likely and
39724         unlikely counters.
39725         (struct format_result): Replace number_chars, number_chars_min,
39726         and number_chars_max with a single member of struct result_range.
39727         Remove bounded.
39728         (format_result::operator+=): Adjust.
39729         (struct fmtresult): Remove bounded.  Handle likely and unlikely
39730         counters.
39731         (fmtresult::adjust_for_width_or_precision): New function.
39732         (fmtresult:type_max_digits): New function.
39733         (bytes_remaining): Handle likely and unlikely counters.
39734         (min_bytes_remaining): Remove.
39735         (format_percent): Simplify.
39736         (format_integer, format_floating): Set likely and unlikely counters.
39737         (get_string_length, format_character, format_string): Same.
39738         (format_plain, should_warn_p): New function.
39739         (maybe_warn): Call should_warn_p.  Update diagnostic messages
39740         and handle those for all directives, including plain strings.
39741         (format_directive): Handle likely and unlikely counters.
39742         Remove unnecessary quoting from diagnostics.  Add an informational
39743         note.
39744         (add_bytes): Remove.
39745         (pass_sprintf_length::compute_format_length): Simplify.
39746         (try_substitute_return_value): Handle likely and unlikely counters.
39748 2017-01-26  Carl Love  <cel@us.ibm.com>
39750         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39751         bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39753 2017-01-26  Vladimir Makarov  <vmakarov@redhat.com>
39755         PR target/79131
39756         * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39757         endianess for subregs into account.
39758         * lra-constraints.c (lra_constraints): Do risky transformations
39759         always on the first iteration.
39760         * lra-lives.c (check_pseudos_live_through_calls): Add arg
39761         last_call_used_reg_set.
39762         (process_bb_lives): Define and use last_call_used_reg_set.
39763         * lra.c (lra): Always continue after lra_constraints on the first
39764         iteration.
39766 2017-01-26  Kirill Yukhin  <kirill.yukhin@gmail.com>
39768         * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39769         constant.
39770         * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39772 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39774         * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39775         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39776         * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39777         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39778         * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39779         __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39780         OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39781         * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39782         (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39784 2017-01-26  Marek Polacek  <polacek@redhat.com>
39786         PR c/79199
39787         * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39788         for the third operand.
39790 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39792         PR middle-end/79236
39793         * omp-low.c (struct omp_context): Add simt_stmt field.
39794         (scan_omp_for): Return omp_context *.
39795         (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39796         context to the _simt_ SIMD stmt.
39797         (lower_omp_for): For combined SIMD with sibling _simt_
39798         SIMD, make sure to use the same decls in _looptemp_
39799         clauses as in the sibling.
39801 2017-01-26  David Sherwood  <david.sherwood@arm.com>
39803         PR middle-end/79212
39804         * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39805         all contexts.
39807 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39809         PR target/70465
39810         * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39811         emit fld b; fld a; if possible.
39813         * brig-builtins.def: Update copyright years.
39814         * config/arm/arm_acle_builtins.def: Update copyright years.
39816 2017-01-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
39818         PR target/79179
39819         * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39820         constraint instead of o for the stxsd instruction.
39822 2017-01-25  Carl Love  <cel@us.ibm.com>
39824         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39825         of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39827 2017-01-25  Jonathan Wakely  <jwakely@redhat.com>
39829         * doc/invoke.texi (C++ Dialect Options): Fix typo.
39831 2017-01-25  Richard Biener  <rguenther@suse.de>
39833         PR tree-optimization/69264
39834         * target.def (vector_alignment_reachable): Improve documentation.
39835         * doc/tm.texi: Regenerate.
39836         * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39837         and add a comment.
39838         * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39839         earlier changes with respect to TYPE_USER_ALIGN.
39840         (vector_alignment_reachable_p): Likewise.  Improve dumping.
39842 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39844         PR target/79145
39845         * config/arm/arm.md (xordi3): Force constant operand into a register
39846         for TARGET_IWMMXT.
39848 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39850         * doc/invoke.texi (-fstore-merging): Correct default optimization
39851         levels at which it is enabled.
39852         (-O): Move -fstore-merging from list to...
39853         (-O2): ... Here.
39855 2017-01-25  Richard Biener  <rguenther@suse.de>
39857         PR debug/78363
39858         * omp-expand.c: Include debug.h.
39859         (expand_omp_taskreg): Make sure to generate early debug before
39860         outlining anything from a function.
39861         (expand_omp_target): Likewise.
39862         (grid_expand_target_grid_body): Likewise.
39864 2017-01-25  Maxim Ostapenko  <m.ostapenko@samsung.com>
39866         PR lto/79061
39867         * asan.c (get_translation_unit_decl): New function.
39868         (asan_add_global): Extract modules file name from globals
39869         TRANSLATION_UNIT_DECL name.
39871 2017-01-24  Eric Botcazou  <ebotcazou@adacore.com>
39873         PR target/77439
39874         * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39875         for long calls with APCS frame and VFP.
39877 2017-01-24  David Malcolm  <dmalcolm@redhat.com>
39879         * cfg.c (original_copy_tables_initialized_p): New function.
39880         * cfg.h (original_copy_tables_initialized_p): New decl.
39881         * cfgrtl.c (relink_block_chain): Guard the call to
39882         free_original_copy_tables with a call to
39883         original_copy_tables_initialized_p.
39884         * cgraph.h (symtab_node::native_rtl_p): New decl.
39885         * cgraphunit.c (symtab_node::native_rtl_p): New function.
39886         (symtab_node::needed_p): Don't assert for early assembly output
39887         for __RTL functions.
39888         (cgraph_node::finalize_function): Set "force_output" for __RTL
39889         functions.
39890         (cgraph_node::analyze): Bail out early for __RTL functions.
39891         (analyze_functions): Update assertion to support __RTL functions.
39892         (cgraph_node::expand): Bail out early for __RTL functions.
39893         * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39894         __RTL functions.
39895         * function.h (struct function): Update comment for field
39896         "pass_startwith".
39897         * gimple-expr.c: Include "tree-pass.h".
39898         (gimple_has_body_p): Return false for __RTL functions.
39899         * Makefile.in (OBJS): Add run-rtl-passes.o.
39900         * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
39901         accessor.
39902         (gcc::pass_manager::get_clean_slate): New accessor.
39903         * passes.c: Include "insn-addr.h".
39904         (should_skip_pass_p): Add logging.  Update logic for running
39905         "expand" to be compatible with both __GIMPLE and __RTL.  Guard
39906         property-provider override so it is only done for gimple passes.
39907         Don't skip dfinit.
39908         (skip_pass): New function.
39909         (execute_one_pass): Call skip_pass when skipping passes.
39910         * read-md.c (md_reader::read_char): Support filtering
39911         the input to a subset of line numbers.
39912         (md_reader::md_reader): Initialize fields
39913         m_first_line and m_last_line.
39914         (md_reader::read_file_fragment): New function.
39915         * read-md.h (md_reader::read_file_fragment): New decl.
39916         (md_reader::m_first_line): New field.
39917         (md_reader::m_last_line): New field.
39918         * read-rtl-function.c (function_reader::create_function): Only
39919         create cfun if it doesn't already exist.  Set PROP_rtl on cfun's
39920         curr_properties.  Set DECL_INITIAL to a dummy block.
39921         (read_rtl_function_body_from_file_range): New function.
39922         * read-rtl-function.h (read_rtl_function_body_from_file_range):
39923         New decl.
39924         * run-rtl-passes.c: New file.
39925         * run-rtl-passes.h: New file.
39927 2017-01-24  Jeff Law  <law@redhat.com>
39929         * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
39930         buffer size.
39932 2017-01-24  Bin Cheng  <bin.cheng@arm.com>
39934         PR tree-optimization/79159
39935         * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
39936         (record_nonwrapping_iv): Improve boundary using above function if no
39937         value range information.
39939 2017-01-24  Pekka Jääskeläinen  <pekka@parmance.com>
39940             Martin Jambor  <mjambor@suse.cz>
39942         * brig-builtins.def: New file.
39943         * builtins.def (DEF_HSAIL_BUILTIN): New macro.
39944         (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
39945         (DEF_HSAIL_SAT_BUILTIN): Likewise.
39946         (DEF_HSAIL_INTR_BUILTIN): Likewise.
39947         (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
39948         * builtin-types.def (BT_INT8): New.
39949         (BT_INT16): Likewise.
39950         (BT_UINT8): Likewise.
39951         (BT_UINT16): Likewise.
39952         (BT_FN_ULONG): Likewise.
39953         (BT_FN_UINT_INT): Likewise.
39954         (BT_FN_UINT_ULONG): Likewise.
39955         (BT_FN_UINT_LONG): Likewise.
39956         (BT_FN_UINT_PTR): Likewise.
39957         (BT_FN_ULONG_PTR): Likewise.
39958         (BT_FN_INT8_FLOAT): Likewise.
39959         (BT_FN_INT16_FLOAT): Likewise.
39960         (BT_FN_UINT32_FLOAT): Likewise.
39961         (BT_FN_UINT16_FLOAT): Likewise.
39962         (BT_FN_UINT8_FLOAT): Likewise.
39963         (BT_FN_UINT64_FLOAT): Likewise.
39964         (BT_FN_UINT16_UINT32): Likewise.
39965         (BT_FN_UINT32_UINT16): Likewise.
39966         (BT_FN_UINT16_UINT16_UINT16): Likewise.
39967         (BT_FN_INT_PTR_INT): Likewise.
39968         (BT_FN_UINT_PTR_UINT): Likewise.
39969         (BT_FN_LONG_PTR_LONG): Likewise.
39970         (BT_FN_ULONG_PTR_ULONG): Likewise.
39971         (BT_FN_VOID_UINT64_UINT64): Likewise.
39972         (BT_FN_UINT8_UINT8_UINT8): Likewise.
39973         (BT_FN_INT8_INT8_INT8): Likewise.
39974         (BT_FN_INT16_INT16_INT16): Likewise.
39975         (BT_FN_INT_INT_INT): Likewise.
39976         (BT_FN_UINT_FLOAT_UINT): Likewise.
39977         (BT_FN_FLOAT_UINT_UINT): Likewise.
39978         (BT_FN_ULONG_UINT_UINT): Likewise.
39979         (BT_FN_ULONG_UINT_PTR): Likewise.
39980         (BT_FN_ULONG_ULONG_ULONG): Likewise.
39981         (BT_FN_UINT_UINT_UINT): Likewise.
39982         (BT_FN_VOID_UINT_PTR): Likewise.
39983         (BT_FN_UINT_UINT_PTR: Likewise.
39984         (BT_FN_UINT32_UINT64_PTR): Likewise.
39985         (BT_FN_INT_INT_UINT_UINT): Likewise.
39986         (BT_FN_UINT_UINT_UINT_UINT): Likewise.
39987         (BT_FN_UINT_UINT_UINT_PTR): Likewise.
39988         (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
39989         (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
39990         (BT_FN_LONG_LONG_UINT_UINT): Likewise.
39991         (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
39992         (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
39993         (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
39994         (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
39995         (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
39996         (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
39997         * doc/frontends.texi: List BRIG FE.
39998         * doc/install.texi (Testing): Add BRIG tesring requirements.
39999         * doc/invoke.texi (Overall Options): Mention BRIG.
40000         * doc/standards.texi (Standards): Doucment BRIG HSA version.
40002 2017-01-24  Richard Biener  <rguenther@suse.de>
40004         PR translation/79208
40005         * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
40007 2017-01-24  Martin Jambor  <mjambor@suse.cz>
40009         PR bootstrap/79198
40010         * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
40011         * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
40012         and known_contexts.
40014 2017-01-24  Aldy Hernandez  <aldyh@redhat.com>
40016         PR middle-end/79123
40017         * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
40018         casts from signed to unsigned really don't have a range.
40020 2017-01-24  Markus Trippelsdorf  <markus@trippelsdorf.de>
40022         * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
40023         GMP_RNDx for compatiblity.
40025 2017-01-24  Martin Liska  <mliska@suse.cz>
40027         PR bootstrap/79132
40028         * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
40029         that would prevent us to call alloca with -1 as argument.
40031 2017-01-24  Jakub Jelinek  <jakub@redhat.com>
40033         * dwarf2out.c (output_compilation_unit_header, output_file_names):
40034         Avoid -Wformat-security warning.
40036 2017-01-23  Andrew Pinski  <apinski@cavium.com>
40038         * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
40039         cost table.
40041 2017-01-23  Martin Sebor  <msebor@redhat.com>
40043         PR middle-end/78703
40044         * gimple-ssa-sprintf.c (warn_level): New global.
40045         (format_integer): Use it here and throughout the rest of the file.
40046         Use the same switch to compute sign as base.
40047         (maybe_warn): New function.
40048         (format_directive): Factor out warnings into maybe_warn.
40049         Add debugging output.  Use warn_level.
40050         (add_bytes): Use warn_level.
40051         (pass_sprintf_length::compute_format_length): Add debugging output.
40052         (try_substitute_return_value): Same.
40053         (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
40055         PR middle-end/78703
40056         * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
40057         (struct fmtresult, format_integer, format_floating): Adjust.
40058         (fmtresult::fmtresult): Set max correctly in two argument ctor.
40059         (get_string_length, format_string,format_directive): Same.
40060         (pass_sprintf_length::compute_format_length): Same.
40061         (try_substitute_return_value): Simplify slightly.
40063         PR middle-end/78703
40064         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
40065         (fmtresult::operator+=): Outlined.
40066         (struct fmtresult): Add ctors.
40067         (struct conversion_spec): Rename...
40068         (struct directive): ...to this.  Add and remove data members.
40069         (directive::set_width, directive::set_precision): New functions.
40070         (format_percent): Use fmtresult ctor.
40071         (get_width_and_precision): Remove.
40072         (format_integer): Make naming changes.  Avoid computing width and
40073         precision.
40074         (format_floating): Same.  Adjust indentation.
40075         (format_character, format_none): New functions.
40076         (format_string): Moved character handling to format_character.
40077         (format_directive): Remove arguments, change return type.
40078         (parse_directive): New function.
40079         (pass_sprintf_length::compute_format_length): Move directive
40080         parsing to parse_directive.
40082 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40084         * tree.h (assign_assembler_name_if_neeeded): Rename to ...
40085         (assign_assembler_name_if_needed): ... this.
40086         * tree.c (assign_assembler_name_if_neeeded): Rename to ...
40087         (assign_assembler_name_if_needed): ... this.
40088         (free_lang_data_in_cgraph): Adjust callers.
40089         * cgraphunit.c (cgraph_node::analyze): Likewise.
40090         * omp-expand.c (expand_omp_taskreg, expand_omp_target):
40091         Likewise.
40093 2017-01-23  Richard Biener  <rguenther@suse.de>
40095         PR tree-optimization/79088
40096         PR tree-optimization/79188
40097         * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
40098         resetting loop bounds after last path deletion.  Reset loop
40099         bounds of the target loop, make code match the comments.
40100         * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
40101         Make sure loops need no fixups.
40103 2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
40105         * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
40106         exponent support with double type for first argument.
40107         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
40108         type returned by __builtin_vec_extract_sig,
40109         __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
40110         functions from "vector int" to "vector unsigned int" or from
40111         "vector long long int" to "vector unsigned long long int".
40112         Changed type returned by __builtin_vec_extract_exp,
40113         __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
40114         functions from "vector int" to "vector unsigned int" or from
40115         "vector long long int" to "vector unsigned long long int".
40116         Changed return type of __builtin_vec_test_data_class,
40117         __builtin_vec_test_data_class_sp, and
40118         __builtin_vec_test_data_class_dp from "vector int" to
40119         "vector bool int" or from "vector long long int" to "vector bool
40120         long long int" and changed second argument type from "unsigned
40121         int" to "int".  Added new overloaded function forms "vector float
40122         __builtin_vec_insert_exp (vector float, vector unsigned int)" and
40123         "vector float __builtin_vec_insert_exp_sp (vector float, vector
40124         unsigned int)" and "vector double __builtin_vec_insert_exp (vector
40125         double, vector unsigned long long int)" and "vector double
40126         __builtin_vec_insert_exp_dp (vector double, vector unsigned long
40127         long int)".  Changed return type of
40128         __builtin_scalar_test_data_class and
40129         __builtin_scalar_test_data_class_sp and
40130         __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
40131         int" and changed second argument from "unsigned int" to "int".
40132         Changed type returned by __builtin_scalar_test_neg,
40133         __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
40134         from "int" to "bool int".  Added new overloaded function form
40135         "double __builtin_scalar_insert_exp (double, unsigned long long int)".
40136         * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
40137         exponent double-precision with floating point first argument.
40138         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
40139         documentation of scalar_test_data_class, scalar_test_neg,
40140         scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
40141         vector_extract_exp, vec_extract_sig, vec_insert_exp, and
40142         vec_test_data_class built-in functions to reflect refinements in
40143         their type signatures.
40145 2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>
40147         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
40148         size of buf.
40149         (aarch64_elf_asm_destructor): Likewise.
40151 2017-01-23  Bernd Schmidt  <bschmidt@redhat.com>
40153         PR rtl-optimization/78634
40154         * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
40155         (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
40156         * ifcvt.c (noce_try_cmove): Add missing cost check.
40158         PR rtl-optimization/71724
40159         * combine.c (if_then_else_cond): Look for situations where it is
40160         beneficial to undo the work of one of the recursive calls.
40162 2017-01-23  Bin Cheng  <bin.cheng@arm.com>
40164         PR tree-optimization/70754
40165         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
40166         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
40167         combined stmt before it if not NULL.
40168         (combine_chains): Process refs reversely and compute dominance point
40169         for root ref.
40171 2017-01-23  Martin Liska  <mliska@suse.cz>
40173         PR tree-optimization/79196
40174         * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
40175         (fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
40176         instead of memcmp.
40177         (strlen_optimize_stmt): Call the renamed function.
40179 2017-01-23  Michael Matz  <matz@suse.de>
40181         PR tree-optimization/78384
40182         * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
40184 2017-01-23  Richard Biener  <rguenther@suse.de>
40186         PR tree-optimization/79186
40187         * tree-vrp.c (register_new_assert_for): Make sure we've seen
40188         both incoming edges before moving an assert.
40190 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40192         * ipa-prop.c (load_from_param_1): Removed.
40193         (load_from_unmodified_param): Bits from load_from_param_1 put back
40194         here.
40195         (load_from_param): Removed.
40196         (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
40197         with stmt.  Reverted back to use of load_from_unmodified_param.
40199 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40201         PR ipa/79108
40202         * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
40203         (ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
40204         field a pointer to garbage collected vector, mark lattices and
40205         ipcp_orig_node with GTY((skip)).
40206         (ipa_get_param_count): Adjust to descriptors being a pointer.
40207         (ipa_get_param): Likewise.
40208         (ipa_get_type): Likewise.
40209         (ipa_get_param_move_cost): Likewise.
40210         (ipa_set_param_used): Likewise.
40211         (ipa_get_controlled_uses): Likewise.
40212         (ipa_set_controlled_uses): Likewise.
40213         (ipa_is_param_used): Likewise.
40214         (ipa_node_params_t): Move into garbage collector.  New methods insert
40215         and remove.
40216         (ipa_node_params_sum): Annotate wth GTY(()).
40217         (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
40218         garbage collected.
40219         (ipa_load_from_parm_agg): Adjust declaration.
40220         * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
40221         * ipa-profile.c (ipa_profile): Likewise.
40222         * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
40223         (ipa_populate_param_decls): Make descriptors parameter garbage
40224         collected.
40225         (ipa_dump_param): Adjust to descriptors being a pointer.
40226         (ipa_alloc_node_params): Likewise.
40227         (ipa_initialize_node_params): Likewise.
40228         (load_from_param_1): Make descriptors parameter garbage collected.
40229         (load_from_unmodified_param): Likewise.
40230         (load_from_param): Likewise.
40231         (ipa_load_from_parm_agg): Likewise.
40232         (ipa_node_params::~ipa_node_params): Removed.
40233         (ipa_free_all_node_params): Remove call to delete operator.
40234         (ipa_node_params_t::insert): New.
40235         (ipa_node_params_t::remove): Likewise.
40236         (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40237         copy known_csts and known_contexts vectors.
40238         (ipa_read_node_info): Adjust to descriptors being a pointer.
40239         (ipcp_modif_dom_walker): Make m_descriptors field garbage
40240         collected.
40241         (ipcp_transform_function): Make descriptors variable garbage
40242         collected.
40244 2017-01-23  Andrew Senkevich  <andrew.senkevich@intel.com>
40246         * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40247         * config/i386/avx512dqintrin.h: Ditto.
40248         * config/i386/avx512fintrin.h: Ditto.
40249         * config/i386/i386.c: Handle new builtins.
40250         * config/i386/i386-builtin.def: Add new builtins.
40251         * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40252         (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40254 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40255             Martin Liska  <mliska@suse.cz>
40257         * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40258         * asan.c (asan_expand_poison_ifn): Support stores and use
40259         appropriate ASAN report function.
40260         * internal-fn.c (expand_ASAN_POISON_USE): New function.
40261         * internal-fn.def (ASAN_POISON_USE): Declare.
40262         * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40263         (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40264         * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40265         ASAN_POISON calls w/o LHS.
40266         * tree-ssa.c (execute_update_addresses_taken): Create clobber
40267         for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40268         from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40269         * gimplify.c (asan_poison_variables): Add attribute
40270         use_after_scope_memory to variables that really needs to live
40271         in memory.
40272         * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40273         having the attribute.
40275 2017-01-23  Martin Liska  <mliska@suse.cz>
40277         * asan.c (create_asan_shadow_var): New function.
40278         (asan_expand_poison_ifn): Likewise.
40279         * asan.h (asan_expand_poison_ifn): New declaration.
40280         * internal-fn.c (expand_ASAN_POISON): Likewise.
40281         * internal-fn.def (ASAN_POISON): New builtin.
40282         * sanopt.c (pass_sanopt::execute): Expand
40283         asan_expand_poison_ifn.
40284         * tree-inline.c (copy_decl_for_dup_finish): Make function
40285         external.
40286         * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40287         * tree-ssa.c (is_asan_mark_p): New function.
40288         (execute_update_addresses_taken): Rewrite local variables
40289         (identified just by use-after-scope as addressable) into SSA.
40291 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40293         * doc/install.texi (Specific): opensource.apple.com uses https
40294         now. Remove trailing slash.
40296 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40298         * README.Portability: Remove note on an Irix compatibility issue.
40300 2017-01-22  Dimitry Andric <dim@FreeBSD.org>
40302         * gcov.c (INCLUDE_ALGORITHM): Define.
40303         (INCLUDE_VECTOR): Define.
40304         No longer include <vector> and <algorithm> directly.
40306 2017-01-21  Gerald Pfeifer  <gerald@pfeifer.com>
40308         * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40309         to https.
40310         * doc/invoke.texi (Code Gen Options): Ditto.
40312 2017-01-21  Jan Hubicka  <hubicka@ucw.cz>
40314         PR lto/78407
40315         * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40317 2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>
40319         rtl-optimization/79125
40320         * cprop.c (local_cprop_pass): Handle cases where we make an
40321         unconditional trap.
40323 2017-01-20  Segher Boessenkool  <segher@kernel.crashing.org>
40325         PR target/61729
40326         PR target/77850
40327         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40328         read from, for big endian.
40330 2017-01-20  Jiong Wang  <jiong.wang@arm.com>
40332         * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40333         register pauth builtins for LP64 only.
40335 2017-01-20  Marek Polacek  <polacek@redhat.com>
40337         PR c/79152
40338         * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40339         non-case labels.
40341 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40343         * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40344         of safelen status.
40345         * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40346         * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40347         * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40349 2017-01-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40351         PR target/71270
40352         * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40353         in big-endian mode when they are not a single duplicated value.
40355 2017-01-20  Richard Biener  <rguenther@suse.de>
40357         * BASE-VER: Bump to 7.0.1.
40359 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40361         * omp-low.c (omplow_simd_context): New struct.  Use it...
40362         (lower_rec_simd_input_clauses): ...here and...
40363         (lower_rec_input_clauses): ...here to hold common data.  Adjust all
40364         references to idx, lane, max_vf, is_simt.
40366 2017-01-20  Graham Markall  <graham.markall@embecosm.com>
40368         * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40369         mcpu=nps400.
40371 2017-01-20  Martin Jambor  <mjambor@suse.cz>
40373         * hsa.h: Renaed to hsa-common.h.  Adjusted a comment.
40374         * hsa.c: Renaed to hsa-common.c.  Change include of gt-hsa.h to
40375         gt-hsa-common.h.
40376         * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40377         (GTFILES): Rename hsa.c to hsa-common.c.
40378         * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40379         * hsa-dump.c: Likewise.
40380         * hsa-gen.c: Likewise.
40381         * hsa-regalloc.c: Likewise.
40382         * ipa-hsa.c: Likewise.
40383         * omp-expand.c: Likewise.
40384         * omp-low.c: Likewise.
40385         * toplev.c: Likewise.
40387 2017-01-20  Marek Polacek  <polacek@redhat.com>
40389         PR c/64279
40390         * doc/invoke.texi: Document -Wduplicated-branches.
40391         * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40392         COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40393         POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40394         STATEMENT_LIST, and RETURN_EXPR.  For non-pure non-const functions
40395         return 0 only when not OEP_LEXICOGRAPHIC.
40396         (fold_build_cleanup_point_expr): Use the expression
40397         location when building CLEANUP_POINT_EXPR.
40398         * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40399         * tree.c (add_expr): Handle error_mark_node.
40401 2017-01-20  Martin Liska  <mliska@suse.cz>
40403         PR lto/69188
40404         * tree-profile.c (init_ic_make_global_vars): Do not call
40405         finalize_decl.
40406         (gimple_init_gcov_profiler): Likewise.
40408 2017-01-20  Martin Liska  <mliska@suse.cz>
40410         PR ipa/71190
40411         * cgraph.h (maybe_create_reference): Remove argument and
40412         update comment.
40413         * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40414         argument.
40415         * ipa-cp.c (create_specialized_node): Likewise.
40416         * symtab.c (symtab_node::maybe_create_reference): Handle
40417         VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40419 2017-01-20  Martin Liska  <mliska@suse.cz>
40421         * read-rtl-function.c (function_reader::create_function): Use
40422         build_decl instread of build_decl_stat.
40424 2017-01-20  Andrew Senkevich  <andrew.senkevich@intel.com>
40426         * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40427         * config/i386/avx512dqintrin.h: Ditto.
40428         * config/i386/avx512fintrin.h: Ditto.
40429         * config/i386/i386-builtin-types.def: Add new types.
40430         * config/i386/i386.c: Handle new types.
40431         * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40432         (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40433         (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40434         (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40435         (__builtin_ia32_kshiftridi): New.
40436         * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40438 2017-01-19  Segher Boessenkool  <segher@kernel.crashing.org>
40440         PR target/78875
40441         PR target/79140
40442         * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40443         define to rs6000_init_stack_protect_guard.
40444         (rs6000_init_stack_protect_guard): New function.
40446 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40447             Yunqiang Su  <yunqiang.su@imgtec.com>
40449         * config.gcc (supported_defaults): Add madd4.
40450         (with_madd4): Add validation.
40451         (all_defaults): Add madd4.
40452         * config/mips/mips.opt (mmadd4): New option.
40453         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40454         mmadd4.
40455         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40456         __mips_no_madd4.
40457         (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40458         (ISA_HAS_FUSED_MADD4): Likewise.
40459         * doc/invoke.texi (-mmadd4): Document the new option.
40460         * doc/install.texi (--with-madd4): Document the new option.
40462 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40464         * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40465         entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40466         AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40467         (aarch64_init_pauth_hint_builtins): New.
40468         (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40469         (aarch64_expand_builtin): Expand new builtins.
40471 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40473         * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40474         * combine-stack-adj.c (no_unhandled_cfa): Handle
40475         REG_CFA_TOGGLE_RA_MANGLE.
40476         * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40477         * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40478         info for return address signing.
40479         (aarch64_expand_epilogue): Likewise.
40481 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40483         * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40484         * config/aarch64/aarch64-protos.h
40485         (aarch64_return_address_signing_enabled): New declaration.
40486         * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40487         New function.
40488         (aarch64_expand_prologue): Sign return address before it's pushed onto
40489         stack.
40490         (aarch64_expand_epilogue): Authenticate return address fetched from
40491         stack.
40492         (aarch64_override_options): Sanity check for ILP32 and ISA level.
40493         (aarch64_attributes): New function attributes for "sign-return-address".
40494         * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40495         UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40496         ("*do_return"): Generate combined instructions according to key index.
40497         ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40498         * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40499         iterators.
40500         (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40501         * config/aarch64/aarch64.opt (msign-return-address=): New.
40502         * doc/extend.texi (AArch64 Function Attributes): Documents
40503         "sign-return-address=".
40504         * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40506 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40508         * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40509         overall option summary.
40511 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40513         * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40514         * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40515         AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40516         * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40518 2017-01-19  Michael Meissner  <meissner@linux.vnet.ibm.com>
40520         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40521         -mpower9-minmax by default for -mcpu=power9.
40522         (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40523         128-bit floating point.
40525 2017-01-20  Alan Modra  <amodra@gmail.com>
40527         * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40528         optimizing for size.
40530 2017-01-20  Alan Modra  <amodra@gmail.com>
40532         PR target/79144
40533         * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40534         for strcmp and strncmp from corresponding builtin decl.
40536 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40538         * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40539         instead of i386/rtems-64.h.
40540         * config/i386/rtems-64.h: Remove.
40542 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40544         PR target/78478
40545         Revert:
40546         2013-11-05  Uros Bizjak  <ubizjak@gmail.com>
40548         * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40550 2017-01-19  Tamar Christina  <tamar.christina@arm.com>
40552         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40553         Change int to HOST_WIDE_INT.
40554         * config/aarch64/aarch64-protos.h
40555         (aarch64_simd_gen_const_vector_dup): Likewise.
40556         * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40558 2017-01-19  David Malcolm  <dmalcolm@redhat.com>
40560         * langhooks-def.h (lhd_type_for_size): New decl.
40561         (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40562         * langhooks.c (lhd_type_for_size): New function, taken from
40563         lto_type_for_size.
40565 2017-01-19  Pat Haugen  <pthaugen@us.ibm.com>
40567         * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40568         define_bypass for CR latency.
40569         (power9-cracked-alu): Update bypass latency and remove power9-branch.
40570         (power9-alu2): Add define_bypass for CR latency.
40571         (power9-cmp): New.
40572         (power9-mul): Update insn latency.
40573         (power9-mul-compare): Update insn latency, bypass latency and remove
40574         power9-branch.
40576 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40578         * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40579         Delete.
40580         * config/aarch64/aarch64.md
40581         (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40582         aarch64_nopcrelative_literal_loads.
40583         (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40585 2017-01-19  Chenghua Xu  <paul.hua.gm@gmail.com>
40587         * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40588         TARGET_LOONGSON_3A.
40589         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40591 2017-01-19  Doug Gilmore  <doug.gilmore@imgtec.com>
40593         PR target/78176
40594         * config.gcc (supported_defaults): Add lxc1-sxc1.
40595         (with_lxc1_sxc1): Add validation.
40596         (all_defaults): Add lxc1-sxc1.
40597         * config/mips/mips.opt (mlxc1-sxc1): New option.
40598         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40599         mlxc1-sxc1.
40600         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40601         __mips_no_lxc1_sxc1.
40602         (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40603         * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40604         * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40606 2017-01-19  Richard Biener  <rguenther@suse.de>
40608         PR tree-optimization/72488
40609         * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40610         sure to restore SSA info.
40611         * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40613 2017-01-19  Richard Earnshaw  <rearnsha@arm.com>
40615         PR rtl-optimization/79121
40616         * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40617         of the inner type when shifting an extended value.
40619 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40621         PR lto/78407
40622         * symtab.c (symtab_node::equal_address_to): Fix comparing of
40623         interposable aliases.
40625 2017-01-18  Peter Bergner  <bergner@vnet.ibm.com>
40627         PR target/78516
40628         * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40629         Use the evmergelohi instruction.
40630         (mov_si<mode>_e500_subreg4_2_le): Likewise.
40631         (mov_sitf_e500_subreg8_2_be): Likewise.
40632         (mov_sitf_e500_subreg12_2_le): Likewise.
40633         (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40634         (mov_si<mode>_e500_subreg4_2_be): Likewise.
40635         (mov_sitf_e500_subreg8_2_le): Likewise.
40636         (mov_sitf_e500_subreg12_2_be): Likewise.
40638 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40640         * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40641         attribute from vecsimple to vecperm.
40642         (altivec_vbpermq2): Likewise.
40644 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40646         PR target/79040
40647         * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40649 2017-01-18  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
40650         * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40651         * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40652         strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40653         case where N arg is SIZE_MAX.
40654         * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40655         (cmpstrsi): Add pattern.
40657 2017-01-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
40659         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40660         __builtin_vec_revb builtins.
40661         * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40662         built-in functions to support generation of the ISA 3.0 XXBR<x>
40663         vector byte reverse instructions.
40664         (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40665         (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40666         (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40667         (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40668         (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40669         (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40670         (P9V_BUILTIN_VEC_REVB): Likewise.
40671         * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40672         generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40673         (p9_xxbrq_v16qi): Likewise.
40674         (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40675         (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40676         (p9_xxbrh_v8hi): Likewise.
40677         * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40678         * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40679         vec_revb built-in functions.
40681 2017-01-18  Uros Bizjak  <ubizjak@gmail.com>
40683         PR rtl-optimization/78952
40684         * config/i386/i386.md (any_extract): New code iterator.
40685         (*insvqi_2): Use any_extract for source operand.
40686         (*insvqi_3): Use any_shiftrt for source operand.
40688 2017-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
40690         * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40691         New function.
40692         (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40694 2017-01-18  Matthias Klose  <doko@ubuntu.com>
40696         * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40698 2016-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40700         * config/rs6000/altivec.h (vec_bperm): Change #define.
40701         * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40702         (altivec_vbpermq2): New define_insn.
40703         (altivec_vbpermd): Likewise.
40704         * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40705         function interface.
40706         (VBPERMD): Likewise.
40707         (VBPERM): New polymorphic function interface.
40708         * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40709         Add entries for P9V_BUILTIN_VEC_VBPERM.
40710         * doc/extend.texi: Add interfaces for vec_bperm.
40712 2017-01-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
40714         * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40715         first letter of error messages.
40716         (s390_resolve_overloaded_builtin): Likewise.
40717         * config/s390/s390.c (s390_expand_builtin): Likewise.
40718         (s390_invalid_arg_for_unprototyped_fn): Likewise.
40719         (s390_valid_target_attribute_inner_p): Likewise.
40720         * config/s390/s390.md ("tabort"): Likewise.
40722 2017-01-18  Toma Tabacu  <toma.tabacu@imgtec.com>
40724         * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40725         (ISA_AVOID_DIV_HILO): New macro.
40726         (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40727         (ISA_HAS_DDIV): Likewise.
40729 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40731         * doc/invoke.texi (fabi-version): Correct number of occurrences.
40733 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40735         * doc/invoke.texi (fabi-version): Spelling fix.
40737 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40739         PR c++/70182
40740         * doc/invoke.texi (fabi-version): Mention mangling fix for
40741         operator names.
40743 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40745         PR c++/77489
40746         * doc/invoke.texi (fabi-version): Document discriminator mangling.
40748 2017-01-17  Segher Boessenkool  <segher@kernel.crashing.org>
40750         PR target/78875
40751         * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40752         * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40753         the new options.
40754         * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40755         flexible settings.
40756         (stack_protect_test): Ditto.
40757         * config/rs6000/rs6000.opt (mstack-protector-guard=,
40758         mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40759         options.
40760         * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40761         Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40762         -mstack-protector-guard-offset=.
40763         (RS/6000 and PowerPC Options): Ditto.
40765 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40767         * config/i386/i386.h (MASK_CLASS_P): New define.
40768         * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40769         there are no registers from different register sets also when
40770         mask registers are used.  Update function comment.
40771         * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40772         to (*k/*r) and (*k/*km) alternatives.
40774 2017-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
40776         * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40777         * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40778         (EH_RETURN_HANDLER_RTX): New define.
40779         * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40780         Force frame pointer in EH return functions.
40781         (aarch64_expand_epilogue): Add barrier for eh_return.
40782         (aarch64_final_eh_return_addr): Remove.
40783         (aarch64_eh_return_handler_rtx): New function.
40784         * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40785         Remove.
40786         (aarch64_eh_return_handler_rtx): New prototype.
40788 2017-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40790         * config/rs6000/altivec.h (vec_rlmi): New #define.
40791         (vec_vrlnm): Likewise.
40792         (vec_rlnm): Likewise.
40793         * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40794         (UNSPEC_VRLNM): Likewise.
40795         (VIlong): New mode iterator.
40796         (altivec_vrl<VI_char>mi): New define_insn.
40797         (altivec_vrl<VI_char>nm): Likewise.
40798         * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40799         function entry.
40800         (VRLDNM): Likewise.
40801         (RLNM): New polymorphic function entry.
40802         (VRLWMI): New monomorphic function entry.
40803         (VRLDMI): Likewise.
40804         (RLMI): New polymorphic function entry.
40805         * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40806         new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40807         * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40808         vec_vrlnm.
40810 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40812         PR debug/78839
40813         * dwarf2out.c (field_byte_offset): Restore the
40814         PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40815         and DECL_FIELD_BIT_OFFSET.  Use fold_build2 instead of build2 + fold.
40816         (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40817         of build2 + fold.
40819 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40821         PR ada/67205
40822         * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40824 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40826         PR debug/71669
40827         * dwarf2out.c (add_data_member_location_attribute): For constant
40828         offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40829         instead of DW_AT_data_member_location, DW_AT_bit_offset and
40830         DW_AT_byte_size attributes.
40832 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40834         * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40835         after forcing to constant memory when the code model is medium.
40837 2017-01-17  Julia Koval  <julia.koval@intel.com>
40839         PR target/76731
40840         * config/i386/avx512fintrin.h
40841         (_mm512_i32gather_ps): Change __addr type to void const*.
40842         (_mm512_mask_i32gather_ps): Ditto.
40843         (_mm512_i32gather_pd): Ditto.
40844         (_mm512_mask_i32gather_pd): Ditto.
40845         (_mm512_i64gather_ps): Ditto.
40846         (_mm512_mask_i64gather_ps): Ditto.
40847         (_mm512_i64gather_pd): Ditto.
40848         (_mm512_mask_i64gather_pd): Ditto.
40849         (_mm512_i32gather_epi32): Ditto.
40850         (_mm512_mask_i32gather_epi32): Ditto.
40851         (_mm512_i32gather_epi64): Ditto.
40852         (_mm512_mask_i32gather_epi64): Ditto.
40853         (_mm512_i64gather_epi32): Ditto.
40854         (_mm512_mask_i64gather_epi32): Ditto.
40855         (_mm512_i64gather_epi64): Ditto.
40856         (_mm512_mask_i64gather_epi64): Ditto.
40857         (_mm512_i32scatter_ps): Change __addr type to void*.
40858         (_mm512_mask_i32scatter_ps): Ditto.
40859         (_mm512_i32scatter_pd): Ditto.
40860         (_mm512_mask_i32scatter_pd): Ditto.
40861         (_mm512_i64scatter_ps): Ditto.
40862         (_mm512_mask_i64scatter_ps): Ditto.
40863         (_mm512_i64scatter_pd): Ditto.
40864         (_mm512_mask_i64scatter_pd): Ditto.
40865         (_mm512_i32scatter_epi32): Ditto.
40866         (_mm512_mask_i32scatter_epi32): Ditto.
40867         (_mm512_i32scatter_epi64): Ditto.
40868         (_mm512_mask_i32scatter_epi64): Ditto.
40869         (_mm512_i64scatter_epi32): Ditto.
40870         (_mm512_mask_i64scatter_epi32): Ditto.
40871         (_mm512_i64scatter_epi64): Ditto.
40872         (_mm512_mask_i64scatter_epi64): Ditto.
40873         * config/i386/avx512pfintrin.h
40874         (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40875         (_mm512_mask_prefetch_i32gather_ps): Ditto.
40876         (_mm512_mask_prefetch_i64gather_pd): Ditto.
40877         (_mm512_mask_prefetch_i64gather_ps): Ditto.
40878         (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40879         (_mm512_prefetch_i32scatter_ps): Ditto.
40880         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40881         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40882         (_mm512_prefetch_i64scatter_pd): Ditto.
40883         (_mm512_prefetch_i64scatter_ps): Ditto.
40884         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40885         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40886         * config/i386/avx512vlintrin.h
40887         (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40888         (_mm_mmask_i32gather_ps): Ditto.
40889         (_mm256_mmask_i32gather_pd): Ditto.
40890         (_mm_mmask_i32gather_pd): Ditto.
40891         (_mm256_mmask_i64gather_ps): Ditto.
40892         (_mm_mmask_i64gather_ps): Ditto.
40893         (_mm256_mmask_i64gather_pd): Ditto.
40894         (_mm_mmask_i64gather_pd): Ditto.
40895         (_mm256_mmask_i32gather_epi32): Ditto.
40896         (_mm_mmask_i32gather_epi32): Ditto.
40897         (_mm256_mmask_i32gather_epi64): Ditto.
40898         (_mm_mmask_i32gather_epi64): Ditto.
40899         (_mm256_mmask_i64gather_epi32): Ditto.
40900         (_mm_mmask_i64gather_epi32): Ditto.
40901         (_mm256_mmask_i64gather_epi64): Ditto.
40902         (_mm_mmask_i64gather_epi64): Ditto.
40903         (_mm256_i32scatter_ps): Change __addr type to void*.
40904         (_mm256_mask_i32scatter_ps): Ditto.
40905         (_mm_i32scatter_ps): Ditto.
40906         (_mm_mask_i32scatter_ps): Ditto.
40907         (_mm256_i32scatter_pd): Ditto.
40908         (_mm256_mask_i32scatter_pd): Ditto.
40909         (_mm_i32scatter_pd): Ditto.
40910         (_mm_mask_i32scatter_pd): Ditto.
40911         (_mm256_i64scatter_ps): Ditto.
40912         (_mm256_mask_i64scatter_ps): Ditto.
40913         (_mm_i64scatter_ps): Ditto.
40914         (_mm_mask_i64scatter_ps): Ditto.
40915         (_mm256_i64scatter_pd): Ditto.
40916         (_mm256_mask_i64scatter_pd): Ditto.
40917         (_mm_i64scatter_pd): Ditto.
40918         (_mm_mask_i64scatter_pd): Ditto.
40919         (_mm256_i32scatter_epi32): Ditto.
40920         (_mm256_mask_i32scatter_epi32): Ditto.
40921         (_mm_i32scatter_epi32): Ditto.
40922         (_mm_mask_i32scatter_epi32): Ditto.
40923         (_mm256_i32scatter_epi64): Ditto.
40924         (_mm256_mask_i32scatter_epi64): Ditto.
40925         (_mm_i32scatter_epi64): Ditto.
40926         (_mm_mask_i32scatter_epi64): Ditto.
40927         (_mm256_i64scatter_epi32): Ditto.
40928         (_mm256_mask_i64scatter_epi32): Ditto.
40929         (_mm_i64scatter_epi32): Ditto.
40930         (_mm_mask_i64scatter_epi32): Ditto.
40931         (_mm256_i64scatter_epi64): Ditto.
40932         (_mm256_mask_i64scatter_epi64): Ditto.
40933         (_mm_i64scatter_epi64): Ditto.
40934         (_mm_mask_i64scatter_epi64): Ditto.
40935         * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
40936         (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
40937         (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
40938         (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
40939         (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
40940         (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
40941         (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
40942         (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
40943         (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
40944         (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
40945         (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
40946         (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
40947         (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
40948         (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
40949         (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
40950         (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
40951         (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
40952         (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
40953         (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
40954         (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
40955         (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
40956         (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
40957         (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
40958         (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
40959         (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
40960         (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
40961         (VOID_QI_V8DI_PCINT_INT_INT): Remove.
40962         (V16SF_V16SF_PCVOID_V16SI_HI_INT,  V8DF_V8DF_PCVOID_V8SI_QI_INT)
40963         (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
40964         (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
40965         (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
40966         (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
40967         (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
40968         (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
40969         (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
40970         (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
40971         (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
40972         (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
40973         (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
40974         (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
40975         (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
40976         (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
40977         (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
40978         (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
40979         (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
40980         (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
40981         (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
40982         (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
40983         (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
40984         (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
40985         (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
40986         (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
40987         (VOID_QI_V8DI_PCVOID_INT_INT): Add.
40988         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
40989         definitions accordingly.
40991 2017-01-17  Kito Cheng  <kito.cheng@gmail.com>
40992             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
40994         PR target/79079
40995         * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
40996         gen_lowpart.
40998 2017-01-17  Vladimir Makarov  <vmakarov@redhat.com>
41000         PR target/79058
41001         * ira-conflicts.c (ira_build_conflicts): Update total conflict
41002         hard regs for inner regno.
41004 2017-01-17  Martin Liska  <mliska@suse.cz>
41006         PR ipa/71207
41007         * ipa-polymorphic-call.c (contains_type_p): Fix wrong
41008         assumption and add comment.
41010 2017-01-17  Nathan Sidwell  <nathan@acm.org>
41012         * ipa-visibility.c (localize_node): New function, broken out of ...
41013         (function_and_variable_visibility): ... here. Call it.
41015 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
41017         PR middle-end/77445
41018         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
41019         correctly set frequency of oudgoing edge.
41020         (duplicate_thread_path): Fix profile updating.
41022 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
41024         PR other/79046
41025         * configure.ac: Add GCC_BASE_VER.
41026         * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
41027         version from BASE-VER file.
41028         (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
41029         (gcc.o): Depend on $(BASEVER).
41030         * common.opt (dumpfullversion): New option.
41031         * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
41032         * doc/invoke.texi: Document -dumpfullversion.
41033         * doc/install.texi: Document --with-gcc-major-version-only.
41034         * configure: Regenerated.
41036 2017-01-17  Richard Biener  <rguenther@suse.de>
41038         PR tree-optimization/71433
41039         * tree-vrp.c (register_new_assert_for): Merge same asserts
41040         on all incoming edges.
41041         (process_assert_insertions_for): Handle insertions at the
41042         beginning of BBs.
41044 2017-01-17  Gerald Pfeifer  <gerald@pfeifer.com>
41046         * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
41047         * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
41049 2017-01-17  Kaz Kojima  <kkojima@gcc.gnu.org>
41051         PR target/78633
41052         * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
41053         RTL sharing.
41055 2017-01-17  Alan Modra  <amodra@gmail.com>
41057         PR target/79066
41058         * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
41059         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
41060         symbolic stack limit when pic.
41062 2017-01-16  Martin Sebor  <msebor@redhat.com>
41064         PR tree-optimization/78608
41065         * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
41067 2017-01-16  Jeff Law  <law@redhat.com>
41069         Revert:
41070         2016-12-02  Tadek Kijkowski  <tkijkowski@gmail.com>
41071         * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
41072         for several include directories that may be relative to sysroot.
41073         * config/i386/x-mingw32 (gplus_includedir): Define.
41074         (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
41075         (native_system_includedir): Likewise.
41076         * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
41077         override if TARGET_SYSTEM_ROOT is defined.
41078         (NATIVE_SYSTEM_HEADER_DIR): Likewise.
41080         PR tree-optimization/79090
41081         PR tree-optimization/33562
41082         PR tree-optimization/61912
41083         PR tree-optimization/77485
41084         * tree-ssa-dse.c (compute_trims): Accept STMT argument.  Dump STMT
41085         and computed trims into the dump file.
41087 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
41089         * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
41091 2017-01-16  Jakub Jelinek  <jakub@redhat.com>
41093         PR c/79089
41094         * gimplify.c (gimplify_init_constructor): If want_value and
41095         object == lhs, unshare lhs to avoid invalid tree sharing.  Formatting
41096         fix.
41098         PR target/79080
41099         * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
41100         sequence.  Formatting fixes.
41101         (doloop_optimize): Formatting fixes.
41103         PR driver/49726
41104         * gcc.c (debug_level_greater_than_spec_func): New function.
41105         (static_spec_functions): Add debug-level-gt spec function.
41106         (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
41107         !g0.
41108         * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41109         * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41110         * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
41111         gpubnames, ggnu-pubnames, gno-record-gcc-switches,
41112         grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
41113         gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
41115 2017-01-16  Uros Bizjak  <ubizjak@gmail.com>
41117         * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
41118         QImode fixups to general and mask registers only.
41120 2017-01-16  Carl Love  <cel@us.ibm.com>
41122         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
41123         for built-in functions
41124         vector signed char vec_nabs (vector signed char)
41125         vector signed short vec_nabs (vector signed short)
41126         vector signed int vec_nabs (vector signed int)
41127         vector signed long long vec_nabs (vector signed long long)
41128         vector float vec_nabs (vector float)
41129         vector double vec_nabs (vector double)
41130         * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
41131         and NABS overload.
41132         * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
41133         * config/rs6000/altivec.h: New define for vec_nabs built-in function.
41134         * doc/extend.texi: Update the documentation file for the new built-in
41135         functions.
41137 2017-01-16  Martin Sebor  <msebor@redhat.com>
41139         * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
41140         message.
41142 2017-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41144         * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
41145         UNSPEC_VSX__XXSPLTD to require special splat handling.
41147 2017-01-16  David Malcolm  <dmalcolm@redhat.com>
41149         PR bootstrap/78616
41150         * system.h: Poison strndup.
41152 2017-01-16  Alan Modra  <amodra@gmail.com>
41154         PR target/79098
41155         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
41156         use a switch.
41158 2017-01-16  Georg-Johann Lay  <avr@gjlay.de>
41160         * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
41162 2017-01-15  Uros Bizjak  <ubizjak@gmail.com>
41164         * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
41165         call recog here.  Assert that INSN_CODE (insn) is non-negative.
41167 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41169         PR target/72749
41170         * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
41171         fallthrough.
41172         * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
41173         in the currently scheduled RTL fragment.
41175 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41177         PR rtl-optimization/78751
41178         * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
41179         give up.
41181 2017-01-14  Jeff Law  <law@redhat.com>
41183         PR tree-optimization/79090
41184         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
41185         variable length stores.
41186         (compute_trims): Delete dead assignment to *trim_tail.
41187         (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
41188         zero length.
41190 2017-01-14  Bernd Schmidt  <bschmidt@redhat.com>
41192         PR rtl-optimization/78626
41193         PR rtl-optimization/78727
41194         * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
41195         of a block, and split such blocks after everything else is finished.
41197 2017-01-14  Alan Modra  <amodra@gmail.com>
41199         PR target/72749
41200         * combine.c (recog_for_combine_1): Set INSN_CODE before calling
41201         target legitimate_combined_insn.
41202         * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
41203         (rs6000_legitimate_combined_insn): New function.
41204         * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
41205         all uses.
41206         (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
41207         (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
41208         (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
41210 2017-01-14  Gerald Pfeifer  <gerald@pfeifer.com>
41212         * doc/frontends.texi (G++ and GCC): Remove references to Java.
41214 2017-01-13  Jeff Law  <law@redhat.com>
41216         PR tree-optimization/33562
41217         PR tree-optimization/61912
41218         PR tree-optimization/77485
41219         * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
41220         a statement.
41221         (delete_dead_assignment): Likewise.
41222         (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
41223         statement to delete_dead_call and delete_dead_assignment.
41225 2017-01-13  David Malcolm  <dmalcolm@redhat.com>
41227         PR c/78304
41228         * substring-locations.c (format_warning_va): Strengthen case 1 so
41229         that both endpoints of the substring must be within the format
41230         range for just the substring to be printed.
41232 2017-01-13  Uros Bizjak  <ubizjak@gmail.com>
41234         * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41235         * config/i386/i386.c (ix86_target_string): Add missing options
41236         to isa_opts and reorder options by implied ISAs.  Rename isa_opts2 to
41237         isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41238         flags_other and ix86_target_other to flags2_other.  Display unknown
41239         isa2 options.
41240         (ix86_valid_target_attribute_inner_p): Add missing options and
41241         reorder options by implied ISAs, as in ix86_target_string.
41243 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41245         * hash-table.h (hash_table::too_empty_p): New function.
41246         (hash_table::expand): Use it.
41247         (hash_table::traverse): Likewise.
41248         (hash_table::empty_slot): Use sizeof (value_type) instead of
41249         sizeof (PTR) to convert bytes to elements.  Shrink the table
41250         if the current size is excessive for the current number of
41251         elements.
41253 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41255         * ira-costs.c (record_reg_classes): Break from the inner loop
41256         early once alt_fail is known to be true.  Update outer loop
41257         handling accordingly.
41259 2017-01-13  Jeff Law  <law@redhat.com>
41261         * tree-ssa-dse.c (decrement_count): New function.
41262         (increment_start_addr, maybe_trim_memstar_call): Likewise.
41263         (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41264         when we know the partially dead statement is a mem* function.
41266         PR tree-optimization/61912
41267         PR tree-optimization/77485
41268         * tree-ssa-dse.c: Include expr.h.
41269         (maybe_trim_constructor_store): New function.
41270         (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41272         PR tree-optimization/33562
41273         PR tree-optimization/61912
41274         PR tree-optimization/77485
41275         * doc/invoke.texi: Document new dse-max-object-size param.
41276         * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41277         * tree-ssa-dse.c: Include params.h.
41278         (dse_store_status): New enum.
41279         (initialize_ao_ref_for_dse): New, partially extracted from
41280         dse_optimize_stmt.
41281         (valid_ao_ref_for_dse, normalize_ref): New.
41282         (setup_live_bytes_from_ref, compute_trims): Likewise.
41283         (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41284         (maybe_trim_partially_dead_store): Likewise.
41285         (maybe_trim_complex_store): Likewise.
41286         (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41287         Track what bytes live from the original store.  Return tri-state
41288         for dead, partially dead or live.
41289         (dse_dom_walker): Add constructor, destructor and new private members.
41290         (delete_dead_call, delete_dead_assignment): New extracted from
41291         dse_optimize_stmt.
41292         (dse_optimize_stmt): Make a member of dse_dom_walker.
41293         Use initialize_ao_ref_for_dse.
41295         PR tree-optimization/33562
41296         PR tree-optimization/61912
41297         PR tree-optimization/77485
41298         * sbitmap.h (bitmap_count_bits): Prototype.
41299         (bitmap_clear_range, bitmap_set_range): Likewise.
41300         * sbitmap.c (bitmap_clear_range): New function.
41301         (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41303 2017-01-13  Martin Liska  <mliska@suse.cz>
41305         PR ipa/79043
41306         * function.c (set_cfun): Add new argument force.
41307         * function.h (set_cfun): Likewise.
41308         * ipa-inline-transform.c (inline_call): Use the function when
41309         strict alising from is dropped for function we inline to.
41311 2017-01-13  Richard Biener  <rguenther@suse.de>
41313         * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41314         for dumping GIMPLE INTEGER_CSTs.
41316 2017-01-13  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
41318         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41319         to 201112L since C++17.
41321 2017-01-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
41323         PR sanitizer/78887
41324         * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41325         if -fsanitize=kernel-address is present.
41327 2017-01-13  Richard Biener  <rguenther@suse.de>
41329         * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41330         as _Literal ( type ) number in case usual suffixes do not
41331         preserve all information.
41333 2017-01-13  Richard Biener  <rguenther@suse.de>
41335         PR tree-optimization/77283
41336         * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41337         and ssa-iterators.h.
41338         (is_feasible_trace): Implement a cost model based on joiner
41339         PHI node uses.
41341 2017-01-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
41343         PR target/79004
41344         * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41345         char or short to __float128/_Float128 directly.
41347 2017-01-12  Martin Sebor  <msebor@redhat.com>
41349         to -Wformat-overflow.
41350         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41351         (min_bytes_remaining): Same.
41352         (get_string_length): Same.
41353         (format_string): Same.
41354         (format_directive): Same.
41355         (add_bytes): Same.
41356         (pass_sprintf_length::handle_gimple_call): Same.
41358 2017-01-12  Jakub Jelinek  <jakub@redhat.com>
41360         * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41361         info.nowrite calls with no lhs that can't throw.  Return bool
41362         whether gsi_remove has been called or not.
41363         (pass_sprintf_length::handle_gimple_call): Return bool whether
41364         try_substitute_return_value called gsi_remove.  Formatting fix.
41365         (pass_sprintf_length::execute): Don't use gsi_remove if
41366         handle_gimple_call returned true.
41368         PR bootstrap/79069
41369         * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41370         be removed due to side-effects, don't remove following barrier nor
41371         turn the successor edge into fallthru edge.
41373 2017-01-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41375         PR target/79044
41376         * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41377         element-reversing loads and stores as not swappable.
41379 2017-01-12  Nathan Sidwell  <nathan@acm.org>
41380             Nicolai Stange  <nicstange@gmail.com>
41382         * combine.c (try_combine): Don't ignore result of overlap checking
41383         loop.  Combine overlap & asm check into single loop.
41385 2017-01-12  Richard Biener  <rguenther@suse.de>
41387         * tree-pretty-print.c (dump_generic_node): Provide -gimple
41388         variant for MEM_REF.  Sanitize INTEGER_CST for -gimple.
41390 2017-01-12  Richard Biener  <rguenther@suse.de>
41392         * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41393         and TS_TARGET_OPTION directly derive from TS_BASE.
41394         * tree-core.h (tree_optimization_option): Derive from tree_base.
41395         (tree_target_option): Likewise.
41397 2017-01-11  Uros Bizjak  <ubizjak@gmail.com>
41399         * config/i386/i386.c (memory_address_length): Increase len
41400         only when rip_relative_addr_p returns false.
41402 2017-01-11  Julia Koval  <julia.koval@intel.com>
41404         * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41405         (OPTION_MASK_ISA_SGX_SET): New.
41406         (ix86_handle_option): Handle OPT_msgx.
41407         * config.gcc: Added sgxintrin.h.
41408         * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41409         * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41410         * config/i386/i386.c (ix86_target_string): Add -msgx.
41411         (PTA_SGX): New.
41412         (ix86_option_override_internal): Handle new options.
41413         (ix86_valid_target_attribute_inner_p): Add sgx.
41414         * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41415         * config/i386/i386.opt: Add msgx.
41416         * config/i386/sgxintrin.h: New file.
41417         * config/i386/x86intrin.h: Add sgxintrin.h.
41419 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41421         PR c++/71537
41422         * fold-const.c (maybe_nonzero_address): Return 1 for function
41423         local objects.
41424         (tree_single_nonzero_warnv_p): Don't handle function local objects
41425         here.
41427         PR c++/72813
41428         * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41429         of c-header.
41431 2017-01-11  David Malcolm  <dmalcolm@redhat.com>
41433         PR driver/78877
41434         * opts.c: Include "spellcheck.h"
41435         (struct string_fragment): New struct.
41436         (struct edit_distance_traits<const string_fragment &>): New
41437         struct.
41438         (get_closest_sanitizer_option): New function.
41439         (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41441 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41443         * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41444         by 12.
41445         (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41446         DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41447         (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41448         (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41449         for initial die_offset if dwarf_split_debug_info.
41450         (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41451         initial next_die_offset if dwo_id is non-NULL.  Don't emit padding
41452         fields.
41453         (output_skeleton_debug_sections): Formatting fix.  Use
41454         DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41455         DWARF_COMPILE_UNIT_HEADER_SIZE.  Don't emit padding.
41457 2017-01-11  Wilco Dijkstra  <wdijkstr@arm.com>
41459         * config/arm/cortex-a53.md: Add bypasses for
41460         cortex_a53_r2f_cvt.
41461         (cortex_a53_r2f): Only use for transfers.
41462         (cortex_a53_f2r): Likewise.
41463         (cortex_a53_r2f_cvt): Add reservation for conversions.
41464         (cortex_a53_f2r_cvt): Likewise.
41466 2017-01-11  Tamar Christina  <tamar.christina@arm.com>
41468         * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41469         to all inlined functions, change static to extern.
41471 2017-01-11  Christophe Lyon  <christophe.lyon@linaro.org>
41473         PR target/78253
41474         * config/arm/arm.c (legitimize_pic_address): Handle reference to
41475         weak symbol.
41476         (arm_assemble_integer): Likewise.
41478 2017-01-11  Richard Earnshaw  <rearnsha@arm.com>
41480         * config.gcc: Use new awk script to check CPU, FPU and architecture
41481         parameters for --with-... options.
41482         * config/arm/parsecpu.awk: New file
41483         * config/arm/arm-cpus.in: New file.
41484         * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41485         files.
41486         * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41487         files.
41488         * config/arm/t-arm: Update dependency rules.
41489         * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41490         of processing .def files.
41491         * config/arm/genopt.sh: Deleted.
41492         * config/arm/gentune.sh: Deleted.
41493         * config/arm/arm-cores.def: Deleted.
41494         * config/arm/arm-arches.def: Deleted.
41495         * config/arm/arm-fpus.def: Deleted.
41496         * config/arm/arm-tune.md: Regenerated.
41497         * config/arm/arm-tables.opt: Regenerated.
41498         * config/arm/arm-cpu.h: New generated file.
41499         * config/arm/arm-cpu-data.h: New generated file.
41500         * config/arm/arm-cpu-cdata.h: New generated file.
41502 2017-01-11  Maxim Ostapenko  <m.ostapenko@samsung.com>
41504         PR lto/79042
41505         * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41506         bit.
41507         (input_varpool_node): Unpack dynamically_initialized bit.
41509 2017-01-11  Eric Botcazou  <ebotcazou@adacore.com>
41511         PR rtl-optimization/79032
41512         * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41513         the alignment of the adjusted memory reference against that of MODE,
41514         instead of the alignment of the original memory reference.
41516 2017-01-11  Martin Jambor  <mjambor@suse.cz>
41518         * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41519         test.
41520         * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41521         decorated functions.
41523 2017-01-11  Richard Biener  <rguenther@suse.de>
41525         * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41526         set range/nonnull info for PHI results.  Do not set it on
41527         stmts marked for removal.
41529 2017-01-10  Eric Botcazou  <ebotcazou@adacore.com>
41531         * expr.c (store_field): In the bitfield case, fetch the return value
41532         from the registers before applying a single big-endian adjustment.
41533         Always do a final load for a BLKmode value not larger than a word.
41535 2017-01-10  David Malcolm  <dmalcolm@redhat.com>
41537         PR c++/77949
41538         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41539         that we correctly handle column numbers greater than
41540         LINE_MAP_MAX_COLUMN_NUMBER.
41542 2017-01-10  Martin Sebor  <msebor@redhat.com>
41544         PR middle-end/78245
41545         * gimple-ssa-sprintf.c (get_destination_size): Call
41546         {init,fini}object_sizes.
41547         * tree-object-size.c (addr_object_size): Adjust.
41548         (pass_through_call): Adjust.
41549         (pass_object_sizes::execute): Adjust.
41550         * tree-object-size.h (fini_object_sizes): Declare.
41552 2017-01-10  Martin Sebor  <msebor@redhat.com>
41554         PR tree-optimization/78775
41555         * builtins.c (get_size_range): Move...
41556         * calls.c: ...to here.
41557         (alloc_max_size): Accept zero argument.
41558         (operand_signed_p): Remove.
41559         (maybe_warn_alloc_args_overflow): Call get_size_range.
41560         * calls.h (get_size_range): Declare.
41562 2017-01-10  Joe Seymour  <joe.s@somniumtech.com>
41564         * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41565         from TI's devices.csv file as of September 2016.
41566         * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41568 2017-01-10  Sandra Loosemore  <sandra@codesourcery.com>
41570         * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41571         * doc/invoke.texi: Likewise.
41572         * doc/md.texi: Likewise.
41573         * doc/objc.texi: Likewise.
41575 2017-01-10  Joshua Conner  <joshconner@google.com>
41577         * config/arm/fuchsia-elf.h: New file.
41578         * config/fuchsia.h: New file.
41579         * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41580         (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41581         targets.
41582         * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41584 2016-01-10  Richard Biener  <rguenther@suse.de>
41586         PR tree-optimization/79034
41587         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41588         Propagate out degenerate PHIs in the joiner.
41590 2017-01-10  Martin Liska  <mliska@suse.cz>
41592         * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41593         (sort_congruence_classes_by_decl_uid): Likewise.
41594         (sort_congruence_class_groups_by_decl_uid): Likewise.
41595         (sem_item_optimizer::merge_classes): Sort class, groups in these
41596         classes and members in the groups by DECL_UID of declarations.
41597         This would make merge operations stable.
41599 2017-01-10  Martin Liska  <mliska@suse.cz>
41601         * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41602         usage of m_classes_vec.
41603         (sem_item_optimizer::~sem_item_optimizer):  Likewise.
41604         (sem_item_optimizer::get_group_by_hash): Likewise.
41605         (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41606         (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41607         (sem_item_optimizer::verify_classes): Likewise.
41608         (sem_item_optimizer::process_cong_reduction): Likewise.
41609         (sem_item_optimizer::dump_cong_classes): Likewise.
41610         (sem_item_optimizer::merge_classes): Likewise.
41611         * ipa-icf.h (congruence_class_hash): Rename from
41612         congruence_class_group_hash.  Remove declaration of m_classes_vec.
41614 2017-01-10  Andrew Senkevich  <andrew.senkevich@intel.com>
41616         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41617         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41618         * config.gcc: Add avx512vpopcntdqintrin.h.
41619         * config/i386/avx512vpopcntdqintrin.h: New.
41620         * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41621         * config/i386/i386-builtin-types.def: Add new types.
41622         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41623         __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41624         __builtin_ia32_vpopcountq_v8di_mask): New.
41625         * config/i386/i386-c.c (ix86_target_macros_internal): Define
41626         __AVX512VPOPCNTDQ__.
41627         * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41628         (PTA_AVX512VPOPCNTDQ): Define.
41629         * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41630         TARGET_AVX512VPOPCNTDQ_P): Define.
41631         * config/i386/i386.opt: Add mavx512vpopcntdq.
41632         * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41633         * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41635 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41637         PR middle-end/77484
41638         * predict.def (PRED_CALL): Set to 67.
41640 2017-01-09  Eric Botcazou  <ebotcazou@adacore.com>
41642         * expr.c (store_field): In the bitfield case, if the value comes from
41643         a function call and is of an aggregate type returned in registers, do
41644         not modify the field mode; extract the value in all cases if the mode
41645         is BLKmode and the size is not larger than a word.
41647 2017-01-09  Dominique d'Humieres  <dominiq@lps.ens.fr>
41649         PR target/71017
41650         * config/i386/cpuid.h: Fix undefined behavior.
41652 2017-01-04  Jeff Law  <law@redhat.com>
41654         PR tree-optimization/79007
41655         PR tree-optimization/67955
41656         * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41657         conservative for pt.null when flag_non_call_exceptions is on.
41659 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41661         PR translation/79019
41662         PR translation/79020
41663         * params.def (PARAM_INLINE_MIN_SPEEDUP,
41664         PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41665         PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41666         in descriptions.
41667         * config/avr/avr.opt (maccumulate-args): Likewise.
41668         * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41669         * common.opt (freport-bug): Likewise.
41670         * cif-code.def (CIF_FINAL_ERROR): Likewise.
41671         * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41672         * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41673         translatable string.
41674         * config/i386/i386.c (function_value_32): Likewise.
41675         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41676         * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41677         Likewise.
41678         * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41679         * common/config/msp430/msp430-common.c (msp430_handle_option):
41680         Likewise.
41681         * symtab.c (symtab_node::verify_base): Likewise.
41682         * opts.c (set_debug_level): Likewise.
41683         * tree.c (verify_type_variant): Likewise.  Fix typo in comment.
41684         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41685         missing whitespace to translatable strings.
41686         * config/avr/avr.md (bswapsi2): Fix typo in comment.
41687         * config/sh/superh.h: Likewise.
41688         * config/i386/xopintrin.h: Likewise.
41689         * config/i386/znver1.md: Likewise.
41690         * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41691         * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41692         * double-int.h (struct double_int): Likewise.
41693         * double-int.c (div_and_round_double): Likewise.
41694         * wide-int.cc: Likewise.
41695         * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41696         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41697         * cfgcleanup.c (crossjumps_occured): Renamed to ...
41698         (crossjumps_occurred): ... this.
41699         (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41700         Adjust all uses.
41702         PR tree-optimization/78899
41703         * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41704         returning bool return struct loop *, NULL for failure and the new
41705         loop on success.
41706         (versionable_outer_loop_p): Don't version outer loop if it has
41707         dont_vectorized bit set.
41708         (tree_if_conversion): When versioning outer loop, ensure
41709         tree_if_conversion is performed also on the inner loop of the
41710         non-vectorizable outer loop copy.
41711         * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix.  Fold
41712         LOOP_VECTORIZED in inner loop of the scalar outer loop and
41713         prevent vectorization of it.
41714         (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41715         the outer loop vectorization of the non-scalar version is attempted
41716         before vectorization of the inner loop in scalar version.  If
41717         outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41718         vectorization of its inner loop.
41719         * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41720         has 2 inner loops, rename also on edges from bb whose single pred
41721         is outer_loop->header.  Fix typo in function comment.
41723 2017-01-09  Martin Sebor  <msebor@redhat.com>
41725         PR bootstrap/79033
41726         * asan.c (asan_emit_stack_protection): Increase local buffer size
41727         to avoid snprintf truncation warning.
41729 2017-01-09  Andrew Pinski  <apinski@cavium.com>
41731         * config/aarch64/aarch64-cores.def: Add thunderx2t99.  Change vulcan
41732         to reference thunderx2t99 for the tuning structure
41733         * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41734         Rename to ...
41735         (thunderx2t99_extra_costs): This.
41736         * config/aarch64/aarch64-tune.md: Regenerate.
41737         * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41738         (vulcan_addrcost_table): This.
41739         (vulcan_regmove_cost): Rename to ...
41740         (thunderx2t99_regmove_cost): This.
41741         (vulcan_vector_cost): Rename to ...
41742         (thunderx2t99_vector_cost): this.
41743         (vulcan_branch_cost): Rename to ...
41744         (thunderx2t99_branch_cost): This.
41745         (vulcan_tunings): Rename to ...
41746         (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41747         * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41749 2017-01-09  Martin Jambor  <mjambor@suse.cz>
41751         PR ipa/78365
41752         PR ipa/78599
41753         * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41754         * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41755         (propagate_vr_accross_jump_function): Use the above function for all
41756         value range computations for pass-through jump functions and type
41757         converasion from explicit value range values.
41758         (ipcp_propagate_stage): Do not attempt to deduce types of formal
41759         parameters from TYPE_ARG_TYPES.
41760         * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41761         (ipa_write_node_info): Stream type of the actual argument.
41762         (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41764 2017-01-09  Martin Liska  <mliska@suse.cz>
41766         PR pch/78970
41767         * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41768         (lookup_compiler): Do not show error message with have_E.
41770 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41772         PR tree-optimization/78938
41773         * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41774         where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41775         BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41776         {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR.  Formatting
41777         fixes.
41779 2017-01-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
41781         * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41782         is const0_rtx.
41784 2017-01-09  Richard Biener  <rguenther@suse.de>
41786         PR tree-optimization/78997
41787         * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41788         name condition properly.
41790 2017-01-09  Richard Biener  <rguenther@suse.de>
41792         PR debug/79000
41793         * dwarf2out.c (is_cxx): New overload with context.
41794         (is_naming_typedef_decl): Use it.
41796 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41798         * invoke.texi (Option Summary): Correct spacing in option lists
41799         and add line breaks to fix over-long lines.
41801 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41803         PR middle-end/17660
41805         * extend.texi (Common Variable Attributes): Add xref to GCC
41806         Internals manual to explain mode attribute keywords.
41808 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41810         PR other/16519
41811         * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41812         and Preprocessor Options.
41813         (Options for Linking): Document -pthread here....
41814         (RS/6000 and PowerPC Options): ...not here.
41815         (Solaris 2 Options): ...or here.
41816         * doc/cppopts.texi: Document -pthread.
41818 2017-01-08  Martin Sebor  <msebor@redhat.com>
41820         PR middle-end/77708
41821         * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41822         * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41823         New member functions.
41824         (format_directive): Used them.
41825         (add_bytes): Same.
41826         (pass_sprintf_length::handle_gimple_call): Same.
41827         * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41828         to avoid truncation for any argument.
41829         (extract_affine_mul): Same.
41830         * tree.c (get_file_function_name): Same.
41832 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41834         PR middle-end/77484
41835         * predict.def (PRED_INDIR_CALL): Set to 86.
41837 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41839         PR preprocessor/54124
41840         * doc/cppopts.texi: Reformat -d subtable to list the full name
41841         of the options.  Add cross-reference to the docs for the general
41842         compiler -d options.
41843         * doc/invoke.texi (Developer Options): Add cross-reference to the
41844         preprocessor-specific -d option documentation.
41846 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41848         PR preprocessor/13498
41849         * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41850         redudant material, and reflect new command-line options.
41851         (System Headers): Likewise.
41853 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41855         * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41856         -isystem, and -idirafter.  Copy-edit.
41857         * doc/cppopts.texi: Copy-edit.  Remove contradiction about
41858         default for -ftrack-macro-expansion.  Delete obsolete and
41859         badly-formatted implementation details about -fdebug-cpp output.
41860         * doc/cppwarnopts.texi: Copy-edit.
41862 2017-01-07  David Malcolm  <dmalcolm@redhat.com>
41864         PR c++/72803
41865         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41866         that the transition from a max line width >= 1<<10 to narrower
41867         lines works correctly.
41869 2017-01-07  Alexandre Oliva  <aoliva@redhat.com>
41871         * doc/options.texi (PerFunction): New.
41872         * opt-functions.awk (switch_flags): Map both Optimization and
41873         PerFunction to CL_OPTIMIZATION.
41874         * opth-gen.awk: Test for PerFunction flag along with
41875         Optimization.
41876         * optc-save-gen.awk: Likewise.  Introduce var_opt_hash and set
41877         it only when the latter is present.  Skip those that don't in
41878         the hash function generator.
41879         * common.opt (fvar-tracking): Mark as PerFunction instead of
41880         Optimization.
41881         (fvar-tracking-assignments): Likewise.
41882         (fvar-tracking-assignments-toggle): Likewise.
41883         (fvar-tracking-uninit): Likewise.
41885 2017-01-07  Jakub Jelinek  <jakub@redhat.com>
41887         PR translation/79018
41888         * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41889         the and store.
41891 2017-01-06  Mikael Pettersson  <mikpelinux@gmail.com>
41893         PR target/57583
41894         * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41895         * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41896         TARGET_LONG_JUMP_TABLE_OFFSETS.
41897         * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
41898         * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
41899         * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
41900         (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
41901         * config/m68k/m68k.md (tablejump expander): Likewise.
41902         (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
41903         TARGET_LONG_JUMP_TABLE_OFFSETS.
41904         (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
41905         * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
41907 2017-01-06  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
41908             David Holsgrove <david.holsgrove@xilinx.com>
41910         * common/config/microblaze/microblaze-common.c
41911         (TARGET_EXCEPT_UNWIND_INFO): Remove.
41912         * config/microblaze/microblaze-protos.h (microblaze_eh_return):
41913         New prototype.
41914         * config/microblaze/microblaze.c (microblaze_must_save_register)
41915         (microblaze_expand_epilogue, microblaze_return_addr): Handle
41916         calls_eh_return.
41917         (microblaze_eh_return): New function.
41918         * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
41919         (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
41920         (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
41921         * config/microblaze/microblaze.md (eh_return): New pattern.
41923 2017-01-06  Jakub Jelinek  <jakub@redhat.com>
41925         * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
41926         GCC_DIAGNOSTIC_STRINGIFY): Define.
41928         * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
41930 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41932         * config/arm/arm.md (<mcrr>): New.
41933         (<mrrc>): New.
41934         * config/arm/arm.c (arm_arch5te): New.
41935         (arm_option_override): Set arm_arch5te.
41936         (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
41937         and mrrc2.
41938         * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
41939         (arm_mcrr_qualifiers): ... this. New.
41940         (MRRC_QUALIFIERS): Define to...
41941         (arm_mrrc_qualifiers): ... this. New.
41942         * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
41943         __arm_mrrc2): New.
41944         * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
41945         * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
41946         (MRRCI, mrrc, MRRC): New.
41947         * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
41948         VUNSPEC_MRRC2): New.
41950 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41952         * config/arm/arm.md (<mcr>): New.
41953         (<mrc>): New.
41954         * config/arm/arm.c (arm_coproc_builtin_available): Add
41955         support for mcr, mrc, mcr2 and mrc2.
41956         * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
41957         (arm_mcr_qualifiers): ... this. New.
41958         (MRC_QUALIFIERS): Define to ...
41959         (arm_mrc_qualifiers): ... this. New.
41960         (MCR_QUALIFIERS): Define to ...
41961         (arm_mcr_qualifiers): ... this. New.
41962         * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
41963         __arm_mrc2): New.
41964         * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
41965         * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
41966         * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
41967         VUNSPEC_MRC2): New.
41969 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41971         * config/arm/arm.md (*ldc): New.
41972         (*stc): New.
41973         (<ldc>): New.
41974         (<stc>): New.
41975         * config/arm/arm.c (arm_coproc_builtin_available): Add
41976         support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
41977         (arm_coproc_ldc_stc_legitimate_address): New.
41978         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
41979         'qualifier_const_pointer'.
41980         (LDC_QUALIFIERS): Define to...
41981         (arm_ldc_qualifiers): ... this. New.
41982         (STC_QUALIFIERS): Define to...
41983         (arm_stc_qualifiers): ... this. New.
41984         * config/arm/arm-protos.h
41985         (arm_coproc_ldc_stc_legitimate_address): New.
41986         * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
41987         __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
41988         * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
41989         stc2, stcl, stc2l): New.
41990         * config/arm/constraints.md (Uz): New.
41991         * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
41992         * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
41993         VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
41994         VUNSPEC_STC2L): New.
41996 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41998         * config/arm/arm.md (<cdp>): New.
41999         * config/arm/arm.c (neon_const_bounds): Rename this ...
42000         (arm_const_bounds): ... this.
42001         (arm_coproc_builtin_available): New.
42002         * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
42003         (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
42004         (CDP_QUALIFIERS): Define to...
42005         (arm_cdp_qualifiers): ... this. New.
42006         (void_UP): Define.
42007         (arm_expand_builtin_args): Add case for 6 arguments.
42008         * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
42009         (arm_const_bounds): ... this.
42010         (arm_coproc_builtin_available): New.
42011         * config/arm/arm_acle.h (__arm_cdp): New.
42012         (__arm_cdp2): New.
42013         * config/arm/arm_acle_builtins.def (cdp): New.
42014         (cdp2): New.
42015         * config/arm/iterators.md (CDPI,CDP,cdp): New.
42016         * config/arm/neon.md: Rename all 'neon_const_bounds' to
42017         'arm_const_bounds'.
42018         * config/arm/types.md (coproc): New.
42019         * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
42020         * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
42021         * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
42022         arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
42024 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42026         * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
42027         (UBINOP_QUALIFIERS): New.
42028         (si_UP): Define.
42029         (acle_builtin_data): New. Change comment.
42030         (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
42031         ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
42032         ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
42033         arm_acle_builtins.def.
42034         (ARM_BUILTIN_ACLE_PATTERN_START): Define.
42035         (arm_init_acle_builtins): New.
42036         (CRC32_BUILTIN): Remove.
42037         (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
42038         crc32cb, crc32ch and crc32cw.
42039         (arm_init_crc32_builtins): Remove.
42040         (arm_init_builtins): Use arm_init_acle_builtins rather
42041         than arm_init_crc32_builtins.
42042         (arm_expand_acle_builtin): New.
42043         (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
42044         * config/arm/arm_acle_builtins.def: New.
42046 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
42048         * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
42049         (arm_builtin_datum): ... this.
42050         (arm_init_neon_builtin): Rename to ...
42051         (arm_init_builtin): ... this. Add a new parameters PREFIX
42052         and USE_SIG_IN_NAME.
42053         (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
42054         'arm_init_builtin'. Replace type 'neon_builtin_datum' with
42055         'arm_builtin_datum'.
42056         (arm_init_vfp_builtins): Likewise.
42057         (builtin_arg): Rename enum's replacing 'NEON_ARG' with
42058         'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
42059         (arm_expand_neon_args): Rename to ...
42060         (arm_expand_builtin_args): ... this. Rename builtin_arg
42061         enum values and differentiate between ARG_BUILTIN_MEMORY
42062         and ARG_BUILTIN_NEON_MEMORY.
42063         (arm_expand_neon_builtin_1): Rename to ...
42064         (arm_expand_builtin_1): ... this. Rename builtin_arg enum
42065         values, arm_expand_builtin_args and add bool parameter NEON.
42066         (arm_expand_neon_builtin): Use arm_expand_builtin_1.
42067         (arm_expand_vfp_builtin): Likewise.
42068         (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
42070 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42072         PR middle-end/77484
42073         * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
42074         * predict.c (tree_estimate_probability_bb): Reverse direction of
42075         polymorphic call predictor.
42077 2017-01-06  David Malcolm  <dmalcolm@redhat.com>
42079         * passes.c (execute_one_pass): Split out pass-skipping logic into...
42080         (determine_pass_name_match): ...this new function and...
42081         (should_skip_pass_p): ...this new function.
42083 2017-01-06  Nathan Sidwell  <nathan@acm.org>
42085         * ipa-visibility.c (function_and_variable_visibility): Reformat
42086         comments and long lines.  Remove extrneous if.
42087         * symtab.c (symtab_node::make_decl_local): Fix code format.
42088         (symtab_node::set_section_for_node): Fix comment typo.
42090 2017-01-06  Martin Liska  <mliska@suse.cz>
42092         PR bootstrap/79003
42093         * lra-constraints.c: Rename invariant to lra_invariant.
42094         * predict.c (set_even_probabilities): Initialize e to NULL.
42096 2017-01-05  Martin Sebor  <msebor@redhat.com>
42098         PR tree-optimization/78910
42099         * gimple-ssa-sprintf.c (tree_digits): Add an argument.
42100         (format_integer): Correct off-by-one error in the handling
42101         of precision with negative numbers in signed conversions..
42103 2017-01-05  Eric Botcazou  <ebotcazou@adacore.com>
42105         * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
42107 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42109         PR tree-optimization/71016
42110         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
42111         factor_out_conditional_conversion.  Formatting fix.
42112         (factor_out_conditional_conversion): Add cond_stmt argument.
42113         If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
42114         cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
42115         Formatting fix.
42117 2017-01-05  David Malcolm  <dmalcolm@redhat.com>
42119         * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
42120         read-rtl-function.o, and selftest-rtl.o.
42121         * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
42122         (selftest::aarch64_test_loading_full_dump): New function.
42123         (selftest::aarch64_run_selftests): New function.
42124         (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
42125         selftest::aarch64_run_selftests.
42126         * config/i386/i386.c
42127         (selftest::ix86_test_loading_dump_fragment_1): New function.
42128         (selftest::ix86_test_loading_call_insn): New function.
42129         (selftest::ix86_test_loading_full_dump): New function.
42130         (selftest::ix86_test_loading_unspec): New function.
42131         (selftest::ix86_run_selftests): Call the new functions.
42132         * emit-rtl.c (maybe_set_max_label_num): New function.
42133         * emit-rtl.h (maybe_set_max_label_num): New decl.
42134         * function.c (instantiate_decls): Guard call to
42135         instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
42136         * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
42137         "static".
42138         * gensupport.c (gen_reader::gen_reader): Pass "false"
42139         for new "compact" param of rtx_reader.
42140         * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
42141         rather than an empty string for NULL strings.
42142         * read-md.c: Potentially include config.h rather than bconfig.h.
42143         Wrap include of errors.h with #ifdef GENERATOR_FILE.
42144         (have_error): New global, copied from errors.c.
42145         (md_reader::read_name): Rename to...
42146         (md_reader::read_name_1): ...this, adding "out_loc" param,
42147         and converting "missing name or number" to returning false, rather
42148         than failing.
42149         (md_reader::read_name): Reimplement in terms of read_name_1.
42150         (md_reader::read_name_or_nil): New function.
42151         (md_reader::read_string): Handle "(nil)" by returning NULL.
42152         (md_reader::md_reader): Add new param "compact".
42153         (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
42154         (md_reader::read_file): New method.
42155         * read-md.h (md_reader::md_reader): Add new param "compact".
42156         (md_reader::read_file): New method.
42157         (md_reader::is_compact): New accessor.
42158         (md_reader::read_name): Convert return type from void to file_location.
42159         (md_reader::read_name_or_nil): New decl.
42160         (md_reader::read_name_1): New decl.
42161         (md_reader::m_compact): New field.
42162         (noop_reader::noop_reader): Pass "false" for new "compact" param
42163         of rtx_reader.
42164         (rtx_reader::rtx_reader): Add new "compact" param.
42165         (rtx_reader::read_rtx_operand): Make virtual and convert return
42166         type from void to rtx.
42167         (rtx_reader::read_until): New decl.
42168         (rtx_reader::handle_any_trailing_information): New virtual function.
42169         (rtx_reader::postprocess): New virtual function.
42170         (rtx_reader::finalize_string): New virtual function.
42171         (rtx_reader::m_in_call_function_usage): New field.
42172         (rtx_reader::m_reuse_rtx_by_id): New field.
42173         * read-rtl-function.c: New file.
42174         * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
42175         * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
42176         (selftest::verify_three_block_rtl_cfg): New decl.
42177         * read-rtl-function.h: New file.
42178         * read-rtl.c: Potentially include config.h rather than bconfig.h.
42179         For host, include function.h, memmodel.h, and emit-rtl.h.
42180         (one_time_initialization): New function.
42181         (struct compact_insn_name): New struct.
42182         (compact_insn_names): New array.
42183         (find_code): Handle insn codes in compact dumps.
42184         (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
42185         (bind_subst_iter_and_attr): Likewise.
42186         (add_condition_to_string): Likewise.
42187         (add_condition_to_rtx): Likewise.
42188         (apply_attribute_uses): Likewise.
42189         (add_current_iterators): Likewise.
42190         (apply_iterators): Likewise.
42191         (initialize_iterators): Guard usage of apply_subst_iterator with
42192         #ifdef GENERATOR_FILE.
42193         (read_conditions): Wrap with #ifdef GENERATOR_FILE.
42194         (md_reader::read_mapping): Likewise.
42195         (add_define_attr_for_define_subst): Likewise.
42196         (add_define_subst_attr): Likewise.
42197         (read_subst_mapping): Likewise.
42198         (check_code_iterator): Likewise.
42199         (rtx_reader::read_rtx): Likewise.  Move one-time initialization
42200         logic to...
42201         (one_time_initialization): New function.
42202         (rtx_reader::read_until): New method.
42203         (read_flags): New function.
42204         (parse_reg_note_name): New function.
42205         (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
42206         Handle reuse_rtx ids.
42207         Wrap iterator lookup within #ifdef GENERATOR_FILE.
42208         Add parsing support for RTL dumps, mirroring the special-cases in
42209         print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
42210         values, and calling handle_any_trailing_information.
42211         (rtx_reader::read_rtx_operand): Convert return type from void
42212         to rtx, returning return_rtx.  Handle case 'e'.  Call
42213         finalize_string on XSTR and XTMPL fields.
42214         (rtx_reader::read_nested_rtx):  Handle dumps in which trailing
42215          "(nil)" values were omitted.  Call the postprocess vfunc on the
42216         return_rtx.
42217         (rtx_reader::rtx_reader): Add new "compact" param and pass to base
42218         class ctor.  Initialize m_in_call_function_usage.  Call
42219         one_time_initialization.
42220         * rtl-tests.c (selftest::test_uncond_jump): Call
42221         set_new_first_and_last_insn.
42222         * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
42223         * selftest-rtl.c: New file.
42224         * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42225         (selftest::get_insn_by_uid): New decl.
42226         * selftest-run-tests.c (selftest::run_tests): Call
42227         read_rtl_function_c_tests.
42228         * selftest.h  (selftest::read_rtl_function_c_tests): New decl.
42229         * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42230         dumps.
42232 2017-01-05  Uros Bizjak  <ubizjak@gmail.com>
42234         * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42235         operands in a special way.  Assert that pos+len <= mode precision.
42237 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42239         * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42240         3 argument Alias with unlimited for the negative form.
42241         (fno-vect-cost-model): Removed.
42243 2017-01-05  Martin Liska  <mliska@suse.cz>
42245         * hsa-gen.c (gen_hsa_divmod): New function.
42246         (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42248 2017-01-05  Martin Liska  <mliska@suse.cz>
42250         PR pch/78970
42251         * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42252         header.
42254 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42256         * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42257         small constant length operands.
42259 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42261         * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42262         between loop iterations.
42264 2017-01-05  Martin Liska  <mliska@suse.cz>
42266         PR sanitizer/78815
42267         * gimplify.c (gimplify_decl_expr): Compare to
42268         asan_poisoned_variables instread of checking flags.
42269         (gimplify_target_expr): Likewise.
42270         (gimplify_expr): Likewise.
42271         (gimplify_function_tree): Conditionally initialize
42272         asan_poisoned_variables.
42274 2017-01-04  Jeff Law  <law@redhat.com>
42276         PR tree-optimizatin/78812
42277         * rtl.h (contains_mem_rtx_p): Prototype.
42278         * ifcvt.c (containts_mem_rtx_p): Move from here to...
42279         * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42280         * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42281         and prune MEMs that are not at the toplevel of a SET_SRC rtx.  Look
42282         through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42284 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42286         * input.c (assert_char_at_range): Default-initialize actual_range.
42288 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42290         * df-scan.c (df_ref_create_structure): Make regno unsigned,
42291         to match the caller.
42293 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42295         * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42296         insns after final jump in test to emit dummy move.
42298 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42300         * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42301         * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42303 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42305         * multiple_target.c (create_dispatcher_calls): Init e_next.
42306         * tree-ssa-loop-split.c (split_loop): Init border.
42307         * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42308         scalar_type.
42310 2017-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
42312         PR target/71977
42313         PR target/70568
42314         PR target/78823
42315         * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42316         (altivec_register_operand): Do not return true if the operand
42317         contains a SUBREG mixing SImode and SFmode.
42318         (vsx_register_operand): Likewise.
42319         (vsx_reg_sfsubreg_ok): New predicate.
42320         (vfloat_operand): Do not return true if the operand contains a
42321         SUBREG mixing SImode and SFmode.
42322         (vint_operand): Likewise.
42323         (vlogical_operand): Likewise.
42324         (gpc_reg_operand): Likewise.
42325         (int_reg_operand): Likewise.
42326         * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42327         * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42328         determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42329         SImode and SFmode.
42330         (rs6000_emit_move_si_sf_subreg): New helper function.
42331         (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42332         fixup SUBREGs involving SImode and SFmode.
42333         * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42334         numbers for the new peephole2 optimization.
42335         (peephole2 for SFmode unions): New peephole2 to optimize cases in
42336         the GLIBC math library that do AND/IOR/XOR operations on single
42337         precision floating point.
42338         * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42339         target macros to say whether we need to avoid SUBREGs mixing
42340         SImode and SFmode.
42341         (TARGET_ALLOW_SF_SUBREG): Likewise.
42342         * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42343         (UNSPEC_SI_FROM_SF): Likewise.
42344         (iorxor): Change spacing.
42345         (and_ior_xor): New iterator for AND, IOR, and XOR.
42346         (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42347         (movdi_from_sf_zero_ext): Likewise.
42348         (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42349         instead of gpc_reg_operand.  Add SImode/SFmode SUBREG support.
42350         (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42351         (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42352         (fms<mode>4): Likewise.
42353         (fnma<mode>4): Likewise.
42354         (fnms<mode>4): Likewise.
42355         (nfma<mode>4): Likewise.
42356         (nfms<mode>4): Likewise.
42358 2017-01-04  Marek Polacek  <polacek@redhat.com>
42360         PR c++/64767
42361         * doc/invoke.texi: Document -Wpointer-compare.
42363 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42365         * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42366         RejectNegative.
42368         * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42369         descriptions for -gdwarf-5 and emit them as uleb128 instead of
42370         2-byte data.
42372 2017-01-04  Kelvin Nilsen  <kelvin@gcc.gnu.org>
42374         PR target/78056
42375         * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42376         documentation of the powerpc_popcntb_ok attribute.
42377         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42378         code to issue warning messages if a requested CPU configuration is
42379         not supported by the binary (assembler and loader) toolchain.
42380         (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42381         made to define a built-in function that has been disabled.
42382         (paired_init_builtins): Add assertion to prevent ICE if attempt is
42383         made to define a built-in function that has been disabled.
42384         (altivec_init_builtins): Add comment explaining why definition
42385         of the DST built-in functions is not preceded by an assertion
42386         check.  Add assertions to prevent ICE if attempts are made to
42387         define an altivec predicate or an abs* built-in function that has
42388         been disabled.
42389         (htm_init_builtins): Add comment explaining why definition of the
42390         htm built-in functions is not preceded by an assertion check.
42392 2017-01-04  Jeff Law  <law@redhat.com>
42394         PR tree-optimizatin/67955
42395         * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42396         Allow any SSA_VAR_P as the base objects.  Use integer_zerop.  Verify
42397         the points-to solution does not include pt_null.  Use DECL_PT_UID
42398         unconditionally.
42400 2017-01-04  Uros Bizjak  <ubizjak@gmail.com>
42402         * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42403         Use gen_int_mode instead of gen_lopwart for const_int operands.
42405 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42407         PR tree-optimization/71563
42408         * match.pd: Simplify X << Y into X if Y is known to be 0 or
42409         out of range value - has low bits known to be zero.
42411 2017-01-04  Alan Modra  <amodra@gmail.com>
42413         * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42414         * configure: Regenerate.
42415         * config.in: Regenerate.
42417 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42419         PR bootstrap/77569
42420         * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42421         a substring of the message, but strcmp with the whole message.  Ifdef
42422         ENABLE_NLS, translate the message first using dgettext.
42424 2017-01-03  Jeff Law  <law@redhat.com>
42426         PR tree-optimizatin/78856
42427         * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42428         (mark_threaded_blocks): Remove code to truncate thread paths that
42429         cross multiple loop headers.  Instead invalidate the cached loop
42430         iteration information and handle case of a thread path walking
42431         into an irreducible region.
42433 2017-01-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
42435         PR target/78900
42436         * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42437         assertions.  Add support for doing the signbit if the IEEE 128-bit
42438         floating point value is in a GPR.
42439         * config/rs6000/rs6000.md (Fsignbit): Delete.
42440         (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42441         Update the length attribute if the value is in a GPR.
42442         (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42443         the sign or zero extension instruction, since the value is always 0/1.
42444         (signbit<mode>2_dm2): Delete using <Fsignbit>.
42446         PR target/78953
42447         * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42448         extracting SImode to a GPR register so that we can generate a
42449         store, limit the vector to be in a traditional Altivec register
42450         for the vextuwrx instruction.
42452 2017-01-03  Ian Lance Taylor  <iant@google.com>
42454         * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42456 2017-01-03  Martin Sebor  <msebor@redhat.com>
42458         PR tree-optimization/78696
42459         * gimple-ssa-sprintf.c (format_floating): Correct handling of
42460         precision.  Use MPFR for %f for greater fidelity.  Correct handling
42461         of %g.
42462         (pass_sprintf_length::compute_format_length): Set width and precision
42463         specified by asrerisk to void_node for vararg functions.
42464         (try_substitute_return_value): Adjust dump output.
42466 2017-01-03  David Edelsohn  <dje.gcc@gmail.com>
42468         * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42470 2017-01-03  Eric Botcazou  <ebotcazou@adacore.com>
42472         * doc/invoke.texi (SPARC options): Document -mlra as the default.
42473         * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42474         -mlra/-mno-lra was passed to the compiler.
42476 2017-01-03  James Cowgill  <James.Cowgill@imgtec.com>
42478         PR rtl-optimization/65618
42479         * emit-rtl.c (try_split): Move initialization of "before" and
42480         "after" to just before the call to emit_insn_after_setloc.
42482 2017-01-03  Gerald Pfeifer  <gerald@pfeifer.com>
42484         * doc/md.texi (Standard Names): Remove reference to Java frontend.
42486 2017-01-03  Pierre-Marie de Rodat  <derodat@adacore.com>
42488         * dwarf2out.c (gen_enumeration_type_die): When
42489         -gno-strict-dwarf, add a DW_AT_encoding attribute.
42491 2017-01-03  Jakub Jelinek  <jakub@redhat.com>
42493         PR tree-optimization/78965
42494         * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42495         Change first argument from const call_info & to call_info &.  For %n
42496         set info.nowrite to false.
42498         PR middle-end/78901
42499         * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42500         possibly throwing calls.
42502         * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42503         and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42504         and fns handling, rather than in a separate case SSA_NAME.
42506 2017-01-02  Jeff Law  <law@redhat.com>
42508         * config/darwin-driver.c (darwin_driver_init): Const-correctness
42509         fixes for first_period and second_period variables.
42511 2017-01-02  Uros Bizjak  <ubizjak@gmail.com>
42513         PR target/78967
42514         * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42515         (*insvqi_1): New insn pattern.
42516         (*insvqi_1_mem_rex64): Ditto.
42517         (*insvqi_2): Ditto.
42518         (*insvqi_3): Rename from *insvqi.
42520         (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42522 2017-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
42524         * doc/cfg.texi (Edges): Remove reference to Java.
42525         (Maintaining the CFG): Ditto.
42527 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42529         PR middle-end/77674
42530         * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42531         transparent aliases.
42533 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42535         PR middle-end/77484
42536         * predict.def (PRED_CALL): Update hitrate.
42537         (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42538         * predict.c (tree_estimate_probability_bb): Split CALL predictor
42539         into direct/indirect/polymorphic variants.
42541 2017-01-01  Jakub Jelinek  <jakub@redhat.com>
42543         Update copyright years.
42545         * gcc.c (process_command): Update copyright notice dates.
42546         * gcov-dump.c (print_version): Ditto.
42547         * gcov.c (print_version): Ditto.
42548         * gcov-tool.c (print_version): Ditto.
42549         * gengtype.c (create_file): Ditto.
42550         * doc/cpp.texi: Bump @copying's copyright year.
42551         * doc/cppinternals.texi: Ditto.
42552         * doc/gcc.texi: Ditto.
42553         * doc/gccint.texi: Ditto.
42554         * doc/gcov.texi: Ditto.
42555         * doc/install.texi: Ditto.
42556         * doc/invoke.texi: Ditto.
42558 Copyright (C) 2017 Free Software Foundation, Inc.
42560 Copying and distribution of this file, with or without modification,
42561 are permitted in any medium without royalty provided the copyright
42562 notice and this notice are preserved.