poly_int: vectorizable_call
[official-gcc.git] / gcc / ChangeLog
blob9aded12aadcdacbdf65a42f6f28d592094aa550f
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         * tree-vect-stmts.c (vectorizable_call): Treat the number of
6         vectors as polynomial.  Use build_index_vector for
7         IFN_GOMP_SIMD_LANE.
9 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
10             Alan Hayward  <alan.hayward@arm.com>
11             David Sherwood  <david.sherwood@arm.com>
13         * tree-vect-stmts.c (get_load_store_type): Treat the number of
14         units as polynomial.  Reject VMAT_ELEMENTWISE and VMAT_STRIDED_SLP
15         for variable-length vectors.
16         (vectorizable_mask_load_store): Treat the number of units as
17         polynomial, asserting that it is constant if the condition has
18         already been enforced.
19         (vectorizable_store, vectorizable_load): Likewise.
21 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
22             Alan Hayward  <alan.hayward@arm.com>
23             David Sherwood  <david.sherwood@arm.com>
25         * tree-vect-loop.c (vectorizable_live_operation): Treat the number
26         of units as polynomial.  Punt if we can't tell at compile time
27         which vector contains the final result.
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-loop.c (vectorizable_induction): Treat the number
34         of units as polynomial.  Punt on SLP inductions.  Use an integer
35         VEC_SERIES_EXPR for variable-length integer reductions.  Use a
36         cast of such a series for variable-length floating-point
37         reductions.
39 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
40             Alan Hayward  <alan.hayward@arm.com>
41             David Sherwood  <david.sherwood@arm.com>
43         * tree.h (build_index_vector): Declare.
44         * tree.c (build_index_vector): New function.
45         * tree-vect-loop.c (get_initial_defs_for_reduction): Treat the number
46         of units as polynomial, forcibly converting it to a constant if
47         vectorizable_reduction has already enforced the condition.
48         (vect_create_epilog_for_reduction): Likewise.  Use build_index_vector
49         to create a {1,2,3,...} vector.
50         (vectorizable_reduction): Treat the number of units as polynomial.
51         Choose vectype_in based on the largest scalar element size rather
52         than the smallest number of units.  Enforce the restrictions
53         relied on above.
55 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
56             Alan Hayward  <alan.hayward@arm.com>
57             David Sherwood  <david.sherwood@arm.com>
59         * tree-vect-data-refs.c (vector_alignment_reachable_p): Treat the
60         number of units as polynomial.
62 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
63             Alan Hayward  <alan.hayward@arm.com>
64             David Sherwood  <david.sherwood@arm.com>
66         * target.h (vector_sizes, auto_vector_sizes): New typedefs.
67         * target.def (autovectorize_vector_sizes): Return the vector sizes
68         by pointer, using vector_sizes rather than a bitmask.
69         * targhooks.h (default_autovectorize_vector_sizes): Update accordingly.
70         * targhooks.c (default_autovectorize_vector_sizes): Likewise.
71         * config/aarch64/aarch64.c (aarch64_autovectorize_vector_sizes):
72         Likewise.
73         * config/arc/arc.c (arc_autovectorize_vector_sizes): Likewise.
74         * config/arm/arm.c (arm_autovectorize_vector_sizes): Likewise.
75         * config/i386/i386.c (ix86_autovectorize_vector_sizes): Likewise.
76         * config/mips/mips.c (mips_autovectorize_vector_sizes): Likewise.
77         * omp-general.c (omp_max_vf): Likewise.
78         * omp-low.c (omp_clause_aligned_alignment): Likewise.
79         * optabs-query.c (can_vec_mask_load_store_p): Likewise.
80         * tree-vect-loop.c (vect_analyze_loop): Likewise.
81         * tree-vect-slp.c (vect_slp_bb): Likewise.
82         * doc/tm.texi: Regenerate.
83         * tree-vectorizer.h (current_vector_size): Change from an unsigned int
84         to a poly_uint64.
85         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Take
86         the vector size as a poly_uint64 rather than an unsigned int.
87         (current_vector_size): Change from an unsigned int to a poly_uint64.
88         (get_vectype_for_scalar_type): Update accordingly.
89         * tree.h (build_truth_vector_type): Take the size and number of
90         units as a poly_uint64 rather than an unsigned int.
91         (build_vector_type): Add a temporary overload that takes
92         the number of units as a poly_uint64 rather than an unsigned int.
93         * tree.c (make_vector_type): Likewise.
94         (build_truth_vector_type): Take the number of units as a poly_uint64
95         rather than an unsigned int.
97 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
98             Alan Hayward  <alan.hayward@arm.com>
99             David Sherwood  <david.sherwood@arm.com>
101         * target.def (get_mask_mode): Take the number of units and length
102         as poly_uint64s rather than unsigned ints.
103         * targhooks.h (default_get_mask_mode): Update accordingly.
104         * targhooks.c (default_get_mask_mode): Likewise.
105         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
106         * doc/tm.texi: Regenerate.
108 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
109             Alan Hayward  <alan.hayward@arm.com>
110             David Sherwood  <david.sherwood@arm.com>
112         * omp-general.h (omp_max_vf): Return a poly_uint64 instead of an int.
113         * omp-general.c (omp_max_vf): Likewise.
114         * omp-expand.c (omp_adjust_chunk_size): Update call to omp_max_vf.
115         (expand_omp_simd): Handle polynomial safelen.
116         * omp-low.c (omplow_simd_context): Add a default constructor.
117         (omplow_simd_context::max_vf): Change from int to poly_uint64.
118         (lower_rec_simd_input_clauses): Update accordingly.
119         (lower_rec_input_clauses): Likewise.
121 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
122             Alan Hayward  <alan.hayward@arm.com>
123             David Sherwood  <david.sherwood@arm.com>
125         * tree-vectorizer.h (vect_nunits_for_cost): New function.
126         * tree-vect-loop.c (vect_model_reduction_cost): Use it.
127         * tree-vect-slp.c (vect_analyze_slp_cost_1): Likewise.
128         (vect_analyze_slp_cost): Likewise.
129         * tree-vect-stmts.c (vect_model_store_cost): Likewise.
130         (vect_model_load_cost): Likewise.
132 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
133             Alan Hayward  <alan.hayward@arm.com>
134             David Sherwood  <david.sherwood@arm.com>
136         * tree-vect-slp.c (vect_record_max_nunits, vect_build_slp_tree_1)
137         (vect_build_slp_tree_2, vect_build_slp_tree): Change max_nunits
138         from an unsigned int * to a poly_uint64_pod *.
139         (calculate_unrolling_factor): New function.
140         (vect_analyze_slp_instance): Use it.  Track polynomial max_nunits.
142 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
143             Alan Hayward  <alan.hayward@arm.com>
144             David Sherwood  <david.sherwood@arm.com>
146         * tree-vectorizer.h (_slp_instance::unrolling_factor): Change
147         from an unsigned int to a poly_uint64.
148         (_loop_vec_info::slp_unrolling_factor): Likewise.
149         (_loop_vec_info::vectorization_factor): Change from an int
150         to a poly_uint64.
151         (MAX_VECTORIZATION_FACTOR): Bump from 64 to INT_MAX.
152         (vect_get_num_vectors): New function.
153         (vect_update_max_nunits, vect_vf_for_cost): Likewise.
154         (vect_get_num_copies): Use vect_get_num_vectors.
155         (vect_analyze_data_ref_dependences): Change max_vf from an int *
156         to an unsigned int *.
157         (vect_analyze_data_refs): Change min_vf from an int * to a
158         poly_uint64 *.
159         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
160         than an unsigned HOST_WIDE_INT.
161         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr)
162         (vect_analyze_data_ref_dependence): Change max_vf from an int *
163         to an unsigned int *.
164         (vect_analyze_data_ref_dependences): Likewise.
165         (vect_compute_data_ref_alignment): Handle polynomial vf.
166         (vect_enhance_data_refs_alignment): Likewise.
167         (vect_prune_runtime_alias_test_list): Likewise.
168         (vect_shift_permute_load_chain): Likewise.
169         (vect_supportable_dr_alignment): Likewise.
170         (dependence_distance_ge_vf): Take the vectorization factor as a
171         poly_uint64 rather than an unsigned HOST_WIDE_INT.
172         (vect_analyze_data_refs): Change min_vf from an int * to a
173         poly_uint64 *.
174         * tree-vect-loop-manip.c (vect_gen_scalar_loop_niters): Take
175         vfm1 as a poly_uint64 rather than an int.  Make the same change
176         for the returned bound_scalar.
177         (vect_gen_vector_loop_niters): Handle polynomial vf.
178         (vect_do_peeling): Likewise.  Update call to
179         vect_gen_scalar_loop_niters and handle polynomial bound_scalars.
180         (vect_gen_vector_loop_niters_mult_vf): Assert that the vf must
181         be constant.
182         * tree-vect-loop.c (vect_determine_vectorization_factor)
183         (vect_update_vf_for_slp, vect_analyze_loop_2): Handle polynomial vf.
184         (vect_get_known_peeling_cost): Likewise.
185         (vect_estimate_min_profitable_iters, vectorizable_reduction): Likewise.
186         (vect_worthwhile_without_simd_p, vectorizable_induction): Likewise.
187         (vect_transform_loop): Likewise.  Use the lowest possible VF when
188         updating the upper bounds of the loop.
189         (vect_min_worthwhile_factor): Make static.  Return an unsigned int
190         rather than an int.
191         * tree-vect-slp.c (vect_attempt_slp_rearrange_stmts): Cope with
192         polynomial unroll factors.
193         (vect_analyze_slp_cost_1, vect_analyze_slp_instance): Likewise.
194         (vect_make_slp_decision): Likewise.
195         (vect_supported_load_permutation_p): Likewise, and polynomial
196         vf too.
197         (vect_analyze_slp_cost): Handle polynomial vf.
198         (vect_slp_analyze_node_operations): Likewise.
199         (vect_slp_analyze_bb_1): Likewise.
200         (vect_transform_slp_perm_load): Take the vf as a poly_uint64 rather
201         than an unsigned HOST_WIDE_INT.
202         * tree-vect-stmts.c (vectorizable_simd_clone_call, vectorizable_store)
203         (vectorizable_load): Handle polynomial vf.
204         * tree-vectorizer.c (simduid_to_vf::vf): Change from an int to
205         a poly_uint64.
206         (adjust_simduid_builtins, shrink_simd_arrays): Update accordingly.
208 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
209             Alan Hayward  <alan.hayward@arm.com>
210             David Sherwood  <david.sherwood@arm.com>
212         * match.pd: Handle bit operations involving three constants
213         and try to fold one pair.
215 2018-01-03  Richard Sandiford  <richard.sandiford@linaro.org>
217         * tree-vect-loop-manip.c: Include gimple-fold.h.
218         (slpeel_make_loop_iterate_ntimes): Add step, final_iv and
219         niters_maybe_zero parameters.  Handle other cases besides a step of 1.
220         (vect_gen_vector_loop_niters): Add a step_vector_ptr parameter.
221         Add a path that uses a step of VF instead of 1, but disable it
222         for now.
223         (vect_do_peeling): Add step_vector, niters_vector_mult_vf_var
224         and niters_no_overflow parameters.  Update calls to
225         slpeel_make_loop_iterate_ntimes and vect_gen_vector_loop_niters.
226         Create a new SSA name if the latter choses to use a ste other
227         than zero, and return it via niters_vector_mult_vf_var.
228         * tree-vect-loop.c (vect_transform_loop): Update calls to
229         vect_do_peeling, vect_gen_vector_loop_niters and
230         slpeel_make_loop_iterate_ntimes.
231         * tree-vectorizer.h (slpeel_make_loop_iterate_ntimes, vect_do_peeling)
232         (vect_gen_vector_loop_niters): Update declarations after above changes.
234 2018-01-02  Michael Meissner  <meissner@linux.vnet.ibm.com>
236         * config/rs6000/rs6000.md (floor<mode>2): Add support for IEEE
237         128-bit round to integer instructions.
238         (ceil<mode>2): Likewise.
239         (btrunc<mode>2): Likewise.
240         (round<mode>2): Likewise.
242 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
244         * config/rs6000/rs6000-string.c (expand_block_move): Allow the use of
245         unaligned VSX load/store on P8/P9.
246         (expand_block_clear): Allow the use of unaligned VSX
247         load/store on P8/P9.
249 2018-01-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
251         * config/rs6000/rs6000-p8swap.c (swap_feeds_both_load_and_store):
252         New function.
253         (rs6000_analyze_swaps): Mark a web unoptimizable if it contains a
254         swap associated with both a load and a store.
256 2018-01-02  Andrew Waterman  <andrew@sifive.com>
258         * config/riscv/linux.h (ICACHE_FLUSH_FUNC): New.
259         * config/riscv/riscv.md (clear_cache): Use it.
261 2018-01-02  Artyom Skrobov  <tyomitch@gmail.com>
263         * web.c: Remove out-of-date comment.
265 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
267         * expr.c (fixup_args_size_notes): Check that any existing
268         REG_ARGS_SIZE notes are correct, and don't try to re-add them.
269         (emit_single_push_insn_1): Move stack_pointer_delta adjustment to...
270         (emit_single_push_insn): ...here.
272 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
274         * rtl.h (CONST_VECTOR_ELT): Redefine to const_vector_elt.
275         (const_vector_encoded_nelts): New function.
276         (CONST_VECTOR_NUNITS): Redefine to use GET_MODE_NUNITS.
277         (const_vector_int_elt, const_vector_elt): Declare.
278         * emit-rtl.c (const_vector_int_elt_1): New function.
279         (const_vector_elt): Likewise.
280         * simplify-rtx.c (simplify_immed_subreg): Avoid taking the address
281         of CONST_VECTOR_ELT.
283 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
285         * expr.c: Include rtx-vector-builder.h.
286         (const_vector_mask_from_tree): Use rtx_vector_builder and operate
287         directly on the tree encoding.
288         (const_vector_from_tree): Likewise.
289         * optabs.c: Include rtx-vector-builder.h.
290         (expand_vec_perm_var): Use rtx_vector_builder and create a repeating
291         sequence of "u" values.
292         * vec-perm-indices.c: Include rtx-vector-builder.h.
293         (vec_perm_indices_to_rtx): Use rtx_vector_builder and operate
294         directly on the vec_perm_indices encoding.
296 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
298         * doc/rtl.texi (const_vector): Describe new encoding scheme.
299         * Makefile.in (OBJS): Add rtx-vector-builder.o.
300         * rtx-vector-builder.h: New file.
301         * rtx-vector-builder.c: Likewise.
302         * rtl.h (rtx_def::u2): Add a const_vector field.
303         (CONST_VECTOR_NPATTERNS): New macro.
304         (CONST_VECTOR_NELTS_PER_PATTERN): Likewise.
305         (CONST_VECTOR_DUPLICATE_P): Likewise.
306         (CONST_VECTOR_STEPPED_P): Likewise.
307         (CONST_VECTOR_ENCODED_ELT): Likewise.
308         (const_vec_duplicate_p): Check for a duplicated vector encoding.
309         (unwrap_const_vec_duplicate): Likewise.
310         (const_vec_series_p): Check for a non-duplicated vector encoding.
311         Say that the function only returns true for integer vectors.
312         * emit-rtl.c: Include rtx-vector-builder.h.
313         (gen_const_vec_duplicate_1): Delete.
314         (gen_const_vector): Call gen_const_vec_duplicate instead of
315         gen_const_vec_duplicate_1.
316         (const_vec_series_p_1): Operate directly on the CONST_VECTOR encoding.
317         (gen_const_vec_duplicate): Use rtx_vector_builder.
318         (gen_const_vec_series): Likewise.
319         (gen_rtx_CONST_VECTOR): Likewise.
320         * config/powerpcspe/powerpcspe.c: Include rtx-vector-builder.h.
321         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
322         Build a new vector rather than modifying a CONST_VECTOR in-place.
323         (handle_special_swappables): Update call accordingly.
324         * config/rs6000/rs6000-p8swap.c: Include rtx-vector-builder.h.
325         (swap_const_vector_halves): Take an rtx pointer rather than rtx.
326         Build a new vector rather than modifying a CONST_VECTOR in-place.
327         (handle_special_swappables): Update call accordingly.
329 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
331         * simplify-rtx.c (simplify_const_binary_operation): Use
332         CONST_VECTOR_ELT instead of XVECEXP.
334 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
336         * tree-cfg.c (verify_gimple_assign_ternary): Allow the size of
337         the selector elements to be different from the data elements
338         if the selector is a VECTOR_CST.
339         * tree-vect-stmts.c (vect_gen_perm_mask_any): Use a vector of
340         ssizetype for the selector.
342 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
344         * optabs.c (shift_amt_for_vec_perm_mask): Try using series_p
345         before testing each element individually.
346         * tree-vect-generic.c (lower_vec_perm): Likewise.
348 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
350         * selftest.h (selftest::vec_perm_indices_c_tests): Declare.
351         * selftest-run-tests.c (selftest::run_tests): Call it.
352         * vector-builder.h (vector_builder::operator ==): New function.
353         (vector_builder::operator !=): Likewise.
354         * vec-perm-indices.h (vec_perm_indices::series_p): Declare.
355         (vec_perm_indices::all_from_input_p): New function.
356         * vec-perm-indices.c (vec_perm_indices::series_p): Likewise.
357         (test_vec_perm_12, selftest::vec_perm_indices_c_tests): Likewise.
358         * fold-const.c (fold_ternary_loc): Use tree_to_vec_perm_builder
359         instead of reading the VECTOR_CST directly.  Detect whether both
360         vector inputs are the same before constructing the vec_perm_indices,
361         and update the number of inputs argument accordingly.  Use the
362         utility functions added above.  Only construct sel2 if we need to.
364 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
366         * optabs.c (expand_vec_perm_var): Use an explicit encoding for
367         the broadcast of the low byte.
368         (expand_mult_highpart): Use an explicit encoding for the permutes.
369         * optabs-query.c (can_mult_highpart_p): Likewise.
370         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
371         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
372         (vectorizable_bswap): Likewise.
373         * tree-vect-data-refs.c (vect_grouped_store_supported): Use an
374         explicit encoding for the power-of-2 permutes.
375         (vect_permute_store_chain): Likewise.
376         (vect_grouped_load_supported): Likewise.
377         (vect_permute_load_chain): Likewise.
379 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
381         * vec-perm-indices.h (vec_perm_indices_to_tree): Declare.
382         * vec-perm-indices.c (vec_perm_indices_to_tree): New function.
383         * tree-ssa-forwprop.c (simplify_vector_constructor): Use it.
384         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
385         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
386         (vect_gen_perm_mask_any): Likewise.
388 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
390         * int-vector-builder.h: New file.
391         * vec-perm-indices.h: Include int-vector-builder.h.
392         (vec_perm_indices): Redefine as an int_vector_builder.
393         (auto_vec_perm_indices): Delete.
394         (vec_perm_builder): Redefine as a stand-alone class.
395         (vec_perm_indices::vec_perm_indices): New function.
396         (vec_perm_indices::clamp): Likewise.
397         * vec-perm-indices.c: Include fold-const.h and tree-vector-builder.h.
398         (vec_perm_indices::new_vector): New function.
399         (vec_perm_indices::new_expanded_vector): Update for new
400         vec_perm_indices class.
401         (vec_perm_indices::rotate_inputs): New function.
402         (vec_perm_indices::all_in_range_p): Operate directly on the
403         encoded form, without computing elided elements.
404         (tree_to_vec_perm_builder): Operate directly on the VECTOR_CST
405         encoding.  Update for new vec_perm_indices class.
406         * optabs.c (expand_vec_perm_const): Create a vec_perm_indices for
407         the given vec_perm_builder.
408         (expand_vec_perm_var): Update vec_perm_builder constructor.
409         (expand_mult_highpart): Use vec_perm_builder instead of
410         auto_vec_perm_indices.
411         * optabs-query.c (can_mult_highpart_p): Use vec_perm_builder and
412         vec_perm_indices instead of auto_vec_perm_indices.  Use a single
413         or double series encoding as appropriate.
414         * fold-const.c (fold_ternary_loc): Use vec_perm_builder and
415         vec_perm_indices instead of auto_vec_perm_indices.
416         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
417         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
418         (vect_permute_store_chain): Likewise.
419         (vect_grouped_load_supported): Likewise.
420         (vect_permute_load_chain): Likewise.
421         (vect_shift_permute_load_chain): Likewise.
422         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
423         (vect_transform_slp_perm_load): Likewise.
424         (vect_schedule_slp_instance): Likewise.
425         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
426         (vectorizable_mask_load_store): Likewise.
427         (vectorizable_bswap): Likewise.
428         (vectorizable_store): Likewise.
429         (vectorizable_load): Likewise.
430         * tree-vect-generic.c (lower_vec_perm): Use vec_perm_builder and
431         vec_perm_indices instead of auto_vec_perm_indices.  Use
432         tree_to_vec_perm_builder to read the vector from a tree.
433         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Take a
434         vec_perm_builder instead of a vec_perm_indices.
435         (have_whole_vector_shift): Use vec_perm_builder and
436         vec_perm_indices instead of auto_vec_perm_indices.  Leave the
437         truncation to calc_vec_perm_mask_for_shift.
438         (vect_create_epilog_for_reduction): Likewise.
439         * config/aarch64/aarch64.c (expand_vec_perm_d::perm): Change
440         from auto_vec_perm_indices to vec_perm_indices.
441         (aarch64_expand_vec_perm_const_1): Use rotate_inputs on d.perm
442         instead of changing individual elements.
443         (aarch64_vectorize_vec_perm_const): Use new_vector to install
444         the vector in d.perm.
445         * config/arm/arm.c (expand_vec_perm_d::perm): Change
446         from auto_vec_perm_indices to vec_perm_indices.
447         (arm_expand_vec_perm_const_1): Use rotate_inputs on d.perm
448         instead of changing individual elements.
449         (arm_vectorize_vec_perm_const): Use new_vector to install
450         the vector in d.perm.
451         * config/powerpcspe/powerpcspe.c (rs6000_expand_extract_even):
452         Update vec_perm_builder constructor.
453         (rs6000_expand_interleave): Likewise.
454         * config/rs6000/rs6000.c (rs6000_expand_extract_even): Likewise.
455         (rs6000_expand_interleave): Likewise.
457 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
459         * optabs-query.c (can_vec_perm_var_p): Check whether lowering
460         to qimode could truncate the indices.
461         * optabs.c (expand_vec_perm_var): Likewise.
463 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
465         * Makefile.in (OBJS): Add vec-perm-indices.o.
466         * vec-perm-indices.h: New file.
467         * vec-perm-indices.c: Likewise.
468         * target.h (vec_perm_indices): Replace with a forward class
469         declaration.
470         (auto_vec_perm_indices): Move to vec-perm-indices.h.
471         * optabs.h: Include vec-perm-indices.h.
472         (expand_vec_perm): Delete.
473         (selector_fits_mode_p, expand_vec_perm_var): Declare.
474         (expand_vec_perm_const): Declare.
475         * target.def (vec_perm_const_ok): Replace with...
476         (vec_perm_const): ...this new hook.
477         * doc/tm.texi.in (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Replace with...
478         (TARGET_VECTORIZE_VEC_PERM_CONST): ...this new hook.
479         * doc/tm.texi: Regenerate.
480         * optabs.def (vec_perm_const): Delete.
481         * doc/md.texi (vec_perm_const): Likewise.
482         (vec_perm): Refer to TARGET_VECTORIZE_VEC_PERM_CONST.
483         * expr.c (expand_expr_real_2): Use expand_vec_perm_const rather than
484         expand_vec_perm for constant permutation vectors.  Assert that
485         the mode of variable permutation vectors is the integer equivalent
486         of the mode that is being permuted.
487         * optabs-query.h (selector_fits_mode_p): Declare.
488         * optabs-query.c: Include vec-perm-indices.h.
489         (selector_fits_mode_p): New function.
490         (can_vec_perm_const_p): Check whether targetm.vectorize.vec_perm_const
491         is defined, instead of checking whether the vec_perm_const_optab
492         exists.  Use targetm.vectorize.vec_perm_const instead of
493         targetm.vectorize.vec_perm_const_ok.  Check whether the indices
494         fit in the vector mode before using a variable permute.
495         * optabs.c (shift_amt_for_vec_perm_mask): Take a mode and a
496         vec_perm_indices instead of an rtx.
497         (expand_vec_perm): Replace with...
498         (expand_vec_perm_const): ...this new function.  Take the selector
499         as a vec_perm_indices rather than an rtx.  Also take the mode of
500         the selector.  Update call to shift_amt_for_vec_perm_mask.
501         Use targetm.vectorize.vec_perm_const instead of vec_perm_const_optab.
502         Use vec_perm_indices::new_expanded_vector to expand the original
503         selector into bytes.  Check whether the indices fit in the vector
504         mode before using a variable permute.
505         (expand_vec_perm_var): Make global.
506         (expand_mult_highpart): Use expand_vec_perm_const.
507         * fold-const.c: Includes vec-perm-indices.h.
508         * tree-ssa-forwprop.c: Likewise.
509         * tree-vect-data-refs.c: Likewise.
510         * tree-vect-generic.c: Likewise.
511         * tree-vect-loop.c: Likewise.
512         * tree-vect-slp.c: Likewise.
513         * tree-vect-stmts.c: Likewise.
514         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm_const):
515         Delete.
516         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>): Delete.
517         * config/aarch64/aarch64.c (aarch64_expand_vec_perm_const)
518         (aarch64_vectorize_vec_perm_const_ok): Fuse into...
519         (aarch64_vectorize_vec_perm_const): ...this new function.
520         (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
521         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
522         * config/arm/arm-protos.h (arm_expand_vec_perm_const): Delete.
523         * config/arm/vec-common.md (vec_perm_const<mode>): Delete.
524         * config/arm/arm.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
525         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
526         (arm_expand_vec_perm_const, arm_vectorize_vec_perm_const_ok): Merge
527         into...
528         (arm_vectorize_vec_perm_const): ...this new function.  Explicitly
529         check for NEON modes.
530         * config/i386/i386-protos.h (ix86_expand_vec_perm_const): Delete.
531         * config/i386/sse.md (VEC_PERM_CONST, vec_perm_const<mode>): Delete.
532         * config/i386/i386.c (ix86_expand_vec_perm_const_1): Update comment.
533         (ix86_expand_vec_perm_const, ix86_vectorize_vec_perm_const_ok): Merge
534         into...
535         (ix86_vectorize_vec_perm_const): ...this new function.  Incorporate
536         the old VEC_PERM_CONST conditions.
537         * config/ia64/ia64-protos.h (ia64_expand_vec_perm_const): Delete.
538         * config/ia64/vect.md (vec_perm_const<mode>): Delete.
539         * config/ia64/ia64.c (ia64_expand_vec_perm_const)
540         (ia64_vectorize_vec_perm_const_ok): Merge into...
541         (ia64_vectorize_vec_perm_const): ...this new function.
542         * config/mips/loongson.md (vec_perm_const<mode>): Delete.
543         * config/mips/mips-msa.md (vec_perm_const<mode>): Delete.
544         * config/mips/mips-ps-3d.md (vec_perm_constv2sf): Delete.
545         * config/mips/mips-protos.h (mips_expand_vec_perm_const): Delete.
546         * config/mips/mips.c (mips_expand_vec_perm_const)
547         (mips_vectorize_vec_perm_const_ok): Merge into...
548         (mips_vectorize_vec_perm_const): ...this new function.
549         * config/powerpcspe/altivec.md (vec_perm_constv16qi): Delete.
550         * config/powerpcspe/paired.md (vec_perm_constv2sf): Delete.
551         * config/powerpcspe/spe.md (vec_perm_constv2si): Delete.
552         * config/powerpcspe/vsx.md (vec_perm_const<mode>): Delete.
553         * config/powerpcspe/powerpcspe-protos.h (altivec_expand_vec_perm_const)
554         (rs6000_expand_vec_perm_const): Delete.
555         * config/powerpcspe/powerpcspe.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK):
556         Delete.
557         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
558         (altivec_expand_vec_perm_const_le): Take each operand individually.
559         Operate on constant selectors rather than rtxes.
560         (altivec_expand_vec_perm_const): Likewise.  Update call to
561         altivec_expand_vec_perm_const_le.
562         (rs6000_expand_vec_perm_const): Delete.
563         (rs6000_vectorize_vec_perm_const_ok): Delete.
564         (rs6000_vectorize_vec_perm_const): New function.
565         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
566         an element count and rtx array.
567         (rs6000_expand_extract_even): Update call accordingly.
568         (rs6000_expand_interleave): Likewise.
569         * config/rs6000/altivec.md (vec_perm_constv16qi): Delete.
570         * config/rs6000/paired.md (vec_perm_constv2sf): Delete.
571         * config/rs6000/vsx.md (vec_perm_const<mode>): Delete.
572         * config/rs6000/rs6000-protos.h (altivec_expand_vec_perm_const)
573         (rs6000_expand_vec_perm_const): Delete.
574         * config/rs6000/rs6000.c (TARGET_VECTORIZE_VEC_PERM_CONST_OK): Delete.
575         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
576         (altivec_expand_vec_perm_const_le): Take each operand individually.
577         Operate on constant selectors rather than rtxes.
578         (altivec_expand_vec_perm_const): Likewise.  Update call to
579         altivec_expand_vec_perm_const_le.
580         (rs6000_expand_vec_perm_const): Delete.
581         (rs6000_vectorize_vec_perm_const_ok): Delete.
582         (rs6000_vectorize_vec_perm_const): New function.  Remove stray
583         reference to the SPE evmerge intructions.
584         (rs6000_do_expand_vec_perm): Take a vec_perm_builder instead of
585         an element count and rtx array.
586         (rs6000_expand_extract_even): Update call accordingly.
587         (rs6000_expand_interleave): Likewise.
588         * config/sparc/sparc.md (vec_perm_constv8qi): Delete in favor of...
589         * config/sparc/sparc.c (sparc_vectorize_vec_perm_const): ...this
590         new function.
591         (TARGET_VECTORIZE_VEC_PERM_CONST): Redefine.
593 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
595         * optabs.c (expand_vec_perm_1): Assert that SEL has an integer
596         vector mode and that that mode matches the mode of the data
597         being permuted.
598         (expand_vec_perm): Split handling of non-CONST_VECTOR selectors
599         out into expand_vec_perm_var.  Do all CONST_VECTOR handling here,
600         directly using expand_vec_perm_1 when forcing selectors into
601         registers.
602         (expand_vec_perm_var): New function, split out from expand_vec_perm.
604 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
606         * optabs-query.h (can_vec_perm_p): Delete.
607         (can_vec_perm_var_p, can_vec_perm_const_p): Declare.
608         * optabs-query.c (can_vec_perm_p): Split into...
609         (can_vec_perm_var_p, can_vec_perm_const_p): ...these two functions.
610         (can_mult_highpart_p): Use can_vec_perm_const_p to test whether a
611         particular selector is valid.
612         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
613         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
614         (vect_grouped_load_supported): Likewise.
615         (vect_shift_permute_load_chain): Likewise.
616         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
617         (vect_transform_slp_perm_load): Likewise.
618         * tree-vect-stmts.c (perm_mask_for_reverse): Likewise.
619         (vectorizable_bswap): Likewise.
620         (vect_gen_perm_mask_checked): Likewise.
621         * fold-const.c (fold_ternary_loc): Likewise.  Don't take
622         implementations of variable permutation vectors into account
623         when deciding which selector to use.
624         * tree-vect-loop.c (have_whole_vector_shift): Don't check whether
625         vec_perm_const_optab is supported; instead use can_vec_perm_const_p
626         with a false third argument.
627         * tree-vect-generic.c (lower_vec_perm): Use can_vec_perm_const_p
628         to test whether the constant selector is valid and can_vec_perm_var_p
629         to test whether a variable selector is valid.
631 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
633         * optabs-query.h (can_vec_perm_p): Take a const vec_perm_indices *.
634         * optabs-query.c (can_vec_perm_p): Likewise.
635         * fold-const.c (fold_vec_perm): Take a const vec_perm_indices &
636         instead of vec_perm_indices.
637         * tree-vectorizer.h (vect_gen_perm_mask_any): Likewise,
638         (vect_gen_perm_mask_checked): Likewise,
639         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise,
640         (vect_gen_perm_mask_checked): Likewise,
642 2018-01-02  Richard Sandiford  <richard.sandiford@linaro.org>
644         * optabs-query.h (qimode_for_vec_perm): Declare.
645         * optabs-query.c (can_vec_perm_p): Split out qimode search to...
646         (qimode_for_vec_perm): ...this new function.
647         * optabs.c (expand_vec_perm): Use qimode_for_vec_perm.
649 2018-01-02  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
651         * rtlanal.c (canonicalize_condition): Return 0 if final rtx
652         does not have a conditional at the top.
654 2018-01-02  Richard Biener  <rguenther@suse.de>
656         * ipa-inline.c (big_speedup_p): Fix expression.
658 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
660         PR target/81616
661         * config/i386/x86-tune-costs.h: Increase cost of integer load costs
662         for generic 4->6.
664 2018-01-02  Jan Hubicka  <hubicka@ucw.cz>
666         PR target/81616
667         Generic tuning.
668         * x86-tune-costs.h (generic_cost): Reduce cost of FDIV 20->17,
669         cost of sqrt 20->14, DIVSS 18->13, DIVSD 32->17, SQRtSS 30->14
670         and SQRTsD 58->18, cond_not_taken_branch_cost. 2->1. Increase
671         cond_taken_branch_cost 3->4.
673 2018-01-01  Jakub Jelinek  <jakub@redhat.com>
675         PR tree-optimization/83581
676         * tree-loop-distribution.c (pass_loop_distribution::execute): Return
677         TODO_cleanup_cfg if any changes have been made.
679         PR middle-end/83608
680         * expr.c (store_expr_with_bounds): Use simplify_gen_subreg instead of
681         convert_modes if target mode has the right side, but different mode
682         class.
684         PR middle-end/83609
685         * expr.c (expand_assignment): Fix up a typo in simplify_gen_subreg
686         last argument when extracting from CONCAT.  If either from_real or
687         from_imag is NULL, use expansion through memory.  If result is not
688         a CONCAT and simplify_gen_subreg fails, try to simplify_gen_subreg
689         the parts directly to inner mode, if even that fails, use expansion
690         through memory.
692         PR middle-end/83623
693         * expmed.c (expand_shift_1): For 2-byte rotates by BITS_PER_UNIT,
694         check for bswap in mode rather than HImode and use that in expand_unop
695         too.
697 2017-12-31  Jakub Jelinek  <jakub@redhat.com>
699         PR target/83536
700         * config/i386/i386.c (ix86_attribute_table): Remove excess
701         initializer for "shared" attribute.
703 2017-12-30  Tom de Vries  <tom@codesourcery.com>
705         PR libgomp/83046
706         * omp-expand.c (expand_omp_target): If in_lto_p, mark offload_funcs with
707         DECL_PRESERVE_P.
708         * lto-streamer-out.c (prune_offload_funcs): New function.  Remove
709         offload_funcs entries that no longer have a corresponding cgraph_node.
710         Mark the remaining ones as DECL_PRESERVE_P.
711         (output_lto): Call prune_offload_funcs.
713 2017-12-30  Jakub Jelinek  <jakub@redhat.com>
715         * config/i386/sse.md (vgf2p8affineinvqb_<mode><mask_name>,
716         vgf2p8affineqb_<mode><mask_name>, vgf2p8mulb_<mode><mask_name>,
717         vpshrd_<mode><mask_name>, vpshld_<mode><mask_name>,
718         vpshrdv_<mode>, vpshrdv_<mode>_mask, vpshrdv_<mode>_maskz,
719         vpshrdv_<mode>_maskz_1, vpshldv_<mode>, vpshldv_<mode>_mask,
720         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1, vpdpbusd_<mode>,
721         vpdpbusd_<mode>_mask, vpdpbusd_<mode>_maskz, vpdpbusd_<mode>_maskz_1,
722         vpdpbusds_<mode>, vpdpbusds_<mode>_mask, vpdpbusds_<mode>_maskz,
723         vpdpbusds_<mode>_maskz_1, vpdpwssd_<mode>, vpdpwssd_<mode>_mask,
724         vpdpwssd_<mode>_maskz, vpdpwssd_<mode>_maskz_1, vpdpwssds_<mode>,
725         vpdpwssds_<mode>_mask, vpdpwssds_<mode>_maskz,
726         vpdpwssds_<mode>_maskz_1, vaesdec_<mode>, vaesdeclast_<mode>,
727         vaesenc_<mode>, vpclmulqdq_<mode>,
728         avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): Formatting fixes.
730 2017-12-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
732         * builtins.def: (_Float<N> and _Float<N>X BUILT_IN_CEIL): Add
733         _Float<N> and _Float<N>X variants for rounding built-in
734         functions.
735         (_Float<N> and _Float<N>X BUILT_IN_FLOOR): Likewise.
736         (_Float<N> and _Float<N>X BUILT_IN_NEARBYINT): Likewise.
737         (_Float<N> and _Float<N>X BUILT_IN_RINT): Likewise.
738         (_Float<N> and _Float<N>X BUILT_IN_ROUND): Likewise.
739         (_Float<N> and _Float<N>X BUILT_IN_TRUNC): Likewise.
740         * builtins.c (mathfn_built_in_2): Likewise.
741         * internal-fn.def (CEIL): Likewise.
742         (FLOOR): Likewise.
743         (NEARBYINT): Likewise.
744         (RINT): Likewise.
745         (ROUND): Likewise.
746         (TRUNC): Likewise.
747         * convert.c (convert_to_integer_1): Likewise.
748         * fold-const.c (tree_call_nonnegative_warnv_p): Likewise.
749         (integer_valued_real_call_p): Likewise.
750         * fold-const-call.c (fold_const_call_ss): Likewise.
751         * gencfn-macros.c (print_case_cfn): Change CFN and operator
752         printers to take a const char * suffix instead of a bool.
753         (print_define_operator_list): Likewise.
754         (fltall_suffixes): New list of suffixes, that include the
755         traditional suffixes as well as all of the _Float<N> and
756         _Float<N>X suffixes.
757         (main): For _Float<N> and _Float<N>X functions, emit both
758         <name>_FN and <name>_ALL variants.  The <macro>_FN variant only
759         has the _Float<N> and _Float<N>X case names or operators.  The
760         <name>_ALL variant has both the traditional and the
761         _Float<N>/_Float<N>X case names or operators.
762         * match.pd (COPYSIGN optimizations): Provide optimizations for
763         _Float<N> and _Float<N>X types where possible.
764         (MIN/MAX optimizations): Likewise.
765         (sqrt optimizations): Likewise.
766         (rounding optimizations): Likewise.
768 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
770         * config/powerpcspe/powerpcspe.md (ieee_128bit_negative_zero): Use
771         gen_int_mode rather than GEN_INT.
772         * config/rs6000/rs6000.md (ieee_128bit_negative_zero): Likewise.
774 2017-12-28  Richard Sandiford  <richard.sandiford@linaro.org>
776         * emit-rtl.c (gen_const_vec_series): Use valid_for_const_vector_p
777         instead of CONSTANT_P.
778         (gen_vec_series): Likewise.
779         * simplify-rtx.c (simplify_binary_operation_1): Likewise.
781 2017-12-28  Andreas Schwab  <schwab@linux-m68k.org>
783         * config/m68k/m68k.md (ashrdi3_const1, lshrdi3_const1): Add
784         CC_STATUS_INIT.
786 2017-12-27  Kugan Vivekanandarajah  <kuganv@linaro.org>
788         * config/aarch64/aarch64-simd.md (aarch64_ld1x2<VQ:mode>): New.
789         (aarch64_ld1x2<VDC:mode>): Likewise.
790         (aarch64_simd_ld1<mode>_x2): Likewise.
791         (aarch64_simd_ld1<mode>_x2): Likewise.
792         * config/aarch64/arm_neon.h (vld1_u8_x2): New.
793         (vld1_s8_x2): Likewise.
794         (vld1_u16_x2): Likewise.
795         (vld1_s16_x2): Likewise.
796         (vld1_u32_x2): Likewise.
797         (vld1_s32_x2): Likewise.
798         (vld1_u64_x2): Likewise.
799         (vld1_s64_x2): Likewise.
800         (vld1_f16_x2): Likewise.
801         (vld1_f32_x2): Likewise.
802         (vld1_f64_x2): Likewise.
803         (vld1_p8_x2): Likewise.
804         (vld1_p16_x2): Likewise.
805         (vld1_p64_x2): Likewise.
806         (vld1q_u8_x2): Likewise.
807         (vld1q_s8_x2): Likewise.
808         (vld1q_u16_x2): Likewise.
809         (vld1q_s16_x2): Likewise.
810         (vld1q_u32_x2): Likewise.
811         (vld1q_s32_x2): Likewise.
812         (vld1q_u64_x2): Likewise.
813         (vld1q_s64_x2): Likewise.
814         (vld1q_f16_x2): Likewise.
815         (vld1q_f32_x2): Likewise.
816         (vld1q_f64_x2): Likewise.
817         (vld1q_p8_x2): Likewise.
818         (vld1q_p16_x2): Likewise.
819         (vld1q_p64_x2): Likewise.
821 2017-12-27  Martin Liska  <mliska@suse.cz>
823         PR tree-optimization/83552
824         * tree-ssa-strlen.c (fold_strstr_to_strncmp): Assign result
825         of get_string_lenth to a SSA_NAME if not a GIMPLE value.
827 2017-12-27  Tom de Vries  <tom@codesourcery.com>
829         * config/nvptx/nvptx.c (nvptx_option_override): Disable
830         -gstatement-frontiers.
832 2017-12-26  Alexander Monakov  <amonakov@ispras.ru>
834         PR rtl-optimization/83513
835         * sel-sched.c (sel_rank_for_schedule): Order by non-zero usefulness
836         before priority comparison.
838 2017-12-25  Jakub Jelinek  <jakub@redhat.com>
840         PR target/83488
841         * config/i386/i386.opt (-mavx512vpopcntdq, -mmavx512bitalg): Move from
842         ix86_isa_flags2 to ix86_isa_flags.
843         * config/i386/i386-c.c (ix86_target_macros_internal): Test
844         OPTION_MASK_ISA_AVX512BITALG and OPTION_MASK_ISA_AVX512VPOPCNTDQ in
845         isa_flags rather than isa_flags2.
846         * config/i386/i386.c (ix86_target_string): Move -mavx512vpopcntdq
847         and -mavx512bitalg from isa2_opts to isa_opts.
848         (ix86_option_override_internal): Test OPTION_MASK_ISA_AVX512VPOPCNTDQ
849         in x_ix86_isa_flags_explicit rather than x_ix86_isa_flags2_explicit
850         and set it in x_ix86_isa_flags rather than x_ix86_isa_flags2.
851         Formatting fixes.
852         (def_builtin): Treat OPTION_MASK_ISA_AVX512BW or
853         OPTION_MASK_ISA_AVX512F ored with another option similarly to
854         OPTION_MASK_ISA_AVX512VL.  Even for OPTION_MASK_ISA_AVX512VL don't
855         clear it if mask is just OPTION_MASK_ISA_AVX512VL itself.
856         (ix86_expand_builtin): Don't handle OPTION_MASK_ISA_GFNI and
857         OPTION_MASK_ISA_VPCLMULQDQ specially, instead handle
858         OPTION_MASK_ISA_AVX512BW and OPTION_MASK_ISA_AVX512F that way.
859         * config/i386/i386-builtin.def: Move AVX512VPOPCNTDQ and AVX512BITALG
860         builtins from bdesc_args2 to bdesc_args section.
861         (__builtin_ia32_compressstoreuqi512_mask,
862         __builtin_ia32_compressstoreuhi512_mask,
863         __builtin_ia32_compressstoreuqi256_mask,
864         __builtin_ia32_expandloadqi512_mask,
865         __builtin_ia32_expandloadqi512_maskz,
866         __builtin_ia32_expandloadhi512_mask,
867         __builtin_ia32_expandloadhi512_maskz,
868         __builtin_ia32_compressqi512_mask, __builtin_ia32_compresshi512_mask,
869         __builtin_ia32_compressqi256_mask, __builtin_ia32_expandqi512_mask,
870         __builtin_ia32_expandqi512_maskz, __builtin_ia32_expandhi512_mask,
871         __builtin_ia32_expandhi512_maskz, __builtin_ia32_expandqi256_mask,
872         __builtin_ia32_expandqi256_maskz, __builtin_ia32_vpshrd_v32hi_mask,
873         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshrdv_v32hi_mask,
874         __builtin_ia32_vpshrdv_v32hi_maskz, __builtin_ia32_vpshldv_v32hi_mask,
875         __builtin_ia32_vpshldv_v32hi_maskz,
876         __builtin_ia32_vpopcountb_v64qi_mask,
877         __builtin_ia32_vpopcountw_v32hi_mask,
878         __builtin_ia32_vpshufbitqmb512_mask,
879         __builtin_ia32_vpshufbitqmb256_mask): Add
880         " | OPTION_MASK_ISA_AVX512BW".
881         (__builtin_ia32_expandloadqi256_mask,
882         __builtin_ia32_expandloadqi256_maskz,
883         __builtin_ia32_vpopcountb_v32qi_mask): Add
884         " | OPTION_MASK_ISA_AVX512VL | OPTION_MASK_ISA_AVX512BW".
885         (__builtin_ia32_expandloadhi256_mask,
886         __builtin_ia32_expandloadhi256_maskz,
887         __builtin_ia32_expandloadqi128_mask,
888         __builtin_ia32_expandloadqi128_maskz,
889         __builtin_ia32_expandloadhi128_mask,
890         __builtin_ia32_expandloadhi128_maskz,
891         __builtin_ia32_vpshrd_v16hi, __builtin_ia32_vpshrd_v16hi_mask,
892         __builtin_ia32_vpshrd_v8hi, __builtin_ia32_vpshrd_v8hi_mask,
893         __builtin_ia32_vpshrd_v8si, __builtin_ia32_vpshrd_v8si_mask,
894         __builtin_ia32_vpshrd_v4si, __builtin_ia32_vpshrd_v4si_mask,
895         __builtin_ia32_vpshrd_v4di, __builtin_ia32_vpshrd_v4di_mask,
896         __builtin_ia32_vpshrd_v2di, __builtin_ia32_vpshrd_v2di_mask,
897         __builtin_ia32_vpshld_v16hi, __builtin_ia32_vpshld_v16hi_mask,
898         __builtin_ia32_vpshld_v8hi, __builtin_ia32_vpshld_v8hi_mask,
899         __builtin_ia32_vpshld_v8si, __builtin_ia32_vpshld_v8si_mask,
900         __builtin_ia32_vpshld_v4si, __builtin_ia32_vpshld_v4si_mask,
901         __builtin_ia32_vpshld_v4di, __builtin_ia32_vpshld_v4di_mask,
902         __builtin_ia32_vpshld_v2di, __builtin_ia32_vpshld_v2di_mask,
903         __builtin_ia32_vpshrdv_v16hi, __builtin_ia32_vpshrdv_v16hi_mask,
904         __builtin_ia32_vpshrdv_v16hi_maskz, __builtin_ia32_vpshrdv_v8hi,
905         __builtin_ia32_vpshrdv_v8hi_mask, __builtin_ia32_vpshrdv_v8hi_maskz,
906         __builtin_ia32_vpshrdv_v8si, __builtin_ia32_vpshrdv_v8si_mask,
907         __builtin_ia32_vpshrdv_v8si_maskz, __builtin_ia32_vpshrdv_v4si,
908         __builtin_ia32_vpshrdv_v4si_mask, __builtin_ia32_vpshrdv_v4si_maskz,
909         __builtin_ia32_vpshrdv_v4di, __builtin_ia32_vpshrdv_v4di_mask,
910         __builtin_ia32_vpshrdv_v4di_maskz, __builtin_ia32_vpshrdv_v2di,
911         __builtin_ia32_vpshrdv_v2di_mask, __builtin_ia32_vpshrdv_v2di_maskz,
912         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
913         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
914         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
915         __builtin_ia32_vpshldv_v8si, __builtin_ia32_vpshldv_v8si_mask,
916         __builtin_ia32_vpshldv_v8si_maskz, __builtin_ia32_vpshldv_v4si,
917         __builtin_ia32_vpshldv_v4si_mask, __builtin_ia32_vpshldv_v4si_maskz,
918         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
919         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
920         __builtin_ia32_vpshldv_v2di_mask, __builtin_ia32_vpshldv_v2di_maskz,
921         __builtin_ia32_vpopcountb_v32qi, __builtin_ia32_vpopcountb_v16qi,
922         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v16hi,
923         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
924         __builtin_ia32_vpopcountw_v8hi_mask): Add
925         " | OPTION_MASK_ISA_AVX512VL".
926         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
927         _mm512_shrdi_epi32, _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32,
928         _mm512_shrdi_epi64, _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64,
929         _mm512_shldi_epi16, _mm512_shldi_epi32, _mm512_mask_shldi_epi32,
930         _mm512_maskz_shldi_epi32, _mm512_shldi_epi64, _mm512_mask_shldi_epi64,
931         _mm512_maskz_shldi_epi64, _mm512_shrdv_epi16, _mm512_shrdv_epi32,
932         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
933         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64, _mm512_shldv_epi16,
934         _mm512_shldv_epi32, _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32,
935         _mm512_shldv_epi64, _mm512_mask_shldv_epi64,
936         _mm512_maskz_shldv_epi64): Don't require avx512bw for these intrinsics.
937         * config/i386/avx512bitalgintrin.h (_mm_bitshuffle_epi64_mask,
938         _mm_mask_bitshuffle_epi64_mask): Likewise.
939         * common/config/i386/i386-common.c
940         (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
941         OPTION_MASK_ISA_AVX512BITALG_SET): Or in OPTION_MASK_ISA_AVX512F_SET.
942         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
943         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET and
944         OPTION_MASK_ISA_AVX512BITALG_UNSET.
945         (OPTION_MASK_ISA2_AVX512F_UNSET,
946         OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET): Define.
947         (ix86_handle_option): For -mno-general-regs-only, clear from
948         ix86_isa_flags2 OPTION_MASK_ISA2_GENERAL_REGS_ONLY_UNSET rather than
949         just OPTION_MASK_ISA_MPX.  For -mno-sse{,2,3,4,4.1,4.2,avx,avx2} and
950         -mno-ssse3 clear OPTION_MASK_ISA2_AVX512F_UNSET bits from
951         ix86_isa_flags2.  For -mno-avx512f likewise, instead of masking
952         individually listed ISAs.  For -m{,no-}avx512{vpopcntdq,bitalg} adjust
953         for moving from ix86_isa_flags2 to ix86_isa_flags.
955 2017-12-23  Jakub Jelinek  <jakub@redhat.com>
957         PR c++/83553
958         * fold-const.c (struct contains_label_data): New type.
959         (contains_label_1): Return non-NULL even for CASE_LABEL_EXPR, unless
960         inside of a SWITCH_BODY seen during the walk.
961         (contains_label_p): Use walk_tree instead of
962         walk_tree_without_duplicates, prepare data for contains_label_1 and
963         provide own pset.
965 2017-12-22  Jakub Jelinek  <jakub@redhat.com>
967         PR debug/83547
968         * tree-iterator.c (alloc_stmt_list): Start with cleared
969         TREE_SIDE_EFFECTS regardless whether a new STATEMENT_LIST is allocated
970         or old one reused.
972         PR target/83488
973         * config/i386/avx512vnniintrin.h: Don't check for __AVX512F__ nor
974         enable avx512f explicitly in #pragma GCC target.
975         * config/i386/i386-builtin.def (__builtin_ia32_vpdpbusd_v8si,
976         __builtin_ia32_vpdpbusd_v8si_mask, __builtin_ia32_vpdpbusd_v8si_maskz,
977         __builtin_ia32_vpdpbusd_v4si, __builtin_ia32_vpdpbusd_v4si_mask,
978         __builtin_ia32_vpdpbusd_v4si_maskz, __builtin_ia32_vpdpbusds_v8si,
979         __builtin_ia32_vpdpbusds_v8si_mask,
980         __builtin_ia32_vpdpbusds_v8si_maskz, __builtin_ia32_vpdpbusds_v4si,
981         __builtin_ia32_vpdpbusds_v4si_mask,
982         __builtin_ia32_vpdpbusds_v4si_maskz, __builtin_ia32_vpdpwssd_v8si,
983         __builtin_ia32_vpdpwssd_v8si_mask, __builtin_ia32_vpdpwssd_v8si_maskz,
984         __builtin_ia32_vpdpwssd_v4si, __builtin_ia32_vpdpwssd_v4si_mask,
985         __builtin_ia32_vpdpwssd_v4si_maskz, __builtin_ia32_vpdpwssds_v8si,
986         __builtin_ia32_vpdpwssds_v8si_mask,
987         __builtin_ia32_vpdpwssds_v8si_maskz, __builtin_ia32_vpdpwssds_v4si,
988         __builtin_ia32_vpdpwssds_v4si_mask,
989         __builtin_ia32_vpdpwssds_v4si_maskz): Use
990         OPTION_MASK_ISA_AVX512VNNI | OPTION_MASK_ISA_AVX512VL instead of
991         just OPTION_MASK_ISA_AVX512VNNI.
993 2017-12-22  Martin Jambor  <mjambor@suse.cz>
995         PR lto/82027
996         * lto-cgraph.c (output_cgraph_opt_summary_p): Also check former
997         clones.
999 2017-12-22  Julia Koval  <julia.koval@intel.com>
1000             Sebastian Peryt  <sebastian.peryt@intel.com>
1002         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512BITALG_SET,
1003         OPTION_MASK_ISA_AVX512BITALG_UNSET): New.
1004         (ix86_handle_option): Handle -mavx512bitalg, fix 4VNNIW formatting.
1005         * config.gcc: Add avx512vpopcntdqvlintrin.h and avx512bitalgintrin.h.
1006         * config/i386/avx512bitalgintrin.h (_mm512_popcnt_epi8, _mm512_popcnt_epi16,
1007         _mm512_mask_popcnt_epi8, _mm512_maskz_popcnt_epi8, _mm512_mask_popcnt_epi16,
1008         _mm512_maskz_popcnt_epi16, _mm512_bitshuffle_epi64_mask, _mm256_popcnt_epi8,
1009         _mm512_mask_bitshuffle_epi64_mask, _mm256_mask_popcnt_epi8, _mm_popcnt_epi8,
1010         _mm256_maskz_popcnt_epi8, _mm_bitshuffle_epi64_mask, _mm256_popcnt_epi16,
1011         _mm_mask_bitshuffle_epi64_mask, _mm256_bitshuffle_epi64_mask,
1012         _mm256_mask_bitshuffle_epi64_mask, _mm_popcnt_epi16, _mm_maskz_popcnt_epi8,
1013         _mm256_mask_popcnt_epi16, _mm256_maskz_popcnt_epi16, _mm_mask_popcnt_epi8,
1014         _mm_mask_popcnt_epi16, _mm_maskz_popcnt_epi16): New intrinsics.
1015         * config/i386/avx512vpopcntdqvlintrin.h (_mm_popcnt_epi32, _mm_popcnt_epi64,
1016         _mm_mask_popcnt_epi32, _mm_maskz_popcnt_epi32, _mm256_popcnt_epi32,
1017         _mm256_mask_popcnt_epi32, _mm256_maskz_popcnt_epi32, _mm_mask_popcnt_epi64,
1018         _mm_maskz_popcnt_epi64, _mm256_popcnt_epi64, _mm256_mask_popcnt_epi64,
1019         _mm256_maskz_popcnt_epi64): New intrinsics.
1020         * config/i386/cpuid.h (bit_AVX512BITALG): New bit.
1021         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mavx512bitalg.
1022         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI, V64QI_FTYPE_V64QI,
1023         V4DI_FTYPE_V4DI, UHI_FTYPE_V2DI_V2DI_UHI, USI_FTYPE_V4DI_V4DI_USI,
1024         V4SI_FTYPE_V4SI_V4SI_UHI, V8SI_FTYPE_V8SI_V8SI_UHI): New types.
1025         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountq_v4di,
1026         __builtin_ia32_vpopcountq_v4di_mask, __builtin_ia32_vpopcountq_v2di,
1027         __builtin_ia32_vpopcountq_v2di_mask, __builtin_ia32_vpopcountd_v4si,
1028         __builtin_ia32_vpopcountd_v4si_mask, __builtin_ia32_vpopcountd_v8si,
1029         __builtin_ia32_vpopcountd_v8si_mask, __builtin_ia32_vpopcountb_v64qi,
1030         __builtin_ia32_vpopcountb_v64qi_mask, __builtin_ia32_vpopcountb_v32qi,
1031         __builtin_ia32_vpopcountb_v32qi_mask, __builtin_ia32_vpopcountb_v16qi,
1032         __builtin_ia32_vpopcountb_v16qi_mask, __builtin_ia32_vpopcountw_v32hi,
1033         __builtin_ia32_vpopcountw_v32hi_mask, __builtin_ia32_vpopcountw_v16hi,
1034         __builtin_ia32_vpopcountw_v16hi_mask, __builtin_ia32_vpopcountw_v8hi,
1035         __builtin_ia32_vpopcountw_v8hi_mask, __builtin_ia32_vpshufbitqmb128_mask,
1036         __builtin_ia32_vpshufbitqmb256_mask,
1037         __builtin_ia32_vpshufbitqmb512_mask): New builtins.
1038         * config/i386/i386-c.c (__AVX512BITALG__): New.
1039         * config/i386/i386.c (isa2_opts): Add -mavx512bitalg.
1040         (ix86_valid_target_attribute_inner_p): Ditto.
1041         (ix86_expand_args_builtin): Handle new types.
1042         * config/i386/i386.h (TARGET_AVX512BITALG, TARGET_AVX512BITALG_P): New.
1043         * config/i386/i386.opt: Add -mavx512bitalg.
1044         * config/i386/immintrin.h: Add avx512vpopcntdqvlintrin.h and
1045         avx512bitalgintrin.h.
1046         * config/i386/sse.md (VI48_AVX512VLBW): New iterator.
1047         (vpopcount<mode><mask_name>): Add more types.
1048         (avx512vl_vpshufbitqmb<mode><mask_scalar_merge_name>): New.
1049         * doc/invoke.texi: Add -mavx512bitalg and -mavx512vpopcntdq.
1051 2017-12-22  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
1053         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET):
1054         Or in OPTION_MASK_ISA_AVX512F_SET.
1055         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1056         OPTION_MASK_ISA_AVX512VNNI_UNSET.
1057         (ix86_handle_option): Adjust for
1058         OPTION_MASK_ISA_AVX512VNNI_*SET being in ix86_isa_flags.
1059         * config/i386/i386-builtin.def: Move VNNI builtins from ARGS2
1060         section to ARGS.
1061         * config/i386/i386-c.c: Check for OPTION_MASK_ISA_AVX512VNNI in
1062         isa_flag instead of isa_flag2.
1063         * config/i386/i386.c (ix86_target_string): Move -mavx512vnni from
1064         isa_opts2 to isa_opts.
1065         * config/i386/i386.opt (mavx512vnni): Move from ix86_isa_flags2
1066         to ix86_isa_flags.
1068 2017-12-22  Mike Stump  <mikestump@comcast.net>
1069             Eric Botcazou  <ebotcazou@adacore.com>
1071         * doc/extend.texi (Loop-Specific Pragmas): Document pragma GCC unroll.
1073 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1075         PR middle-end/83487
1076         * config/i386/i386.c (ix86_function_arg_boundary): Return
1077         PARM_BOUNDARY for TYPE_EMPTY_P types.
1079         PR c/83448
1080         * gimple-ssa-sprintf.c (maybe_warn): Don't call set_caret_index
1081         if navail is >= dir.len.
1083 2017-12-21  Steve Ellcey  <sellcey@cavium.com>
1085         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Fix
1086         triplet for ilp32.
1088 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1090         PR rtl-optimization/80747
1091         PR rtl-optimization/83512
1092         * cfgrtl.c (force_nonfallthru_and_redirect): When splitting
1093         succ edge from ENTRY, copy partition from e->dest to the newly
1094         created bb.
1095         * bb-reorder.c (reorder_basic_blocks_simple): If last_tail is
1096         ENTRY, use BB_PARTITION of its successor block as current_partition.
1097         Don't copy partition when splitting succ edge from ENTRY.
1099         PR tree-optimization/83523
1100         * tree-ssa-math-opts.c (is_widening_mult_p): Return false if
1101         for INTEGER_TYPE TYPE_OVERFLOW_TRAPS.
1102         (convert_mult_to_fma): Likewise.
1104         PR tree-optimization/83521
1105         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Use
1106         gimple_build_assign without code on result of
1107         fold_build1 (VIEW_CONVERT_EXPR, ...), as it might not create
1108         a VIEW_CONVERT_EXPR.
1110 2017-12-21  Andrew Pinski  <apinski@cavium.com>
1111             Steve Ellcey  <sellcey@cavium.com>
1113         * config/aarch64/t-aarch64-linux (MULTILIB_OSDIRNAMES): Handle
1114         multi-arch for ilp32.
1116 2017-12-21  Uros Bizjak  <ubizjak@gmail.com>
1118         PR target/83467
1119         * config/i386/i386.md (*ashl<mode>3_mask): Add operand
1120         constraints to operand 2.
1121         (*ashl<mode>3_mask_1): Ditto.
1122         (*<shift_insn><mode>3_mask): Ditto.
1123         (*<shift_insn><mode>3_mask_1): Ditto.
1124         (*<rotate_insn><mode>3_mask): Ditto.
1125         (*<rotate_insn><mode>3_mask_1): Ditto.
1127 2017-12-21  Alexandre Oliva  <aoliva@redhat.com>
1129         * reorg.c (make_return_insns): Reemit each insn with its own location.
1131 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1133         * config/aarch64/aarch64.c (aarch64_expand_vector_init): Modify code
1134         generation for cases where splatting a value is not useful.
1135         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
1136         across a vec_duplicate and a paradoxical subreg forming a vector
1137         mode to a vec_concat.
1139 2017-12-21  James Greenhalgh  <james.greenhalgh@arm.com>
1141         * combine.c (simplify_set): Do not transform subregs to zero_extends
1142         if the destination is not a scalar int mode.
1144 2017-12-21  Eric Botcazou  <ebotcazou@adacore.com>
1146         PR c++/82872
1147         * convert.c (convert_to_integer_1) <POINTER_TYPE>: Do not return the
1148         shared zero if the input has overflowed.
1150 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1152         * config/arm/driver-arm.c (arm_cpu_table): Specify dotprod
1153         support for Cortex-A55 and Cortex-A75.
1155 2017-12-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
1157         * common/config/arm/arm-common.c (compare_opt_names): Add function
1158         comment.  Use strcmp instead of manual loop.
1160 2017-12-21  Martin Liska  <mliska@suse.cz>
1162         PR gcov-profile/83509
1163         * gcov-dump.c (dump_gcov_file): Do not read info about
1164         support_unexecuted_blocks for gcda files.
1166 2017-12-21  Jakub Jelinek  <jakub@redhat.com>
1168         PR rtl-optimization/82973
1169         * emit-rtl.h (valid_for_const_vec_duplicate_p): Rename to ...
1170         (valid_for_const_vector_p): ... this.
1171         * emit-rtl.c (valid_for_const_vec_duplicate_p): Rename to ...
1172         (valid_for_const_vector_p): ... this.  Adjust function comment.
1173         (gen_vec_duplicate): Adjust caller.
1174         * optabs.c (expand_vector_broadcast): Likewise.
1175         * simplify-rtx.c (simplify_const_unary_operation): Don't optimize into
1176         CONST_VECTOR if some element isn't simplified valid_for_const_vector_p
1177         constant.
1178         (simplify_const_binary_operation): Likewise.  Use CONST_FIXED_P macro
1179         instead of GET_CODE == CONST_FIXED.
1180         (simplify_subreg): Use CONST_FIXED_P macro instead of
1181         GET_CODE == CONST_FIXED.
1183         PR target/83488
1184         * config/i386/i386.c (ix86_target_string): Move -mavx512vbmi2 and
1185         -mshstk entries from isa_opts2 to isa_opts and -mhle, -mmovbe,
1186         -mclzero and -mmwaitx entries from isa_opts to isa_opts2.
1187         (ix86_option_override_internal): Adjust for
1188         OPTION_MASK_ISA_{HLE,MOVBE,CLZERO,MWAITX} moving to ix86_isa_flags2
1189         and OPTION_MASK_ISA_SHSTK moving to ix86_isa_flags.
1190         (BDESC_VERIFYS): Remove SPECIAL_ARGS2 related checks.
1191         (ix86_init_mmx_sse_builtins): Remove bdesc_special_args2 handling.
1192         Use def_builtin2 instead of def_builtin for OPTION_MASK_ISA_MWAITX
1193         and OPTION_MASK_ISA_CLZERO builtins.  Use def_builtin instead of
1194         def_builtin2 for CET builtins.
1195         (ix86_expand_builtin): Remove bdesc_special_args2 handling.  Fix
1196         up formatting in IX86_BUILTIN_RDPID code.
1197         * config/i386/i386-builtin.def: Move VBMI2 builtins from SPECIAL_ARGS2
1198         section to SPECIAL_ARGS and from ARGS2 section to ARGS.
1199         * config/i386/i386.opt (mavx512vbmi2, mshstk): Move from
1200         ix86_isa_flags2 to ix86_isa_flags.
1201         (mhle, mmovbe, mclzero, mmwaitx): Move from ix86_isa_flags to
1202         ix86_isa_flags2.
1203         * config/i386/i386-c.c (ix86_target_macros_internal): Check for
1204         OPTION_MASK_ISA_{CLZERO,MWAITX} in isa_flag2 instead of isa_flag.
1205         Check for OPTION_MASK_ISA_{SHSTK,AVX512VBMI2} in isa_flag instead
1206         of isa_flag2.
1207         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET):
1208         Or in OPTION_MASK_ISA_AVX512F_SET.
1209         (OPTION_MASK_ISA_AVX512F_UNSET): Or in
1210         OPTION_MASK_ISA_AVX512VBMI2_UNSET.
1211         (ix86_handle_option): Adjust for
1212         OPTION_MASK_ISA_{SHSTK,AVX512VBMI2}_*SET being in ix86_isa_flags
1213         and OPTION_MASK_ISA_{MOVBE,MWAITX,CLZERO}_*SET in ix86_isa_flags2.
1215 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1216             Alan Hayward  <alan.hayward@arm.com>
1217             David Sherwood  <david.sherwood@arm.com>
1219         * tree-data-ref.h (prune_runtime_alias_test_list): Take the
1220         factor as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1221         * tree-data-ref.c (prune_runtime_alias_test_list): Likewise.
1222         Track polynomial offsets.
1224 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1225             Alan Hayward  <alan.hayward@arm.com>
1226             David Sherwood  <david.sherwood@arm.com>
1228         * tree-vect-data-refs.c (vect_compute_data_ref_alignment):
1229         Treat drb->init as a poly_int.  Fail if its misalignment wrt
1230         vector_alignment isn't known.
1232 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1233             Alan Hayward  <alan.hayward@arm.com>
1234             David Sherwood  <david.sherwood@arm.com>
1236         * tree-vectorizer.h (_loop_vec_info): Add a versioning_threshold
1237         field.
1238         (LOOP_VINFO_VERSIONING_THRESHOLD): New macro
1239         (vect_loop_versioning): Take the loop versioning threshold as a
1240         separate parameter.
1241         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
1242         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
1243         versioning_threshold.
1244         (vect_analyze_loop_2): Compute the loop versioning threshold
1245         whenever loop versioning is needed, and store it in the new
1246         field rather than combining it with the cost model threshold.
1247         (vect_transform_loop): Update call to vect_loop_versioning.
1248         Try to combine the loop versioning and cost thresholds here.
1250 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1251             Alan Hayward  <alan.hayward@arm.com>
1252             David Sherwood  <david.sherwood@arm.com>
1254         * tree-ssa-loop-ivopts.h (strip_offset): Return the offset as
1255         poly_uint64_pod rather than an unsigned HOST_WIDE_INT.
1256         * tree-loop-distribution.c (classify_builtin_st): Update accordingly.
1257         * tree-ssa-loop-ivopts.c (iv_use::addr_offset): Change from
1258         an unsigned HOST_WIDE_INT to a poly_uint64_pod.
1259         (group_compare_offset): Update accordingly.
1260         (split_small_address_groups_p): Likewise.
1261         (record_use): Take addr_offset as a poly_uint64 rather than
1262         an unsigned HOST_WIDE_INT.
1263         (strip_offset): Return the offset as a poly_uint64 rather than
1264         an unsigned HOST_WIDE_INT.
1265         (record_group_use, split_address_groups): Track polynomial offsets.
1266         (add_iv_candidate_for_use): Likewise.
1267         (addr_offset_valid_p): Take the offset as a poly_int64 rather
1268         than a HOST_WIDE_INT.
1269         (strip_offset_1): Return the offset as a poly_int64 rather than
1270         a HOST_WIDE_INT.
1272 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1273             Alan Hayward  <alan.hayward@arm.com>
1274             David Sherwood  <david.sherwood@arm.com>
1276         * tree.h (get_binfo_at_offset): Take the offset as a poly_int64
1277         rather than a HOST_WIDE_INT.
1278         * tree.c (get_binfo_at_offset): Likewise.
1280 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1281             Alan Hayward  <alan.hayward@arm.com>
1282             David Sherwood  <david.sherwood@arm.com>
1284         * ipa-prop.h (build_ref_for_offset): Take the offset as a poly_int64
1285         rather than a HOST_WIDE_INT.
1286         * tree-sra.c (build_ref_for_offset): Likewise.
1288 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1289             Alan Hayward  <alan.hayward@arm.com>
1290             David Sherwood  <david.sherwood@arm.com>
1292         * fold-const.h (mem_ref_offset): Return a poly_offset_int rather
1293         than an offset_int.
1294         * tree.c (mem_ref_offset): Likewise.
1295         (build_simple_mem_ref_loc): Treat MEM_REF offsets as poly_ints.
1296         * builtins.c (get_object_alignment_2): Likewise.
1297         * expr.c (get_inner_reference, expand_expr_real_1): Likewise.
1298         * gimple-fold.c (get_base_constructor): Likewise.
1299         * gimple-ssa-strength-reduction.c (restructure_reference): Likewise.
1300         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1301         Likewise.
1302         * ipa-polymorphic-call.c
1303         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
1304         * ipa-prop.c (compute_complex_assign_jump_func): Likewise.
1305         (get_ancestor_addr_info): Likewise.
1306         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Likewise.
1307         * match.pd: Likewise.
1308         * tree-data-ref.c (dr_analyze_innermost): Likewise.
1309         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1310         * tree-eh.c (tree_could_trap_p): Likewise.
1311         * tree-object-size.c (addr_object_size): Likewise.
1312         * tree-ssa-address.c (copy_ref_info): Likewise.
1313         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p): Likewise.
1314         (indirect_refs_may_alias_p): Likewise.
1315         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1316         * tree-ssa.c (maybe_rewrite_mem_ref_base): Likewise.
1317         (non_rewritable_mem_ref_base): Likewise.
1318         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1319         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1320         * varasm.c (decode_addr_const): Likewise.
1322 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1323             Alan Hayward  <alan.hayward@arm.com>
1324             David Sherwood  <david.sherwood@arm.com>
1326         * gimple-ssa-stor-merging.c (find_bswap_or_nop_load): Track polynomial
1327         offsets for MEM_REFs.
1329 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1330             Alan Hayward  <alan.hayward@arm.com>
1331             David Sherwood  <david.sherwood@arm.com>
1333         * tree-ssanames.h (adjust_ptr_info_misalignment): Take the increment
1334         as a poly_uint64 rather than an unsigned int.
1335         * tree-ssanames.c (adjust_ptr_info_misalignment): Likewise.
1337 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1338             Alan Hayward  <alan.hayward@arm.com>
1339             David Sherwood  <david.sherwood@arm.com>
1341         * varasm.c (addr_const::offset): Change from HOST_WIDE_INT
1342         to poly_int64.
1343         (decode_addr_const): Update accordingly.
1345 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1346             Alan Hayward  <alan.hayward@arm.com>
1347             David Sherwood  <david.sherwood@arm.com>
1349         * tree.h (bit_field_size, bit_field_offset): New functions.
1350         * hsa-gen.c (gen_hsa_addr): Use them.
1351         * tree-ssa-forwprop.c (simplify_bitfield_ref): Likewise.
1352         (simplify_vector_constructor): Likewise.
1353         * tree-ssa-sccvn.c (copy_reference_ops_from_ref): Likewise.
1354         * tree-cfg.c (verify_expr): Require the sizes and offsets of a
1355         BIT_FIELD_REF to be poly_uint64s rather than uhwis.
1356         * fold-const.c (fold_ternary_loc): Protect tree_to_uhwi with
1357         tree_fits_uhwi_p.
1359 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1360             Alan Hayward  <alan.hayward@arm.com>
1361             David Sherwood  <david.sherwood@arm.com>
1363         * expr.h (emit_group_load, emit_group_load_into_temps)
1364         (emit_group_store): Take the size as a poly_int64 rather than an int.
1365         * expr.c (emit_group_load_1, emit_group_load): Likewise.
1366         (emit_group_load_into_temp, emit_group_store): Likewise.
1368 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1369             Alan Hayward  <alan.hayward@arm.com>
1370             David Sherwood  <david.sherwood@arm.com>
1372         * ira-int.h (ira_spilled_reg_stack_slot::width): Change from
1373         an unsigned int to a poly_uint64.
1374         * ira.h (ira_reuse_stack_slot, ira_mark_new_stack_slot): Take the
1375         sizes as poly_uint64s rather than unsigned ints.
1376         * ira-color.c (ira_reuse_stack_slot, ira_mark_new_stack_slot):
1377         Likewise.
1379 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1380             Alan Hayward  <alan.hayward@arm.com>
1381             David Sherwood  <david.sherwood@arm.com>
1383         * lra-constraints.c (emit_inc): Change inc_amount from an int
1384         to a poly_int64.
1386 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1387             Alan Hayward  <alan.hayward@arm.com>
1388             David Sherwood  <david.sherwood@arm.com>
1390         * cfgexpand.c (stack_var::size): Change from a HOST_WIDE_INT
1391         to a poly_uint64.
1392         (add_stack_var, stack_var_cmp, partition_stack_vars)
1393         (dump_stack_var_partition): Update accordingly.
1394         (alloc_stack_frame_space): Take the size as a poly_int64 rather
1395         than a HOST_WIDE_INT.
1396         (expand_stack_vars, expand_one_stack_var_1): Handle polynomial sizes.
1397         (defer_stack_allocation, estimated_stack_frame_size): Likewise.
1398         (account_stack_vars, expand_one_var): Likewise.  Return a poly_uint64
1399         rather than a HOST_WIDE_INT.
1401 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1402             Alan Hayward  <alan.hayward@arm.com>
1403             David Sherwood  <david.sherwood@arm.com>
1405         * target.def (return_pops_args): Treat both the input and output
1406         sizes as poly_int64s rather than HOST_WIDE_INTS.
1407         * targhooks.h (default_return_pops_args): Update accordingly.
1408         * targhooks.c (default_return_pops_args): Likewise.
1409         * doc/tm.texi: Regenerate.
1410         * emit-rtl.h (incoming_args): Change pops_args, size and
1411         outgoing_args_size from int to poly_int64_pod.
1412         * function.h (expr_status): Change x_pending_stack_adjust and
1413         x_stack_pointer_delta from int to poly_int64.
1414         (args_size::constant): Change from HOST_WIDE_INT to poly_int64.
1415         (ARGS_SIZE_RTX): Update accordingly.
1416         * calls.c (highest_outgoing_arg_in_use): Change from int to
1417         unsigned int.
1418         (stack_usage_watermark, stored_args_watermark): New variables.
1419         (stack_region_maybe_used_p, mark_stack_region_used): New functions.
1420         (emit_call_1): Change the stack_size and rounded_stack_size
1421         parameters from HOST_WIDE_INT to poly_int64.  Track n_popped
1422         as a poly_int64.
1423         (save_fixed_argument_area): Check stack_usage_watermark.
1424         (initialize_argument_information): Change old_pending_adj from
1425         a HOST_WIDE_INT * to a poly_int64_pod *.
1426         (compute_argument_block_size): Return the size as a poly_int64
1427         rather than an int.
1428         (finalize_must_preallocate): Track polynomial argument sizes.
1429         (compute_argument_addresses): Likewise.
1430         (internal_arg_pointer_based_exp): Track polynomial offsets.
1431         (mem_overlaps_already_clobbered_arg_p): Rename to...
1432         (mem_might_overlap_already_clobbered_arg_p): ...this and take the
1433         size as a poly_uint64 rather than an unsigned HOST_WIDE_INT.
1434         Check stored_args_used_watermark.
1435         (load_register_parameters): Update accordingly.
1436         (check_sibcall_argument_overlap_1): Likewise.
1437         (combine_pending_stack_adjustment_and_call): Take the unadjusted
1438         args size as a poly_int64 rather than an int.  Return a bool
1439         indicating whether the optimization was possible and return
1440         the new adjustment by reference.
1441         (check_sibcall_argument_overlap): Track polynomail argument sizes.
1442         Update stored_args_watermark.
1443         (can_implement_as_sibling_call_p): Handle polynomial argument sizes.
1444         (expand_call): Likewise.  Maintain stack_usage_watermark and
1445         stored_args_watermark.  Update calls to
1446         combine_pending_stack_adjustment_and_call.
1447         (emit_library_call_value_1): Handle polynomial argument sizes.
1448         Call stack_region_maybe_used_p and mark_stack_region_used.
1449         Maintain stack_usage_watermark.
1450         (store_one_arg): Likewise.  Update call to
1451         mem_overlaps_already_clobbered_arg_p.
1452         * config/arm/arm.c (arm_output_function_prologue): Add a cast to
1453         HOST_WIDE_INT.
1454         * config/avr/avr.c (avr_outgoing_args_size): Likewise.
1455         * config/microblaze/microblaze.c (microblaze_function_prologue):
1456         Likewise.
1457         * config/cr16/cr16.c (cr16_return_pops_args): Update for new
1458         TARGET_RETURN_POPS_ARGS interface.
1459         (cr16_compute_frame, cr16_initial_elimination_offset): Add casts
1460         to HOST_WIDE_INT.
1461         * config/ft32/ft32.c (ft32_compute_frame): Likewise.
1462         * config/i386/i386.c (ix86_return_pops_args): Update for new
1463         TARGET_RETURN_POPS_ARGS interface.
1464         (ix86_expand_split_stack_prologue): Add a cast to HOST_WIDE_INT.
1465         * config/moxie/moxie.c (moxie_compute_frame): Likewise.
1466         * config/m68k/m68k.c (m68k_return_pops_args): Update for new
1467         TARGET_RETURN_POPS_ARGS interface.
1468         * config/vax/vax.c (vax_return_pops_args): Likewise.
1469         * config/pa/pa.h (STACK_POINTER_OFFSET): Add a cast to poly_int64.
1470         (EXIT_IGNORE_STACK): Update reference to crtl->outgoing_args_size.
1471         * config/arm/arm.h (CALLER_INTERWORKING_SLOT_SIZE): Likewise.
1472         * config/powerpcspe/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1473         * config/powerpcspe/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1474         * config/powerpcspe/powerpcspe.h (STACK_DYNAMIC_OFFSET): Likewise.
1475         * config/rs6000/aix.h (STACK_DYNAMIC_OFFSET): Likewise.
1476         * config/rs6000/darwin.h (STACK_DYNAMIC_OFFSET): Likewise.
1477         * config/rs6000/rs6000.h (STACK_DYNAMIC_OFFSET): Likewise.
1478         * dojump.h (saved_pending_stack_adjust): Change x_pending_stack_adjust
1479         and x_stack_pointer_delta from int to poly_int64.
1480         * dojump.c (do_pending_stack_adjust): Update accordingly.
1481         * explow.c (allocate_dynamic_stack_space): Handle polynomial
1482         stack_pointer_deltas.
1483         * function.c (STACK_DYNAMIC_OFFSET): Add a cast to poly_int64.
1484         (pad_to_arg_alignment): Track polynomial offsets.
1485         (assign_parm_find_stack_rtl): Likewise.
1486         (assign_parms, locate_and_pad_parm): Handle polynomial argument sizes.
1487         * toplev.c (output_stack_usage): Update reference to
1488         current_function_pushed_stack_size.
1490 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1491             Alan Hayward  <alan.hayward@arm.com>
1492             David Sherwood  <david.sherwood@arm.com>
1494         * function.c (in_arg_offset, var_offset, dynamic_offset)
1495         (out_arg_offset, cfa_offset): Change from int to poly_int64.
1496         (instantiate_new_reg): Return the new offset as a poly_int64_pod
1497         rather than a HOST_WIDE_INT.
1498         (instantiate_virtual_regs_in_rtx): Track polynomial offsets.
1499         (instantiate_virtual_regs_in_insn): Likewise.
1501 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1502             Alan Hayward  <alan.hayward@arm.com>
1503             David Sherwood  <david.sherwood@arm.com>
1505         * rtl.h (get_args_size, add_args_size_note): New functions.
1506         (find_args_size_adjust): Return a poly_int64 rather than a
1507         HOST_WIDE_INT.
1508         (fixup_args_size_notes): Likewise.  Make the same change to the
1509         end_args_size parameter.
1510         * rtlanal.c (get_args_size, add_args_size_note): New functions.
1511         * builtins.c (expand_builtin_trap): Use add_args_size_note.
1512         * calls.c (emit_call_1): Likewise.
1513         * explow.c (adjust_stack_1): Likewise.
1514         * cfgcleanup.c (old_insns_match_p): Update use of
1515         find_args_size_adjust.
1516         * combine.c (distribute_notes): Track polynomial arg sizes.
1517         * dwarf2cfi.c (dw_trace_info): Change beg_true_args_size,
1518         end_true_args_size, beg_delay_args_size and end_delay_args_size
1519         from HOST_WIDE_INT to poly_int64.
1520         (add_cfi_args_size): Take the args_size as a poly_int64 rather
1521         than a HOST_WIDE_INT.
1522         (notice_args_size, notice_eh_throw, maybe_record_trace_start)
1523         (maybe_record_trace_start_abnormal, scan_trace, connect_traces): Track
1524         polynomial arg sizes.
1525         * emit-rtl.c (try_split): Use get_args_size.
1526         * recog.c (peep2_attempt): Likewise.
1527         * reload1.c (reload_as_needed): Likewise.
1528         * expr.c (find_args_size_adjust): Return the adjustment as a
1529         poly_int64 rather than a HOST_WIDE_INT.
1530         (fixup_args_size_notes): Change end_args_size from a HOST_WIDE_INT
1531         to a poly_int64 and change the return type in the same way.
1532         (emit_single_push_insn): Track polynomial arg sizes.
1534 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1535             Alan Hayward  <alan.hayward@arm.com>
1536             David Sherwood  <david.sherwood@arm.com>
1538         * expr.h (push_block, emit_push_insn): Change the "extra" parameter
1539         from HOST_WIDE_INT to poly_int64.
1540         * expr.c (push_block, emit_push_insn): Likewise.
1542 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1543             Alan Hayward  <alan.hayward@arm.com>
1544             David Sherwood  <david.sherwood@arm.com>
1546         * function.h (frame_space): Change start and length from HOST_WIDE_INT
1547         to poly_int64.
1548         (get_frame_size): Return the size as a poly_int64 rather than a
1549         HOST_WIDE_INT.
1550         (frame_offset_overflow): Take the offset as a poly_int64 rather
1551         than a HOST_WIDE_INT.
1552         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1553         (assign_stack_temp): Likewise for the size.
1554         * function.c (get_frame_size): Return a poly_int64 rather than
1555         a HOST_WIDE_INT.
1556         (frame_offset_overflow): Take the offset as a poly_int64 rather
1557         than a HOST_WIDE_INT.
1558         (try_fit_stack_local): Take the start, length and size as poly_int64s
1559         rather than HOST_WIDE_INTs.  Return the offset as a poly_int64_pod
1560         rather than a HOST_WIDE_INT.
1561         (add_frame_space): Take the start and end as poly_int64s rather than
1562         HOST_WIDE_INTs.
1563         (assign_stack_local_1, assign_stack_local, assign_stack_temp_for_type)
1564         (assign_stack_temp): Likewise for the size.
1565         (temp_slot): Change size, base_offset and full_size from HOST_WIDE_INT
1566         to poly_int64.
1567         (find_temp_slot_from_address): Handle polynomial offsets.
1568         (combine_temp_slots): Likewise.
1569         * emit-rtl.h (rtl_data::x_frame_offset): Change from HOST_WIDE_INT
1570         to poly_int64.
1571         * cfgexpand.c (alloc_stack_frame_space): Return the offset as a
1572         poly_int64 rather than a HOST_WIDE_INT.
1573         (expand_one_stack_var_at): Take the offset as a poly_int64 rather
1574         than a HOST_WIDE_INT.
1575         (expand_stack_vars, expand_one_stack_var_1, expand_used_vars): Handle
1576         polynomial frame offsets.
1577         * config/m32r/m32r-protos.h (m32r_compute_frame_size): Take the size
1578         as a poly_int64 rather than an int.
1579         * config/m32r/m32r.c (m32r_compute_frame_size): Likewise.
1580         * config/v850/v850-protos.h (compute_frame_size): Likewise.
1581         * config/v850/v850.c (compute_frame_size): Likewise.
1582         * config/xtensa/xtensa-protos.h (compute_frame_size): Likewise.
1583         * config/xtensa/xtensa.c (compute_frame_size): Likewise.
1584         * config/pa/pa-protos.h (pa_compute_frame_size): Likewise.
1585         * config/pa/pa.c (pa_compute_frame_size): Likewise.
1586         * explow.h (get_dynamic_stack_base): Take the offset as a poly_int64
1587         rather than a HOST_WIDE_INT.
1588         * explow.c (get_dynamic_stack_base): Likewise.
1589         * final.c (final_start_function): Use the constant lower bound
1590         of the frame size for -Wframe-larger-than.
1591         * ira.c (do_reload): Adjust for new get_frame_size return type.
1592         * lra.c (lra): Likewise.
1593         * reload1.c (reload): Likewise.
1594         * config/avr/avr.c (avr_asm_function_end_prologue): Likewise.
1595         * config/pa/pa.h (EXIT_IGNORE_STACK): Likewise.
1596         * rtlanal.c (get_initial_register_offset): Return the offset as
1597         a poly_int64 rather than a HOST_WIDE_INT.
1599 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1600             Alan Hayward  <alan.hayward@arm.com>
1601             David Sherwood  <david.sherwood@arm.com>
1603         * reload1.c (elim_table): Change initial_offset, offset and
1604         previous_offset from HOST_WIDE_INT to poly_int64_pod.
1605         (offsets_at): Change the target array's element type from
1606         HOST_WIDE_INT to poly_int64_pod.
1607         (set_label_offsets, eliminate_regs_1, eliminate_regs_in_insn)
1608         (elimination_costs_in_insn, update_eliminable_offsets)
1609         (verify_initial_elim_offsets, set_offsets_for_label)
1610         (init_eliminable_invariants): Update after above changes.
1612 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1613             Alan Hayward  <alan.hayward@arm.com>
1614             David Sherwood  <david.sherwood@arm.com>
1616         * reload.h (reload::inc): Change from an int to a poly_int64_pod.
1617         * reload.c (combine_reloads, debug_reload_to_stream): Likewise.
1618         (decomposition): Change start and end from HOST_WIDE_INT
1619         to poly_int64_pod.
1620         (decompose, immune_p): Update accordingly.
1621         (find_inc_amount): Return a poly_int64 rather than an int.
1622         * reload1.c (inc_for_reload): Take the inc_amount as a poly_int64
1623         rather than an int.
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         * tree.h (get_inner_reference): Return the bitsize and bitpos
1630         as poly_int64_pods rather than HOST_WIDE_INT.
1631         * fold-const.h (ptr_difference_const): Return the pointer difference
1632         as a poly_int64_pod rather than a HOST_WIDE_INT.
1633         * expr.c (get_inner_reference): Return the bitsize and bitpos
1634         as poly_int64_pods rather than HOST_WIDE_INT.
1635         (expand_expr_addr_expr_1, expand_expr_real_1): Track polynomial
1636         offsets and sizes.
1637         * fold-const.c (make_bit_field_ref): Take the bitpos as a poly_int64
1638         rather than a HOST_WIDE_INT.  Update call to get_inner_reference.
1639         (optimize_bit_field_compare): Update call to get_inner_reference.
1640         (decode_field_reference): Likewise.
1641         (fold_unary_loc): Track polynomial offsets and sizes.
1642         (split_address_to_core_and_offset): Return the bitpos as a
1643         poly_int64_pod rather than a HOST_WIDE_INT.
1644         (ptr_difference_const): Likewise for the pointer difference.
1645         * asan.c (instrument_derefs): Track polynomial offsets and sizes.
1646         * config/mips/mips.c (r10k_safe_mem_expr_p): Likewise.
1647         * dbxout.c (dbxout_expand_expr): Likewise.
1648         * dwarf2out.c (loc_list_for_address_of_addr_expr_of_indirect_ref)
1649         (loc_list_from_tree_1, fortran_common): Likewise.
1650         * gimple-laddress.c (pass_laddress::execute): Likewise.
1651         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Likewise.
1652         * gimplify.c (gimplify_scan_omp_clauses): Likewise.
1653         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
1654         * tree-affine.c (tree_to_aff_combination): Likewise.
1655         (get_inner_reference_aff): Likewise.
1656         * tree-data-ref.c (split_constant_offset_1): Likewise.
1657         (dr_analyze_innermost): Likewise.
1658         * tree-scalar-evolution.c (interpret_rhs_expr): Likewise.
1659         * tree-sra.c (ipa_sra_check_caller): Likewise.
1660         * tree-vect-data-refs.c (vect_check_gather_scatter): Likewise.
1661         * ubsan.c (maybe_instrument_pointer_overflow): Likewise.
1662         (instrument_bool_enum_load, instrument_object_size): Likewise.
1663         * gimple-ssa-strength-reduction.c (slsr_process_ref): Update call
1664         to get_inner_reference.
1665         * hsa-gen.c (gen_hsa_addr): Likewise.
1666         * sanopt.c (maybe_optimize_ubsan_ptr_ifn): Likewise.
1667         * tsan.c (instrument_expr): Likewise.
1668         * match.pd: Update call to ptr_difference_const.
1670 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1671             Alan Hayward  <alan.hayward@arm.com>
1672             David Sherwood  <david.sherwood@arm.com>
1674         * fold-const.c (fold_comparison): Track sizes and offsets as
1675         poly_int64s rather than HOST_WIDE_INTs when folding address
1676         comparisons.
1678 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1679             Alan Hayward  <alan.hayward@arm.com>
1680             David Sherwood  <david.sherwood@arm.com>
1682         * expr.h (get_bit_range): Return the bitstart and bitend as
1683         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Return the bitpos
1684         as a poly_int64 rather than a HOST_WIDE_INT.
1685         * expr.c (get_bit_range): Likewise.
1686         (expand_assignment): Update call accordingly.
1687         * fold-const.c (optimize_bit_field_compare): Likewise.
1689 2017-12-21  Richard Sandiford  <richard.sandiford@linaro.org>
1690             Alan Hayward  <alan.hayward@arm.com>
1691             David Sherwood  <david.sherwood@arm.com>
1693         * config/aarch64/aarch64-protos.h (aarch64_addr_query_type): New enum.
1694         (aarch64_legitimate_address_p): Use it instead of an rtx code,
1695         as an optional final parameter.
1696         * config/aarch64/aarch64.c (aarch64_classify_address): Likewise.
1697         (aarch64_legitimate_address_p): Likewise.
1698         (aarch64_print_address_internal): Take an aarch64_addr_query_type
1699         instead of an rtx code.
1700         (aarch64_address_valid_for_prefetch_p): Update calls accordingly.
1701         (aarch64_legitimate_address_hook_p): Likewise.
1702         (aarch64_print_ldpstp_address): Likewise.
1703         (aarch64_print_operand_address): Likewise.
1704         (aarch64_address_cost): Likewise.
1705         * config/aarch64/constraints.md (Uml, Umq, Ump, Utq): Likewise.
1706         * config/aarch64/predicates.md (aarch64_mem_pair_operand): Likewise.
1707         (aarch64_mem_pair_lanes_operand): Likewise.
1709 2017-12-20  Richard Biener  <rguenther@suse.de>
1711         * tree-ssa-dom.c (dom_opt_dom_walker::optimize_stmt): Call
1712         update_stmt_if_modified.
1714 2017-12-20  Wilco Dijkstra  <wdijkstr@arm.com>
1716         PR tree-optimization/83491
1717         * tree-ssa-math-opts.c (execute_cse_reciprocals_1): Check for SSA_NAME
1718         before walking uses.  Improve coding style and comments.
1720 2017-12-20  Tom de Vries  <tom@codesourcery.com>
1722         * gimple-fold.c (fold_internal_goacc_dim): Simplify.
1724 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
1726         PR ipa/83506
1727         * ipa-fnsummary.c (pass_data_ipa_free_fn_summary): Use 0 for
1728         todo_flags_finish.
1729         (pass_ipa_free_fn_summary): Add small_p private data member,
1730         initialize to false in the ctor.
1731         (pass_ipa_free_fn_summary::clone,
1732         pass_ipa_free_fn_summary::set_pass_param,
1733         pass_ipa_free_fn_summary::gate): New methods.
1734         (pass_ipa_free_fn_summary::execute): Return TODO_remove_functions
1735         | TODO_dump_symtab if small_p.
1736         * passes.def: Add true parm for the existing pass_ipa_free_fn_summary
1737         entry and add another instance of the pass with false parm after
1738         ipa-pure-const.
1739         * ipa-pure-const.c (pass_ipa_pure_const): Don't call
1740         ipa_free_fn_summary here.
1742 2017-12-20  Paolo Carlini  <paolo.carlini@oracle.com>
1744         * gimplify.c (gimplify_return_expr): Remove dead error_mark_node check.
1746 2017-12-20  Martin Sebor  <msebor@redhat.com>
1748         PR testsuite/83131
1749         * builtins.c (expand_builtin_strlen): Use get_callee_fndecl.
1750         (expand_builtin_strcmp): Call maybe_warn_nonstring_arg.
1751         (expand_builtin_strncmp): Same.
1753 2017-12-20  Alexandre Oliva  <aoliva@redhat.com>
1755         PR bootstrap/83396
1756         * cfgexpand.c (label_rtx_for_bb): Revert SFN changes that
1757         allowed debug stmts before labels.
1758         (expand_gimple_basic_block): Likewise.
1759         * gimple-iterator.c (gimple_find_edge_insert_loc): Likewise.
1760         * gimple-iterator.h (gsi_after_labels): Likewise.
1761         * tree-cfgcleanup (remove_forwarder_block): Likewise, but
1762         rename reused variable, and simplify using gsi_move_before.
1763         * tree-ssa-tail-merge.c (find_duplicate): Likewise.
1764         * tree-cfg.c (make_edges, cleanup_dead_labels): Likewise.
1765         (gimple_can_merge_blocks_p, verify_gimple_in_cfg): Likewise.
1766         (gimple_verify_flow_info, gimple_block_label): Likewise.
1767         (make_blocks): Move debug markers after adjacent labels.
1768         * cfgrtl.c (skip_insns_after_block): Revert SFN changes that
1769         allowed debug insns outside blocks.
1770         * df-scan.c (df_insn_delete): Likewise.
1771         * lra-constraints.c (update_ebb_live_info): Likewise.
1772         * var-tracking.c (get_first_insn, vt_emit_notes): Likewise.
1773         (vt_initialize, delete_vta_debug_insns): Likewise.
1774         (reemit_marker_as_note): Drop BB parm.  Adjust callers.
1776 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1777             Alan Hayward  <alan.hayward@arm.com>
1778             David Sherwood  <david.sherwood@arm.com>
1780         * poly-int-types.h (round_down_to_byte_boundary): New macro.
1781         (round_up_to_byte_boundary): Likewise.
1782         * expr.h (get_bit_range): Add temporary shim.
1783         * gimple-ssa-store-merging.c (store_operand_info): Change the
1784         bitsize, bitpos, bitregion_start and bitregion_end fields from
1785         unsigned HOST_WIDE_INT to poly_uint64.
1786         (merged_store_group): Likewise load_align_base.
1787         (compatible_load_p, compatible_load_p): Update accordingly.
1788         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
1789         (split_group, imm_store_chain_info::output_merged_store): Likewise.
1790         (mem_valid_for_store_merging): Return the bitsize, bitpos,
1791         bitregion_start and bitregion_end as poly_uint64s rather than
1792         unsigned HOST_WIDE_INTs.  Track polynomial offsets internally.
1793         (handled_load): Take the bitsize, bitpos,
1794         bitregion_start and bitregion_end as poly_uint64s rather than
1795         unsigned HOST_WIDE_INTs.
1796         (pass_store_merging::process_store): Update call to
1797         mem_valid_for_store_merging.
1799 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1800             Alan Hayward  <alan.hayward@arm.com>
1801             David Sherwood  <david.sherwood@arm.com>
1803         * builtins.c (get_object_alignment_2): Track polynomial offsets
1804         and sizes.  Update the alignment handling.
1806 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1807             Alan Hayward  <alan.hayward@arm.com>
1808             David Sherwood  <david.sherwood@arm.com>
1810         * tree.h (get_inner_reference): Add a version that returns the
1811         offset and size as poly_int64_pods rather than HOST_WIDE_INTs.
1812         * cfgexpand.c (expand_debug_expr): Track polynomial offsets.  Simply
1813         the case in which bitpos is not associated with the first byte.
1815 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1816             Alan Hayward  <alan.hayward@arm.com>
1817             David Sherwood  <david.sherwood@arm.com>
1819         * tree-affine.h (get_inner_reference_aff): Return the size as a
1820         poly_widest_int.
1821         * tree-affine.c (get_inner_reference_aff): Likewise.
1822         * tree-data-ref.c (dr_may_alias_p): Update accordingly.
1823         * tree-ssa-loop-im.c (mem_refs_may_alias_p): Likewise.
1825 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1826             Alan Hayward  <alan.hayward@arm.com>
1827             David Sherwood  <david.sherwood@arm.com>
1829         * fold-const.c (pointer_may_wrap_p): Take the offset as a
1830         HOST_WIDE_INT rather than a poly_int64.
1832 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1833             Alan Hayward  <alan.hayward@arm.com>
1834             David Sherwood  <david.sherwood@arm.com>
1836         * gimple-ssa-store-merging.c (symbolic_number::bytepos): Change from
1837         HOST_WIDE_INT to poly_int64_pod.
1838         (perform_symbolic_merge): Update accordingly.
1839         (bswap_replace): Likewise.
1841 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1842             Alan Hayward  <alan.hayward@arm.com>
1843             David Sherwood  <david.sherwood@arm.com>
1845         * tree-affine.h (aff_tree::offset): Change from widest_int
1846         to poly_widest_int.
1847         (wide_int_ext_for_comb): Delete.
1848         (aff_combination_const, aff_comb_cannot_overlap_p): Take the
1849         constants as poly_widest_int rather than widest_int.
1850         (aff_combination_constant_multiple_p): Return the multiplier
1851         as a poly_widest_int.
1852         (aff_combination_zero_p, aff_combination_singleton_var_p): Handle
1853         polynomial offsets.
1854         * tree-affine.c (wide_int_ext_for_comb): Make original widest_int
1855         version static and add an overload for poly_widest_int.
1856         (aff_combination_const, aff_combination_add_cst)
1857         (wide_int_constant_multiple_p, aff_comb_cannot_overlap_p): Take
1858         the constants as poly_widest_int rather than widest_int.
1859         (tree_to_aff_combination): Generalize INTEGER_CST case to
1860         poly_int_tree_p.
1861         (aff_combination_to_tree): Track offsets as poly_widest_ints.
1862         (aff_combination_add_product, aff_combination_mult): Handle
1863         polynomial offsets.
1864         (aff_combination_constant_multiple_p): Return the multiplier
1865         as a poly_widest_int.
1866         * tree-predcom.c (determine_offset): Return the offset as a
1867         poly_widest_int.
1868         (split_data_refs_to_components, suitable_component_p): Update
1869         accordingly.
1870         (valid_initializer_p): Update call to
1871         aff_combination_constant_multiple_p.
1872         * tree-ssa-address.c (addr_to_parts): Handle polynomial offsets.
1873         * tree-ssa-loop-ivopts.c (get_address_cost_ainc): Take the step
1874         as a poly_int64 rather than a HOST_WIDE_INT.
1875         (get_address_cost): Handle polynomial offsets.
1876         (iv_elimination_compare_lt): Likewise.
1877         (rewrite_use_nonlinear_expr): Likewise.
1879 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1880             Alan Hayward  <alan.hayward@arm.com>
1881             David Sherwood  <david.sherwood@arm.com>
1883         * tree-dfa.h (get_addr_base_and_unit_offset_1): Return the offset
1884         as a poly_int64_pod rather than a HOST_WIDE_INT.
1885         (get_addr_base_and_unit_offset): Likewise.
1886         * tree-dfa.c (get_addr_base_and_unit_offset_1): Likewise.
1887         (get_addr_base_and_unit_offset): Likewise.
1888         * doc/match-and-simplify.texi: Change off from HOST_WIDE_INT
1889         to poly_int64 in example.
1890         * fold-const.c (fold_binary_loc): Update call to
1891         get_addr_base_and_unit_offset.
1892         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
1893         (maybe_canonicalize_mem_ref_addr): Likewise.
1894         (gimple_fold_stmt_to_constant_1): Likewise.
1895         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
1896         Likewise.
1897         * ipa-param-manipulation.c (ipa_modify_call_arguments): Likewise.
1898         * match.pd: Likewise.
1899         * omp-low.c (lower_omp_target): Likewise.
1900         * tree-sra.c (build_ref_for_offset): Likewise.
1901         (build_debug_ref_for_model): Likewise.
1902         * tree-ssa-address.c (maybe_fold_tmr): Likewise.
1903         * tree-ssa-alias.c (ao_ref_init_from_ptr_and_size): Likewise.
1904         * tree-ssa-ccp.c (optimize_memcpy): Likewise.
1905         * tree-ssa-forwprop.c (forward_propagate_addr_expr_1): Likewise.
1906         (constant_pointer_difference): Likewise.
1907         * tree-ssa-loop-niter.c (expand_simple_operations): Likewise.
1908         * tree-ssa-phiopt.c (jump_function_from_stmt): Likewise.
1909         * tree-ssa-pre.c (create_component_ref_by_pieces_1): Likewise.
1910         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
1911         (vn_reference_maybe_forwprop_address, vn_reference_lookup_3): Likewise.
1912         (set_ssa_val_to): Likewise.
1913         * tree-ssa-strlen.c (get_addr_stridx, addr_stridxptr)
1914         (maybe_diag_stxncpy_trunc): Likewise.
1915         * tree-vrp.c (vrp_prop::check_array_ref): Likewise.
1916         * tree.c (build_simple_mem_ref_loc): Likewise.
1917         (array_at_struct_end_p): Likewise.
1919 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1920             Alan Hayward  <alan.hayward@arm.com>
1921             David Sherwood  <david.sherwood@arm.com>
1923         * tree-dfa.h (get_ref_base_and_extent): Return the base, size and
1924         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1925         (get_ref_base_and_extent_hwi): Declare.
1926         * tree-dfa.c (get_ref_base_and_extent): Return the base, size and
1927         max_size as poly_int64_pods rather than HOST_WIDE_INTs.
1928         (get_ref_base_and_extent_hwi): New function.
1929         * cfgexpand.c (expand_debug_expr): Update call to
1930         get_ref_base_and_extent.
1931         * dwarf2out.c (add_var_loc_to_decl): Likewise.
1932         * gimple-fold.c (get_base_constructor): Return the offset as a
1933         poly_int64_pod rather than a HOST_WIDE_INT.
1934         (fold_const_aggregate_ref_1): Track polynomial sizes and offsets.
1935         * ipa-polymorphic-call.c
1936         (ipa_polymorphic_call_context::set_by_invariant)
1937         (extr_type_from_vtbl_ptr_store): Track polynomial offsets.
1938         (ipa_polymorphic_call_context::ipa_polymorphic_call_context)
1939         (check_stmt_for_type_change): Use get_ref_base_and_extent_hwi
1940         rather than get_ref_base_and_extent.
1941         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
1942         * ipa-prop.c (ipa_load_from_parm_agg, compute_complex_assign_jump_func)
1943         (get_ancestor_addr_info, determine_locally_known_aggregate_parts):
1944         Likewise.
1945         * ipa-param-manipulation.c (ipa_get_adjustment_candidate): Update
1946         call to get_ref_base_and_extent.
1947         * tree-sra.c (create_access, get_access_for_expr): Likewise.
1948         * tree-ssa-alias.c (ao_ref_base, aliasing_component_refs_p)
1949         (stmt_kills_ref_p): Likewise.
1950         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
1951         * tree-ssa-scopedtables.c (avail_expr_hash, equal_mem_array_ref_p):
1952         Likewise.
1953         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
1954         Use get_ref_base_and_extent_hwi rather than get_ref_base_and_extent
1955         when calling native_encode_expr.
1956         * tree-ssa-structalias.c (get_constraint_for_component_ref): Update
1957         call to get_ref_base_and_extent.
1958         (do_structure_copy): Use get_ref_base_and_extent_hwi rather than
1959         get_ref_base_and_extent.
1960         * var-tracking.c (track_expr_p): 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         * ipa-param-manipulation.h (ipa_parm_adjustment::offset): Change from
1967         HOST_WIDE_INT to poly_int64_pod.
1968         * ipa-param-manipulation.c (ipa_modify_call_arguments): Track
1969         polynomail parameter offsets.
1971 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1972             Alan Hayward  <alan.hayward@arm.com>
1973             David Sherwood  <david.sherwood@arm.com>
1975         * gengtype.c (main): Handle poly_int64_pod.
1976         * dwarf2out.h (dw_cfi_oprnd_cfa_loc): New dw_cfi_oprnd_type.
1977         (dw_cfi_oprnd::dw_cfi_cfa_loc): New field.
1978         (dw_cfa_location::offset, dw_cfa_location::base_offset): Change
1979         from HOST_WIDE_INT to poly_int64_pod.
1980         * dwarf2cfi.c (queued_reg_save::cfa_offset): Likewise.
1981         (copy_cfa): New function.
1982         (lookup_cfa_1): Use the cached dw_cfi_cfa_loc, if it exists.
1983         (cfi_oprnd_equal_p): Handle dw_cfi_oprnd_cfa_loc.
1984         (cfa_equal_p, dwarf2out_frame_debug_adjust_cfa)
1985         (dwarf2out_frame_debug_cfa_offset, dwarf2out_frame_debug_expr)
1986         (initial_return_save): Treat offsets as poly_ints.
1987         (def_cfa_0): Likewise.  Cache the CFA in dw_cfi_cfa_loc if either
1988         offset is nonconstant.
1989         (reg_save): Take the offset as a poly_int64.  Fall back to
1990         DW_CFA_expression for nonconstant offsets.
1991         (queue_reg_save): Take the offset as a poly_int64.
1992         * dwarf2out.c (dw_cfi_oprnd2_desc): Handle DW_CFA_def_cfa_expression.
1994 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
1995             Alan Hayward  <alan.hayward@arm.com>
1996             David Sherwood  <david.sherwood@arm.com>
1998         * rtl.h (operand_subword, operand_subword_force): Take the offset
1999         as a poly_uint64 an unsigned int.
2000         * emit-rtl.c (operand_subword, operand_subword_force): 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         * doc/rtl.texi: Update documentation of SUBREG_BYTE.  Document the
2007         'p' format code.  Use INT_LIST rather than SUBREG as the example of
2008         a code with an XINT and an XEXP.  Remove the implication that
2009         accessing an rtx field using XINT is expected to work.
2010         * rtl.def (SUBREG): Change format from "ei" to "ep".
2011         * rtl.h (rtunion::rt_subreg): New field.
2012         (XCSUBREG): New macro.
2013         (SUBREG_BYTE): Use it.
2014         (subreg_shape): Change offset from an unsigned int to a poly_uint16.
2015         Update constructor accordingly.
2016         (subreg_shape::operator ==): Update accordingly.
2017         (subreg_shape::unique_id): Return an unsigned HOST_WIDE_INT rather
2018         than an unsigned int.
2019         (subreg_lsb, subreg_lowpart_offset, subreg_highpart_offset): Return
2020         a poly_uint64 rather than an unsigned int.
2021         (subreg_lsb_1): Likewise.  Take the offset as a poly_uint64 rather
2022         than an unsigned int.
2023         (subreg_size_offset_from_lsb, subreg_size_lowpart_offset)
2024         (subreg_size_highpart_offset): Return a poly_uint64 rather than
2025         an unsigned int.  Take the sizes as poly_uint64s.
2026         (subreg_offset_from_lsb): Return a poly_uint64 rather than
2027         an unsigned int.  Take the shift as a poly_uint64 rather than
2028         an unsigned int.
2029         (subreg_regno_offset, subreg_offset_representable_p): Take the offset
2030         as a poly_uint64 rather than an unsigned int.
2031         (simplify_subreg_regno): Likewise.
2032         (byte_lowpart_offset): Return the memory offset as a poly_int64
2033         rather than an int.
2034         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2035         poly_uint64 rather than an unsigned int.
2036         (simplify_subreg, simplify_gen_subreg, subreg_get_info)
2037         (gen_rtx_SUBREG, validate_subreg): Take the subreg offset as a
2038         poly_uint64 rather than an unsigned int.
2039         * rtl.c (rtx_format): Describe 'p' in comment.
2040         (copy_rtx, rtx_equal_p_cb, rtx_equal_p): Handle 'p'.
2041         * emit-rtl.c (validate_subreg, gen_rtx_SUBREG): Take the subreg
2042         offset as a poly_uint64 rather than an unsigned int.
2043         (byte_lowpart_offset): Return the memory offset as a poly_int64
2044         rather than an int.
2045         (subreg_memory_offset): Likewise.  Take the subreg offset as a
2046         poly_uint64 rather than an unsigned int.
2047         (subreg_size_lowpart_offset, subreg_size_highpart_offset): Take the
2048         mode sizes as poly_uint64s rather than unsigned ints.  Return a
2049         poly_uint64 rather than an unsigned int.
2050         (subreg_lowpart_p): Treat subreg offsets as poly_ints.
2051         (copy_insn_1): Handle 'p'.
2052         * rtlanal.c (set_noop_p): Treat subregs offsets as poly_uint64s.
2053         (subreg_lsb_1): Take the subreg offset as a poly_uint64 rather than
2054         an unsigned int.  Return the shift in the same way.
2055         (subreg_lsb): Return the shift as a poly_uint64 rather than an
2056         unsigned int.
2057         (subreg_size_offset_from_lsb): Take the sizes and shift as
2058         poly_uint64s rather than unsigned ints.  Return the offset as
2059         a poly_uint64.
2060         (subreg_get_info, subreg_regno_offset, subreg_offset_representable_p)
2061         (simplify_subreg_regno): Take the offset as a poly_uint64 rather than
2062         an unsigned int.
2063         * rtlhash.c (add_rtx): Handle 'p'.
2064         * genemit.c (gen_exp): Likewise.
2065         * gengenrtl.c (type_from_format, gendef): Likewise.
2066         * gensupport.c (subst_pattern_match, get_alternatives_number)
2067         (collect_insn_data, alter_predicate_for_insn, alter_constraints)
2068         (subst_dup): Likewise.
2069         * gengtype.c (adjust_field_rtx_def): Likewise.
2070         * genrecog.c (find_operand, find_matching_operand, validate_pattern)
2071         (match_pattern_2): Likewise.
2072         (rtx_test::SUBREG_FIELD): New rtx_test::kind_enum.
2073         (rtx_test::subreg_field): New function.
2074         (operator ==, safe_to_hoist_p, transition_parameter_type)
2075         (print_nonbool_test, print_test): Handle SUBREG_FIELD.
2076         * genattrtab.c (attr_rtx_1): Say that 'p' is deliberately not handled.
2077         * genpeep.c (match_rtx): Likewise.
2078         * print-rtl.c (print_poly_int): Include if GENERATOR_FILE too.
2079         (rtx_writer::print_rtx_operand): Handle 'p'.
2080         (print_value): Handle SUBREG.
2081         * read-rtl.c (apply_int_iterator): Likewise.
2082         (rtx_reader::read_rtx_operand): Handle 'p'.
2083         * alias.c (rtx_equal_for_memref_p): Likewise.
2084         * cselib.c (rtx_equal_for_cselib_1, cselib_hash_rtx): Likewise.
2085         * caller-save.c (replace_reg_with_saved_mem): Treat subreg offsets
2086         as poly_ints.
2087         * calls.c (expand_call): Likewise.
2088         * combine.c (combine_simplify_rtx, expand_field_assignment): Likewise.
2089         (make_extraction, gen_lowpart_for_combine): Likewise.
2090         * loop-invariant.c (hash_invariant_expr_1, invariant_expr_equal_p):
2091         Likewise.
2092         * cse.c (remove_invalid_subreg_refs): Take the offset as a poly_uint64
2093         rather than an unsigned int.  Treat subreg offsets as poly_ints.
2094         (exp_equiv_p): Handle 'p'.
2095         (hash_rtx_cb): Likewise.  Treat subreg offsets as poly_ints.
2096         (equiv_constant, cse_insn): Treat subreg offsets as poly_ints.
2097         * dse.c (find_shift_sequence): Likewise.
2098         * dwarf2out.c (rtl_for_decl_location): Likewise.
2099         * expmed.c (extract_low_bits): Likewise.
2100         * expr.c (emit_group_store, undefined_operand_subword_p): Likewise.
2101         (expand_expr_real_2): Likewise.
2102         * final.c (alter_subreg): Likewise.
2103         (leaf_renumber_regs_insn): Handle 'p'.
2104         * function.c (assign_parm_find_stack_rtl, assign_parm_setup_stack):
2105         Treat subreg offsets as poly_ints.
2106         * fwprop.c (forward_propagate_and_simplify): Likewise.
2107         * ifcvt.c (noce_emit_move_insn, noce_emit_cmove): Likewise.
2108         * ira.c (get_subreg_tracking_sizes): Likewise.
2109         * ira-conflicts.c (go_through_subreg): Likewise.
2110         * ira-lives.c (process_single_reg_class_operands): Likewise.
2111         * jump.c (rtx_renumbered_equal_p): Likewise.  Handle 'p'.
2112         * lower-subreg.c (simplify_subreg_concatn): Take the subreg offset
2113         as a poly_uint64 rather than an unsigned int.
2114         (simplify_gen_subreg_concatn, resolve_simple_move): Treat
2115         subreg offsets as poly_ints.
2116         * lra-constraints.c (operands_match_p): Handle 'p'.
2117         (match_reload, curr_insn_transform): Treat subreg offsets as poly_ints.
2118         * lra-spills.c (assign_mem_slot): Likewise.
2119         * postreload.c (move2add_valid_value_p): Likewise.
2120         * recog.c (general_operand, indirect_operand): Likewise.
2121         * regcprop.c (copy_value, maybe_mode_change): Likewise.
2122         (copyprop_hardreg_forward_1): Likewise.
2123         * reginfo.c (simplifiable_subregs_hasher::hash, simplifiable_subregs)
2124         (record_subregs_of_mode): Likewise.
2125         * rtlhooks.c (gen_lowpart_general, gen_lowpart_if_possible): Likewise.
2126         * reload.c (operands_match_p): Handle 'p'.
2127         (find_reloads_subreg_address): Treat subreg offsets as poly_ints.
2128         * reload1.c (alter_reg, choose_reload_regs): Likewise.
2129         (compute_reload_subreg_offset): Likewise, and return an poly_int64.
2130         * simplify-rtx.c (simplify_truncation, simplify_binary_operation_1):
2131         (test_vector_ops_duplicate): Treat subreg offsets as poly_ints.
2132         (simplify_const_poly_int_tests<N>::run): Likewise.
2133         (simplify_subreg, simplify_gen_subreg): Take the subreg offset as
2134         a poly_uint64 rather than an unsigned int.
2135         * valtrack.c (debug_lowpart_subreg): Likewise.
2136         * var-tracking.c (var_lowpart): Likewise.
2137         (loc_cmp): Handle 'p'.
2139 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2140             Alan Hayward  <alan.hayward@arm.com>
2141             David Sherwood  <david.sherwood@arm.com>
2143         * ira.c (get_subreg_tracking_sizes): New function.
2144         (init_live_subregs): Take an integer size rather than a register.
2145         (build_insn_chain): Use get_subreg_tracking_sizes.  Update calls
2146         to init_live_subregs.
2148 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2149             Alan Hayward  <alan.hayward@arm.com>
2150             David Sherwood  <david.sherwood@arm.com>
2152         * expr.c (store_constructor_field): Change bitsize from a
2153         unsigned HOST_WIDE_INT to a poly_uint64 and bitpos from a
2154         HOST_WIDE_INT to a poly_int64.
2155         (store_constructor): Change size from a HOST_WIDE_INT to
2156         a poly_int64.
2157         (store_field): Likewise bitsize and bitpos.
2159 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2160             Alan Hayward  <alan.hayward@arm.com>
2161             David Sherwood  <david.sherwood@arm.com>
2163         * expmed.h (store_bit_field): Change bitregion_start and
2164         bitregion_end from unsigned HOST_WIDE_INT to poly_uint64.
2165         * expmed.c (adjust_bit_field_mem_for_reg, strict_volatile_bitfield_p)
2166         (store_bit_field_1, store_integral_bit_field, store_bit_field)
2167         (store_fixed_bit_field, store_split_bit_field): Likewise.
2168         * expr.c (store_constructor_field, store_field): Likewise.
2169         (optimize_bitfield_assignment_op): Likewise.  Make the same change
2170         to bitsize and bitpos.
2171         * machmode.h (bit_field_mode_iterator): Change m_bitregion_start
2172         and m_bitregion_end from HOST_WIDE_INT to poly_int64.  Make the
2173         same change in the constructor arguments.
2174         (get_best_mode): Change bitregion_start and bitregion_end from
2175         unsigned HOST_WIDE_INT to poly_uint64.
2176         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
2177         Change bitregion_start and bitregion_end from HOST_WIDE_INT to
2178         poly_int64.
2179         (bit_field_mode_iterator::next_mode): Update for new types
2180         of m_bitregion_start and m_bitregion_end.
2181         (get_best_mode): Change bitregion_start and bitregion_end from
2182         unsigned HOST_WIDE_INT to poly_uint64.
2184 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2185             Alan Hayward  <alan.hayward@arm.com>
2186             David Sherwood  <david.sherwood@arm.com>
2188         * rtl.h (simplify_gen_subreg): Add a temporary overload that
2189         accepts poly_uint64 offsets.
2190         * expmed.h (extract_bit_field): Take bitsize and bitnum as
2191         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2192         * expmed.c (lowpart_bit_field_p): Likewise.
2193         (extract_bit_field_as_subreg): New function, split out from...
2194         (extract_bit_field_1): ...here.  Take bitsize and bitnum as
2195         poly_uint64s rather than unsigned HOST_WIDE_INTs.  For vector
2196         extractions, check that BITSIZE matches the size of the extracted
2197         value and that BITNUM is an exact multiple of that size.
2198         If all else fails, try forcing the value into memory if
2199         BITNUM is variable, and adjusting the address so that the
2200         offset is constant.  Split the part that can only handle constant
2201         bitsize and bitnum out into...
2202         (extract_integral_bit_field): ...this new function.
2203         (extract_bit_field): Take bitsize and bitnum as poly_uint64s
2204         rather than unsigned HOST_WIDE_INTs.
2206 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2207             Alan Hayward  <alan.hayward@arm.com>
2208             David Sherwood  <david.sherwood@arm.com>
2210         * expmed.h (store_bit_field): Take bitsize and bitnum as
2211         poly_uint64s rather than unsigned HOST_WIDE_INTs.
2212         * expmed.c (simple_mem_bitfield_p): Likewise.  Add a parameter
2213         that returns the byte size.
2214         (store_bit_field_1): Take bitsize and bitnum as
2215         poly_uint64s rather than unsigned HOST_WIDE_INTs.  Update call
2216         to simple_mem_bitfield_p.  Split the part that can only handle
2217         constant bitsize and bitnum out into...
2218         (store_integral_bit_field): ...this new function.
2219         (store_bit_field): Take bitsize and bitnum as poly_uint64s rather
2220         than unsigned HOST_WIDE_INTs.
2221         (extract_bit_field_1): Update call to simple_mem_bitfield_p.
2223 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2224             Alan Hayward  <alan.hayward@arm.com>
2225             David Sherwood  <david.sherwood@arm.com>
2227         * lra-int.h (lra_reg): Change offset from int to poly_int64.
2228         (lra_insn_recog_data): Change sp_offset from HOST_WIDE_INT
2229         to poly_int64.
2230         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2231         update_sp_offset from a HOST_WIDE_INT to a poly_int64.
2232         (lra_update_reg_val_offset, lra_reg_val_equal_p): Take the
2233         offset as a poly_int64 rather than an int.
2234         * lra-assigns.c (find_hard_regno_for_1): Handle poly_int64 offsets.
2235         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
2236         * lra-constraints.c (equiv_address_substitution): Track offsets
2237         as poly_int64s.
2238         (emit_inc): Check poly_int_rtx_p instead of CONST_INT_P.
2239         (curr_insn_transform): Handle the new form of sp_offset.
2240         * lra-eliminations.c (lra_elim_table): Change previous_offset
2241         and offset from HOST_WIDE_INT to poly_int64.
2242         (print_elim_table, update_reg_eliminate): Update accordingly.
2243         (self_elim_offsets): Change from HOST_WIDE_INT to poly_int64_pod.
2244         (get_elimination): Update accordingly.
2245         (form_sum): Check poly_int_rtx_p instead of CONST_INT_P.
2246         (lra_eliminate_regs_1, eliminate_regs_in_insn): Change
2247         update_sp_offset from a HOST_WIDE_INT to a poly_int64.  Handle
2248         poly_int64 offsets generally.
2249         (curr_sp_change): Change from HOST_WIDE_INT to poly_int64.
2250         (mark_not_eliminable, init_elimination): Update accordingly.
2251         (remove_reg_equal_offset_note): Return a bool and pass the new
2252         offset back by pointer as a poly_int64.
2253         * lra-remat.c (change_sp_offset): Take sp_offset as a poly_int64
2254         rather than a HOST_WIDE_INT.
2255         (do_remat): Track offsets poly_int64s.
2256         * lra.c (lra_update_insn_recog_data, setup_sp_offset): Likewise.
2258 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2259             Alan Hayward  <alan.hayward@arm.com>
2260             David Sherwood  <david.sherwood@arm.com>
2262         * rtl.h (mem_attrs): Add a default constructor.  Change size and
2263         offset from HOST_WIDE_INT to poly_int64.
2264         * emit-rtl.h (set_mem_offset, set_mem_size, adjust_address_1)
2265         (adjust_automodify_address_1, set_mem_attributes_minus_bitpos)
2266         (widen_memory_access): Take the sizes and offsets as poly_int64s
2267         rather than HOST_WIDE_INTs.
2268         * alias.c (ao_ref_from_mem): Handle the new form of MEM_OFFSET.
2269         (offset_overlap_p): Take poly_int64s rather than HOST_WIDE_INTs
2270         and ints.
2271         (adjust_offset_for_component_ref): Change the offset from a
2272         HOST_WIDE_INT to a poly_int64.
2273         (nonoverlapping_memrefs_p): Track polynomial offsets and sizes.
2274         * cfgcleanup.c (merge_memattrs): Update after mem_attrs changes.
2275         * dce.c (find_call_stack_args): Likewise.
2276         * dse.c (record_store): Likewise.
2277         * dwarf2out.c (tls_mem_loc_descriptor, dw_sra_loc_expr): Likewise.
2278         * print-rtl.c (rtx_writer::print_rtx): Likewise.
2279         * read-rtl-function.c (test_loading_mem): Likewise.
2280         * rtlanal.c (may_trap_p_1): Likewise.
2281         * simplify-rtx.c (delegitimize_mem_from_attrs): Likewise.
2282         * var-tracking.c (int_mem_offset, track_expr_p): Likewise.
2283         * emit-rtl.c (mem_attrs_eq_p, get_mem_align_offset): Likewise.
2284         (mem_attrs::mem_attrs): New function.
2285         (set_mem_attributes_minus_bitpos): Change bitpos from a
2286         HOST_WIDE_INT to poly_int64.
2287         (set_mem_alias_set, set_mem_addr_space, set_mem_align, set_mem_expr)
2288         (clear_mem_offset, clear_mem_size, change_address)
2289         (get_spill_slot_decl, set_mem_attrs_for_spill): Directly
2290         initialize mem_attrs.
2291         (set_mem_offset, set_mem_size, adjust_address_1)
2292         (adjust_automodify_address_1, offset_address, widen_memory_access):
2293         Likewise.  Take poly_int64s rather than HOST_WIDE_INT.
2295 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2296             Alan Hayward  <alan.hayward@arm.com>
2297             David Sherwood  <david.sherwood@arm.com>
2299         * rtlanal.c (rtx_addr_can_trap_p_1): Take the offset and size
2300         as poly_int64s rather than HOST_WIDE_INTs.  Use a size of -1
2301         rather than 0 to represent an unknown size.  Assert that the size
2302         is known when the mode isn't BLKmode.
2303         (may_trap_p_1): Use -1 for unknown sizes.
2304         (rtx_addr_can_trap_p): Likewise.  Pass BLKmode rather than VOIDmode.
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         * dse.c (store_info): Change offset and width from HOST_WIDE_INT
2311         to poly_int64.  Update commentary for positions_needed.large.
2312         (read_info_type): Change offset and width from HOST_WIDE_INT
2313         to poly_int64.
2314         (set_usage_bits): Likewise.
2315         (canon_address): Return the offset as a poly_int64 rather than
2316         a HOST_WIDE_INT.  Use strip_offset_and_add.
2317         (set_all_positions_unneeded, any_positions_needed_p): Use
2318         positions_needed.large to track stores with non-constant widths.
2319         (all_positions_needed_p): Likewise.  Take the offset and width
2320         as poly_int64s rather than ints.  Assert that rhs is nonnull.
2321         (record_store): Cope with non-constant offsets and widths.
2322         Nullify the rhs of an earlier store if we can't tell which bytes
2323         of it are needed.
2324         (find_shift_sequence): Take the access_size and shift as poly_int64s
2325         rather than ints.
2326         (get_stored_val): Take the read_offset and read_width as poly_int64s
2327         rather than HOST_WIDE_INTs.
2328         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Handle
2329         non-constant offsets and widths.
2331 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2332             Alan Hayward  <alan.hayward@arm.com>
2333             David Sherwood  <david.sherwood@arm.com>
2335         * inchash.h (inchash::hash::add_poly_int): New function.
2336         * tree-ssa-alias.h (ao_ref::offset, ao_ref::size, ao_ref::max_size):
2337         Use poly_int64 rather than HOST_WIDE_INT.
2338         (ao_ref::max_size_known_p): New function.
2339         * tree-ssa-sccvn.h (vn_reference_op_struct::off): Use poly_int64_pod
2340         rather than HOST_WIDE_INT.
2341         * tree-ssa-alias.c (ao_ref_base): Apply get_ref_base_and_extent
2342         to temporaries until its interface is adjusted to match.
2343         (ao_ref_init_from_ptr_and_size): Handle polynomial offsets and sizes.
2344         (aliasing_component_refs_p, decl_refs_may_alias_p)
2345         (indirect_ref_may_alias_decl_p, indirect_refs_may_alias_p): Take
2346         the offsets and max_sizes as poly_int64s instead of HOST_WIDE_INTs.
2347         (refs_may_alias_p_1, stmt_kills_ref_p): Adjust for changes to
2348         ao_ref fields.
2349         * alias.c (ao_ref_from_mem): Likewise.
2350         * tree-ssa-dce.c (mark_aliased_reaching_defs_necessary_1): Likewise.
2351         * tree-ssa-dse.c (valid_ao_ref_for_dse, normalize_ref)
2352         (clear_bytes_written_by, setup_live_bytes_from_ref, compute_trims)
2353         (maybe_trim_complex_store, maybe_trim_constructor_store)
2354         (live_bytes_read, dse_classify_store): Likewise.
2355         * tree-ssa-sccvn.c (vn_reference_compute_hash, vn_reference_eq):
2356         (copy_reference_ops_from_ref, ao_ref_init_from_vn_reference)
2357         (fully_constant_vn_reference_p, valueize_refs_1): Likewise.
2358         (vn_reference_lookup_3): Likewise.
2359         * tree-ssa-uninit.c (warn_uninitialized_vars): Likewise.
2361 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2362             Alan Hayward  <alan.hayward@arm.com>
2363             David Sherwood  <david.sherwood@arm.com>
2365         * tree-ssa-alias.c (indirect_ref_may_alias_decl_p)
2366         (indirect_refs_may_alias_p): Use ranges_may_overlap_p
2367         instead of ranges_overlap_p.
2369 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2370             Alan Hayward  <alan.hayward@arm.com>
2371             David Sherwood  <david.sherwood@arm.com>
2373         * tree-ssa-alias.c (same_addr_size_stores_p): Take the offsets and
2374         sizes as poly_int64s rather than HOST_WIDE_INTs.
2376 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2377             Alan Hayward  <alan.hayward@arm.com>
2378             David Sherwood  <david.sherwood@arm.com>
2380         * gimple-fold.h (fold_ctor_reference): Take the offset and size
2381         as poly_uint64 rather than unsigned HOST_WIDE_INT.
2382         * gimple-fold.c (fold_ctor_reference): Likewise.
2384 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2385             Alan Hayward  <alan.hayward@arm.com>
2386             David Sherwood  <david.sherwood@arm.com>
2388         * target.def (dwarf_poly_indeterminate_value): New hook.
2389         * targhooks.h (default_dwarf_poly_indeterminate_value): Declare.
2390         * targhooks.c (default_dwarf_poly_indeterminate_value): New function.
2391         * doc/tm.texi.in (TARGET_DWARF_POLY_INDETERMINATE_VALUE): Document.
2392         * doc/tm.texi: Regenerate.
2393         * dwarf2out.h (build_cfa_loc, build_cfa_aligned_loc): Take the
2394         offset as a poly_int64.
2395         * dwarf2out.c (new_reg_loc_descr): Move later in file.  Take the
2396         offset as a poly_int64.
2397         (loc_descr_plus_const, loc_list_plus_const, build_cfa_aligned_loc):
2398         Take the offset as a poly_int64.
2399         (build_cfa_loc): Likewise.  Use loc_descr_plus_const.
2400         (frame_pointer_fb_offset): Change to a poly_int64.
2401         (int_loc_descriptor): Take the offset as a poly_int64.  Use
2402         targetm.dwarf_poly_indeterminate_value for polynomial offsets.
2403         (based_loc_descr): Take the offset as a poly_int64.
2404         Use strip_offset_and_add to handle (plus X (const)).
2405         Use new_reg_loc_descr instead of an open-coded version of the
2406         previous implementation.
2407         (mem_loc_descriptor): Handle CONST_POLY_INT.
2408         (compute_frame_pointer_to_fb_displacement): Take the offset as a
2409         poly_int64.  Use strip_offset_and_add to handle (plus X (const)).
2411 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2412             Alan Hayward  <alan.hayward@arm.com>
2413             David Sherwood  <david.sherwood@arm.com>
2415         * rtl.h (reg_attrs::offset): Change from HOST_WIDE_INT to poly_int64.
2416         (gen_rtx_REG_offset): Take the offset as a poly_int64.
2417         * inchash.h (inchash::hash::add_poly_hwi): New function.
2418         * gengtype.c (main): Register poly_int64.
2419         * emit-rtl.c (reg_attr_hasher::hash): Use inchash.  Treat the
2420         offset as a poly_int.
2421         (reg_attr_hasher::equal): Use must_eq to compare offsets.
2422         (get_reg_attrs, update_reg_offset, gen_rtx_REG_offset): Take the
2423         offset as a poly_int64.
2424         (set_reg_attrs_from_value): Treat the offset as a poly_int64.
2425         * print-rtl.c (print_poly_int): New function.
2426         (rtx_writer::print_rtx_operand_code_r): Treat REG_OFFSET as
2427         a poly_int.
2428         * var-tracking.c (track_offset_p, get_tracked_reg_offset): New
2429         functions.
2430         (var_reg_set, var_reg_delete_and_set, var_reg_delete): Use them.
2431         (same_variable_part_p, track_loc_p): Take the offset as a poly_int64.
2432         (vt_get_decl_and_offset): Return the offset as a poly_int64.
2433         Enforce track_offset_p for parts of a PARALLEL.
2434         (vt_add_function_parameter): Use const_offset for the final
2435         offset to track.  Use get_tracked_reg_offset for the parts
2436         of a PARALLEL.
2438 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2439             Alan Hayward  <alan.hayward@arm.com>
2440             David Sherwood  <david.sherwood@arm.com>
2442         * target.def (truly_noop_truncation): Take poly_uint64s instead of
2443         unsigned ints.  Change default to hook_bool_puint64_puint64_true.
2444         * doc/tm.texi: Regenerate.
2445         * hooks.h (hook_bool_uint_uint_true): Delete.
2446         (hook_bool_puint64_puint64_true): Declare.
2447         * hooks.c (hook_bool_uint_uint_true): Delete.
2448         (hook_bool_puint64_puint64_true): New function.
2449         * config/mips/mips.c (mips_truly_noop_truncation): Take poly_uint64s
2450         instead of unsigned ints.
2451         * config/spu/spu.c (spu_truly_noop_truncation): Likewise.
2452         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): Likewise.
2454 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2455             Alan Hayward  <alan.hayward@arm.com>
2456             David Sherwood  <david.sherwood@arm.com>
2458         * optabs.h (expand_operand): Add an int_value field.
2459         (create_expand_operand): Add an int_value parameter and use it
2460         to initialize the new expand_operand field.
2461         (create_integer_operand): Replace with a declaration of a function
2462         that accepts poly_int64s.  Move the implementation to...
2463         * optabs.c (create_integer_operand): ...here.
2464         (maybe_legitimize_operand): For EXPAND_INTEGER, check whether
2465         the mode preserves the value of int_value, instead of calling
2466         const_int_operand on the rtx.  Use gen_int_mode to generate
2467         the new rtx.
2469 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2470             Alan Hayward  <alan.hayward@arm.com>
2471             David Sherwood  <david.sherwood@arm.com>
2473         * dumpfile.h (dump_dec): Declare.
2474         * dumpfile.c (dump_dec): New function.
2475         * pretty-print.h (pp_wide_integer): Turn into a function and
2476         declare a poly_int version.
2477         * pretty-print.c (pp_wide_integer): New function for poly_ints.
2479 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2480             Alan Hayward  <alan.hayward@arm.com>
2481             David Sherwood  <david.sherwood@arm.com>
2483         * doc/generic.texi (POLY_INT_CST): Document.
2484         * tree.def (POLY_INT_CST): New tree code.
2485         * treestruct.def (TS_POLY_INT_CST): New tree layout.
2486         * tree-core.h (tree_poly_int_cst): New struct.
2487         (tree_node): Add a poly_int_cst field.
2488         * tree.h (POLY_INT_CST_P, POLY_INT_CST_COEFF): New macros.
2489         (wide_int_to_tree, force_fit_type): Take a poly_wide_int_ref
2490         instead of a wide_int_ref.
2491         (build_int_cst, build_int_cst_type): Take a poly_int64 instead
2492         of a HOST_WIDE_INT.
2493         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2494         instead of an unsigned HOST_WIDE_INT.
2495         (build_poly_int_cst, tree_fits_poly_int64_p, tree_fits_poly_uint64_p)
2496         (ptrdiff_tree_p): Declare.
2497         (tree_to_poly_int64, tree_to_poly_uint64): Likewise.  Provide
2498         extern inline implementations if the target doesn't use POLY_INT_CST.
2499         (poly_int_tree_p): New function.
2500         (wi::unextended_tree): New class.
2501         (wi::int_traits <unextended_tree>): New override.
2502         (wi::extended_tree): Add a default constructor.
2503         (wi::extended_tree::get_tree): New function.
2504         (wi::widest_extended_tree, wi::offset_extended_tree): New typedefs.
2505         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): Use them.
2506         (wi::tree_to_poly_widest_ref, wi::tree_to_poly_offset_ref)
2507         (wi::tree_to_poly_wide_ref): New typedefs.
2508         (wi::ints_for): Provide overloads for extended_tree and
2509         unextended_tree.
2510         (poly_int_cst_value, wi::to_poly_widest, wi::to_poly_offset)
2511         (wi::to_wide): New functions.
2512         (wi::fits_to_boolean_p, wi::fits_to_tree_p): Handle poly_ints.
2513         * tree.c (poly_int_cst_hasher): New struct.
2514         (poly_int_cst_hash_table): New variable.
2515         (tree_node_structure_for_code, tree_code_size, simple_cst_equal)
2516         (valid_constant_size_p, add_expr, drop_tree_overflow): Handle
2517         POLY_INT_CST.
2518         (initialize_tree_contains_struct): Handle TS_POLY_INT_CST.
2519         (init_ttree): Initialize poly_int_cst_hash_table.
2520         (build_int_cst, build_int_cst_type, build_invariant_address): Take
2521         a poly_int64 instead of a HOST_WIDE_INT.
2522         (build_int_cstu, build_array_type_nelts): Take a poly_uint64
2523         instead of an unsigned HOST_WIDE_INT.
2524         (wide_int_to_tree): Rename to...
2525         (wide_int_to_tree_1): ...this.
2526         (build_new_poly_int_cst, build_poly_int_cst): New functions.
2527         (force_fit_type): Take a poly_wide_int_ref instead of a wide_int_ref.
2528         (wide_int_to_tree): New function that takes a poly_wide_int_ref.
2529         (ptrdiff_tree_p, tree_to_poly_int64, tree_to_poly_uint64)
2530         (tree_fits_poly_int64_p, tree_fits_poly_uint64_p): New functions.
2531         * lto-streamer-out.c (DFS::DFS_write_tree_body, hash_tree): Handle
2532         TS_POLY_INT_CST.
2533         * tree-streamer-in.c (lto_input_ts_poly_tree_pointers): Likewise.
2534         (streamer_read_tree_body): Likewise.
2535         * tree-streamer-out.c (write_ts_poly_tree_pointers): Likewise.
2536         (streamer_write_tree_body): Likewise.
2537         * tree-streamer.c (streamer_check_handled_ts_structures): Likewise.
2538         * asan.c (asan_protect_global): Require the size to be an INTEGER_CST.
2539         * cfgexpand.c (expand_debug_expr): Handle POLY_INT_CST.
2540         * expr.c (expand_expr_real_1, const_vector_from_tree): Likewise.
2541         * gimple-expr.h (is_gimple_constant): Likewise.
2542         * gimplify.c (maybe_with_size_expr): Likewise.
2543         * print-tree.c (print_node): Likewise.
2544         * tree-data-ref.c (data_ref_compare_tree): Likewise.
2545         * tree-pretty-print.c (dump_generic_node): Likewise.
2546         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
2547         * tree-vect-data-refs.c (dr_group_sort_cmp): Likewise.
2548         * tree-vrp.c (compare_values_warnv): Likewise.
2549         * tree-ssa-loop-ivopts.c (determine_base_object, constant_multiple_of)
2550         (get_loop_invariant_expr, add_candidate_1, get_computation_aff_1)
2551         (force_expr_to_var_cost): Likewise.
2552         * tree-ssa-loop.c (for_each_index): Likewise.
2553         * fold-const.h (build_invariant_address, size_int_kind): Take a
2554         poly_int64 instead of a HOST_WIDE_INT.
2555         * fold-const.c (fold_negate_expr_1, const_binop, const_unop)
2556         (fold_convert_const, multiple_of_p, fold_negate_const): Handle
2557         POLY_INT_CST.
2558         (size_binop_loc): Likewise.  Allow int_const_binop_1 to fail.
2559         (int_const_binop_2): New function, split out from...
2560         (int_const_binop_1): ...here.  Handle POLY_INT_CST.
2561         (size_int_kind): Take a poly_int64 instead of a HOST_WIDE_INT.
2562         * expmed.c (make_tree): Handle CONST_POLY_INT_P.
2563         * gimple-ssa-strength-reduction.c (slsr_process_add)
2564         (slsr_process_mul): Check for INTEGER_CSTs before using them
2565         as candidates.
2566         * stor-layout.c (bits_from_bytes): New function.
2567         (bit_from_pos): Use it.
2568         (layout_type): Likewise.  For vectors, multiply the TYPE_SIZE_UNIT
2569         by BITS_PER_UNIT to get the TYPE_SIZE.
2570         * tree-cfg.c (verify_expr, verify_types_in_gimple_reference): Allow
2571         MEM_REF and TARGET_MEM_REF offsets to be a POLY_INT_CST.
2573 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2574             Alan Hayward  <alan.hayward@arm.com>
2575             David Sherwood  <david.sherwood@arm.com>
2577         * doc/rtl.texi (const_poly_int): Document.  Also document the
2578         rtl sharing behavior.
2579         * gengenrtl.c (excluded_rtx): Return true for CONST_POLY_INT.
2580         * rtl.h (const_poly_int_def): New struct.
2581         (rtx_def::u): Add a cpi field.
2582         (CASE_CONST_UNIQUE, CASE_CONST_ANY): Add CONST_POLY_INT.
2583         (CONST_POLY_INT_P, CONST_POLY_INT_COEFFS): New macros.
2584         (wi::rtx_to_poly_wide_ref): New typedef
2585         (const_poly_int_value, wi::to_poly_wide, rtx_to_poly_int64)
2586         (poly_int_rtx_p): New functions.
2587         (trunc_int_for_mode): Declare a poly_int64 version.
2588         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2589         (immed_wide_int_const): Take a poly_wide_int_ref rather than
2590         a wide_int_ref.
2591         (strip_offset): Declare.
2592         (strip_offset_and_add): New function.
2593         * rtl.def (CONST_POLY_INT): New rtx code.
2594         * rtl.c (rtx_size): Handle CONST_POLY_INT.
2595         (shared_const_p): Use poly_int_rtx_p.
2596         * emit-rtl.h (gen_int_mode): Take a poly_int64 instead of a
2597         HOST_WIDE_INT.
2598         (gen_int_shift_amount): Likewise.
2599         * emit-rtl.c (const_poly_int_hasher): New class.
2600         (const_poly_int_htab): New variable.
2601         (init_emit_once): Initialize it when NUM_POLY_INT_COEFFS > 1.
2602         (const_poly_int_hasher::hash): New function.
2603         (const_poly_int_hasher::equal): Likewise.
2604         (gen_int_mode): Take a poly_int64 instead of a HOST_WIDE_INT.
2605         (immed_wide_int_const): Rename to...
2606         (immed_wide_int_const_1): ...this and make static.
2607         (immed_wide_int_const): New function, taking a poly_wide_int_ref
2608         instead of a wide_int_ref.
2609         (gen_int_shift_amount): Take a poly_int64 instead of a HOST_WIDE_INT.
2610         (gen_lowpart_common): Handle CONST_POLY_INT.
2611         * cse.c (hash_rtx_cb, equiv_constant): Likewise.
2612         * cselib.c (cselib_hash_rtx): Likewise.
2613         * dwarf2out.c (const_ok_for_output_1): Likewise.
2614         * expr.c (convert_modes): Likewise.
2615         * print-rtl.c (rtx_writer::print_rtx, print_value): Likewise.
2616         * rtlhash.c (add_rtx): Likewise.
2617         * explow.c (trunc_int_for_mode): Add a poly_int64 version.
2618         (plus_constant): Take a poly_int64 instead of a HOST_WIDE_INT.
2619         Handle existing CONST_POLY_INT rtxes.
2620         * expmed.h (expand_shift): Take a poly_int64 instead of a
2621         HOST_WIDE_INT.
2622         * expmed.c (expand_shift): Likewise.
2623         * rtlanal.c (strip_offset): New function.
2624         (commutative_operand_precedence): Give CONST_POLY_INT the same
2625         precedence as CONST_DOUBLE and put CONST_WIDE_INT between that
2626         and CONST_INT.
2627         * rtl-tests.c (const_poly_int_tests): New struct.
2628         (rtl_tests_c_tests): Use it.
2629         * simplify-rtx.c (simplify_const_unary_operation): Handle
2630         CONST_POLY_INT.
2631         (simplify_const_binary_operation): Likewise.
2632         (simplify_binary_operation_1): Fold additions of symbolic constants
2633         and CONST_POLY_INTs.
2634         (simplify_subreg): Handle extensions and truncations of
2635         CONST_POLY_INTs.
2636         (simplify_const_poly_int_tests): New struct.
2637         (simplify_rtx_c_tests): Use it.
2638         * wide-int.h (storage_ref): Add default constructor.
2639         (wide_int_ref_storage): Likewise.
2640         (trailing_wide_ints): Use GTY((user)).
2641         (trailing_wide_ints::operator[]): Add a const version.
2642         (trailing_wide_ints::get_precision): New function.
2643         (trailing_wide_ints::extra_size): Likewise.
2645 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2646             Alan Hayward  <alan.hayward@arm.com>
2647             David Sherwood  <david.sherwood@arm.com>
2649         * emit-rtl.h (gen_int_shift_amount): Declare.
2650         * emit-rtl.c (gen_int_shift_amount): New function.
2651         * asan.c (asan_emit_stack_protection): Use gen_int_shift_amount
2652         instead of GEN_INT.
2653         * calls.c (shift_return_value): Likewise.
2654         * cse.c (fold_rtx): Likewise.
2655         * dse.c (find_shift_sequence): Likewise.
2656         * expmed.c (init_expmed_one_mode, store_bit_field_1, expand_shift_1)
2657         (expand_shift, expand_smod_pow2): Likewise.
2658         * lower-subreg.c (shift_cost): Likewise.
2659         * optabs.c (expand_superword_shift, expand_doubleword_mult)
2660         (expand_unop, expand_binop, shift_amt_for_vec_perm_mask)
2661         (expand_vec_perm_var): Likewise.
2662         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
2663         (simplify_binary_operation_1): Likewise.
2664         * combine.c (try_combine, find_split_point, force_int_to_mode)
2665         (simplify_shift_const_1, simplify_shift_const): Likewise.
2666         (change_zero_ext): Likewise.  Use simplify_gen_binary.
2668 2017-12-20  Richard Sandiford  <richard.sandiford@linaro.org>
2670         * poly-int.h (multiple_p): Fix handling of two non-poly_ints.
2672 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2674         * doc/invoke.texi (ARM Options): Document accepted extension options
2675         for -march=armv8.3-a.
2677 2017-12-20  Richard Earnshaw  <rearnsha@arm.com>
2679         PR target/83105
2680         * config.gcc (arm*-*-linux*): When configured with --with-float=hard
2681         or --with-float=softfp, set the default CPU to arm10e.
2683 2017-12-20  Eric Botcazou  <ebotcazou@adacore.com>
2685         * config/visium/constraints.md (J, K, L): Use IN_RANGE macro.
2686         * config/visium/predicates.md (const_shift_operand): Likewise.
2687         * config/visium/visium.c (visium_legitimize_address): Fix oversight.
2688         (visium_legitimize_reload_address): Likewise.
2690 2017-12-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2692         PR target/82975
2693         * config/arm/arm.h (TEST_REGNO): Adjust comment as expected in r255830.
2695 2017-12-20  Jakub Jelinek  <jakub@redhat.com>
2697         PR c++/83490
2698         * calls.c (compute_argument_addresses): Ignore TYPE_EMPTY_P arguments.
2700 2017-12-20  Julia Koval  <julia.koval@intel.com>
2702         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VPCLMULQDQ_SET,
2703         OPTION_MASK_ISA_VPCLMULQDQ_UNSET): New.
2704         (ix86_handle_option): Handle -mvpclmulqdq, move cx6 to flags2.
2705         * config.gcc: Include vpclmulqdqintrin.h.
2706         * config/i386/cpuid.h: Handle bit_VPCLMULQDQ.
2707         * config/i386/driver-i386.c (host_detect_local_cpu): Handle
2708         -mvpclmulqdq.
2709         * config/i386/i386-builtin.def (__builtin_ia32_vpclmulqdq_v2di,
2710         __builtin_ia32_vpclmulqdq_v4di, __builtin_ia32_vpclmulqdq_v8di): New.
2711         * config/i386/i386-c.c (__VPCLMULQDQ__): New.
2712         * config/i386/i386.c (isa2_opts): Add -mcx16.
2713         (isa_opts): Add -mpclmulqdq, remove -mcx16.
2714         (ix86_option_override_internal): Move mcx16 to flags2.
2715         (ix86_valid_target_attribute_inner_p): Add vpclmulqdq.
2716         (ix86_expand_builtin): Handle OPTION_MASK_ISA_VPCLMULQDQ.
2717         * config/i386/i386.h (TARGET_VPCLMULQDQ, TARGET_VPCLMULQDQ_P): New.
2718         * config/i386/i386.opt: Add mvpclmulqdq, move mcx16 to flags2.
2719         * config/i386/immintrin.h: Include vpclmulqdqintrin.h.
2720         * config/i386/sse.md (vpclmulqdq_<mode>): New pattern.
2721         * config/i386/vpclmulqdqintrin.h (_mm512_clmulepi64_epi128,
2722         _mm_clmulepi64_epi128, _mm256_clmulepi64_epi128): New intrinsics.
2723         * doc/invoke.texi: Add -mvpclmulqdq.
2725 2017-12-20  Tom de Vries  <tom@codesourcery.com>
2727         PR middle-end/83423
2728         * config/i386/i386.c (ix86_static_chain): Move
2729         DECL_STATIC_CHAIN test ...
2730         * calls.c (rtx_for_static_chain): ... here.  New function.
2731         * calls.h (rtx_for_static_chain): Declare.
2732         * builtins.c (expand_builtin_setjmp_receiver): Use rtx_for_static_chain
2733         instead of targetm.calls.static_chain.
2734         * df-scan.c (df_get_entry_block_def_set): Same.
2736 2017-12-19  Tom de Vries  <tom@codesourcery.com>
2738         PR tree-optimization/83493
2739         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for): Unshare ub
2740         and lb.
2742 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2744         * gimple-ssa-sprintf.c (format_directive): Use inform_n instead of
2745         inform with hardcoded english plural handling.
2747 2017-12-18  Jeff Law  <law@redhat.com>
2749         PR tree-optimization/83477
2750         * tree-ssa-threadedge.c (record_temporary_equivalences_from_phis): For
2751         a non-virtual PHI, always push a new range.
2753 2017-12-19  Martin Sebor  <msebor@redhat.com>
2755         PR middle-end/77608
2756         * builtins.c (compute_objsize): Handle non-constant offsets.
2758 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2760         PR tree-optimization/83444
2761         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): For the
2762         character load case, if get_stridx on MEM_REF's operand doesn't
2763         look usable, retry with get_addr_stridx.
2765 2017-12-19  Alexandre Oliva  <aoliva@redhat.com>
2767         PR debug/83422
2768         * var-tracking.c (vt_debug_insns_local): Do not drop markers.
2769         (variable_tracking_main_1): Keep markers even when VTA fails.
2771         PR bootstrap/83396
2772         * cfgexpand.c (expand_gimple_basic_block): Expand label first,
2773         even if there are markers before it.
2774         * cfgrtl.c (rtl_verify_bb_layout): Reject DEBUG_INSNs outside BBs.
2776 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2778         * read-rtl.c (parse_reg_note_name): Replace Yoda conditions with
2779         typical order conditions.
2780         * sel-sched.c (extract_new_fences_from): Likewise.
2781         * config/visium/constraints.md (J, K, L): Likewise.
2782         * config/visium/predicates.md (const_shift_operand): Likewise.
2783         * config/visium/visium.c (visium_legitimize_address,
2784         visium_legitimize_reload_address): Likewise.
2785         * config/m68k/m68k.c (output_reg_adjust, emit_reg_adjust): Likewise.
2786         * config/arm/arm.c (arm_block_move_unaligned_straight): Likewise.
2787         * config/avr/constraints.md (Y01, Ym1, Y02, Ym2): Likewise.
2788         * config/avr/avr-log.c (avr_vdump, avr_log_set_avr_log,
2789         SET_DUMP_DETAIL): Likewise.
2790         * config/avr/predicates.md (const_8_16_24_operand): Likewise.
2791         * config/avr/avr.c (STR_PREFIX_P, avr_popcount_each_byte,
2792         avr_is_casesi_sequence, avr_casei_sequence_check_operands,
2793         avr_set_core_architecture, avr_set_current_function,
2794         avr_legitimize_reload_address, avr_asm_len, avr_print_operand,
2795         output_movqi, output_movsisf, avr_out_plus, avr_out_bitop,
2796         avr_out_fract, avr_adjust_insn_length, avr_encode_section_info,
2797         avr_2word_insn_p, output_reload_in_const, avr_has_nibble_0xf,
2798         avr_map_decompose, avr_fold_builtin): Likewise.
2799         * config/avr/driver-avr.c (avr_devicespecs_file): Likewise.
2800         * config/avr/gen-avr-mmcu-specs.c (str_prefix_p, print_mcu): Likewise.
2801         * config/i386/i386.c (ix86_parse_stringop_strategy_string): Likewise.
2802         * config/m32c/m32c-pragma.c (m32c_pragma_memregs): Likewise.
2803         * config/m32c/m32c.c (m32c_conditional_register_usage,
2804         m32c_address_cost): Likewise.
2805         * config/m32c/predicates.md (shiftcount_operand,
2806         longshiftcount_operand): Likewise.
2807         * config/iq2000/iq2000.c (iq2000_expand_prologue): Likewise.
2808         * config/nios2/nios2.c (nios2_handle_custom_fpu_insn_option,
2809         can_use_cdx_ldstw): Likewise.
2810         * config/nios2/nios2.h (CDX_REG_P): Likewise.
2811         * config/cr16/cr16.h (RETURN_ADDR_RTX, REGNO_MODE_OK_FOR_BASE_P):
2812         Likewise.
2813         * config/cr16/cr16.md (*mov<mode>_double): Likewise.
2814         * config/cr16/cr16.c (cr16_create_dwarf_for_multi_push): Likewise.
2815         * config/h8300/h8300.c (h8300_rtx_costs, get_shift_alg): Likewise.
2816         * config/vax/constraints.md (U06, U08, U16, CN6, S08, S16): Likewise.
2817         * config/vax/vax.c (adjacent_operands_p): Likewise.
2818         * config/ft32/constraints.md (L, b, KA): Likewise.
2819         * config/ft32/ft32.c (ft32_load_immediate, ft32_expand_prologue):
2820         Likewise.
2821         * cfgexpand.c (expand_stack_alignment): Likewise.
2822         * gcse.c (insert_expr_in_table): Likewise.
2823         * print-rtl.c (rtx_writer::print_rtx_operand_codes_E_and_V): Likewise.
2824         * cgraphunit.c (cgraph_node::expand): Likewise.
2825         * ira-build.c (setup_min_max_allocno_live_range_point): Likewise.
2826         * emit-rtl.c (add_insn): Likewise.
2827         * input.c (dump_location_info): Likewise.
2828         * passes.c (NEXT_PASS): Likewise.
2829         * read-rtl-function.c (parse_note_insn_name,
2830         function_reader::read_rtx_operand_r, function_reader::parse_mem_expr):
2831         Likewise.
2832         * sched-rgn.c (sched_rgn_init): Likewise.
2833         * diagnostic-show-locus.c (layout::show_ruler): Likewise.
2834         * combine.c (find_split_point, simplify_if_then_else, force_to_mode,
2835         if_then_else_cond, simplify_shift_const_1, simplify_comparison):
2836         Likewise.
2837         * explow.c (eliminate_constant_term): Likewise.
2838         * final.c (leaf_renumber_regs_insn): Likewise.
2839         * cfgrtl.c (print_rtl_with_bb): Likewise.
2840         * genhooks.c (emit_init_macros): Likewise.
2841         * poly-int.h (maybe_ne, maybe_le, maybe_lt): Likewise.
2842         * tree-data-ref.c (conflict_fn): Likewise.
2843         * selftest.c (assert_streq): Likewise.
2844         * expr.c (store_constructor_field, expand_expr_real_1): Likewise.
2845         * fold-const.c (fold_range_test, extract_muldiv_1, fold_truth_andor,
2846         fold_binary_loc, multiple_of_p): Likewise.
2847         * reload.c (push_reload, find_equiv_reg): Likewise.
2848         * et-forest.c (et_nca, et_below): Likewise.
2849         * dbxout.c (dbxout_symbol_location): Likewise.
2850         * reorg.c (relax_delay_slots): Likewise.
2851         * dojump.c (do_compare_rtx_and_jump): Likewise.
2852         * gengtype-parse.c (type): Likewise.
2853         * simplify-rtx.c (simplify_gen_ternary, simplify_gen_relational,
2854         simplify_const_relational_operation): Likewise.
2855         * reload1.c (do_output_reload): Likewise.
2856         * dumpfile.c (get_dump_file_info_by_switch): Likewise.
2857         * gengtype.c (type_for_name): Likewise.
2858         * gimple-ssa-sprintf.c (format_directive): Likewise.
2860 2017-12-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
2862         PR target/82975
2863         * config/arm/arm.h (TEST_REGNO): Check reg_renumber is set before
2864         accessing it.  Adjust comment.
2866 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2868         PR middle-end/81914
2869         * predict.c (zero_one_minusone): New function.
2870         (apply_return_prediction): Avoid return prediction for functions
2871         returning only -1, 0 and 1 values, unless they only return -1 and 0
2872         or 0 and 1.
2874 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2876         * config/arc/arc.c (legitimate_scaled_address_p): Clean
2877         fall-through warning.
2878         (arc_compute_frame_size): Remove unused variables.
2879         (arc_print_operand): Fix fprintif format.
2880         (arc_can_follow_jump): Clean fall-through warning.
2882 2017-12-19  Marek Polacek  <polacek@redhat.com>
2884         PR c++/83489
2885         * config/i386/i386.c (init_cumulative_args): Don't check TYPE_EMPTY_P
2886         on an error node.
2888 2017-12-19  Claudiu Zissulescu  <claziss@synopsys.com>
2890         * config/arc/arc.c (overriderregs): New variable.
2891         (arc_override_options): Track fixed/call saved/call options.
2892         (arc_conditional_register_usage): Check against overrideregs
2893         variable whenever we change register properties.
2895 2017-12-19  Nathan Sidwell  <nathan@acm.org>
2897         * opts.c (finish_options): Don't prefix dump_base_name if it
2898         already contains directories.
2900 2017-12-19  Martin Liska  <mliska@suse.cz>
2902         PR rtl-optimization/82675
2903         * loop-unroll.c (unroll_loop_constant_iterations): Allocate one
2904         more element in sbitmap.
2906 2017-12-19  Paolo Carlini  <paolo.carlini@oracle.com>
2908         * gimplify.c (gimplify_expr): Use error_operand_p.
2910 2017-12-19  Sebastian Huber  <sebastian.huber@embedded-brains.de>
2912         PR target/83387
2913         * config/rs6000/t-rtems (MULTILIB_REQUIRED): Remove 64-bit soft-float
2914         multilib.
2916 2017-12-19  Daniel Cederman  <cederman@gaisler.com>
2918         * config/sparc/sparc.c (sparc_do_work_around_errata): Make sure
2919         the jump is to a label.
2921 2017-12-19  Jakub Jelinek  <jakub@redhat.com>
2923         PR tree-optimization/83444
2924         * tree-ssa-strlen.c (strlen_check_and_optimize_stmt): Optimize
2925         character loads.
2927         PR ipa/82801
2928         PR ipa/83346
2929         * ipa-inline.c (flatten_remove_node_hook): New function.
2930         (ipa_inline): Keep only nodes with flatten attribute at the end of
2931         the array in the order from ipa_reverse_postorder, only walk that
2932         portion of array for flattening, if there is more than one such
2933         node, temporarily register a removal hook and ignore removed nodes.
2935         PR tree-optimization/80631
2936         * tree-vect-loop.c (vect_create_epilog_for_reduction): Compare
2937         induc_code against MAX_EXPR or MIN_EXPR instead of reduc_fn against
2938         IFN_REDUC_MAX or IFN_REDUC_MIN.
2940 2017-12-18  Jeff Law  <law@redhat.com>
2942         * tree-ssa-dom.c (record_equivalences_from_phis): Do not
2943         record symbolic equivalences from backedges in the CFG.
2945         Revert
2946         2017-11-19  Jeff Law  <law@redhat.com>
2948         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
2949         of degenerates resulting from ignoring an edge.
2951 2017-12-18  Martin Sebor  <msebor@redhat.com>
2953         PR middle-end/83373
2954         PR tree-optimization/78450
2955         * tree-ssa-strlen.c (maybe_set_strlen_range): New function.
2956         (handle_builtin_strlen): Call it.
2958 2017-12-18  Segher Boessenkool  <segher@kernel.crashing.org>
2960         PR rtl-optimization/83424
2961         * rtlanal.c (dead_or_set_regno_p): Handle CLOBBER just like SET.
2963 2017-12-18  Marek Polacek  <polacek@redhat.com>
2965         PR middle-end/83463
2966         * gimple-ssa-warn-restrict.c (builtin_memref::builtin_memref):
2967         Check if TYPE is INTEGRAL_TYPE_P before accessing its min/max
2968         values.
2970 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
2972         * config/arc/arc.md (maddsidi4, maddsidi4_split): Update pattern.
2973         (umaddsidi4, umaddsidi_split): Likewise.
2975 2017-12-18  Claudiu Zissulescu  <claziss@synopsys.com>
2977         * config/arc/arc.c (arc_legitimate_constant_p): Always check all
2978         constants.
2980 2017-12-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
2982         PR target/83420
2983         * config/s390/s390.c (s390_option_override): Avoid strncpy.
2985 2017-12-18  Richard Biener  <rguenther@suse.de>
2987         PR tree-optimization/81877
2988         * tree-ssa-loop-im.c (ref_indep_loop_p): Remove safelen parameters.
2989         (outermost_indep_loop): Adjust.
2990         (ref_indep_loop_p_1): Likewise.  Remove safelen handling again.
2991         (can_sm_ref_p): Adjust.
2993 2017-12-18  Richard Biener  <rguenther@suse.de>
2995         PR middle-end/77291
2996         * tree.c (array_at_struct_end_p): Return true if the underlying
2997         object has space for at least one element in excess of what
2998         the array domain specifies.
3000 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3002         * doc/extend.texi (x86 Function Attributes): Reformat nocf_check
3003         example to avoid overfull hbox.
3004         * doc/invoke.texi (Option Summary): Add missing @gol.
3005         (C++ Dialect Options): Reformat -Wnoexcept-type example to avoid
3006         overfull hbox.
3008 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3009             Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3011         * doc/invoke.texi (Option Summary): Add -mverbose-cost-dump
3012         to AArch64 and ARM lists, plus missing -mflip-thumb for ARM.
3013         (AArch64 Options): Document -mverbose-cost-dump.
3014         (ARM Options): Likewise, plus -mflip-thumb.
3016 2017-12-17  Martin Sebor  <msebor@redhat.com>
3018         PR bootstrap/83446
3019         * gimple-ssa-warn-restrict.c
3020         (builtin_memref::offset_out_of_bounds): Correct the handling of
3021         anti-ranges.
3023 2017-12-17  Sandra Loosemore  <sandra@codesourcery.com>
3025         * doc/invoke.texi (ARC Options): Add missing -mlra entry.
3027 2017-12-17  John David Anglin  <danglin@gcc.gnu.org>
3029         * config/pa/pa.c (pa_som_asm_init_sections): Fix comment.
3031 2017-12-17  Markus Trippelsdorf  <markus@trippelsdorf.de>
3033         * x86-tune-costs.h (skylake_cost, core_cost): Decrease r64 multiply
3034         latencies.
3036 2017-12-16  Sandra Loosemore  <sandra@codesourcery.com>
3038         * doc/invoke.texi: Fix some typos.
3040 2017-12-16  Martin Sebor  <msebor@redhat.com>
3042         PR tree-optimization/78918
3043         * Makefile.in (OBJS): Add gimple-ssa-warn-restrict.o.
3044         * builtins.c (check_sizes): Rename...
3045         (check_access): ...to this.  Rename function arguments for clarity.
3046         (check_memop_sizes): Adjust names.
3047         (expand_builtin_memchr, expand_builtin_memcpy): Same.
3048         (expand_builtin_memmove, expand_builtin_mempcpy): Same.
3049         (expand_builtin_strcat, expand_builtin_stpncpy): Same.
3050         (check_strncat_sizes, expand_builtin_strncat): Same.
3051         (expand_builtin_strncpy, expand_builtin_memset): Same.
3052         (expand_builtin_bzero, expand_builtin_memcmp): Same.
3053         (expand_builtin_memory_chk, maybe_emit_chk_warning): Same.
3054         (maybe_emit_sprintf_chk_warning): Same.
3055         (expand_builtin_strcpy): Adjust.
3056         (expand_builtin_stpcpy): Same.
3057         (expand_builtin_with_bounds): Detect out-of-bounds accesses
3058         in pointer-checking forms of memcpy, memmove, and mempcpy.
3059         (gcall_to_tree_minimal, max_object_size): Define new functions.
3060         * builtins.h (max_object_size): Declare.
3061         * calls.c (alloc_max_size): Call max_object_size instead of
3062         hardcoding ssizetype limit.
3063         (get_size_range): Handle new argument.
3064         * calls.h (get_size_range): Add a new argument.
3065         * cfgexpand.c (expand_call_stmt): Propagate no-warning bit.
3066         * doc/invoke.texi (-Wrestrict): Adjust, add example.
3067         * gimple-fold.c (gimple_fold_builtin_memory_op): Detect overlapping
3068         operations.
3069         (gimple_fold_builtin_memory_chk): Same.
3070         (gimple_fold_builtin_stxcpy_chk): New function.
3071         * gimple-ssa-warn-restrict.c: New source.
3072         * gimple-ssa-warn-restrict.h: New header.
3073         * gimple.c (gimple_build_call_from_tree): Propagate location.
3074         * passes.def (pass_warn_restrict): Add new pass.
3075         * tree-pass.h (make_pass_warn_restrict): Declare.
3076         * tree-ssa-strlen.c (handle_builtin_strcpy): Detect overlapping
3077         operations.
3078         (handle_builtin_strcat): Same.
3079         (strlen_optimize_stmt): Rename...
3080         (strlen_check_and_optimize_stmt): ...to this.  Handle strncat,
3081         stpncpy, strncpy, and their checking forms.
3083 2017-12-16  Jan Hubicka  <hubicka@ucw.cz>
3085         PR rtl-optimization/82849
3086         * modulo-sched.c (sms_schedule): Use get_estimated_loop_iterations_int
3087         and get_max_loop_iterations_int.
3089 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3090             Alan Hayward  <alan.hayward@arm.com>
3091             David Sherwood  <david.sherwood@arm.com>
3093         * machmode.h (mode_for_size, int_mode_for_size, float_mode_for_size)
3094         (smallest_mode_for_size, smallest_int_mode_for_size): Take the mode
3095         size as a poly_uint64.
3096         (mode_for_vector, mode_for_int_vector): Take the number of vector
3097         elements as a poly_uint64.
3098         * stor-layout.c (mode_for_size, smallest_mode_for_size): Take the mode
3099         size as a poly_uint64.
3100         (mode_for_vector, mode_for_int_vector): Take the number of vector
3101         elements as a poly_uint64.
3103 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3104             Alan Hayward  <alan.hayward@arm.com>
3105             David Sherwood  <david.sherwood@arm.com>
3107         * machmode.h (MACRO_MODE): New macro.
3108         * addresses.h (base_reg_class, ok_for_base_p_1): Use it.
3109         * targhooks.c (default_libcall_value, default_secondary_reload)
3110         (default_memory_move_cost, default_register_move_cost)
3111         (default_class_max_nregs): Likewise.
3113 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3114             Alan Hayward  <alan.hayward@arm.com>
3115             David Sherwood  <david.sherwood@arm.com>
3117         * doc/sourcebuild.texi: Document IN_TARGET_CODE.
3118         * genattrtab.c (write_header): Define IN_TARGET_CODE to 1 in the
3119         target C file.
3120         * genautomata.c (main): Likewise.
3121         * genconditions.c (write_header): Likewise.
3122         * genemit.c (main): Likewise.
3123         * genextract.c (print_header): Likewise.
3124         * genopinit.c (main): Likewise.
3125         * genoutput.c (output_prologue): Likewise.
3126         * genpeep.c (main): Likewise.
3127         * genpreds.c (write_insn_preds_c): Likewise.
3128         * genrecog.c (writer_header): Likewise.
3129         * config/aarch64/aarch64-builtins.c (IN_TARGET_CODE): Define.
3130         * config/aarch64/aarch64-c.c (IN_TARGET_CODE): Likewise.
3131         * config/aarch64/aarch64.c (IN_TARGET_CODE): Likewise.
3132         * config/aarch64/cortex-a57-fma-steering.c (IN_TARGET_CODE): Likewise.
3133         * config/aarch64/driver-aarch64.c (IN_TARGET_CODE): Likewise.
3134         * config/alpha/alpha.c (IN_TARGET_CODE): Likewise.
3135         * config/alpha/driver-alpha.c (IN_TARGET_CODE): Likewise.
3136         * config/arc/arc-c.c (IN_TARGET_CODE): Likewise.
3137         * config/arc/arc.c (IN_TARGET_CODE): Likewise.
3138         * config/arc/driver-arc.c (IN_TARGET_CODE): Likewise.
3139         * config/arm/aarch-common.c (IN_TARGET_CODE): Likewise.
3140         * config/arm/arm-builtins.c (IN_TARGET_CODE): Likewise.
3141         * config/arm/arm-c.c (IN_TARGET_CODE): Likewise.
3142         * config/arm/arm.c (IN_TARGET_CODE): Likewise.
3143         * config/arm/driver-arm.c (IN_TARGET_CODE): Likewise.
3144         * config/avr/avr-c.c (IN_TARGET_CODE): Likewise.
3145         * config/avr/avr-devices.c (IN_TARGET_CODE): Likewise.
3146         * config/avr/avr-log.c (IN_TARGET_CODE): Likewise.
3147         * config/avr/avr.c (IN_TARGET_CODE): Likewise.
3148         * config/avr/driver-avr.c (IN_TARGET_CODE): Likewise.
3149         * config/avr/gen-avr-mmcu-specs.c (IN_TARGET_CODE): Likewise.
3150         * config/bfin/bfin.c (IN_TARGET_CODE): Likewise.
3151         * config/c6x/c6x.c (IN_TARGET_CODE): Likewise.
3152         * config/cr16/cr16.c (IN_TARGET_CODE): Likewise.
3153         * config/cris/cris.c (IN_TARGET_CODE): Likewise.
3154         * config/darwin.c (IN_TARGET_CODE): Likewise.
3155         * config/epiphany/epiphany.c (IN_TARGET_CODE): Likewise.
3156         * config/epiphany/mode-switch-use.c (IN_TARGET_CODE): Likewise.
3157         * config/epiphany/resolve-sw-modes.c (IN_TARGET_CODE): Likewise.
3158         * config/fr30/fr30.c (IN_TARGET_CODE): Likewise.
3159         * config/frv/frv.c (IN_TARGET_CODE): Likewise.
3160         * config/ft32/ft32.c (IN_TARGET_CODE): Likewise.
3161         * config/h8300/h8300.c (IN_TARGET_CODE): Likewise.
3162         * config/i386/djgpp.c (IN_TARGET_CODE): Likewise.
3163         * config/i386/driver-i386.c (IN_TARGET_CODE): Likewise.
3164         * config/i386/driver-mingw32.c (IN_TARGET_CODE): Likewise.
3165         * config/i386/host-cygwin.c (IN_TARGET_CODE): Likewise.
3166         * config/i386/host-i386-darwin.c (IN_TARGET_CODE): Likewise.
3167         * config/i386/host-mingw32.c (IN_TARGET_CODE): Likewise.
3168         * config/i386/i386-c.c (IN_TARGET_CODE): Likewise.
3169         * config/i386/i386.c (IN_TARGET_CODE): Likewise.
3170         * config/i386/intelmic-mkoffload.c (IN_TARGET_CODE): Likewise.
3171         * config/i386/msformat-c.c (IN_TARGET_CODE): Likewise.
3172         * config/i386/winnt-cxx.c (IN_TARGET_CODE): Likewise.
3173         * config/i386/winnt-stubs.c (IN_TARGET_CODE): Likewise.
3174         * config/i386/winnt.c (IN_TARGET_CODE): Likewise.
3175         * config/i386/x86-tune-sched-atom.c (IN_TARGET_CODE): Likewise.
3176         * config/i386/x86-tune-sched-bd.c (IN_TARGET_CODE): Likewise.
3177         * config/i386/x86-tune-sched-core.c (IN_TARGET_CODE): Likewise.
3178         * config/i386/x86-tune-sched.c (IN_TARGET_CODE): Likewise.
3179         * config/ia64/ia64-c.c (IN_TARGET_CODE): Likewise.
3180         * config/ia64/ia64.c (IN_TARGET_CODE): Likewise.
3181         * config/iq2000/iq2000.c (IN_TARGET_CODE): Likewise.
3182         * config/lm32/lm32.c (IN_TARGET_CODE): Likewise.
3183         * config/m32c/m32c-pragma.c (IN_TARGET_CODE): Likewise.
3184         * config/m32c/m32c.c (IN_TARGET_CODE): Likewise.
3185         * config/m32r/m32r.c (IN_TARGET_CODE): Likewise.
3186         * config/m68k/m68k.c (IN_TARGET_CODE): Likewise.
3187         * config/mcore/mcore.c (IN_TARGET_CODE): Likewise.
3188         * config/microblaze/microblaze-c.c (IN_TARGET_CODE): Likewise.
3189         * config/microblaze/microblaze.c (IN_TARGET_CODE): Likewise.
3190         * config/mips/driver-native.c (IN_TARGET_CODE): Likewise.
3191         * config/mips/frame-header-opt.c (IN_TARGET_CODE): Likewise.
3192         * config/mips/mips.c (IN_TARGET_CODE): Likewise.
3193         * config/mmix/mmix.c (IN_TARGET_CODE): Likewise.
3194         * config/mn10300/mn10300.c (IN_TARGET_CODE): Likewise.
3195         * config/moxie/moxie.c (IN_TARGET_CODE): Likewise.
3196         * config/msp430/driver-msp430.c (IN_TARGET_CODE): Likewise.
3197         * config/msp430/msp430-c.c (IN_TARGET_CODE): Likewise.
3198         * config/msp430/msp430.c (IN_TARGET_CODE): Likewise.
3199         * config/nds32/nds32-cost.c (IN_TARGET_CODE): Likewise.
3200         * config/nds32/nds32-fp-as-gp.c (IN_TARGET_CODE): Likewise.
3201         * config/nds32/nds32-intrinsic.c (IN_TARGET_CODE): Likewise.
3202         * config/nds32/nds32-isr.c (IN_TARGET_CODE): Likewise.
3203         * config/nds32/nds32-md-auxiliary.c (IN_TARGET_CODE): Likewise.
3204         * config/nds32/nds32-memory-manipulation.c (IN_TARGET_CODE): Likewise.
3205         * config/nds32/nds32-pipelines-auxiliary.c (IN_TARGET_CODE): Likewise.
3206         * config/nds32/nds32-predicates.c (IN_TARGET_CODE): Likewise.
3207         * config/nds32/nds32.c (IN_TARGET_CODE): Likewise.
3208         * config/nios2/nios2.c (IN_TARGET_CODE): Likewise.
3209         * config/nvptx/mkoffload.c (IN_TARGET_CODE): Likewise.
3210         * config/nvptx/nvptx.c (IN_TARGET_CODE): Likewise.
3211         * config/pa/pa.c (IN_TARGET_CODE): Likewise.
3212         * config/pdp11/pdp11.c (IN_TARGET_CODE): Likewise.
3213         * config/powerpcspe/driver-powerpcspe.c (IN_TARGET_CODE): Likewise.
3214         * config/powerpcspe/host-darwin.c (IN_TARGET_CODE): Likewise.
3215         * config/powerpcspe/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3216         * config/powerpcspe/powerpcspe-c.c (IN_TARGET_CODE): Likewise.
3217         * config/powerpcspe/powerpcspe-linux.c (IN_TARGET_CODE): Likewise.
3218         * config/powerpcspe/powerpcspe.c (IN_TARGET_CODE): Likewise.
3219         * config/riscv/riscv-builtins.c (IN_TARGET_CODE): Likewise.
3220         * config/riscv/riscv-c.c (IN_TARGET_CODE): Likewise.
3221         * config/riscv/riscv.c (IN_TARGET_CODE): Likewise.
3222         * config/rl78/rl78-c.c (IN_TARGET_CODE): Likewise.
3223         * config/rl78/rl78.c (IN_TARGET_CODE): Likewise.
3224         * config/rs6000/driver-rs6000.c (IN_TARGET_CODE): Likewise.
3225         * config/rs6000/host-darwin.c (IN_TARGET_CODE): Likewise.
3226         * config/rs6000/host-ppc64-darwin.c (IN_TARGET_CODE): Likewise.
3227         * config/rs6000/rs6000-c.c (IN_TARGET_CODE): Likewise.
3228         * config/rs6000/rs6000-linux.c (IN_TARGET_CODE): Likewise.
3229         * config/rs6000/rs6000-p8swap.c (IN_TARGET_CODE): Likewise.
3230         * config/rs6000/rs6000-string.c (IN_TARGET_CODE): Likewise.
3231         * config/rs6000/rs6000.c (IN_TARGET_CODE): Likewise.
3232         * config/rx/rx.c (IN_TARGET_CODE): Likewise.
3233         * config/s390/driver-native.c (IN_TARGET_CODE): Likewise.
3234         * config/s390/s390-c.c (IN_TARGET_CODE): Likewise.
3235         * config/s390/s390.c (IN_TARGET_CODE): Likewise.
3236         * config/sh/sh-c.c (IN_TARGET_CODE): Likewise.
3237         * config/sh/sh-mem.cc (IN_TARGET_CODE): Likewise.
3238         * config/sh/sh.c (IN_TARGET_CODE): Likewise.
3239         * config/sh/sh_optimize_sett_clrt.cc (IN_TARGET_CODE): Likewise.
3240         * config/sh/sh_treg_combine.cc (IN_TARGET_CODE): Likewise.
3241         * config/sparc/driver-sparc.c (IN_TARGET_CODE): Likewise.
3242         * config/sparc/sparc-c.c (IN_TARGET_CODE): Likewise.
3243         * config/sparc/sparc.c (IN_TARGET_CODE): Likewise.
3244         * config/spu/spu-c.c (IN_TARGET_CODE): Likewise.
3245         * config/spu/spu.c (IN_TARGET_CODE): Likewise.
3246         * config/stormy16/stormy16.c (IN_TARGET_CODE): Likewise.
3247         * config/tilegx/mul-tables.c (IN_TARGET_CODE): Likewise.
3248         * config/tilegx/tilegx-c.c (IN_TARGET_CODE): Likewise.
3249         * config/tilegx/tilegx.c (IN_TARGET_CODE): Likewise.
3250         * config/tilepro/mul-tables.c (IN_TARGET_CODE): Likewise.
3251         * config/tilepro/tilepro-c.c (IN_TARGET_CODE): Likewise.
3252         * config/tilepro/tilepro.c (IN_TARGET_CODE): Likewise.
3253         * config/v850/v850-c.c (IN_TARGET_CODE): Likewise.
3254         * config/v850/v850.c (IN_TARGET_CODE): Likewise.
3255         * config/vax/vax.c (IN_TARGET_CODE): Likewise.
3256         * config/visium/visium.c (IN_TARGET_CODE): Likewise.
3257         * config/vms/vms-c.c (IN_TARGET_CODE): Likewise.
3258         * config/vms/vms-f.c (IN_TARGET_CODE): Likewise.
3259         * config/vms/vms.c (IN_TARGET_CODE): Likewise.
3260         * config/xtensa/xtensa.c (IN_TARGET_CODE): Likewise.
3262 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3263             Alan Hayward  <alan.hayward@arm.com>
3264             David Sherwood  <david.sherwood@arm.com>
3266         * tree-ssa-sccvn.c (vn_reference_lookup_3): Avoid repeated
3267         checks for MEM_REF.
3269 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3270             Alan Hayward  <alan.hayward@arm.com>
3271             David Sherwood  <david.sherwood@arm.com>
3273         * doc/generic.texi (VEC_SERIES_EXPR): Document.
3274         * doc/md.texi (vec_series@var{m}): Document.
3275         * tree.def (VEC_SERIES_EXPR): New tree code.
3276         * tree.h (build_vec_series): Declare.
3277         * tree.c (build_vec_series): New function.
3278         * cfgexpand.c (expand_debug_expr): Handle VEC_SERIES_EXPR.
3279         * tree-pretty-print.c (dump_generic_node): Likewise.
3280         * gimple-pretty-print.c (dump_binary_rhs): Likewise.
3281         * tree-inline.c (estimate_operator_cost): Likewise.
3282         * expr.c (expand_expr_real_2): Likewise.
3283         * optabs-tree.c (optab_for_tree_code): Likewise.
3284         * tree-cfg.c (verify_gimple_assign_binary): Likewise.
3285         * fold-const.c (const_binop): Fold VEC_SERIES_EXPRs of constants.
3286         * expmed.c (make_tree): Handle VEC_SERIES.
3287         * optabs.def (vec_series_optab): New optab.
3288         * optabs.h (expand_vec_series_expr): Declare.
3289         * optabs.c (expand_vec_series_expr): New function.
3290         * tree-vect-generic.c (expand_vector_operations_1): Check that
3291         the operands also have vector type.
3293 2017-12-16  Richard Sandiford  <richard.sandiford@linaro.org>
3294             Alan Hayward  <alan.hawyard@arm.com>
3295             David Sherwood  <david.sherwood@arm.com>
3297         * doc/generic.texi (VEC_DUPLICATE_EXPR): Document.
3298         (VEC_COND_EXPR): Add missing @tindex.
3299         * doc/md.texi (vec_duplicate@var{m}): Document.
3300         * tree.def (VEC_DUPLICATE_EXPR): New tree codes.
3301         * tree.c (build_vector_from_val): Add stubbed-out handling of
3302         variable-length vectors, using VEC_DUPLICATE_EXPR.
3303         (uniform_vector_p): Handle VEC_DUPLICATE_EXPR.
3304         * cfgexpand.c (expand_debug_expr): Likewise.
3305         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
3306         * tree-inline.c (estimate_operator_cost): Likewise.
3307         * tree-pretty-print.c (dump_generic_node): Likewise.
3308         * tree-vect-generic.c (ssa_uniform_vector_p): Likewise.
3309         * fold-const.c (const_unop): Fold VEC_DUPLICATE_EXPRs of a constant.
3310         (test_vec_duplicate_folding): New function.
3311         (fold_const_c_tests): Call it.
3312         * optabs.def (vec_duplicate_optab): New optab.
3313         * optabs-tree.c (optab_for_tree_code): Handle VEC_DUPLICATE_EXPR.
3314         * optabs.h (expand_vector_broadcast): Declare.
3315         * optabs.c (expand_vector_broadcast): Make non-static.  Try using
3316         vec_duplicate_optab.
3317         * expr.c (store_constructor): Try using vec_duplicate_optab for
3318         uniform vectors.
3319         (expand_expr_real_2): Handle VEC_DUPLICATE_EXPR.
3321 2017-12-15  Markus Trippelsdorf  <markus@trippelsdorf.de>
3323         PR target/83358
3324         * config/i386/x86-tune-costs.h (skylake_cost, core_cost): Increase
3325         div/mod latencies a bit.
3327 2017-12-15  Jeff Law  <law@redhat.com>
3329         PR tree-optimization/36550
3330         * tree-ssa-threadupdate.c (count_stmts_and_phis_in_block): New.
3331         (mark_threaded_blocks): Rewrite code to avoid block copying when
3332         optimizing for size.  Don't pessimize blocks which will be
3333         copied, but all the statements will be dead.
3335 2017-12-15  Alexandre Oliva  <aoliva@redhat.com>
3337         PR tree-optimization/81165
3338         * tree-ssa-threadupdate.c (uses_in_bb): New.
3339         (estimate_threading_killed_stmts): New.
3340         * tree-ssa-threadupdate.h (estimate_threading_killed_stmts): Prototype.
3341         * tree-ssa-threadedge.c
3342         (record_temporary_equivalences_from_stmts_at_dest): Expand limit
3343         when its hit.
3345 2017-12-15  Jeff Law  <law@redhat.com>
3347         PR tree-optimization/83410
3348         * tree-ssa-threadupdate.c (thread_block_1): Avoid certain jump
3349         threads when parallelizing loops.
3351 2017-12-15  Jakub Jelinek  <jakub@redhat.com>
3353         * tree-core.h (struct attribute_spec): Swap affects_type_identity and
3354         handler fields.
3355         * config/alpha/alpha.c (vms_attribute_table): Swap
3356         affects_type_identity and handler fields, adjust comments.
3357         * config/mips/mips.c (mips_attribute_table): Likewise.
3358         * config/visium/visium.c (visium_attribute_table): Likewise.
3359         * config/epiphany/epiphany.c (epiphany_attribute_table): Likewise.
3360         * config/microblaze/microblaze.c (microblaze_attribute_table):
3361         Likewise.
3362         * config/spu/spu.c (spu_attribute_table): Likewise.
3363         * config/mcore/mcore.c (mcore_attribute_table): Likewise.
3364         * config/arc/arc.c (arc_attribute_table): Likewise.
3365         * config/m68k/m68k.c (m68k_attribute_table): Likewise.
3366         * config/v850/v850.c (v850_handle_interrupt_attribute,
3367         v850_handle_data_area_attribute): Formatting fixes.
3368         (v850_attribute_table): Swap affects_type_identity and handler
3369         fields, adjust comments.
3370         * config/m32r/m32r.c (m32r_attribute_table): Likewise.
3371         * config/arm/arm.c (arm_attribute_table): Likewise.
3372         * config/avr/avr.c (avr_attribute_table): Likewise.
3373         * config/s390/s390.c (s390_attribute_table): Likewise.
3374         * config/sh/sh.c (sh_attribute_table): Likewise.
3375         * config/i386/i386.c (ix86_handle_cconv_attribute,
3376         ix86_handle_callee_pop_aggregate_return): Formatting fixes.
3377         (ix86_attribute_table): Swap affects_type_identity and handler
3378         fields, adjust comments.
3379         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3380         * config/sparc/sparc.c (sparc_attribute_table): Likewise.
3381         * config/m32c/m32c.c (m32c_attribute_table): Likewise.
3382         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Likewise.
3383         * config/ia64/ia64.c (ia64_attribute_table): Likewise.
3384         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
3385         * config/rx/rx.c (rx_attribute_table): Likewise.
3386         * config/cr16/cr16.c (cr16_attribute_table): Likewise.
3387         * config/h8300/h8300.c (h8300_attribute_table): Likewise.
3388         * config/nvptx/nvptx.c (nvptx_attribute_table): Likewise.
3389         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Likewise.
3390         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
3391         * config/stormy16/stormy16.c (xstormy16_attribute_table): Likewise.
3392         * config/bfin/bfin.c (bfin_attribute_table): Likewise.
3393         * config/rs6000/rs6000.c (rs6000_attribute_table): Likewise.
3394         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
3395         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
3396         * doc/plugins.texi (user_attr): Likewise.  Add NULL for
3397         exclude.
3398         * attribs.c (empty_attribute_table): Swap affects_type_identity and
3399         handler fields.
3400         (register_scoped_attributes, decl_attributes): Formatting fixes.
3402         PR tree-optimization/83269
3403         * fold-const.c (fold_binary_loc): Perform (-A) - B -> (-B) - A
3404         subtraction in arg0's type if type is signed and arg0 is unsigned.
3405         Formatting fix.
3407         PR sanitizer/81281
3408         * match.pd ((T)(P + A) - (T)P -> (T) A): Use @@0 instead of @0 and
3409         convert? on @0 instead of convert.  Check type of @1, not @0.
3410         ((T)P - (T)(P + A) -> -(T) A): Use @@0 instead of @0 and
3411         convert? on @0 instead of convert.  Check type of @1, not @0.
3412         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Use @@0 instead of @0,
3413         only optimize if either both @1 and @2 types are narrower
3414         precision, or both are wider or equal precision, and in the former
3415         case only if both have undefined overflow.
3417 2017-12-15  Richard Biener  <rguenther@suse.de>
3419         PR lto/83388
3420         * internal-fn.def (IFN_NOP): Add.
3421         * internal-fn.c (expand_NOP): Do nothing.
3422         * lto-streamer-in.c (input_function): Instead of removing
3423         sanitizer calls replace them with IFN_NOP calls.
3425 2017-12-15  Richard Sandiford  <richard.sandiford@linaro.org>
3426             Alan Hayward  <alan.hayward@arm.com>
3427             David Sherwood  <david.sherwood@arm.com>
3429         * dse.c (store_info, read_info_type): Replace begin and end with
3430         offset and width.
3431         (print_range): New function.
3432         (set_all_positions_unneeded, any_positions_needed_p)
3433         (check_mem_read_rtx, scan_stores, scan_reads, dse_step5): Update
3434         accordingly.
3435         (record_store): Likewise.  Optimize the case in which all positions
3436         are unneeded.
3437         (get_stored_val): Replace read_begin and read_end with read_offset
3438         and read_width.
3439         (replace_read): Update call accordingly.
3441 2017-12-15  Bin Cheng  <bin.cheng@arm.com>
3443         * gimple-loop-interchange.cc (STMT_COST_RATIO): New macro.
3444         (loop_cand::m_num_stmts, loop_cand::m_const_init_reduc): New members.
3445         (loop_cand::loop_cand): Initialize above members.
3446         (loop_cand::supported_operations): Delete.
3447         (loop_cand::can_interchange_p): Inline above function.
3448         (loop_cand::classify_simple_reduction): Record number of constant
3449         initialized simple reductions.
3450         (should_interchange_loops): New parameters.  Check stmt cost of loops
3451         to be interchange.
3452         (tree_loop_interchange::interchange): Prepare stmt cost of outer loop.
3453         Update call to should_interchange_loops.
3454         (should_interchange_loop_nest): Update call to
3455         should_interchange_loops.
3457 2017-12-15  Eric Botcazou  <ebotcazou@adacore.com>
3459         PR target/66488
3460         * ggc-page.c (HOST_BITS_PER_PTR): Do not define here...
3461         * hwint.h (HOST_BITS_PER_PTR): ...but here instead.
3462         * config/i386/xm-mingw32.h (HOST_BITS_PER_PTR): Delete.
3464 2017-12-15  Julia Koval  <julia.koval@intel.com>
3466         * config/i386/i386-builtin.def (__builtin_ia32_vaesenclast_v16qi,
3467         __builtin_ia32_vaesenclast_v32qi, __builtin_ia32_vaesenclast_v64qi):
3468         New.
3469         * config/i386/sse.md (vaesenclast_<mode>): New pattern.
3470         * config/i386/vaesintrin.h (_mm256_aesenclast_epi128,
3471         _mm512_aesenclast_epi128, _mm_aesenclast_epi128): New intrinsics.
3473 2017-12-15  Julia Koval  <julia.koval@intel.com>
3475         * config/i386/i386-builtin.def (__builtin_ia32_vaesenc_v16qi,
3476         __builtin_ia32_vaesenc_v32qi, __builtin_ia32_vaesenc_v64qi): New.
3477         * config/i386/sse.md (vaesenc_<mode>): New pattern.
3478         * config/i386/vaesintrin.h (_mm256_aesenc_epi128, _mm512_aesenc_epi128,
3479         _mm_aesenc_epi128): New intrinsics.
3481 2017-12-15  Julia Koval  <julia.koval@intel.com>
3483         * config/i386/i386-builtin.def (__builtin_ia32_vaesdeclast_v16qi,
3484         __builtin_ia32_vaesdeclast_v32qi, __builtin_ia32_vaesdeclast_v64qi):
3485         New.
3486         * config/i386/sse.md (vaesdeclast_<mode>): New pattern.
3487         * config/i386/vaesintrin.h (_mm256_aesdeclast_epi128,
3488         _mm512_aesdeclast_epi128, _mm_aesdeclast_epi128): New intrinsics.
3490 2017-12-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3492         * gimple-ssa-strength-reduction.c (analyze_increments):
3493         Distinguish replacement costs for constant strides from those for
3494         unknown strides.
3496 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3498         * var-tracking.c (variable_tracking_main_1): Formatting fix.
3500 2017-12-14  Bernd Edlinger  <bernd.edlinger@hotmail.de>
3502         * doc/invoke.texi: Document -Wcast-function-type.
3503         * recog.h (stored_funcptr): Change signature.
3504         * tree-dump.c (dump_node): Avoid warning.
3505         * typed-splay-tree.h (typed_splay_tree): Avoid warning.
3507 2017-12-14  Qing Zhao  <qing.zhao@oracle.com>
3509         PR middle_end/79538
3510         * gimple-fold.c (get_range_strlen): Add the handling of non-member 
3511         array.
3513 2017-12-14  David Malcolm  <dmalcolm@redhat.com>
3515         PR tree-optimization/83312
3516         * domwalk.h (dom_walker::dom_walker): Fix typo in comment.
3517         * tree-cfg.c (find_taken_edge): Update to handle NULL_TREE for
3518         "val" param, and to cope with arbitrary basic blocks.
3519         (find_taken_edge_cond_expr): Add "cond_stmt" param and use it to
3520         handle NULL_TREE for "val", dropping "bb" param.
3521         (find_taken_edge_switch_expr): Make "switch_stmt" param const and
3522         drop "bb" param.  Handle NULL_TREE for "val".
3523         (find_case_label_for_value): Make "switch_stmt" param const.
3524         * tree-vrp.c (class check_array_bounds_dom_walker): New subclass
3525         of dom_walker.
3526         (vrp_prop::check_all_array_refs): Reimplement as...
3527         (check_array_bounds_dom_walker::before_dom_children): ...this new
3528         vfunc.  Replace linear search through BB block list, excluding
3529         those with non-executable in-edges via dominator walk.
3531 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3533         * config/arm/arm.opt (mverbose-cost-dump): New option.
3534         * config/arm/arm.c (arm_rtx_costs): Use it.
3536 2017-12-14  Andreas Schwab  <schwab@linux-m68k.org>
3538         PR bootstrap/83396
3539         * reload1.c (emit_input_reload_insns): Skip debug markers.
3541 2017-12-14  Alexandre Oliva  <aoliva@redhat.com>
3543         * config/i386/i386.c (rest_of_insert_endbranch): Use call loc
3544         for its nop_endbr.
3546         PR bootstrap/83396
3547         * config/arc/arc.c (hwloop_optimize): Skip debug insns.
3548         * config/sh/sh-protos.h (sh_find_set_of_reg): Adjust.
3549         * config/sh/sh.c: Skip debug insns besides notes.
3550         * config/sh/sh.md: Likewise.
3551         * config/sh/sh_treg_combine.cc: Likewise.
3552         * config/sh/sync.md: Likewise.
3554 2017-12-14  Tom de Vries  <tom@codesourcery.com>
3556         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
3557         Add item for weak_undefined.
3559 2017-12-14  Richard Biener  <rguenther@suse.de>
3561         PR tree-optimization/67842
3562         * sese.h (bb_in_region): Remove #if 0'ed code.
3564 2017-12-14  Richard Biener  <rguenther@suse.de>
3566         PR tree-optimization/83326
3567         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
3568         may_be_zero parameter and handle it by not marking the first
3569         peeled copy as not exiting the loop.
3570         (try_peel_loop): Likewise.
3571         (canonicalize_loop_induction_variables): Use number_of_iterations_exit
3572         to handle the case of constant or zero iterations and perform
3573         loop header copying on-the-fly.
3575 2017-12-14  Richard Biener  <rguenther@suse.de>
3577         PR tree-optimization/83418
3578         * vr-values.c (vr_values::extract_range_for_var_from_comparison_expr):
3579         Instead of asserting we don't get unfolded comparisons deal with them.
3581 2017-12-14  Jakub Jelinek  <jakub@redhat.com>
3583         PR bootstrap/83396
3584         * var-tracking.c (vt_initialize): Ignore non-DEBUG_INSNs outside of
3585         basic blocks.  Assert debug bind insns don't appear outside of bbs,
3586         don't reset them.  Assert insns without BLOCK_FOR_INSN are outside of
3587         bb.  Simplify.
3589         PR tree-optimization/83198
3590         * gimple-ssa-sprintf.c (format_floating): Set type solely based on
3591         dir.modifier, regardless of TREE_TYPE (arg).  Assume non-REAL_CST
3592         value if arg is a REAL_CST with incompatible type.
3594 2017-12-14  Sudakshina Das  <sudi.das@arm.com>
3595             Bin Cheng  <bin.cheng@arm.com>
3597         PR target/81228
3598         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Move LTGT
3599         to CCFPEmode.
3600         * config/aarch64/aarch64-simd.md (vec_cmp<mode><v_int_equiv>): Add
3601         LTGT.
3603 2017-12-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
3605         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75,
3606         cortex-a75.cortex-a55): Specify AARCH64_FL_F16 in the arch features.
3608 2017-12-14  Richard Sandiford  <richard.sandiford@linaro.org>
3609             Alan Hayward  <alan.hayward@arm.com>
3610             David Sherwood  <david.sherwood@arm.com>
3612         * poly-int.h: New file.
3613         * poly-int-types.h: Likewise.
3614         * coretypes.h: Include them.
3615         (POLY_INT_CONVERSION): Define.
3616         * target.def (estimated_poly_value): New hook.
3617         * doc/tm.texi.in (TARGET_ESTIMATED_POLY_VALUE): New hook.
3618         * doc/tm.texi: Regenerate.
3619         * doc/poly-int.texi: New file.
3620         * doc/gccint.texi: Include it.
3621         * doc/rtl.texi: Describe restrictions on subreg modes.
3622         * Makefile.in (TEXI_GCCINT_FILES): Add poly-int.texi.
3623         * genmodes.c (NUM_POLY_INT_COEFFS): Provide a default definition.
3624         (emit_insn_modes_h): Emit a definition of NUM_POLY_INT_COEFFS.
3625         * targhooks.h (default_estimated_poly_value): Declare.
3626         * targhooks.c (default_estimated_poly_value): New function.
3627         * target.h (estimated_poly_value): Likewise.
3628         * wide-int.h (WI_UNARY_RESULT): Use wi::binary_traits.
3629         (wi::unary_traits): Delete.
3630         (wi::binary_traits::signed_shift_result_type): Define for
3631         offset_int << HOST_WIDE_INT, etc.
3632         (generic_wide_int::operator <<=): Define for all types and use
3633         wi::lshift instead of <<.
3634         (wi::hwi_with_prec): Add a default constructor.
3635         (wi::ints_for): New class.
3636         (operator <<): Define for all wide-int types.
3637         (operator /): New function.
3638         (operator %): Likewise.
3639         * selftest.h (ASSERT_KNOWN_EQ, ASSERT_KNOWN_EQ_AT, ASSERT_MAYBE_NE)
3640         (ASSERT_MAYBE_NE_AT): New macros.
3642 2017-12-13  Eric Botcazou  <ebotcazou@adacore.com>
3643             Dominik Vogt  <vogt@linux.vnet.ibm.com>
3645         PR middle-end/78468
3646         * emit-rtl.c (init_emit): Remove ??? comment.
3647         * explow.c (get_dynamic_stack_size): Take known alignment of stack
3648         pointer + STACK_DYNAMIC_OFFSET into account in lieu of STACK_BOUNDARY.
3649         * config/sparc/sparc.h (INIT_EXPANDERS): In 32-bit mode, lower the
3650         alignment of 3 virtual registers to BITS_PER_WORD.
3652         * config/sparc/sparc.c (sparc_compute_frame_size): Simplify.
3654 2017-12-13  Peter Bergner  <bergner@vnet.ibm.com>
3656         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_HTM_NO_SUSPEND): New define.
3657         * config/rs6000/rs6000.c (cpu_supports_info): Use it.
3659 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3661         PR bootstrap/83396
3662         * reload1.c (eliminate_regs_in_insn): Skip debug markers.
3664 2017-12-13  Jakub Jelinek  <jakub@redhat.com>
3666         * tree-cfg.c (verify_gimple_in_cfg): Verify no non-label stmts
3667         with the exception of debug begin stmt markers appear before
3668         labels.
3670         PR bootstrap/83396
3671         * final.c (rest_of_handle_final): Call variable_tracking_main only
3672         if !flag_var_tracking.
3674 2017-12-13  Alexandre Oliva  <aoliva@redhat.com>
3675             Jakub Jelinek  <jakub@redhat.com>
3677         PR bootstrap/83396
3678         PR debug/83391
3679         * tree-cfgcleanup.c (remove_forwarder_block): Keep after
3680         labels debug stmts that can only appear after labels.
3682 2017-12-13  Alexander Monakov  <amonakov@ispras.ru>
3684         PR rtl-optimization/82398
3685         * sel-sched.c (sel_rank_for_schedule): Fix check for zero
3686         EXPR_USEFULNESS in priority comparison.
3688 2017-12-13  Segher Boessenkool  <segher@kernel.crashing.org>
3690         PR rtl-optimization/83393
3691         * combine.c (move_deaths): If reg_stat points to a too new insn in
3692         last_death, do not use it: find the proper insn instead.
3694 2017-12-12  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
3696         PR target/82190
3697         * config/rs6000/rs6000-string.c (expand_block_compare,
3698         expand_strn_compare): Fix set_mem_size() calls.
3700 2017-12-12  Jeff Law  <law@redhat.com>
3702         PR tree-optimization/83298
3703         PR tree-optimization/83362
3704         PR tree-optimization/83383
3705         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Make
3706         push_value_range a public interface.  Add new argument to
3707         record_ranges_from_stmt.
3708         * gimple-ssa-evrp-analyze.c
3709         (evrp_range_analyzer::record_ranges_from_stmt): Add new argument.
3710         Update comments.  Handle recording temporary equivalences.
3711         * tree-ssa-dom.c (dom_opt_opt_walker::before_dom_children): Add
3712         new argument to call to evrp_range_analyzer::record_ranges_from_stmt.
3713         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Likewise.
3714         * tree-ssa-threadedge.c: Include alloc-pool.h, vr-values.h and
3715         gimple-ssa-evrp-analyze.h.
3716         (record_temporary_equivalences_from_phis): Add new argument.  When
3717         the PHI arg is an SSA_NAME, set the result's range to the range
3718         of the PHI arg.
3719         (record_temporary_equivalences_from_stmts_at_dest): Record ranges
3720         from statements too.
3721         (thread_through_normal_block): Accept new argument, evrp_range_analyzer.
3722         Pass it down to children as needed.
3723         (thread_outgoing_edges): Likewise.
3724         (thread_across_edge): Likewise.   Push/pop range state as needed.
3725         * tree-ssa-threadedge.h (thread_outgoing_edges): Update prototype.
3727 2017-12-12  Julia Koval  <julia.koval@intel.com>
3729         * config/i386/i386.c (PTA_SKYLAKE_AVX512): Add PTA_CLWB.
3730         (PTA_CANNONLAKE): Remove PTA_CLWB.
3732 2017-12-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
3734         PR target/83332
3735         * config/rs6000/vector.md (vcondv2dfv2di): New define_expand.
3736         (vcondv2div2df): Likewise.
3737         (vconduv2dfv2di): Likewise.
3739 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3741         * builtins.def (BUILT_IN_NEXTAFTER, BUILT_IN_NEXTAFTERF,
3742         BUILT_IN_NEXTAFTERL, BUILT_IN_NEXTTOWARD, BUILT_IN_NEXTTOWARDF,
3743         BUILT_IN_NEXTTOWARDL): Use ATTR_MATHFN_ERRNO instead of
3744         ATTR_MATHFN_FPROUNDING_ERRNO.
3746 2017-12-12  Richard Biener  <rguenther@suse.de>
3748         PR tree-optimization/83385
3749         * graphite-scop-detection.c (get_order, order): Remove.
3750         (bb_to_rpo): New global.
3751         (cmp_pbbs): Adjust.
3752         (build_scops): Sort pbbs in RPO order.
3754 2017-12-12  James Greenhalgh  <james.greenhalgh@arm.com>
3756         * combine.c (simplify_set): Do not transform subregs to zero_extends
3757         if the destination mode is a vector mode.
3759 2017-12-12  Jakub Jelinek  <jakub@redhat.com>
3761         PR tree-optimization/83359
3762         * tree-cfg.h (fold_loop_internal_call): Declare.
3763         * tree-vectorizer.c (fold_loop_internal_call): Moved to ...
3764         * tree-cfg.c (fold_loop_internal_call): ... here.  No longer static.
3765         (find_loop_dist_alias): New function.
3766         (move_sese_region_to_fn): If any dloop->orig_loop_num value is
3767         updated, also adjust any corresponding LOOP_DIST_ALIAS internal
3768         calls.
3770         PR tree-optimization/80631
3771         * tree-vect-loop.c (get_initial_def_for_reduction): Fix comment typo.
3772         (vect_create_epilog_for_reduction): Add INDUC_VAL and INDUC_CODE
3773         arguments, for INTEGER_INDUC_COND_REDUCTION use INDUC_VAL instead of
3774         hardcoding zero as the value if COND_EXPR is never true.  For
3775         INTEGER_INDUC_COND_REDUCTION don't emit the final COND_EXPR if
3776         INDUC_VAL is equal to INITIAL_DEF, and use INDUC_CODE instead of
3777         hardcoding MAX_EXPR as the reduction operation.
3778         (is_nonwrapping_integer_induction): Allow negative step.
3779         (vectorizable_reduction): Compute INDUC_VAL and INDUC_CODE for
3780         vect_create_epilog_for_reduction, if no value is suitable, don't
3781         use INTEGER_INDUC_COND_REDUCTION for now.  Formatting fixes.
3783 2017-12-12  Richard Biener  <rguenther@suse.de>
3785         PR tree-optimization/81889
3786         * tree-ssa-loop-niter.c (infer_loop_bounds_from_signedness): Use
3787         range info from the non-wrapping IV instead of just the range
3788         of the type.
3790 2017-12-12  Julia Koval  <julia.koval@intel.com>
3792         * config.gcc: Add vaesintrin.h.
3793         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
3794         New type.
3795         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
3796         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
3797         New builtins.
3798         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
3799         * config/i386/immintrin.h: Include vaesintrin.h.
3800         * config/i386/sse.md (vaesdec_<mode>): New pattern.
3801         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
3802         _mm_aesdec_epi128): New intrinsics.
3804 2017-12-12  Julia Koval  <julia.koval@intel.com>
3806         * common/config/i386/i386-common.c (OPTION_MASK_ISA_VAES_SET,
3807         OPTION_MASK_ISA_VAES_UNSET): New.
3808         (ix86_handle_option): Handle -mvaes.
3809         * config/i386/cpuid.h: Define bit_VAES.
3810         * config/i386/driver-i386.c (host_detect_local_cpu): Detect -mvaes.
3811         * config/i386/i386-c.c (__VAES__): New.
3812         * config/i386/i386.c (ix86_target_string): Add -mvaes.
3813         (ix86_valid_target_attribute_inner_p): Ditto.
3814         * config/i386/i386.h (TARGET_VAES, TARGET_VAES_P): New.
3815         * config/i386/i386.opt: Add -mvaes.
3816         * doc/invoke.texi: Ditto.
3818 2017-12-12  Alexandre Oliva  <aoliva@redhat.com>
3820         * debug.h (gcc_debug_hooks): Add inline_entry.
3821         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
3822         * debug.c (do_nothing_debug_hooks): Likewise.
3823         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
3824         * dwarf2out.c (dwarf2_debug_hooks): Likewise.
3825         (dwarf2_lineno_debug_hooks): Likewise.
3827         * common.opt (gstatement-frontiers): New, setting
3828         debug_nonbind_markers_p.
3829         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): Activate.
3830         * toplev.c (process_options): Autodetect value for debug statement
3831         frontiers option.
3832         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): Activate.
3833         * doc/invoke.texi (gstatement-frontiers, gno-statement-frontiers): New.
3835         * cfgexpand.c (expand_gimple_basic_block): Handle begin stmt
3836         markers.  Integrate source bind into debug stmt expand loop.
3837         (pass_expand::execute): Check debug marker limit.  Avoid deep
3838         TER and expand debug locations for debug bind insns only.
3839         * cse.c (insn_live_p): Keep nonbind markers and debug bindings
3840         followed by them.
3841         * df-scan.c (df_insn_delete): Accept out-of-block debug insn.
3842         * final.c (reemit_insn_block_notes): Take current block from
3843         nonbind markers.  Declare note where it's first set.
3844         (final_scan_insn): Handle begin stmt notes.  Emit is_stmt according to
3845         begin stmt markers if enabled.
3846         (notice_source_line): Handle nonbind markers.  Fail if their
3847         location is unknown or that of builtins.
3848         (rest_of_handle_final): Convert begin stmt markers to notes if
3849         var-tracking didn't run.
3850         (rest_of_clean_state): Skip begin stmt markers.
3851         * gimple-pretty-print.c (dump_gimple_debug): Handle begin stmt
3852         markers.
3853         * function.c (allocate_struct_function): Set begin_stmt_markers.
3854         * function.h (struct function): Add debug_marker_count counter
3855         and debug_nonbind_markers flag.
3856         * gimple-iterator.c (gsi_remove): Adjust debug_marker_count.
3857         * gimple-low.c (lower_function_body): Adjust
3858         debug_nonbind_markers.
3859         (lower_stmt): Drop or skip gimple debug stmts.
3860         (lower_try_catch): Skip debug stmts.
3861         * gimple.c (gimple_build_debug_begin_stmt): New.
3862         (gimple_copy): Increment debug_marker_count if copying one.
3863         * gimple.h (gimple_build_debug_begin_stmt): Declare.
3864         * gimplify.c (rexpr_location): New.
3865         (rexpr_has_location): New.
3866         (warn_switch_unreachable_r): Handle gimple debug stmts.
3867         (shortcut_cond_r): Call expr_location.
3868         (find_goto): New.
3869         (find_goto_label): New.
3870         (shortcut_cond_expr): Call expr_has_location, expr_location, and
3871         find_goto_label.
3872         (gimplify_cond_expr): Call find_goto_label, expr_has_location, and
3873         expr_location.
3874         (gimplify_expr): Handle begin stmt markers.  Reject debug expr decls.
3875         * langhooks-def.h (LANG_HOOKS_EMITS_BEGIN_STMT): New.  Add to...
3876         (LANG_HOOKS_INITIALIZER): ... this.
3877         * langhooks.h (struct lang_hooks): Add emits_begin_stmt.
3878         * lra-contraints.c (inherit_reload_reg): Tolerate between-blocks
3879         debug insns.
3880         (update_ebb_live_info): Skip debug insn markers.
3881         * lra.c (debug_insn_static_data): Rename to...
3882         (debug_bind_static_data): ... this.
3883         (debug_marker_static_data): New.
3884         (lra_set_insn_recog_data): Select one of the above depending
3885         on debug insn kind.
3886         (lra_update_isn_regno_info): Don't assume debug insns have
3887         freqs.
3888         (push_insns): Skip debug insns.
3889         * lto-streamer-in.c (input_function): Drop debug stmts
3890         depending on active options.  Adjust debug_nonbind_markers.
3891         * params.def (PARAM_MAX_DEBUG_MARKER_COUNT): New.
3892         * print-rtl.c (rtx_writer::print_rtx_operand_code_0): Handle
3893         begin stmt marker notes.
3894         (print_insn): Likewise.
3895         * recog.c (extract_insn): Recognize rtl for debug markers.
3896         * rtl.def (DEBUG_MARKER): New.
3897         * tree-inline.c: Include params.h.
3898         (remap_gimple_stmt): Handle nonbind markers.
3899         (maybe_move_debug_stmts_to_successors): Likewise.
3900         (copy_debug_stmt): Likewise.
3901         * tree-iterator.c (append_to_statement_list_1): Append begin stmt
3902         markers regardless of no side effects.
3903         (tsi_link_before): Don't update container's side effects when adding
3904         a begin stmt marker.
3905         (tsi_link_after): Likewise.
3906         (expr_first): Skip begin stmt markers.
3907         (expr_last): Likewise.
3908         * tree-pretty-print (dump_generic_node): Handle begin stmt markers.
3909         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
3910         Disregard nonbind markers.
3911         * tree.c (make_node_stat): Don't set side effects for begin stmt
3912         markers.
3913         (build1_stat): Likewise.
3914         * tree.def (DEBUG_BEGIN_STMT): New.
3915         * tree.h (GOTO_DESTINATION): Require a GOTO_EXPR.
3916         * var-tracking.c (delete_debug_insns): Renamed to...
3917         (delete_vta_debug_insns): ... this.
3918         (reemit_marker_as_note): New.
3919         (vt_initialize): Reemit markers.
3920         (delete_vta_debug_insns): Likewise.
3921         (vt_debug_insns_local): Reemit or delete markers.
3922         (variable_tracking_main_1): Likewise.
3923         * doc/generic.texi (DEBUG_BEGIN_STMT): Document.
3924         * doc/gimple.texi (gimple_debug_begin_stmt_p): New.
3925         (gimple_debug_nonbind_marker_p): New.
3926         (gimple_build_debug_bind): Adjust.
3927         (gimple_build_debug_begin_stmt): New.
3928         * doc/invoke.texi (max-debug-marker-count): New param.
3929         * doc/rtl.texi (debug_implicit_ptr, entry_value): New.
3930         (debug_parameter_ref, debug_marker): New.
3931         (NOTE_INSN_BEGIN_STMT): New.
3932         (DEBUG_INSN): Describe begin stmt markers.
3934         * cfgbuild.c (find_bb_boundaries): Don't purge dead edges if,
3935         without debug insns, we wouldn't, but clean up debug insns
3936         after a control flow insn nevertheless.
3938         * cfgcleanup.c (delete_unreachable_blocks): Use alternate
3939         block removal order if MAY_HAVE_DEBUG_BIND_INSNS.
3940         * cfgexpand.c (label_rtx_for_bb): Skip debug insns.
3941         * cfgrtl.c (try_redirect_by_replacing_jump): Skip debug insns.
3942         (rtl_tidy_fallthru_edge): Likewise.
3943         (rtl_verify_fallthru): Likewise.
3944         (rtl_verify_bb_layout): Likewise.
3945         (skip_insns_after_block): Likewise.
3946         (duplicate_insn_chain): Use DEBUG_BIND_INSN_P.
3947         * dwarf2out.c: Include print-rtl.h.
3948         (dwarf2out_next_real_insn): New.
3949         (dwarf2out_var_location): Call it.  Disregard begin stmt markers.
3950         Dump debug binds in asm comments.
3951         * gimple-iterator.c (gimple_find_edge_insert_loc): Skip debug stmts.
3952         * gimple-iterator.h (gsi_start_bb_nondebug): Remove; adjust
3953         callers to use gsi_start_nondebug_bb instead.
3954         (gsi_after_labels): Skip gimple debug stmts.
3955         (gsi_start_nondebug): New.
3956         * gimple-loop-interchange.c (find_deps_in_bb_for_stmt): Adjust.
3957         (proper_loop_form_for_interchange): Adjust.
3958         * gimple-low.c (gimple_seq_may_fallthru): Take last nondebug stmt.
3959         * gimple.h (gimple_seq_last_nondebug_stmt): New.
3960         * gimplify.c (last_stmt_in_scope): Skip debug stmts.
3961         (collect_fallthrough_labels): Likewise.
3962         (should_warn_for_implicit_fallthrough): Likewise.
3963         (warn_implicit_fallthrough_r): Likewise.
3964         (expand_FALLTHROUGH_r): Likewise.
3965         * graphite-isl-ast-to-gimple.c (gsi_insert_earliest): Adjust.
3966         (graphite_copy_stmts_from_block): Skip nonbind markers.
3967         * haifa-sched.c (sched_extend_bb): Skip debug insns.
3968         * ipa-icf-gimple.c (func_checker::compare_bb): Adjust.
3969         * jump.c (clean_barriers): Skip debug insns.
3970         * omp-expand.c (expand_parallel_call): Skip debug insns.
3971         (expand_task_call): Likewise.
3972         (remove_exit_barrier): Likewise.
3973         (expand_omp_taskreg): Likewise.
3974         (expand_omp_for_init_counts): Likewise.
3975         (expand_omp_for_generic): Likewise.
3976         (expand_omp_for_static_nochunk): Likewise.
3977         (expand_omp_for_static_chunk): Likewise.
3978         (expand_omp_simd): Likewise.
3979         (expand_omp_taskloop_for_outer): Likewise.
3980         (expand_omp_taskloop_for_inner): Likewise.
3981         (expand_oacc_for): Likewise.
3982         (expand_omp_sections): Likewise.
3983         (expand_omp_single): Likewise.
3984         (expand_omp_synch): Likewise.
3985         (expand_omp_atomic_load): Likewise.
3986         (expand_omp_atomic_store): Likewise.
3987         (expand_omp_atomic_fetch_op): Likewise.
3988         (expand_omp_atomic_pipeline): Likewise.
3989         (expand_omp_atomic_mutex): Likewise.
3990         (expand_omp_target): Likewise.
3991         (grid_expand_omp_for_loop): Likewise.
3992         (grid_expand_target_grid_body): Likewise.
3993         (build_omp_regions_1): Likewise.
3994         * omp-low.c (check_combined_parallel): Skip debug stmts.
3995         * postreload.c (fixup_debug_insns): Skip nonbind debug insns.
3996         * regcprop.c (find_oldest_value_reg): Ensure REGNO is not a pseudo.
3997         * sese.c (sese_trivially_empty_bb_p): Call is_gimple_debug in
3998         test.
3999         * tree-cfg.c (make_blobs_1): Skip debug stmts.
4000         (make_edges): Likewise.
4001         (cleanup_dead_labels): Likewise.
4002         (gimple_can_merge_blocks_p): Likewise.
4003         (stmt_starts_bb_p): Likewise.
4004         (gimple_block_label): Likewise.
4005         (gimple_redirect_edge_and_branch): Likewise.
4006         * tree-cfgcleanup.c (remove_forwarder_block): Rearrange skipping
4007         of debug stmts.
4008         (execute_cleanup_cfg_post_optimizing): Dump enumerated decls with
4009         TDF_SLIM.
4010         * tree-pretty-print (print_declaration): Omit initializer in slim
4011         dumps.
4012         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Mark begin stmt
4013         markers.
4014         (eliminate_unnecessary_stmts): Stabilize block removal order.
4015         * tree-ssa-tail-merge.c (find_duplicate): Skip debug stmts.
4016         * var-tracking.c (get_first_insn): New.
4017         (vt_emit_notes): Call it.
4018         (vt_initialize): Walk any insns before the first BB.
4019         (delete_debug_insns): Likewise.
4021         * gimple.h (enum gimple_debug_subcode): Add
4022         GIMPLE_DEBUG_BEGIN_STMT.
4023         (gimple_debug_begin_stmt_p): New.
4024         (gimple_debug_nonbind_marker_p): New.
4025         * tree.h (MAY_HAVE_DEBUG_MARKER_STMTS): New.
4026         (MAY_HAVE_DEBUG_BIND_STMTS): Renamed from....
4027         (MAY_HAVE_DEBUG_STMTS): ... this.  Check both.
4028         * insn-notes.def (BEGIN_STMT): New.
4029         * rtl.h (MAY_HAVE_DEBUG_MARKER_INSNS): New.
4030         (MAY_HAVE_DEBUG_BIND_INSNS): Renamed from....
4031         (MAY_HAVE_DEBUG_INSNS): ... this.  Check both.
4032         (NOTE_MARKER_LOCATION, NOTE_MARKER_P): New.
4033         (DEBUG_BIND_INSN_P, DEBUG_MARKER_INSN_P): New.
4034         (INSN_DEBUG_MARKER_KIND): New.
4035         (GEN_RTX_DEBUG_MARKER_BEGIN_STMT_PAT): New.
4036         (INSN_VAR_LOCATION): Check for VAR_LOCATION.
4037         (INSN_VAR_LOCATION_PTR): New.
4038         * cfgexpand.c (expand_debug_locations): Handle debug bind insns
4039         only.
4040         (expand_gimple_basic_block): Likewise.  Emit debug temps for TER
4041         deps only if debug bind insns are enabled.
4042         (pass_expand::execute): Avoid deep TER and expand
4043         debug locations for debug bind insns only.
4044         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Narrow
4045         debug stmts special handling down to debug bind stmts.
4046         * combine.c (try_combine): Narrow debug insns special handling
4047         down to debug bind insns.
4048         * cse.c (delete_trivially_dead_insns): Handle debug bindings.
4049         Narrow debug insns preexisting special handling down to debug
4050         bind insns.
4051         * dce.c (rest_of_handle_ud_dce): Narrow debug insns special
4052         handling down to debug bind insns.
4053         * function.c (instantiate_virtual_regs): Skip debug markers,
4054         adjust handling of debug binds.
4055         * gimple-ssa-backprop.c (backprop::prepare_change): Try debug
4056         temp insertion iff MAY_HAVE_DEBUG_BIND_STMTS.
4057         * haifa-sched.c (schedule_insn): Narrow special handling of debug
4058         insns to debug bind insns.
4059         * ipa-param-manipulation.c (ipa_modify_call_arguments): Narrow
4060         special handling of debug stmts to debug bind stmts.
4061         * ipa-split.c (split_function): Likewise.
4062         * ira.c (combine_and_move_insns): Adjust debug bind insns only.
4063         * loop-unroll.c (apply_opt_in_copies): Adjust tests on bind
4064         debug insns.
4065         * reg-stack.c (convert_regs_1): Use DEBUG_BIND_INSN_P.
4066         * regrename.c (build_def_use): Likewise.
4067         * regcprop.c (copyprop_hardreg_forward_1): Likewise.
4068         (pass_cprop_hardreg): Narrow special casing of debug insns to
4069         debug bind insns.
4070         * regstat.c (regstat_init_n_sets_and_refs): Likewise.
4071         * reload1.c (reload): Likewise.
4072         * sese.c (sese_insert_phis_for_liveouts): Narrow special
4073         casing of debug stmts to debug bind stmts.
4074         * shrink-wrap.c (move_insn_for_shrink_wrap): Likewise.
4075         * ssa-iterators.h (num_imm_uses): Likewise.
4076         * tree-cfg.c (gimple_merge_blocks): Narrow special casing of
4077         debug stmts to debug bind stmts.
4078         * tree-inline.c (tree_function_versioning): Narrow special casing
4079         of debug stmts to debug bind stmts.
4080         * tree-loop-distribution.c (generate_loops_for_partition):
4081         Narrow special casing of debug stmts to debug bind stmts.
4082         * tree-sra.c (analyze_access_subtree): Narrow special casing
4083         of debug stmts to debug bind stmts.
4084         * tree-ssa-dce.c (remove_dead_stmt): Narrow special casing of debug
4085         stmts to debug bind stmts.
4086         * tree-ssa-loop-ivopt.c (remove_unused_ivs): Narrow special
4087         casing of debug stmts to debug bind stmts.
4088         * tree-ssa-reassoc.c (reassoc_remove_stmt): Likewise.
4089         * tree-ssa-tail-merge.c (tail_merge_optimize): Narrow special
4090         casing of debug stmts to debug bind stmts.
4091         * tree-ssa-threadedge.c (propagate_threaded_block_debug_info):
4092         Likewise.
4093         * tree-ssa.c (flush_pending_stmts): Narrow special casing of
4094         debug stmts to debug bind stmts.
4095         (gimple_replace_ssa_lhs): Likewise.
4096         (insert_debug_temp_for_var_def): Likewise.
4097         (insert_debug_temps_for_defs): Likewise.
4098         (reset_debug_uses): Likewise.
4099         * tree-ssanames.c (release_ssa_name_fn): Likewise.
4100         * tree-vect-loop-manip.c (adjust_debug_stmts_now): Likewise.
4101         (adjust_debug_stmts): Likewise.
4102         (adjust_phi_and_debug_stmts): Likewise.
4103         (vect_do_peeling): Likewise.
4104         * tree-vect-loop.c (vect_transform_loop): Likewise.
4105         * valtrack.c (propagate_for_debug): Use BIND_DEBUG_INSN_P.
4106         * var-tracking.c (adjust_mems): Narrow special casing of debug
4107         insns to debug bind insns.
4108         (dv_onepart_p, dataflow_set_clar_at_call, use_type): Likewise.
4109         (compute_bb_dataflow, vt_find_locations): Likewise.
4110         (vt_expand_loc, emit_notes_for_changes): Likewise.
4111         (vt_init_cfa_base): Likewise.
4112         (vt_emit_notes): Likewise.
4113         (vt_initialize): Likewise.
4114         (vt_finalize): Likewise.
4116         * emit-rtl.c (next_nondebug_insn, prev_nondebug_insn): Reorder.
4117         (next_nonnote_nondebug_insn, prev_nonnote_nondebug_insn): Reorder.
4118         (next_nonnote_nondebug_insn_bb): New.
4119         (prev_nonnote_nondebug_insn_bb): New.
4120         (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove.
4121         * rtl.h (prev_nonnote_insn_bb, next_nonnote_insn_bb): Remove decls.
4122         (prev_nonnote_nondebug_insn_bb): Declare.
4123         (next_nonnote_nondebug_insn_bb): Declare.
4124         * cfgbuild.c (find_bb_boundaries): Adjust to skip debug insns.
4125         * cfgrtl.c (get_last_bb_insn): Likewise.
4126         * lra.c (push_insns): Likewise.
4128 2017-12-11  David Malcolm  <dmalcolm@redhat.com>
4130         PR c/82050
4131         * selftest-run-tests.c (selftest::run_tests): Move start/finish code
4132         to...
4133         * selftest.c (selftest::test_runner::test_runner): New ctor.
4134         (selftest::test_runner::~test_runner): New dtor.
4135         * selftest.h (class selftest::test_runner): New class.
4137 2017-12-11  Carl Love  <cel@us.ibm.com>
4139         * config/rs6000/altivec.h (vec_extract_fp32_from_shorth,
4140         vec_extract_fp32_from_shortl]): Add #defines.
4141         * config/rs6000/rs6000-builtin.def (VSLDOI_2DI): Add macro expansion.
4142         * config/rs6000/rs6000-c.c (ALTIVEC_BUILTIN_VEC_UNPACKH,
4143         ALTIVEC_BUILTIN_VEC_UNPACKL, ALTIVEC_BUILTIN_VEC_AND,
4144         ALTIVEC_BUILTIN_VEC_SLD, ALTIVEC_BUILTIN_VEC_SRL,
4145         ALTIVEC_BUILTIN_VEC_SRO, ALTIVEC_BUILTIN_VEC_SLD,
4146         ALTIVEC_BUILTIN_VEC_SLL): Add expansions.
4147         * doc/extend.texi: Add documentation for the added builtins.
4149 2017-12-11  Alexandre Oliva  <aoliva@redhat.com>
4151         PR rtl-optimization/80693
4152         PR rtl-optimization/81019
4153         PR rtl-optimization/81020
4154         * combine.c (distribute_notes): Reset any REG_UNUSED REGs that
4155         are not mentioned in i3.  Place the REG_UNUSED note on i2,
4156         possibly modified to REG_DEAD, if it did not originate in i3.
4158 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4160         * recog.c (store_data_bypass_p_1): New function.
4161         (store_data_bypass_p): Handle USE in a PARALLEL like CLOBBER.  Use
4162         store_data_bypass_p_1 to avoid code duplication.  Formatting fixes.
4164 2017-12-11  Segher Boessenkool  <segher@kernel.crashing.org>
4166         PR rtl-optimization/83361
4167         * ifcvt.c (if_convert): Call fixup_partitions.
4169 2017-12-11  Will Schmidt  <will_schmidt@vnet.ibm.com>
4171         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
4172         early folding of splat_u{8,16,32}.
4174 2017-12-11  Jakub Jelinek  <jakub@redhat.com>
4176         * config/aarch64/aarch64.c (aarch64_print_operand): Don't start
4177         output_operand_lossage first argument with capital letter.
4178         (aarch64_override_options): Don't start error and sorry first argument
4179         with capital letter.
4181 2017-12-11  Andi Kleen  <ak@linux.intel.com>
4183         PR gcov-profile/83355
4184         * auto-profile.c (string_table::get_index_by_decl): Don't
4185         recurse when abstract origin points to itself.
4187 2017-12-11  Bin Cheng  <bin.cheng@arm.com>
4189         PR tree-optimization/83320
4190         * gimple-loop-interchange.cc (free_data_refs_with_aux): Use delete.
4191         (prune_datarefs_not_in_loop): Ditto.
4193 2017-12-10  Gerald Pfeifer  <gerald@pfeifer.com>
4195         * doc/install.texi (Specific): Tweak link to mkssoftware.com.
4197 2017-12-10  Jakub Jelinek  <jakub@redhat.com>
4199         PR tree-optimization/83337
4200         * gimple-loop-interchange.cc (compute_access_stride): Handle
4201         bitfield DRs properly.
4203 2017-12-09  Jakub Jelinek  <jakub@redhat.com>
4205         PR tree-optimization/83338
4206         * tree-vect-stmts.c (vectorizable_operation): Handle POINTER_DIFF_EXPR
4207         vectorization as MINUS_EXPR with a subsequent VIEW_CONVERT_EXPR from
4208         vector of unsigned integers to vector of signed integers.
4210 2017-12-08  Vladimir Makarov  <vmakarov@redhat.com>
4212         PR rtl-optimization/83317
4213         * lra-constraints.c (process_address_1): Add insn code check.
4215 2017-12-08  Michael Matz  <matz@suse.de>
4217         Fix PR tree-optimization/83323
4218         * gimple-loop-jam (unroll_jam_possible_p): Correct test for
4219         head-controlled loops and loop BBs.
4220         * common.opt (funroll-and-jam): Remove, instead ...
4221         (floop-unroll-and-jam): ... reuse this option.
4222         * opts.c (default_options_table): Use OPT_floop_unroll_and_jam.
4223         * doc/invoke.texi (-funroll-and-jam): Move docu to ...
4224         (-floop-unroll-and-jam): ... this option.
4226 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4228         * ipa-polymorphic-call.c (noncall_stmt_may_be_vtbl_ptr_store): Fix
4229         a comment typo, get_base_ref_and_offset -> get_ref_base_and_extent.
4230         * ipa-prop.c (stmt_may_be_vtbl_ptr_store): Likewise.
4232 2017-12-08  Richard Biener  <rguenther@suse.de>
4234         PR middle-end/81782
4235         * tree-ssa-uninit.c (warn_uninitialized_vars): Properly
4236         handle accesses outside of zero-sized vars.
4238 2017-12-08  Martin Jambor  <mjambor@suse.cz>
4240         PR tree-optimization/83141
4241         * tree-sra.c (contains_vce_or_bfcref_p): Move up in the file, also
4242         test for MEM_REFs implicitely changing types with padding.  Remove
4243         inline keyword.
4244         (build_accesses_from_assign): Added contains_vce_or_bfcref_p checks.
4246 2017-12-08  Jakub Jelinek  <jakub@redhat.com>
4248         * config/arc/arc.c (arc_attribute_table): Add exclusions to
4249         the comment.
4250         * config/avr/avr.c (avr_attribute_table): Likewise.
4251         * config/msp430/msp430.c (msp430_attribute_table): Likewise.
4252         * config/rl78/rl78.c (rl78_attribute_table): Likewise.
4253         * config/nds32/nds32.c (nds32_attribute_table): Likewise.
4254         * config/darwin.h (SUBTARGET_ATTRIBUTE_TABLE): Initialize new member
4255         of struct attribute_spec.
4256         * config/i386/cygming.h (SUBTARGET_ATTRIBUTE_TABLE): Likewise.
4258 2017-12-08  Ulrich Weigand  <uweigand@de.ibm.com>
4260         PR target/82960
4261         * config/spu/spu.c (pad_bb): Only check INSN_CODE when INSN_P is true.
4263 2017-12-08  Jan Hubicka  <hubicka@ucw.cz>
4265         PR middle-end/83609
4266         * profile-count.c (profile_count::from_gcov_type): Move from
4267         profile-count.h; handle overflow.
4268         * profile-count.h (profile_count::from_gcov_type): Move offline.
4270 2017-12-08  Segher Boessenkool  <segher@kernel.crashing.org>
4272         PR rtl-optimization/83304
4273         * combine.c (move_deaths): If we do not know where a register died,
4274         search for it.
4276 2017-12-08  Richard Biener  <rguenther@suse.de>
4278         * gimple-loop-interchange.cc (tree_loop_interchange::interchange):
4279         Provide -fopt-info-loop feedback when we interchange in a nest.
4281 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4283         * config/arm/driver-arm.c (arm_cpu_table): Use fp-capable product names
4284         for armv6 ARM CPU IDs.
4286 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4288         * common/config/arm/arm-common.c: Include <algorithm>.
4289         (INCLUDE_VECTOR): Define.
4290         (compare_opt_names): New function.
4291         (arm_rewrite_selected_arch): Only strip out extensions that can be
4292         expressed through -mfpu.  Sort the remaining extensions
4293         alphabetically.
4295 2017-06-08  Richard Earnshaw  <rearnsha@arm.com>
4297         * config/arm/arm.h (arm_asm_auto_mfpu): Declare.
4298         (ASM_CPU_SPEC_FUNCTIONS): Add new rule asm_auto_mfpu.
4299         (ASM_CPU_SPEC): Use it if -mfpu is set to auto.
4300         * common/config/arm/arm-common.c (arm_asm_auto_mfpu): New function.
4302 2017-06-08  Tristan Gingold  <gindold@adacore.com>
4304         PR ada/81470
4305         * dwarf2out.c (dwarf2out_do_cfi_startproc): Only emit
4306         .cfi_personality or .cfi_lsda if the eh data format is dwarf2.
4308 2017-12-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4310         * config/sol2.h (SOLARIS_ATTRIBUTE_TABLE): Initialize new member
4311         of struct attribute_spec.
4313 2017-12-08  Julia Koval  <julia.koval@intel.com>
4315         * config/i386/avx512vnniintrin.h (_mm512_dpwssds_epi32,
4316         _mm512_mask_dpwssds_epi32, _mm512_maskz_dpwssds_epi32): New intrinsics.
4317         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssds_epi32,
4318         _mm256_mask_dpwssds_epi32, _mm256_maskz_dpwssds_epi32,
4319         _mm_dpwssds_epi32, _mm_mask_dpwssds_epi32,
4320         _mm_maskz_dpwssds_epi32): Ditto.
4322 2017-12-08  Richard Biener  <rguenther@suse.de>
4324         PR tree-optimization/81303
4325         * tree-vect-stmts.c (vect_is_simple_cond): For invariant
4326         conditions try to create a comparison vector type matching
4327         the data vector type.
4328         (vectorizable_condition): Adjust.
4329         * tree-vect-patterns.c (vect_recog_mask_conversion_pattern):
4330         Leave invariant conditions alone in case we can vectorize those.
4332 2017-12-08  Julia Koval  <julia.koval@intel.com>
4334         * config/i386/avx512vnniintrin.h (_mm512_dpwssd_epi32,
4335         _mm512_mask_dpwssd_epi32, _mm512_maskz_dpwssd_epi32): New intrinsics.
4336         * config/i386/avx512vnnivlintrin.h (_mm256_dpwssd_epi32,
4337         _mm256_mask_dpwssd_epi32, _mm256_maskz_dpwssd_epi32, _mm_dpwssd_epi32,
4338         _mm_mask_dpwssd_epi32, _mm_maskz_dpwssd_epi32): Ditto.
4340 2017-12-08  Julia Koval  <julia.koval@intel.com>
4342         * config/i386/avx512vnniintrin.h (_mm512_dpbusds_epi32,
4343         _mm512_mask_dpbusds_epi32, _mm512_maskz_dpbusds_epi32): New.
4344         * config/i386/avx512vnnivlintrin.h (_mm256_dpbusds_epi32,
4345         _mm256_mask_dpbusds_epi32, _mm256_maskz_dpbusds_epi32,
4346         _mm_dpbusds_epi32, _mm_mask_dpbusds_epi32,
4347         _mm_maskz_dpbusds_epi32): New intrinsics.
4349 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4351         * config/nios2/nios2.md (ld<bhw_uns>io): Add splitter for memory
4352         operand.
4353         (ld<bh>io_signed): Likewise.
4354         (st<bhw>io): Likewise.
4355         * config/nios2/predicates.md (ldstio_memory_operand): Allow
4356         SMALL_INT12 constant integer operand.
4358 2017-12-07  Sandra Loosemore  <sandra@codesourcery.com>
4360         * config/nios2/nios2.c (nios2_symbolic_constant_allowed):
4361         Rename to...
4362         (nios2_large_constant_allowed): ...this.  Adjust uses.
4363         (nios2_plus_symbolic_constant_p): Rename to...
4364         (nios2_plus_large_constant_p): ...this.  Adjust uses.
4365         (nios2_legitimate_address_p): Correct CONST_INT handling.
4366         (nios2_symbolic_memory_operand_p): Rename to...
4367         (nios2_large_constant_memory_operand_p): ...this.  Adjust uses.
4368         (nios2_large_constant_p): Check for large constant integers too.
4369         (nios2_split_large_constant): Handle constant integers.
4370         (nios2_split_symbolic_memory_operand): Rename to...
4371         (nios2_split_large_constant_memory_operand): ...this.  Adjust uses.
4372         (nios2_legitimize_constant_address): Handle constant integers.
4373         (r0rel_constant_p): Handle small constant integers.
4374         (nios2_print_operand_address): Handle r0-relative integer addresses.
4375         * config/nios2/nios2-protos.h: Adjust for renamed functions.
4376         * config/nios2/nios2.md: Adjust for renamed functions.
4378 2017-12-07  Andrew Waterman  <andrew@sifive.com>
4380         * config/riscv/riscv.c (TARGET_ASM_SELECT_SECTION): New define.
4381         (TARGET_HAVE_SRODATA_SECTION): New define.
4382         (riscv_select_section): New function.
4384 2017-12-08  Joseph Myers  <joseph@codesourcery.com>
4385             Alexander Monakov  <amonakov@ispras.ru>
4386             Jakub Jelinek  <jakub@redhat.com>
4388         PR target/81906
4389         * config/i386/i386.c (ix86_expand_rint): Handle flag_rounding_math.
4391 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4393         * config/aarch64/aarch64.c (aarch64_print_address_internal): Return
4394         a bool success value.  Don't call output_operand_lossage here.
4395         (aarch64_print_ldpstp_address): Return a bool success value.
4396         (aarch64_print_operand_address): Call output_addr_const if
4397         aarch64_print_address_internal fails.
4398         (aarch64_print_operand): Don't assert that the mode is 16 bytes for
4399         'y'; call output_operand_lossage instead.  Call output_operand_lossage
4400         if aarch64_print_ldpstp_address fails.
4402 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4404         * tree-vector-builder.h
4405         (tree_vector_builder::binary_encoded_nelts): Declare.
4406         * tree-vector-builder.c
4407         (tree_vector_builder::binary_encoded_nelts): New function.
4408         * fold-const.c (negate_expr_p): Likewise.
4409         (operand_equal_p, fold_checksum_tree): Likewise.
4410         * tree-loop-distribution.c (const_with_all_bytes_same): Likewise.
4411         * tree.c (integer_zerop, integer_onep, integer_all_onesp, real_zerop)
4412         (real_onep, real_minus_onep, add_expr, initializer_zerop): Likewise.
4413         (uniform_vector_p): Likewise.
4414         * varasm.c (const_hash_1, compare_constant): Likewise.
4415         * tree-ssa-ccp.c: Include tree-vector-builder.h.
4416         (valid_lattice_transition): Operate directly on the VECTOR_CST
4417         encoding.
4418         * ipa-icf.c: Include tree-vector-builder.h.
4419         (sem_variable::equals): Operate directly on the VECTOR_CST encoding.
4420         * print-tree.c (print_node): Print encoding of VECTOR_CSTs.
4422 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4424         * tree.c (build_vector): Delete.
4425         * tree.h (build_vector): Make static and move into the self-testing
4426         block.
4428 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4430         * vector-builder.h (vector_builder::derived): New const overload.
4431         (vector_builder::elt): New function.
4432         * tree-vector-builder.h (tree_vector_builder::type): New function.
4433         (tree_vector_builder::apply_step): Declare.
4434         * tree-vector-builder.c (tree_vector_builder::apply_step): New
4435         function.
4436         * gimple-fold.h (tree_vector_builder): Declare.
4437         (gimple_build_vector): Take a tree_vector_builder instead of a
4438         type and vector of elements.
4439         * gimple-fold.c (gimple_build_vector): Likewise.
4440         * tree-vect-loop.c (get_initial_def_for_reduction): Update call
4441         accordingly.
4442         (get_initial_defs_for_reduction): Likewise.
4443         (vectorizable_induction): Likewise.
4445 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4447         * tree-vector-builder.h
4448         (tree_vector_builder::new_binary_operation): Declare.
4449         * tree-vector-builder.c
4450         (tree_vector_builder::new_binary_operation): New function.
4451         * fold-const.c (fold_relational_const): Use it.
4452         (const_binop): Likewise.  Check that both input vectors have
4453         the same number of elements, thus excluding things like WIDEN_SUM.
4454         Check whether it is possible to operate directly on the encodings
4455         of stepped inputs.
4457 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4459         * fold-const.c (fold_negate_expr_1): Use tree_vector_builder and
4460         new_unary_operation, operating only on the encoded elements.
4461         (const_unop): Likewise.
4462         (exact_inverse): Likewise.
4463         (distributes_over_addition_p): New function.
4464         (const_binop): Use tree_vector_builder and new_unary_operation
4465         for combinations of VECTOR_CST and INTEGER_CST.  Operate only
4466         on the encoded elements unless the encoding is strided and the
4467         operation does not distribute over addition.
4468         (fold_convert_const):  Use tree_vector_builder and
4469         new_unary_operation.  Operate only on the encoded elements
4470         for truncating integer conversions, or for non-stepped encodings.
4472 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4474         * config/sparc/sparc.c: Include tree-vector-builder.h.
4475         (sparc_fold_builtin): Use tree_vector_builder instead of build_vector.
4476         * expmed.c: Include tree-vector-builder.h.
4477         (make_tree): Use tree_vector_builder instead of build_vector.
4478         * fold-const.c: Include tree-vector-builder.h.
4479         (const_binop): Use tree_vector_builder instead of build_vector.
4480         (const_unop): Likewise.
4481         (native_interpret_vector): Likewise.
4482         (fold_vec_perm): Likewise.
4483         (fold_ternary_loc): Likewise.
4484         * gimple-fold.c: Include tree-vector-builder.h.
4485         (gimple_fold_stmt_to_constant_1): Use tree_vector_builder instead
4486         of build_vector.
4487         * tree-ssa-forwprop.c: Include tree-vector-builder.h.
4488         (simplify_vector_constructor): Use tree_vector_builder instead
4489         of build_vector.
4490         * tree-vect-generic.c: Include tree-vector-builder.h.
4491         (add_rshift): Use tree_vector_builder instead of build_vector.
4492         (expand_vector_divmod): Likewise.
4493         (optimize_vector_constructor): Likewise.
4494         * tree-vect-loop.c: Include tree-vector-builder.h.
4495         (vect_create_epilog_for_reduction): Use tree_vector_builder instead
4496         of build_vector.  Explicitly use a stepped encoding for
4497         { 1, 2, 3, ... }.
4498         * tree-vect-slp.c: Include tree-vector-builder.h.
4499         (vect_get_constant_vectors): Use tree_vector_builder instead
4500         of build_vector.
4501         (vect_transform_slp_perm_load): Likewise.
4502         (vect_schedule_slp_instance): Likewise.
4503         * tree-vect-stmts.c: Include tree-vector-builder.h.
4504         (vectorizable_bswap): Use tree_vector_builder instead of build_vector.
4505         (vect_gen_perm_mask_any): Likewise.
4506         (vectorizable_call): Likewise.  Explicitly use a stepped encoding.
4507         * tree.c: (build_vector_from_ctor): Use tree_vector_builder instead
4508         of build_vector.
4509         (build_vector_from_val): Likewise.  Explicitly use a duplicate
4510         encoding.
4512 2017-12-07  Richard Sandiford  <richard.sandiford@arm.com>
4514         * doc/generic.texi (VECTOR_CST): Describe new representation of
4515         vector constants.
4516         * vector-builder.h: New file.
4517         * tree-vector-builder.h: Likewise.
4518         * tree-vector-builder.c: Likewise.
4519         * Makefile.in (OBJS): Add tree-vector-builder.o.
4520         * tree.def (VECTOR_CST): Update comment to refer to generic.texi.
4521         * tree-core.h (tree_base): Add a vector_cst field to the u union.
4522         (tree_vector): Change the number of elements to
4523         vector_cst_encoded_nelts.
4524         * tree.h (VECTOR_CST_NELTS): Redefine using TYPE_VECTOR_SUBPARTS.
4525         (VECTOR_CST_ELTS): Delete.
4526         (VECTOR_CST_ELT): Redefine using vector_cst_elt.
4527         (VECTOR_CST_LOG2_NPATTERNS, VECTOR_CST_NPATTERNS): New macros.
4528         (VECTOR_CST_NELTS_PER_PATTERN, VECTOR_CST_DUPLICATE_P): Likewise.
4529         (VECTOR_CST_STEPPED_P, VECTOR_CST_ENCODED_ELTS): Likewise.
4530         (VECTOR_CST_ENCODED_ELT): Likewise.
4531         (vector_cst_encoded_nelts): New function.
4532         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4533         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4534         (vector_cst_int_elt, vector_cst_elt): Declare.
4535         * tree.c: Include tree-vector-builder.h.
4536         (tree_code_size): Abort if passed VECTOR_CST.
4537         (tree_size): Update for new VECTOR_CST layout.
4538         (make_vector): Take the values of VECTOR_CST_LOG2_NPATTERNS and
4539         VECTOR_CST_NELTS_PER_PATTERN as arguments.
4540         (build_vector): Use tree_vector_builder.
4541         (vector_cst_int_elt, vector_cst_elt): New functions.
4542         (drop_tree_overflow): For VECTOR_CST, drop the TREE_OVERFLOW from the
4543         encoded elements and then create the vector in the canonical form.
4544         (check_vector_cst, check_vector_cst_duplicate, check_vector_cst_fill)
4545         (check_vector_cst_stepped, test_vector_cst_patterns): New functions.
4546         (tree_c_tests): Call test_vector_cst_patterns.
4547         * lto-streamer-out.c (DFS::DFS_write_tree_body): Handle the new
4548         VECTOR_CST fields.
4549         (hash_tree): Likewise.
4550         * tree-streamer-out.c (write_ts_vector_tree_pointers): Likewise.
4551         (streamer_write_tree_header): Likewise.
4552         * tree-streamer-in.c (lto_input_ts_vector_tree_pointers): Likewise.
4553         (streamer_alloc_tree): Likewise.  Update call to make_vector.
4554         * fold-const.c (fold_ternary_loc): Avoid using VECTOR_CST_ELTS.
4556 2017-12-07  Richard Sandiford  <richard.sandiford@linaro.org>
4558         * selftest.h (ASSERT_TRUE_AT, ASSERT_FALSE_AT, ASSERT_EQ_AT)
4559         (ASSERT_NE, ASSERT_PRED1): Add underscores to local variable names
4560         * selftest-rtl.h (ASSERT_RTX_EQ, ASSERT_RTX_PTR_EQ): Likewise.
4562 2017-12-07  Bin Cheng  <bin.cheng@arm.com>
4563             Richard Biener  <rguenther@suse.de>
4565         PR tree-optimization/81303
4566         * Makefile.in (gimple-loop-interchange.o): New object file.
4567         * common.opt (floop-interchange): Reuse the option from graphite.
4568         * doc/invoke.texi (-floop-interchange): Ditto.  New document for
4569         -floop-interchange and mention it for -O3.
4570         * opts.c (default_options_table): Enable -floop-interchange at -O3.
4571         * gimple-loop-interchange.cc: New file.
4572         * params.def (PARAM_LOOP_INTERCHANGE_MAX_NUM_STMTS): New parameter.
4573         (PARAM_LOOP_INTERCHANGE_STRIDE_RATIO): New parameter.
4574         * passes.def (pass_linterchange): New pass.
4575         * timevar.def (TV_LINTERCHANGE): New time var.
4576         * tree-pass.h (make_pass_linterchange): New declaration.
4577         * tree-ssa-loop-ivcanon.c (create_canonical_iv): Change to external
4578         interchange.  Record IV before/after increment in new parameters.
4579         * tree-ssa-loop-ivopts.h (create_canonical_iv): New declaration.
4580         * tree-vect-loop.c (vect_is_simple_reduction): Factor out reduction
4581         path check into...
4582         (check_reduction_path): ...New function here.
4583         * tree-vectorizer.h (check_reduction_path): New declaration.
4585 2017-12-07  Vladimir Makarov  <vmakarov@redhat.com>
4587         PR target/83252
4588         PR rtl-optimization/80818
4589         * lra.c (add_regs_to_insn_regno_info): Make a hard reg in CLOBBER
4590         always early clobbered.
4591         * lra-lives.c (process_bb_lives): Check input hard regs for early
4592         clobbered non-operand hard reg.
4594 2017-12-07  Jakub Jelinek  <jakub@redhat.com>
4596         PR middle-end/83164
4597         * tree-cfg.c (verify_gimple_assign_binary): Don't require
4598         types_compatible_p, just that TYPE_MODE is the same.
4600 2017-12-07  Martin Sebor  <msebor@redhat.com>
4602         PR c/81544
4603         * attribs.c (empty_attribute_table): Initialize new member of
4604         struct attribute_spec.
4605         (decl_attributes): Add argument.  Handle mutually exclusive
4606         combinations of attributes.
4607         (selftests::test_attribute_exclusions): New function.
4608         (selftests::attribute_c_tests): Ditto.
4609         * attribs.h (decl_attributes): Add default argument.
4610         * selftest.h (attribute_c_tests): Declare.
4611         * selftest-run-tests.c (selftest::run_tests): Call attribute_c_tests.
4612         * tree-core.h (attribute_spec::exclusions, exclude): New type and
4613         member.
4614         * doc/extend.texi (Common Function Attributes): Update const and pure.
4615         * config/alpha/alpha.c (vms_attribute_table): Initialize new member
4616         of struct attribute_spec.
4617         * config/arc/arc.c (arc_attribute_table): Same.
4618         * config/arm/arm.c (arm_attribute_table): Same.
4619         * config/avr/avr.c ( avr_attribute_table): Same.
4620         * config/bfin/bfin.c (bfin_attribute_table): Same.
4621         * config/cr16/cr16.c (cr16_attribute_table): Same.
4622         * config/epiphany/epiphany.c (epiphany_attribute_table): Same.
4623         * config/h8300/h8300.c (h8300_attribute_table): Same.
4624         * config/i386/i386.c (ix86_attribute_table): Same.
4625         * config/ia64/ia64.c (ia64_attribute_table): Same.
4626         * config/m32c/m32c.c (m32c_attribute_table): Same.
4627         * config/m32r/m32r.c (m32r_attribute_table): Same.
4628         * config/m68k/m68k.c (m68k_attribute_table): Same.
4629         * config/mcore/mcore.c (mcore_attribute_table): Same.
4630         * config/microblaze/microblaze.c (microblaze_attribute_table): Same.
4631         * config/mips/mips.c (mips_attribute_table): Same.
4632         * config/msp430/msp430.c (msp430_attribute_table): Same.
4633         * config/nds32/nds32.c (nds32_attribute_table): Same.
4634         * config/nvptx/nvptx.c (nvptx_attribute_table): Same.
4635         * config/powerpcspe/powerpcspe.c (rs6000_attribute_table): Same.
4636         * config/rl78/rl78.c (rl78__attribute_table): Same.
4637         * config/rs6000/rs6000.c (rs6000_attribute_table): Same.
4638         * onfig/rx/rx.c (rx_attribute_table): Same.
4639         * config/s390/s390.c (s390_handle_vectorbool_attribute): Same.
4640         * config/sh/sh.c (sh_attribute_table): Same.
4641         * config/sparc/sparc.c (sparc_attribute_table): Same.
4642         * config/spu/spu.c (spu_attribute_table): Same.
4643         * config/stormy16/stormy16.c (xstormy16_attribute_table): Same.
4644         * config/v850/v850.c (v850_attribute_table): Same.
4645         * config/visium/visium.c (visium_attribute_table): Same.
4647 2017-12-07  Tamar Christina  <tamar.christina@arm.com>
4649         PR target/82641
4650         * config/arm/arm.c (INCLUDE_STRING): Define.
4651         (arm_last_printed_arch_string, arm_last_printed_fpu_string): New.
4652         (arm_declare_function_name): Conservatively emit .arch, .arch_extensions
4653         and .fpu.
4655 2017-12-07  Michael Matz  <matz@suse.de>
4657         Add unroll and jam pass
4659         * gimple-loop-jam.c: New file.
4660         * Makefile.in (OBJS): Add gimple-loop-jam.o.
4661         * common.opt (funroll-and-jam): New option.
4662         * opts.c (default_options_table): Add unroll-and-jam at -O3.
4663         * params.def (PARAM_UNROLL_JAM_MIN_PERCENT): New param.
4664         (PARAM_UNROLL_JAM_MAX_UNROLL): Ditto.
4665         * passes.def: Add pass_loop_jam.
4666         * timevar.def (TV_LOOP_JAM): Add.
4667         * tree-pass.h (make_pass_loop_jam): Declare.
4668         * cfgloop.c (flow_loop_tree_node_add): Add AT argument.
4669         * cfgloop.h (flow_loop_tree_node_add): Adjust declaration.
4670         * cfgloopmanip.c (duplicate_loop): Add AT argument, adjust call
4671         to flow_loop_tree_node_add.
4672         (duplicate_subloops, copy_loops_to): Append to sibling list.
4673         * cfgloopmanip.h: (duplicate_loop): Adjust declaration.
4674         * doc/invoke.texi (-funroll-and-jam): Document new option.
4675         (unroll-jam-min-percent, unroll-jam-max-unroll): Document new params.
4677 2017-12-07  Richard Biener  <rguenther@suse.de>
4679         PR tree-optimization/83296
4680         PR tree-optimization/67769
4681         * tree-ssa-phiopt.c (conditional_replacement): Do not reset
4682         flow sensitive info in an unrelated BB.
4683         (value_replacement): Use reset_flow_sensitive_info.
4684         (minmax_replacement): Reset flow sensitive info on the def
4685         we move.  Do not reset flow sensitive info in the whole BB
4686         we move the stmt to.
4687         (abs_replacement): Likewise.
4689 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4691         PR target/43871
4692         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4693         rs6000_cpu to the given -mcpu=, or to the default processor.
4695 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4697         * config/rs6000/rs6000.h (rs6000_cpu_attr): Delete.
4698         * config/rs6000/rs6000.c (rs6000_variable_issue_1): Use rs6000_tune
4699         instead of rs6000_cpu_attr.
4700         (rs6000_adjust_cost): Ditto.
4701         (is_microcoded_insn): Ditto.
4702         (rs6000_adjust_priority): Ditto.
4703         (rs6000_issue_rate): Ditto.
4704         (rs6000_use_sched_lookahead): Ditto.
4705         (rs6000_use_sched_lookahead_guard): Ditto.
4706         (rs6000_sched_reorder): Ditto.
4707         (force_new_group): Ditto.
4708         * config/rs6000/rs6000.md (cpu attribute): Ditto.
4709         (group_ending_nop): Ditto.
4711 2017-12-07  Segher Boessenkool  <segher@kernel.crashing.org>
4713         * config/rs6000/rs6000.opt (rs6000_tune): New variable.
4714         * config/rs6000/rs6000.c (rs6000_option_override_internal): Also set
4715         rs6000_tune.  Use rs6000_tune instead of rs6000_cpu where appropriate.
4716         (rs6000_loop_align): Use rs6000_tune instead of rs6000_cpu where
4717         appropriate.
4718         (rs6000_reassociation_width): Ditto.
4719         (rs6000_emit_epilogue): Ditto.
4720         (rs6000_adjust_cost): Ditto.
4721         (is_microcoded_insn): Ditto.
4722         (is_cracked_insn): Ditto.
4723         (rs6000_adjust_priority): Ditto.
4724         (rs6000_sched_reorder): Ditto.
4725         (rs6000_sched_reorder2): Ditto.
4726         (insn_must_be_first_in_group): Ditto.
4727         (insn_must_be_last_in_group): Ditto.
4728         (rs6000_register_move_cost): Ditto.
4729         * config/rs6000/rs6000.h (rs6000_cpu_attr): Use rs6000_tune instead of
4730         rs6000_cpu.
4732 2017-12-07  Julia Koval  <julia.koval@intel.com>
4734         * config.gcc: Add vaesintrin.h.
4735         * config/i386/i386-builtin-types.def (V64QI_FTYPE_V64QI_V64QI):
4736         New type.
4737         * config/i386/i386-builtin.def (__builtin_ia32_vaesdec_v16qi,
4738         __builtin_ia32_vaesdec_v32qi, __builtin_ia32_vaesdec_v64qi):
4739         New builtins.
4740         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
4741         * config/i386/immintrin.h: Include vaesintrin.h.
4742         * config/i386/sse.md (vaesdec_<mode>): New pattern.
4743         * config/i386/vaesintrin.h (_mm256_aesdec_epi128, _mm512_aesdec_epi128,
4744         _mm_aesdec_epi128): New intrinsics.
4746 2017-12-06  David Malcolm  <dmalcolm@redhat.com>
4748         * Makefile.in (C_COMMON_OBJS): Add c-family/c-spellcheck.o.
4749         * spellcheck-tree.c (find_closest_macro_cpp_cb): Move to
4750         c-family/c-spellcheck.cc.
4751         (best_macro_match::best_macro_match): Likewise.
4752         * spellcheck-tree.h
4753         (struct edit_distance_traits<cpp_hashnode *>): Move to
4754         c-family/c-spellcheck.h.
4755         (class best_macro_match): Likewise.
4757 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4759         PR tree-optimization/83293
4760         * gimple-ssa-strength-reduction.c (insert_initializers): Use
4761         GSI_NEW_STMT instead of GSI_SAME_STMT in gsi_insert_after that
4762         might insert into empty bb.
4764         PR sanitizer/81281
4765         * match.pd ((T)(P + A) - (T)P -> (T) A): Split into separate
4766         simplify for plus with :c added, and pointer_plus without that.
4767         ((T)P - (T)(P + A) -> -(T) A): Likewise.  If type is integral
4768         with undefined overflow and the conversion is not widening,
4769         perform negation in utype and only convert to type afterwards.
4770         ((T)(P + A) - (T)(P + B) -> (T)A - (T)B): Split into separate
4771         simplify for plus with :c added, and pointer_plus without that.
4772         If type is integral with undefined overflow and the conversion is
4773         not widening, perform minus in utype and only convert to type
4774         afterwards.  Move the last pointer_diff_expr simplify into the
4775         two outermost ifs.
4777 2017-12-06  Martin Sebor  <msebor@redhat.com>
4779         PR tree-optimization/82646
4780         * builtins.c (maybe_emit_chk_warning): Use size as the bound for
4781         strncpy, not maxlen.
4783 2017-12-06  Martin Sebor  <msebor@redhat.com>
4785         * doc/invoke.texi (-Wstringop-truncation): Mention attribute
4786         nonstring.
4788         PR tree-optimization/83075
4789         * tree-ssa-strlen.c (handle_builtin_stxncpy): Avoid assuming
4790         strncat/strncpy don't change length of source string.
4792 2017-12-06  Eric Botcazou  <ebotcazou@adacore.com>
4794         Revert
4795         2017-11-29  Martin Aberg  <maberg@gaisler.com>
4797         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
4798         to prevent b2bst errata sequence.
4799         (sqrtdf2_fix): Likewise.
4801 2017-12-06  Jakub Jelinek  <jakub@redhat.com>
4803         PR tree-optimization/81945
4804         * cfgloop.h (FOR_EACH_LOOP_FN): Use FN instead of hardcoding fn.
4805         * tree-cfg.c (move_sese_region_to_fn): If any of the loops moved
4806         to dest_cfun has orig_loop_num set, either remap it to the new
4807         loop number if the loop got moved too, or clear it.
4809 2017-12-05  Steve Ellcey  <sellcey@cavium.com>
4811         * config/aarch64/thunderx2-t99.md (thunderx2t99_branch): Add trap
4812         to reservation.
4813         (thunderx2t99_nothing): New insn reservation.
4814         (thunderx2t99_mrs): New insn reservation.
4815         (thunderx2t99_multiple): New insn reservation.
4816         (thunderx2t99_alu_basi): Add bfx to reservation.
4817         (thunderx2t99_fp_cmp): Add fccmps and fccmpd to reservation.
4819 2017-12-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
4821         PR target/82248
4822         * config/arm/arm.md (probe_stack) : Use the 'o' constraint.
4824 2017-12-05  Bin Cheng  <bin.cheng@arm.com>
4826         * tree-ssa-dce.c (simple_dce_from_worklist): Move and rename from
4827         tree-ssa-pre.c::remove_dead_inserted_code.
4828         * tree-ssa-dce.h: New file.
4829         * tree-ssa-pre.c (tree-ssa-dce.h): Include new header file.
4830         (remove_dead_inserted_code): Move and rename to function
4831         tree-ssa-dce.c::simple_dce_from_worklist.
4832         (pass_pre::execute): Update use.
4834 2017-12-05  Richard Biener  <rguenther@suse.de>
4836         PR tree-optimization/83277
4837         * graphite-isl-ast-to-gimple.c (should_copy_to_new_region): Make sure
4838         to code-gen liveout vars.
4840 2017-12-05  Richard Sandiford  <richard.sandiford@linaro.org>
4842         * config/aarch64/aarch64-simd.md (aarch64_simd_bsldi_internal)
4843         (aarch64_simd_bsldi_alt): Check REG_P before GP_REGNUM_P.
4844         (aarch64_cm<optab>di, aarch64_cmtstdi): Add leading "&&" to
4845         split condition.
4847 2017-12-05  Max Filippov  <jcmvbkbc@gmail.com>
4849         * config/xtensa/xtensa.c (xtensa_asan_shadow_offset): New
4850         function.
4851         (TARGET_ASAN_SHADOW_OFFSET): New macro definition.
4852         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): Set to 1 if
4853         ASAN is enabled.
4855 2017-12-05  Richard Biener  <rguenther@suse.de>
4857         * timevar.def (TV_TREE_RECIP, TV_TREE_SINCOS, TV_TREE_WIDEN_MUL): Add.
4858         * tree-ssa-math-opts.c (pass_data_cse_reciprocal): Use TV_TREE_RECIP.
4859         (pass_data_cse_sincos): Use TV_TREE_SINCOS.
4860         (pass_data_optimize_widening_mul): Use TV_TREE_WIDEN_MUL.
4862 2017-12-05  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
4864         * dbxout.c (dbxout_block): Grow buf to 30 bytes.
4866 2017-12-05  Martin Liska  <mliska@suse.cz>
4867             Jakub Jelinek  <jakub@redhat.com>
4869         * doc/invoke.texi: Document the options.
4870         * flag-types.h (enum sanitize_code): Add
4871         SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4872         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Add handling
4873         of SANITIZE_POINTER_COMPARE and SANITIZE_POINTER_SUBTRACT.
4874         * opts.c: Define new sanitizer options.
4875         * sanitizer.def (BUILT_IN_ASAN_POINTER_COMPARE): Likewise.
4876         (BUILT_IN_ASAN_POINTER_SUBTRACT): Likewise.
4878 2017-12-05  Julia Koval  <julia.koval@intel.com>
4880         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VNNI_SET,
4881         OPTION_MASK_ISA_AVX512VNNI_UNSET): New.
4882         (ix86_handle_option): Handle -mavx512vnni.
4883         * config/i386/cpuid.h (bit_AVX512VNNI): New bit.
4884         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
4885         * config/i386/i386-c (__AVX512VNNI__): New.
4886         * config/i386/i386.c (ix86_target_string): Handle new option.
4887         (ix86_valid_target_attribute_inner_p): Handle new option.
4888         * config/i386/i386.h (TARGET_AVX512VNNI, TARGET_AVX512VNNI_P): New.
4889         * config/i386/i386.opt (mavx512vnni): New option.
4891 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4893         PR target/81616
4894         * athlon.md: Disable for generic.
4895         * haswell.md: Enable for generic.
4896         * i386.c (ix86_sched_init_global): Add core hooks for generic.
4897         * x86-tune-sched.c (ix86_issue_rate): Increase issue rate for generic
4898         to 4.
4899         (ix86_adjust_cost): Move generic to haswell path.
4901 2017-12-04  Eric Botcazou  <ebotcazou@adacore.com>
4903         * config/sparc/sparc.c (sparc_do_work_around_errata): Use mem_ref
4904         instead of MEM_P in a couple more places.  Fix formatting issues.
4906 2017-12-04  Jim Wilson  <jimw@sifive.com>
4908         * config/riscv/riscv.c (riscv_for_each_saved_reg): Use GP_REG_LAST
4909         instead of GP_REG_LAST-1.
4910         (riscv_adjust_libcall_cfi_prologue): Likewise.
4911         (riscv_adjust_libcall_cri_epilogue): Likewise.
4912         * config/riscv/riscv.h (CALL_USED_REGISTERS): Change a6 to t6 in
4913         comment.
4915 2017-12-04  Luis Machado  <luis.machado@linaro.org>
4917         * ipa-pure-const.c (check_decl): Add missing newline.
4918         (state_from_flags): Likewise.
4920 2017-12-04  Jeff Law  <law@redhat.com>
4922         PR tree-optimizatin/78496
4923         * gimple-ssa-evrp-analyze.h
4924         (evrp_range_analyzer::get_vr_values): Simplify.
4925         * gimple-ssa-evrp-analyze.c: Corresponding changes.
4926         * tree-ssa-dom.c: Include alloc-pool.h, tree-vrp.h, vr-values.h
4927         and gimple-ssa-evrp-analyze.h.
4928         (dom_opt_dom_walker class): Add evrp_range_analyzer member.
4929         (simplify_stmt_for_jump_threading): Copy a blob of code from
4930         tree-vrp.c to use ranges to simplify statements.
4931         (dom_opt_dom_walker::before_dom_children): Call
4932         evrp_range_analyzer::{enter,record_ranges_from_stmt} methods.
4933         (dom_opt_dom_walker::after_dom_children): Similarly for
4934         evrp_range_analyzer::leave.
4935         (dom_opt_dom_walker::optimize_stmt): Use EVRP ranges to optimize
4936         conditionals.
4938         * gimple-ssa-evrp-analyze.c
4939         (evrp_range_analyzer::extract_range_from_stmt):  Always use
4940         vr_values::update_value_range so preexisting range info is
4941         medged with new range info, even if the new range is VR_VARYING.
4943 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4945         * combine.c: Adjust comment.
4946         (use_crosses_set_p): Delete.
4947         (can_combine_p): Use modified_between_p instead of use_crosses_set_p.
4948         (try_combine): Ditto.
4950 2017-12-04  Richard Biener  <rguenther@suse.de>
4952         PR tree-optimization/83255
4953         * graphite-isl-ast-to-gimple.c (translate_isl_ast_node_for):
4954         Re-add zero-iteration check.
4956 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4958         PR rtl-optimization/83245
4959         * lra.c (collect_non_operand_hard_regs): Treat clobbers of non-operand
4960         hard registers as earlyclobber, also if not in an asm.
4962 2017-12-04  Segher Boessenkool  <segher@kernel.crashing.org>
4964         PR bootstrap/83265
4965         Revert
4966         2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
4968         PR target/43871
4969         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
4970         rs6000_cpu based on cpu_index, not tune_index.
4972 2017-12-04  Richard Biener  <rguenther@suse.de>
4974         PR tree-optimization/83238
4975         * graphite-scop-detection.c (scop_detection::merge_sese): Make
4976         code match comment, rejecting invalid SESE regions.
4978 2017-12-03  John David Anglin  <danglin@gcc.gnu.org>
4980         * config/pa/pa.c (pa_legitimate_address_p): For scaled indexing,
4981         require base operand is a REG_POINTER prior to reload on targets
4982         with non-equivalent space registers.
4984 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4986         * ipa-cp.c (ipcp_lattice<valtype>::print): Update dumping.
4987         (update_specialized_profile): Fix updating of counts.
4988         (perhaps_add_new_callers): Likewise.
4990 2017-12-01  Jan Hubicka  <hubicka@ucw.cz>
4992         PR target/81616
4993         * x86-tune.def: Remove obsolette FIXMEs.
4994         (X86_TUNE_PARTIAL_FLAG_REG_STALL): Disable for generic
4995         (X86_TUNE_FUSE_CMP_AND_BRANCH_32, X86_TUNE_FUSE_CMP_AND_BRANCH_64,
4996         X86_TUNE_FUSE_CMP_AND_BRANCH_SOFLAGS, X86_TUNE_FUSE_ALU_AND_BRANCH):
4997         Enable for generic.
4998         (X86_TUNE_PAD_RETURNS): Disable for generic.
5000 2017-12-02  Jakub Jelinek  <jakub@redhat.com>
5002         PR tree-optimization/83170
5003         PR tree-optimization/83241
5004         * gimple-ssa-store-merging.c
5005         (imm_store_chain_info::try_coalesce_bswap): Update vuse field from
5006         gimple_vuse (ins_stmt) in case it has changed.
5007         (imm_store_chain_info::output_merged_store): Likewise.
5009         * tree-chkp.c (chkp_compute_bounds_for_assignment): Handle
5010         POINTER_DIFF_EXPR.
5012         PR c++/81212
5013         * tree-cfg.c (pass_warn_function_return::execute): Handle
5014         __builtin_ubsan_handle_missing_return like __builtin_unreachable
5015         with BUILTINS_LOCATION.
5017         PR target/78643
5018         PR target/80583
5019         * expr.c (get_inner_reference): If DECL_MODE of a non-bitfield
5020         is BLKmode for vector field with vector raw mode, use TYPE_MODE
5021         instead of DECL_MODE.
5023         * config/i386/i386-protos.h (standard_sse_constant_opcode): Change
5024         last argument to rtx pointer.
5025         * config/i386/i386.c (standard_sse_constant_opcode): Replace X argument
5026         with OPERANDS.  For AVX+ 128-bit VEX encoded instructions over 256-bit
5027         or 512-bit.  If setting EXT_REX_SSE_REG_P, use EVEX encoded insn
5028         depending on the chosen ISAs.
5029         * config/i386/i386.md (*movxi_internal_avx512f, *movoi_internal_avx,
5030         *movti_internal, *movdi_internal, *movsi_internal, *movtf_internal,
5031         *movdf_internal, *movsf_internal): Adjust standard_sse_constant_opcode
5032         callers.
5033         * config/i386/sse.md (mov<mode>_internal): Likewise.
5034         * config/i386/mmx.md (*mov<mode>_internal): Likewise.
5036 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5038         * doc/invoke.texi (-dp): Say that instruction cost is printed as well.
5040 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5042         * config/rs6000/rs6000-protos.h (emit_fusion_addis): Remove last two
5043         parameters from prototype.
5044         * config/rs6000/rs6000.c (emit_fusion_addis): Remove last two
5045         parameters.  Don't print a comment.
5046         (emit_fusion_gpr_load): Adjust.
5047         (emit_fusion_load_store): Adjust.
5048         * config/rs6000/rs6000.md (*fusion_p9_<mode>_constant): Adjust.
5049         * config/rs6000/vsx.md (two peepholes): Print the "vector load fusion"
5050         comment on the second line.
5052 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5054         PR target/43871
5055         * config/rs6000/rs6000.c (rs6000_option_override_internal): Set
5056         rs6000_cpu based on cpu_index, not tune_index.
5058 2017-12-01  Segher Boessenkool  <segher@kernel.crashing.org>
5060         * final.c (output_asm_name): Print insn_cost.  Shorten output.  Print
5061         which_alternative instead of which_alternative + 1.
5062         (output_asm_insn): Print an extra tab if the template is short.
5064 2017-12-01  Jim Wilson  <jimw@sifive.com>
5066         * common.opt (use_gnu_debug_info_extensions): Delete DWARF_DEBUG from
5067         comment.
5068         * config/vx-common.h (DWARF_DEBUGGING_INFO): Delete undef.
5069         * doc/tm.texi.in (PREFERRED_DEBUGGING_TYPE): Delete DWARF_DEBUG
5070         reference.
5071         * doc/tm.texi: Regenerate.
5073 2017-12-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
5075         PR target/81959
5076         * config/rs6000/rs6000.c (rs6000_address_for_fpconvert): Check for
5077         whether we can allocate pseudos before trying to fix an address.
5078         * config/rs6000/rs6000.md (float_<mode>si2_hw): Make sure the
5079         memory address is indexed or indirect.
5080         (floatuns_<mode>si2_hw2): Likewise.
5082 2017-12-01  Jason Merrill  <jason@redhat.com>
5084         * Makefile.in (TAGS): Add c-family/*.cc.
5086 2017-12-01  Wilco Dijkstra  <wdijkstr@arm.com>
5088         * config/aarch64/aarch64.md (call_insn): Use %c rather than %a.
5089         (call_value_insn): Likewise.
5090         (sibcall_insn): Likewise.
5091         (sibcall_value_insn): Likewise.
5092         (movsi_aarch64): Likewise.
5093         (movdi_aarch64): Likewise.
5094         (add_losym_): Likewise.
5095         (ldr_got_small_): Likewise.
5096         (ldr_got_small_sidi): Likewise.
5097         (ldr_got_small_28k_): Likewise.
5098         (ldr_got_small_28k_sidi): Likewise.
5099         * config/aarch64/aarch64.c (aarch64_print_address_internal):
5100         Move output_addr_const to symbolic case. Add error check.
5102 2017-12-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
5104         * config/s390/predicates.md (plus16_Q_operand): New predicate.
5105         * config/s390/s390.md: Disable MVC merging peephole if it would
5106         disable operand forwarding.
5107         (new peephole2): Split MVCs if it would turn them into up to 2
5108         forwardable MVCs.
5110 2017-12-01  Richard Biener  <rguenther@suse.de>
5112         PR tree-optimization/83232
5113         * tree-vect-data-refs.c (vect_analyze_data_ref_accesses): Fix
5114         detection of same access. Instead of breaking the group here
5115         do not consider the duplicate.  Add comment explaining real fix.
5117 2017-12-01  Jonathan Wakely  <jwakely@redhat.com>
5119         * doc/md.texi (Insn Splitting): Fix "central flowgraph" typo.
5121 2017-12-01  Sudakshina Das  <sudi.das@arm.com>
5123         * config/arm/vfp.md (*movhf_vfp_fp16): Add conds attribute.
5125 2017-12-01  Jakub Jelinek  <jakub@redhat.com>
5127         * function.h (struct function): Remove cilk_frame_decl,
5128         is_cilk_function and calls_cilk_spawn fields.
5129         * tree-inline.h (struct copy_body_data): Remove remap_var_for_cilk
5130         field.
5131         * omp-simd-clone.c (simd_clone_clauses_extract): Don't clear
5132         cilk_elemental field.
5133         * cgraph.h (struct cgraph_simd_clone): Remove cilk_elemental field.
5134         * target.def: Adjust comment.
5135         * config/i386/i386.c (ix86_simd_clone_compute_vecsize_and_simdlen):
5136         Don't test cilk_elemental.
5138         PR tree-optimization/83233
5139         * gimple-ssa-store-merging.c (nop_stats, bswap_stats): Use
5140         bswap_stat name for the struct.
5142         PR c/79153
5143         * tree.h (SWITCH_BREAK_LABEL_P): Define.
5144         * gimplify.c (collect_fallthrough_labels): Handle GIMPLE_BIND
5145         starting with a GIMPLE_SWITCH and ending with GIMPLE_LABEL with
5146         SWITCH_BREAK_LABEL_P set on the label.
5147         (gimplify_switch_expr): Set SWITCH_BREAK_LABEL_P on the label
5148         added for default case if it was missing and not all cases covered.
5149         Wrap GIMPLE_SWITCH and the switch_body_seq into a GIMPLE_BIND if
5150         switch_body_seq ends with a GIMPLE_LABEL with SWITCH_BREAK_LABEL_P
5151         set on the label.
5152         * tree-chrec.c (evolution_function_is_univariate_p): Add return true;
5153         to avoid -Wimplicit-fallthrough warning.
5154         * config/i386/i386.c (ix86_expand_special_args_builtin): Add
5155         FALLTHRU comment to avoid -Wimplicit-fallthrough warning.
5157         PR tree-optimization/83221
5158         * tree-ssa-reassoc.c (sort_by_operand_rank): Shift bb_rank
5159         down by 16.
5160         (init_reassoc): Formatting fix.
5162         PR sanitizer/81275
5163         * tree-cfg.c (group_case_labels_stmt): Don't optimize away
5164         C++ FE implicitly added __builtin_unreachable () until -Wreturn-type
5165         is diagnosed.
5167         PR sanitizer/83219
5168         * tree-cfg.c: Include asan.h.
5169         (gimple_seq_unreachable_p): Return false for -fsanitize=unreachable.
5171 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5173         * config/i386/i386.md: Fix AVX512 register width in AVX512 instruction.
5175 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5177         * config/i386/i386.c (standard_sse_constant_opcode): Fix wrong form for
5178         vpcmpeqd instruction.
5180 2017-12-01  Sergey Shalnov  <Sergey.Shalnov@intel.com>
5182         * config/i386/i386.c (standard_sse_constant_opcode): Fix
5183         registers type for 128bit mode.
5185 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5187         * spellcheck-tree.c (test_find_closest_identifier): Use ; instead
5188         of ;;.
5189         * gengtype-state.c (read_state_pair): Likewise.
5190         * gimple-fold.c (gimple_fold_builtin_string_compare): Likewise.
5191         * sel-sched-dump.c (dump_insn_rtx_1): Likewise.
5192         * ipa-cp.c (intersect_aggregates_with_edge): Likewise.
5193         * ifcvt.c (noce_try_store_flag_constants): Likewise.
5194         * tree-ssa-ccp.c (ccp_finalize): Likewise.
5195         * omp-grid.c (grid_process_kernel_body_copy): Likewise.
5196         * builtins.c (fold_builtin_3): Likewise.
5197         * graphite-scop-detection.c
5198         (scop_detection::stmt_has_simple_data_refs_p): Likewise.
5199         * hsa-gen.c (hsa_function_representation::hsa_function_representation):
5200         Likewise.
5202 2017-12-01  Maxim Ostapenko  <m.ostapenko@samsung.com>
5204         PR sanitizer/81697
5205         * asan.c (asan_protect_global): Add new ignore_decl_rtl_set_p
5206         parameter. Return true if ignore_decl_rtl_set_p is true and other
5207         conditions are satisfied.
5208         * asan.h (asan_protect_global): Add new parameter.
5209         * varasm.c (categorize_decl_for_section): Pass true as second parameter
5210         to asan_protect_global calls.
5212 2017-11-30  Jim Wilson  <jimw@sifive.com>
5214         * doc/invoke.texi (RISC-V Options): Delete nonexistent -mmemcpy and
5215         -mno-memcpy options.  For -mplt, -mfdiv, -mdiv, -msave-restore, and
5216         -mstrict-align, add info on default value.  Delete redundant lines for
5217         -mabi.  Add missing -mexplicit-relocs docs.
5219 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5221         * config/arc/arc.md (trap): New pattern.
5223 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5225         * config/arc/arc.c (hwloop_optimize): Prevent the last ZOL
5226         instruction to end into a delay slot.
5227         * config/arc/arc.md (cond_delay_insn): Check if the instruction
5228         can be placed into a delay slot against reg_note.
5230 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5232         * config/arc/arc.c (hwloop_optimize): Update hw-loop's end/start
5233         labels number of usages.
5235 2017-11-30  Claudiu Zissulescu  <claziss@synopsys.com>
5237         * config/arc/arc.c (arc_cannot_substitue_mem_equiv_p): New
5238         function.
5239         (TARGET_CANNOT_SUBSTITUTE_MEM_EQUIV_P): Define.
5241 2017-11-30  Jakub Jelinek  <jakub@redhat.com>
5243         PR target/83210
5244         * internal-fn.c (expand_mul_overflow): Optimize unsigned
5245         multiplication by power of 2 constant into two shifts + comparison.
5247 2017-11-30  Jan Hubicka  <hubicka@ucw.cz>
5249         PR target/81616
5250         * config/i386/x86-tune-costs.h (generic_cost): Revise for modern CPUs.
5252 2017-11-30  Richard Biener  <rguenther@suse.de>
5254         PR tree-optimization/83202
5255         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Add
5256         allow_peel argument and guard peeling.
5257         (canonicalize_loop_induction_variables): Likewise.
5258         (canonicalize_induction_variables): Pass false.
5259         (tree_unroll_loops_completely_1): Pass unroll_outer to disallow
5260         peeling from cunrolli.
5262 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5264         * combine.c (try_combine): Print a message to dump file whenever
5265         I0, I1, or I2 cannot be combined into I3.
5267 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5269         PR rtl-optimization/83156
5270         PR rtl-optimization/82621
5271         * combine.c (try_combine): Don't split an I2 if one of the dests is
5272         set again before I3.  Allow unused dests.
5274 2017-11-29  Segher Boessenkool  <segher@kernel.crashing.org>
5276         * config/rs6000/rs6000.md (*add<mode>3_carry_in_internal2): New.
5278 2017-11-29  Vladimir Makarov  <vmakarov@redhat.com>
5280         PR rtl-optimization/80818
5281         * lra.c (collect_non_operand_hard_regs): New arg insn.  Pass it
5282         recursively.  Use insn code for clobber.
5283         (lra_set_insn_recog_data): Pass the new arg to
5284         collect_non_operand_hard_regs.
5285         (add_regs_to_insn_regno_info): Pass insn instead of uid.  Use insn
5286         code for clobber.
5287         (lra_update_insn_regno_info): Pass insn to
5288         add_regs_to_insn_regno_info.
5290 2017-11-29  Jim Wilson  <jimw@sifive.com>
5291             Andrew Waterman  <andrew@sifive.com>
5293         * config/riscv/riscv.c (SINGLE_SHIFT_COST): New.
5294         (riscv_rtx_costs): Case ZERO_EXTRACT, match new pattern, and return
5295         SINGLE_SHIFT_COST.  Case LT and ZERO_EXTEND, likewise.  Case ASHIFT,
5296         use SINGLE_SHIFT_COST.
5297         * config/riscv/riscv.md (lshrsi3_zero_extend_1): New.
5298         (lshrsi3_zero_extend_2, lshrsi3_zero_extend_3): New.
5300 2017-11-29  Julia Koval  <julia.koval@intel.com>
5302         * config/i386/avx512vbmi2intrin.h (_mm512_shldv_epi16,
5303         _mm512_mask_shldv_epi16, _mm512_maskz_shldv_epi16, _mm512_shldv_epi32,
5304         _mm512_mask_shldv_epi32, _mm512_maskz_shldv_epi32, _mm512_shldv_epi64,
5305         _mm512_mask_shldv_epi64, _mm512_maskz_shldv_epi64): New intrinsics.
5306         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldv_epi16,
5307         _mm256_mask_shldv_epi16, _mm256_maskz_shldv_epi16, _mm256_shldv_epi32,
5308         _mm256_mask_shldv_epi32, _mm256_maskz_shldv_epi32, _mm256_shldv_epi64,
5309         _mm256_mask_shldv_epi64, _mm256_maskz_shldv_epi64, _mm_shldv_epi16,
5310         _mm_mask_shldv_epi16, _mm_maskz_shldv_epi16, _mm_shldv_epi32,
5311         _mm_mask_shldv_epi32, _mm_maskz_shldv_epi32, _mm_shldv_epi64,
5312         _mm_mask_shldv_epi64, _mm_maskz_shldv_epi64): Ditto.
5313         * config/i386/i386-builtin.def (__builtin_ia32_vpshldv_v32hi,
5314         __builtin_ia32_vpshldv_v32hi_mask, __builtin_ia32_vpshldv_v32hi_maskz,
5315         __builtin_ia32_vpshldv_v16hi, __builtin_ia32_vpshldv_v16hi_mask,
5316         __builtin_ia32_vpshldv_v16hi_maskz, __builtin_ia32_vpshldv_v8hi,
5317         __builtin_ia32_vpshldv_v8hi_mask, __builtin_ia32_vpshldv_v8hi_maskz,
5318         __builtin_ia32_vpshldv_v16si, __builtin_ia32_vpshldv_v16si_mask,
5319         __builtin_ia32_vpshldv_v16si_maskz, __builtin_ia32_vpshldv_v8si,
5320         __builtin_ia32_vpshldv_v8si_mask, __builtin_ia32_vpshldv_v8si_maskz,
5321         __builtin_ia32_vpshldv_v4si, __builtin_ia32_vpshldv_v4si_mask,
5322         __builtin_ia32_vpshldv_v4si_maskz, __builtin_ia32_vpshldv_v8di,
5323         __builtin_ia32_vpshldv_v8di_mask, __builtin_ia32_vpshldv_v8di_maskz,
5324         __builtin_ia32_vpshldv_v4di, __builtin_ia32_vpshldv_v4di_mask,
5325         __builtin_ia32_vpshldv_v4di_maskz, __builtin_ia32_vpshldv_v2di,
5326         __builtin_ia32_vpshldv_v2di_mask,
5327         __builtin_ia32_vpshldv_v2di_maskz): New builtins.
5328         * config/i386/sse.md (vpshldv_<mode>, vpshldv_<mode>_mask,
5329         vpshldv_<mode>_maskz, vpshldv_<mode>_maskz_1): New patterns.
5331 2017-11-29  Julia Koval  <julia.koval@intel.com>
5333         * config/i386/avx512vbmi2intrin.h (_mm512_shrdv_epi16,
5334         _mm512_mask_shrdv_epi16, _mm512_maskz_shrdv_epi16, _mm512_shrdv_epi32,
5335         _mm512_mask_shrdv_epi32, _mm512_maskz_shrdv_epi32, _mm512_shrdv_epi64,
5336         _mm512_mask_shrdv_epi64, _mm512_maskz_shrdv_epi64): New intrinsics.
5337         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdv_epi16,
5338         _mm256_mask_shrdv_epi16, _mm256_maskz_shrdv_epi16, _mm256_shrdv_epi32,
5339         _mm256_mask_shrdv_epi32, _mm256_maskz_shrdv_epi32, _mm256_shrdv_epi64,
5340         _mm256_mask_shrdv_epi64, _mm256_maskz_shrdv_epi64, _mm_shrdv_epi16,
5341         _mm_mask_shrdv_epi16, _mm_maskz_shrdv_epi16, _mm_shrdv_epi32,
5342         _mm_mask_shrdv_epi32, _mm_maskz_shrdv_epi32, _mm_shrdv_epi64,
5343         _mm_mask_shrdv_epi64, _mm_maskz_shrdv_epi64): Ditto.
5344         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_V32HI,
5345         V32HI_FTYPE_V32HI_V32HI_V32HI_INT, V16HI_FTYPE_V16HI_V16HI_V16HI_INT,
5346         V8HI_FTYPE_V8HI_V8HI_V8HI_INT, V8SI_FTYPE_V8SI_V8SI_V8SI_INT,
5347         V4SI_FTYPE_V4SI_V4SI_V4SI_INT, V8DI_FTYPE_V8DI_V8DI_V8DI,
5348         V8DI_FTYPE_V8DI_V8DI_V8DI_INT, V4DI_FTYPE_V4DI_V4DI_V4DI_INT,
5349         V16SI_FTYPE_V16SI_V16SI_V16SI, V16SI_FTYPE_V16SI_V16SI_V16SI_INT,
5350         V2DI_FTYPE_V2DI_V2DI_V2DI_INT): New types.
5351         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5352         * config/i386/sse.md (vpshrdv_<mode>, vpshrdv_<mode>_mask,
5353         vpshrdv_<mode>_maskz, vpshrdv_<mode>_maskz_1): New pattern.
5355 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5357         * config/sparc/sparc.c (sparc_do_work_around_errata): Treat the
5358         movsi_pic_gotdata_op instruction as a load for the UT699 errata
5359         workaround.
5361 2017-11-29  Martin Aberg  <maberg@gaisler.com>
5363         * config/sparc/sparc.md (divdf3_fix): Add NOP and adjust length
5364         to prevent b2bst errata sequence.
5365         (sqrtdf2_fix): Likewise.
5367 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5369         * config/sparc/sparc.c (fpop_reg_depend_p): New function.
5370         (div_sqrt_insn_p): New function.
5371         (sparc_do_work_around_errata): Insert NOP instructions to
5372         prevent sequences that could trigger the TN-0013 errata for
5373         certain LEON3 processors.
5374         (pass_work_around_errata::gate): Also test sparc_fix_lost_divsqrt.
5375         (sparc_option_override): Set sparc_fix_lost_divsqrt appropriately.
5376         * config/sparc/sparc.md (fix_lost_divsqrt): New attribute.
5377         (in_branch_delay): Prevent div and sqrt in delay slot if
5378         fix_lost_divsqrt.
5379         * config/sparc/sparc.opt (sparc_fix_lost_divsqrt): New variable.
5381 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5383         * config/sparc/sparc.c (atomic_insn_p): New function.
5384         (sparc_do_work_around_errata): Insert NOP instructions to
5385         prevent sequences that could trigger the TN-0010 errata for
5386         UT700.
5387         * config/sparc/sync.md (atomic_compare_and_swap_leon3_1): Make
5388         instruction referable in atomic_insns_p.
5390 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5392         * config/sparc/sync.md (swapsi): 16-byte align if sparc_fix_gr712rc.
5393         (atomic_compare_and_swap_leon3_1): Likewise.
5394         (ldstub): Likewise.
5396 2017-11-29  Daniel Cederman  <cederman@gaisler.com>
5398         * config/sparc/sparc.c (fpop_insn_p): New function.
5399         (sparc_do_work_around_errata): Insert NOP instructions to
5400         prevent sequences that could trigger the TN-0012 errata for
5401         GR712RC.
5402         (pass_work_around_errata::gate): Also test sparc_fix_gr712rc.
5403         * config/sparc/sparc.md (fix_gr712rc): New attribute.
5404         (in_branch_annul_delay): Prevent floating-point instructions
5405         in delay slot of annulled integer branch.
5407 2017-11-29  Richard Biener  <rguenther@suse.de>
5409         PR tree-optimization/83202
5410         * tree-vect-slp.c (scalar_stmts_set_t): New typedef.
5411         (bst_fail): Use it.
5412         (vect_analyze_slp_cost_1): Add visited set, do not account SLP
5413         nodes vectorized to the same stmts multiple times.
5414         (vect_analyze_slp_cost): Allocate a visited set and pass it down.
5415         (vect_analyze_slp_instance): Adjust.
5416         (scalar_stmts_to_slp_tree_map_t): New typedef.
5417         (vect_schedule_slp_instance): Add a map recording the SLP node
5418         representing the vectorized stmts for a set of scalar stmts.
5419         Avoid code-generating redundancies.
5420         (vect_schedule_slp): Allocate map and pass it down.
5422 2017-11-29  Nathan Sidwell  <nathan@acm.org>
5424         PR c++/83187
5425         * tree.c (build_complex_type): Fix canonicalization.  Only fill in
5426         type if it is new.
5428 2017-11-29  Wilco Dijkstra  <wdijkstr@arm.com>
5430         * config/aarch64/aarch64.c (aarch64_print_operand): Add new
5431         cases for printing LDP/STP memory addresses.
5432         (aarch64_print_address_internal): Renamed from
5433         aarch64_print_operand_address, added parameter, add Pmode check.
5434         (aarch64_print_ldpstp_address): New function for LDP/STP addresses.
5435         (aarch64_print_operand_address): Indirect to
5436         aarch64_print_address_internal.
5437         * config/aarch64/aarch64-simd.md (store_pair_lanes): Use new
5438         'y' operand output specifier.
5440 2017-11-29  Jakub Jelinek  <jakub@redhat.com>
5442         PR middle-end/83185
5443         * tree.c (build_simple_mem_ref_loc): Handle
5444         get_addr_base_and_unit_offset returning a MEM_REF.
5446         PR middle-end/80929
5447         * rtlanal.c (seq_cost): For non-single_set insns try to use insn_cost.
5449         PR target/80819
5450         * config/i386/sse.md (vec_concatv2di): Remove * from (=Yr,0,*rm)
5451         alternative.
5453 2017-11-28  David Malcolm  <dmalcolm@redhat.com>
5455         * diagnostic-show-locus.c (layout::print_trailing_fixits): Handle
5456         m_x_offset.
5457         (layout::move_to_column): Likewise.
5459 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5461         PR sanitizer/81275
5462         * tree.c (block_may_fallthru): Return false if SWITCH_ALL_CASES_P
5463         is set on SWITCH_EXPR and !block_may_fallthru (SWITCH_BODY ()).
5465 2017-11-28  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
5466             Martin Jambor  <mjambor@suse.cz>
5468         PR ipa/82808
5469         * tree.h (expr_type_first_operand_type_p): Declare
5470         * tree.c (expr_type_first_operand_type_p): New function.
5471         * ipa-prop.h (ipa_get_type): Allow i to be out of bounds.
5472         (ipa_value_from_jfunc): Adjust declaration.
5473         * ipa-cp.c (ipa_get_jf_pass_through_result): New parameter RES_TYPE.
5474         Use it as result type for arithmetics, unless it is NULL in which case
5475         be more conservative.
5476         (ipa_value_from_jfunc): New parameter PARM_TYPE, pass it to
5477         ipa_get_jf_pass_through_result.
5478         (propagate_vals_across_pass_through): Likewise.
5479         (propagate_scalar_across_jump_function): New parameter PARM_TYPE, pass
5480         is to propagate_vals_across_pass_through.
5481         (propagate_constants_across_call): Pass PARM_TYPE to
5482         propagate_scalar_across_jump_function.
5483         (find_more_scalar_values_for_callers_subset): Pass parameter type to
5484         ipa_value_from_jfunc.
5485         (cgraph_edge_brings_all_scalars_for_node): Likewise.
5486         * ipa-fnsummary.c (evaluate_properties_for_edge): Renamed parms_info
5487         to caller_parms_info, pass parameter type to ipa_value_from_jfunc.
5488         * ipa-prop.c (try_make_edge_direct_simple_call): New parameter
5489         target_type, pass it to ipa_value_from_jfunc.
5490         (update_indirect_edges_after_inlining): Pass parameter type to
5491         try_make_edge_direct_simple_call.
5493 2017-11-28  Jeff Law  <law@redhat.com>
5495         * gimple-ssa-evrp-analyze.c
5496         (evrp_range_analyzer::record_ranges_from_phis): Only use SCEV to
5497         refine ranges if scev_initialized_p returns true.
5498         * vr-values.c (vr_values::extract_range_from_phi_node): Likewise.
5500 2017-11-28  Julia Koval  <julia.koval@intel.com>
5502         * config/i386/avx512vbmi2intrin.h (_mm512_shrdi_epi16,
5503         _mm512_mask_shrdi_epi16, _mm512_maskz_shrdi_epi16, _mm512_shrdi_epi32,
5504         _mm512_mask_shrdi_epi32, _mm512_maskz_shrdi_epi32, _mm512_shrdi_epi64,
5505         _mm512_mask_shrdi_epi64, _mm512_maskz_shrdi_epi64): New intrinsics.
5506         * config/i386/avx512vbmi2vlintrin.h (_mm256_shrdi_epi16,
5507         _mm256_mask_shrdi_epi16, _mm256_maskz_shrdi_epi16,
5508         _mm256_mask_shrdi_epi32, _mm256_maskz_shrdi_epi32, _mm256_shrdi_epi32,
5509         _mm256_mask_shrdi_epi64, _mm256_maskz_shrdi_epi64, _mm256_shrdi_epi64,
5510         _mm_mask_shrdi_epi16, _mm_maskz_shrdi_epi16, _mm_shrdi_epi16,
5511         _mm_mask_shrdi_epi32, _mm_maskz_shrdi_epi32, _mm_shrdi_epi32,
5512         _mm_mask_shrdi_epi64, _mm_maskz_shrdi_epi64, _mm_shrdi_epi64): Ditto.
5513         * config/i386/i386-builtin.def (__builtin_ia32_vpshrd_v32hi,
5514         __builtin_ia32_vpshrd_v32hi_mask, __builtin_ia32_vpshrd_v16hi,
5515         __builtin_ia32_vpshrd_v16hi_mask, __builtin_ia32_vpshrd_v8hi,
5516         __builtin_ia32_vpshrd_v8hi_mask, __builtin_ia32_vpshrd_v16si,
5517         __builtin_ia32_vpshrd_v16si_mask, __builtin_ia32_vpshrd_v8si,
5518         __builtin_ia32_vpshrd_v8si_mask, __builtin_ia32_vpshrd_v4si,
5519         __builtin_ia32_vpshrd_v4si_mask, __builtin_ia32_vpshrd_v8di,
5520         __builtin_ia32_vpshrd_v8di_mask, __builtin_ia32_vpshrd_v4di,
5521         __builtin_ia32_vpshrd_v4di_mask, __builtin_ia32_vpshrd_v2di,
5522         __builtin_ia32_vpshrd_v2di_mask): New builtins.
5523         * config/i386/sse.md (vpshrd_<mode><mask_name>): New pattern.
5525 2017-11-28  Julia Koval  <julia.koval@intel.com>
5527         * config/i386/avx512vbmi2intrin.h (_mm512_shldi_epi16,
5528         _mm512_mask_shldi_epi16, _mm512_maskz_shldi_epi16, _mm512_shldi_epi32,
5529         _mm512_mask_shldi_epi32, _mm512_maskz_shldi_epi32, _mm512_shldi_epi64,
5530         _mm512_mask_shldi_epi64, _mm512_maskz_shldi_epi64): New intrinsics.
5531         * config/i386/avx512vbmi2vlintrin.h (_mm256_shldi_epi16,
5532         _mm256_mask_shldi_epi16, _mm256_maskz_shldi_epi16,
5533         _mm256_mask_shldi_epi32, _mm256_maskz_shldi_epi32, _mm256_shldi_epi32,
5534         _mm256_mask_shldi_epi64, _mm256_maskz_shldi_epi64, _mm256_shldi_epi64,
5535         _mm_mask_shldi_epi16, _mm_maskz_shldi_epi16, _mm_shldi_epi16,
5536         _mm_mask_shldi_epi32, _mm_maskz_shldi_epi32, _mm_shldi_epi32,
5537         _mm_mask_shldi_epi64, _mm_maskz_shldi_epi64, _mm_shldi_epi64): Ditto.
5538         * config/i386/i386-builtin-types.def (V32HI_FTYPE_V32HI_V32HI_INT,
5539         V32HI_FTYPE_V32HI_V32HI_INT_V32HI_INT, V16SI_FTYPE_V16SI_V16SI_INT,
5540         V16SI_FTYPE_V16SI_V16SI_INT_V16SI_INT,
5541         V8DI_FTYPE_V8DI_V8DI_INT_V8DI_INT, V8SI_FTYPE_V8SI_V8SI_INT_V8SI_INT,
5542         V16HI_FTYPE_V16HI_V16HI_INT_V16HI_INT,
5543         V4DI_FTYPE_V4DI_V4DI_INT_V4DI_INT,
5544         V8HI_FTYPE_V8HI_V8HI_INT_V8HI_INT,
5545         V4SI_FTYPE_V4SI_V4SI_INT_V4SI_INT,
5546         V2DI_FTYPE_V2DI_V2DI_INT_V2DI_INT): New types.
5547         * config/i386/i386-builtin.def (__builtin_ia32_vpshld_v32hi,
5548         __builtin_ia32_vpshld_v32hi_mask, __builtin_ia32_vpshld_v16hi,
5549         __builtin_ia32_vpshld_v16hi_mask, __builtin_ia32_vpshld_v8hi,
5550         __builtin_ia32_vpshld_v8hi_mask, __builtin_ia32_vpshld_v16si,
5551         __builtin_ia32_vpshld_v16si_mask, __builtin_ia32_vpshld_v8si,
5552         __builtin_ia32_vpshld_v8si_mask, __builtin_ia32_vpshld_v4si,
5553         __builtin_ia32_vpshld_v4si_mask, __builtin_ia32_vpshld_v8di,
5554         __builtin_ia32_vpshld_v8di_mask, __builtin_ia32_vpshld_v4di,
5555         __builtin_ia32_vpshld_v4di_mask, __builtin_ia32_vpshld_v2di,
5556         __builtin_ia32_vpshld_v2di_mask): New builtins.
5557         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
5558         * config/i386/sse.md (vpshld_<mode><mask_name>): New pattern.
5560 2017-11-28  Richard Biener  <rguenther@suse.de>
5562         PR tree-optimization/80776
5563         * gimple-ssa-evrp-analyze.h (evrp_range_analyzer::set_ssa_range_info):
5564         Declare.
5565         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::set_ssa_range_info):
5566         New function.
5567         (evrp_range_analyzer::record_ranges_from_incoming_edges):
5568         If the incoming edge is an effective fallthru because the other
5569         edge only reaches a __builtin_unreachable () then record ranges
5570         derived from the controlling condition in SSA info.
5571         (evrp_range_analyzer::record_ranges_from_phis): Use set_ssa_range_info.
5572         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
5574 2017-11-28  Olivier Hainque  <hainque@adacore.com>
5576         * Makefile.in (SELFTEST_FLAGS): Use nul instead of /dev/null
5577         on mingw build hosts.
5579 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5581         PR debug/81307
5582         * dbxout.c (dbx_block_with_cold_children): Fix function comment.
5584 2017-11-28  Richard Biener  <rguenther@suse.de>
5586         PR middle-end/83141
5587         * gimple-fold.c (gimple_fold_builtin_memory_op): For aggregate
5588         copies generated from memcpy use a character array as reference
5589         type.
5591 2017-11-28  Julia Koval  <julia.koval@intel.com>
5592             Sebastian Peryt  <sebastian.peryt@intel.com>
5594         * Makefile.in (cilkplus.def, cilk-builtins.def, c-family/cilk.o,
5595         c-family/c-cilkplus.o, c-family/array-notation-common.o,
5596         cilk-common.o, cilk.h, cilk-common.c): Remove.
5597         * builtin-types.def
5598         (BT_FN_INT_PTR_PTR_PTR_FTYPE_BT_INT_BT_PTR_BT_PTR_BT_PTR): Remove.
5599         * builtins.c (is_builtin_name): Remove cilkplus condition.
5600         (BUILT_IN_CILK_DETACH, BUILT_IN_CILK_POP_FRAME): Remove.
5601         * builtins.def (DEF_CILK_BUILTIN_STUB, DEF_CILKPLUS_BUILTIN,
5602         cilk-builtins.def, cilkplus.def): Remove.
5603         * cif-code.def (CILK_SPAWN): Remove.
5604         * cilk-builtins.def: Delete.
5605         * cilk-common.c: Ditto.
5606         * cilk.h: Ditto.
5607         * cilkplus.def: Ditto.
5608         * config/darwin.h (fcilkplus): Delete.
5609         * cppbuiltin.c: Ditto.
5610         * doc/extend.texi: Remove cilkplus doc.
5611         * doc/generic.texi: Ditto.
5612         * doc/invoke.texi: Ditto.
5613         * doc/passes.texi: Ditto.
5614         * gcc.c (fcilkplus): Remove.
5615         * gengtype.c (cilk.h): Remove.
5616         * gimple-pretty-print.c (dump_gimple_omp_for): Remove cilkplus
5617         support.
5618         * gimple.h (GF_OMP_FOR_KIND_CILKFOR, GF_OMP_FOR_KIND_CILKSIMD):
5619         Remove.
5620         * gimplify.c (gimplify_return_expr, maybe_fold_stmt,
5621         gimplify_call_expr,
5622         is_gimple_stmt, gimplify_modify_expr, gimplify_scan_omp_clauses,
5623         gimplify_adjust_omp_clauses, gimplify_omp_for, gimplify_expr): Remove
5624         cilkplus conditions.
5625         * ipa-fnsummary.c (ipa_dump_fn_summary, compute_fn_summary,
5626         inline_read_section): Ditto.
5627         * ipa-inline-analysis.c (cilk.h): Remove.
5628         * ira.c (ira_setup_eliminable_regset): Remove cilkplus support.
5629         * lto-wrapper.c (merge_and_complain, append_compiler_options,
5630         append_linker_options): Remove condition for fcilkplus.
5631         * lto/lto-lang.c (cilk.h): Remove.
5632         (lto_init): Remove condition for fcilkplus.
5633         * omp-expand.c (expand_cilk_for_call): Delete.
5634         (expand_omp_taskreg, expand_omp_for_static_chunk,
5635         expand_omp_for): Remove cilkplus
5636         conditions.
5637         (expand_cilk_for): Delete.
5638         * omp-general.c (omp_extract_for_data): Remove cilkplus support.
5639         * omp-low.c (scan_sharing_clauses, create_omp_child_function,
5640         execute_lower_omp, diagnose_sb_0): Ditto.
5641         * omp-simd-clone.c (simd_clone_clauses_extract): Ditto.
5642         * tree-core.h (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5643         * tree-nested.c: Ditto.
5644         * tree-pretty-print.c (dump_omp_clause): Remove cilkplus support.
5645         (dump_generic_node): Ditto.
5646         * tree.c (OMP_CLAUSE__CILK_FOR_COUNT_): Delete.
5647         * tree.def (cilk_simd, cilk_for, cilk_spawn_stmt, cilk_sync_stmt):
5648         Delete.
5649         * tree.h (CILK_SPAWN_FN, EXPR_CILK_SPAWN): Delete.
5651 2017-11-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
5653         * config/aarch64/aarch64.md (div<mode>3): Change check to TARGET_FLOAT.
5654         * config/aarch64/aarch64.c (aarch64_emit_approx_div): Add early exit
5655         for vector mode and !TARGET_SIMD.
5657 2017-11-28  Jakub Jelinek  <jakub@redhat.com>
5659         * tree.def (SWITCH_EXPR): Change from 3 operand to 2 operand tree.
5660         Adjust comment.
5661         * tree.h (SWITCH_LABELS): Remove.
5662         * gimplify.c (gimplify_switch_expr): Don't test SWITCH_LABELS,
5663         assert SWITCH_BODY is non-NULL.
5664         * tree-pretty-print.c (dump_generic_node): Remove SWITCH_LABELS
5665         handling.
5666         * tree.c (block_may_fallthru): Always return true; for SWITCH_EXPR.
5668         PR tree-optimization/80788
5669         * match.pd (X +- C1 CMP C2 -> X CMP C2 -+ C1): If res
5670         has TREE_OVERFLOW set, call drop_tree_overflow.
5672 2017-11-28  Richard Biener  <rguenther@suse.de>
5674         PR tree-optimization/83158
5675         * tree-vrp.c (intersect_ranges): Prefer ~[0, 0] in a few more cases.
5677 2017-11-28  Segher Boessenkool  <segher@kernel.crashing.org>
5679         PR 81288/target
5680         * config/rs6000/rs6000.c (rs6000_rtx_costs): Do not handle
5681         TARGET_ISEL && !TARGET_MFCRF differently.  Simplify code.
5683 2017-11-27  Segher Boessenkool  <segher@kernel.crashing.org>
5685         * config/rs6000/rs6000.md (<code><GPR:mode><GPR2:mode>2_isel): Change
5686         LT/GT/LTU/GTU to LE/GE/LEU/GEU where possible.
5688 2017-11-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
5690         PR middle_end/82333
5691         * varasm.c (compare_constant): Take the mode of the constants into
5692         account when comparing floating point constants.
5694 2017-11-27  Gerald Pfeifer  <gerald@pfeifer.com>
5696         * hash-set.h (DEFINE_DEBUG_HASH_SET): Remove static qualifier
5697         from explicit instantiation of debug_helper.
5698         * vec.h (DEFINE_DEBUG_VEC): Ditto.
5700 2017-11-27  Richard Biener  <rguenther@suse.de>
5702         * gimple-fold.c (gimple_fold_builtin_memory_op): Remove dead code,
5703         refactor a bit.
5705 2017-11-27  Richard Biener  <rguenther@suse.de>
5707         * tree.c (wide_int_to_tree): Free discarded INTEGER_CST.
5708         (type_hash_canon): Also clear int_cst_hash_table entry for
5709         TYPE_MIN/MAX_VALUE.
5710         (build_nonstandard_integer_type): Hash all TYPE_MAX_VALUEs.
5712 2017-11-27  Tamar Christina  <tamar.christina@arm.com>
5714         * doc/extend.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5715         * doc/invoke.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5716         * doc/sourcebuild.texi: Add -A suffix (ARMv8*-A, ARMv7-A).
5718 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5720         * hash-map.h (gt_cleare_cache): Avoid UB.
5722 2017-11-27  Eric Botcazou  <ebotcazou@adacore.com>
5724         * cfgloop.h (struct loop): Document usage of USHRT_MAX for unroll.
5725         * loop-unroll.c (decide_unroll_constant_iterations): Implement it.
5726         (decide_unroll_runtime_iterations): Likewise.
5727         (decide_unroll_stupid): Likewise.
5729 2017-11-27  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
5731         PR target/83109
5732         * config/i386/i386.md: Add a loop with incssp.
5734 2017-11-27  Martin Jambor  <mjambor@suse.cz>
5736         PR tree-optimization/81248
5737         * tree-sra.c (splice_param_accesses): Remove size check.
5738         (decide_one_param_reduction): Fix size check.
5739         * gimple-pretty-print.c (dump_profile): Silence warning.
5740         * params.def (PARAM_IPA_SRA_PTR_GROWTH_FACTOR): Adjust description.
5742 2017-11-27  Jakub Jelinek  <jakub@redhat.com>
5744         PR debug/81307
5745         * dbxout.c (lastlineno): New variable.
5746         (dbx_debug_hooks): Use dbxout_switch_text_section as
5747         switch_text_section debug hook.
5748         (dbxout_function_end): Switch to current_function_section
5749         rather than function_section.  If crtl->has_bb_partition,
5750         output just one N_FUN, depending on in_cold_section_p.
5751         (dbxout_source_line): Remember last lineno in lastlineno.
5752         (dbxout_switch_text_section): New function.
5753         (dbxout_function_decl): Adjust dbxout_block caller.
5754         (dbx_block_with_cold_children): New function.
5755         (dbxout_block): Return true if any LBRAC/RBRAC have been
5756         emitted.  Use dbx_block_with_cold_children at depth == 0
5757         in second partition.  Add PARENT_BLOCKNUM argument, pass
5758         it optionally adjusted to children.  Output LBRAC/RBRAC
5759         around recursive call only if the block is in the current
5760         partition, if not and anything was output, emit empty
5761         range LBRAC/RBRAC.
5762         * final.c (final_scan_insn): Compute cold_function_name
5763         before calling switch_text_section debug hook.  Call
5764         that hook even if dwarf2out_do_frame if not emitting
5765         dwarf debug info.
5767         PR target/83100
5768         * varasm.c (bss_initializer_p): Return true for DECL_COMMON
5769         TREE_READONLY decls.
5771 2017-11-27  Markus Trippelsdorf  <markus@trippelsdorf.de>
5773         PR rtl-optimization/82488
5774         * expr.c (fixup_args_size_notes): Avoid signed integer overflow.
5776 2017-11-26  Julia Koval  <julia.koval@intel.com>
5778         * config/i386/i386.c (processor_target_table): Add skylake_cost for
5779         skylake-avx512.
5780         * config/i386/x86-tune-costs.h (skylake_memcpy, skylake_memset,
5781         skylake_cost): New.
5783 2017-11-26  Julia Koval  <julia.koval@intel.com>
5785         * config/i386/driver-i386.c (host_detect_local_cpu):
5786         Detect skylake-avx512.
5788 2017-11-26  Julia Koval  <julia.koval@intel.com>
5790         * config.gcc: Add -march=cannonlake.
5791         * config/i386/driver-i386.c (host_detect_local_cpu): Detect cannonlake.
5792         * config/i386/i386-c.c (ix86_target_macros_internal): Handle cannonlake.
5793         * config/i386/i386.c (processor_costs): Add m_CANNONLAKE.
5794         (PTA_CANNONLAKE): New.
5795         (processor_target_table): Add cannonlake.
5796         (ix86_option_override_internal): Ditto.
5797         (fold_builtin_cpu): Ditto.
5798         (get_builtin_code_for_version): Handle cannonlake.
5799         (M_INTEL_COREI7_CANNONLAKE): New.
5800         * config/i386/i386.h (TARGET_CANNONLAKE, PROCESSOR_CANNONLAKE): New.
5801         * doc/invoke.texi: Add -march=cannonlake.
5803 2017-11-14  Boris Kolpackov  <boris@codesynthesis.com>
5805         * plugin.c (add_new_plugin): Use platform-specific library extensions.
5806         (try_init_one_plugin): Alternative implementation for MinGW.
5807         * Makefile.in (plugin_implib): New.
5808         (gengtype-lex.c): Fix broken AIX workaround.
5809         * configure: Regenerate.
5810         * doc/plugins.texi: Document support for MinGW.
5812 2017-11-25  Jakub Jelinek  <jakub@redhat.com>
5814         PR rtl-optimization/81553
5815         * combine.c (simplify_if_then_else): In (if_then_else COND (OP Z C1) Z)
5816         to (OP Z (mult COND (C1 * STORE_FLAG_VALUE))) optimization, if OP
5817         is a shift where C1 has different mode than the whole shift, use C1's
5818         mode for MULT rather than the shift's mode.
5820         PR target/82848
5821         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Don't fold
5822         builtins not enabled in the currently selected ISA.
5824 2017-11-24  Jackson Woodruff  <jackson.woodruff@arm.com>
5826         PR tree-optimization/71026
5827         * tree-ssa-math-opts (is_division_by_square, is_square_of): New.
5828         (insert_reciprocals): Change to insert reciprocals before a division
5829         by a square and to insert the square of a reciprocal.
5830         (execute_cse_reciprocals_1): Change to consider division by a square.
5831         (register_division_in): Add importance parameter.
5833 2017-11-24  Richard Biener  <rguenther@suse.de>
5835         PR tree-optimization/82402
5836         * tree-vect-loop-manip.c (create_lcssa_for_virtual_phi): Properly
5837         set SSA_NAME_OCCURS_IN_ABNORMAL_PHI.
5839 2017-11-24  Marc Glisse  <marc.glisse@inria.fr>
5841         * match.pd (0-ptr): New transformation.
5843 2017-11-24  Jan Hubicka  <hubicka@ucw.cz>
5845         PR bootstrap/83015
5846         * ipa-inline.c (inline_small_functions): Set current badnes correctly
5847         when skipping checking.
5849 2017-11-24  Richard Biener  <rguenther@suse.de>
5851         PR tree-optimization/83128
5852         * tree-ssa-sccvn.c (fully_constant_vn_reference_p): Handle STRING_CSTs.
5853         (vn_reference_lookup_3): Likewise.
5855 2017-11-24  Jakub Jelinek  <jakub@redhat.com>
5857         PR sanitizer/83014
5858         * ubsan.c (ubsan_type_descriptor): Use pp_unsigned_wide_integer
5859         instead of pp_printf with HOST_WIDE_INT_PRINT_DEC.  Avoid calling
5860         tree_to_uhwi twice.
5862         * tree-object-size.c (pass_through_call): Use gimple_call_return_flags
5863         ERF_RETURN*ARG* for builtins other than BUILT_IN_ASSUME_ALIGNED,
5864         check for the latter with gimple_call_builtin_p.  Do not handle
5865         BUILT_IN_STPNCPY_CHK which is not a pass through call.
5867 2017-11-24  Christophe Lyon  <christophe.lyon@linaro.org>
5869         * config/arm/arm_neon.h: Fix pragma GCC push_options before
5870         vdot_u32.
5872 2017-11-23  Julia Koval  <julia.koval@intel.com>
5874         * config/i386/avx512vbmi2intrin.h (_mm512_mask_expand_epi8,
5875         _mm512_maskz_expand_epi8, _mm512_mask_expandloadu_epi8,
5876         _mm512_maskz_expandloadu_epi8, _mm512_mask_expand_epi16,
5877         _mm512_maskz_expand_epi16, _mm512_mask_expandloadu_epi16,
5878         _mm512_maskz_expandloadu_epi16): New intrinsics.
5879         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_expand_epi8,
5880         _mm_maskz_expand_epi8, _mm_mask_expandloadu_epi8,
5881         _mm_maskz_expandloadu_epi8, _mm_mask_expand_epi16,
5882         _mm_maskz_expand_epi16, _mm_mask_expandloadu_epi16,
5883         _mm_maskz_expandloadu_epi16, _mm256_mask_expand_epi16,
5884         _mm256_maskz_expand_epi16, _mm256_mask_expandloadu_epi16,
5885         _mm256_maskz_expandloadu_epi16, _mm256_mask_expand_epi8,
5886         _mm256_maskz_expand_epi8, _mm256_mask_expandloadu_epi8,
5887         _mm256_maskz_expandloadu_epi8): New intrinsics.
5888         * config/i386/i386-builtin-types.def (V64QI_FTYPE_PCV64QI_V64QI_UDI,
5889         V32HI_FTYPE_PCV32HI_V32HI_USI, V32QI_FTYPE_PCV32QI_V32QI_USI,
5890         V16HI_FTYPE_PCV16HI_V16HI_UHI, V16QI_FTYPE_PCV16QI_V16QI_UHI,
5891         V8HI_FTYPE_PCV8HI_V8HI_UQI): New types.
5892         * config/i386/i386.c (ix86_expand_special_args_builtin): Use new types.
5893         * config/i386/sse.md (VI248_VLBW): New iterator.
5894         (expand<mode>_mask, expand<mode>_maskz): New patterns.
5896 2017-11-23  Julia Koval  <julia.koval@intel.com>
5898         * config.gcc (avx512vbmi2intrin.h, avx512vbmi2vlintrin): New headers.
5899         * config/i386/avx512vbmi2intrin.h (_mm512_mask_compress_epi8,
5900         _mm512_maskz_compress_epi8, _mm512_mask_compressstoreu_epi8,
5901         _mm512_mask_compress_epi16, _mm512_maskz_compress_epi16,
5902         _mm512_mask_compressstoreu_epi16): New.
5903         * config/i386/avx512vbmi2vlintrin.h (_mm_mask_compress_epi8,
5904         _mm_maskz_compress_epi8, _mm256_mask_compressstoreu_epi16,
5905         _mm_mask_compress_epi16, _mm_maskz_compress_epi16,
5906         _mm256_mask_compress_epi16, _mm256_maskz_compress_epi16,
5907         _mm_mask_compressstoreu_epi8, _mm_mask_compressstoreu_epi16,
5908         _mm256_mask_compress_epi8, _mm256_maskz_compress_epi8,
5909         _mm256_mask_compressstoreu_epi8): New.
5910         * config/i386/i386-builtin-types.def (VOID_FTYPE_PV64QI_V64QI_UDI,
5911         VOID_FTYPE_PV32HI_V32HI_USI, VOID_FTYPE_PV32QI_V32QI_USI,
5912         VOID_FTYPE_PV16QI_V16QI_UHI, VOID_FTYPE_PV16HI_V16HI_UHI,
5913         VOID_FTYPE_PV8HI_V8HI_UQI): New types.
5914         * config/i386/i386-builtin.def (__builtin_ia32_compressqi512_mask,
5915         __builtin_ia32_compresshi512_mask, __builtin_ia32_compressqi256_mask,
5916         __builtin_ia32_compressqi128_mask, __builtin_ia32_compresshi256_mask,
5917         __builtin_ia32_compresshi128_mask,
5918         __builtin_ia32_compressstoreuqi512_mask,
5919         __builtin_ia32_compressstoreuhi512_mask,
5920         __builtin_ia32_compressstoreuqi256_mask,
5921         __builtin_ia32_compressstoreuqi128_mask,
5922         __builtin_ia32_compressstoreuhi256_mask,
5923         __builtin_ia32_compressstoreuhi128_mask): New builtins.
5924         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Create special args
5925         array for flags2.
5926         (ix86_expand_special_args_builtin): Handle new types.
5927         (s4fma_expand): Handle new builtin array.
5928         * config/i386/immintrin.h: Include new headers.
5929         * config/i386/sse.md (VI12_AVX512VLBW): New iterator.
5930         (compress<mode>_mask, compressstore<mode>_mask): New patterns.
5932 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
5934         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Formatting
5935         fixes.  Declare temp and g variables at the top in order to avoid
5936         {} in most of the cases.
5938 2017-11-23  Marc Glisse  <marc.glisse@inria.fr>
5940         * match.pd (ptr-0): New transformation.
5942 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
5944         * config/arm/arm-protos.h (enum arm_addr_mode_op): New.
5945         (struct addr_mode_cost_table): New.
5946         (struct tune_params): Add field addr_mode_costs.
5947         * config/arm/arm.c (generic_addr_mode_costs): New.
5948         (arm_slowmul_tune): Initialise addr_mode_costs field.
5949         (arm_fastmul_tune): Likewise.
5950         (arm_strongarm_tune): Likewise.
5951         (arm_xscale_tune): Likewise.
5952         (arm_9e_tune): Likewise.
5953         (arm_marvell_pj4_tune): Likewise.
5954         (arm_v6t2_tune): Likewise.
5955         (arm_cortex_tune): Likewise.
5956         (arm_cortex_a8_tune): Likewise.
5957         (arm_cortex_a7_tune): Likewise.
5958         (arm_cortex_a15_tune): Likewise.
5959         (arm_cortex_a35_tune): Likewise.
5960         (arm_cortex_a53_tune): Likewise.
5961         (arm_cortex_a57_tune): Likewise.
5962         (arm_exynosm1_tune): Likewise.
5963         (arm_xgene1_tune): Likewise.
5964         (arm_cortex_a5_tune): Likewise.
5965         (arm_cortex_a9_tune): Likewise.
5966         (arm_cortex_a12_tune): Likewise.
5967         (arm_cortex_a73_tune): Likewise.
5968         (arm_v7m_tune): Likewise.
5969         (arm_cortex_m7_tune): Likewise.
5970         (arm_v6m_tune): Likewise.
5971         (arm_fa726te_tune): Likewise.
5972         (arm_mem_costs): Use table lookup to calculate cost of addressing mode.
5974 2017-11-23  Charles Baylis  <charles.baylis@linaro.org>
5976         * config/arm/arm.c (arm_mem_costs): New function.
5977         (arm_rtx_costs_internal): Use arm_mem_costs.
5979 2017-11-23  Mark Wielaard  <mark@klomp.org>
5981         * dwarf2out.c (init_sections_and_labels): Use generation to create
5982         unique ranges_section_label and ranges_base_label. Return generation.
5983         (output_rnglists): Add generation argument. Use generation to create
5984         unique ranges labels.
5985         (dwarf2out_finish): Get generation from init_sections_and_labels
5986         and pass generation to output_rnglists.
5988 2017-11-23  Mike Stump  <mikestump@comcast.net>
5989             Eric Botcazou  <ebotcazou@adacore.com>
5991         * doc/generic.texi (ANNOTATE_EXPR): Document 3rd operand.
5992         * cfgloop.h (struct loop): Add unroll field.
5993         * function.h (struct function): Add has_unroll bitfield.
5994         * gimplify.c (gimple_boolify) <ANNOTATE_EXPR>: Deal with unroll kind.
5995         (gimplify_expr) <ANNOTATE_EXPR>: Propagate 3rd operand.
5996         * loop-init.c (pass_loop2::gate): Return true if cfun->has_unroll.
5997         (pass_rtl_unroll_loops::gate): Likewise.
5998         * loop-unroll.c (decide_unrolling): Tweak note message.  Skip loops
5999         for which loop->unroll==1.
6000         (decide_unroll_constant_iterations): Use note for consistency and
6001         take loop->unroll into account.  Return early if loop->unroll is set.
6002         Fix thinko in existing test.
6003         (decide_unroll_runtime_iterations): Use note for consistency and
6004         take loop->unroll into account.
6005         (decide_unroll_stupid): Likewise.
6006         * lto-streamer-in.c (input_cfg): Read loop->unroll.
6007         * lto-streamer-out.c (output_cfg): Write loop->unroll.
6008         * tree-cfg.c (replace_loop_annotate_in_block) <annot_expr_unroll_kind>:
6009         New case.
6010         (replace_loop_annotate) <annot_expr_unroll_kind>: Likewise.
6011         (print_loop): Print loop->unroll if set.
6012         * tree-core.h (enum annot_expr_kind): Add annot_expr_unroll_kind.
6013         * tree-inline.c (copy_loops): Copy unroll and set cfun->has_unroll.
6014         * tree-pretty-print.c (dump_generic_node) <annot_expr_unroll_kind>:
6015         New case.
6016         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Bail out if
6017         loop->unroll is set and smaller than the trip count.  Otherwise bypass
6018         entirely the heuristics if loop->unroll is set.  Remove dead note.
6019         Fix off-by-one bug in other note.
6020         (try_peel_loop): Bail out if loop->unroll is set.  Fix formatting.
6021         (tree_unroll_loops_completely_1): Force unrolling if loop->unroll
6022         is greater than 1.
6023         (tree_unroll_loops_completely): Make static.
6024         (pass_complete_unroll::execute): Use correct type for variable.
6025         (pass_complete_unrolli::execute): Fix formatting.
6026         * tree.def (ANNOTATE_EXPR): Add 3rd operand.
6028 2017-11-23  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6030         * config/i386/i386.h (TARGET_PREFER_AVX256): Also
6031         enable when TARGET_PREFER_AVX128 is set.
6033 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6035         * ipa-profile.c (ipa_propagate_frequency_1): Use count instead of
6036         frequency.
6037         * cgraph.c (cgraph_edge::dump_edge_flags): Dump sreal frequencies.
6038         (cgraph_edge::maybe_hot_p): Use sreal frequencies.
6040 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6042         * ipa-fnsummary.c (record_modified_bb_info): Use sreal
6043         frequencies. Fix estimation of aggregate parameters.
6045 2017-11-23  Jan Hubicka  <hubicka@ucw.cz>
6047         * cgraphclones.c (cgraph_node::create_clone): Fix updating of profile
6048         when inlining.
6050 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6052         * config/sh/sh.h (ASM_OUTPUT_ADDR_VEC_ELT): Wrap in "do {} while (0)".
6054 2017-11-23  Tom de Vries  <tom@codesourcery.com>
6056         * config/ft32/ft32.h (ASM_OUTPUT_ADDR_VEC_ELT): Remove semicolon after
6057         macro.
6059 2017-11-23  Oleg Endo  <olegendo@gcc.gnu.org>
6061         PR target/83111
6062         * config/sh/sh.md (udivsi3, divsi3, sibcall_value_pcrel,
6063         sibcall_value_pcrel_fdpic): Use local variable instead of
6064         operands[3].
6065         (calli_tbr_rel): Add missing operand 2.
6066         (call_valuei_tbr_rel): Add missing operand 3.
6068 2017-11-23  Jakub Jelinek  <jakub@redhat.com>
6070         PR middle-end/82253
6071         * expr.c (expand_assignment): For CONCAT to_rtx, complex type from and
6072         bitpos/bitsize covering the whole destination, use store_expr only if
6073         the complex mode is the same.  Otherwise, use expand_normal and if
6074         it returns CONCAT, subreg each part separately instead of trying to
6075         subreg the whole result.
6077 2017-11-23  Richard Biener  <rguenther@suse.de>
6079         PR tree-optimization/23094
6080         * tree-ssa-sccvn.c (vuse_ssa_val): Handle VN_TOP when we
6081         come here from walking over backedges in the first iteration.
6082         (vn_reference_lookup_3): Skip clobbers that store the same value.
6084 2017-11-23  Richard Biener  <rguenther@suse.de>
6086         PR tree-optimization/81403
6087         * tree-ssa-pre.c (get_representative_for): Add parameter specifying
6088         a block we need a leader relative to.
6089         (phi_translate_1): For nary processing require a leader from
6090         get_representative_for given we run expression simplification
6091         using match-and-simplify.  Remove previous fix.
6093 2017-11-22  Jeff Law  <law@redhat.com>
6095         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::try_find_new_range):
6096         Use new method allocate_value_range rather than accessing the
6097         vrp_value_range_pool data member directly.
6098         * tree-vrp.c (simplify_stmt_for_jump_threading): Tweak slightly
6099         to use extract_range_from_stmt method to avoid need for
6100         extract_range_from_assignment method.
6101         (vrp_prop::vrp_finalize): Use set_lattice_propagation_complete
6102         method rather than setting values_propgated data member directly.
6103         * vr-values.h (class vr_values): Privatize vrp_value_range_pool,
6104         and values propagated data members and extract_range_from_assignment
6105         method.  Reorder private data members to conform to standards.
6106         Add new methods set_lattice_propagation_complete and
6107         allocate_value_range.
6109 2017-11-22  Eric Botcazou  <ebotcazou@adacore.com>
6111         PR rtl-optimization/83030
6112         * doc/rtl.texi (Flags in an RTL Expression): Alphabetize, add entry
6113         for CROSSING_JUMP_P and mention usage of 'jump' for JUMP_INSNs.
6114         (Insns): Delete entry for REG_CROSSING_JUMP in register notes.
6115         * bb-reorder.c (update_crossing_jump_flags): Do not test whether the
6116         CROSSING_JUMP_P flag is already set before setting it.
6117         * cfgrtl.c (fixup_partition_crossing): Likewise.
6118         * reorg.c (relax_delay_slots): Do not consider a CROSSING_JUMP_P
6119         insn as useless.
6121 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6123         * simplify-rtx.c (simplify_binary_operation_1) <case VEC_SERIES>:
6124         Handle the case where both arguments are using gen_const_vec_series.
6126 2017-11-22  David Malcolm  <dmalcolm@redhat.com>
6128         PR c++/62170
6129         * pretty-print.c (pp_format): Move quoting implementation to
6130         pp_begin_quote and pp_end_quote.  Update pp_format_decoder call
6131         to pass address of "quote" local.
6132         (pp_begin_quote): New function.
6133         (pp_end_quote): New function.
6134         * pretty-print.h (printer_fn): Convert penultimate param from bool
6135         to bool *.
6136         (pp_begin_quote): New decl.
6137         (pp_end_quote): New decl.
6138         * tree-diagnostic.c (default_tree_printer): Convert penultimate
6139         param from bool to bool *.
6140         * tree-diagnostic.h (default_tree_printer): Likewise.
6142 2017-11-22  Jeff Law  <law@redhat.com>
6144         * gimple-ssa-evrp-analyze.c (evrp_range_analyzer::evrp_range_analyzer)
6145         Initialize vr_values.
6146         (evrp_range_analyzer::try_find_new_range): Call methods attached to
6147         vr_values via vr_values class instance rather than delegators.
6148         (evrp_range_analyzer::record_ranges_from_phis): Likewise.
6149         (evrp_range_analyzer::record_ranges_from_stmt): Likewise.
6150         (evrp_range_analyzer::push_value_range): Likewise.
6151         (evrp_range_analyzer::pop_value_range): Likewise.
6152         * gimple-ssa-evrp-analyze.h (class evrp_range_analyzer): Remove
6153         most delegators.  Those remaining are exposed as public interfaces.
6154         Make vr_values a pointer and private.
6155         (evrp_range_analyzer::~evrp_range_analyzer): Delete the attached
6156         vr_values.
6157         (evrp_range_analyzer::get_vr_value): New method.
6158         * gimple-ssa-evrp.c (class evrp_folder): Use DISABLE_COPY_AND_ASSIGN.
6159         (evrp_folder::evrp_folder): New ctor to initialize vr_values.
6160         (class evrp_dom_walker): Attach evrp_folder class, initialize
6161         it in the ctor.  Remove temporary delegators.
6162         (evrp_dom_walker::before_dom_children): Call methods in attached
6163         evrp_range_analyzer class via class instance pointer.  Use
6164         free value_range_constant_singleton to remove need for
6165         op_with_constant_singleton_value delegator method.  Do not
6166         create a vrp_prop class instance for every call!  Narrow
6167         scope of a couple variables.
6168         (evrp_dom_walker::cleanup): Call methods in attached
6169         evrp_range_analyzer class via class instance pointer.
6170         * vr-values.h (class vr_values): Privatize many methods and
6171         data members.
6173 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6175         * tree.def (REDUC_MAX_EXPR, REDUC_MIN_EXPR, REDUC_PLUS_EXPR): Delete.
6176         * cfgexpand.c (expand_debug_expr): Remove handling for them.
6177         * expr.c (expand_expr_real_2): Likewise.
6178         * fold-const.c (const_unop): Likewise.
6179         * optabs-tree.c (optab_for_tree_code): Likewise.
6180         * tree-cfg.c (verify_gimple_assign_unary): Likewise.
6181         * tree-inline.c (estimate_operator_cost): Likewise.
6182         * tree-pretty-print.c (dump_generic_node): Likewise.
6183         (op_code_prio): Likewise.
6184         (op_symbol_code): Likewise.
6185         * internal-fn.def (DEF_INTERNAL_SIGNED_OPTAB_FN): Define.
6186         (IFN_REDUC_PLUS, IFN_REDUC_MAX, IFN_REDUC_MIN): New internal functions.
6187         * internal-fn.c (direct_internal_fn_optab): New function.
6188         (direct_internal_fn_array, direct_internal_fn_supported_p
6189         (internal_fn_expanders): Handle DEF_INTERNAL_SIGNED_OPTAB_FN.
6190         * fold-const-call.c (fold_const_reduction): New function.
6191         (fold_const_call): Handle CFN_REDUC_PLUS, CFN_REDUC_MAX and
6192         CFN_REDUC_MIN.
6193         * tree-vect-loop.c: Include internal-fn.h.
6194         (reduction_code_for_scalar_code): Rename to...
6195         (reduction_fn_for_scalar_code): ...this and return an internal
6196         function.
6197         (vect_model_reduction_cost): Take an internal_fn rather than
6198         a tree_code.
6199         (vect_create_epilog_for_reduction): Likewise.  Build calls rather
6200         than assignments.
6201         (vectorizable_reduction): Use internal functions rather than tree
6202         codes for the reduction operation.  Update calls to the functions
6203         above.
6204         * config/aarch64/aarch64-builtins.c (aarch64_gimple_fold_builtin):
6205         Use calls to internal functions rather than REDUC tree codes.
6206         * config/aarch64/aarch64-simd.md: Update comment accordingly.
6208 2017-11-22  Olivier Hainque  <hainque@adacore.com>
6210         * config/vxworks.c (vxworks_override_options): Pick default
6211         dwarf version from macro value, VXWORKS_DWARF_VERSION_DEFAULT.
6212         * config/vxworks.h: Define VXWORKS_DWARF_VERSION_DEFAULT and
6213         DWARF_GNAT_ENCODINGS_DEFAULT.
6214         * config/vxworksae.h: Likewise.
6216 2017-11-22  Marc Glisse  <marc.glisse@inria.fr>
6218         PR tree-optimization/83104
6219         * vr-values.c (simplify_stmt_using_ranges): Check integral argument,
6220         not result.
6222 2017-11-22  Marek Polacek  <polacek@redhat.com>
6223             H.J. Lu  <hongjiu.lu@intel.com>
6224             Jason Merrill  <jason@redhat.com>
6226         PR c++/60336
6227         PR middle-end/67239
6228         PR target/68355
6229         * calls.c (initialize_argument_information): Call
6230         warn_parameter_passing_abi target hook.
6231         (store_one_arg): Use 0 for empty record size.  Don't push 0 size
6232         argument onto stack.
6233         (must_pass_in_stack_var_size_or_pad): Return false for empty types.
6234         * common.opt: Update -fabi-version description.
6235         * config/i386/i386.c (init_cumulative_args): Set cum->warn_empty.
6236         (ix86_gimplify_va_arg): Call arg_int_size_in_bytes instead of
6237         int_size_in_bytes.
6238         (ix86_is_empty_record): New function.
6239         (ix86_warn_parameter_passing_abi): New function.
6240         (TARGET_EMPTY_RECORD_P): Redefine.
6241         (TARGET_WARN_PARAMETER_PASSING_ABI): Redefine.
6242         * config/i386/i386.h (CUMULATIVE_ARGS): Add warn_empty.
6243         * doc/tm.texi: Regenerated.
6244         * doc/tm.texi.in (TARGET_EMPTY_RECORD_P,
6245         TARGET_WARN_PARAMETER_PASSING_ABI): Add.
6246         * dwarf2out.c (get_ultimate_context): Move to tree.c.
6247         * explow.c (hard_function_value): Call arg_int_size_in_bytes
6248         instead of int_size_in_bytes.
6249         * expr.c (copy_blkmode_to_reg): Likewise.
6250         * function.c (aggregate_value_p): Return 0 for empty types.
6251         (assign_parm_find_entry_rtl): Call warn_parameter_passing_abi
6252         target hook.
6253         (locate_and_pad_parm): Call arg size_in_bytes instead
6254         size_in_bytes.
6255         * lto-streamer-out.c (hash_tree): Hash TYPE_EMPTY_P and DECL_PADDING_P.
6256         * stor-layout.c (finalize_type_size): Set TYPE_EMPTY_P.
6257         * target.def (empty_record_p, warn_parameter_passing_abi): New target
6258         hooks.
6259         * targhooks.c (hook_void_CUMULATIVE_ARGS_tree): New hook.
6260         (std_gimplify_va_arg_expr): Skip empty records.  Call
6261         arg_size_in_bytes instead size_in_bytes.
6262         * targhooks.h (hook_void_CUMULATIVE_ARGS_tree): Declare.
6263         * tree-core.h (tree_type_common): Add empty_flag.
6264         (tree_decl_common): Update comments.
6265         * tree-streamer-in.c (unpack_ts_decl_common_value_fields): Stream
6266         DECL_PADDING_P.
6267         (unpack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6268         * tree-streamer-out.c (pack_ts_decl_common_value_fields): Stream
6269         DECL_PADDING_P.
6270         (pack_ts_type_common_value_fields): Stream TYPE_EMPTY_P.
6271         * tree.c (default_is_empty_type): New function.
6272         (default_is_empty_record): New function.
6273         (arg_int_size_in_bytes): New function.
6274         (arg_size_in_bytes): New function.
6275         (get_ultimate_context): New function.
6276         * tree.h: Define TYPE_EMPTY_P, DECL_PADDING_P and
6277         TRANSLATION_UNIT_WARN_EMPTY_P.
6278         (default_is_empty_record, arg_int_size_in_bytes,
6279         arg_size_in_bytes, get_ultimate_context): Declare.
6281 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6283         * config/arm/arm.c (cmse_clear_registers): New function.
6284         (cmse_nonsecure_call_clear_caller_saved): Replace register clearing
6285         code by call to cmse_clear_registers.
6286         (cmse_nonsecure_entry_clear_before_return): Likewise.
6288 2017-11-22  Tamar Christina  <tamar.christina@arm.com>
6290         * config/arm/arm_neon.h (vdot_u32, vdotq_u32)
6291         (vdot_s32, vdotq_s32): New.
6292         (vdot_lane_u32, vdotq_lane_u32): New.
6293         (vdot_lane_s32, vdotq_lane_s32): New.
6296 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6298         PR middle-end/82547
6299         * wide-int.cc (wi::add_large, wi::sub_large): Fix overflow detection
6300         for unsigned values with fewer HWIs than the precision.
6301         (test_overflow): New function.
6302         (wide_int_cc_tests): Call it.
6304 2017-11-22  Richard Sandiford  <richard.sandiford@linaro.org>
6306         * emit-rtl.c (init_derived_machine_modes): Make sure ptr_mode
6307         has the same mode class as Pmode.
6309 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6311         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Get rid of
6312         padding_bits_to_clear_ptr.
6313         (cmse_nonsecure_entry_clear_before_return): Likewise.
6315 2017-11-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6317         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Use
6318         auto_sbitap instead of integer bitfield to control register needing
6319         clearing.
6321 2017-11-22  Jakub Jelinek  <jakub@redhat.com>
6323         PR tree-optimization/83044
6324         * tree-vrp.c (vrp_prop::check_array_ref): If eltsize is not
6325         INTEGER_CST or is 0, clear up_bound{,_p1} and later ignore tests
6326         that need the upper bound.  Subtract offset from
6327         get_addr_base_and_unit_offset only if positive and subtract it
6328         before division by eltsize rather than after it.
6330         PR debug/83084
6331         * valtrack.c (propagate_for_debug_subst, propagate_for_debug): Reset
6332         debug insns if they would contain UNSPEC_VOLATILE or volatile asm.
6333         (dead_debug_insert_temp): Likewise, but also ignore even non-volatile
6334         asm.
6336         PR middle-end/82875
6337         * optabs.c (expand_doubleword_mult, expand_binop): Before calling
6338         expand_binop with *mul_widen_optab, make sure at least one of the
6339         operands doesn't have VOIDmode.
6341         PR debug/83034
6342         * dwarf2out.c (mem_loc_descriptor): Handle VEC_SERIES.
6344         PR rtl-optimization/82044
6345         PR tree-optimization/82042
6346         * dse.c (record_store): Check for overflow.
6347         (check_mem_read_rtx): Properly check for overflow if width == -1, call
6348         add_wild_read instead of clear_rhs_from_active_local_stores on
6349         overflow and log it into dump_file.
6351 2017-11-22  Richard Biener  <rguenther@suse.de>
6353         * gimple-iterator.c (gimple_find_edge_insert_loc): Ignore
6354         fake edges to exit when looking for a place to insert.
6355         * tree-ssa-pre.c (clear_expression_ids): Inline into callers
6356         and remove.
6357         (insert_into_preds_of_block): Commit edge insertion immediately,
6358         assert that doesn't require new BBs.
6359         (fini_pre): Release expressions.
6360         (pass_pre::execute): Shuffle things around a bit, if the fn
6361         is too large do not compute AVAIL either as this is really the
6362         quadratic bit.
6364 2017-11-22  Richard Biener  <rguenther@suse.de>
6366         PR tree-optimization/83089
6367         * tree-if-conv.c (pass_if_conversion::execute): If anything
6368         changed reset SCEV and free the number of iteration estimates.
6370 2017-11-21  Martin Sebor  <msebor@redhat.com>
6372         PR tree-optimization/82945
6373         * calls.h (warn_nonstring_bound): Remove unused function.
6375 2017-11-21  Martin Sebor  <msebor@redhat.com>
6377         PR tree-optimization/82945
6378         * builtins.c (expand_builtin_strlen): Call maybe_warn_nonstring_arg.
6379         * calls.h (maybe_warn_nonstring_arg): Declare new function.
6380         * calls.c (get_attr_nonstring_decl, maybe_warn_nonstring_arg): New
6381         functions.
6382         (initialize_argument_information): Call maybe_warn_nonstring_arg.
6383         * calls.h (get_attr_nonstring_decl): Declare new function.
6384         * doc/extend.texi (attribute nonstring): Update.
6385         * gimple-fold.c (gimple_fold_builtin_strncpy): Call
6386         get_attr_nonstring_decl and handle it.
6387         * tree-ssa-strlen.c (maybe_diag_stxncpy_trunc): Same.  Improve
6388         detection of nul-termination.
6389         (strlen_to_stridx): Change to a pointer.
6390         (handle_builtin_strlen, handle_builtin_stxncpy): Adjust.
6391         (pass_strlen::execute): Same.
6393 2017-11-21  Sergey Shalnov  <Sergey.Shalnov@intel.com>
6395         * config/i386/i386-opts.h (enum prefer_vector_width): Added new enum
6396         for the new option -mprefer-vector-width=[none|128|256|512].
6397         * config/i386/i386.c (ix86_target_string): remove old style options
6398         -mprefer-avx256 and make -mprefer-avx128 as alias.
6399         (ix86_option_override_internal):  Apply defaults for the
6400         -mprefer-vector-width=[128|256] option.
6401         * config/i386/i386.h (TARGET_PREFER_AVX128, TARGET_PREFER_AVX256):
6402         Implement macros to work with -mprefer-vector-width=.
6403         * config/i386/i386.opt: Implemented option
6404         -mprefer-vector-width=[none|128|256|512].
6405         * doc/invoke.texi: Documentation for
6406         -mprefer-vector-width=[none|128|256|512].
6408 2017-11-21  Pat Haugen  <pthaugen@us.ibm.com>
6410         * config/rs6000/ppc-asm.h (f50, vs50): Fix values.
6412 2017-11-21  Jonathan Wakely  <jwakely@redhat.com>
6414         * doc/invoke.texi (-Wmaybe-uninitialized): Rephrase for clarity.
6416 2017-11-21  Marc Glisse  <marc.glisse@inria.fr>
6418         * doc/generic.texi: Document POINTER_DIFF_EXPR, update
6419         POINTER_PLUS_EXPR.
6420         * cfgexpand.c (expand_debug_expr): Handle POINTER_DIFF_EXPR.
6421         * expr.c (expand_expr_real_2): Likewise.
6422         * fold-const.c (const_binop, fold_addr_of_array_ref_difference,
6423         fold_binary_loc): Likewise.
6424         * match.pd (X-X, P+(Q-P), &D-P, (P+N)-P, P-(P+N), (P+M)-(P+N),
6425         P-Q==0, -(A-B), X-Z<Y-Z, (X-Z)-(Y-Z), Z-X<Z-Y, (Z-X)-(Z-Y),
6426         (A-B)+(C-A)): New transformations for POINTER_DIFF_EXPR, based on
6427         MINUS_EXPR transformations.
6428         * optabs-tree.c (optab_for_tree_code): Handle POINTER_DIFF_EXPR.
6429         * tree-cfg.c (verify_expr, verify_gimple_assign_binary): Likewise.
6430         * tree-inline.c (estimate_operator_cost): Likewise.
6431         * tree-pretty-print.c (dump_generic_node, op_code_prio,
6432         op_symbol_code): Likewise.
6433         * tree-vect-stmts.c (vectorizable_operation): Likewise.
6434         * vr-values.c (extract_range_from_binary_expr): Likewise.
6435         * varasm.c (initializer_constant_valid_p_1): Likewise.
6436         * tree.def: New tree code POINTER_DIFF_EXPR.
6438 2017-11-21  Uros Bizjak  <ubizjak@gmail.com>
6440         * config/i386/i386.md (*bswap<mode>2_movbe): Add
6441         integer suffix to movbe mnemonic.
6442         (*bswaphi2_movbe): Ditto.
6443         (bswaphi_lowpart): Merge with *bswaphi_lowpart_1.
6445 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6447         PR c++/83045
6448         * tree-cfg.c (pass_warn_function_return::execute): Formatting fix.
6449         Also warn if seen __builtin_unreachable () call with BUILTINS_LOCATION.
6450         Use LOCATION_LOCUS when comparing against UNKNOWN_LOCATION.
6452 2017-11-21  Martin Liska  <mliska@suse.cz>
6454         * tree-inline.c (expand_call_inline): Remove not needed
6455         xstrdup_for_dump.
6457 2017-11-21  James Cowgill  <James.Cowgill@imgtec.com>
6458             Jakub Jelinek  <jakub@redhat.com>
6460         PR target/82880
6461         * config/mips/frame-header-opt.c (mips_register_frame_header_opt):
6462         Remove static keyword from f variable.
6464 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6466         PR tree-optimization/83086
6467         * gimple-ssa-store-merging.c
6468         (imm_store_chain_info::try_coalesce_bswap): Test this_n.base_addr
6469         rather than n.base_addr.
6471 2017-11-21  Martin Liska  <mliska@suse.cz>
6473         PR rtl-optimization/82044
6474         PR tree-optimization/82042
6475         * dse.c (check_mem_read_rtx): Check for overflow.
6477 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6479         * tree-ssa-threadbackward.c (find_jump_threads_backwards): Fix
6480         typo in comment.
6482 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6484         * cfgrtl.c (force_nonfallthru_and_redirect): Don't split a call
6485         and its corresponding call arg location note.
6487 2017-11-21  Claudiu Zissulescu  <claziss@synopsys.com>
6488             Andrew Burgess  <andrew.burgess@embecosm.com>
6490         * config/arc/arc-protos.h (arc_compute_frame_size): Delete
6491         declaration.
6492         (arc_return_slot_offset): Likewise.
6493         (arc_eh_return_address_location): New declaration.
6494         * config/arc/arc.c (TARGET_BUILTIN_SETJMP_FRAME_VALUE): Define.
6495         (MUST_SAVE_REGISTER): Add exception handler case.
6496         (MUST_SAVE_RETURN_ADDR): Likewise.
6497         (arc_frame_pointer_required): Likewise.
6498         (arc_frame_pointer_needed): New function.
6499         (arc_compute_frame_size): Changed.
6500         (arc_expand_prologue): Likewise.
6501         (arc_expand_epilogue): Likewise.
6502         (arc_initial_elimination_offset): Likewise.
6503         (arc_return_slot_offset): Delete.
6504         (arc_eh_return_address_location): New function.
6505         (arc_builtin_setjmp_frame_value): Likewise.
6506         * config/arc/arc.h (EH_RETURN_DATA_REGNO): Use 2 registers.
6507         (EH_RETURN_STACKADJ_RTX): Define.
6508         (EH_RETURN_HANDLER_RTX): Likewise.
6509         * config/arc/arc.md (eh_return): Delete.
6511 2017-11-21  Aldy Hernandez  <aldyh@redhat.com>
6513         * print-tree.h (debug_vec_tree): Remove prototype.
6514         * gdbinit.in (pvt): Remove macro.
6516 2017-11-21  Jakub Jelinek  <jakub@redhat.com>
6518         PR tree-optimization/83047
6519         * gimple-ssa-store-merging.c
6520         (imm_store_chain_info::output_merged_store): If the loads with the
6521         same vuse are in different basic blocks, for load_gsi pick a load
6522         location that is dominated by the other loads.
6524         PR c++/83059
6525         * config/i386/i386.c (ix86_memmodel_check): Start
6526         -Winvalid-memory-model diagnostics with lowercase letter.
6528         PR debug/82718
6529         * dwarf2out.c (dw_loc_list): If crtl->has_bb_partition, temporarily
6530         set in_cold_section_p to the partition containing loc_list->first.
6531         When seeing loc_list->last_before_switch node, update secname and
6532         perform range_across_switch second partition handling only after that.
6534         PR debug/82933
6535         * run-rtl-passes.c: Include debug.h.
6536         (run_rtl_passes): Call debug_hooks->assembly_start.
6537         * dwarf2out.c (dwarf2out_assembly_start): Return early if invoked
6538         multiple times.
6540         PR target/82981
6541         * internal-fn.c (expand_mul_overflow): Use OPTAB_WIDEN instead of
6542         OPTAB_DIRECT in calls to expand_simple_binop.
6544 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6546         PR c/81404
6547         * Makefile.in (C_COMMON_OBJS): Add c-family/known-headers.o.
6549 2017-11-20  David Malcolm  <dmalcolm@redhat.com>
6551         PR c++/72786
6552         * spellcheck.h (best_match::blithely_get_best_candidate): New
6553         accessor.
6555 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6557         * config/i386/i386.c (parse_mtune_ctrl_str): Start diagnostics
6558         with lower case letter.
6560 2017-11-20  Uros Bizjak  <ubizjak@gmail.com>
6562         * config/i386/i386.md (bswaphi2): New expander.
6563         (*bswaphi2_movbe): New insn pattern.
6564         (bswaphi -> rorhi pepehole2): New peephole pattern.
6566 2017-11-20  Jeff Law  <law@redhat.com>
6568         * Makefile.in (OBJS): Add gimple-ssa-evrp-analyze.o.
6569         * gimple-ssa-evrp-analyze.c: New file pulled from gimple-ssa-evrp.c.
6570         * gimple-ssa-evrp-analyze.h: New file pulled from gimple-ssa-evrp.c.
6571         * gimple-ssa-evrp.c: Remove bits moved into new files.  Include
6572         gimple-ssa-evrp-analyze.h.
6574         * gimple-ssa-evrp.c (evrp_dom_walker::before_dom_children): Do not
6575         set BB_VISITED here.
6576         (evrp_range_analyzer::enter): Set BB_VISITED here instead.
6578 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6580         * tree-predcom.c: Add general comment on Store-Store chains.
6581         (split_data_refs_to_components): Postpone clearing eliminate_store_p
6582         flag in component.
6583         (get_chain_last_ref_at): Rename into...
6584         (get_chain_last_write_at): ...this.
6585         (get_chain_last_write_before_load): New function.
6586         (add_ref_to_chain): Promote type of chain from CT_STORE_LOAD to
6587         CT_STORE_STORE when write reference is added.
6588         (determine_roots_comp): Support load ref in CT_STORE_STORE chains.
6589         (is_inv_store_elimination_chain): Update get_chain_last_write_at call.
6590         (initialize_root_vars_store_elim_1): Ditto.
6591         (initialize_root_vars_store_elim_2): Ditto.  Replace rhs once default
6592         definition is created.
6593         (execute_pred_commoning_chain): Support load ref in CT_STORE_STORE
6594         chain by replacing it with dominant stored value.
6596 2017-11-20  Bin Cheng  <bin.cheng@arm.com>
6598         * tree-predcom.c (add_ref_to_chain): Remove check on distance.
6600 2017-11-20  Marc Glisse  <marc.glisse@inria.fr>
6602         * vr-values.c (extract_range_from_binary_expr): Use a full range
6603         for VR_VARYING.
6605 2017-11-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6607         * config/arm/arm.md (R4_REGNUM): Define constant.
6608         (nonsecure_call_internal): Remove r4 clobber.
6609         (nonsecure_call_value_internal): Likewise.
6610         * config/arm/thumb1.md (nonsecure_call_reg_thumb1_v5): Remove second
6611         clobber and resequence match_operands.
6612         (nonsecure_call_value_reg_thumb1_v5): Likewise.
6613         * config/arm/thumb2.md (nonsecure_call_reg_thumb2): Likewise.
6614         (nonsecure_call_value_reg_thumb2): Likewise.
6616 2017-11-20  Jakub Jelinek  <jakub@redhat.com>
6618         PR tree-optimization/78821
6619         * gimple-ssa-store-merging.c (find_bswap_or_nop_load): Give up
6620         if base is TARGET_MEM_REF.  If base is not MEM_REF, set base_addr
6621         to the address of the base rather than the base itself.
6622         (find_bswap_or_nop_1): Just use pointer comparison for vuse check.
6623         (find_bswap_or_nop_finalize): New function.
6624         (find_bswap_or_nop): Use it.
6625         (bswap_replace): Return a tree rather than bool, change first
6626         argument from gimple * to gimple_stmt_iterator, allow inserting
6627         into an empty sequence, allow ins_stmt to be NULL - then emit
6628         all stmts into gsi.  Fix up MEM_REF address gimplification.
6629         (pass_optimize_bswap::execute): Adjust bswap_replace caller.
6630         (struct store_immediate_info): Add N and INS_STMT non-static
6631         data members.
6632         (store_immediate_info::store_immediate_info): Initialize them
6633         from newly added ctor args.
6634         (merged_store_group::apply_stores): Formatting fixes.  Sort by
6635         bitpos at the end.
6636         (stmts_may_clobber_ref_p): For stores call also
6637         refs_anti_dependent_p.
6638         (gather_bswap_load_refs): New function.
6639         (imm_store_chain_info::try_coalesce_bswap): New method.
6640         (imm_store_chain_info::coalesce_immediate_stores): Use it.
6641         (split_group): Handle LROTATE_EXPR and NOP_EXPR rhs_code specially.
6642         (imm_store_chain_info::output_merged_store): Fail if number of
6643         new estimated stmts is bigger or equal than old.  Handle LROTATE_EXPR
6644         and NOP_EXPR rhs_code.
6645         (pass_store_merging::process_store): Compute n and ins_stmt, if
6646         ins_stmt is non-NULL and the store rhs is otherwise invalid, use
6647         LROTATE_EXPR rhs_code.  Pass n and ins_stmt to store_immediate_info
6648         ctor.
6649         (pass_store_merging::execute): Calculate dominators.
6651         * tree-ssa-math-opts.c (nop_stats, bswap_stats, struct symbolic_number,
6652         BITS_PER_MARKER, MARKER_MASK, MARKER_BYTE_UNKNOWN, HEAD_MARKER, CMPNOP,
6653         CMPXCHG, do_shift_rotate, verify_symbolic_number_p,
6654         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6655         find_bswap_or_nop_1, find_bswap_or_nop, pass_data_optimize_bswap,
6656         class pass_optimize_bswap, bswap_replace,
6657         pass_optimize_bswap::execute): Moved to ...
6658         * gimple-ssa-store-merging.c: ... this file.
6659         Include optabs-tree.h.
6660         (nop_stats, bswap_stats, do_shift_rotate, verify_symbolic_number_p,
6661         init_symbolic_number, find_bswap_or_nop_load, perform_symbolic_merge,
6662         find_bswap_or_nop_1, find_bswap_or_nop, bswap_replace): Put into
6663         anonymous namespace, remove static keywords.
6664         (pass_optimize_bswap::gate): Test BITS_PER_UNIT == 8 here...
6665         (pass_optimize_bswap::execute): ... rather than here.  Formatting fix.
6667 2017-11-20  Jan Hubicka  <hubicka@ucw.cz>
6669         PR bootstrap/83062
6670         * ipa-inline.c (can_inline_edge_p): Fix typo in previous patch.
6672 2017-11-20  Aldy Hernandez  <aldyh@redhat.com>
6674         * vec.h (debug_helper): New function.
6675         (DEFINE_DEBUG_VEC): New macro.
6676         * hash-set.h (debug_helper): New function.
6677         (DEFINE_DEBUG_HASH_SET): New macro.
6678         * cfg.c (debug_slim (edge)): New function.
6679         Call DEFINE_DEBUG_VEC for edges.
6680         Call DEFINE_DEBUG_HASH_SET for edges.
6681         * cfghooks.c (debug_slim (basic_block)): New function.
6682         Call DEFINE_DEBUG_VEC for basic blocks.
6683         Call DEFINE_DEBUG_HASH_SET for basic blocks.
6684         * print-tree.c (debug_slim): New function to handle trees.
6685         Call DEFINE_DEBUG_VEC for trees.
6686         Call DEFINE_DEBUG_HASH_SET for trees.
6687         (debug (vec<tree, va_gc>) &): Remove.
6688         (debug (<vec<tree, va_gc>) *): Remove.
6689         * print-rtl.c (debug_slim): New function to handle const_rtx.
6690         Call DEFINE_DEBUG_VEC for rtx_def.
6691         Call DEFINE_DEBUG_VEC for rtx_insn.
6692         Call DEFINE_DEBUG_HASH_SET for rtx_def.
6693         Call DEFINE_DEBUG_HASH_SET for rtx_insn.
6694         * sel-sched-dump.c (debug (vec<rtx_insn *> &): Remove.
6695         (debug (vec<rtx_insn *> *ptr): Remove.
6696         (debug_insn_vector): Remove.
6697         * stor-layout.c (debug_rli): Call debug() instead of debug_vec_tree.
6699 2017-11-20  Tom de Vries  <tom@codesourcery.com>
6701         PR rtl-optimization/82020
6702         * simplify-rtx.c (simplify_ternary_operation): Fix comparison mode of
6703         IF_THEN_ELSE condition.
6705 2017-11-19  Jeff Law  <law@redhat.com>
6707         * tree-ssa-dom.c (record_equivalences_from_phis): Fix handling
6708         of degenerates resulting from ignoring an edge.
6710 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6712         PR ipa/81360
6713         * ipa-inline.c (can_inline_edge_p): Also check that caller is optimized
6715 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6717         PR ipa/83001
6718         * profile-count.c (profile_count::to_sreal_scale): Fix return value
6719         for uninitialied counts.
6721 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6723         PR ipa/60243
6724         * tree-inline.c (estimate_num_insns): Set to 1 at least.
6726 2017-11-19  Jan Hubicka  <hubicka@ucw.cz>
6728         PR target/82713
6729         * i386.c (ix86_builtin_vectorization_cost): Be ready for insane types.
6731 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6733         * config/arc/arc.h (FUNCTION_PROFILER): Remove semicolon after
6734         "do while (0)".
6736 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6738         * config/phoenix.h (TARGET_OS_CPP_BUILTINS): Remove semicolon after
6739         "do {} while (0)".
6741 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6743         * config/visium/visium.h (ASM_OUTPUT_CASE_END): Remove semicolon after
6744         macro body.
6746 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6748         * config/ft32/ft32.h (REGISTER_TARGET_PRAGMAS): Remove semicolon after
6749         "do {} while (0)".
6750         * config/spu/spu.h (REGISTER_TARGET_PRAGMAS): Same.
6752 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6754         * config/mcore/mcore-elf.h (MCORE_EXPORT_NAME): Remove semicolon after
6755         "do {} while (0)".
6756         * config/mcore/mcore.h (ASM_OUTPUT_ALIGNED_COMMON): After missing
6757         semicolon after MCORE_EXPORT_NAME call.
6759 2017-11-19  Tom de Vries  <tom@codesourcery.com>
6761         PR target/82961
6762         * vmsdbgout.c (vmsdbgout_early_finish): New function.
6763         (vmsdbg_debug_hooks): Set early_finish field to vmsdbgout_early_finish.
6765 2017-11-18  Jan Hubicka  <hubicka@ucw.cz>
6767         * cgraphclones.c (cgraph_edge::clone): Rename gcov_count to prof_count.
6768         (cgraph_edge::clone): Cleanup updating of profile.
6769         * ipa-cp.c (update_profiling_info): Likewise.
6770         * ipa-inline-transform.c (inline_transform): Likewise.
6771         * ipa-inline.c (inline_small_functions): Add missing space to dump.
6772         * ipa-split.c (execute_split_functions): Do not split when function
6773         is cold.
6774         * predict.c (estimate_bb_frequencies): Cleanup updating of profile.
6775         * profile-count.c (profile_count::dump): Add global0.
6776         (profile_count::to_cgraph_frequency): Do not ICE when entry is
6777         undefined.
6778         (profile_count::to_sreal_scale): Likewise.
6779         (profile_count::adjust_for_ipa_scaling): Fix typo in comment.
6780         (profile_count::combine_with_ipa_count): New function.
6781         * profile-count.h (profile_guessed_global0adjusted): New.
6782         (profile_count::adjusted_zero): New.
6783         (profile_count::global0adjusted): New.
6784         (profile_count::combine_with_ipa_count): New.
6785         * tree-inline.c (copy_edges_for_bb): Add NUM/DEN arugment;
6786         correct profile of return block of split functions.
6787         (copy_cfg_body): Remove unused profile_count.
6788         (copy_body): Likewise.
6789         (expand_call_inline): Update.
6790         (tree_function_versioning): Update.
6792 2017-11-18  Aldy Hernandez  <aldyh@redhat.com>
6794         * hash-set.h (hash_set::empty): New.
6795         * tree-ssa-threadbackward.h: Delete.
6796         * tree-ssa-threadbackward.c (class thread_jumps): New.
6797         Move max_threaded_paths into class.
6798         (fsm_find_thread_path): Remove arguments that are now in class.
6799         (profitable_jump_thread_path): Rename to...
6800         (thread_jumps::profitable_jump_thread_path): ...this.
6801         (convert_and_register_jump_thread_path): Rename to...
6802         (thread_jumps::convert_and_register_current_path): ...this.
6803         (check_subpath_and_update_thread_path): Rename to...
6804         (thread_jumps::check_subpath_and_update_thread_path): ...this.
6805         (register_jump_thread_path_if_profitable): Rename to...
6806         (thread_jumps::register_jump_thread_path_if_profitable): ...this.
6807         (handle_phi): Rename to...
6808         (thread_jumps::handle_phi): ...this.
6809         (handle_assignment): Rename to...
6810         (thread_jumps::handle_assignment): ...this.
6811         (fsm_find_control_statement_thread_paths): Rename to...
6812         (thread_jumps::fsm_find_control_statement_thread_paths): ...this.
6813         (find_jump_threads_backwards): Rename to...
6814         (thread_jumps::find_jump_threads_backwards): ...this.
6815         Initialize path local data.
6816         (pass_thread_jumps::execute): Call find_jump_threads_backwards
6817         from within thread_jumps class.
6818         (pass_early_thread_jumps::execute): Same.
6820 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6822         * cgraph.c (cgraph_node::dump): Do IPA sanity checking on IPA counts.
6824 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6826         * config/aarch64/aarch64-simd.md (fnma<mode>4): Move neg operator
6827         to canonical location.
6829 2017-11-17  Steve Ellcey  <sellcey@cavium.com>
6831         PR target/81356
6832         * config/aarch64/aarch64.c (aarch64_use_by_pieces_infrastructure_p):
6833         Remove.
6834         (TARGET_USE_BY_PIECES_INFRASTRUCTURE_P): Remove define.
6836 2017-11-17  H.J. Lu  <hongjiu.lu@intel.com>
6838         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Use
6839         rcrt1.o%s/grcrt1.o%s for -static-pie.
6841 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6843         * i386.c (ix86_multiplication_cost, ix86_division_cost,
6844         ix86_shift_rotate_cost): Break out from ...
6845         (ix86_rtx_costs): ... here.
6846         (ix86_add_stmt_cost): Use rtx cost machinery to compute cost of
6847         vector operations.
6849 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6851         * predict.c (determine_unlikely_bbs): Set cgraph node count to 0
6852         when entry block was promoted unlikely.
6853         (estimate_bb_frequencies): Increase frequency scale.
6854         * profile-count.h (profile_count): Export precision info.
6856 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6858         * tree-tailcall.c (eliminate_tail_call): Be more careful about not
6859         disturbin profile of entry block.
6861 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6863         * ipa-fnsummary.c (estimate_node_size_and_time): Be more tolerant for
6864         roundoff errors.
6866 2017-11-17  Jan Hubicka  <hubicka@ucw.cz>
6868         * ipa-cp.c (update_profiling_info): Handle conversion to local profile.
6869         * tree-cfg.c (execute_fixup_cfg): Do fixup same way as inliner does.
6871 2017-11-17  Jeff Law  <law@redhat.com>
6873         * gimple-ssa-evrp.c (class evrp_range_analyzer): New class extracted
6874         from evrp_dom_walker class.  Various methods moved into new class.
6875         (evrp_range_analyzer::evrp_range_analyzer): Constructor for new class.
6876         (evrp_range_analyzer::enter): New method.
6877         (evrp_range_analyzer::leave): New method.
6878         (evrp_dom_walker): Remove delegators no longer needed by this class.
6879         Replace vr_values data member with evrp_range_analyzer
6881         * gimple-ssa-evrp.c (evrp_dom_walker::record_ranges_from_phis): New
6882         method extracted from evrp_dom_walker::before_dom_children.
6883         (evrp_dom_walker::record_ranges_from_stmt): Likewise.
6884         (evrp_dom_walker::record_ranges_from_incoming_edge): Likewise.
6886         * gimple-ssa-evrp.c (evrp_dom_walker): Add cleanup method.
6887         Add private copy constructor and move assignment operators.
6888         Privatize methods and class data where trivially possible.
6889         (evrp_dom_walker::cleanup): New function, extracted from
6890         execute_early_vrp.  Simplify access to class data.
6892         * vr-values.h (get_output_for_vrp): Prototype.
6893         * vr-values.c (get_output_for_vrp): New function extracted from
6894         vrp_visit_assignment_or_call and extract_range_from_stmt.
6895         (vrp_visit_assignment_or_call): Use get_output_for_vrp.  Simplify.
6897 2017-11-17  Luis Machado  <luis.machado@linaro.org>
6899         * config/aarch64/aarch64.c
6900         (qdf24xx_prefetch_tune) <default_opt_level>: Set to -1.
6901         (qdf24xx_tunings) <autoprefetcher_model>: Set to
6902         tune_params::AUTOPREFETCHER_WEAK.
6904 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6906         PR target/82641
6907         * config/arm/arm.c (arm_valid_target_attribute_rec):
6908         Parse "arch=" and "+<ext>".
6909         (arm_valid_target_attribute_tree): Re-init global options.
6910         (arm_option_override): Make non-static.
6911         (arm_options_perform_arch_sanity_checks): Make errors fatal.
6912         * gcc/config/arm/arm-c.c (__ARM_FEATURE_CMSE): Support undef.
6913         (__ARM_FEATURE_CRC32): Support undef.
6914         * config/arm/arm_acle.h (__ARM_FEATURE_CRC32): Replace with pragma.
6915         * doc/extend.texi (ARM Function Attributes): Add pragma and target.
6917 2017-11-17  David Malcolm  <dmalcolm@redhat.com>
6919         * gdbinit.in (break-on-diagnostic): New command.
6921 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6923         * config/i386/i386.c (ix86_expand_epilogue): Change simple
6924         return to indirect jump for EH return if control-flow
6925         protection is enabled. Change explicit 'false' argument in
6926         pro_epilogue_adjust_stack with a value of
6927         flag_cf_protection.
6928         * config/i386/i386.md (simple_return_indirect_internal):
6929         Remove SImode restriction to support 64-bit.
6931 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6933         * combine.c (added_notes_insn): New.
6934         (try_combine): Handle added_notes_insn like added_links_insn.
6935         Rewrite return value code.
6936         (distribute_notes): Set added_notes_insn to the earliest insn we added
6937         a note to.
6939 2017-11-17  Segher Boessenkool  <segher@kernel.crashing.org>
6941         PR rtl-optimization/82621
6942         * combine.c (try_combine): Do not split PARALLELs of two SETs if the
6943         dest of one of those SETs is unused.
6945 2017-11-17  Richard Biener  <rguenther@suse.de>
6947         PR fortran/83017
6948         * tree-core.h (enum annot_expr_kind): Add annot_expr_parallel_kind.
6949         * tree-pretty-print.c (dump_generic_node): Handle
6950         annot_expr_parallel_kind.
6951         * tree-cfg.c (replace_loop_annotate_in_block): Likewise.
6952         * gimplify.c (gimple_boolify): Likewise.
6954 2017-11-17  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
6956         * config.gcc (extra_headers): Add cet.h for x86 targets.
6957         * config/i386/cet.h: New file.
6958         * doc/install.texi: Add --enable-cet/--disable-cet.
6960 2017-11-17  Richard Biener  <rguenther@suse.de>
6962         PR tree-optimization/83017
6963         * tree-parloops.c (MIN_PER_THREAD): Use --param parloops-min-per-thread.
6964         (gen_parallel_loop): Properly count iterations.
6965         (parallelize_loops): Handle loop->can_be_parallel independent
6966         of flag_loop_parallelize_all.  Make static profitability test match
6967         the runtime one.
6968         * params.def (PARAM_PARLOOPS_MIN_PER_THREAD): New.
6969         * invoke.texi (parloops-min-per-thread): Document.
6971 2017-11-17  Vineet Gupta  <vgupta@synopsys.com>
6973         * config/arc/linux.h: GLIBC_DYNAMIC_LINKER update per glibc
6974         upstreaming review comments.
6976 2017-11-17  Tamar Christina  <tamar.christina@arm.com>
6978         * expr.c (copy_blkmode_to_reg): Fix bitsize for targets
6979         with fast unaligned access.
6980         * doc/sourcebuild.texi (word_mode_no_slow_unalign): New.
6982 2017-11-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
6984         * config/arm/arm.c (cmse_nonsecure_entry_clear_before_return): Allocate
6985         to_clear_arg_regs_bitmap to the same size as to_clear_bitmap.
6987 2017-11-17  Richard Biener  <rguenther@suse.de>
6989         * tree-ssa-pre.c (phi_translate_1): Remove redundant constant
6990         folding of references.
6992 2017-11-17  Qing Zhao  <qing.zhao@oracle.com>
6994         PR middle-end/78809
6995         * gimple-fold.c (gimple_fold_builtin_string_compare): Add handling
6996         of replacing call to strncmp with corresponding call to strcmp when
6997         meeting conditions.
6999 2017-11-17  Sergey Shalnov  <Sergey.Shalnov@intel.com>
7001         * config/i386/x86-tune.def (X86_TUNE_AVX256_OPTIMAL): Add tuning
7002         option prefer-avx256 for skylake-avx512 configuration.
7003         * config/i386/i386.c (ix86_option_override_internal): Ditto.
7004         (get_builtin_code_for_version): Ditto.
7006 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7007             Monk Chiang  <sh.chiang04@gmail.com>
7009         * config/nds32/nds32.h (FIRST_PSEUDO_REGISTER): Modify.
7010         (FIXED_REGISTERS): Reserve more register numbers.
7011         (CALL_USED_REGISTERS): Likewise.
7012         (REG_ALLOC_ORDER): Likewise.
7013         (REG_CLASS_CONTENTS): Likewise.
7014         (REGISTER_NAMES): Likewise.
7016 2017-11-17  Chung-Ju Wu  <jasonwucj@gmail.com>
7017             Kito Cheng  <kito.cheng@gmail.com>
7019         * config/nds32/nds32-modes.def: Add vector mode V4QI V2HI V8QI V4HI
7020         V2SI.
7021         * config/nds32/iterators.md: Add vector mode iterators and attributes.
7023 2017-11-16  Steven Munroe  <munroesj@gcc.gnu.org>
7025         * config/rs6000/mmintrin.h (_mm_add_pi32[_ARCH_PWR]): Correct
7026         parameter list for vec_splats.
7028 2017-11-16  Joseph Myers  <joseph@codesourcery.com>
7030         * doc/invoke.texi (-std=c17): Refer to 2018 expected publication
7031         date of C17.
7032         (-std=c18, -std=iso9899:2018, -std=gnu18): Document option aliases.
7034 2017-11-16  Thomas Koenig  <tkoenig@gcc.gnu.org>
7036         PR bootstrap/82856
7037         * doc/install.texi: Document incompatibility of Perl >=5.6.26
7038         with the required version of automake 1.11.6.
7040 2017-11-16  Pat Haugen  <pthaugen@us.ibm.com>
7042         * rs6000/power9.md (power9fpdiv): New automaton and cpu_unit defined
7043         for it.
7044         (DU_C2_3_power9): Correct reservation combinations.
7045         (FP_DIV_power9, VEC_DIV_power9): New.
7046         (power9-alu): Split out rotate/shift...
7047         (power9-rot): ...to here, correct dispatch resource.
7048         (power9-cracked-alu, power9-mul, power9-mul-compare): Correct dispatch
7049         resource.
7050         (power9-fp): Correct latency.
7051         (power9-sdiv): Add div/sqrt resource.
7052         (power9-ddiv): Correct latency, add div/sqrt resource.
7053         (power9-sqrt, power9-dsqrt): Add div/sqrt resource.
7054         (power9-vecfdiv, power9-vecdiv): Correct latency, add div/sqrt
7055         resource.
7056         (power9-qpdiv, power9-qpmul): Adjust resource usage.
7058 2017-11-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
7060         * config/rs6000/rs6000.c (rs6000_expand_builtin): Do not do the
7061         switch statement mapping KF built-ins to TF built-ins if we don't
7062         have the proper ISA 3.0 assembler support.
7064 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7066         * tree-emutls.c (lower_emutls_data): Remove unused bb_freq.
7067         (lower_emutls_function_body): Do not compute it.
7069 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7071         * ipa-split.c (split_bb_info): Turn time to sreal.
7072         (split_point): Likewise.
7073         (dump_split_point): Likewise.
7074         (fine_split_points): Likewise.
7075         (execute_split_functions): Only zero split_bbs; turn time to sreals.
7077 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7079         * ipa-fnsummary.c (analyze_function_body): Accumulate time consistently
7080         in sreal.
7082 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7084         * predict.c (combine_predictions_for_bb): Preserve zero predicted
7085         edges.
7086         (expensive_function_p): Remove useless assert.
7087         (determine_unlikely_bbs): Propagate also forward; determine cold blocks
7089 2017-11-16  Martin Sebor  <msebor@redhat.com>
7091         PR tree-optimization/82588
7092         PR tree-optimization/82583
7093         * tree-vrp.c (check_array_ref): Handle flexible array members,
7094         string literals, and inner indices.
7095         (search_for_addr_array): Add detail to diagnostics.
7097 2017-11-16  Nathan Sidwell  <nathan@acm.org>
7099         PR c++/82836
7100         PR c++/82737
7101         * tree.h (COPY_DECL_RTL): Rename parms for clarity.
7102         (SET_DECL_ASSEMBLER_NAME): Forward to
7103         overwrite_decl_assembler_name.
7104         (COPY_DECL_ASSEMBLER_NAME): Rename parms for clarity.
7105         (overwrite_decl_assembler_name): Declare.
7106         * tree.c (overwrite_decl_assembler_name): New.
7107         * langhooks-def.h (lhd_overwrite_decl_assembler_name): Declare.
7108         (LANG_HOOKS_OVERWRITE_DECL_ASSEMBLER_NAME): Provide default.
7109         (LANG_HOOKS_INITIALIZER): Add it.
7110         * langhooks.h (struct lang_hooks): Add overwrite_decl_assembler_name.
7111         * langhooks.c (lhd_set_decl_assembler_name): Use
7112         SET_DECL_ASSEMBLER_NAME.
7113         (lhd_overwrite_decl_assembler_name): Default implementation.
7115 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7116             Jackson Woodruff  <jackson.woodruff@arm.com>
7118         PR tree-optimization/71026
7119         * match.pd: Canonicalize constant multiplies in division.
7121 2017-11-16  Wilco Dijkstra  <wdijkstr@arm.com>
7123         * opts.c (default_options_table): Add OPT_fomit_frame_pointer entry.
7124         * common/config/alpha/alpha-common.c (TARGET_OPTION_OPTIMIZATION_TABLE):
7125         Remove OPT_fomit_frame_pointer entry.
7126         * common/config/arc/arc-common.c: Likewise.
7127         * common/config/arm/arm-common.c: Likewise.
7128         * common/config/avr/avr-common.c: Likewise.
7129         * common/config/c6x/c6x-common.c: Likewise.
7130         * common/config/cr16/cr16-common.c: Likewise.
7131         * common/config/cris/cris-common.c: Likewise.
7132         * common/config/epiphany/epiphany-common.c: Likewise.
7133         * common/config/fr30/fr30-common.c: Likewise.
7134         * common/config/frv/frv-common.c: Likewise.
7135         * common/config/ia64/ia64-common.c: Likewise.
7136         * common/config/iq2000/iq2000-common.c: Likewise.
7137         * common/config/lm32/lm32-common.c: Likewise.
7138         * common/config/m32r/m32r-common.c: Likewise.
7139         * common/config/mcore/mcore-common.c: Likewise.
7140         * common/config/microblaze/microblaze-common.c: Likewise.
7141         * common/config/mips/mips-common.c: Likewise.
7142         * common/config/mmix/mmix-common.c: Likewise.
7143         * common/config/mn10300/mn10300-common.c: Likewise.
7144         * common/config/nios2/nios2-common.c: Likewise.
7145         * common/config/pa/pa-common.c: Likewise.
7146         * common/config/pdp11/pdp11-common.c: Likewise.
7147         * common/config/powerpcspe/powerpcspe-common.c: Likewise.
7148         * common/config/riscv/riscv-common.c: Likewise.
7149         * common/config/rs6000/rs6000-common.c: Likewise.
7150         * common/config/rx/rx-common.c: Likewise.
7151         * common/config/s390/s390-common.c: Likewise.
7152         * common/config/sh/sh-common.c: Likewise.
7153         * common/config/sparc/sparc-common.c: Likewise.
7154         * common/config/tilegx/tilegx-common.c: Likewise.
7155         * common/config/tilepro/tilepro-common.c: Likewise.
7156         * common/config/v850/v850-common.c: Likewise.
7157         * common/config/visium/visium-common.c: Likewise.
7158         * common/config/xstormy16/xstormy16-common.c: Likewise.
7159         * common/config/xtensa/xtensa-common.c: Likewise.
7160         * invoke.texi (-fomit-frame-pointer): Update documentation.
7162 2017-11-16  Jan Hubicka  <hubicka@ucw.cz>
7164         * tree-cfg.c (gimple_find_sub_bbs): Do not compute freq.
7166 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7168         * cfg.c (scale_bbs_frequencies_int,
7169         cale_bbs_frequencies_gcov_type): Remove.
7170         * cfg.h (scale_bbs_frequencies_int,
7171         cale_bbs_frequencies_gcov_type): Remove.
7173 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7175         * tree-ssa-loop-manip.c
7176         (scale_dominated_blocks_in_loop): Update to profile counts.
7177         (tree_transform_and_unroll_loop): Likewise.
7179 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7181         * tree-vect-loop-manip.c (vect_do_peeling): Do not use
7182         scale_bbs_frequencies_int.
7184 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7186         * final.c (compute_alignments): Use counts rather than frequencies.
7188 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7190         * cfgloopanal.c: Include sreal.h
7191         (average_num_loop_insns): Use counts and sreal for accounting.
7193 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7195         * cfgloopmanip.c (duplicate_loop_to_header_edge): Cleanup profile
7196         manipulation.
7198 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7199             Kito Cheng  <kito.cheng@gmail.com>
7201         * config/nds32/constraints.md: Provide more constraints.
7202         * config/nds32/nds32.h (enum reg_class, REG_CLASS_NAMES,
7203         REG_CLASS_CONTENTS): Define R5_REG, R8_REG, and FRAME_POINTER_REG to
7204         support constraints usage.
7206 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7208         * config/nds32/constants.md (UNSPEC_VOLATILE_FUNC_RETURN): Remove.
7210 2017-11-16  Chung-Ju Wu  <jasonwucj@gmail.com>
7211             Kito Cheng  <kito.cheng@gmail.com>
7213         * config/nds32/nds32.opt: Add mext-perf, mext-perf2, mext-string.
7214         * config/nds32/nds32.opt: Refine the layout.
7215         * config/nds32/nds32.c (TARGET_EXT_PERF, TARGET_EXT_PERF2,
7216         TARGET_EXT_STRING): Support new options.
7217         * config/nds32/nds32.h: Likewise.
7218         * config/nds32/nds32.md: Likewise.
7219         * config/nds32/nds32-predicates.c: Likewise.
7220         * config/nds32/constraints.md: Likewise.
7221         * common/config/nds32/nds32-common.c: Likewise.
7223 2017-11-16  Julia Koval  <julia.koval@intel.com>
7225         PR target/82983
7226         * config/i386/gfniintrin.h: Add sse check.
7227         * config/i386/i386.c (ix86_expand_builtin): Fix gfni check.
7229 2017-11-16  Julia Koval  <julia.koval@intel.com>
7231         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VBMI2_SET,
7232         OPTION_MASK_ISA_AVX512VBMI2_UNSET): New.
7233         (ix86_handle_option): Handle -mavx512vbmi2.
7234         * config/i386/cpuid.h: Add bit_AVX512VBMI2.
7235         * config/i386/driver-i386.c (host_detect_local_cpu): Handle new bit.
7236         * config/i386/i386-c.c (__AVX512VBMI2__): New.
7237         * config/i386/i386.c (ix86_target_string): Handle -mavx512vbmi2.
7238         (ix86_valid_target_attribute_inner_p): Ditto.
7239         * config/i386/i386.h (TARGET_AVX512VBMI2, TARGET_AVX512VBMI2_P): New.
7240         * config/i386/i386.opt (mavx512vbmi2): New option.
7241         * doc/invoke.texi: Add new option.
7243 2017-11-16  Julia Koval  <julia.koval@intel.com>
7245         * config/i386/gfniintrin.h (_mm_gf2p8mul_epi8, _mm256_gf2p8mul_epi8,
7246         _mm_mask_gf2p8mul_epi8, _mm_maskz_gf2p8mul_epi8,
7247         _mm256_mask_gf2p8mul_epi8, _mm256_maskz_gf2p8mul_epi8,
7248         _mm512_mask_gf2p8mul_epi8, _mm512_maskz_gf2p8mul_epi8,
7249         _mm512_gf2p8mul_epi8): New intrinsics.
7250         * config/i386/i386-builtin-types.def
7251         (V64QI_FTYPE_V64QI_V64QI): New type.
7252         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8mulb_v64qi,
7253         __builtin_ia32_vgf2p8mulb_v64qi_mask, __builtin_ia32_vgf2p8mulb_v32qi,
7254         __builtin_ia32_vgf2p8mulb_v32qi_mask, __builtin_ia32_vgf2p8mulb_v16qi,
7255         __builtin_ia32_vgf2p8mulb_v16qi_mask): New builtins.
7256         * config/i386/sse.md (vgf2p8mulb_*): New pattern.
7257         * config/i386/i386.c (ix86_expand_args_builtin): Handle new type.
7259 2017-11-15  Uros Bizjak  <ubizjak@gmail.com>
7261         * config/i386/i386.c (x86_print_call_or_nop): Emit 5 byte nop
7262         explicitly as a stream of bytes.
7264 2017-11-15  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
7266         * config/rs6000/altivec.h (vec_xst_be): New #define.
7267         * config/rs6000/altivec.md (altivec_vperm_<mode>_direct): Rename
7268         and externalize from *altivec_vperm_<mode>_internal.
7269         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI): Remove macro
7270         instantiation.
7271         (XL_BE_V8HI): Likewise.
7272         (XL_BE_V4SI): Likewise.
7273         (XL_BE_V4SI): Likewise.
7274         (XL_BE_V2DI): Likewise.
7275         (XL_BE_V4SF): Likewise.
7276         (XL_BE_V2DF): Likewise.
7277         (XST_BE): Add BU_VSX_OVERLOAD_X macro instantiation.
7278         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Correct
7279         all array entries with these keys: VSX_BUILTIN_VEC_XL,
7280         VSX_BUILTIN_VEC_XL_BE, VSX_BUILTIN_VEC_XST.  Add entries for key
7281         VSX_BUILTIN_VEC_XST_BE.
7282         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Remove.
7283         (altivec_expand_builtin): Remove handling for VSX_BUILTIN_XL_BE_*
7284         built-ins.
7285         (altivec_init_builtins): Replace conditional calls to def_builtin
7286         for __builtin_vsx_ld_elemrev_{v8hi,v16qi} and
7287         __builtin_vsx_st_elemrev_{v8hi,v16qi} based on TARGET_P9_VECTOR
7288         with unconditional calls.  Remove calls to def_builtin for
7289         __builtin_vsx_le_be_<mode>.  Add a call to def_builtin for
7290         __builtin_vec_xst_be.
7291         * config/rs6000/vsx.md (vsx_ld_elemrev_v8hi): Convert define_insn
7292         to define_expand, and add alternate RTL generation for P8.
7293         (*vsx_ld_elemrev_v8hi_internal): New define_insn based on
7294         vsx_ld_elemrev_v8hi.
7295         (vsx_ld_elemrev_v16qi): Convert define_insn to define_expand, and
7296         add alternate RTL generation for P8.
7297         (*vsx_ld_elemrev_v16qi_internal): New define_insn based on
7298         vsx_ld_elemrev_v16qi.
7299         (vsx_st_elemrev_v8hi): Convert define_insn
7300         to define_expand, and add alternate RTL generation for P8.
7301         (*vsx_st_elemrev_v8hi_internal): New define_insn based on
7302         vsx_st_elemrev_v8hi.
7303         (vsx_st_elemrev_v16qi): Convert define_insn to define_expand, and
7304         add alternate RTL generation for P8.
7305         (*vsx_st_elemrev_v16qi_internal): New define_insn based on
7306         vsx_st_elemrev_v16qi.
7308 2017-11-15  H.J. Lu  <hongjiu.lu@intel.com>
7310         PR target/82990
7311         * config/i386/i386.c (pass_insert_vzeroupper::gate): Remove
7312         TARGET_AVX512ER check.
7313         (ix86_option_override_internal): Set MASK_VZEROUPPER if
7314         neither -mzeroupper nor -mno-zeroupper is used and
7315         TARGET_EMIT_VZEROUPPER is set.
7316         * config/i386/i386.h (TARGET_EMIT_VZEROUPPER): New.
7317         * config/i386/x86-tune.def: Add X86_TUNE_EMIT_VZEROUPPER.
7319 2017-11-15  Will Schmidt  <will_schmidt@vnet.ibm.com>
7321         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
7322         folding of vector compares.
7323         (fold_build_vec_cmp): New helper function.
7324         (fold_compare_helper): New helper function.
7325         (builtin_function_type): Add compare builtins to the list of functions
7326         having unsigned arguments.  Cosmetic updates to comment indentation.
7327         * config/rs6000/vsx.md (vcmpneb, vcmpneh, vcmpnew): Update to specify
7328         the not+eq combination.
7330 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7332         PR tree-optimization/82726
7333         PR tree-optimization/70754
7334         * tree-predcom.c (order_drefs_by_pos): New function.
7335         (combine_chains): Move code setting has_max_use_after to...
7336         (try_combine_chains): ...here.  New parameter.  Sort combined chains
7337         according to position information.
7338         (tree_predictive_commoning_loop): Update call to above function.
7339         (update_pos_for_combined_chains, pcom_stmt_dominates_stmt_p): New.
7341 2017-11-15  Bin Cheng  <bin.cheng@arm.com>
7343         PR tree-optimization/82726
7344         Revert
7345         2017-01-23  Bin Cheng  <bin.cheng@arm.com>
7347         PR tree-optimization/70754
7348         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
7349         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
7350         combined stmt before it if not NULL.
7351         (combine_chains): Process refs reversely and compute dominance point
7352         for root ref.
7354         Revert
7355         2017-02-23  Bin Cheng  <bin.cheng@arm.com>
7357         PR tree-optimization/79663
7358         * tree-predcom.c (combine_chains): Process refs in reverse order
7359         only for ZERO length chains, and add explaining comment.
7361 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7363         * config/arm/arm-cpus.in (armv8_3, ARMv8_3a, armv8.3-a): New
7364         * config/arm/arm-tables.opt (armv8.3-a): Regenerated.
7365         * doc/invoke.texi (ARM Options): Add armv8.3-a.
7367 2017-11-15  Tamar Christina  <tamar.christina@arm.com>
7369         * config/arm/arm.h (TARGET_DOTPROD): Add arm_arch8_2.
7371 2017-11-15  Martin Liska  <mliska@suse.cz>
7373         * tree-cfg.c (pass_warn_function_return::execute):
7374         Compare warn_return_type for greater than zero.
7376 2017-11-15  Sebastian Peryt  <sebastian.peryt@intel.com>
7378         PR target/82941
7379         PR target/82942
7380         * config/i386/i386.c (pass_insert_vzeroupper): Modify gate condition
7381         to return true on Xeon and not on Xeon Phi.
7382         (ix86_check_avx256_register): Changed to ...
7383         (ix86_check_avx_upper_register): ... this. Add extra check for
7384         VALID_AVX512F_REG_OR_XI_MODE.
7385         (ix86_avx_u128_mode_needed): Changed
7386         ix86_check_avx256_register to ix86_check_avx_upper_register.
7387         (ix86_check_avx256_stores): Changed to ...
7388         (ix86_check_avx_upper_stores): ... this. Changed
7389         ix86_check_avx256_register to ix86_check_avx_upper_register.
7390         (ix86_avx_u128_mode_after): Changed
7391         avx_reg256_found to avx_upper_reg_found. Changed
7392         ix86_check_avx256_stores to ix86_check_avx_upper_stores.
7393         (ix86_avx_u128_mode_entry): Changed
7394         ix86_check_avx256_register to ix86_check_avx_upper_register.
7395         (ix86_avx_u128_mode_exit): Ditto.
7396         * config/i386/i386.h: (host_detect_local_cpu): New define.
7398 2017-11-15  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
7400         * config/arm/xgene1.md (xgene1): Split into automatons
7401         xgene1_main, xgene1_decoder, xgene1_div, xgene1_simd.
7402         (xgene1_f_load): Adjust reservations and/or types.
7403         (xgene1_f_store): Likewise.
7404         (xgene1_load_pair): Likewise.
7405         (xgene1_store_pair): Likewise.
7406         (xgene1_fp_load1): Likewise.
7407         (xgene1_load1): Likewise.
7408         (xgene1_store1): Likewise.
7409         (xgene1_move): Likewise.
7410         (xgene1_alu): Likewise.
7411         (xgene1_simd): Likewise.
7412         (xgene1_bfm): Likewise.
7413         (xgene1_neon_load1): Likewise.
7414         (xgene1_neon_store1): Likewise.
7415         (xgene1_neon_logic): Likewise.
7416         (xgene1_neon_st1): Likewise.
7417         (xgene1_neon_ld1r): Likewise.
7418         (xgene1_alu_cond): Added.
7419         (xgene1_shift_reg): Likwise.
7420         (xgene1_bfx): Likewise.
7421         (xgene1_mul): Split into xgene1_mul32, xgene1_mul64.
7423 2017-11-15  Jakub Jelinek  <jakub@redhat.com>
7425         PR target/82981
7426         * internal-fn.c: Include gimple-ssa.h, tree-phinodes.h and
7427         ssa-iterators.h.
7428         (can_widen_mult_without_libcall): New function.
7429         (expand_mul_overflow): If only checking unsigned mul overflow,
7430         not result, and can do efficiently MULT_HIGHPART_EXPR, emit that.
7431         Don't use WIDEN_MULT_EXPR if it would involve a libcall, unless
7432         no other way works.  Add MULT_HIGHPART_EXPR + MULT_EXPR support.
7433         (expand_DIVMOD): Formatting fix.
7434         * expmed.h (expand_mult): Add NO_LIBCALL argument.
7435         * expmed.c (expand_mult): Likewise.  Use OPTAB_WIDEN rather
7436         than OPTAB_LIB_WIDEN if NO_LIBCALL is true, and allow it to fail.
7438         PR tree-optimization/82977
7439         * tree-ssa-strlen.c (strlen_optimize_stmt): Pass a reference to a copy
7440         constructed temporary to strlen_to_stridx.put.
7442 2017-11-15  Martin Liska  <mliska@suse.cz>
7444         * configure.ac: Remove -fkeep-inline-functions from coverage_flags.
7445         * configure: Regenerate.
7447 2017-11-15  Martin Liska  <mliska@suse.cz>
7449         PR target/82927
7450         * config/sh/sh-mem.cc: Use proper probability for
7451         REG_BR_PROB_NOTE.
7453 2017-11-14  Jeff Law  <law@redhat.com>
7455         * explow.c (anti_adjust_stack_and_probe_stack_clash): Avoid probing
7456         the red zone for stack_clash_protection_final_dynamic_probe targets
7457         when the total dynamic stack size is zero bytes.
7459         * tree-ssa-threadupdate.c (thread_through_all_blocks): Thread
7460         blocks is post order.
7462 2017-11-15  Alexandre Oliva  <aoliva@redhat.com>
7464         * dumpfile.h (TDF_COMPARE_DEBUG): New.
7465         * final.c (rest_of_clean_state): Set it for the
7466         -fcompare-debug dump.
7467         * tree-pretty-print.c (dump_generic_node): Omit OBJ_TYPE_REF
7468         class when TDF_COMPARE_DEBUG is set.
7470         * dwarf2out.c (gen_producer_string): Discard
7471         OPT_fcompare_debug.
7473 2017-11-15  Joseph Myers  <joseph@codesourcery.com>
7475         PR c/81156
7476         * doc/extend.texi (Other Builtins): Document __builtin_tgmath.
7477         * ginclude/tgmath.h (__tg_cplx, __tg_ldbl, __tg_dbl, __tg_choose)
7478         (__tg_choose_2, __tg_choose_3, __TGMATH_REAL_1_2)
7479         (__TGMATH_REAL_2_3): Remove macros.
7480         (__TGMATH_CPLX, __TGMATH_CPLX_2, __TGMATH_REAL, __TGMATH_REAL_2)
7481         (__TGMATH_REAL_3, __TGMATH_CPLX_ONLY): Define using
7482         __builtin_tgmath.
7483         (frexp, ldexp, nexttoward, scalbn, scalbln): Define using
7484         __TGMATH_REAL_2.
7485         (remquo): Define using __TGMATH_REAL_3.
7487 2017-11-14  Jeff Law  <law@redhat.com>
7489         * vr-values.c: New file with contents extracted from tree-vrp.c.
7490         * Makefile.in (OBJS): Add vr-values.o
7491         * tree-vrp.h (set_value_range_to_nonnull): Prototype.
7492         (set_value_range, set_and_canonicalize_value_range): Likewise.
7493         (vrp_bitmap_equal_p, range_is_nonnull): Likewise.
7494         (value_range_constant_singleton, symbolic_range_p): Likewise.
7495         (compare_values, compare_values_warnv, vrp_val_is_min): Likewise.
7496         (vrp_val_is_max, copy_value_range, set_value_range_to_value): Likewise.
7497         (extract_range_from_binary_expr_1, vrp_val_min, vrp_val_max): Likewise.
7498         (set_value_range_to_null, range_int_cst_p, opreand_less_p): Likewise.
7499         (find_case_label_range, find_case_label_index): Likewise.
7500         (zero_nonzero_bits_from_vr, overflow_comparison_p): Likewise.
7501         (range_int_cst_singleton_p, value_inside_range): Likewise.
7502         (get_single_symbol): Likewise.
7503         (switch_update): Move structure definition here.
7504         (to_remove_edges, to_update_switch_stmts): Provide externs.
7505         * tree-vrp.c: Move all methods for vr-values class to vr-values.c
7506         (vrp_val_max, vrp_val_min, vrp_val_is_max): Make externally visible.
7507         (vrp_val_is_min, set_value_range): Likewise.
7508         (set_and_canonicalize_value_range, copy_value_range): Likewise.
7509         (set_value_range_to_value, set_value_range_to_nonnull): Likewise.
7510         (set_value_range_to_null, vrp_bitmap_equal_p): Likewise.
7511         (range_is_nonnull, range_int_cst_p): Likewwise.
7512         (range_int_cst_singleton_p, symbolic_range_p): Likewise.
7513         (get_single_symbol, operand_less_p): Likewise
7514         (compare_values_warnv, compare_values): Likewise.
7515         (value_inside_range, value_range_constant_singleton): Likewise.
7516         (zero_nonzero_bitgs_from_vr): Likewise.
7517         (extract_range_from_binary_expr_1): Likewise.
7518         (overflow_comparison_p): Likewise.
7519         (to_remove_edges, to_update_switch_stmts): Likewise.
7520         (find_case_label-index, find_case_label_range): Likewise.
7521         (switch_update, set_value_range_to_nonnegative): Remove.
7522         (set_value_range_to_truthvalue): Likewise.
7523         (symbolic_range_based_on_p, gimple_assign_nonzero_p): Likewise.
7524         (gimple_stmt_nonzero_p, compare_ranges): Likewise.
7525         (compare_range_with_value, vrp_valueize, vrp_valueize_1): Likewise.
7526         (find_case_label_ranges, test_for_singularity): Likewise.
7527         (range_fits_type_p, simplify_conversion_using_ranges): LIkewise.
7528         (x_vr_values): Move to its remaining use site.
7530 2017-11-10  Jeff Law  <law@redhat.com>
7532         * vr-values.h (VR_INITIALIZER): Move #define here.
7533         * gimple-ssa-evrp.c: New file with contents extracted from tree-vrp.c
7534         * Makefile.in (OBJS): Add tree-evrp.o
7535         * tree-vrp.h (assert_info): Move structure definition here.
7536         (set_value_range_to_varying): Prototype.
7537         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7538         (infer_value_range, register_edge_assert_for): Likewise.
7539         (stmt_interesting_for_vrp): Likewise.
7540         * tree-vrp.c: Move all methods for evrp class into gimple-ssa-evrp.c.
7541         (set_value_range_to_varying): No longer static.
7542         (vrp_operand_equal_p, range_includes_zero_p): Likewise.
7543         (infer_value_range, register_edge_assert_for): Likewise.
7545 2017-11-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
7547         * config/rs6000/rs6000.md (bswapdi2): On 32-bit ISA 3.0, don't
7548         generate the XXBRD instruction.
7550         * config/rs6000/rs6000-c.c (is_float128_p): New helper function.
7551         (rs6000_builtin_type_compatible): Treat _Float128 and long double
7552         as being compatible if -mabi=ieeelongdouble.
7553         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_HW_1): New macros
7554         to setup float128 built-ins with hardware support.
7555         (BU_FLOAT128_HW_2): Likewise.
7556         (BU_FLOAT128_HW_3): Likewise.
7557         (BU_FLOAT128_HW_VSX_1): Likewise.
7558         (BU_FLOAT128_HW_VSX_2): Likewise.
7559         (scalar_extract_expq): Change float128 built-in functions to
7560         accommodate having both KFmode and TFmode functions.  Use the
7561         KFmode variant as the default.
7562         (scalar_extract_sigq): Likewise.
7563         (scalar_test_neg_qp): Likewise.
7564         (scalar_insert_exp_q): Likewise.
7565         (scalar_insert_exp_qp): Likewise.
7566         (scalar_test_data_class_qp): Likewise.
7567         (sqrtf128_round_to_odd): Delete processing the round to odd
7568         built-in functions as special built-in functions, and define them
7569         as float128 built-ins.  Use the KFmode variant as the default.
7570         (truncf128_round_to_odd): Likewise.
7571         (addf128_round_to_odd): Likewise.
7572         (subf128_round_to_odd): Likewise.
7573         (mulf128_round_to_odd): Likewise.
7574         (divf128_round_to_odd): Likewise.
7575         (fmaf128_round_to_odd): Likewise.
7576         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
7577         support for KFmode and TFmode xststdcqp calls.
7578         (rs6000_expand_builtin): If long double is IEEE 128-bit floating
7579         point, switch the built-in handlers for the get/set float128
7580         exponent, get float128 mantissa, float128 test built-ins, and the
7581         float128 round to odd built-in functions.  Eliminate creating the
7582         float128 round to odd built-in functions as special built-ins.
7583         (rs6000_init_builtins): Eliminate special creation of the float128
7584         round to odd built-in functions.
7585         * config/rs6000/vsx.md (xsxexpqp_<mode>): Change float128 built-in
7586         function insns to support both TFmode and KFmode variants.
7587         (xsxsigqp_<mode>): Likewise.
7588         (xsiexpqpf_<mode>): Likewise.
7589         (xsiexpqp_<mode>): Likewise.
7590         (xststdcqp_<mode>): Likewise.
7591         (xststdcnegqp_<mode>): Likewise.
7592         (xststdcqp_<mode>): Likewise.
7594 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7596         * tree-ssa-threadupdate.c (compute_path_counts): Remove
7597         unused path_in_freq_ptr parameter.
7598         (ssa_fix_duplicate_block_edges): Do not pass around path_in_freq
7600 2017-11-14  Jan Hubicka  <hubicka@ucw.cz>
7602         * ipa-inline.c (edge_badness): Dump sreal frequency.
7603         (compute_inlined_call_time): Match natural implementaiton ...
7604         * ipa-fnsummary.c (estimate_edge_size_and_time): ... here; remove
7605         forgotten division by CGRAPH_FREQ_BASE.
7607 2017-11-14  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
7609         * config.gcc (*-*-solaris2*): Enable default_use_cxa_atexit since
7610         Solaris 11.  Update comment.
7611         * configure.ac (gcc_cv_ld_pid): Adapt comment for Solaris 12
7612         renaming.
7613         * config/sol2.h (STARTFILE_SPEC): Likewise.
7614         * configure: Regenerate.
7616 2017-11-14  Carl Love  <cel@us.ibm.com>
7618         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Remove
7619         le_ and be_ prefixes to swap* variables.  Remove
7620         if (VECTOR_ELT_ORDER_BIG) statement. Remove E_V16QImode case
7621         statements.
7623 2017-11-14  Jason Merrill  <jason@redhat.com>
7625         Support GTY((cache)) on hash_map.
7626         * hash-traits.h (ggc_remove): Add ggc_maybe_mx member function.
7627         (ggc_cache_remove): Override it instead of ggc_mx.
7628         * hash-table.h (gt_ggc_mx): Call it instead of ggc_mx.
7629         (gt_cleare_cache): Call ggc_mx instead of gt_ggc_mx.
7630         * hash-map-traits.h (simple_hashmap_traits): Add maybe_mx member.
7631         (simple_cache_map_traits): Override maybe_mx.
7632         * hash-map.h (hash_entry): Add ggc_maybe_mx and keep_cache_entry.
7633         (hash_map): Friend gt_cleare_cache.
7634         (gt_cleare_cache): New.
7635         * tree.h (tree_cache_traits): New hash_map traits class.
7636         (tree_cache_map): New typedef.
7638 2017-11-14  Richard Biener  <rguenther@suse.de>
7640         * tree-cfgcleanup.c (cleanup_control_expr_graph): Remove first_p
7641         paramter and handling.
7642         (cleanup_control_flow_bb): Likewise.
7643         (cleanup_control_flow_pre): New helper performing a DFS walk
7644         to call cleanup_control_flow_bb in PRE order.
7645         (cleanup_tree_cfg_1): Do the first phase of cleanup_control_flow_bb
7646         via cleanup_control_flow_pre.
7648 2017-11-14  James Greenhalgh  <james.greenhalgh@arm.com>
7650         * config/aarch64/aarch64-simd.md
7651         (aarch64_simd_bsl<mode>_internal): Remove DImode.
7652         (*aarch64_simd_bsl<mode>_alt): Likewise.
7653         (aarch64_simd_bsldi_internal): New.
7654         (aarch64_simd_bsldi_alt): Likewise.
7656 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7658         * tracer.c (better_p): Do not compare frequencies.
7659         * reg-stack.c (better_edge): Likewise.
7660         * shrink-wrap.c (try_shrink_wrapping): Do not convert to gcov counts
7661         and back.
7663 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7665         * auto-profile.c (afdo_annotate_cfg): Use update_max_bb_count.
7666         * cgraphunit.c (cgraph_node::expand_thunk): Use update_max_bb_count.
7667         * ipa-utils.c (ipa_merge_profiles): Use update_max_bb_count.
7668         * lto-streamer-in.c (input_function): Use update_max_bb_count.
7669         * omp-expand.c (expand_omp_taskreg): Use update_max_bb_count.
7670         * predict.c (maybe_hot_frequency_p): Inline to ...
7671         (maybe_hot_count_p): ... here; rewrite to counts.
7672         (counts_to_freqs): Rename to ...
7673         (update_max_bb_count): ... this one.
7674         (expensive_function_p): Use counts.
7675         (estimate_bb_frequencies): Update.
7676         (rebuild_frequencies): Update.
7677         * predict.h (counts_to_freqs): Rename to ...
7678         (update_max_bb_count): ... this one.
7679         * profile.c (compute_branch_probabilities): Add debug info
7680         * tree-inline.c (expand_call_inline): Update debug info.
7681         (optimize_inline_calls): Use update_max_bb_count..
7682         (tree_function_versioning): Use update_max_bb_count..
7683         * value-prof.c (gimple_value_profile_transformations):
7684         Do not use update_max_bb_count.
7686 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7688         * ipa-inline.c (compute_uninlined_call_time, compute_inlined_call_time):
7689         always use frequencies.
7691 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7693         * bb-reorder.c: Remove frequencies from comments.
7694         (better_edge_p): Use profile counts.
7695         (find_traces): Dump profile counts.
7696         (rotate_loop): Use profile counts.
7697         (find_traces_1_round): Likewise.
7698         (connect_better_edge_p): Use counts instead of probabilities for
7699         reverse walk.
7700         (copy_bb_p): Drop early check for non-0 frequency.
7701         (sanitize_hot_paths): Update comments.
7703 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7705         * ipa-split.c (struct split_point): Add count.
7706         (consider_split): Do not compute incoming frequency; compute incoming
7707         count and store it to split_point.
7708         (split_function): Set count of the call to split part correctly.
7710 2017-11-13  Carl Love  <cel@us.ibm.com>
7712         * config/rs6000/altivec.md (altivec_vsumsws_be): Add define_expand.
7714 2017-11-13  Tom Tromey  <tom@tromey.com>
7716         * doc/cpp.texi (Variadic Macros): Document __VA_OPT__.
7718 2017-11-13  Carl Love  <cel@us.ibm.com>
7720         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins):
7721         Add support for builtins:
7722         unsigned int vec_first_{,miss}_match_{,or_eos}index,
7723         vector {un,}signed {char,int,short},
7724         vector {un,}signed {char,int,short}) arguments.
7725         * config/rs6000/rs6000-builtin.def (VFIRSTMATCHINDEX,
7726         VFIRSTMATCHOREOSINDEX, VFIRSTMISMATCHINDEX, VFIRSTMISMATCHOREOSINDEX):
7727         Add BU_P9V_AV_2 expansions for the builtins.
7728         * config/rs6000/altivec.h (vec_first_match_index,
7729         vec_first_mismatch_index, vec_first_match_or_eos_index,
7730         vec_first_mismatch_or_eos_index): Add #defines for the builtins.
7731         * config/rs6000/rs6000-protos.h (bytes_in_mode): Add
7732         new extern declaration.
7733         * config/rs6000/rs6000.c (bytes_in_mode): Add new function.
7734         * config/rs6000/vsx.md (first_match_index_<mode>,
7735         first_match_or_eos_index_<mode>, first_mismatch_index_<mode>,
7736         first_mismatch_or_eos_index_<mode>): Add define expand.
7737         (vctzlsbb_<mode>): Add mode field to define_insn for vctzlsbb.
7738         * doc/extend.texi: Update the built-in documenation file for the new
7739         built-in functions.
7741 2017-11-13  Michael Meissner  <meissner@linux.vnet.ibm.com>
7743         * match.pd: Convert fminf<N>, fminf<N>x, fmax<N>, and fmax<N>x
7744         into the min/max operations for _Float<N> and _Float<N>X types.
7746 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7748         PR lto/81351
7749         * dwarf2out.c (do_eh_frame): New static variable.
7750         (dwarf2out_begin_prologue): Set it.
7751         (dwarf2out_frame_finish): Test it instead of dwarf2out_do_eh_frame.
7753 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7755         * tree-ssa-coalesce.c (coalesce_cost): Fix formating.
7757         * tree-ssa-sink.c (select_best_block): Do not use frequencies.
7759 2017-11-13  Eric Botcazou  <ebotcazou@adacore.com>
7761         PR lto/81351
7762         * debug.h (dwarf2out_do_eh_frame): Declare.
7763         * dwarf2cfi.c (dwarf2out_do_eh_frame): New predicate.
7764         (dwarf2out_do_frame): Use it.
7765         (dwarf2out_do_cfi_asm): Likewise.
7766         * dwarf2out.c (dwarf2out_frame_finish): Likewise.
7767         (dwarf2out_assembly_start): Likewise.
7768         (dwarf2out_begin_prologue): Fix comment.
7769         * toplev.c (compile_file): Always call dwarf2out_frame_finish
7770         if the target needs either debug or unwind DWARF2 info.
7771         * lto-opts.c (lto_write_options): Do not save -fexceptions,
7772         -fnon-call-exceptions, -ffp-contract, -fmath-errno, -fsigned-zeros,
7773         -ftrapping-math, -ftrapv and -fwrapv.
7775 2017-11-13  Jan Hubicka  <hubicka@ucw.cz>
7777         * cgraph.c (cgraph_edge::sreal_frequency): New function.
7778         * cgraph.h (cgraph_edge::sreal_frequency): Declare.
7779         * ipa-fnsummary.c (dump_ipa_call_summary): Use sreal_frequency.
7780         (estimate_edge_size_and_time): Likewise.
7781         (ipa_merge_fn_summary_after_inlining): Likewise.
7782         * ipa-inline.c (cgraph_freq_base_rec): Remove.
7783         (compute_uninlined_call_time): Use sreal_frequency.
7784         (compute_inlined_call_time): Likewise.
7785         (ipa_inline): Do not initialize cgraph_freq_base_rec.
7786         * profile-count.c: Include sreal.h.
7787         (profile_count::to_sreal_scale): New.
7788         * profile-count.h: Forward declare sreal.
7789         (profile_count::to_sreal_scale): Declare.
7791 2017-11-13  Nathan Sidwell  <nathan@acm.org>
7793         * diagnostic.c (maybe_line_and_column): New.
7794         (diagnostic_get_location_text): Use it.
7795         (diagnostic_report_current_module): Likewise.
7796         (test_diagnostic_get_location_text): Add tests.
7798 2017-11-13  Luis Machado  <luis.machado@linaro.org>
7800         * doc/md.texi (Specifying processor pipeline description): Fix
7801         incorrect latency for the div instruction example.
7803 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7805         PR tree-optimization/78821
7806         * gimple-ssa-store-merging.c (compatible_load_p): Don't require
7807         that bit_not_p is the same.
7808         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
7809         (split_group): Count precisely bit_not_p bits in each statement.
7810         (invert_op): New function.
7811         (imm_store_chain_info::output_merged_store): Use invert_op to
7812         emit BIT_XOR_EXPR with a xor_mask instead of BIT_NOT_EXPR if some
7813         but not all orig_stores have BIT_NOT_EXPR in the corresponding spots.
7815 2017-11-13  Martin Liska  <mliska@suse.cz>
7817         * gcov.c (struct coverage_info): Remove typedef of coverage_t.
7818         (struct source_info): Likewise.
7819         (add_branch_counts): Likewise.
7820         (add_line_counts): Likewise.
7821         (function_summary): Likewise.
7822         (output_intermediate_line): Likewise.
7823         (generate_results): Likewise.
7825 2017-11-13  Martin Liska  <mliska@suse.cz>
7827         * gcov.c (struct block_info): Remove typedef for block_t.
7828         (struct line_info): Likewise.
7829         (line_info::has_block): Likewise.
7830         (EXIT_BLOCK): Likewise.
7831         (unblock): Likewise.
7832         (circuit): Likewise.
7833         (get_cycles_count): Likewise.
7834         (process_file): Likewise.
7835         (read_graph_file): Likewise.
7836         (solve_flow_graph): Likewise.
7837         (find_exception_blocks): Likewise.
7838         (add_line_counts): Likewise.
7839         (accumulate_line_info): Likewise.
7840         (output_line_details): Likewise.
7842 2017-11-13  Martin Liska  <mliska@suse.cz>
7844         * gcov.c (struct arc_info): Remove typedef for arc_t.
7845         (struct line_info): Likewise.
7846         (add_branch_counts): Likewise.
7847         (output_branch_count): Likewise.
7848         (function_info::~function_info): Likewise.
7849         (circuit): Likewise.
7850         (output_intermediate_line): Likewise.
7851         (read_graph_file): Likewise.
7852         (solve_flow_graph): Likewise.
7853         (find_exception_blocks): Likewise.
7854         (add_line_counts): Likewise.
7855         (accumulate_line_info): Likewise.
7856         (output_line_details): Likewise.
7857         (output_function_details): Likewise.
7859 2017-11-13  Martin Liska  <mliska@suse.cz>
7861         * gcov.c (struct function_info): Remove typedef for function_t.
7862         (struct source_info): Likewise.
7863         (source_info::get_functions_at_location): Likewise.
7864         (solve_flow_graph): Likewise.
7865         (find_exception_blocks): Likewise.
7866         (add_line_counts): Likewise.
7867         (output_intermediate_file): Likewise.
7868         (process_file): Likewise.
7869         (generate_results): Likewise.
7870         (release_structures): Likewise.
7871         (read_graph_file): Likewise.
7872         (read_count_file): Likewise.
7873         (accumulate_line_counts): Likewise.
7874         (output_lines): Likewise.
7876 2017-11-13  Martin Liska  <mliska@suse.cz>
7878         * gcov.c (function_info::function_info): Remove num_counts
7879         and add vector<gcov_type>.
7880         (function_info::~function_info): Use the vector.
7881         (process_file): Likewise.
7882         (read_graph_file): Likewise.
7883         (read_count_file): Likewise.
7884         (solve_flow_graph): Likewise.
7886 2017-11-13  Martin Liska  <mliska@suse.cz>
7888         * gcov.c (function_info::is_artificial): New function.
7889         (process_file): Erase all artificial early.
7890         (generate_results): Skip as all artificial are already
7891         removed.
7893 2017-11-13  Martin Liska  <mliska@suse.cz>
7895         * gcov.c (read_graph_file): Store to global vector of functions.
7896         (read_count_file): Iterate the vector.
7897         (process_file): Likewise.
7898         (generate_results): Likewise.
7899         (release_structures): Likewise.
7901 2017-11-13  Jakub Jelinek  <jakub@redhat.com>
7903         PR tree-optimization/82954
7904         * gimple-ssa-store-merging.c
7905         (imm_store_chain_info::coalesce_immediate_stores): If
7906         !infof->ops[N].base_addr, split group if info->ops[N].base_addr.
7908 2017-11-13  Richard Sandiford  <richard.sandiford@linaro.org>
7910         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
7911         Upddate call to ENDIAN_LANE_N.
7912         (aarch64_<sur>dot_lane<vsi2qi>): Use aarch64_endian_lane_rtx.
7913         (aarch64_<sur>dot_laneq<vsi2qi>): Likewise.
7914         (*aarch64_simd_vec_copy_lane<mode>): Update calls to ENDIAN_LANE_N
7915         and use aarch64_endian_lane_rtx.
7916         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Likewise.
7918 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7920         * config/riscv/riscv.h (ASM_OUTPUT_LABELREF): Wrap in do {} while (0).
7922 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7924         * config/elfos.h (ASM_OUTPUT_ASCII): Remove semicolon after macro body.
7926 2017-11-12  Tom de Vries  <tom@codesourcery.com>
7928         * config/cr16/cr16.h (ASM_OUTPUT_LABELREF): Remove semicolon after macro
7929         body.
7930         * config/powerpcspe/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7931         * config/rs6000/xcoff.h (ASM_OUTPUT_LABELREF): Same.
7932         * defaults.h (ASM_OUTPUT_LABELREF): Same.
7934 2017-11-11  Martin Sebor  <msebor@redhat.com>
7936         PR c/81117
7937         * doc/extend.texi (attribute nonstring): Remove spurious argument.
7939         PR bootstrap/82948
7940         * prefic.c (translate_name): Replace strncpy with memcpy to
7941         avoid -Wstringop-truncation.
7943 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7945         * tree-ssa-loop-im.c (execute_sm_if_changed): Do not compute freq_sum.
7947 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7949         * predict.c (maybe_hot_frequency_p): Do not use cfun.
7951 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7953         * tree-ssa-tail-merge.c (replace_block_by): Fix and re-enable profile
7954         merging.
7956 2017-11-10  Michael Meissner  <meissner@linux.vnet.ibm.com>
7958         * config/rs6000/rs6000.md (bswaphi2_reg): On ISA 3.0 systems,
7959         enable generating XXBRH if the value is in a vector register.
7960         (bswapsi2_reg): On ISA 3.0 systems, enable generating XXBRW if the
7961         value is in a vector register.
7962         (bswapdi2_reg): On ISA 3.0 systems, always use XXBRD to do
7963         register to register bswap64's instead of doing the GPR sequence
7964         used on previous machines.
7965         (bswapdi2_xxbrd): New insn.
7966         (bswapdi2_reg): Disallow on ISA 3.0.
7967         (register to register bswap64 splitter): Do not split the insn on
7968         ISA 3.0 systems that use XXBRD.
7970 2017-11-10  Martin Sebor  <msebor@redhat.com>
7972         PR c/81117
7973         * config/darwin-c.c (framework_construct_pathname): Replace strncpy
7974         with memcpy.
7975         (find_subframework_file): Same.
7977 2017-11-10  Jan Hubicka  <hubicka@ucw.cz>
7979         * auto-profile.c (afdo_indirect_call): Drop frequency.
7980         * cgraph.c (symbol_table::create_edge): Drop frequency argument.
7981         (cgraph_node::create_edge): Drop frequency argument.
7982         (cgraph_node::create_indirect_edge): Drop frequency argument.
7983         (cgraph_edge::make_speculative): Drop frequency arguments.
7984         (cgraph_edge::resolve_speculation): Do not update frequencies
7985         (cgraph_edge::dump_edge_flags): Do not dump frequency.
7986         (cgraph_node::dump): Check consistency in IPA mode.
7987         (cgraph_edge::maybe_hot_p): Use IPA counter.
7988         (cgraph_edge::verify_count_and_frequency): Rename to ...
7989         (cgraph_edge::verify_count): ... this one; drop frequency checking.
7990         (cgraph_node::verify_node): Update.
7991         * cgraph.h (struct cgraph_edge): Drop frequency.
7992         (cgraph_edge::frequency): New function.
7993         * cgraphbuild.c (pass_build_cgraph_edges::execute): Donot pass
7994         frequencies.
7995         (cgraph_edge::rebuild_edges): Likewise.
7996         * cgraphclones.c (cgraph_edge::clone): Scale only counts.
7997         (duplicate_thunk_for_node): Do not pass frequency.
7998         (cgraph_node::create_clone): Scale only counts.
7999         (cgraph_node::create_virtual_clone): Do not pass frequency.
8000         (cgraph_node::create_edge_including_clones): Do not pass frequency.
8001         (cgraph_node::create_version_clone): Do not pass frequency.
8002         * cgraphunit.c (cgraph_node::analyze): Do not pass frequency.
8003         (cgraph_node::expand_thunk): Do not pass frequency.
8004         (cgraph_node::create_wrapper): Do not pass frequency.
8005         * gimple-iterator.c (update_call_edge_frequencies): Do not pass
8006         frequency.
8007         * gimple-streamer-in.c (input_bb): Scale only IPA counts.
8008         * ipa-chkp.c (chkp_produce_thunks): Do not pass frequency.
8009         * ipa-cp.c (ipcp_lattice::print): Use frequency function.
8010         (gather_caller_stats): Use frequency function.
8011         (ipcp_cloning_candidate_p): Use frequency function.
8012         (ipcp_propagate_stage): Use frequency function.
8013         (get_info_about_necessary_edges): Use frequency function.
8014         (update_profiling_info): Update only IPA profile.
8015         (update_specialized_profile): Use frequency functoin.
8016         (perhaps_add_new_callers): Update only IPA profile.
8017         * ipa-devirt.c (ipa_devirt): Use IPA profile.
8018         * ipa-fnsummary.c (redirect_to_unreachable): Do not set frequrency.
8019         (dump_ipa_call_summary): Use frequency function.
8020         (estimate_edge_size_and_time): Use frequency function.
8021         (ipa_merge_fn_summary_after_inlining): Use frequency function.
8022         * ipa-inline-analysis.c (do_estimate_edge_time): Use IPA profile.
8023         * ipa-inline-transform.c (update_noncloned_frequencies): Rename to ..
8024         (update_noncloned_counts): ... ths one; scale counts only.
8025         (clone_inlined_nodes): Do not scale frequency.
8026         (inline_call): Do not pass frequency.
8027         * ipa-inline.c (compute_uninlined_call_time): Use IPA profile.
8028         (compute_inlined_call_time): Use IPA profile.
8029         (want_inline_small_function_p): Use IPA profile.
8030         (want_inline_self_recursive_call_p): Use IPA profile.
8031         (edge_badness): Use IPA profile.
8032         (lookup_recursive_calls): Use IPA profile.
8033         (recursive_inlining): Do not pass frequency.
8034         (resolve_noninline_speculation): Do not update frequency.
8035         (inline_small_functions): Collect max of IPA profile.
8036         (dump_overall_stats): Dump IPA porfile.
8037         (dump_inline_stats): Dump IPA porfile.
8038         (ipa_inline): Collect IPA stats.
8039         * ipa-inline.h (clone_inlined_nodes): Update prototype.
8040         * ipa-profile.c (ipa_propagate_frequency_1): Use frequency function.
8041         (ipa_propagate_frequency): Use frequency function.
8042         (ipa_profile): Cleanup.
8043         * ipa-prop.c (ipa_make_edge_direct_to_target): Do not pass frequency
8044         * ipa-utils.c (ipa_merge_profiles): Merge all profiles.
8045         * lto-cgraph.c (lto_output_edge): Do not stream frequency.
8046         (input_node): Do not stream frequency.
8047         (input_edge): Do not stream frequency.
8048         (merge_profile_summaries): Scale only IPA profiles.
8049         * omp-simd-clone.c (simd_clone_adjust): Do not pass frequency.
8050         * predict.c (drop_profile): Do not recompute frequency.
8051         * trans-mem.c (ipa_tm_insert_irr_call): Do not pass frequency.
8052         (ipa_tm_insert_gettmclone_call): Do not pass frequency.
8053         * tree-cfg.c (execute_fixup_cfg): Drop profile to global0 if needed.
8054         * tree-chkp.c (chkp_copy_bounds_for_assign): Do not pass frequency.
8055         * tree-emutls.c (gen_emutls_addr): Do not pass frequency.
8056         * tree-inline.c (copy_bb): Do not scale frequency.
8057         (expand_call_inline): Do not scale frequency.
8058         (tree_function_versioning): Do not scale frequency.
8059         * ubsan.c (ubsan_create_edge): Do not pass frequency.
8061 2017-11-10  Julia Koval  <julia.koval@intel.com>
8063         * config/i386/gfniintrin.h (_mm_gf2p8affine_epi64_epi8)
8064         (_mm256_gf2p8affine_epi64_epi8, _mm_mask_gf2p8affine_epi64_epi8)
8065         (_mm_maskz_gf2p8affine_epi64_epi8, _mm256_mask_gf2p8affine_epi64_epi8)
8066         (_mm256_maskz_gf2p8affine_epi64_epi8)
8067         (_mm512_mask_gf2p8affine_epi64_epi8, _mm512_gf2p8affine_epi64_epi8)
8068         (_mm512_maskz_gf2p8affine_epi64_epi8): New intrinsics.
8069         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineqb_v64qi)
8070         (__builtin_ia32_vgf2p8affineqb_v32qi)
8071         (__builtin_ia32_vgf2p8affineqb_v16qi): New builtins.
8072         * config/i386/sse.md (vgf2p8affineqb_<mode><mask_name>): New pattern.
8074 2017-11-10  Tamar Christina  <tamar.christina@arm.com>
8076         PR target/82641
8077         * config/arm/arm.c
8078         (arm_option_override): Refactor.
8079         (arm_option_reconfigure_globals): New.
8080         (arm_options_perform_arch_sanity_checks): New.
8081         * config/arm/arm-protos.h (arm_option_reconfigure_globals):
8082         New prototype.
8083         (arm_options_perform_arch_sanity_checks): Likewise
8085 2017-11-10  Pat Haugen  <pthaugen@us.ibm.com>
8087         * rs6000/power9.md (power9-qpdiv): Correct DFU pipe usage.
8088         (power9-qpmul): New.
8089         * rs6000/rs6000.md ("type" attr): Add qmul.
8090         (mul<mode>3, fma<mode>4_hw, *fms<mode>4_hw, *nfma<mode>4_hw,
8091         *nfms<mode>4_hw, mul<mode>3_odd, fma<mode>4_odd, *fms<mode>4_odd,
8092         *nfma<mode>4_odd, *nfms<mode>4_odd): Change type to qmul.
8094 2017-11-10  Martin Sebor  <msebor@redhat.com>
8096         PR c/81117
8097         * builtins.c (compute_objsize): Handle arrays that
8098         compute_builtin_object_size likes to fail for.  Make extern.
8099         * builtins.h (compute_objsize): Declare.
8100         (check_strncpy_sizes): New function.
8101         (expand_builtin_strncpy): Call check_strncpy_sizes.
8102         * gimple-fold.c (gimple_fold_builtin_strncpy): Implement
8103         -Wstringop-truncation.
8104         (gimple_fold_builtin_strncat): Same.
8105         * gimple.c (gimple_build_call_from_tree): Set call location.
8106         * tree-ssa-strlen.c (strlen_to_stridx): New global variable.
8107         (maybe_diag_bound_equal_length, is_strlen_related_p): New functions.
8108         (handle_builtin_stxncpy, handle_builtin_strncat): Same.
8109         (handle_builtin_strlen): Use strlen_to_stridx.
8110         (strlen_optimize_stmt): Handle flavors of strncat, strncpy, and
8111         stpncpy.
8112         Use strlen_to_stridx.
8113         (pass_strlen::execute): Release strlen_to_stridx.
8114         * doc/invoke.texi (-Wsizeof-pointer-memaccess): Document enhancement.
8115         (-Wstringop-truncation): Document new option.
8117 2017-11-10  Martin Liska  <mliska@suse.cz>
8119         PR gcov-profile/82702
8120         * gcov.c (main): Handle intermediate files in a different
8121         way.
8122         (get_gcov_intermediate_filename): New function.
8123         (output_gcov_file): Remove support of intermediate files.
8124         (generate_results): Allocate intermediate file.
8125         (release_structures): Clean-up properly fn_end.
8126         (output_intermediate_file): Start iterating with line 1.
8128 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8130         PR tree-optimization/82929
8131         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8132         ops_swapped_p non-static data member.
8133         (store_immediate_info::store_immediate_info): Clear it.
8134         (imm_store_chain_info::coalesce_immediate_stores): If swapping
8135         ops set ops_swapped_p.
8136         (count_multiple_uses): Handle ops_swapped_p.
8138 2017-11-10  Martin Liska  <mliska@suse.cz>
8140         * coverage.c (coverage_init): Stream information about
8141         support of has_unexecuted_blocks.
8142         * doc/gcov.texi: Document that.
8143         * gcov-dump.c (dump_gcov_file): Support it in gcov_dump tool.
8144         * gcov.c (read_graph_file): Likewise.
8145         (output_line_beginning): Fix a small issue with
8146         color output.
8148 2017-11-10  Bin Cheng  <bin.cheng@arm.com>
8150         * tree-predcom.c (determine_roots_comp): Avoid memory leak by freeing
8151         reference of trivial component.
8153 2017-11-10  Jakub Jelinek  <jakub@redhat.com>
8155         PR bootstrap/82916
8156         * gimple-ssa-store-merging.c
8157         (pass_store_merging::terminate_all_aliasing_chains): For
8158         gimple_store_p stmts also call refs_output_dependent_p.
8160         PR rtl-optimization/82913
8161         * compare-elim.c (try_merge_compare): Punt if def_insn is not
8162         single set.
8164 2017-11-09  Jeff Law  <law@redhat.com>
8166         * vr-values.h: New file with vr_values class.
8167         * tree-vrp.c: Include vr-values.h
8168         (vrp_value_range_pool, vrp_equiv_obstack, num_vr_values): Move static
8169         data objects into the vr_values class.
8170         (vr_value, values_propagated, vr_phi_edge_counts): Likewise.
8171         (get_value_range): Make it a member function within vr_values class.
8172         (set_defs_to_varying, update_value_range, add_equivalence): Likewise.
8173         (vrp_stmt_computes_nonzero_p, op_with_boolean_value_range_p): Likewise.
8174         (op_with_constant_singleton_value_range): Likewise.
8175         (extract_range_for_var_from_comparison_expr): Likewise.
8176         (extract_range_from_assert, extract_range_from_ssa_name): Likewise.
8177         (extract_range_from_binary_expr): Likewise.
8178         (extract_range_from_unary_expr): Likewise.
8179         (extract_range_from_cond_expr, extrat_range_from_comparison): Likewise.
8180         (check_for_binary_op_overflow, extract_range_basic): Likewise.
8181         (extract_range_from_assignment, adjust_range_with_scev): Likewise.
8182         (dump_all_value_ranges, get_vr_for_comparison): Likewise.
8183         (compare_name_with_value, compare_names): Likewise.
8184         (vrp_evaluate_conditional_warnv_with_ops_using_ranges): Likewise.
8185         (vrp_evaluate_conditional_warnv_with_ops): Likewise.  Remove prototype.
8186         (vrp_evaluate_conditional, vrp_visit_cond_stmt): Likewise.
8187         (vrp_visit_switch_stmt, extract_range_from_stmt): Likewise.
8188         (extract_range_from_phi_node): Likewise.
8189         (simplify_truth_ops_using_ranges): Likewise.
8190         (simplify_div_or_mod_using_ranges): Likewise.
8191         (simplify_min_or_max_using_ranges, simplify_abs_using_ranges): Likewise.
8192         (simplify_bit_ops_using_ranges, simplify_cond_using_ranges_1): Likewise.
8193         (simplify_cond_using_ranges_2, simplify_switch_using_ranges): Likewise.
8194         (simplify_float_conversion_using_ranges): Likewise.
8195         (simplify_internal_call_using_ranges): Likewise.
8196         (two_valued_val_range_p, simplify_stmt_using_ranges): Likewise.
8197         (vrp_visit_assignment_or_call): Likewise.  Smuggle class instance
8198         poitner via x_vr_values for calls into gimple folder.
8199         (vrp_initialize_lattice): Make this the vr_values ctor.
8200         (vrp_free_lattice): Make this the vr_values dtor.
8201         (set_vr_value): New function.
8202         (class vrp_prop): Add vr_values data member.  Add various member
8203         functions as well as member functions that delegate to vr_values.
8204         (check_array_ref): Make a member function within vrp_prop class.
8205         (search_for_addr_array, vrp_initialize): Likewise.
8206         (vrp_finalize): Likewise.  Revamp to avoid direct access to
8207         vr_value, values_propagated, etc.
8208         (check_array_bounds): Extract vrp_prop class instance pointer from
8209         walk info structure.  Use it to call member functions.
8210         (check_all_array_refs): Make a member function within vrp_prop class.
8211         Smuggle class instance pointer via walk info structure.
8212         (x_vr_values): New local static.
8213         (vrp_valueize): Use x_vr_values to get class instance.
8214         (vr_valueize_1): Likewise.
8215         (class vrp_folder): Add vr_values data member.  Add various member
8216         functions as well as member functions that delegate to vr_values.
8217         (fold_predicate_in): Make a mber fucntion within vrp_folder class.
8218         (simplify_stmt_for_jump_threading): Extract smuggled vr_values
8219         class instance from vr_values.  Use it to call member functions.
8220         (vrp_dom_walker): Add vr_values data member.
8221         (vrp_dom_walker::after_dom_children): Smuggle vr_values class
8222         instance via x_vr_values.
8223         (identify_jump_threads): Accept vr_values as argument.  Store
8224         it into the walker structure.
8225         (evrp_dom_walker): Add vr_values class data member.  Add various
8226         delegators.
8227         (evrp_dom_walker::try_find_new_range): Use vr_values data
8228         member to access the memory allocator.
8229         (evrp_dom_walker::before_dom_children): Store vr_values class
8230         instance into the vrp_folder class.
8231         (evrp_dom_walker::push_value_range): Rework to avoid direct
8232         access to num_vr_values and vr_value.
8233         (evrp_dom_walker::pop_value_range): Likewise.
8234         (execute_early_vrp): Remove call to vrp_initialize_lattice.
8235         Use vr_values to get to dump_all_value_ranges member function.
8236         Remove call to vrp_free_lattice.  Call vrp_initialize, vrp_finalize,
8237         and simplify_cond_using_ranges_2 via vrp_prop class instance.
8238         Pass vr_values class instance down to identify_jump_threads.
8239         Remove call to vrp_free_lattice.
8240         (debug_all_value_ranges): Remove.
8242         * tree-vrp.c (vrp_prop): Move class to earlier point in the file.
8243         (vrp_folder): Likewise.
8245         * tree-vrp.c (set_value_range): Do not reference vrp_equiv_obstack.
8246         Get it from the existing bitmap instead.
8247         (vrp_intersect_ranges_1): Likewise.
8249 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8251         * gimple-ssa-store-merging.c (struct store_immediate_info): Add
8252         bit_not_p field.
8253         (store_immediate_info::store_immediate_info): Add bitnotp argument,
8254         set bit_not_p to it.
8255         (imm_store_chain_info::coalesce_immediate_stores): Break group
8256         if bit_not_p is different.
8257         (count_multiple_uses, split_group,
8258         imm_store_chain_info::output_merged_store): Handle info->bit_not_p.
8259         (handled_load): Avoid multiple chained BIT_NOT_EXPRs.
8260         (pass_store_merging::process_store): Handle BIT_{AND,IOR,XOR}_EXPR
8261         result inverted using BIT_NOT_EXPR, compute bit_not_p, pass it
8262         to store_immediate_info ctor.
8264 2017-11-09  Jim Wilson  <jimw@sifive.com>
8266         * collect2.c (OBJECT_FORMAT_COFF): Remove EXTENDED_COFF support.
8267         (scan_prog_file): Likewise.
8269 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8271         * bb-reorder.c (max_entry_frequency): Remove.
8272         (find_traces, rotate_loop, mark_bb_visited, connect_better_edge_p,
8273         connect_traces, push_to_next_round_p): Remove prototypes.
8274         (find_traces_1_round): Use counts only.
8275         (push_to_next_round_p): Likewise.
8276         (find_traces): Likewise.
8277         (rotate_loop): Likewise.
8278         (find_traces_1_round): Likewise.
8279         (connect_traces): Likewise.
8280         (edge_order): Likewise.
8282 2017-11-09  Thomas Preud'homme  <thomas.preudhomme@arm.com>
8284         * config/arm/arm.c (output_return_instruction): Add comments to
8285         indicate requirement for cmse_nonsecure_entry return to account
8286         for the size of clearing instruction output here.
8287         (thumb_exit): Likewise.
8288         * config/arm/thumb2.md (thumb2_cmse_entry_return): Fix length for
8289         return in hardfloat mode.
8291 2017-11-09  Segher Boessenkool  <segher@kernel.crashing.org>
8293         * config/rs6000/rs6000.c (machine_function): Add a bool,
8294         "toc_is_wrapped_separately".
8295         (rs6000_option_override_internal): Enable OPTION_MASK_SAVE_TOC_INDIRECT
8296         if it wasn't explicitly set or unset, we are optimizing for speed, and
8297         doing separate shrink-wrapping.
8298         (rs6000_get_separate_components): Enable the TOC component if
8299         saving the TOC register in the prologue.
8300         (rs6000_components_for_bb): Handle the TOC component.
8301         (rs6000_emit_prologue_components): Store the TOC register where needed.
8302         (rs6000_set_handled_components): Mark TOC as handled, if handled.
8303         (rs6000_emit_prologue): Don't save the TOC if that is already done.
8305 2017-11-09  Martin Jambor  <mjambor@suse.cz>
8307         * ipa-param-manipulation.c: New file.
8308         * ipa-param-manipulation.h: Likewise.
8309         * Makefile.in (OBJS): Add ipa-param-manipulation.o.
8310         (PLUGIN_HEADERS): Addded ipa-param-manipulation.h
8311         * ipa-param.h (ipa_parm_op): Moved to ipa-param-manipulation.h.
8312         (ipa_parm_adjustment): Likewise.
8313         (ipa_parm_adjustment_vec): Likewise.
8314         (ipa_get_vector_of_formal_parms): Moved declaration to
8315         ipa-param-manipulation.h.
8316         (ipa_get_vector_of_formal_parm_types): Likewise.
8317         (ipa_modify_formal_parameters): Likewise.
8318         (ipa_modify_call_arguments): Likewise.
8319         (ipa_combine_adjustments): Likewise.
8320         (ipa_dump_param_adjustments): Likewise.
8321         (ipa_modify_expr): Likewise.
8322         (ipa_get_adjustment_candidate): Likewise.
8323         * ipa-prop.c (ipa_get_vector_of_formal_parms): Moved to
8324         ipa-param-manipulation.c.
8325         (ipa_get_vector_of_formal_parm_types): Likewise.
8326         (ipa_modify_formal_parameters): Likewise.
8327         (ipa_modify_call_arguments): Likewise.
8328         (ipa_modify_expr): Likewise.
8329         (get_ssa_base_param): Likewise.
8330         (ipa_get_adjustment_candidate): Likewise.
8331         (index_in_adjustments_multiple_times_p): Likewise.
8332         (ipa_combine_adjustments): Likewise.
8333         (ipa_dump_param_adjustments): Likewise.
8334         * tree-sra.c: Also include ipa-param-manipulation.h
8335         * omp-simd-clone.c: Include ipa-param-manipulation.h instead of
8336         ipa-param.h.
8338 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8339             Alan Hayward  <alan.hayward@arm.com>
8340             David Sherwood  <david.sherwood@arm.com>
8342         * doc/sourcebuild.texi (vect_masked_store): Document.
8344 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8345             Alan Hayward  <alan.hayward@arm.com>
8346             David Sherwood  <david.sherwood@arm.com>
8348         * doc/sourcebuild.texi (vect_align_stack_vars): Document.
8350 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8351             Alan Hayward  <alan.hayward@arm.com>
8352             David Sherwood  <david.sherwood@arm.com>
8354         * doc/sourcebuild.texi (vect_variable_length): Document.
8356 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8357             Alan Hayward  <alan.hayward@arm.com>
8358             David Sherwood  <david.sherwood@arm.com>
8360         * doc/sourcebuild.texi (vect_unaligned_possible): Document.
8362 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8363             Alan Hayward  <alan.hayward@arm.com>
8364             David Sherwood  <david.sherwood@arm.com>
8366         * doc/sourcebuild.texi (vect_element_align_preferred): Document.
8368 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8369             Alan Hayward  <alan.hayward@arm.com>
8370             David Sherwood  <david.sherwood@arm.com>
8372         * doc/sourcebuild.texi (vect_perm_short, vect_perm_byte): Document
8373         previously undocumented selectors.
8374         (vect_perm3_byte, vect_perm3_short, vect_perm3_int): Document.
8376 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8378         * doc/rtl.texi (const_vector): Say that elements can be
8379         const_wide_ints too.
8380         * emit-rtl.h (valid_for_const_vec_duplicate_p): Declare.
8381         * emit-rtl.c (valid_for_const_vec_duplicate_p): New function.
8382         (gen_vec_duplicate): Use it instead of CONSTANT_P.
8383         * optabs.c (expand_vector_broadcast): Likewise.
8385 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8386             Alan Hayward  <alan.hayward@arm.com>
8387             David Sherwood  <david.sherwood@arm.com>
8389         * tree-ssa-loop-ivopts.c (get_address_cost): Try using a
8390         scaled index even if the unscaled address was invalid.
8391         Don't increase the complexity of using a scale in that case.
8393 2017-11-09  Richard Sandiford  <richard.sandiford@linaro.org>
8394             Alan Hayward  <alan.hayward@arm.com>
8395             David Sherwood  <david.sherwood@arm.com>
8397         * doc/rtl.texi: Rewrite the subreg rules so that they partition
8398         the inner register into REGMODE_NATURAL_SIZE bytes rather than
8399         UNITS_PER_WORD bytes.
8400         * emit-rtl.c (validate_subreg): Divide subregs into blocks
8401         based on REGMODE_NATURAL_SIZE of the inner mode.
8402         (gen_lowpart_common): Split the SCALAR_FLOAT_MODE_P and
8403         !SCALAR_FLOAT_MODE_P cases.  Use REGMODE_NATURAL_SIZE for the latter.
8404         * expmed.c (lowpart_bit_field_p): Divide the value up into
8405         chunks of REGMODE_NATURAL_SIZE rather than UNITS_PER_WORD.
8406         * expr.c (store_constructor): Use REGMODE_NATURAL_SIZE to test
8407         whether something is likely to occupy more than one register.
8409 2017-11-09  Jan Hubicka  <hubicka@ucw.cz>
8411         PR ipa/82879
8412         * ipa-inline-transform.c (update_noncloned_frequencies): Use
8413         profile_count::adjust_for_ipa_scaling.
8414         * tree-inline.c (copy_bb, copy_cfg_body): Likewise.
8415         * profile-count.c (profile_count::adjust_for_ipa_scaling): New member
8416         function.
8417         * profile-count.h (profile_count::adjust_for_ipa_scaling): Declare.
8419 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8421         * gimple-ssa-store-merging.c (count_multiple_uses): New function.
8422         (split_group): Add total_orig and total_new arguments, estimate the
8423         number of statements related to the store group without store merging
8424         and with store merging.
8425         (imm_store_chain_info::output_merged_store): Adjust split_group
8426         callers, punt if estimated number of statements with store merging
8427         is not smaller than estimated number of statements without it.
8428         Formatting fix.
8429         (handled_load): Remove has_single_use checks.
8430         (pass_store_merging::process_store): Likewise.
8432 2017-11-09  Richard Biener  <rguenther@suse.de>
8434         PR tree-optimization/82902
8435         * tree-ssa-phiprop.c (propagate_with_phi): Test proper type.
8437 2017-11-09  Martin Liska  <mliska@suse.cz>
8439         PR target/82863
8440         * emit-rtl.c (init_emit_regs): Initialize split_branch_probability to
8441         uninitialized.
8443 2017-11-09  Martin Liska  <mliska@suse.cz>
8445         PR tree-optimization/82669
8446         * sbitmap.h (bmp_iter_set_init): Remove non needed check.
8448 2017-11-09  Martin Liska  <mliska@suse.cz>
8450         PR gcov-profile/48463
8451         * coverage.c (coverage_begin_function): Output also end locus
8452         of a function and information whether the function is
8453         artificial.
8454         * gcov-dump.c (tag_function): Parse and print the information.
8455         * gcov.c (INCLUDE_MAP): Add include.
8456         (INCLUDE_SET): Likewise.
8457         (struct line_info): Move earlier in the source file because
8458         of vector<line_info> in function_info structure.
8459         (line_info::line_info): Likewise.
8460         (line_info::has_block): Likewise.
8461         (struct source_info): Add new member index.
8462         (source_info::get_functions_at_location): New function.
8463         (function_info::group_line_p): New function.
8464         (output_intermediate_line): New function.
8465         (output_intermediate_file): Use the mentioned function.
8466         (struct function_start): New.
8467         (struct function_start_pair_hash): Likewise.
8468         (process_file): Add code that identifies group functions.
8469         Assign lines either to global or function scope.
8470         (generate_results): Skip artificial functions.
8471         (find_source): Assign index for each source file.
8472         (read_graph_file): Read new flag artificial and end_line.
8473         (add_line_counts): Assign it either to global of function scope.
8474         (accumulate_line_counts): Isolate core of the function to
8475         accumulate_line_info and call it for both function and global
8476         scope lines.
8477         (accumulate_line_info): New function.
8478         (output_line_beginning): Fix GNU coding style.
8479         (print_source_line): New function.
8480         (output_line_details): Likewise.
8481         (output_function_details): Likewise.
8482         (output_lines): Iterate both source (global) scope and function
8483         scope.
8484         (struct function_line_start_cmp): New class.
8485         * doc/gcov.texi: Reflect changes in documentation.
8487 2017-11-09  Jakub Jelinek  <jakub@redhat.com>
8489         PR debug/82837
8490         * dwarf2out.c (const_ok_for_output_1): Reject NEG in addition to NOT.
8491         (mem_loc_descriptor): Handle (const (neg (...))) as (neg (const (...)))
8492         and similarly for not instead of neg.
8494 2017-11-08  Andi Kleen  <ak@linux.intel.com>
8496         * config/i386/i386.opt: Add -mforce-indirect-call.
8497         * config/i386/predicates.md: Check for flag_force_indirect_call.
8498         * doc/invoke.texi: Document -mforce-indirect-call
8500 2017-11-08  Kito Cheng  <kito.cheng@gmail.com>
8502         * config/riscv/riscv-protos.h (riscv_slow_unaligned_access_p):
8503         New extern.
8504         (MOVE_RATIO): Use riscv_slow_unaligned_access_p.
8505         config/riscv/riscv.c (predict.h): New include.
8506         (riscv_slow_unaligned_access_p): No longer static.
8507         (riscv_block_move_straight): Add require.
8508         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
8510 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8512         PR target/82855
8513         * config/i386/sse.md (<avx512>_eq<mode>3<mask_scalar_merge_name>,
8514         <avx512>_eq<mode>3<mask_scalar_merge_name>_1): Use
8515         nonimmediate_operand predicate for operand 1 instead of
8516         register_operand.
8518 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8520         * config/aarch64/aarch64-simd.md (store_pair_lanes<mode>):
8521         New pattern.
8522         * config/aarch64/constraints.md (Uml): New constraint.
8523         * config/aarch64/predicates.md (aarch64_mem_pair_lanes_operand): New
8524         predicate.
8526 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8528         * simplify-rtx.c (simplify_ternary_operation): Simplify vec_merge
8529         of two vec_duplicates into a vec_concat.
8531 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8533         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8534         Simplify vec_merge of vec_duplicate and vec_concat.
8535         * config/aarch64/constraints.md (Utq): New constraint.
8536         * config/aarch64/aarch64-simd.md (load_pair_lanes<mode>): New
8537         define_insn.
8539 2017-11-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
8541         * simplify-rtx.c (simplify_ternary_operation, VEC_MERGE):
8542         Simplify vec_merge of vec_duplicate and const_vector.
8543         * config/aarch64/predicates.md (aarch64_simd_or_scalar_imm_zero):
8544         New predicate.
8545         * config/aarch64/aarch64-simd.md (*aarch64_combinez<mode>): Use VDC
8546         mode iterator.  Update predicate on operand 1 to
8547         handle non-const_vec constants.  Delete constraints.
8548         (*aarch64_combinez_be<mode>): Likewise for operand 2.
8550 2017-11-08  Jakub Jelinek  <jakub@redhat.com>
8552         PR tree-optimization/78821
8553         * gimple-ssa-store-merging.c (struct store_operand_info): Add bit_not_p
8554         data member.
8555         (store_operand_info::store_operand_info): Initialize it to false.
8556         (pass_store_merging::terminate_all_aliasing_chains): Rewritten to use
8557         ref_maybe_used_by_stmt_p and stmt_may_clobber_ref_p on lhs of each
8558         store in the group, and if chain_info is non-NULL, to ignore altogether
8559         that chain.
8560         (compatible_load_p): Fail if bit_not_p does not match.
8561         (imm_store_chain_info::output_merged_store): Handle bit_not_p loads.
8562         (handled_load): Fill in bit_not_p.  Handle BIT_NOT_EXPR.
8563         (pass_store_merging::process_store): Adjust
8564         terminate_all_aliasing_chains calls to pass NULL in all current spots,
8565         call terminate_all_aliasing_chains newly when adding a store into
8566         a chain with non-NULL chain_info.
8568 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8570         * config/aarch64/aarch64.c (aarch64_can_eliminate): Simplify logic.
8572 2017-11-08  Wilco Dijkstra  <wdijkstr@arm.com>
8574         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
8575         Remove.
8576         (aarch64_layout_frame): Initialise emit_frame_chain.
8577         (aarch64_can_eliminate): Remove omit leaf frame pointer code.
8578         (TARGET_FRAME_POINTER_REQUIRED): Remove define.
8580 2017-11-08  Martin Liska  <mliska@suse.cz>
8582         * gimplify.c (expand_FALLTHROUGH_r): Simplify usage
8583         of gimple_call_internal_p.
8585 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8587         * config/mips/mips.h (ASM_OUTPUT_LABELREF): Wrap in "do {} while (0)".
8589 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8591         * config/mips/mips.h (ASM_OUTPUT_CASE_END): Remove semicolon after
8592         "do {} while (0)".
8594 2017-11-08  Martin Liska  <mliska@suse.cz>
8596         PR sanitizer/82792
8597         * gimplify.c (expand_FALLTHROUGH_r): Skip IFN_ASAN_MARK.
8599 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8601         * gimple-pretty-print.c (dump_profile): Return "" instead of NULL.
8603 2017-11-07  Jakub Jelinek  <jakub@redhat.com>
8605         PR target/82855
8606         * config/i386/i386.md (SWI1248_AVX512BWDQ2_64): New mode iterator.
8607         (*cmp<mode>_ccz_1): New insn with $k alternative.
8609         PR target/82855
8610         * config/i386/i386.c (ix86_swap_binary_operands_p): Treat
8611         RTX_COMM_COMPARE as commutative as well.
8612         (ix86_binary_operator_ok): Formatting fix.
8613         * config/i386/sse.md (*mul<mode>3<mask_name><round_name>,
8614         *<code><mode>3<mask_name><round_saeonly_name>,
8615         *<code><mode>3<mask_name>, *<code>tf3, *mul<mode>3<mask_name>,
8616         *<s>mul<mode>3_highpart<mask_name>,
8617         *vec_widen_umult_even_v16si<mask_name>,
8618         *vec_widen_umult_even_v8si<mask_name>,
8619         *vec_widen_umult_even_v4si<mask_name>,
8620         *vec_widen_smult_even_v16si<mask_name>,
8621         *vec_widen_smult_even_v8si<mask_name>, *sse4_1_mulv2siv2di3<mask_name>,
8622         *avx2_pmaddwd, *sse2_pmaddwd, *<sse4_1_avx2>_mul<mode>3<mask_name>,
8623         *avx2_<code><mode>3, *avx512f_<code><mode>3<mask_name>,
8624         *sse4_1_<code><mode>3<mask_name>, *<code>v8hi3,
8625         *sse4_1_<code><mode>3<mask_name>, *<code>v16qi3, *avx2_eq<mode>3,
8626         <avx512>_eq<mode>3<mask_scalar_merge_name>_1, *sse4_1_eqv2di3,
8627         *sse2_eq<mode>3, <mask_codefor><code><mode>3<mask_name>,
8628         *<code><mode>3, *<sse2_avx2>_uavg<mode>3<mask_name>,
8629         *<ssse3_avx2>_pmulhrsw<mode>3<mask_name>, *ssse3_pmulhrswv4hi3): Use
8630         !(MEM_P (operands[1]) && MEM_P (operands[2])) condition instead of
8631         ix86_binary_operator_ok.  Formatting fixes.
8632         (*<plusminus_insn><mode>3<mask_name><round_name>,
8633         *<plusminus_insn><mode>3, *<plusminus_insn><mode>3_m): Formatting
8634         fixes.
8636 2017-11-07  Segher Boessenkool  <segher@kernel.crashing.org>
8638         * config/rs6000/rs6000.md (GPR2): New mode_iterator.
8639         ("cstore<mode>4"): Don't always expand with rs6000_emit_int_cmove for
8640         eq and ne if TARGET_ISEL.
8641         (cmp): New code_iterator.
8642         (UNS, UNSU_, UNSIK): New code_attrs.
8643         (<code><GPR:mode><GPR2:mode>2_isel): New define_insn_and_split.
8644         ("eq<mode>3"): New define_expand, rename the define_insn_and_split
8645         to...
8646         ("eq<mode>3"): ... this.
8647         ("ne<mode>3"): New define_expand, rename the define_insn_and_split
8648         to...
8649         ("ne<mode>3"): ... this.
8651 2017-11-07  Julia Koval  <julia.koval@intel.com>
8653         PR target/82812
8654         * common/config/i386/i386-common.c
8655         (OPTION_MASK_ISA_GENERAL_REGS_ONLY_UNSET): Remove MPX from flag.
8656         (ix86_handle_option): Move MPX to isa_flags2 and GFNI to isa_flags.
8657         * config/i386/i386-c.c (ix86_target_macros_internal): Ditto.
8658         * config/i386/i386.opt: Ditto.
8659         * config/i386/i386.c (ix86_target_string): Ditto.
8660         (ix86_option_override_internal): Ditto.
8661         (ix86_init_mpx_builtins): Move MPX to args2.
8662         (ix86_expand_builtin): Special handling for OPTION_MASK_ISA_GFNI.
8663         * config/i386/i386-builtin.def (__builtin_ia32_vgf2p8affineinvqb_v64qi,
8664         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
8665         __builtin_ia32_vgf2p8affineinvqb_v32qi,
8666         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
8667         __builtin_ia32_vgf2p8affineinvqb_v16qi,
8668         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): Move to ARGS array.
8670 2017-11-07  Uros Bizjak  <ubizjak@gmail.com>
8672         PR target/80425
8673         * config/i386.i386.md (*zero_extendsidi2): Change (?r,*Yj), (?*Yi,r)
8674         and (*x,m) to ($r,Yj), ($Yi,r) and ($x,m).
8675         (zero-extendsidi peephole2): Remove peephole.
8677 2017-11-07  Eric Botcazou  <ebotcazou@adacore.com>
8679         PR c/53037
8680         * stor-layout.c: Include attribs.h.
8681         (handle_warn_if_not_align): Replace test on TYPE_USER_ALIGN with
8682         explicit lookup of "aligned" attribute.
8684 2017-11-07  Andrew Waterman  <andrew@sifive.com>
8686         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): New prototype.
8687         (riscv_expand_block_move): Likewise.
8688         * config/riscv/riscv.h (MOVE_RATIO): Tune cost to movmemsi
8689         implementation.
8690         (RISCV_MAX_MOVE_BYTES_PER_LOOP_ITER): New define.
8691         (RISCV_MAX_MOVE_BYTES_STRAIGHT): New define.
8692         * config/riscv/riscv.c (riscv_block_move_straight): New function.
8693         (riscv_adjust_block_mem): Likewise.
8694         (riscv_block_move_loop): Likewise.
8695         (riscv_expand_block_move): Likewise.
8696         * config/riscv/riscv.md (movmemsi): New pattern.
8698 2017-11-07  Michael Clark  <michaeljclark@mac.com>
8700         * config/riscv/linux.h (MUSL_ABI_SUFFIX): New define.
8701         (MUSL_DYNAMIC_LINKER): Likewise.
8703 2017-11-07  Richard Sandiford  <richard.sandiford@linaro.org>
8705         * config/aarch64/aarch64.md (ashl<mode>3, ashr<mode>3, lshr<mode>3)
8706         (rotr<mode>3, rotl<mode>3): Use aarch64_reg_or_imm instead of
8707         nonmmory_operand.
8709 2017-11-07  Richard Biener  <rguenther@suse.de>
8711         * match.pd: Fix build.
8713 2017-11-07  Wilco Dijkstra  <wdijkstr@arm.com>
8714             Jackson Woodruff  <jackson.woodruff@arm.com>
8716         PR tree-optimization/71026
8717         * match.pd: Canonicalize negate in division.
8719 2017-11-07  Sudakshina Das  <sudi.das@arm.com>
8721         PR middle-end/80131
8722         * match.pd: Simplify 1 << (C - x) where C = precision (x) - 1.
8724 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8726         * match.pd ((a&~b)|(a^b),(a&~b)^~a,(a|b)&~(a^b),a|~(a^b),
8727         (a|b)|(a&^b),(a&b)|~(a^b),~(~a&b),~X^Y): New transformations.
8729 2017-11-07  Marc Glisse  <marc.glisse@inria.fr>
8731         * fold-const.c (negate_expr_p) [PLUS_EXPR, MINUS_EXPR]: Handle
8732         non-scalar integral types.
8733         * match.pd (negate_expr_p): Handle MINUS_EXPR.
8734         (-(A-B), -(~A)): New transformations.
8736 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8738         * config/powerpcspe/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove
8739         semicolon after "do {} while (0)".
8740         * config/powerpcspe/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8741         * config/powerpcspe/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8742         * config/powerpcspe/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8743         * config/powerpcspe/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8744         * config/powerpcspe/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8746 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8748         * config/rs6000/aix43.h (SUBTARGET_OVERRIDE_OPTIONS): Remove semicolon
8749         after "do {} while (0)".
8750         * config/rs6000/aix51.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8751         * config/rs6000/aix52.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8752         * config/rs6000/aix53.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8753         * config/rs6000/aix61.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8754         * config/rs6000/aix71.h (SUBTARGET_OVERRIDE_OPTIONS): Same.
8756 2017-11-07  Tom de Vries  <tom@codesourcery.com>
8758         PR other/82784
8759         * config/arm/arm.c (HANDLE_NARROW_SHIFT_ARITH): Remove semicolon after
8760         "while {} do (0)".
8761         (arm_rtx_costs_internal): Add missing semicolon after
8762         HANDLE_NARROW_SHIFT_ARITH call.
8764 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
8766         * config/rs6000/rs6000.c (rs6000_option_override_internal): Don't
8767         disable isel if it was not set explicitly.
8769 2017-11-06  James Bowman  <james.bowman@ftdichip.com>
8771         * gcc/dwarf2out.c (modified_type_die): Retain ADDR_SPACE qualifiers.
8772         (add_type_attribute) likewise.
8774 2017-11-06  H.J. Lu  <hongjiu.lu@intel.com>
8776         * config/i386/i386.c (ix86_can_use_return_insn_p): Use reference
8777         of struct ix86_frame.
8778         (ix86_initial_elimination_offset): Likewise.
8779         (ix86_expand_split_stack_prologue): Likewise.
8781 2017-11-06  Marc Glisse  <marc.glisse@inria.fr>
8783         * tree-vrp.h (enum value_range_type): Update stale comment.
8785 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8786             Alan Hayward  <alan.hayward@arm.com>
8787             David Sherwood  <david.sherwood@arm.com>
8789         * config/aarch64/aarch64-protos.h (aarch64_expand_vec_perm)
8790         (aarch64_expand_vec_perm_const): Take the number of units too.
8791         * config/aarch64/aarch64.c (aarch64_expand_vec_perm)
8792         (aarch64_expand_vec_perm_const): Likewise.
8793         * config/aarch64/aarch64-simd.md (vec_perm_const<mode>)
8794         (vec_perm<mode>): Update accordingly.
8796 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8797             Alan Hayward  <alan.hayward@arm.com>
8798             David Sherwood  <david.sherwood@arm.com>
8800         * config/aarch64/aarch64-protos.h (aarch64_simd_vect_par_cnst_half):
8801         Take the number of units too.
8802         * config/aarch64/aarch64.c (aarch64_simd_vect_par_cnst_half): Likewise.
8803         (aarch64_simd_check_vect_par_cnst_half): Update call accordingly,
8804         but check for a vector mode before rather than after the call.
8805         * config/aarch64/aarch64-simd.md (aarch64_split_simd_mov<mode>)
8806         (move_hi_quad_<mode>, vec_unpack<su>_hi_<mode>)
8807         (vec_unpack<su>_lo_<mode, vec_widen_<su>mult_lo_<mode>)
8808         (vec_widen_<su>mult_hi_<mode>, vec_unpacks_lo_<mode>)
8809         (vec_unpacks_hi_<mode>, aarch64_saddl2<mode>, aarch64_uaddl2<mode>)
8810         (aarch64_ssubl2<mode>, aarch64_usubl2<mode>, widen_ssum<mode>3)
8811         (widen_usum<mode>3, aarch64_saddw2<mode>, aarch64_uaddw2<mode>)
8812         (aarch64_ssubw2<mode>, aarch64_usubw2<mode>, aarch64_sqdmlal2<mode>)
8813         (aarch64_sqdmlsl2<mode>, aarch64_sqdmlal2_lane<mode>)
8814         (aarch64_sqdmlal2_laneq<mode>, aarch64_sqdmlsl2_lane<mode>)
8815         (aarch64_sqdmlsl2_laneq<mode>, aarch64_sqdmlal2_n<mode>)
8816         (aarch64_sqdmlsl2_n<mode>, aarch64_sqdmull2<mode>)
8817         (aarch64_sqdmull2_lane<mode>, aarch64_sqdmull2_laneq<mode>)
8818         (aarch64_sqdmull2_n<mode>): Update accordingly.
8820 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8821             Alan Hayward  <alan.hayward@arm.com>
8822             David Sherwood  <david.sherwood@arm.com>
8824         * config/aarch64/aarch64-protos.h (aarch64_reverse_mask): Take
8825         the number of units too.
8826         * config/aarch64/aarch64.c (aarch64_reverse_mask): Likewise.
8827         * config/aarch64/aarch64-simd.md (vec_load_lanesoi<mode>)
8828         (vec_store_lanesoi<mode>, vec_load_lanesci<mode>)
8829         (vec_store_lanesci<mode>, vec_load_lanesxi<mode>)
8830         (vec_store_lanesxi<mode>): Update accordingly.
8832 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8833             Alan Hayward  <alan.hayward@arm.com>
8834             David Sherwood  <david.sherwood@arm.com>
8836         * config/aarch64/aarch64-protos.h (aarch64_endian_lane_rtx): Declare.
8837         * config/aarch64/aarch64.c (aarch64_endian_lane_rtx): New function.
8838         * config/aarch64/aarch64.h (ENDIAN_LANE_N): Take the number
8839         of units rather than the mode.
8840         * config/aarch64/iterators.md (nunits): New mode attribute.
8841         * config/aarch64/aarch64-builtins.c (aarch64_simd_expand_args):
8842         Use aarch64_endian_lane_rtx instead of GEN_INT (ENDIAN_LANE_N ...).
8843         * config/aarch64/aarch64-simd.md (aarch64_dup_lane<mode>)
8844         (aarch64_dup_lane_<vswap_width_name><mode>, *aarch64_mul3_elt<mode>)
8845         (*aarch64_mul3_elt_<vswap_width_name><mode>): Likewise.
8846         (*aarch64_mul3_elt_to_64v2df, *aarch64_mla_elt<mode>): Likewise.
8847         (*aarch64_mla_elt_<vswap_width_name><mode>, *aarch64_mls_elt<mode>)
8848         (*aarch64_mls_elt_<vswap_width_name><mode>, *aarch64_fma4_elt<mode>)
8849         (*aarch64_fma4_elt_<vswap_width_name><mode>):: Likewise.
8850         (*aarch64_fma4_elt_to_64v2df, *aarch64_fnma4_elt<mode>): Likewise.
8851         (*aarch64_fnma4_elt_<vswap_width_name><mode>): Likewise.
8852         (*aarch64_fnma4_elt_to_64v2df, reduc_plus_scal_<mode>): Likewise.
8853         (reduc_plus_scal_v4sf, reduc_<maxmin_uns>_scal_<mode>): Likewise.
8854         (reduc_<maxmin_uns>_scal_<mode>): Likewise.
8855         (*aarch64_get_lane_extend<GPI:mode><VDQQH:mode>): Likewise.
8856         (*aarch64_get_lane_zero_extendsi<mode>): Likewise.
8857         (aarch64_get_lane<mode>, *aarch64_mulx_elt_<vswap_width_name><mode>)
8858         (*aarch64_mulx_elt<mode>, *aarch64_vgetfmulx<mode>): Likewise.
8859         (aarch64_sq<r>dmulh_lane<mode>, aarch64_sq<r>dmulh_laneq<mode>)
8860         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_lane<mode>): Likewise.
8861         (aarch64_sqrdml<SQRDMLH_AS:rdma_as>h_laneq<mode>): Likewise.
8862         (aarch64_sqdml<SBINQOPS:as>l_lane<mode>): Likewise.
8863         (aarch64_sqdml<SBINQOPS:as>l_laneq<mode>): Likewise.
8864         (aarch64_sqdml<SBINQOPS:as>l2_lane<mode>_internal): Likewise.
8865         (aarch64_sqdml<SBINQOPS:as>l2_laneq<mode>_internal): Likewise.
8866         (aarch64_sqdmull_lane<mode>, aarch64_sqdmull_laneq<mode>): Likewise.
8867         (aarch64_sqdmull2_lane<mode>_internal): Likewise.
8868         (aarch64_sqdmull2_laneq<mode>_internal): Likewise.
8869         (aarch64_vec_load_lanesoi_lane<mode>): Likewise.
8870         (aarch64_vec_store_lanesoi_lane<mode>): Likewise.
8871         (aarch64_vec_load_lanesci_lane<mode>): Likewise.
8872         (aarch64_vec_store_lanesci_lane<mode>): Likewise.
8873         (aarch64_vec_load_lanesxi_lane<mode>): Likewise.
8874         (aarch64_vec_store_lanesxi_lane<mode>): Likewise.
8875         (aarch64_simd_vec_set<mode>): Update use of ENDIAN_LANE_N.
8876         (aarch64_simd_vec_setv2di): Likewise.
8878 2017-11-06  Carl Love  <cel@us.ibm.com>
8880         * config/rs6000/rs6000-c.c (P8V_BUILTIN_VEC_REVB): Add power 8
8881         definitions.
8882         (P9V_BUILTIN_VEC_REVB): Remove the power 9 instance definitions.
8883         * config/rs6000/altivec.h (vec_revb): Change the #define from power 9
8884         to power 8.
8885         * config/rs6000/r6000-protos.h (swap_endian_selector_for_mode): Add new
8886         extern declaration.
8887         * config/rs6000/rs6000.c (swap_endian_selector_for_mode): Add function.
8888         * config/rs6000/rs6000-builtin.def (BU_P8V_VSX_1, BU_P8V_OVERLOAD_1):
8889         Add power 8 macro expansions.
8890         (BU_P9V_OVERLOAD_1): Remove power 9 overload expansion.
8891         * config/rs6000/vsx.md (revb_<mode>): Add define_expand to generate
8892         power 8 instructions.  (VSX_XXBR): Add iterator.
8894 2017-11-06  Wilco Dijkstra  <wdijkstr@arm.com>
8896         * config/arm/arm.md (predicable_short_it): Change default to "no",
8897         improve documentation, remove uses that are identical to the default.
8898         (enabled_for_depr_it): Rename to enabled_for_short_it.
8899         * gcc/config/arm/arm-fixed.md (predicable_short_it):
8900         Remove default uses.
8901         * gcc/config/arm/ldmstm.md (predicable_short_it): Likewise.
8902         * gcc/config/arm/sync.md (predicable_short_it): Likewise.
8903         * gcc/config/arm/thumb2.md (predicable_short_it): Likewise.
8904         * gcc/config/arm/vfp.md (predicable_short_it): Likewise.
8906 2017-11-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
8908         PR target/82748
8909         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1): Delete
8910         float128 helper macros, which are no longer used after deleting
8911         the old 'q' built-in functions, and moving the round to odd
8912         built-in functions to being special built-in functions.
8913         (BU_FLOAT128_2): Likewise.
8914         (BU_FLOAT128_1_HW): Likewise.
8915         (BU_FLOAT128_2_HW): Likewise.
8916         (BU_FLOAT128_3_HW): Likewise.
8917         (FABSQ): Delete old 'q' built-in functions.
8918         (COPYSIGNQ): Likewise.
8919         (SQRTF128_ODD): Move round to odd built-in functions to be
8920         special built-in functions, so that we can handle
8921         -mabi=ieeelongdouble.
8922         (TRUNCF128_ODD): Likewise.
8923         (ADDF128_ODD): Likewise.
8924         (SUBF128_ODD): Likewise.
8925         (MULF128_ODD): Likewise.
8926         (DIVF128_ODD): Likewise.
8927         (FMAF128_ODD): Likewise.
8928         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): Map old 'q'
8929         built-in names to 'f128'.
8930         * config/rs6000/rs6000.c (rs6000_fold_builtin): Remove folding the
8931         old 'q' built-in functions, as the machine independent code for
8932         'f128' built-in functions handles this.
8933         (rs6000_expand_builtin): Add expansion for float128 round to odd
8934         functions, keying off on -mabi=ieeelongdouble of whether to use
8935         the KFmode or TFmode variant.
8936         (rs6000_init_builtins): Initialize the _Float128 round to odd
8937         built-in functions.
8938         * doc/extend.texi (PowerPC Built-in Functions): Document the old
8939         _Float128 'q' built-in functions are now mapped into the new
8940         'f128' built-in functions.
8942 2017-11-06  David Edelsohn  <dje.gcc@gmail.com>
8944         * collect2.c (add_lto_object): Compile for OBJECT_COFF.
8945         (scan_prog_file): Don't skip PASS_LTOINFO. Scan for LTO objects.
8947 2017-11-06  David Malcolm  <dmalcolm@redhat.com>
8949         PR jit/82826
8950         * ipa-fnsummary.c (ipa_fnsummary_c_finalize): New function.
8951         * ipa-fnsummary.h (ipa_fnsummary_c_finalize): New decl.
8952         * toplev.c: Include "ipa-fnsummary.h".
8953         (toplev::finalize): Call ipa_fnsummary_c_finalize.
8955 2017-11-06  Jakub Jelinek  <jakub@redhat.com>
8957         PR tree-optimization/82838
8958         * gimple-ssa-store-merging.c
8959         (imm_store_chain_info::output_merged_store): Call force_gimple_operand_1
8960         on a separate gimple_seq which is then appended to seq.
8962 2017-11-06  Jeff Law  <law@redhat.com>
8964         PR target/82788
8965         * config/i386/i386.c (PROBE_INTERVAL): Remove.
8966         (get_probe_interval): New functions.
8967         (ix86_adjust_stack_and_probe_stack_clash): Use get_probe_interval.
8968         (ix86_adjust_stack_and_probe): Likewise.
8969         (output_adjust_stack_and_probe): Likewise.
8970         (ix86_emit_probe_stack_range): Likewise.
8971         (ix86_expand_prologue): Likewise.
8973 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
8975         PR tree-optimization/82816
8976         * tree-ssa-math-opts.c (convert_mult_to_widen): Return false
8977         if the modes of the two types are the same.
8978         (convert_plusminus_to_widen): Likewise.
8980 2017-11-06  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
8982         * config/rs6000/altivec.md (*p9_vadu<mode>3) Rename to
8983         p9_vadu<mode>3.
8984         (usadv16qi): New define_expand.
8985         (usadv8hi): New define_expand.
8987 2017-11-06  Jan Hubicka  <hubicka@ucw.cz>
8989         PR bootstrap/82832
8990         * ipa-inline-transform.c (update_noncloned_frequencies): Always
8991         scale.
8992         (inline_transform): Likewise.
8993         * predict.c (counts_to_freqs): Remove useless conditional.
8994         * profile-count.h (profile_count::apply_scale): Move sanity check.
8995         * tree-inline.c (copy_bb): Always scale.
8996         (copy_cfg_body): Likewise.
8998 2017-11-06  Christophe Lyon  <christophe.lyon@linaro.org>
9000         PR target/67591
9001         * config/arm/arm.md (*sub_shiftsi): Add predicable_short_it
9002         attribute.
9003         (*cmp_ite0): Add enabled_for_depr_it attribute.
9004         (*cmp_ite1): Likewise.
9006 2017-11-06  Segher Boessenkool  <segher@kernel.crashing.org>
9008         * config/rs6000/rs6000.c (rs6000_insn_cost): Handle TYPE_MFCR and
9009         TYPE_MFCRF.
9011 2017-11-06  Richard Sandiford  <richard.sandiford@linaro.org>
9013         * tree-vrp.c (vrp_int_const_binop): Return true on success and
9014         return the value by pointer.
9015         (extract_range_from_multiplicative_op_1): Update accordingly.
9016         Return as soon as an operation fails.
9018 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9020         PR other/82784
9021         * asan.c (DEF_SANITIZER_BUILTIN_1): Factor out of ...
9022         (DEF_SANITIZER_BUILTIN): ... here.
9023         (initialize_sanitizer_builtins): Use DEF_SANITIZER_BUILTIN_1 instead of
9024         DEF_SANITIZER_BUILTIN in if stmt.  Add missing semicolon.
9026 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9028         PR other/82784
9029         * config/elfos.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon after
9030         macro body.
9031         (ASM_OUTPUT_CASE_LABEL): Add semicolon after
9032         ASM_OUTPUT_BEFORE_CASE_LABEL call.
9033         * config/arc/arc.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove semicolon
9034         after macro body.
9035         * config/m68k/m68kelf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9036         * config/mips/mips.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9037         * config/v850/v850.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Same.
9039 2017-11-05  Tom de Vries  <tom@codesourcery.com>
9041         PR other/82784
9042         * graphite-scop-detection.c (DEBUG_PRINT): Remove semicolon after
9043         "do {} while (0)".
9045 2017-11-04  Michael Clark  <michaeljclark@mac.com>
9047         * config/riscv/riscv.c (riscv_print_operand): Add a 'i' format.
9048         config/riscv/riscv.md (addsi3): Use 'i' for immediates.
9049         (adddi3): Likewise.
9050         (*addsi3_extended): Likewise.
9051         (*addsi3_extended2): Likewise.
9052         (<optab>si3): Likewise.
9053         (<optab>di3): Likewise.
9054         (<optab><mode>3): Likewise.
9055         (<*optabe>si3_internal): Likewise.
9056         (zero_extendqi<SUPERQI:mode>2): Likewise.
9057         (*add<mode>hi3): Likewise.
9058         (*xor<mode>hi3): Likewise.
9059         (<optab>di3): Likewise.
9060         (*<optab>si3_extend): Likewise.
9061         (*sge<u>_<X:mode><GPR:mode>): Likewise.
9062         (*slt<u>_<X:mode><GPR:mode>): Likewise.
9063         (*sle<u>_<X:mode><GPR:mode>): Likewise.
9065 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9067         * config/riscv/riscv.c (riscv_option_override): Conditionally set
9068         TARGET_STRICT_ALIGN based upon -mtune argument.
9070 2017-11-04  Andrew Waterman  <andrew@sifive.com>
9072         * config/riscv/riscv.h (SLOW_BYTE_ACCESS): Change to 1.
9074 2017-11-04  Daniel Santos  <daniel.santos@pobox.com>
9076         * config/i386/i386.c (choose_basereg): Use optional scratch
9077         register and add assertion.
9078         (x86_emit_outlined_ms2sysv_save): Use scratch register when
9079         needed, and don't allocate stack.
9080         (ix86_expand_prologue): Rearrange where SSE saves/stub call is
9081         emitted, correct wrong allocation with -mcall-ms2sysv-xlogues.
9082         (ix86_emit_outlined_ms2sysv_restore): Fix non-immediate offsets.
9084 2017-11-03  Jeff Law  <law@redhat.com>
9086         * config/i386/i386.c (ix86_emit_restore_reg_using_pop): Prototype.
9087         (ix86_adjust_stack_and_probe_stack_clash): Use a push/pop sequence
9088         to probe at the start of a noreturn function.
9090 2017-11-03  Jakub Jelinek  <jakub@redhat.com>
9092         PR tree-optimization/78821
9093         * gimple-ssa-store-merging.c: Update the file comment.
9094         (MAX_STORE_ALIAS_CHECKS): Define.
9095         (struct store_operand_info): New type.
9096         (store_operand_info::store_operand_info): New constructor.
9097         (struct store_immediate_info): Add rhs_code and ops data members.
9098         (store_immediate_info::store_immediate_info): Add rhscode, op0r
9099         and op1r arguments to the ctor, initialize corresponding data members.
9100         (struct merged_store_group): Add load_align_base and load_align
9101         data members.
9102         (merged_store_group::merged_store_group): Initialize them.
9103         (merged_store_group::do_merge): Update them.
9104         (merged_store_group::apply_stores): Pick the constant for
9105         encode_tree_to_bitpos from one of the two operands, or skip
9106         encode_tree_to_bitpos if neither operand is a constant.
9107         (class pass_store_merging): Add process_store method decl.  Remove
9108         bool argument from terminate_all_aliasing_chains method decl.
9109         (pass_store_merging::terminate_all_aliasing_chains): Remove
9110         var_offset_p argument and corresponding handling.
9111         (stmts_may_clobber_ref_p): New function.
9112         (compatible_load_p): New function.
9113         (imm_store_chain_info::coalesce_immediate_stores): Terminate group
9114         if there is overlap and rhs_code is not INTEGER_CST.  For
9115         non-overlapping stores terminate group if rhs is not mergeable.
9116         (get_alias_type_for_stmts): Change first argument from
9117         auto_vec<gimple *> & to vec<gimple *> &.  Add IS_LOAD, CLIQUEP and
9118         BASEP arguments.  If IS_LOAD is true, look at rhs1 of the stmts
9119         instead of lhs.  Compute *CLIQUEP and *BASEP in addition to the
9120         alias type.
9121         (get_location_for_stmts): Change first argument from
9122         auto_vec<gimple *> & to vec<gimple *> &.
9123         (struct split_store): Remove orig_stmts data member, add orig_stores.
9124         (split_store::split_store): Create orig_stores rather than orig_stmts.
9125         (find_constituent_stmts): Renamed to ...
9126         (find_constituent_stores): ... this.  Change second argument from
9127         vec<gimple *> * to vec<store_immediate_info *> *, push pointers
9128         to info structures rather than the statements.
9129         (split_group): Rename ALLOW_UNALIGNED argument to
9130         ALLOW_UNALIGNED_STORE, add ALLOW_UNALIGNED_LOAD argument and handle
9131         it.  Adjust find_constituent_stores caller.
9132         (imm_store_chain_info::output_merged_store): Handle rhs_code other
9133         than INTEGER_CST, adjust split_group, get_alias_type_for_stmts and
9134         get_location_for_stmts callers.  Set MR_DEPENDENCE_CLIQUE and
9135         MR_DEPENDENCE_BASE on the MEM_REFs if they are the same in all stores.
9136         (mem_valid_for_store_merging): New function.
9137         (handled_load): New function.
9138         (pass_store_merging::process_store): New method.
9139         (pass_store_merging::execute): Use process_store method.  Adjust
9140         terminate_all_aliasing_chains caller.
9142 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9144         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
9145         Return true for more constants, symbols and label references.
9146         (aarch64_valid_floating_const): Remove unused function.
9148 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9150         PR target/82786
9151         * config/aarch64/aarch64.c (aarch64_layout_frame):
9152         Undo forcing of LR at bottom of frame.
9154 2017-11-03  Jeff Law  <law@redhat.com>
9156         PR target/82823
9157         * config/i386/i386.c (ix86_expand_prologue): Tighten assert
9158         for int_registers_saved.
9160         * cfganal.c (single_pred_edge_ignoring_loop_edges): New function
9161         extracted from tree-ssa-dom.c.
9162         * cfganal.h (single_pred_edge_ignoring_loop_edges): Prototype.
9163         * tree-ssa-dom.c (single_incoming_edge_ignoring_loop_edges): Remove.
9164         (record_equivalences_from_incoming_edge): Add additional argument
9165         to single_pred_edge_ignoring_loop_edges call.
9166         * tree-ssa-uncprop.c (single_incoming_edge_ignoring_loop_edges): Remove.
9167         (uncprop_dom_walker::before_dom_children): Add additional argument
9168         to single_pred_edge_ignoring_loop_edges call.
9169         * tree-ssa-sccvn.c (sccvn_dom_walker::before_dom_children): Use
9170         single_pred_edge_ignoring_loop_edges rather than open coding.
9171         * tree-vrp.c (evrp_dom_walker::before_dom_children): Similarly.
9173 2017-11-03  Marc Glisse  <marc.glisse@inria.fr>
9175         * match.pd (-(-A)): Rewrite.
9177 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9179         * config/rs6000/rs60000-protos.h (rs6000_emit_sISEL): Delete.
9180         (rs6000_emit_int_cmove): New declaration.
9181         * config/rs6000/rs6000.c (rs6000_emit_int_cmove): Delete declaration.
9182         (rs6000_emit_sISEL): Delete.
9183         (rs6000_emit_int_cmove): Make non-static.
9184         * config/rs6000/rs6000.md (cstore<mode>4): Use rs6000_emit_int_cmove
9185         instead of rs6000_emit_sISEL.
9187 2017-11-03  Jan Hubicka  <hubicka@ucw.cz>
9189         * asan.c (create_cond_insert_point): Maintain profile.
9190         * ipa-utils.c (ipa_merge_profiles): Be sure only IPA profiles are
9191         merged.
9192         * basic-block.h (struct basic_block_def): Remove frequency.
9193         (EDGE_FREQUENCY): Use to_frequency
9194         * bb-reorder.c (push_to_next_round_p): Use only IPA counts for global
9195         heuristics.
9196         (find_traces): Update to use to_frequency.
9197         (find_traces_1_round): Likewise; use only IPA counts.
9198         (bb_to_key): Likewise.
9199         (connect_traces): Use IPA counts only.
9200         (copy_bb_p): Update to use to_frequency.
9201         (fix_up_crossing_landing_pad): Likewise.
9202         (sanitize_hot_paths): Likewise.
9203         * bt-load.c (basic_block_freq): Likewise.
9204         * cfg.c (init_flow): Set count_max to uninitialized.
9205         (check_bb_profile): Remove frequencies; check counts.
9206         (dump_bb_info): Do not dump frequencies.
9207         (update_bb_profile_for_threading): Update counts only.
9208         (scale_bbs_frequencies_int): Likewise.
9209         (MAX_SAFE_MULTIPLIER): Remove.
9210         (scale_bbs_frequencies_gcov_type): Update counts only.
9211         (scale_bbs_frequencies_profile_count): Update counts only.
9212         (scale_bbs_frequencies): Update counts only.
9213         * cfg.h (struct control_flow_graph): Add count-max.
9214         (update_bb_profile_for_threading): Update prototype.
9215         * cfgbuild.c (find_bb_boundaries): Do not update frequencies.
9216         (find_many_sub_basic_blocks): Likewise.
9217         * cfgcleanup.c (try_forward_edges): Likewise.
9218         (try_crossjump_to_edge): Likewise.
9219         * cfgexpand.c (expand_gimple_cond): Likewise.
9220         (expand_gimple_tailcall): Likewise.
9221         (construct_init_block): Likewise.
9222         (construct_exit_block): Likewise.
9223         * cfghooks.c (verify_flow_info): Check consistency of counts.
9224         (dump_bb_for_graph): Do not dump frequencies.
9225         (split_block_1): Do not update frequencies.
9226         (split_edge): Do not update frequencies.
9227         (make_forwarder_block): Do not update frequencies.
9228         (duplicate_block): Do not update frequencies.
9229         (account_profile_record): Do not update frequencies.
9230         * cfgloop.c (find_subloop_latch_edge_by_profile): Use IPA counts
9231         for global heuristics.
9232         * cfgloopanal.c (average_num_loop_insns): Update to use to_frequency.
9233         (expected_loop_iterations_unbounded): Use counts only.
9234         * cfgloopmanip.c (scale_loop_profile): Simplify.
9235         (create_empty_loop_on_edge): Simplify
9236         (loopify): Simplify
9237         (duplicate_loop_to_header_edge): Simplify
9238         * cfgrtl.c (force_nonfallthru_and_redirect): Update profile.
9239         (update_br_prob_note): Take care of removing note when profile
9240         becomes undefined.
9241         (relink_block_chain): Do not dump frequency.
9242         (rtl_account_profile_record): Use to_frequency.
9243         * cgraph.c (symbol_table::create_edge): Convert count to ipa count.
9244         (cgraph_edge::redirect_call_stmt_to_calle): Conver tcount to ipa count.
9245         (cgraph_update_edges_for_call_stmt_node): Likewise.
9246         (cgraph_edge::verify_count_and_frequency): Update.
9247         (cgraph_node::verify_node): Temporarily disable frequency verification.
9248         * cgraphbuild.c (compute_call_stmt_bb_frequency): Use
9249         to_cgraph_frequency.
9250         (cgraph_edge::rebuild_edges): Convert to ipa counts.
9251         * cgraphunit.c (init_lowered_empty_function): Do not initialize
9252         frequencies.
9253         (cgraph_node::expand_thunk): Update profile.
9254         * except.c (dw2_build_landing_pads): Do not update frequency.
9255         * final.c (compute_alignments): Use to_frequency.
9256         (dump_basic_block_info): Do not dump frequency.
9257         * gimple-pretty-print.c (dump_profile): Do not dump frequency.
9258         (dump_gimple_bb_header): Do not dump frequency.
9259         * gimple-ssa-isolate-paths.c (isolate_path): Do not update frequency;
9260         do update count.
9261         * gimple-streamer-in.c (input_bb): Do not stream frequency.
9262         * gimple-streamer-out.c (output_bb): Do not stream frequency.
9263         * haifa-sched.c (sched_pressure_start_bb): Use to_freuqency.
9264         (init_before_recovery): Do not update frequency.
9265         (sched_create_recovery_edges): Do not update frequency.
9266         * hsa-gen.c (convert_switch_statements): Do not update frequency.
9267         * ipa-cp.c (ipcp_propagate_stage): Update search for max_count.
9268         (ipa_cp_c_finalize): Set max_count to uninitialized.
9269         * ipa-fnsummary.c (get_minimal_bb): Use counts.
9270         (param_change_prob): Use counts.
9271         * ipa-profile.c (ipa_profile_generate_summary): Do not summarize
9272         local profiles.
9273         * ipa-split.c (consider_split): Use to_frequency.
9274         (split_function): Use to_frequency.
9275         * ira-build.c (loop_compare_func): Likewise.
9276         (mark_loops_for_removal): Likewise.
9277         (mark_all_loops_for_removal): Likewise.
9278         * loop-doloop.c (doloop_modify): Do not update frequency.
9279         * loop-unroll.c (unroll_loop_runtime_iterations): Do not update
9280         frequency.
9281         * lto-streamer-in.c (input_function): Update count_max.
9282         * omp-expand.c (expand_omp_taskreg): Update count_max.
9283         * omp-simd-clone.c (simd_clone_adjust): Update profile.
9284         * predict.c (maybe_hot_frequency_p): Use to_frequency.
9285         (maybe_hot_count_p): Use ipa counts only.
9286         (maybe_hot_bb_p): Simplify.
9287         (maybe_hot_edge_p): Simplify.
9288         (probably_never_executed): Do not take frequency argument.
9289         (probably_never_executed_bb_p): Do not pass frequency.
9290         (probably_never_executed_edge_p): Likewise.
9291         (combine_predictions_for_bb): Check that profile is nonzero.
9292         (propagate_freq): Do not set frequency.
9293         (drop_profile): Simplify.
9294         (counts_to_freqs): Simplify.
9295         (expensive_function_p): Use to_frequency.
9296         (propagate_unlikely_bbs_forward): Simplify.
9297         (determine_unlikely_bbs): Simplify.
9298         (estimate_bb_frequencies): Add hack to silence graphite issues.
9299         (compute_function_frequency): Use ipa counts.
9300         (pass_profile::execute): Update.
9301         (rebuild_frequencies): Use counts only.
9302         (force_edge_cold): Use counts only.
9303         * profile-count.c (profile_count::dump): Dump new count types.
9304         (profile_count::differs_from_p): Check compatiblity.
9305         (profile_count::to_frequency): New function.
9306         (profile_count::to_cgraph_frequency): New function.
9307         * profile-count.h (struct function): Declare.
9308         (enum profile_quality): Add profile_guessed_local and
9309         profile_guessed_global0.
9310         (class profile_proability): Decrease number of bits to 29;
9311         update from_reg_br_prob_note and to_reg_br_prob_note.
9312         (class profile_count: Update comment; decrease number of bits
9313         to 61. Check compatibility.
9314         (profile_count::compatible_p): New private member function.
9315         (profile_count::ipa_p): New member function.
9316         (profile_count::operator<): Handle global zero correctly.
9317         (profile_count::operator>): Handle global zero correctly.
9318         (profile_count::operator<=): Handle global zero correctly.
9319         (profile_count::operator>=): Handle global zero correctly.
9320         (profile_count::nonzero_p): New member function.
9321         (profile_count::force_nonzero): New member function.
9322         (profile_count::max): New member function.
9323         (profile_count::apply_scale): Handle IPA scalling.
9324         (profile_count::guessed_local): New member function.
9325         (profile_count::global0): New member function.
9326         (profile_count::ipa): New member function.
9327         (profile_count::to_frequency): Declare.
9328         (profile_count::to_cgraph_frequency): Declare.
9329         * profile.c (OVERLAP_BASE): Delete.
9330         (compute_frequency_overlap): Delete.
9331         (compute_branch_probabilities): Do not use compute_frequency_overlap.
9332         * regs.h (REG_FREQ_FROM_BB): Use to_frequency.
9333         * sched-ebb.c (rank): Use counts only.
9334         * shrink-wrap.c (handle_simple_exit): Use counts only.
9335         (try_shrink_wrapping): Use counts only.
9336         (place_prologue_for_one_component): Use counts only.
9337         * tracer.c (find_best_predecessor): Use to_frequency.
9338         (find_trace): Use to_frequency.
9339         (tail_duplicate): Use to_frequency.
9340         * trans-mem.c (expand_transaction): Do not update frequency.
9341         * tree-call-cdce.c: Do not update frequency.
9342         * tree-cfg.c (gimple_find_sub_bbs): Likewise.
9343         (gimple_merge_blocks): Likewise.
9344         (gimple_split_edge): Likewise.
9345         (gimple_duplicate_sese_region): Likewise.
9346         (gimple_duplicate_sese_tail): Likewise.
9347         (move_sese_region_to_fn): Likewise.
9348         (gimple_account_profile_record): Likewise.
9349         (insert_cond_bb): Likewise.
9350         * tree-complex.c (expand_complex_div_wide): Likewise.
9351         * tree-eh.c (lower_resx): Update profile.
9352         * tree-inline.c (copy_bb): Simplify count scaling; do not scale
9353         frequencies.
9354         (initialize_cfun): Do not initialize frequencies
9355         (freqs_to_counts): Delete.
9356         (copy_cfg_body): Ignore count parameter.
9357         (copy_body): Update.
9358         (expand_call_inline): Update count_max.
9359         (optimize_inline_calls): Update count_max.
9360         (tree_function_versioning): Update count_max.
9361         * tree-ssa-coalesce.c (coalesce_cost_bb): Use to_frequency.
9362         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Do not update
9363         frequency.
9364         * tree-ssa-loop-im.c (execute_sm_if_changed): Use counts only.
9365         * tree-ssa-loop-ivcanon.c (unloop_loops): Do not update freuqency.
9366         (try_peel_loop): Likewise.
9367         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Use
9368         to_frequency.
9369         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Pass -1.
9370         (tree_transform_and_unroll_loop): Do not use frequencies
9371         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations):
9372         Use reliable prediction only.
9373         * tree-ssa-loop-unswitch.c (hoist_guard): Do not use frequencies.
9374         * tree-ssa-sink.c (select_best_block): Use to_frequency.
9375         * tree-ssa-tail-merge.c (replace_block_by): Temporarily disable
9376         probability scaling.
9377         * tree-ssa-threadupdate.c (create_block_for_threading): Do
9378         not update frequency
9379         (any_remaining_duplicated_blocks): Likewise.
9380         (update_profile): Likewise.
9381         (estimated_freqs_path): Delete.
9382         (freqs_to_counts_path): Delete.
9383         (clear_counts_path): Delete.
9384         (ssa_fix_duplicate_block_edges): Likewise.
9385         (duplicate_thread_path): Likewise.
9386         * tree-switch-conversion.c (gen_inbound_check): Use counts.
9387         * tree-tailcall.c (decrease_profile): Do not update frequency.
9388         (eliminate_tail_call): Likewise.
9389         * tree-vect-loop-manip.c (vect_do_peeling): Likewise.
9390         * tree-vect-loop.c (scale_profile_for_vect_loop): Likewise.
9391         (optimize_mask_stores): Likewise.
9392         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
9393         * ubsan.c (ubsan_expand_null_ifn): Update profile.
9394         (ubsan_expand_ptr_ifn): Update profile.
9395         * value-prof.c (gimple_ic): Simplify.
9396         * value-prof.h (gimple_ic): Update prototype.
9397         * ipa-inline-transform.c (inline_transform): Fix scaling conditoins.
9398         * ipa-inline.c (compute_uninlined_call_time): Be sure that
9399         counts are nonzero.
9400         (want_inline_self_recursive_call_p): Likewise.
9401         (resolve_noninline_speculation): Only cummulate defined counts.
9402         (inline_small_functions): Use nonzero_p.
9403         (ipa_inline): Do not access freed node.
9405 2017-11-03  Wilco Dijkstra  <wdijkstr@arm.com>
9407         * config/aarch64/aarch64.c (aarch64_override_options_internal):
9408         Set PARAM_SCHED_PRESSURE_ALGORITHM to SCHED_PRESSURE_MODEL.
9410 2017-11-03  Kito Cheng  <kito.cheng@gmail.com>
9412         * config/riscv/riscv.c (riscv_legitimize_move): Handle
9413         non-legitimate address.
9415 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9417         * config/rs6000/rs6000.md (*lt0_disi): Delete.
9418         (*lt0_<mode>di, *lt0_<mode>si): New.
9420 2017-11-03  Segher Boessenkool  <segher@kernel.crashing.org>
9422         * config/rs6000/rs6000.md (move_from_CR_ov_bit): Change condition to
9423         TARGET_PAIRED_FLOAT.
9425 2017-11-03  Siddhesh Poyarekar  <siddhesh.poyarekar@linaro.org>
9426             Jim Wilson  <jim.wilson@linaro.org>
9428         * config/aarch64/aarch64-cores.def (saphira): New CPU.
9429         * config/aarch64/aarch64-tune.md: Regenerated.
9430         * doc/invoke.texi (AArch64 Options/-mtune): Add "saphira".
9431         * gcc/config/aarch64/aarch64.c (saphira_tunings): New tuning table.
9433 2017-11-03  Cupertino Miranda  <cmiranda@synopsys.com>
9435         * config/arc/arc.c (arc_save_restore): Corrected CFA note.
9436         (arc_expand_prologue): Restore blink for millicode.
9437         * config/arc/linux.h (LINK_EH_SPEC): Defined.
9439 2017-11-03  Richard Sandiford  <richard.sandiford@linaro.org>
9441         PR target/82809
9442         * config/i386/i386.c (ix86_vector_duplicate_value): Use
9443         gen_vec_duplicate after forcing the scalar into a register.
9445 2017-11-02  Segher Boessenkool  <segher@kernel.crashing.org>
9447         * combine (try_combine): Print the insns input to try_combine to the
9448         dump file.
9450 2017-11-02  Steve Ellcey  <sellcey@cavium.com>
9452         PR target/79868
9453         * config/aarch64/aarch64-c.c (aarch64_pragma_target_parse):
9454         Remove second argument from aarch64_process_target_attr call.
9455         * config/aarch64/aarch64-protos.h (aarch64_process_target_attr):
9456         Ditto.
9457         * config/aarch64/aarch64.c (aarch64_attribute_info): Change
9458         field type.
9459         (aarch64_handle_attr_arch): Remove second argument.
9460         (aarch64_handle_attr_cpu): Ditto.
9461         (aarch64_handle_attr_tune): Ditto.
9462         (aarch64_handle_attr_isa_flags): Ditto.
9463         (aarch64_process_one_target_attr): Ditto.
9464         (aarch64_process_target_attr): Ditto.
9465         (aarch64_option_valid_attribute_p): Remove second argument.
9466         on aarch64_process_target_attr call.
9468 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9470         * diagnostic.c: Include "selftest-diagnostic.h".
9471         (selftest::assert_location_text): New function.
9472         (selftest::test_diagnostic_get_location_text): New function.
9473         (selftest::diagnostic_c_tests): Call it.
9475 2017-11-02  David Malcolm  <dmalcolm@redhat.com>
9477         * Makefile.in (OBJS-libcommon): Add selftest-diagnostic.o.
9478         * diagnostic-show-locus.c: Include "selftest-diagnostic.h".
9479         (class selftest::test_diagnostic_context): Move to...
9480         * selftest-diagnostic.c: New file.
9481         * selftest-diagnostic.h: New file.
9483 2017-11-02  James Bowman  <james.bowman@ftdichip.com>
9485         * config/ft32/ft32.c (ft32_addr_space_legitimate_address_p): increase
9486         offset range for FT32B.
9487         * config/ft32/ft32.h: option "mcompress" enables relaxation.
9488         * config/ft32/ft32.md: Add TARGET_NOPM.
9489         * config/ft32/ft32.opt: Add mft32b, mcompress, mnopm.
9490         * gcc/doc/invoke.texi: Add mft32b, mcompress, mnopm.
9492 2017-11-02  Wilco Dijkstra  <wdijkstr@arm.com>
9494         * config/aarch64/aarch64.h (MALLOC_ABI_ALIGNMENT): New define.
9496 2017-11-02  Jeff Law  <law@redhat.com>
9498         * gimple-ssa-sprintf.c (sprintf_dom_walker): Remove
9499         virtual keyword on FINAL OVERRIDE members.
9501         * tree-ssa-propagate.h (ssa_propagation_engine): Group
9502         virtuals together.  Add virtual destructor.
9503         (substitute_and_fold_engine): Similarly.
9505 2017-11-02  Jan Hubicka  <hubicka@ucw.cz>
9507         * x86-tune.def (X86_TUNE_USE_INCDEC): Enable for Haswell+.
9509 2017-11-02  Richard Biener  <rguenther@suse.de>
9511         PR tree-optimization/82795
9512         * tree-if-conv.c (predicate_mem_writes): Remove bogus assert.
9514 2017-11-02  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
9516         * acinclude.m4 (gcc_AC_INITFINI_ARRAY): Don't require
9517         gcc_SUN_LD_VERSION.
9518         (gcc_GAS_CHECK_FEATURE): Remove.
9519         * configure.ac (ld_vers) <*-*-solaris2*>: Move comments from
9520         gcc_AC_INITFINI_ARRAY here.  Update for Solaris 11.4 changes.
9521         * configure: Regenerate.
9523 2017-11-02  Claudiu Zissulescu <claziss@synopsys.com>
9525         * config/arc/arc.c (hwloop_optimize): Account for empty
9526         body loops.
9528 2017-11-02  Richard Biener  <rguenther@suse.de>
9530         PR middle-end/82765
9531         * varasm.c (decode_addr_const): Make offset HOST_WIDE_INT.
9532         Truncate ARRAY_REF index and element size.
9534 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9536         * doc/invoke.texi (RISC-V Options): Use "@minus{}2 GB", not "-2 GB".
9538 2017-11-01  Jeff Law  <law@redhat.com>
9540         * tree-ssa-ccp.c (ccp_folder): New class derived from
9541         substitute_and_fold_engine.
9542         (ccp_folder::get_value): New member function.
9543         (ccp_folder::fold_stmt): Renamed from ccp_fold_stmt.
9544         (ccp_fold_stmt): Remove prototype.
9545         (ccp_finalize): Call substitute_and_fold from the ccp_class.
9546         * tree-ssa-copy.c (copy_folder): New class derived from
9547         substitute_and_fold_engine.
9548         (copy_folder::get_value): Renamed from get_value.
9549         (fini_copy_prop): Call substitute_and_fold from copy_folder class.
9550         * tree-vrp.c (vrp_folder): New class derived from
9551         substitute_and_fold_engine.
9552         (vrp_folder::fold_stmt): Renamed from vrp_fold_stmt.
9553         (vrp_folder::get_value): New member function.
9554         (vrp_finalize): Call substitute_and_fold from vrp_folder class.
9555         (evrp_dom_walker::before_dom_children): Similarly for replace_uses_in.
9556         * tree-ssa-propagate.h (substitute_and_fold_engine): New class to
9557         provide a class interface to folder/substitute routines.
9558         (ssa_prop_fold_stmt_fn): Remove typedef.
9559         (ssa_prop_get_value_fn): Likewise.
9560         (subsitute_and_fold): Remove prototype.
9561         (replace_uses_in): Likewise.
9562         * tree-ssa-propagate.c (substitute_and_fold_engine::replace_uses_in):
9563         Renamed from replace_uses_in.  Call the virtual member function
9564         (substitute_and_fold_engine::replace_phi_args_in): Similarly.
9565         (substitute_and_fold_dom_walker): Remove initialization of
9566         data member entries for calbacks.  Add substitute_and_fold_engine
9567         member and initialize it.
9568         (substitute_and_fold_dom_walker::before_dom_children0: Use the
9569         member functions for get_value, replace_phi_args_in c
9570         replace_uses_in, and fold_stmt calls.
9571         (substitute_and_fold_engine::substitute_and_fold): Renamed from
9572         substitute_and_fold.  Remove assert.   Update ctor call.
9574         * tree-ssa-propagate.h (ssa_prop_visit_stmt_fn): Remove typedef.
9575         (ssa_prop_visit_phi_fn): Likewise.
9576         (class ssa_propagation_engine): New class to provide an interface
9577         into ssa_propagate.
9578         * tree-ssa-propagate.c (ssa_prop_visit_stmt): Remove file scoped
9579         variable.
9580         (ssa_prop_visit_phi): Likewise.
9581         (ssa_propagation_engine::simulate_stmt): Moved into class.
9582         Call visit_phi/visit_stmt from the class rather than via
9583         file scoped static variables.
9584         (ssa_propagation_engine::simulate_block): Moved into class.
9585         (ssa_propagation_engine::process_ssa_edge_worklist): Similarly.
9586         (ssa_propagation_engine::ssa_propagate): Similarly.  No longer
9587         set file scoped statics for the visit_stmt/visit_phi callbacks.
9588         * tree-complex.c (complex_propagate): New class derived from
9589         ssa_propagation_engine.
9590         (complex_propagate::visit_stmt): Renamed from complex_visit_stmt.
9591         (complex_propagate::visit_phi): Renamed from complex_visit_phi.
9592         (tree_lower_complex): Call ssa_propagate via the complex_propagate
9593         class.
9594         * tree-ssa-ccp.c: (ccp_propagate): New class derived from
9595         ssa_propagation_engine.
9596         (ccp_propagate::visit_phi): Renamed from ccp_visit_phi_node.
9597         (ccp_propagate::visit_stmt): Renamed from ccp_visit_stmt.
9598         (do_ssa_ccp): Call ssa_propagate from the ccp_propagate class.
9599         * tree-ssa-copy.c (copy_prop): New class derived from
9600         ssa_propagation_engine.
9601         (copy_prop::visit_stmt): Renamed from copy_prop_visit_stmt.
9602         (copy_prop::visit_phi): Renamed from copy_prop_visit_phi_node.
9603         (execute_copy_prop): Call ssa_propagate from the copy_prop class.
9604         * tree-vrp.c (vrp_prop): New class derived from ssa_propagation_engine.
9605         (vrp_prop::visit_stmt): Renamed from vrp_visit_stmt.
9606         (vrp_prop::visit_phi): Renamed from vrp_visit_phi_node.
9607         (execute_vrp): Call ssa_propagate from the vrp_prop class.
9609 2017-11-01  Jakub Jelinek  <jakub@redhat.com>
9611         PR rtl-optimization/82778
9612         PR rtl-optimization/82597
9613         * compare-elim.c (struct comparison): Add in_a_setter field.
9614         (find_comparison_dom_walker::before_dom_children): Remove killed
9615         bitmap and df_simulate_find_defs call, instead walk the defs.
9616         Compute last_setter and initialize in_a_setter.  Merge definitions
9617         with first initialization for a few variables.
9618         (try_validate_parallel): Use insn_invalid_p instead of
9619         recog_memoized.  Return insn rather than just the pattern.
9620         (try_merge_compare): Fix up comment.  Don't uselessly test if
9621         in_a is a REG_P.  Use cmp->in_a_setter instead of walking UD
9622         chains.
9623         (execute_compare_elim_after_reload): Remove df_chain_add_problem
9624         call.
9626 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9627             Alan Hayward  <alan.hayward@arm.com>
9628             David Sherwood  <david.sherwood@arm.com>
9630         * config/aarch64/aarch64.c (aarch64_rtx_costs): Use
9631         aarch64_hard_regno_nregs to get the number of registers
9632         in a mode.
9634 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9635             Alan Hayward  <alan.hayward@arm.com>
9636             David Sherwood  <david.sherwood@arm.com>
9638         * config/aarch64/constraints.md (Upl): Rename to...
9639         (Uaa): ...this.
9640         * config/aarch64/aarch64.md
9641         (*zero_extend<SHORT:mode><GPI:mode>2_aarch64, *addsi3_aarch64_uxtw):
9642         Update accordingly.
9644 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9645             Alan Hayward  <alan.hayward@arm.com>
9646             David Sherwood  <david.sherwood@arm.com>
9648         * config/aarch64/aarch64.c (aarch64_add_constant_internal)
9649         (aarch64_add_constant, aarch64_add_sp, aarch64_sub_sp): Move
9650         earlier in file.
9652 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9653             Alan Hayward  <alan.hayward@arm.com>
9654             David Sherwood  <david.sherwood@arm.com>
9656         * config/aarch64/aarch64.c (aarch64_evpc_trn, aarch64_evpc_uzp)
9657         (aarch64_evpc_zip, aarch64_evpc_ext, aarch64_evpc_rev)
9658         (aarch64_evpc_dup): Generate rtl direcly, rather than using
9659         named expanders.
9660         (aarch64_expand_vec_perm_const_1): Explicitly check for permutes
9661         of a single element.
9662         * config/aarch64/iterators.md: Add a comment above the permute
9663         unspecs to say that they are generated directly by
9664         aarch64_expand_vec_perm_const.
9665         * config/aarch64/aarch64-simd.md: Likewise the permute instructions.
9667 2017-11-01  Nathan Sidwell  <nathan@acm.org>
9669         * tree-dump.c (dequeue_and_dump): Use HAS_DECL_ASSEMBLER_NAME_P.
9671 2017-11-01  Palmer Dabbelt  <palmer@dabbelt.com>
9673         * doc/invoke.texi (RISC-V Options): Explicitly name the medlow
9674         and medany code models, and describe what they do.
9676 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9678         Revert accidental duplicate:
9680         * combine.c (can_change_dest_mode): Reject changes in
9681         REGMODE_NATURAL_SIZE.
9683 2017-11-01  Segher Boessenkool  <segher@kernel.crashing.org>
9685         PR rtl-optimization/64682
9686         PR rtl-optimization/69567
9687         PR rtl-optimization/69737
9688         PR rtl-optimization/82683
9689         * combine.c (distribute_notes) <REG_DEAD>: If the new I2 sets the same
9690         register mentioned in the note, drop the note, unless it came from I3,
9691         in which case it should go to I3 again.
9693 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9695         * tree-ssa-dse.c (normalize_ref): Check whether the ranges overlap
9696         and return false if not.
9697         (clear_bytes_written_by, live_bytes_read): Update accordingly.
9699 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9701         * tree-ssa-alias.h (ranges_overlap_p): Return false if either
9702         range is known to be empty.
9704 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9705             Alan Hayward  <alan.hayward@arm.com>
9706             David Sherwood  <david.sherwood@arm.com>
9708         * simplify-rtx.c (simplify_const_unary_operation): Use GET_MODE_NUNITS
9709         and CONST_VECTOR_NUNITS instead of computing the number of units from
9710         the byte sizes of the vector and element.
9711         (simplify_binary_operation_1): Likewise.
9712         (simplify_const_binary_operation): Likewise.
9713         (simplify_ternary_operation): Likewise.
9715 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9716             Alan Hayward  <alan.hayward@arm.com>
9717             David Sherwood  <david.sherwood@arm.com>
9719         * var-tracking.c (INT_MEM_OFFSET): Replace with...
9720         (int_mem_offset): ...this new function.
9721         (var_mem_set, var_mem_delete_and_set, var_mem_delete)
9722         (find_mem_expr_in_1pdv, dataflow_set_preserve_mem_locs)
9723         (same_variable_part_p, use_type, add_stores, vt_get_decl_and_offset):
9724         Update accordingly.
9726 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9727             Alan Hayward  <alan.hayward@arm.com>
9728             David Sherwood  <david.sherwood@arm.com>
9730         * lower-subreg.c (interesting_mode_p): New function.
9731         (compute_costs, find_decomposable_subregs, decompose_register)
9732         (simplify_subreg_concatn, can_decompose_p, resolve_simple_move)
9733         (resolve_clobber, dump_choices): Use it.
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         * rtlhash.c (add_rtx): Use add_hwi for 'w' and add_int for 'i'.
9741 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9742             Alan Hayward  <alan.hayward@arm.com>
9743             David Sherwood  <david.sherwood@arm.com>
9745         * alias.c (find_base_value, find_base_term): Only process integer
9746         truncations.  Check the precision rather than the size.
9748 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9749             Alan Hayward  <alan.hayward@arm.com>
9750             David Sherwood  <david.sherwood@arm.com>
9752         * machmode.h (is_narrower_int_mode): New function
9753         * optabs.c (expand_float, expand_fix): Use it.
9754         * dwarf2out.c (rotate_loc_descriptor): Likewise.
9756 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9757             Alan Hayward  <alan.hayward@arm.com>
9758             David Sherwood  <david.sherwood@arm.com>
9760         * rtl.h (narrower_subreg_mode): New function.
9761         * ira-color.c (update_costs_from_allocno): Use it.
9763 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9764             Alan Hayward  <alan.hayward@arm.com>
9765             David Sherwood  <david.sherwood@arm.com>
9767         * optabs-query.h (convert_optab_p): New function, split out from...
9768         (convert_optab_handler): ...here.
9769         (widening_optab_handler): Delete.
9770         (find_widening_optab_handler): Remove permit_non_widening parameter.
9771         (find_widening_optab_handler_and_mode): Likewise.  Provide an
9772         override that operates on mode class wrappers.
9773         * optabs-query.c (widening_optab_handler): Delete.
9774         (find_widening_optab_handler_and_mode): Remove permit_non_widening
9775         parameter.  Assert that the two modes are the same class and that
9776         the "from" mode is narrower than the "to" mode.  Use
9777         convert_optab_handler instead of widening_optab_handler.
9778         * expmed.c (expmed_mult_highpart_optab): Use convert_optab_handler
9779         instead of widening_optab_handler.
9780         * expr.c (expand_expr_real_2): Update calls to
9781         find_widening_optab_handler.
9782         * optabs.c (expand_widen_pattern_expr): Likewise.
9783         (expand_binop_directly): Take the insn_code as a parameter.
9784         (expand_binop): Only call find_widening_optab_handler for
9785         conversion optabs; use optab_handler otherwise.  Update calls
9786         to find_widening_optab_handler and expand_binop_directly.
9787         Use convert_optab_handler instead of widening_optab_handler.
9788         * tree-ssa-math-opts.c (convert_mult_to_widen): Update calls to
9789         find_widening_optab_handler and use scalar_mode rather than
9790         machine_mode.
9791         (convert_plusminus_to_widen): Likewise.
9793 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9794             Alan Hayward  <alan.hayward@arm.com>
9795             David Sherwood  <david.sherwood@arm.com>
9797         * machmode.h (fixed_size_mode): New class.
9798         * rtl.h (get_pool_mode): Return fixed_size_mode.
9799         * gengtype.c (main): Add fixed_size_mode.
9800         * target.def (get_raw_result_mode): Return a fixed_size_mode.
9801         (get_raw_arg_mode): Likewise.
9802         * doc/tm.texi: Regenerate.
9803         * targhooks.h (default_get_reg_raw_mode): Return a fixed_size_mode.
9804         * targhooks.c (default_get_reg_raw_mode): Likewise.
9805         * config/ia64/ia64.c (ia64_get_reg_raw_mode): Likewise.
9806         * config/mips/mips.c (mips_get_reg_raw_mode): Likewise.
9807         * config/msp430/msp430.c (msp430_get_raw_arg_mode): Likewise.
9808         (msp430_get_raw_result_mode): Likewise.
9809         * config/avr/avr-protos.h (regmask): Use as_a <fixed_side_mode>
9810         * dbxout.c (dbxout_parms): Require fixed-size modes.
9811         * expr.c (copy_blkmode_from_reg, copy_blkmode_to_reg): Likewise.
9812         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
9813         * omp-low.c (lower_oacc_reductions): Likewise.
9814         * simplify-rtx.c (simplify_immed_subreg): Take fixed_size_modes.
9815         (simplify_subreg): Update accordingly.
9816         * varasm.c (constant_descriptor_rtx::mode): Change to fixed_size_mode.
9817         (force_const_mem): Update accordingly.  Return NULL_RTX for modes
9818         that aren't fixed-size.
9819         (get_pool_mode): Return a fixed_size_mode.
9820         (output_constant_pool_2): Take a fixed_size_mode.
9822 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9823             Alan Hayward  <alan.hayward@arm.com>
9824             David Sherwood  <david.sherwood@arm.com>
9826         * doc/rtl.texi (vec_series): Document.
9827         (const): Say that the operand can be a vec_series.
9828         * rtl.def (VEC_SERIES): New rtx code.
9829         * rtl.h (const_vec_series_p_1): Declare.
9830         (const_vec_series_p): New function.
9831         * emit-rtl.h (gen_const_vec_series): Declare.
9832         (gen_vec_series): Likewise.
9833         * emit-rtl.c (const_vec_series_p_1, gen_const_vec_series)
9834         (gen_vec_series): Likewise.
9835         * optabs.c (expand_mult_highpart): Use gen_const_vec_series.
9836         * simplify-rtx.c (simplify_unary_operation): Handle negations
9837         of vector series.
9838         (simplify_binary_operation_series): New function.
9839         (simplify_binary_operation_1): Use it.  Handle VEC_SERIES.
9840         (test_vector_ops_series): New function.
9841         (test_vector_ops): Call it.
9842         * config/powerpcspe/altivec.md (altivec_lvsl): Use
9843         gen_const_vec_series.
9844         (altivec_lvsr): Likewise.
9845         * config/rs6000/altivec.md (altivec_lvsl, altivec_lvsr): Likewise.
9847 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9848             Alan Hayward  <alan.hayward@arm.com>
9849             David Sherwood  <david.sherwood@arm.com>
9851         * doc/rtl.texi (const): Update description of address constants.
9852         Say that vector constants are allowed too.
9853         * common.md (E, F): Use CONSTANT_P instead of checking for
9854         CONST_VECTOR.
9855         * emit-rtl.c (gen_lowpart_common): Use const_vec_p instead of
9856         checking for CONST_VECTOR.
9857         * expmed.c (make_tree): Use build_vector_from_val for a CONST
9858         VEC_DUPLICATE.
9859         * expr.c (expand_expr_real_2): Check for vector modes instead
9860         of checking for CONST_VECTOR.
9861         * rtl.h (const_vec_p): New function.
9862         (const_vec_duplicate_p): Check for a CONST VEC_DUPLICATE.
9863         (unwrap_const_vec_duplicate): Handle them here too.
9865 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9866             David Malcolm  <dmalcolm@redhat.com>
9867             Alan Hayward  <alan.hayward@arm.com>
9868             David Sherwood  <david.sherwood@arm.com>
9870         * rtl.h (vec_duplicate_p): New function.
9871         * selftest-rtl.c (assert_rtx_eq_at): New function.
9872         * selftest-rtl.h (ASSERT_RTX_EQ): New macro.
9873         (assert_rtx_eq_at): Declare.
9874         * selftest.h (selftest::simplify_rtx_c_tests): Declare.
9875         * selftest-run-tests.c (selftest::run_tests): Call it.
9876         * simplify-rtx.c: Include selftest.h and selftest-rtl.h.
9877         (simplify_unary_operation_1): Recursively handle vector duplicates.
9878         (simplify_binary_operation_1): Likewise.  Handle VEC_SELECTs of
9879         vector duplicates.
9880         (simplify_subreg): Handle subregs of vector duplicates.
9881         (make_test_reg, test_vector_ops_duplicate, test_vector_ops)
9882         (selftest::simplify_rtx_c_tests): New functions.
9884 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9885             Alan Hayward  <alan.hayward@arm.com>
9886             David Sherwood  <david.sherwood@arm.com>
9888         * emit-rtl.h (gen_const_vec_duplicate): Declare.
9889         (gen_vec_duplicate): Likewise.
9890         * emit-rtl.c (gen_const_vec_duplicate_1): New function, split
9891         out from...
9892         (gen_const_vector): ...here.
9893         (gen_const_vec_duplicate, gen_vec_duplicate): New functions.
9894         (gen_rtx_CONST_VECTOR): Use gen_const_vec_duplicate for constants
9895         whose elements are all equal.
9896         * optabs.c (expand_vector_broadcast): Use gen_const_vec_duplicate.
9897         * simplify-rtx.c (simplify_const_unary_operation): Likewise.
9898         (simplify_relational_operation): Likewise.
9899         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
9900         Likewise.
9901         (aarch64_simd_dup_constant): Use gen_vec_duplicate.
9902         (aarch64_expand_vector_init): Likewise.
9903         * config/arm/arm.c (neon_vdup_constant): Likewise.
9904         (neon_expand_vector_init): Likewise.
9905         (arm_expand_vec_perm): Use gen_const_vec_duplicate.
9906         (arm_block_set_unaligned_vect): Likewise.
9907         (arm_block_set_aligned_vect): Likewise.
9908         * config/arm/neon.md (neon_copysignf<mode>): Likewise.
9909         * config/i386/i386.c (ix86_expand_vec_perm): Likewise.
9910         (expand_vec_perm_even_odd_pack): Likewise.
9911         (ix86_vector_duplicate_value): Use gen_vec_duplicate.
9912         * config/i386/sse.md (one_cmpl<mode>2): Use CONSTM1_RTX.
9913         * config/ia64/ia64.c (ia64_expand_vecint_compare): Use
9914         gen_const_vec_duplicate.
9915         * config/ia64/vect.md (addv2sf3, subv2sf3): Use CONST1_RTX.
9916         * config/mips/mips.c (mips_gen_const_int_vector): Use
9917         gen_const_vec_duplicate.
9918         (mips_expand_vector_init): Use CONST0_RTX.
9919         * config/powerpcspe/altivec.md (abs<mode>2, nabs<mode>2): Likewise.
9920         (define_split): Use gen_const_vec_duplicate.
9921         * config/rs6000/altivec.md (abs<mode>2, nabs<mode>2): Use CONST0_RTX.
9922         (define_split): Use gen_const_vec_duplicate.
9923         * config/s390/vx-builtins.md (vec_genmask<mode>): Likewise.
9924         (vec_ctd_s64, vec_ctd_u64, vec_ctsl, vec_ctul): Likewise.
9925         * config/spu/spu.c (spu_const): Likewise.
9927 2017-11-01  Richard Sandiford  <richard.sandiford@linaro.org>
9928             Alan Hayward  <alan.hayward@arm.com>
9929             David Sherwood  <david.sherwood@arm.com>
9931         * combine.c (can_change_dest_mode): Reject changes in
9932         REGMODE_NATURAL_SIZE.
9934 2017-10-31  Sandra Loosemore  <sandra@codesourcery.com>
9936         * configure.ac (--enable-libssp): New.
9937         (gcc_cv_libc_provides_ssp): Check for explicit setting before
9938         trying to determine target-specific default.  Adjust indentation.
9939         * configure: Regenerated.
9940         * doc/install.texi (Configuration): Expand --disable-libssp
9941         documentation.
9943 2017-10-31  Daniel Santos  <daniel.santos@pobox.com>
9945         config/i386/i386.c (ix86_expand_epilogue): Correct stack
9946         calculation.
9948 2017-10-31  Martin Jambor  <mjambor@suse.cz>
9950         PR c++/81702
9951         * gimple-fold.c (gimple_get_virt_method_for_vtable): Remove assert.
9953 2017-10-31  David Malcolm  <dmalcolm@redhat.com>
9955         * auto-profile.c (autofdo_source_profile::read): Use
9956         UNKNOWN_LOCATION rather than 0.
9957         * diagnostic-core.h (warning_at_rich_loc): Rename to...
9958         (warning_at): ...this overload.
9959         (warning_at_rich_loc_n): Rename to...
9960         (warning_n): ...this overload.
9961         (error_at_rich_loc): Rename to...
9962         (error_at): ...this overload.
9963         (pedwarn_at_rich_loc): Rename to...
9964         (pedwarn): ...this overload.
9965         (permerror_at_rich_loc): Rename to...
9966         (permerror): ...this overload.
9967         (inform_at_rich_loc): Rename to...
9968         (inform): ...this overload.
9969         * diagnostic.c: (diagnostic_n_impl): Delete location_t-based decl.
9970         (diagnostic_n_impl_richloc): Rename to...
9971         (diagnostic_n_impl): ...this rich_location *-based decl.
9972         (inform_at_rich_loc): Rename to...
9973         (inform): ...this, and add an assertion.
9974         (inform_n): Update for removal of location_t-based diagnostic_n_impl.
9975         (warning_at_rich_loc): Rename to...
9976         (warning_at): ...this, and add an assertion.
9977         (warning_at_rich_loc_n): Rename to...
9978         (warning_n): ...this, and add an assertion.
9979         (warning_n): Update location_t-based implementation for removal of
9980         location_t-based diagnostic_n_impl.
9981         (pedwarn_at_rich_loc): Rename to...
9982         (pedwarn): ...this, and add an assertion.
9983         (permerror_at_rich_loc): Rename to...
9984         (permerror): ...this, and add an assertion.
9985         (error_n): Update for removal of location_t-based diagnostic_n_impl.
9986         (error_at_rich_loc): Rename to...
9987         (error_at): ...this, and add an assertion.
9988         * gcc.c (do_spec_1): Use UNKNOWN_LOCATION rather than 0.
9989         (driver::do_spec_on_infiles): Likewise.
9990         * substring-locations.c (format_warning_va): Update for renaming
9991         of inform_at_rich_loc.
9993 2017-10-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
9995         * builtins.def (DEF_FLOATN_BUILTIN): Change most _Float<N> and
9996         _Float<N>X built-in functions so that the variant without the
9997         "__builtin_" prefix is only enabled for the GNU C and Objective C
9998         languages when they are in non-strict ANSI/ISO mode.
9999         (DEF_EXT_LIB_FLOATN_NX_BUILTINS): Likewise.
10000         * target.def (floatn_builtin_p): Add a target hook to control
10001         whether _Float<N> and _Float<N>X built-in functions without the
10002         "__builtin_" prefix are enabled, and return true for C and
10003         Objective C in the default hook.  Include langhooks.h in
10004         targhooks.c.
10005         * targhooks.h (default_floatn_builtin_p): Likewise.
10006         * targhooks.c (default_floatn_builtin_p): Likewise.
10007         * doc/tm.texi.in (TARGET_FLOATN_BUILTIN_P): Document the
10008         floatn_builtin_p target hook.
10009         * doc/tm.texi (TARGET_FLOATN_BUILTIN_P): Likewise.
10011 2017-10-31  Matthew Fortune  <matthew.fortune@imgtec.com>
10012             Eric Botcazou  <ebotcazou@adacore.com>
10014         PR rtl-optimization/81803
10015         * lra-constraints.c (curr_insn_transform): Also reload the whole
10016         register for a strict subreg no wider than a word if this is for
10017         a WORD_REGISTER_OPERATIONS target.
10019 2017-10-31  Jason Merrill  <jason@redhat.com>
10021         * gdbinit.in: Skip over inlines from timevar.h.
10023 2017-10-31  Martin Liska  <mliska@suse.cz>
10025         * doc/gcov.texi: Document new option.
10026         * gcov.c (print_usage): Likewise print it.
10027         (process_args): Support the argument.
10028         (format_count): New function.
10029         (format_gcov): Use the function.
10031 2017-10-31  Martin Liska  <mliska@suse.cz>
10033         * gcov.c (struct name_map): do not use typedef.
10034         Define operator== and operator<.
10035         (name_search): Remove.
10036         (name_sort): Remove.
10037         (main): Do not allocate names.
10038         (process_file): Add vertical space.
10039         (generate_results): Use std::find.
10040         (release_structures): Do not release memory.
10041         (find_source): Use std::find.
10043 2017-10-31  Martin Liska  <mliska@suse.cz>
10045         * gcov.c (struct line_info): Remove it's typedef.
10046         (line_info::line_info): Add proper ctor.
10047         (line_info::has_block): Do not use a typedef.
10048         (struct source_info): Do not use typedef.
10049         (circuit): Likewise.
10050         (get_cycles_count): Likewise.
10051         (output_intermediate_file): Iterate via vector iterator.
10052         (add_line_counts): Use std::vector methods.
10053         (accumulate_line_counts): Likewise.
10054         (output_lines): Likewise.
10056 2017-10-31  Martin Liska  <mliska@suse.cz>
10058         * gcov.c (struct source_info): Remove typedef.
10059         (source_info::source_info): Add proper ctor.
10060         (accumulate_line_counts): Use struct, not it's typedef.
10061         (output_gcov_file): Likewise.
10062         (output_lines): Likewise.
10063         (main): Do not allocate an array.
10064         (output_intermediate_file): Use size of vector container.
10065         (process_file): Resize the vector.
10066         (generate_results): Do not preallocate, use newly added vector
10067         lines.
10068         (release_structures): Do not release sources.
10069         (find_source): Use vector methods.
10070         (add_line_counts): Do not use typedef.
10072 2017-10-31  Martin Liska  <mliska@suse.cz>
10074         * doc/gcov.texi: Document that.
10075         * gcov.c (add_line_counts): Mark lines with a non-executed
10076         statement.
10077         (output_line_beginning): Handle such lines.
10078         (output_lines): Pass new argument.
10079         (output_intermediate_file): Print it in intermediate format.
10081 2017-10-31  Martin Liska  <mliska@suse.cz>
10083         * color-macros.h: New file.
10084         * diagnostic-color.c: Factor out color related to macros to
10085         color-macros.h.
10086         * doc/gcov.texi: Document -k option.
10087         * gcov.c (INCLUDE_STRING): Include string.h.
10088         (print_usage): Add -k option.
10089         (process_args): Parse it.
10090         (pad_count_string): New function.
10091         (output_line_beginning): Likewise.
10092         (DEFAULT_LINE_START): New macro.
10093         (output_lines): Support color output.
10095 2017-10-31  Martin Liska  <mliska@suse.cz>
10097         PR gcov-profile/82633
10098         * doc/gcov.texi: Document -fkeep-{static,inline}-functions and
10099         their interaction with GCOV infrastructure.
10100         * configure.ac: Add -fkeep-{inline,static}-functions to
10101         coverage_flags.
10102         * configure: Regenerate.
10104 2017-10-31  Uros Bizjak  <ubizjak@gmail.com>
10106         PR target/82772
10107         * config/alpha/sync.md (fetchop_constr) <and>: Change to "rINM".
10109 2017-10-31  Segher Boessenkool  <segher@kernel.crashing.org>
10111         PR target/82674
10112         * config/rs6000/rs6000.md (allocate_stack): Force update interval
10113         into a register if it does not fit into an immediate offset field.
10115 2017-10-31  Olivier Hainque  <hainque@adacore.com>
10117         * gcc/Makefile.in (FLAGS_TO_PASS): Pass libsubdir as well.
10119 2017-10-31  Julia Koval  <julia.koval@intel.com>
10121         * config.gcc: Add gfniintrin.h.
10122         * config/i386/gfniintrin.h: New.
10123         * config/i386/i386-builtin-types.def
10124         (__builtin_ia32_vgf2p8affineinvqb_v64qi,
10125         __builtin_ia32_vgf2p8affineinvqb_v64qi_mask,
10126         __builtin_ia32_vgf2p8affineinvqb_v32qi,
10127         __builtin_ia32_vgf2p8affineinvqb_v32qi_mask,
10128         __builtin_ia32_vgf2p8affineinvqb_v16qi,
10129         __builtin_ia32_vgf2p8affineinvqb_v16qi_mask): New builtins.
10130         * config/i386/i386-builtin.def (V64QI_FTYPE_V64QI_V64QI_INT_V64QI_UDI,
10131         V32QI_FTYPE_V32QI_V32QI_INT_V32QI_USI,
10132         V16QI_FTYPE_V16QI_V16QI_INT_V16QI_UHI,
10133         V64QI_FTYPE_V64QI_V64QI_INT): New types.
10134         * config/i386/i386.c (ix86_expand_args_builtin): Handle new types.
10135         * config/i386/immintrin.h: Include gfniintrin.h.
10136         * config/i386/sse.md (vgf2p8affineinvqb_*) New pattern.
10138 2017-10-30  Eric Botcazou  <ebotcazou@adacore.com>
10140         * gcc.c (HAVE_TARGET_EXECUTABLE_SUFFIX): Remove old kludge.
10142 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10144         * config/arm/arm.md (ashldi3): Remove shift by 1 expansion.
10145         (arm_ashldi3_1bit): Remove pattern.
10146         (ashrdi3): Remove shift by 1 expansion.
10147         (arm_ashrdi3_1bit): Remove pattern.
10148         (lshrdi3): Remove shift by 1 expansion.
10149         (arm_lshrdi3_1bit): Remove pattern.
10150         * config/arm/arm.c (arm_rtx_costs_internal): Slightly increase
10151         cost of ashldi3 by 1.
10152         * config/arm/neon.md (ashldi3_neon): Remove shift by 1 expansion.
10153         (<shift>di3_neon): Likewise.
10155 2017-10-30  Dominik Infuehr  <dominik.infuehr@theobroma-systems.com>
10157         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov): Rename
10158         both identically named patterns to (*aarch64_simd_mov<VD:mode>)
10159         and (*aarch64_simd_mov<VQ:mode>).
10160         (*aarch64_simd_mov<VD:mode>): Change type attribute to match
10161         pattern alternative.
10162         (*aarch64_simd_mov<VQ:mode>): Re-order and change type
10163         attributes to match pattern alternative.
10165 2017-10-30  Steven Munroe  <munroesj@gcc.gnu.org>
10167         * config.gcc (powerpc*-*-*): Add emmintrin.h.
10168         * config/rs6000/emmintrin.h: New file.
10169         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include emmintrin.h.
10171 2017-10-30  Wilco Dijkstra  <wdijkstr@arm.com>
10173         * config/arm/vfp.md (movdi_vfp): Merge changes from movdi_vfp_cortexa8.
10174         * (movdi_vfp_cortexa8): Remove pattern.
10176 2017-10-30  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10178         * doc/install.texi (Specific, alpha*-*-*): Remove DEC OSF/1
10179         etc. reference.
10180         (Specific, alpha*-dec-osf5.1): Remove.
10181         (Specific, mips-sgi-irix5): Remove.
10182         (Specific, mips-sgi-irix6): Remove.
10184 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10186         PR middle-end/22141
10187         * gimple-ssa-store-merging.c (merged_store_group::apply_stores): Fix
10188         arguments to clear_bit_region_be.
10190 2017-10-30  Jim Wilson  <wilson@tuliptree.org>
10192         * gimplify.c: Include memmodel.h.
10194 2017-10-30  Martin Jambor  <mjambor@suse.cz>
10196         * omp-grid.c (grid_attempt_target_gridification): Also insert a
10197         condition whether loop should be executed at all.
10199 2017-10-30  Will Schmidt  <will_schmidt@vnet.ibm.com>
10201         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add support for
10202         gimple folding of vec_madd() intrinsics.
10203         * config/rs6000/altivec.md (mulv8hi3): Rename altivec_vmladduhm to
10204         fmav8hi4.  (altivec_vmladduhm): Rename to fmav8hi4.
10205         * config/rs6000/rs6000-builtin.def: Rename vmladduhm to fmav8hi4
10207 2017-10-30  Richard Biener  <rguenther@suse.de>
10209         PR tree-optimization/82762
10210         Revert
10211         2017-10-23  Richard Biener  <rguenther@suse.de>
10213         PR tree-optimization/82129
10214         Revert
10215         2017-08-01  Richard Biener  <rguenther@suse.de>
10217         PR tree-optimization/81181
10218         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
10219         (compute_antic): ... end of iteration here.
10221 2017-10-30  Joseph Myers  <joseph@codesourcery.com>
10223         * doc/invoke.texi (C Dialect Options): Document -std=c17,
10224         -std=iso9899:2017 and -std=gnu17.
10225         * doc/standards.texi (C Language): Document C17 support.
10226         * doc/cpp.texi (Overview): Mention -std=c17.
10227         (Standard Predefined Macros): Document C11 and C17 values of
10228         __STDC_VERSION__.  Do not refer to C99 support as incomplete.
10229         * doc/extend.texi (Inline): Do not list individual options for
10230         standards newer than C99.
10231         * dwarf2out.c (highest_c_language, gen_compile_unit_die): Handle
10232         "GNU C17".
10233         * config/rl78/rl78.c (rl78_option_override): Handle "GNU C17"
10234         language name.
10236 2017-10-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
10238         * asan.c (asan_finish_file): Align asan globals array by shadow
10239         granularity.
10241 2017-10-30  Jakub Jelinek  <jakub@redhat.com>
10243         PR middle-end/22141
10244         * gimple-ssa-store-merging.c: Include rtl.h and expr.h.
10245         (struct store_immediate_info): Add bitregion_start and bitregion_end
10246         fields.
10247         (store_immediate_info::store_immediate_info): Add brs and bre
10248         arguments and initialize bitregion_{start,end} from those.
10249         (struct merged_store_group): Add bitregion_start, bitregion_end,
10250         align_base and mask fields.  Drop unnecessary struct keyword from
10251         struct store_immediate_info.  Add do_merge method.
10252         (clear_bit_region_be): Use memset instead of loop storing zeros.
10253         (merged_store_group::do_merge): New method.
10254         (merged_store_group::merge_into): Use do_merge.  Allow gaps in between
10255         stores as long as the surrounding bitregions have no gaps.
10256         (merged_store_group::merge_overlapping): Use do_merge.
10257         (merged_store_group::apply_stores): Test that bitregion_{start,end}
10258         is byte aligned, rather than requiring that start and width are
10259         byte aligned.  Drop unnecessary struct keyword from
10260         struct store_immediate_info.  Allocate and populate also mask array.
10261         Make start of the arrays relative to bitregion_start rather than
10262         start and size them according to bitregion_{end,start} difference.
10263         (struct imm_store_chain_info): Drop unnecessary struct keyword from
10264         struct store_immediate_info.
10265         (pass_store_merging::gate): Punt if BITS_PER_UNIT or CHAR_BIT is not 8.
10266         (pass_store_merging::terminate_all_aliasing_chains): Drop unnecessary
10267         struct keyword from struct store_immediate_info.
10268         (imm_store_chain_info::coalesce_immediate_stores): Allow gaps in
10269         between stores as long as the surrounding bitregions have no gaps.
10270         Formatting fixes.
10271         (struct split_store): Add orig non-static data member.
10272         (split_store::split_store): Initialize orig to false.
10273         (find_constituent_stmts): Return store_immediate_info *, non-NULL
10274         if there is exactly a single original stmt.  Change stmts argument
10275         to pointer from reference, if NULL, don't push anything to it.  Add
10276         first argument, use it to optimize skipping over orig stmts that
10277         are known to be before bitpos already.  Simplify.
10278         (split_group): Return unsigned int count how many stores are or
10279         would be needed rather than a bool.  Add allow_unaligned argument.
10280         Change split_stores argument from reference to pointer, if NULL,
10281         only do a dry run computing how many stores would be produced.
10282         Rewritten algorithm to use both alignment and misalign if
10283         !allow_unaligned and handle bitfield stores with gaps.
10284         (imm_store_chain_info::output_merged_store): Set start_byte_pos
10285         from bitregion_start instead of start.  Compute allow_unaligned
10286         here, if true, do 2 split_group dry runs to compute which one
10287         produces fewer stores and prefer aligned if equal.  Punt if
10288         new count is bigger or equal than original before emitting any
10289         statements, rather than during that.  Remove no longer needed
10290         new_ssa_names tracking.  Replace num_stmts with
10291         split_stores.length ().  Use 32-bit stack allocated entries
10292         in split_stores auto_vec.  Try to reuse original store lhs/rhs1
10293         if possible.  Handle bitfields with gaps.
10294         (pass_store_merging::execute): Ignore bitsize == 0 stores.
10295         Compute bitregion_{start,end} for the stores and construct
10296         store_immediate_info with that.  Formatting fixes.
10298 2017-10-30  Uros Bizjak  <ubizjak@gmail.com>
10300         PR target/82725
10301         * config/i386/i386.c (legitimate_pic_address_disp_p): Allow
10302         UNSPEC_DTPOFF and UNSPEC_NTPOFF with SImode immediate offset.
10304 2017-10-29  Jim Wilson  <wilson@tuliptree.org>
10306         * gimplify.c: Include tm_p.h.
10308         * common.opt (gcoff): Re-add as ignored option.
10309         (gcoff1, gcoff2, gcoff3): Likewise.
10311         * Makefile.in (OBJS): Delete sdbout.o.
10312         (GTFILES): Delete $(srcdir)/sdbout.c.
10313         * debug.h: Delete sdb_debug_hooks.
10314         * final.c: Delete sdbout.h include.
10315         (final_scan_insn): Delete SDB_DEBUG check.
10316         (rest_of_clean_state): Likewise.
10317         * output.h: Delete sdb_begin_function_line.
10318         * sdbout.c: Delete.
10319         * sdbout.h: Delete.
10320         * toplev.c: Delete sdbout.h include.
10321         (process_options): Delete SDB_DEBUG check.
10322         * tree-core.h (tree_type_common): Delete pointer field of
10323         tree_type_symtab.
10324         * tree.c (copy_node): Clear TYPE_SYMTAB_DIE instead of
10325         TYPE_SYMTAB_POINTER.
10326         * tree.h (TYPE_SYMTAB_POINTER): Delete.
10327         (TYPE_SYMTAB_IS_POINTER): Delete.
10328         (TYPE_SYMTAB_IS_DIE): Renumber.
10329         * xcoffout.c: Refer to former sdbout.c file.
10330         (xcoffout_begin_prologue): Use past tense for sdbout.c reference.
10332         * doc/install.texi (--with-stabs): Delete COFF and ECOFF info.
10333         * doc/invoke.texi (SEEALSO): Delete adb and sdb references.
10334         (Debugging Options): Delete -gcoff.
10335         (-gstabs): Delete SDB reference.
10336         (-gcoff): Delete.
10337         (-gcoff@var{level}): Delete.
10338         * doc/passes.texi (Debugging information output): Delete SDB and
10339         sdbout.c references.
10340         * doc/tm.texi: Regenerate.
10341         * doc/tm.texi.in (DWARF_CIE_DATA_ALIGNMENT): Delete SDB from xref.
10342         (SDB and DWARF): Change node name to DWARF and delete SDB and COFF
10343         references.
10344         (DEBUGGER_AUTO_OFFSET): Delete COFF and SDB references.
10345         (PREFERRED_DEBUGGING_TYPE): Delete SDB_DEBUG and -gcoff references.
10346         (SDB_DEBUGGING_INFO): Delete.
10347         (PUT_SDB_@dots{}, SDB_DELIM, SDB_ALLOW_UNKNOWN_REFERENCES)
10348         SDB_ALLOW_FORWARD_REFERENCES, SDB_OUTPUT_SOURCE_LINE): Delete.
10349         * target.def (output_source_filename): Delete COFF reference.
10351         * common.opt (gcoff): Delete.
10352         (gxcoff+): Update Negative chain.
10353         * defaults.h: Delete all references to SDB_DEBUGGING_INFO and
10354         SDB_DEBUG.
10355         * dwarf2out.c (gen_array_type_die): Change SDB to debuggers.
10356         * flag-types.h (enum debug_info_type): Delete SDB_DEBUG.
10357         * function.c (number_blocks): Delete SDB_DEBUGGING_INFO, SDB_DEBUG,
10358         and SDB references.
10359         (expand_function_start): Change sdb reference to past tense.
10360         (expand_function_end): Change sdb reference to past tense.
10361         * gcc.c (cpp_unique_options): Delete gcoff3 reference.
10362         * opts.c (debug_type_names): Delete coff entry.
10363         (common_handle_option): Delete OPT_gcoff case.
10364         * system.h (SDB_DEBUG, SDB_DEBUGGING_INFO): Poison.
10366         * config/dbxcoff.h (PREFERRED_DEBUGGING_TYPE): Set to DBX_DEBUG.
10367         * config/cris/cris.h: Delete SDB reference in comment.
10368         * config/i386/cygming.h: Don't define SDB_DEBUGGING_INFO.
10369         (ASM_DECLARE_FUNCTION_NAME): Delete SDB reference from comment.
10370         * config/i386/gas.h: Don't define SDB_DEBUGGING_INFO.
10371         * config/i386/i386.c (svr4_dbx_register_map): Change SDB references
10372         to past tense.
10373         (ix86_expand_prologue): Likewise.
10374         * config/i386/winnt.c (i386_pe_start_function): Don't check SDB_DEBUG.
10375         * config/ia64/ia64.h: Likewise.
10376         * config/m68k/m68kelf.h (DBX_REGISTER_NUMBER): Delete SDB reference.
10377         * config/mips/mips.h (SUBTARGET_ASM_DEBUGGING_SPEC): Delete gcoff*
10378         support.
10379         * config/mmix/mmix.h: Likewise.
10380         * config/nds32/nds32.c: Likewise.
10381         * config/stormy/storym16.h: Likewise.
10382         * config/visium/visium.h: Likewise.
10383         * config/vx-common.h (SDB_DEBUGGING_INFO): Delete undef.
10385 2017-10-28  Sandra Loosemore  <sandra@codesourcery.com>
10387         * config/nios2/nios2.h (FRAME_GROWS_DOWNWARD): Define to 1.
10388         * config/nios2/nios2.c (nios2_initial_elimination_offset):  Make
10389         FRAME_POINTER_REGNUM point at high end of local var area.
10391 2017-10-27  Eric Botcazou  <ebotcazou@adacore.com>
10393         * bb-reorder.c (find_traces_1_round): Fix off-by-one index.
10394         Move comment around.  Do not reset best_edge for a copiable
10395         destination if the copy would cause a partition change.
10396         (better_edge_p): Remove redundant check.
10398 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10400         * config/i386/i386-protos.h (ix86_fp_compare_mode): Remove prototype.
10402 2017-10-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
10404         * builtins.c (CASE_MATHFN_FLOATN): New helper macro to add cases
10405         for math functions that have _Float<N> and _Float<N>X variants.
10406         (mathfn_built_in_2): Add support for math functions that have
10407         _Float<N> and _Float<N>X variants.
10408         (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro.
10409         (expand_builtin_mathfn_ternary): Add support for fma with
10410         _Float<N> and _Float<N>X variants.
10411         (expand_builtin): Likewise.
10412         (fold_builtin_3): Likewise.
10413         * builtins.def (DEF_EXT_LIB_FLOATN_NX_BUILTINS): New macro to
10414         create math function _Float<N> and _Float<N>X variants as external
10415         library builtins.
10416         (BUILT_IN_COPYSIGN _Float<N> and _Float<N>X variants) Use
10417         DEF_EXT_LIB_FLOATN_NX_BUILTINS to make built-in functions using
10418         the __builtin_ prefix and if not strict ansi, without the prefix.
10419         (BUILT_IN_FABS _Float<N> and _Float<N>X variants): Likewise.
10420         (BUILT_IN_FMA _Float<N> and _Float<N>X variants): Likewise.
10421         (BUILT_IN_FMAX _Float<N> and _Float<N>X variants): Likewise.
10422         (BUILT_IN_FMIN _Float<N> and _Float<N>X variants): Likewise.
10423         (BUILT_IN_NAN _Float<N> and _Float<N>X variants): Likewise.
10424         (BUILT_IN_SQRT _Float<N> and _Float<N>X variants): Likewise.
10425         * builtin-types.def (BT_FN_FLOAT16_FLOAT16_FLOAT16_FLOAT16): New
10426         function signatures for fma _Float<N> and _Float<N>X variants.
10427         (BT_FN_FLOAT32_FLOAT32_FLOAT32_FLOAT32): Likewise.
10428         (BT_FN_FLOAT64_FLOAT64_FLOAT64_FLOAT64): Likewise.
10429         (BT_FN_FLOAT128_FLOAT128_FLOAT128_FLOAT128): Likewise.
10430         (BT_FN_FLOAT32X_FLOAT32X_FLOAT32X_FLOAT32X): Likewise.
10431         (BT_FN_FLOAT64X_FLOAT64X_FLOAT64X_FLOAT64X): Likewise.
10432         (BT_FN_FLOAT128X_FLOAT128X_FLOAT128X_FLOAT128X): Likewise.
10433         * gencfn-macros.c (print_case_cfn): Add support for math functions
10434         that have _Float<N> and _Float<N>X variants.
10435         (print_define_operator_list): Likewise.
10436         (fltfn_suffixes): Likewise.
10437         (main): Likewise.
10438         * internal-fn.def (DEF_INTERNAL_FLT_FLOATN_FN): New helper macro
10439         for math functions that have _Float<N> and _Float<N>X variants.
10440         (SQRT): Add support for sqrt, copysign, fmin and fmax _Float<N>
10441         and _Float<N>X variants.
10442         (COPYSIGN): Likewise.
10443         (FMIN): Likewise.
10444         (FMAX): Likewise.
10445         * fold-const.c (tree_call_nonnegative_warnv_p): Add support for
10446         copysign, fma, fmax, fmin, and sqrt _Float<N> and _Float<N>X
10447         variants.
10448         (integer_valued_read_call_p): Likewise.
10449         * fold-const-call.c (fold_const_call_ss): Likewise.
10450         (fold_const_call_sss): Add support for copysign, fmin, and fmax
10451         _Float<N> and _Float<N>X variants.
10452         (fold_const_call_ssss): Add support for fma _Float<N> and
10453         _Float<N>X variants.
10454         * gimple-ssa-backprop.c (backprop::process_builtin_call_use): Add
10455         support for copysign and fma _Float<N> and _Float<N>X variants.
10456         (backprop::process_builtin_call_use): Likewise.
10457         * tree-call-cdce.c (can_test_argument_range); Add support for
10458         sqrt _Float<N> and _Float<N>X variants.
10459         (edom_only_function): Likewise.
10460         (get_no_error_domain): Likewise.
10461         * tree-ssa-math-opts.c (internal_fn_reciprocal): Likewise.
10462         * tree-ssa-reassoc.c (attempt_builtin_copysign): Add support for
10463         copysign _Float<N> and _Float<N>X variants.
10464         * config/rs6000/rs6000-builtin.def (SQRTF128): Delete, this is now
10465         handled by machine independent code.
10466         (FMAF128): Likewise.
10467         * doc/cpp.texi (Common Predefined Macros): Document defining
10468         __FP_FAST_FMAF<N> and __FP_FAST_FMAF<N>X if the backend supports
10469         fma _Float<N> and _Float<N>X variants.
10471 2017-10-27  Uros Bizjak  <ubizjak@gmail.com>
10473         PR target/82692
10474         * config/i386/i386-modes.def (CCFPU): Remove definition.
10475         * config/i386/i386.c (put_condition_mode): Remove CCFPU mode handling.
10476         (ix86_cc_modes_compatible): Ditto.
10477         (ix86_expand_carry_flag_compare): Ditto.
10478         (ix86_expand_int_movcc): Ditto.
10479         (ix86_expand_int_addcc): Ditto.
10480         (ix86_reverse_condition): Ditto.
10481         (ix86_unordered_fp_compare): Rename from ix86_fp_compare_mode.
10482         Return true/false for unordered/ordered fp comparisons.
10483         (ix86_cc_mode): Always return CCFPmode for float mode comparisons.
10484         (ix86_prepare_fp_compare_args): Update for rename.
10485         (ix86_expand_fp_compare): Update for rename.  Generate unordered
10486         compare RTXes wrapped with UNSPEC_NOTRAP unspec.
10487         (ix86_expand_sse_compare_and_jump): Ditto.
10488         * config/i386/predicates.md (fcmov_comparison_operator):
10489         Remove CCFPU mode handling.
10490         (ix86_comparison_operator): Ditto.
10491         (ix86_carry_flag_operator): Ditto.
10492         * config/i386/i386.md (UNSPEC_NOTRAP): New unspec.
10493         (*cmpu<mode>_i387): Wrap compare RTX with UNSPEC_NOTRAP unspec.
10494         (*cmpu<mode>_cc_i387): Ditto.
10495         (FPCMP): Remove mode iterator.
10496         (unord): Remove mode attribute.
10497         (unord_subst): New define_subst transformation
10498         (unord): New define_subst attribute.
10499         (unordered): Ditto.
10500         (*cmpi<unord><MODEF:mode>): Rewrite using unord_subst transformation.
10501         (*cmpi<unord>xf_i387): Ditto.
10502         * config/i386/sse.md (<sse>_<unord>comi<round_saeonly_name>): Merge
10503         from <sse>_comi<round_saeonly_name> and <sse>_ucomi<round_saeonly_name>
10504         using unord_subst transformation.
10505         * config/i386/subst.md (SUBST_A): Remove CCFP and CCFPU modes.
10506         (round_saeonly): Also handle CCFP mode.
10507         * reg-stack.c (subst_stack_regs_pat): Handle UNSPEC_NOTRAP unspec.
10508         Remove UNSPEC_SAHF unspec handling.
10510 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10512         * x86-tune.def (X86_TUNE_INTER_UNIT_MOVES_TO_VEC): Disable for Zen.
10514 2017-10-27  Jeff Law  <law@redhat.com>
10516         * gimple-ssa-sprintf.c: Include domwalk.h.
10517         (class sprintf_dom_walker): New class, derived from dom_walker.
10518         (sprintf_dom_walker::before_dom_children): New function.
10519         (struct call_info): Moved into sprintf_dom_walker class
10520         (compute_formath_length, handle_gimple_call): Likewise.
10521         (sprintf_length::execute): Call the dominator walker rather
10522         than walking the statements.
10524         * tree-vrp.c (check_all_array_refs): Do not use wi->info to smuggle
10525         gimple statement locations.
10526         (check_array_bounds): Corresponding changes.  Get the statement's
10527         location directly from wi->stmt.
10529 2017-10-27  Palmer Dabbelt  <palmer@dabbelt.com>
10531         PR target/82717
10532         * doc/invoke.texi (RISC-V) <-mabi>: Correct and improve.
10534 2017-10-27  Jan Hubicka  <hubicka@ucw.cz>
10536         * config/i386/x86-tune.def (X86_TUNE_PARTIAL_REG_DEPENDENCY,
10537         X86_TUNE_MOVX): Disable for Haswell and newer CPUs.
10539 2017-10-27  Jakub Jelinek  <jakub@redhat.com>
10541         PR target/82703
10542         * config/i386/i386-protos.h (maybe_get_pool_constant): Removed.
10543         * config/i386/i386.c (maybe_get_pool_constant): Removed.
10544         (ix86_split_to_parts): Use avoid_constant_pool_reference instead of
10545         maybe_get_pool_constant.
10546         * config/i386/predicates.md (zero_extended_scalar_load_operand):
10547         Likewise.
10549 2017-10-27  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10551         * doc/install.texi (Specific, i?86-*-solaris2.10): Simplify gas
10552         2.26 caveat.  Update gas and gld versions.
10553         (Specific, *-*-solaris2*): Update binutils version.  Remove caveat
10554         reference.
10556 2017-10-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
10558         * cgraph.h (set_malloc_flag): Declare.
10559         * cgraph.c (set_malloc_flag_1): New function.
10560         (set_malloc_flag): Likewise.
10561         * ipa-fnsummary.h (ipa_call_summary): Add new field is_return_callee.
10562         * ipa-fnsummary.c (ipa_call_summary::reset): Set is_return_callee to
10563         false.
10564         (read_ipa_call_summary): Add support for reading is_return_callee.
10565         (write_ipa_call_summary): Stream is_return_callee.
10566         * ipa-inline.c (ipa_inline): Remove call to ipa_free_fn_summary.
10567         * ipa-pure-const.c: Add headers ssa.h, alloc-pool.h, symbol-summary.h,
10568         ipa-prop.h, ipa-fnsummary.h.
10569         (pure_const_names): Change to static.
10570         (malloc_state_e): Define.
10571         (malloc_state_names): Define.
10572         (funct_state_d): Add field malloc_state.
10573         (varying_state): Set malloc_state to STATE_MALLOC_BOTTOM.
10574         (check_retval_uses): New function.
10575         (malloc_candidate_p): Likewise.
10576         (analyze_function): Add support for malloc attribute.
10577         (pure_const_write_summary): Stream malloc_state.
10578         (pure_const_read_summary): Add support for reading malloc_state.
10579         (dump_malloc_lattice): New function.
10580         (propagate_malloc): New function.
10581         (warn_function_malloc): New function.
10582         (ipa_pure_const::execute): Call propagate_malloc and
10583         ipa_free_fn_summary.
10584         (pass_local_pure_const::execute): Add support for malloc attribute.
10585         * ssa-iterators.h (RETURN_FROM_IMM_USE_STMT): New macro.
10586         * doc/invoke.texi: Document Wsuggest-attribute=malloc.
10588 2017-10-27  Martin Liska  <mliska@suse.cz>
10590         PR gcov-profile/82457
10591         * doc/invoke.texi: Document that one needs a non-strict ISO mode
10592         for fork-like functions to be properly instrumented.
10594 2017-10-27  Richard Biener  <rguenther@suse.de>
10596         PR middle-end/81659
10597         * tree-eh.c (pass_lower_eh_dispatch::execute): Free dominator
10598         info when we redirected EH.
10600 2017-10-26  Michael Collison  <michael.collison@arm.com>
10602         * config/aarch64/aarch64.md(<optab>_trunc><vf><GPI:mode>2):
10603         New pattern.
10604         (<optab>_trunchf<GPI:mode>2: New pattern.
10605         (<optab>_trunc<vgp><GPI:mode>2: New pattern.
10606         * config/aarch64/iterators.md (wv): New mode attribute.
10607         (vf, VF): New mode attributes.
10608         (vgp, VGP): New mode attributes.
10609         (s): Update attribute with SImode and DImode prefixes.
10611 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10613         * config/nios2/constraints.md ("S"): Match r0rel_constant_p too.
10614         * config/nios2/nios2-protos.h (r0rel_constant_p): Declare.
10615         * config/nios2/nios2.c: (nios2_r0rel_sec_regex): New.
10616         (nios2_option_overide): Initialize it.  Don't allow R0-relative
10617         addressing with PIC.
10618         (nios2_rtx_costs): Handle r0rel_constant_p like gprel_constant_p.
10619         (nios2_symbolic_constant_p): Likewise.
10620         (nios2_legitimate_address_p): Likewise.
10621         (nios2_r0rel_section_name_p): New.
10622         (nios2_symbol_ref_in_r0rel_data_p): New.
10623         (nios2_emit_move_sequence): Handle r0rel_constant_p.
10624         (r0rel_constant_p): New.
10625         (nios2_print_operand_address): Handle r0rel_constant_p.
10626         (nios2_cdx_narrow_form_p): Likewise.
10627         * config/nios2/nios2.opt (mr0rel-sec=): New option.
10628         * doc/invoke.texi (Option Summary): Add -mr0rel-sec.
10629         (Nios II Options): Document -mr0rel-sec.
10631 2017-10-26  Sandra Loosemore  <sandra@codesourcery.com>
10633         * config/nios2/nios2.c: Include xregex.h.
10634         (nios2_gprel_sec_regex): New.
10635         (nios2_option_overide): Initialize it.  Don't allow GP-relative
10636         addressing with PIC.
10637         (nios2_small_section_name_p): Check for regex match.
10638         * config/nios2/nios2.opt (mgprel-sec=): New option.
10639         * doc/invoke.texi (Option Summary): Add -mgprel-sec.
10640         (Nios II Options): Document -mgprel-sec.
10642 2017-10-26  Jim Wilson  <wilson@tuliptree.org>
10644         * doc/invoke.texi (-fdebug-prefix-map): Expand documentation.
10646 2017-10-26  Tom de Vries  <tom@codesourcery.com>
10648         PR tree-optimization/82707
10649         * gimple.c (gimple_copy): Fix unsharing of
10650         GIMPLE_OMP_{SINGLE,TARGET,TEAMS}.
10652 2017-10-26  Olga Makhotina  <olga.makhotina@intel.com>
10654         * config/i386/avx512fintrin.h (_mm512_cmpeq_pd_mask,
10655         _mm512_cmple_pd_mask, _mm512_cmplt_pd_mask,
10656         _mm512_cmpneq_pd_mask, _mm512_cmpnle_pd_mask,
10657         _mm512_cmpnlt_pd_mask, _mm512_cmpord_pd_mask,
10658         _mm512_cmpunord_pd_mask, _mm512_mask_cmpeq_pd_mask,
10659         _mm512_mask_cmple_pd_mask, _mm512_mask_cmplt_pd_mask,
10660         _mm512_mask_cmpneq_pd_mask, _mm512_mask_cmpnle_pd_mask,
10661         _mm512_mask_cmpnlt_pd_mask, _mm512_mask_cmpord_pd_mask,
10662         _mm512_mask_cmpunord_pd_mask, _mm512_cmpeq_ps_mask,
10663         _mm512_cmple_ps_mask, _mm512_cmplt_ps_mask,
10664         _mm512_cmpneq_ps_mask, _mm512_cmpnle_ps_mask,
10665         _mm512_cmpnlt_ps_mask, _mm512_cmpord_ps_mask,
10666         _mm512_cmpunord_ps_mask, _mm512_mask_cmpeq_ps_mask,
10667         _mm512_mask_cmple_ps_mask, _mm512_mask_cmplt_ps_mask,
10668         _mm512_mask_cmpneq_ps_mask, _mm512_mask_cmpnle_ps_mask,
10669         _mm512_mask_cmpnlt_ps_mask, _mm512_mask_cmpord_ps_mask,
10670         _mm512_mask_cmpunord_ps_mask): New intrinsics.
10672 2017-10-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
10674         * config/rs6000/aix.h (TARGET_IEEEQUAD_DEFAULT): Set long double
10675         default to IBM.
10676         * config/rs6000/darwin.h (TARGET_IEEEQUAD_DEFAULT): Likewise.
10677         * config/rs6000/rs6000.opt (-mabi=ieeelongdouble): Move the
10678         warning to rs6000.c.  Remove the Undocumented flag, since it has
10679         been documented.
10680         (-mabi=ibmlongdouble): Likewise.
10681         * config/rs6000/rs6000.c (TARGET_IEEEQUAD_DEFAULT): If it is not
10682         already set, set the default format for long double.
10683         (rs6000_debug_reg_global): Print whether long double is IBM or
10684         IEEE.
10685         (rs6000_option_override_internal): Rework setting long double
10686         format.  Only warn if the user is changing the long double default
10687         and they did not use -Wno-psabi.
10688         * doc/invoke.texi (PowerPC options): Update the documentation for
10689         -mabi=ieeelongdouble and -mabi=ibmlongdouble.
10691 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10692             Alan Hayward  <alan.hayward@arm.com>
10693             David Sherwood  <david.sherwood@arm.com>
10695         * rtl.h (wider_subreg_mode): New function.
10696         * ira.h (ira_sort_regnos_for_alter_reg): Take a machine_mode *
10697         rather than an unsigned int *.
10698         * ira-color.c (regno_max_ref_width): Replace with...
10699         (regno_max_ref_mode): ...this new variable.
10700         (coalesced_pseudo_reg_slot_compare): Update accordingly.
10701         Use wider_subreg_mode.
10702         (ira_sort_regnos_for_alter_reg): Likewise.  Take a machine_mode *
10703         rather than an unsigned int *.
10704         * lra-constraints.c (uses_hard_regs_p): Use wider_subreg_mode.
10705         (process_alt_operands): Likewise.
10706         (invariant_p): Likewise.
10707         * lra-spills.c (assign_mem_slot): Likewise.
10708         (add_pseudo_to_slot): Likewise.
10709         * lra.c (collect_non_operand_hard_regs): Likewise.
10710         (add_regs_to_insn_regno_info): Likewise.
10711         * reload1.c (regno_max_ref_width): Replace with...
10712         (regno_max_ref_mode): ...this new variable.
10713         (reload): Update accordingly.  Update call to
10714         ira_sort_regnos_for_alter_reg.
10715         (alter_reg): Update to use regno_max_ref_mode.  Call wider_subreg_mode.
10716         (init_eliminable_invariants): Update to use regno_max_ref_mode.
10717         (scan_paradoxical_subregs): Likewise.
10719 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10721         * config/aarch64/aarch64.h (EXIT_IGNORE_STACK): Set if alloca is used.
10722         (aarch64_frame): Add emit_frame_chain boolean.
10723         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10724         Move eh_return case to aarch64_layout_frame.
10725         (aarch64_layout_frame): Initialize emit_frame_chain.
10726         (aarch64_expand_prologue): Use emit_frame_chain.
10728 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10730         * config/aarch64/aarch64.c (aarch64_layout_frame):
10731         Ensure LR is always stored at the bottom of the callee-saves.
10732         Remove rarely used frame layout which saves callee-saves at top of
10733         frame, so the store of LR can be used as a valid probe in all cases.
10735 2017-10-26  Wilco Dijkstra  <wdijkstr@arm.com>
10737         * config/aarch64/aarch64.c (aarch64_legitimize_address_displacement):
10738         Improve unaligned TImode/TFmode base/offset split.
10740 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10741             Alan Hayward  <alan.hayward@arm.com>
10742             David Sherwood  <david.sherwood@arm.com>
10744         * caller-save.c (mark_referenced_regs):  Use read_modify_subreg_p.
10745         * combine.c (find_single_use_1): Likewise.
10746         (expand_field_assignment): Likewise.
10747         (move_deaths): Likewise.
10748         * lra-constraints.c (simplify_operand_subreg): Likewise.
10749         (curr_insn_transform): Likewise.
10750         * lra.c (collect_non_operand_hard_regs): Likewise.
10751         (add_regs_to_insn_regno_info): Likewise.
10752         * rtlanal.c (reg_referenced_p): Likewise.
10753         (covers_regno_no_parallel_p): Likewise.
10755 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10757         * wide-int-print.cc (print_hex): Loop based on extract_uhwi.
10758         Don't print any bits outside the precision of the value.
10759         * wide-int.cc (test_printing): Add some new tests.
10761 2017-10-26  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
10763         * configure.ac (gcc_cv_as_ix86_xbrace_comment): Check if assembler
10764         supports -xbrace_comment option.
10765         * configure: Regenerate.
10766         * config.in: Regenerate.
10767         * config/i386/sol2.h (ASM_XBRACE_COMMENT_SPEC): Define.
10768         (ASM_CPU_SPEC): Use it.
10770 2017-10-26  Richard Sandiford  <richard.sandiford@linaro.org>
10772         * target.def (static_rtx_alignment): New hook.
10773         * targhooks.h (default_static_rtx_alignment): Declare.
10774         * targhooks.c (default_static_rtx_alignment): New function.
10775         * doc/tm.texi.in (TARGET_STATIC_RTX_ALIGNMENT): New hook.
10776         * doc/tm.texi: Regenerate.
10777         * varasm.c (force_const_mem): Use targetm.static_rtx_alignment
10778         instead of targetm.constant_alignment.  Remove call to
10779         set_mem_attributes.
10780         * config/cris/cris.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10781         (cris_preferred_mininum_alignment): New function, split out from...
10782         (cris_constant_alignment): ...here.
10783         (cris_static_rtx_alignment): New function.
10784         * config/i386/i386.c (ix86_static_rtx_alignment): New function,
10785         split out from...
10786         (ix86_constant_alignment): ...here.
10787         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10788         * config/mmix/mmix.c (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10789         (mmix_static_rtx_alignment): New function.
10790         * config/spu/spu.c (spu_static_rtx_alignment): New function.
10791         (TARGET_STATIC_RTX_ALIGNMENT): Redefine.
10793 2017-10-26  Tamar Christina  <tamar.christina@arm.com>
10795         PR target/81800
10796         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2):
10797         Add flag_trapping_math and flag_fp_int_builtin_inexact.
10799 2017-10-25  Palmer Dabbelt  <palmer@dabbelt.com>
10801         * config/riscv/riscv.md (ZERO_EXTEND_LOAD): Define.
10802         * config/riscv/pic.md (local_pic_load): Rename to local_pic_load_s,
10803         mark as a sign-extending load.
10804         (local_pic_load_u): Define.
10806 2017-10-25  Eric Botcazou  <ebotcazou@adacore.com>
10808         PR middle-end/82062
10809         * fold-const.c (operand_equal_for_comparison_p): Also return true
10810         if ARG0 is a simple variant of ARG1 with narrower precision.
10811         (fold_ternary_loc): Always pass unstripped operands to the predicate.
10813 2017-10-25  Jan Hubicka  <hubicka@ucw.cz>
10815         * i386.c (ix86_builtin_vectorization_cost): Compute scatter/gather
10816         cost correctly.
10817         * i386.h (processor_costs): Add gather_static, gather_per_elt,
10818         scatter_static, scatter_per_elt.
10819         * x86-tune-costs.h: Add new cost entries.
10821 2017-10-25  Richard Biener  <rguenther@suse.de>
10823         * tree-ssa-sccvn.h (vn_eliminate): Declare.
10824         * tree-ssa-pre.c (class eliminate_dom_walker, eliminate,
10825         class pass_fre): Move to ...
10826         * tree-ssa-sccvn.c (class eliminate_dom_walker, vn_eliminate,
10827         class pass_fre): ... here and adjust for statistics.
10829 2017-10-25  Jakub Jelinek  <jakub@redhat.com>
10831         PR libstdc++/81706
10832         * attribs.c (attribute_value_equal): Use omp_declare_simd_clauses_equal
10833         for comparison of OMP_CLAUSEs regardless of flag_openmp{,_simd}.
10834         (duplicate_one_attribute, copy_attributes_to_builtin): New functions.
10835         * attribs.h (duplicate_one_attribute, copy_attributes_to_builtin): New
10836         declarations.
10838 2017-10-25  Richard Biener  <rguenther@suse.de>
10840         * tree-ssa-pre.c (need_eh_cleanup, need_ab_cleanup, el_to_remove,
10841         el_to_fixup, el_todo, el_avail, el_avail_stack, eliminate_avail,
10842         eliminate_push_avail, eliminate_insert): Move inside...
10843         (class eliminate_dom_walker): ... this class in preparation
10844         of move.
10845         (fini_eliminate): Remove by merging with ...
10846         (eliminate): ... this function.  Adjust for class changes.
10847         (pass_pre::execute): Remove fini_eliminate call.
10848         (pass_fre::execute): Likewise.
10850 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10852         PR target/82460
10853         * config/i386/sse.md (UNSPEC_VPERMI2, UNSPEC_VPERMI2_MASK): Remove.
10854         (VPERMI2, VPERMI2I): New mode iterators.
10855         (<avx512>_vpermi2var<mode>3_maskz): Remove 3 define_expand patterns.
10856         (<avx512>_vpermi2var<mode>3<sd_maskz_name>): Remove 3 define_insn
10857         patterns.
10858         (<avx512>_vpermi2var<mode>3_mask): New define_expand using VPERMI2
10859         mode iterator.  Remove 3 old define_insn patterns.
10860         (*<avx512>_vpermi2var<mode>3_mask): 2 new define_insn patterns.
10861         (<avx512>_vpermt2var<mode>3_maskz): Adjust 1 define_expand to use
10862         VPERMI2 mode iterator, remove the other two expanders.
10863         (<avx512>_vpermt2var<mode>3<sd_maskz_name>): Adjust 1 define_insn
10864         to use VPERMI2 mode iterator, add another alternative for vpermi2*
10865         instructions, remove the other two patterns.
10866         (<avx512>_vpermt2var<mode>3_mask): Adjust 1 define_insn to use VPERMI2
10867         mode iterator, remove the other two patterns.
10868         * config/i386/i386.c (ix86_expand_vec_perm_vpermi2): Renamed to ...
10869         (ix86_expand_vec_perm_vpermt2): ... this.  Swap mask and op0
10870         arguments, use gen_*vpermt2* expanders instead of gen_*vpermi2*
10871         and adjust argument order accordingly.
10872         (ix86_expand_vec_perm): Adjust caller.
10873         (expand_vec_perm_1): Likewise.
10874         (expand_vec_perm_vpermi2_vpshub2): Rename to ...
10875         (expand_vec_perm_vpermt2_vpshub2): ... this.
10876         (ix86_expand_vec_perm_const_1): Adjust caller.
10877         (ix86_vectorize_vec_perm_const_ok): Adjust comments.
10879         PR target/82370
10880         * config/i386/sse.md (VIMAX_AVX2): Remove V4TImode.
10881         (VIMAX_AVX2_AVX512BW, VIMAX_AVX512VL): New mode iterators.
10882         (vec_shl_<mode>): Remove unused expander.
10883         (avx512bw_<shift_insn><mode>3): New define_insn.
10884         (<sse2_avx2>_ashl<mode>3, <sse2_avx2>_lshr<mode>3): Replaced by ...
10885         (<sse2_avx2>_<shift_insn><mode>3): ... this.  New define_insn.
10887 2017-10-24  Paolo Carlini  <paolo.carlini@oracle.com>
10889         PR c++/82466
10890         * doc/invoke.texi ([Wbuiltin-declaration-mismatch]): Extend
10891         description.
10893 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10895         PR rtl-optimization/82396
10896         * gcc/haifa-sched.c (ready_sort_real): Remove qsort workaround.
10897         (autopref_multipass_init): Simplify initialization.
10898         (autopref_rank_data): Simplify sort order.
10899         * gcc/sched-int.h (autopref_multipass_data_): Remove
10900         multi_mem_insn_p, min_offset and max_offset.
10902 2017-10-24  Wilco Dijkstra  <wdijkstr@arm.com>
10904         PR middle-end/60580
10905         * config/aarch64/aarch64.c (aarch64_frame_pointer_required)
10906         Check special value of flag_omit_frame_pointer.
10907         (aarch64_can_eliminate): Likewise.
10908         (aarch64_override_options_after_change_1): Simplify handling of
10909         -fomit-frame-pointer and -fomit-leaf-frame-pointer.
10911 2017-10-24  Richard Biener  <rguenther@suse.de>
10913         PR tree-optimization/82697
10914         * tree-ssa-phiopt.c (cond_store_replacement): Use alias-set
10915         zero for conditional load and unconditional store.
10917 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10919         * doc/install.texi: Document bootstrap-cet.
10921 2017-10-24  H.J. Lu  <hongjiu.lu@intel.com>
10923         PR target/82659
10924         * config/i386/i386.c (rest_of_insert_endbranch): Don't insert
10925         ENDBR instruction at function entrance if function is only
10926         called directly.
10928 2017-10-24  Jakub Jelinek  <jakub@redhat.com>
10930         PR target/82628
10931         * config/i386/i386.md (addcarry<mode>, subborrow<mode>): Change
10932         patterns to better describe from which operation the CF is computed.
10933         (addcarry<mode>_0, subborrow<mode>_0): New patterns.
10934         * config/i386/i386.c (ix86_expand_builtin) <case handlecarry>: Pass
10935         one LTU with [DT]Imode and another one with [SD]Imode.  If arg0
10936         is 0, use _0 suffixed expanders instead of emitting a comparison
10937         before it.
10939 2017-10-06  Sergey Shalnov  <Sergey.Shalnov@intel.com>
10941         * config/i386/i386.md(*movsf_internal, *movdf_internal):
10942         Avoid 512-bit AVX modes for TARGET_PREFER_AVX256.
10944 2017-10-24  Eric Botcazou  <ebotcazou@adacore.com>
10946         PR middle-end/82569
10947         * tree-outof-ssa.h (always_initialized_rtx_for_ssa_name_p): Delete.
10948         * expr.c (expand_expr_real_1) <expand_decl_rtl>: Revert latest change.
10949         * loop-iv.c (iv_get_reaching_def): Likewise.
10950         * cfgexpand.c (expand_one_ssa_partition): Initialize the RTX if the
10951         variable is promoted and the partition contains undefined values.
10953 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10955         * config/nios2/nios2.c (nios2_rtx_costs): Make costs better
10956         reflect reality.
10957         (nios2_address_cost): Define.
10958         (nios2_legitimize_address): Recognize (exp + constant) directly.
10959         (TARGET_ADDRESS_COST): Define.
10961 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10963         * config/nios2/nios2-protos.h (nios2_large_constant_p): Declare.
10964         (nios2_symbolic_memory_operand_p): Declare.
10965         (nios2_split_large_constant): Declare.
10966         (nios2_split_symbolic_memory_operand): Declare.
10967         * config/nios2/nios2.c: Adjust includes.
10968         (nios2_symbolic_constant_allowed): New.
10969         (nios2_symbolic_constant_p): New.
10970         (nios2_plus_symbolic_constant_p): New.
10971         (nios2_valid_addr_expr_p): Recognize addresses involving
10972         symbolic constants.
10973         (nios2_legitimate_address_p): Likewise, also LO_SUM.
10974         (nios2_symbolic_memory_operand_p): New.
10975         (nios2_large_constant_p): New.
10976         (nios2_split_large_constant): New.
10977         (nios2_split_plus_large_constant): New.
10978         (nios2_split_symbolic_memory_operand): New.
10979         (nios2_legitimize_address): Code refactoring.  Handle addresses
10980         involving symbolic constants.
10981         (nios2_emit_move_sequence): Likewise.
10982         (nios2_print_operand): Improve error output.
10983         (nios2_print_operand_address): Handle LO_SUM.
10984         (nios2_cdx_narrow_form_p): Likewise.
10985         * config/nios2/nios2.md (movqi_internal): Add splitter for memory
10986         operands involving symbolic constants.
10987         (movhi_internal, movsi_internal): Likewise.
10988         (zero_extendhisi2, zero_extendqi<mode>2): Likewise.
10989         (extendhisi2, extendqi<mode>2): Likewise.
10991 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10993         * tree-pass.h (PROP_rtl_split_insns): Define.
10994         * recog.c (pass_data_split_all_insns): Provide PROP_rtl_split_insns.
10996 2017-10-23  Sandra Loosemore  <sandra@codesourcery.com>
10998         * config/nios2/nios2.c (TARGET_LRA_P): Don't override.
11000 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11002         PR debug/82630
11003         * target.def (const_not_ok_for_debug_p): Default to
11004         default_const_not_ok_for_debug_p instead of hook_bool_rtx_false.
11005         * targhooks.h (default_const_not_ok_for_debug_p): New declaration.
11006         * targhooks.c (default_const_not_ok_for_debug_p): New function.
11007         * dwarf2out.c (const_ok_for_output_1): Only reject UNSPECs for
11008         which targetm.const_not_ok_for_debug_p returned true.
11009         * config/arm/arm.c (arm_const_not_ok_for_debug_p): Return true
11010         for UNSPECs.
11011         * config/powerpcspe/powerpcspe.c (rs6000_const_not_ok_for_debug_p):
11012         Likewise.
11013         * config/rs6000/rs6000.c (rs6000_const_not_ok_for_debug_p): Likewise.
11014         * config/i386/i386.c (ix86_delegitimize_address_1): Don't delegitimize
11015         UNSPEC_GOTOFF with addend into addend - _GLOBAL_OFFSET_TABLE_ + symbol
11016         if !base_term_p.
11017         (ix86_const_not_ok_for_debug_p): New function.
11018         (i386_asm_output_addr_const_extra): Handle UNSPEC_GOTOFF.
11019         (TARGET_CONST_NOT_OK_FOR_DEBUG_P): Redefine.
11021 2017-10-23  David Malcolm  <dmalcolm@redhat.com>
11023         PR bootstrap/82610
11024         * system.h: Conditionally include "unique-ptr.h" if
11025         INCLUDE_UNIQUE_PTR is defined.
11026         * unique-ptr-tests.cc: Remove include of "unique-ptr.h" in favor
11027         of defining INCLUDE_UNIQUE_PTR before including "system.h".
11029 2017-10-23  Sebastian Perta  <sebastian.perta@renesas.com>
11031         * config/rl78/rl78.md: New define_expand "subdi3".
11033 2017-10-23  H.J. Lu  <hongjiu.lu@intel.com>
11035         PR target/82673
11036         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Skip
11037         DF_REF_INSN if DF_REF_INSN_INFO is false.
11039 2017-10-23  Jan Hubicka  <hubicka@ucw.cz>
11041         * i386.c (dimode_scalar_chain::compute_convert_gain): Use
11042         xmm_move instead of sse_move.
11043         (sse_store_index): New function.
11044         (ix86_register_move_cost): Be more sensible about mismatch stall;
11045         model AVX moves correctly; make difference between sse->integer and
11046         integer->sse.
11047         (ix86_builtin_vectorization_cost): Model correctly aligned and unaligned
11048         moves; make difference between SSE and AVX.
11049         * i386.h (processor_costs): Remove sse_move; add xmm_move, ymm_move
11050         and zmm_move. Increase size of sse load and store tables;
11051         add unaligned load and store tables; add ssemmx_to_integer.
11052         * x86-tune-costs.h: Update all entries according to real
11053         move latencies from Agner Fog's manual and chip documentation.
11055 2017-10-23  Jakub Jelinek  <jakub@redhat.com>
11057         PR target/82628
11058         * config/i386/predicates.md (x86_64_dwzext_immediate_operand): New.
11059         * config/i386/constraints.md (Wf): New constraint.
11060         * config/i386/i386.md (UNSPEC_SBB): New unspec.
11061         (cmp<dwi>_doubleword): Removed.
11062         (sub<mode>3_carry_ccc, *sub<mode>3_carry_ccc_1): New patterns.
11063         (sub<mode>3_carry_ccgz): Use unspec instead of compare.
11064         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>: Don't
11065         expand with cmp<dwi>_doubleword.  For LTU and GEU use
11066         sub<mode>3_carry_ccc instead of sub<mode>3_carry_ccgz and use CCCmode.
11068         * common.opt (gcolumn-info): Enable by default.
11069         * doc/invoke.texi (gcolumn-info): Document new default.
11071 2017-10-23  Richard Biener  <rguenther@suse.de>
11073         PR tree-optimization/82672
11074         * graphite-isl-ast-to-gimple.c (graphite_copy_stmts_from_block):
11075         Fold the stmt if we propagated into it.
11077 2017-10-23  Richard Biener  <rguenther@suse.de>
11079         * tree-ssa-pre.c (bitmap_remove_from_set): Rename to...
11080         (bitmap_remove_expr_from_set): ... this.  All callers call this
11081         for non-constant values.
11082         (bitmap_set_subtract): Rename to...
11083         (bitmap_set_subtract_expressions): ... this.  Adjust and
11084         optimize.
11085         (bitmap_set_contains_value): Remove superfluous check.
11086         (bitmap_set_replace_value): Inline into single caller ...
11087         (bitmap_value_replace_in_set): ... here and simplify.
11088         (dependent_clean): Merge into ...
11089         (clean): ... this using an overload.  Adjust.
11090         (prune_clobbered_mems): Adjust.
11091         (compute_antic_aux): Likewise.
11092         (compute_partial_antic_aux): Likewise.
11094 2017-10-23  Richard Biener  <rguenther@suse.de>
11096         PR tree-optimization/82129
11097         Revert
11098         2017-08-01  Richard Biener  <rguenther@suse.de>
11100         PR tree-optimization/81181
11101         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
11102         (compute_antic): ... end of iteration here.
11104 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11106         * target.def (starting_frame_offset): New hook.
11107         * doc/tm.texi (STARTING_FRAME_OFFSET): Remove in favor of...
11108         (TARGET_STARTING_FRAME_OFFSET): ...this new hook.
11109         * doc/tm.texi.in: Regenerate.
11110         * hooks.h (hook_hwi_void_0): Declare.
11111         * hooks.c (hook_hwi_void_0): New function.
11112         * doc/rtl.texi: Refer to TARGET_STARTING_FRAME_OFFSET instead of
11113         STARTING_FRAME_OFFSET.
11114         * builtins.c (expand_builtin_setjmp_receiver): Likewise.
11115         * reload1.c (reload): Likewise.
11116         * cfgexpand.c (expand_used_vars): Use targetm.starting_frame_offset
11117         instead of STARTING_FRAME_OFFSET.
11118         * function.c (try_fit_stack_local): Likewise.
11119         (assign_stack_local_1): Likewise
11120         (instantiate_virtual_regs): Likewise.
11121         * rtlanal.c (rtx_addr_can_trap_p_1): Likewise.
11122         * config/avr/avr.md (nonlocal_goto_receiver): Likewise.
11123         * config/aarch64/aarch64.h (STARTING_FRAME_OFFSET): Delete.
11124         * config/alpha/alpha.h (STARTING_FRAME_OFFSET): Likewise.
11125         * config/arc/arc.h (STARTING_FRAME_OFFSET): Likewise.
11126         * config/arm/arm.h (STARTING_FRAME_OFFSET): Likewise.
11127         * config/bfin/bfin.h (STARTING_FRAME_OFFSET): Likewise.
11128         * config/c6x/c6x.h (STARTING_FRAME_OFFSET): Likewise.
11129         * config/cr16/cr16.h (STARTING_FRAME_OFFSET): Likewise.
11130         * config/cris/cris.h (STARTING_FRAME_OFFSET): Likewise.
11131         * config/fr30/fr30.h (STARTING_FRAME_OFFSET): Likewise.
11132         * config/frv/frv.h (STARTING_FRAME_OFFSET): Likewise.
11133         * config/ft32/ft32.h (STARTING_FRAME_OFFSET): Likewise.
11134         * config/h8300/h8300.h (STARTING_FRAME_OFFSET): Likewise.
11135         * config/i386/i386.h (STARTING_FRAME_OFFSET): Likewise.
11136         * config/ia64/ia64.h (STARTING_FRAME_OFFSET): Likewise.
11137         * config/m32c/m32c.h (STARTING_FRAME_OFFSET): Likewise.
11138         * config/m68k/m68k.h (STARTING_FRAME_OFFSET): Likewise.
11139         * config/mcore/mcore.h (STARTING_FRAME_OFFSET): Likewise.
11140         * config/mn10300/mn10300.h (STARTING_FRAME_OFFSET): Likewise.
11141         * config/moxie/moxie.h (STARTING_FRAME_OFFSET): Likewise.
11142         * config/msp430/msp430.h (STARTING_FRAME_OFFSET): Likewise.
11143         * config/nds32/nds32.h (STARTING_FRAME_OFFSET): Likewise.
11144         * config/nios2/nios2.h (STARTING_FRAME_OFFSET): Likewise.
11145         * config/nvptx/nvptx.h (STARTING_FRAME_OFFSET): Likewise.
11146         * config/pdp11/pdp11.h (STARTING_FRAME_OFFSET): Likewise.
11147         * config/riscv/riscv.h (STARTING_FRAME_OFFSET): Likewise.
11148         * config/rl78/rl78.h (STARTING_FRAME_OFFSET): Likewise.
11149         * config/rx/rx.h (STARTING_FRAME_OFFSET): Likewise.
11150         * config/s390/s390.h (STARTING_FRAME_OFFSET): Likewise.
11151         * config/sh/sh.h (STARTING_FRAME_OFFSET): Likewise.
11152         * config/sparc/sparc.c (sparc_compute_frame_size): Likewise.
11153         * config/sparc/sparc.h (STARTING_FRAME_OFFSET): Likewise.
11154         * config/spu/spu.h (STARTING_FRAME_OFFSET): Likewise.
11155         * config/stormy16/stormy16.h (STARTING_FRAME_OFFSET): Likewise.
11156         * config/tilegx/tilegx.h (STARTING_FRAME_OFFSET): Likewise.
11157         * config/tilepro/tilepro.h (STARTING_FRAME_OFFSET): Likewise.
11158         * config/v850/v850.h (STARTING_FRAME_OFFSET): Likewise.
11159         * config/visium/visium.h (STARTING_FRAME_OFFSET): Likewise.
11160         * config/avr/avr.h (STARTING_FRAME_OFFSET): Likewise.
11161         * config/avr/avr-protos.h (avr_starting_frame_offset): Likewise.
11162         * config/avr/avr.c (avr_starting_frame_offset): Make static and
11163         return a HOST_WIDE_INT.
11164         (avr_builtin_setjmp_frame_value): Use it instead of
11165         STARTING_FRAME_OFFSET.
11166         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11167         * config/epiphany/epiphany.h (STARTING_FRAME_OFFSET): Delete.
11168         * config/epiphany/epiphany.c (epiphany_starting_frame_offset):
11169         New function.
11170         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11171         * config/iq2000/iq2000.h (STARTING_FRAME_OFFSET): Delete.
11172         * config/iq2000/iq2000.c (iq2000_starting_frame_offset): New function.
11173         (TARGET_CONSTANT_ALIGNMENT): Redefine.
11174         * config/lm32/lm32.h (STARTING_FRAME_OFFSET): Delete.
11175         * config/lm32/lm32.c (lm32_starting_frame_offset): New function.
11176         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11177         * config/m32r/m32r.h (STARTING_FRAME_OFFSET): Delete.
11178         * config/m32r/m32r.c (m32r_starting_frame_offset): New function.
11179         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11180         * config/microblaze/microblaze.h (STARTING_FRAME_OFFSET): Delete.
11181         * config/microblaze/microblaze.c (microblaze_starting_frame_offset):
11182         New function.
11183         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11184         * config/mips/mips.h (STARTING_FRAME_OFFSET): Delete.
11185         * config/mips/mips.c (mips_compute_frame_info): Refer to
11186         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11187         (mips_starting_frame_offset): New function.
11188         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11189         * config/mmix/mmix.h (STARTING_FRAME_OFFSET): Delete.
11190         * config/mmix/mmix-protos.h (mmix_starting_frame_offset): Delete.
11191         * config/mmix/mmix.c (mmix_starting_frame_offset): Make static
11192         and return a HOST_WIDE_INT.
11193         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11194         (mmix_initial_elimination_offset): Refer to
11195         TARGET_STARTING_FRAME_OFFSET instead of STARTING_FRAME_OFFSET.
11196         * config/pa/pa.h (STARTING_FRAME_OFFSET): Delete.
11197         * config/pa/pa.c (pa_starting_frame_offset): New function.
11198         (pa_compute_frame_size): Use it instead of STARTING_FRAME_OFFSET.
11199         (pa_expand_prologue): Likewise.
11200         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11201         * config/powerpcspe/aix.h (STARTING_FRAME_OFFSET): Split out
11202         !FRAME_GROWS_DOWNWARD handling to...
11203         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11204         * config/powerpcspe/darwin.h (STARTING_FRAME_OFFSET): Split out
11205         !FRAME_GROWS_DOWNWARD handling to...
11206         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11207         * config/powerpcspe/powerpcspe.h (STARTING_FRAME_OFFSET): Split out
11208         !FRAME_GROWS_DOWNWARD handling to...
11209         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11210         * config/powerpcspe/powerpcspe.c (TARGET_STARTING_FRAME_OFFSET):
11211         Redefine.
11212         (rs6000_starting_frame_offset): New function.
11213         * config/rs6000/aix.h (STARTING_FRAME_OFFSET): Split out
11214         !FRAME_GROWS_DOWNWARD handling to...
11215         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11216         * config/rs6000/darwin.h (STARTING_FRAME_OFFSET): Split out
11217         !FRAME_GROWS_DOWNWARD handling to...
11218         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11219         * config/rs6000/rs6000.h (STARTING_FRAME_OFFSET): Split out
11220         !FRAME_GROWS_DOWNWARD handling to...
11221         (RS6000_STARTING_FRAME_OFFSET): ...this new macro.
11222         * config/rs6000/rs6000.c (TARGET_STARTING_FRAME_OFFSET): Refine.
11223         (rs6000_starting_frame_offset): New function.
11224         * config/vax/elf.h (STARTING_FRAME_OFFSET): Delete.
11225         * config/vax/vax.h (STARTING_FRAME_OFFSET): Delete.
11226         * config/vax/vax.c (vax_starting_frame_offset): New function.
11227         (vax_expand_prologue): Use it instead of STARTING_FRAME_OFFSET.
11228         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11229         * config/xtensa/xtensa.h (STARTING_FRAME_OFFSET): Delete.
11230         * config/xtensa/xtensa.c (xtensa_starting_frame_offset): New function.
11231         (TARGET_STARTING_FRAME_OFFSET): Redefine.
11232         * system.h (STARTING_FRAME_OFFSET): Poison.
11234 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11236         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
11237         SCALAR_TYPE_MODE instead of TYPE_MODE.
11239 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11240             Alan Hayward  <alan.hayward@arm.com>
11241             David Sherwood  <david.sherwood@arm.com>
11243         * dwarf2out.c (loc_list_from_tree_1): Use SCALAR_INT_TYPE_MODE
11245 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11246             Alan Hayward  <alan.hayward@arm.com>
11247             David Sherwood  <david.sherwood@arm.com>
11249         * expmed.c (expand_shift_1): Use scalar_mode for scalar_mode.
11251 2017-10-23  Richard Biener  <rguenther@suse.de>
11253         PR tree-optimization/82129
11254         * tree-ssa-pre.c (bitmap_set_and): Remove.
11255         (compute_antic_aux): Compute ANTIC_OUT intersection in a way
11256         canonicalizing expressions in the set to those with lowest
11257         ID rather than taking that from the first edge.
11259 2017-10-23  Richard Sandiford  <richard.sandiford@linaro.org>
11261         * combine.c (rtx_equal_for_field_assignment_p): Use
11262         byte_lowpart_offset.
11264 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11265             Alan Hayward  <alan.hayward@arm.com>
11266             David Sherwood  <david.sherwood@arm.com>
11268         * internal-fn.c (expand_direct_optab_fn): Don't assign directly
11269         to a SUBREG_PROMOTED_VAR.
11271 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11272             Alan Hayward  <alan.hayward@arm.com>
11273             David Sherwood  <david.sherwood@arm.com>
11275         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_PRECISION.
11276         (expand_debug_source_expr): Likewise.
11277         * combine.c (combine_simplify_rtx): Likewise.
11278         * cse.c (fold_rtx): Likewise.
11279         * optabs.c (expand_float): Likewise.
11280         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
11281         (simplify_binary_operation_1): Likewise.
11283 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11284             Alan Hayward  <alan.hayward@arm.com>
11285             David Sherwood  <david.sherwood@arm.com>
11287         * combine.c (simplify_comparison): Use HWI_COMPUTABLE_MODE_P.
11288         (record_promoted_value): Likewise.
11289         * expr.c (expand_expr_real_2): Likewise.
11290         * ree.c (update_reg_equal_equiv_notes): Likewise.
11291         (combine_set_extension): Likewise.
11292         * rtlanal.c (low_bitmask_len): Likewise.
11293         * simplify-rtx.c (neg_const_int): Likewise.
11294         (simplify_binary_operation_1): Likewise.
11296 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11297             Alan Hayward  <alan.hayward@arm.com>
11298             David Sherwood  <david.sherwood@arm.com>
11300         * lra-spills.c (assign_mem_slot): Use subreg_size_lowpart_offset.
11301         * regcprop.c (maybe_mode_change): Likewise.
11302         * reload1.c (alter_reg): Likewise.
11304 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11306         * inchash.h (inchash::hash::add_wide_int): New function.
11307         * lto-streamer-out.c (hash_tree): Use it.
11309 2017-10-22  Richard Sandiford  <richard.sandiford@linaro.org>
11311         * inchash.h (inchash::hash::add_wide_int): Rename to...
11312         (inchash::hash::add_hwi): ...this.
11313         * ipa-devirt.c (hash_odr_vtable): Update accordingly.
11314         (polymorphic_call_target_hasher::hash): Likewise.
11315         * ipa-icf.c (sem_function::get_hash, sem_function::init): Likewise.
11316         (sem_item::add_expr, sem_item::add_type, sem_variable::get_hash)
11317         (sem_item_optimizer::update_hash_by_addr_refs): Likewise.
11318         * lto-streamer-out.c (hash_tree): Likewise.
11319         * optc-save-gen.awk: Likewise.
11320         * tree.c (add_expr): Likewise.
11322 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11324         PR target/52451
11325         * config/i386/i386.c (ix86_fp_compare_mode): Return CCFPmode
11326         for ordered inequality comparisons even with TARGET_IEEE_FP.
11328 2017-10-22  Uros Bizjak  <ubizjak@gmail.com>
11330         PR target/82628
11331         * config/i386/i386.md (cmp<dwi>_doubleword): New pattern.
11332         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11333         Expand with cmp<dwi>_doubleword.
11335 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11337         * extend.texi: Add x86 specific to 'nocf_check' attribute.
11338         List CET intrinsics.
11339         * invoke.texi: Add -mcet, -mibt, -mshstk options.  Add x86
11340         specific to -fcf-protection option.
11342 2017-10-21  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11344         * common/config/i386/i386-common.c (OPTION_MASK_ISA_IBT_SET): New.
11345         (OPTION_MASK_ISA_SHSTK_SET): Likewise.
11346         (OPTION_MASK_ISA_IBT_UNSET): Likewise.
11347         (OPTION_MASK_ISA_SHSTK_UNSET): Likewise.
11348         (ix86_handle_option): Add -mibt, -mshstk, -mcet handling.
11349         * config.gcc (extra_headers): Add cetintrin.h for x86 targets.
11350         (extra_objs): Add cet.o for Linux/x86 targets.
11351         (tmake_file): Add i386/t-cet for Linux/x86 targets.
11352         * config/i386/cet.c: New file.
11353         * config/i386/cetintrin.h: Likewise.
11354         * config/i386/t-cet: Likewise.
11355         * config/i386/cpuid.h (bit_SHSTK): New.
11356         (bit_IBT): Likewise.
11357         * config/i386/driver-i386.c (host_detect_local_cpu): Detect and
11358         pass IBT and SHSTK bits.
11359         * config/i386/i386-builtin-types.def
11360         (VOID_FTYPE_UNSIGNED_PVOID): New.
11361         (VOID_FTYPE_UINT64_PVOID): Likewise.
11362         * config/i386/i386-builtin.def: Add CET intrinsics.
11363         * config/i386/i386-c.c (ix86_target_macros_internal): Add
11364         OPTION_MASK_ISA_IBT, OPTION_MASK_ISA_SHSTK handling.
11365         * config/i386/i386-passes.def: Add pass_insert_endbranch pass.
11366         * config/i386/i386-protos.h (make_pass_insert_endbranch): New
11367         prototype.
11368         * config/i386/i386.c (rest_of_insert_endbranch): New.
11369         (pass_data_insert_endbranch): Likewise.
11370         (pass_insert_endbranch): Likewise.
11371         (make_pass_insert_endbranch): Likewise.
11372         (ix86_notrack_prefixed_insn_p): Likewise.
11373         (ix86_target_string): Add -mibt, -mshstk flags.
11374         (ix86_option_override_internal): Add flag_cf_protection
11375         processing.
11376         (ix86_valid_target_attribute_inner_p): Set OPT_mibt, OPT_mshstk.
11377         (ix86_print_operand): Add 'notrack' prefix output.
11378         (ix86_init_mmx_sse_builtins): Add CET intrinsics.
11379         (ix86_expand_builtin): Expand CET intrinsics.
11380         (x86_output_mi_thunk): Add 'endbranch' instruction.
11381         * config/i386/i386.h (TARGET_IBT): New.
11382         (TARGET_IBT_P): Likewise.
11383         (TARGET_SHSTK): Likewise.
11384         (TARGET_SHSTK_P): Likewise.
11385         * config/i386/i386.md (unspecv): Add UNSPECV_NOP_RDSSP,
11386         UNSPECV_INCSSP, UNSPECV_SAVEPREVSSP, UNSPECV_RSTORSSP,
11387         UNSPECV_WRSS, UNSPECV_WRUSS, UNSPECV_SETSSBSY, UNSPECV_CLRSSBSY.
11388         (builtin_setjmp_setup): New pattern.
11389         (builtin_longjmp): Likewise.
11390         (rdssp<mode>): Likewise.
11391         (incssp<mode>): Likewise.
11392         (saveprevssp): Likewise.
11393         (rstorssp): Likewise.
11394         (wrss<mode>): Likewise.
11395         (wruss<mode>): Likewise.
11396         (setssbsy): Likewise.
11397         (clrssbsy): Likewise.
11398         (nop_endbr): Likewise.
11399         * config/i386/i386.opt: Add -mcet, -mibt, -mshstk and -mcet-switch
11400         options.
11401         * config/i386/immintrin.h: Include <cetintrin.h>.
11402         * config/i386/linux-common.h
11403         (file_end_indicate_exec_stack_and_cet): New prototype.
11404         (TARGET_ASM_FILE_END): New.
11406 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11408         * i386.c (ix86_builtin_vectorization_cost): Use existing rtx_cost
11409         latencies instead of having separate table; make difference between
11410         integer and float costs.
11411         * i386.h (processor_costs): Remove scalar_stmt_cost,
11412         scalar_load_cost, scalar_store_cost, vec_stmt_cost, vec_to_scalar_cost,
11413         scalar_to_vec_cost, vec_align_load_cost, vec_unalign_load_cost,
11414         vec_store_cost.
11415         * x86-tune-costs.h: Remove entries which has been removed in
11416         procesor_costs from all tables; make cond_taken_branch_cost
11417         and cond_not_taken_branch_cost COST_N_INSNS based.
11419 2017-10-20  Jan Hubicka  <hubicka@ucw.cz>
11421         * x86-tune-costs.h (intel_cost, generic_cost): Fix move costs.
11423 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11425         * config/i386/i386.md (isa): Remove fma_avx512f.
11426         * config/i386/sse.md (<avx512>_fmadd_<mode>_mask<round_name>,
11427         <avx512>_fmadd_<mode>_mask3<round_name>,
11428         <avx512>_fmsub_<mode>_mask<round_name>,
11429         <avx512>_fmsub_<mode>_mask3<round_name>,
11430         <avx512>_fnmadd_<mode>_mask<round_name>,
11431         <avx512>_fnmadd_<mode>_mask3<round_name>,
11432         <avx512>_fnmsub_<mode>_mask<round_name>,
11433         <avx512>_fnmsub_<mode>_mask3<round_name>,
11434         <avx512>_fmaddsub_<mode>_mask<round_name>,
11435         <avx512>_fmaddsub_<mode>_mask3<round_name>,
11436         <avx512>_fmsubadd_<mode>_mask<round_name>,
11437         <avx512>_fmsubadd_<mode>_mask3<round_name>): Remove isa attribute.
11438         (*vec_widen_umult_even_v16si<mask_name>,
11439         *vec_widen_smult_even_v16si<mask_name>): Likewise.
11440         (<mask_codefor>avx512bw_dbpsadbw<mode><mask_name>): Likewise.
11442 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11444         * extend.texi: Add 'nocf_check' documentation.
11445         * gimple.texi: Add second parameter to
11446         gimple_build_call_from_tree.
11447         * invoke.texi: Add -fcf-protection documentation.
11448         * rtl.texi: Add REG_CALL_NOTRACK documenation.
11450 2017-10-20  Richard Biener  <rguenther@suse.de>
11452         PR tree-optimization/82473
11453         * tree-vect-loop.c (vectorizable_reduction): Properly get at
11454         the largest input type.
11456 2017-10-20  Igor Tsimbalist  <igor.v.tsimbalist@intel.com>
11458         * c-attribs.c (handle_nocf_check_attribute): New function.
11459         (c_common_attribute_table): Add 'nocf_check' handling.
11460         * gimple-parser.c: Add second argument NULL to
11461         gimple_build_call_from_tree.
11462         * attrib.c (comp_type_attributes): Check nocf_check attribute.
11463         * cfgexpand.c (expand_call_stmt): Set REG_CALL_NOCF_CHECK for
11464         call insn.
11465         * combine.c (distribute_notes): Add REG_CALL_NOCF_CHECK handling.
11466         * common.opt: Add fcf-protection flag.
11467         * emit-rtl.c (try_split): Add REG_CALL_NOCF_CHECK handling.
11468         * flag-types.h: Add enum cf_protection_level.
11469         * gimple.c (gimple_build_call_from_tree): Add second parameter.
11470         Add 'nocf_check' attribute propagation to gimple call.
11471         * gimple.h (gf_mask): Add GF_CALL_NOCF_CHECK.
11472         (gimple_build_call_from_tree): Update prototype.
11473         (gimple_call_nocf_check_p): New function.
11474         (gimple_call_set_nocf_check): Likewise.
11475         * gimplify.c: Add second argument to gimple_build_call_from_tree.
11476         * ipa-icf.c: Add nocf_check attribute in statement hash.
11477         * recog.c (peep2_attempt): Add REG_CALL_NOCF_CHECK handling.
11478         * reg-notes.def: Add REG_NOTE (CALL_NOCF_CHECK).
11479         * toplev.c (process_options): Add flag_cf_protection handling.
11481 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11483         * x86-tune-costs.h (core_cost): Fix div, move and sqrt latencies.
11485 2017-10-20  Richard Biener  <rguenther@suse.de>
11487         PR tree-optimization/82603
11488         * tree-if-conv.c (predicate_mem_writes): Make sure to only
11489         remove false predicated stores.
11491 2017-10-20  Richard Biener  <rguenther@suse.de>
11493         * graphite-isl-ast-to-gimple.c
11494         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block):
11495         Remove return value and simplify, dump copied stmt after lhs
11496         adjustment.
11497         (translate_isl_ast_to_gimple::translate_isl_ast_node_user):
11498         Reduce dump verbosity.
11499         (gsi_insert_earliest): Likewise.
11500         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Adjust.
11501         * graphite.c (print_global_statistics): Adjust dumping.
11502         (print_graphite_scop_statistics): Likewise.
11503         (print_graphite_statistics): Do not dump loops here.
11504         (graphite_transform_loops): But here.
11506 2017-10-20  Nicolas Roche  <roche@adacore.com>
11508         * configure.ac (ACX_PROG_GNAT): Append "libgnat" to include search dir.
11509         * configure: Regenerate.
11511 2017-10-20  Jakub Jelinek  <jakub@redhat.com>
11513         PR target/82158
11514         * tree-cfg.c (pass_warn_function_return::execute): In noreturn
11515         functions when optimizing replace GIMPLE_RETURN stmts with
11516         calls to __builtin_unreachable ().
11518         PR sanitizer/82595
11519         * config/gnu-user.h (LIBTSAN_EARLY_SPEC): Add libtsan_preinit.o
11520         for -fsanitize=thread link of executables.
11521         (LIBLSAN_EARLY_SPEC): Add liblsan_preinit.o for -fsanitize=leak
11522         link of executables.
11524         PR target/82370
11525         * config/i386/sse.md (VI248_AVX2, VI248_AVX512BW, VI248_AVX512BW_2):
11526         New mode iterators.
11527         (<shift_insn><mode>3<mask_name>): Change the last of the 3
11528         define_insns for logical vector shifts to use VI248_AVX512BW
11529         iterator instead of VI48_AVX512, remove <mask_mode512bit_condition>
11530         condition, useless isa and prefix attributes.  Change the first
11531         2 of these define_insns to ...
11532         (<mask_codefor><shift_insn><mode>3<mask_name>): ... this, new
11533         define_insn for avx512vl.
11534         (<shift_insn><mode>3): ... and this, new define_insn without
11535         masking for non-avx512vl.
11537         PR target/82370
11538         * config/i386/sse.md (*andnot<mode>3,
11539         <mask_codefor><code><mode>3<mask_name>, *<code><mode>3): Split
11540         (=v,v,vm) alternative into (=x,x,xm) and (=v,v,vm), for 128-bit
11541         and 256-bit vectors, the (=x,x,xm) alternative and when mask is
11542         not applied use empty suffix even for TARGET_AVX512VL.
11543         * config/i386/subst.md (mask_prefix3, mask_prefix4): When mask
11544         is applied, supply evex,evex or evex,evex,evex instead of just
11545         evex.
11547 2017-10-20  Julia Koval  <julia.koval@intel.com>
11549         * common/config/i386/i386-common.c (OPTION_MASK_ISA_GFNI_SET,
11550         (OPTION_MASK_ISA_GFNI_UNSET): New.
11551         (ix86_handle_option): Handle OPT_mgfni.
11552         * config/i386/cpuid.h (bit_GFNI): New.
11553         * config/i386/driver-i386.c (host_detect_local_cpu): Detect gfni.
11554         * config/i386/i386-c.c (ix86_target_macros_internal): Define __GFNI__.
11555         * config/i386/i386.c (ix86_target_string): Add -mgfni.
11556         (ix86_valid_target_attribute_inner_p): Add OPT_mgfni.
11557         * config/i386/i386.h (TARGET_GFNI, TARGET_GFNI_P): New.
11558         * config/i386/i386.opt: Add mgfni.
11560 2017-10-20  Orlando Arias  <oarias@knights.ucf.edu>
11562         * config/msp430/msp430.c (msp430_option_override): Disable
11563         -fdelete-null-pointer-checks.
11564         * doc/invoke.text (-fdelete-null-pointer-checks): Document that.
11566 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11568         * x86-tune-costs.h (generic_cost, core_cost): Correct costs
11569         of x87 and SSE instructions.
11571 2017-10-19  Jan Hubicka  <hubicka@ucw.cz>
11573         * asan.c (create_cond_insert_point): Do not update edge count.
11574         * auto-profile.c (afdo_propagate_edge): Update for edge count removal.
11575         (afdo_propagate_circuit): Likewise.
11576         (afdo_calculate_branch_prob): Likewise.
11577         (afdo_annotate_cfg): Likewise.
11578         * basic-block.h (struct edge_def): Remove count.
11579         (edge_def::count): New accessor.
11580         * bb-reorder.c (rotate_loop): Update.
11581         (find_traces_1_round): Update.
11582         (connect_traces): Update.
11583         (sanitize_hot_paths): Update.
11584         * cfg.c (unchecked_make_edge): Update.
11585         (make_single_succ_edge): Update.
11586         (check_bb_profile): Update.
11587         (dump_edge_info): Update.
11588         (update_bb_profile_for_threading): Update.
11589         (scale_bbs_frequencies_int): Update.
11590         (scale_bbs_frequencies_gcov_type): Update.
11591         (scale_bbs_frequencies_profile_count): Update.
11592         (scale_bbs_frequencies): Update.
11593         * cfganal.c (connect_infinite_loops_to_exit): Update.
11594         * cfgbuild.c (compute_outgoing_frequencies): Update.
11595         (find_many_sub_basic_blocks): Update.
11596         * cfgcleanup.c (try_forward_edges): Update.
11597         (try_crossjump_to_edge): Update
11598         * cfgexpand.c (expand_gimple_cond): Update
11599         (expand_gimple_tailcall): Update
11600         (construct_exit_block): Update
11601         * cfghooks.c (verify_flow_info): Update
11602         (redirect_edge_succ_nodup): Update
11603         (split_edge): Update
11604         (make_forwarder_block): Update
11605         (duplicate_block): Update
11606         (account_profile_record): Update
11607         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
11608         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
11609         * cfgloopmanip.c (scale_loop_profile): Update.
11610         (loopify): Update.
11611         (lv_adjust_loop_entry_edge): Update.
11612         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
11613         (force_nonfallthru_and_redirect): Update.
11614         (purge_dead_edges): Update.
11615         (rtl_flow_call_edges_add): Update.
11616         * cgraphunit.c (init_lowered_empty_function): Update.
11617         (cgraph_node::expand_thunk): Update.
11618         * gimple-pretty-print.c (dump_probability): Update.
11619         (dump_edge_probability): Update.
11620         * gimple-ssa-isolate-paths.c (isolate_path): Update.
11621         * haifa-sched.c (sched_create_recovery_edges): Update.
11622         * hsa-gen.c (convert_switch_statements): Update.
11623         * ifcvt.c (dead_or_predicable): Update.
11624         * ipa-inline-transform.c (inline_transform): Update.
11625         * ipa-split.c (split_function): Update.
11626         * ipa-utils.c (ipa_merge_profiles): Update.
11627         * loop-doloop.c (add_test): Update.
11628         * loop-unroll.c (unroll_loop_runtime_iterations): Update.
11629         * lto-streamer-in.c (input_cfg): Update.
11630         (input_function): Update.
11631         * lto-streamer-out.c (output_cfg): Update.
11632         * modulo-sched.c (sms_schedule): Update.
11633         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
11634         * predict.c (maybe_hot_edge_p): Update.
11635         (unlikely_executed_edge_p): Update.
11636         (probably_never_executed_edge_p): Update.
11637         (dump_prediction): Update.
11638         (drop_profile): Update.
11639         (propagate_unlikely_bbs_forward): Update.
11640         (determine_unlikely_bbs): Update.
11641         (force_edge_cold): Update.
11642         * profile.c (compute_branch_probabilities): Update.
11643         * reg-stack.c (better_edge): Update.
11644         * shrink-wrap.c (handle_simple_exit): Update.
11645         * tracer.c (better_p): Update.
11646         * trans-mem.c (expand_transaction): Update.
11647         (split_bb_make_tm_edge): Update.
11648         * tree-call-cdce.c: Update.
11649         * tree-cfg.c (gimple_find_sub_bbs): Update.
11650         (gimple_split_edge): Update.
11651         (gimple_duplicate_sese_region): Update.
11652         (gimple_duplicate_sese_tail): Update.
11653         (gimple_flow_call_edges_add): Update.
11654         (insert_cond_bb): Update.
11655         (execute_fixup_cfg): Update.
11656         * tree-cfgcleanup.c (cleanup_control_expr_graph): Update.
11657         * tree-complex.c (expand_complex_div_wide): Update.
11658         * tree-eh.c (lower_resx): Update.
11659         (unsplit_eh): Update.
11660         (cleanup_empty_eh_move_lp): Update.
11661         * tree-inline.c (copy_edges_for_bb): Update.
11662         (freqs_to_counts): Update.
11663         (copy_cfg_body): Update.
11664         * tree-ssa-dce.c (remove_dead_stmt): Update.
11665         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
11666         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
11667         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
11668         (unloop_loops): Update.
11669         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
11670         * tree-ssa-loop-split.c (connect_loops): Update.
11671         (split_loop): Update.
11672         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
11673         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
11674         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
11675         * tree-ssa-reassoc.c (branch_fixup): Update.
11676         * tree-ssa-tail-merge.c (replace_block_by): Update.
11677         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
11678         (compute_path_counts): Update.
11679         (update_profile): Update.
11680         (recompute_probabilities): Update.
11681         (update_joiner_offpath_counts): Update.
11682         (estimated_freqs_path): Update.
11683         (freqs_to_counts_path): Update.
11684         (clear_counts_path): Update.
11685         (ssa_fix_duplicate_block_edges): Update.
11686         (duplicate_thread_path): Update.
11687         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
11688         (case_bit_test_cmp): Update.
11689         (collect_switch_conv_info): Update.
11690         (gen_inbound_check): Update.
11691         (do_jump_if_equal): Update.
11692         (emit_cmp_and_jump_insns): Update.
11693         * tree-tailcall.c (decrease_profile): Update.
11694         (eliminate_tail_call): Update.
11695         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
11696         (vect_do_peeling): Update.
11697         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
11698         * ubsan.c (ubsan_expand_null_ifn): Update.
11699         (ubsan_expand_ptr_ifn): Update.
11700         * value-prof.c (gimple_divmod_fixed_value): Update.
11701         (gimple_mod_pow2): Update.
11702         (gimple_mod_subtract): Update.
11703         (gimple_ic): Update.
11704         (gimple_stringop_fixed_value): Update.
11706 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11708         PR target/82618
11709         * config/i386/i386.md (sub to cmp): New peephole2 pattern.
11711 2017-10-19  Alexander Monakov  <amonakov@ispras.ru>
11713         PR rtl-optimization/82395
11714         * ira-color.c (allocno_priority_compare_func): Fix comparison step
11715         based on non_spilled_static_chain_regno_p.
11717 2017-10-19  Uros Bizjak  <ubizjak@gmail.com>
11719         * config/i386/i386.c (output_387_binary_op): Rewrite SSE part.
11720         (ix86_emit_mode_set): Rewrite insn mnemonic construction.
11721         (ix86_prepare_fp_compare_args): Redefine is_sse as bool.
11723 2017-10-19  Martin Sebor  <msebor@redhat.com>
11725         PR tree-optimization/82596
11726         * tree.c (array_at_struct_end_p): Handle STRING_CST.
11728 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11730         * asan.c (handle_builtin_alloca): Deal with all alloca variants.
11731         (get_mem_refs_of_builtin_call): Likewise.
11732         * builtins.c (expand_builtin_apply): Adjust call to
11733         allocate_dynamic_stack_space.
11734         (expand_builtin_alloca): For __builtin_alloca_with_align_and_max, pass
11735         the third argument to allocate_dynamic_stack_space, otherwise -1.
11736         (expand_builtin): Deal with all alloca variants.
11737         (is_inexpensive_builtin): Likewise.
11738         * builtins.def (BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX): New.
11739         * calls.c (special_function_p): Deal with all alloca variants.
11740         (initialize_argument_information): Adjust call to
11741         allocate_dynamic_stack_space.
11742         (expand_call): Likewise.
11743         * cfgexpand.c (expand_call_stmt): Deal with all alloca variants.
11744         * doc/extend.texi (Built-ins): Add __builtin_alloca_with_align_and_max
11745         * explow.c (allocate_dynamic_stack_space): Add MAX_SIZE parameter and
11746         use it for the stack usage computation.
11747         * explow.h (allocate_dynamic_stack_space): Adjust prototype.
11748         * function.c (gimplify_parameters): Call build_alloca_call_expr.
11749         * gimple-ssa-warn-alloca.c (alloca_call_type): Simplify control flow.
11750         Take into account 3rd argument of __builtin_alloca_with_align_and_max.
11751         (in_loop_p): Remove first argument and useless check.
11752         (pass_walloca::execute): Remove useless test and adjust call to above.
11753         * gimple.c (gimple_build_call_from_tree): Deal with all alloc variants
11754         * gimplify.c (gimplify_vla_decl): Call build_alloca_call_expr.
11755         (gimplify_call_expr): Deal with all alloca variants.
11756         * hsa-gen.c (gen_hsa_alloca): Likewise.
11757         (gen_hsa_insns_for_call): Likewise.
11758         * ipa-pure-const.c (special_builtin_state): Likewise.
11759         * tree-chkp.c (chkp_build_returned_bound): Likewise.
11760         * tree-object-size.c (alloc_object_size): Likewise.
11761         * tree-ssa-alias.c (ref_maybe_used_by_call_p_1): Likewise.
11762         (call_may_clobber_ref_p_1): Likewise.
11763         * tree-ssa-ccp.c (evaluate_stmt): Likewise.
11764         (ccp_fold_stmt): Likewise.
11765         (optimize_stack_restore): Likewise.
11766         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Likewise.
11767         (mark_all_reaching_defs_necessary_1): Likewise.
11768         (propagate_necessity): Likewise.
11769         (eliminate_unnecessary_stmts): Likewise.
11770         * tree.c (build_common_builtin_nodes): Build
11771         BUILT_IN_ALLOCA_WITH_ALIGN_AND_MAX.
11772         (build_alloca_call_expr): New function.
11773         * tree.h (ALLOCA_FUNCTION_CODE_P): New macro.
11774         (CASE_BUILT_IN_ALLOCA): Likewise.
11775         (build_alloca_call_expr): Declare.
11776         * varasm.c (incorporeal_function_p): Deal with all alloca variants.
11778 2017-10-19  Eric Botcazou  <ebotcazou@adacore.com>
11780         PR debug/82509
11781         * dwarf2out.c (new_die_raw): New static inline function.
11782         (new_die): Use it to create the DIE.
11783         (add_AT_external_die_ref): Likewise.
11784         (clone_die): Likewise.
11785         (clone_as_declaration): Likewise.
11786         (dwarf2out_vms_debug_main_pointer): Likewise.
11787         (base_type_die): Likewise.  Remove early return for corner cases.
11788         Do not call add_pubtype on the DIE here.
11789         (is_base_type): Remove ERROR_MARK and return 0 for VOID_TYPE.
11790         (modified_type_die): Adjust the lookup for reverse order DIEs.  Skip
11791         typedefs for base types with DW_AT_endianity.  Make sure a DIE with
11792         native order exists for base types, attach the DIE manually and call
11793         add_pubtype on it.  Do not equate a reverse order DIE to the type.
11795 2017-10-19  Richard Earnshaw  <rearnsha@arm.com>
11797         * config/arm/arm.c (align_ok_ldrd_strd): New function.
11798         (mem_ok_for_ldrd_strd): New parameter align.  Extract the alignment of
11799         the mem into it.
11800         (gen_operands_ldrd_strd): Validate the alignment of the accesses.
11802 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11804         * flag-types.h (enum sanitize_code): Add SANITIZE_BUILTIN.  Or
11805         SANITIZE_BUILTIN into SANITIZE_UNDEFINED.
11806         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN,
11807         BUILT_IN_UBSAN_HANDLE_INVALID_BUILTIN_ABORT): New builtins.
11808         * opts.c (sanitizer_opts): Add builtin.
11809         * ubsan.c (instrument_builtin): New function.
11810         (pass_ubsan::execute): Call it.
11811         (pass_ubsan::gate): Enable even for SANITIZE_BUILTIN.
11812         * doc/invoke.texi: Document -fsanitize=builtin.
11814         * ubsan.c (ubsan_expand_null_ifn): Use _v1 suffixed type mismatch
11815         builtins, store max (log2 (align), 0) into uchar field instead of
11816         align into uptr field.
11817         (ubsan_expand_objsize_ifn): Use _v1 suffixed type mismatch builtins,
11818         store uchar 0 field instead of uptr 0 field.
11819         (instrument_nonnull_return): Use _v1 suffixed nonnull return builtin,
11820         instead of passing one address of struct with 2 locations pass
11821         two addresses of structs with 1 location each.
11822         * sanitizer.def (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH,
11823         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_ABORT,
11824         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN,
11825         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_ABORT): Removed.
11826         (BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1,
11827         BUILT_IN_UBSAN_HANDLE_TYPE_MISMATCH_V1_ABORT,
11828         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1,
11829         BUILT_IN_UBSAN_HANDLE_NONNULL_RETURN_V1_ABORT): New builtins.
11831 2017-10-19  Martin Liska  <mliska@suse.cz>
11833         PR driver/81829
11834         * file-find.c (remove_prefix): Remove.
11835         * file-find.h (remove_prefix): Likewise.
11836         * gcc-ar.c: Remove smartness of lookup.
11838 2017-10-19  Segher Boessenkool  <segher@kernel.crashing.org>
11840         * config/rs6000/rs6000.md (*call_indirect_aix<mode>,
11841         *call_value_indirect_aix<mode>, *call_indirect_elfv2<mode>,
11842         *call_value_indirect_elfv2<mode>): Add correct mode to the unspec.
11844 2017-10-19  Jakub Jelinek  <jakub@redhat.com>
11846         PR target/82580
11847         * config/i386/i386.md (setcc + movzbl to xor + setcc): New peephole2.
11848         (setcc + and to xor + setcc): New peephole2.
11850 2017-10-19  Tom de Vries  <tom@codesourcery.com>
11852         * doc/sourcebuild.texi (Test Directives, Variants of
11853         dg-require-support): Add dg-require-stack-size.
11855 2017-10-19  Martin Liska  <mliska@suse.cz>
11857         PR sanitizer/82517
11858         * gimplify.c (gimplify_decl_expr): Do not instrument variables
11859         that have a large alignment.
11860         (gimplify_target_expr): Likewise.
11862 2017-10-18  Segher Boessenkool  <segher@kernel.crashing.org>
11864         PR rtl-optimization/82602
11865         * ira.c (rtx_moveable_p): Return false for volatile asm.
11867 2017-10-18  Uros Bizjak  <ubizjak@gmail.com>
11869         PR target/82580
11870         * config/i386/i386-modes.def (CCGZ): New CC mode.
11871         * config/i386/i386.md (sub<mode>3_carry_ccgz): New insn pattern.
11872         * config/i386/predicates.md (ix86_comparison_operator):
11873         Handle CCGZmode.
11874         * config/i386/i386.c (ix86_expand_branch) <case E_TImode>:
11875         Emulate LE, LEU, GT, GTU, LT, LTU, GE and GEU double-word comparisons
11876         with double-word subtraction.
11877         (put_condition_code): Handle CCGZmode.
11879 2017-10-18  Aldy Hernandez  <aldyh@redhat.com>
11881         * wide-int.cc (debug (const wide_int &)): New.
11882         (debug (const wide_int *)): New.
11883         (debug (const widest_int &)): New.
11884         (debug (const widest_int *)): New.
11886 2017-10-18  Vladimir Makarov  <vmakarov@redhat.com>
11888         PR middle-end/82556
11889         * lra-constraints.c (curr_insn_transform): Use non-input operand
11890         instead of output one for matched reload.
11892 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11894         * tree-loop-distribution.c (INCLUDE_ALGORITHM): New header file.
11895         (tree-ssa-loop-ivopts.h): New header file.
11896         (struct builtin_info): New fields.
11897         (classify_builtin_1): Compute and record base and offset parts for
11898         memset builtin partition by calling strip_offset.
11899         (offset_cmp, fuse_memset_builtins): New functions.
11900         (finalize_partitions): Fuse adjacent memset partitions by calling
11901         above function.
11902         * tree-ssa-loop-ivopts.c (strip_offset): Delete static declaration.
11903         Expose the interface.
11904         * tree-ssa-loop-ivopts.h (strip_offset): New declaration.
11906 2017-10-18  Bin Cheng  <bin.cheng@arm.com>
11908         PR tree-optimization/82574
11909         * tree-loop-distribution.c (find_single_drs): New parameter.  Check
11910         that data reference must be executed exactly once per iteration
11911         against the outermost loop in nest.
11912         (classify_partition): Update call to above function.
11914 2017-10-18  Richard Biener  <rguenther@suse.de>
11916         PR tree-optimization/82591
11917         * graphite.c (graphite_transform_loops): Move code gen message
11918         printing ...
11919         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
11920         Here.  Handle scop_to_isl_ast failing.
11921         (scop_to_isl_ast): Limit the number of ISL operations.
11923 2017-10-18  Richard Biener  <rguenther@suse.de>
11925         * graphite-isl-ast-to-gimple.c
11926         (translate_isl_ast_to_gimple::set_rename): Simplify.
11927         (translate_isl_ast_to_gimple::set_rename_for_each_def): Inline...
11928         (graphite_copy_stmts_from_block): ... here.
11929         (copy_bb_and_scalar_dependences): Simplify.
11930         (add_parameters_to_ivs_params): Canonicalize.
11931         (generate_entry_out_of_ssa_copies): Simplify.
11932         * graphite-sese-to-poly.c (extract_affine_name): Simplify
11933         by passing in ISL dimension.
11934         (parameter_index_in_region_1): Rename to ...
11935         (parameter_index_in_region): ... this.
11936         (extract_affine): Adjust assert, pass down parameter index.
11937         (add_param_constraints): Use range-info when available.
11938         (build_scop_context): Adjust.
11939         * sese.c (new_sese_info): Adjust.
11940         (free_sese_info): Likewise.
11941         * sese.h (bb_map_t, rename_map_t, phi_rename, init_back_edge_pair_t):
11942         Remove unused typedefs.
11943         (struct sese_info_t): Simplify rename_map, remove incomplete_phis.
11945 2017-10-18  Martin Liska  <mliska@suse.cz>
11947         * combine.c (simplify_compare_const): Add gcc_fallthrough.
11949 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11951         * config/s390/s390.c (s390_bb_fallthru_entry_likely): New function.
11952         (s390_sched_init): Do not reset s390_sched_state if we entered the
11953         current basic block via a fallthru edge and all others are unlikely.
11955 2017-10-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
11957         * config/s390/s390.c (NUM_SIDES): New variable.
11958         (LONGRUNNING_THRESHOLD): New variable.
11959         (LATENCY_FACTOR): New variable.
11960         (s390_sched_score): Decrease score for long-running instructions on
11961         wrong side.
11962         (s390_sched_variable_issue): Perform bookkeeping for long-running
11963         instructions.
11965 2017-10-18  Richard Biener  <rguenther@suse.de>
11967         * graphite-isl-ast-to-gimple.c (gcc_expression_from_isl_ast_expr_id):
11968         Simplify with removal of the parameter rename map.
11969         (set_rename): Likewise.
11970         (should_copy_to_new_region): Likewise.
11971         (graphite_copy_stmts_from_block): Likewise.
11972         (copy_bb_and_scalar_dependences): Remove initialization of
11973         unused copied_bb_map.
11974         (copy_def): Remove.
11975         (copy_internal_parameters): Likewise.
11976         (graphite_regenerate_ast_isl): Do not call copy_internal_parameters.
11977         * graphite-scop-detection.c (scop_detection::stmt_simple_for_scop_p):
11978         Use INTEGRAL_TYPE_P.
11979         (parameter_index_in_region_1): Rename to ...
11980         (assign_parameter_index_in_region): ... this.  Assert we have
11981         a parameter we handle.
11982         (scan_tree_for_params): Adjust.
11983         * sese.h (parameter_rename_map_t): Remove.
11984         (struct sese_info_t): Remove unused parameter_rename_map and
11985         copied_bb_map members.
11986         * sese.c (new_sese_info): Adjust.
11987         (free_sese_info): Likewise.
11989 2017-10-18  Martin Liska  <mliska@suse.cz>
11991         PR sanitizer/82545
11992         * asan.c (asan_expand_poison_ifn): Do not put gimple stmt
11993         on an abnormal edge.
11995 2017-10-18  Sebastian Huber  <sebastian.huber@embedded-brains.de>
11997         * doc/invoke.texi (ffunction-sections and fdata-sections):
11998         Update.
12000 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12002         * tree-ssa-loop-ivopts.c (add_autoinc_candidates): Bail out only if
12003         the use statement can throw internally.
12005 2017-10-17  Eric Botcazou  <ebotcazou@adacore.com>
12007         * config/visium/visium.c (visium_select_cc_mode): Return CCmode for
12008         any RTX present on the RHS of a SET.
12009         * compare-elim.c (try_eliminate_compare): Restore comment.
12011 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12013         * langhooks.h (struct lang_hooks): Document that tree_size langhook
12014         may be also called on tcc_type nodes.
12015         * langhooks.c (lhd_tree_size): Likewise.
12017 2017-10-17  David Malcolm  <dmalcolm@redhat.com>
12019         * gimple-ssa-sprintf.c (fmtwarn): Update for changed signature of
12020         format_warning_at_substring.
12021         (maybe_warn): Convert source_range * param to a location_t.  Pass
12022         UNKNOWN_LOCATION rather than NULL to fmtwarn.
12023         (format_directive): Remove code to extract source_ranges and
12024         source_range * in favor of just a location_t.
12025         (parse_directive): Pass UNKNOWN_LOCATION rather than NULL to
12026         fmtwarn.
12027         * substring-locations.c (format_warning_va): Convert
12028         source_range * param to a location_t.
12029         (format_warning_at_substring): Likewise.
12030         * substring-locations.h (format_warning_va): Likewise.
12031         (format_warning_at_substring): Likewise.
12033 2017-10-17  Jan Hubicka  <hubicka@ucw.cz>
12035         * target.h (enum vect_cost_for_stmt): Add vec_gather_load and
12036         vec_scatter_store
12037         * tree-vect-stmts.c (record_stmt_cost): Make difference between normal
12038         and scatter/gather ops.
12040         * aarch64/aarch64.c (aarch64_builtin_vectorization_cost): Add
12041         vec_gather_load and vec_scatter_store.
12042         * arm/arm.c (arm_builtin_vectorization_cost): Likewise.
12043         * powerpcspe/powerpcspe.c (rs6000_builtin_vectorization_cost): Likewise.
12044         * rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Likewise.
12045         * s390/s390.c (s390_builtin_vectorization_cost): Likewise.
12046         * spu/spu.c (spu_builtin_vectorization_cost): Likewise.
12047         * i386/i386.c (x86_builtin_vectorization_cost): Likewise.
12049 2017-10-17  Uros Bizjak  <ubizjak@gmail.com>
12051         * reg-stack.c (compare_for_stack_reg): Add bool argument.
12052         Detect FTST instruction and handle its register pops.  Only pop
12053         second operand if can_pop_second_op is true.
12054         (subst_stack_regs_pat) <case COMPARE>: Detect FCOMI instruction to
12055         set can_pop_second_op to false in the compare_for_stack_reg call.
12057         * config/i386/i386.md (*cmpi<FPCMP:unord><MODEF:mode>): Only call
12058         output_fp_compare for stack register operands.
12059         * config/i386/i386.c (output_fp_compare): Do not output SSE compare
12060         instructions here.  Do not emit stack register pops here.  Assert
12061         that FCOMPP pops next to top stack register.  Rewrite function.
12063 2017-10-17  Nathan Sidwell  <nathan@acm.org>
12065         PR middle-end/82577
12066         * alias.c (compare_base_decls): Check HAS_DECL_ASSEMBLER_NAME_P,
12067         use DECL_ASSEMBLER_NAME_RAW.
12069         PR middle-end/82546
12070         * tree.c (tree_code_size): Reformat.  Punt to lang hook for unknown
12071         TYPE nodes.
12073 2017-10-17  Qing Zhao <qing.zhao@oracle.com>
12074             Wilco Dijkstra <wilco.dijkstra@arm.com>
12076         * builtins.c (expand_builtin_update_setjmp_buf): Add a
12077         converstion to Pmode from the buf_addr.
12079 2017-10-17  Richard Biener  <rguenther@suse.de>
12081         * graphite-dependences.c (scop_get_reads_and_writes): Change
12082         output parameters to references.
12084 2017-10-17  Jackson Woodruff  <jackson.woodruff@arm.com>
12086         PR 71026/tree-optimization
12087         * fold-const.c (distribute_real_division): Removed.
12088         (fold_binary_loc): Remove calls to distribute_real_divison.
12090 2017-10-17  Richard Biener  <rguenther@suse.de>
12092         * graphite-scop-detection.c
12093         (scop_detection::stmt_has_simple_data_refs_p): Always use
12094         the full nest as region.
12095         (try_generate_gimple_bb): Likewise.
12096         * sese.c (scalar_evolution_in_region): Simplify now that
12097         SCEV can handle instantiation in regions.
12098         * tree-scalar-evolution.c (instantiate_scev_name): Also instantiate
12099         in the non-loop part of a function if requested.
12101 2017-10-17  Richard Biener  <rguenther@suse.de>
12103         PR tree-optimization/82563
12104         * graphite-isl-ast-to-gimple.c (generate_entry_out_of_ssa_copies):
12105         New function.
12106         (graphite_regenerate_ast_isl): Call it.
12107         * graphite-scop-detection.c (build_scops): Remove entry edge split.
12109 2017-10-17  Jakub Jelinek  <jakub@redhat.com>
12111         PR tree-optimization/82549
12112         * fold-const.c (optimize_bit_field_compare, fold_truth_andor_1):
12113         Formatting fixes.  Instead of calling make_bit_field_ref with negative
12114         bitpos return 0.
12116 2017-10-17  Olga Makhotina  <olga.makhotina@intel.com>
12118         * config/i386/avx512dqintrin.h (_mm_mask_reduce_sd,
12119         _mm_maskz_reduce_sd, _mm_mask_reduce_ss,=20
12120         _mm_maskz_reduce_ss): New.
12121         * config/i386/i386-builtin.def (__builtin_ia32_reducesd_mask,
12122         __builtin_ia32_reducess_mask): Ditto..
12123         (__builtin_ia32_reducesd, __builtin_ia32_reducess): Remove.
12124         * config/i386/sse.md (reduces<mode>): Renamed to ...
12125         (reduces<mode><mask_scalar_name>): ... this.
12126         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0|%0, %1, %2, %3}):
12127         Changed to ...
12128         (vreduce<ssescalarmodesuffix>\t{%3, %2, %1, %0<mask_scalar_operand4>|
12129         %0<mask_scalar_operand4>, %1, %2, %3}): ... this.
12131 2017-10-16  David Malcolm  <dmalcolm@redhat.com>
12133         * Makefile.in (OBJS): Add unique-ptr-tests.o.
12134         * selftest-run-tests.c (selftest::run_tests): Call
12135         selftest::unique_ptr_tests_cc_tests.
12136         * selftest.h (selftest::unique_ptr_tests_cc_tests): New decl.
12137         * unique-ptr-tests.cc: New file.
12139 2017-10-16  Vladimir Makarov  <vmakarov@redhat.com>
12141         PR sanitizer/82353
12142         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12143         locations.
12144         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12145         (make_hard_regno_born, make_hard_regno_dead): Update
12146         bb_killed_pseudos and bb_gen_pseudos for fixed regs.
12148 2017-10-16  Jeff Law  <law@redhat.com>
12150         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12152 2017-10-16  Jan Hubicka  <hubicka@ucw.cz>
12154         * x86-tune-costs.h (znver1_cost): Fix move cost tables.
12156 2017-10-16  Olivier Hainque  <hainque@adacore.com>
12158         * gcc/config.gcc (powerpc*-*-*spe*): Pick 8548 as the default
12159         with_cpu if we were configured for an e500v2 target cpu name.
12161 2017-10-16  Thomas Preud'homme  <thomas.preudhomme@arm.com>
12163         * config/arm/arm-cpus.in (cortex-m33): Add nodsp option.
12164         * doc/invoke.texi: Document +nodsp as a valid extension for
12165         -mcpu=cortex-m33.
12167 2017-10-16  Martin Liska  <mliska@suse.cz>
12169         * sbitmap.c (bitmap_bit_in_range_p_checking): New function.
12170         (test_set_range): Likewise.
12171         (test_range_functions): Rename to ...
12172         (test_bit_in_range): ... this.
12173         (sbitmap_c_tests): Add new test.
12175 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12177         * config/aarch64/arm_neon.h (vdot_u32, vdotq_u32, vdot_s32, vdotq_s32):
12178         New.
12179         (vdot_lane_u32, vdot_laneq_u32, vdotq_lane_u32, vdotq_laneq_u32): New.
12180         (vdot_lane_s32, vdot_laneq_s32, vdotq_lane_s32, vdotq_laneq_s32): New.
12182 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12184         * config/aarch64/aarch64-builtins.c
12185         (aarch64_types_quadopu_lane_qualifiers): New.
12186         (TYPES_QUADOPU_LANE): New.
12187         * config/aarch64/aarch64-simd.md (aarch64_<sur>dot<vsi2qi>): New.
12188         (<sur>dot_prod<vsi2qi>, aarch64_<sur>dot_lane<vsi2qi>): New.
12189         (aarch64_<sur>dot_laneq<vsi2qi>): New.
12190         * config/aarch64/aarch64-simd-builtins.def (sdot, udot): New.
12191         (sdot_lane, udot_lane, sdot_laneq, udot_laneq): New.
12192         * config/aarch64/iterators.md (sur): Add UNSPEC_SDOT, UNSPEC_UDOT.
12193         (Vdottype, DOTPROD): New.
12194         (sur): Add SDOT and UDOT.
12196 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12198         * config/aarch64/aarch64.h (AARCH64_FL_DOTPROD): New.
12199         (AARCH64_ISA_DOTPROD, TARGET_DOTPROD): New.
12200         * config/aarch64/aarch64-c.c (aarch64_update_cpp_builtins):
12201         Add TARGET_DOTPROD.
12202         * config/aarch64/aarch64-option-extensions.def (dotprod): New.
12203         * config/aarch64/aarch64-cores.def (cortex-a55, cortex-a75):
12204         Enable TARGET_DOTPROD.
12205         (cortex-a75.cortex-a55): Likewise.
12206         * doc/invoke.texi (aarch64-feature-modifiers): Document dotprod.
12208 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12210         * config/arm/arm-builtins.c (arm_unsigned_uternop_qualifiers): New.
12211         (UTERNOP_QUALIFIERS, arm_umac_lane_qualifiers, UMAC_LANE_QUALIFIERS):
12212         New.
12213         * config/arm/arm_neon_builtins.def (sdot, udot, sdot_lane, udot_lane):
12214         New.
12215         * config/arm/iterators.md (DOTPROD, VSI2QI, vsi2qi): New.
12216         (UNSPEC_DOT_S, UNSPEC_DOT_U, opsuffix): New.
12217         * config/arm/neon.md (neon_<sup>dot<vsi2qi>): New.
12218         (neon_<sup>dot_lane<vsi2qi>, <sup>dot_prod<vsi2qi>): New.
12219         * config/arm/types.md (neon_dot, neon_dot_q): New.
12220         * config/arm/unspecs.md (sup): Add UNSPEC_DOT_S, UNSPEC_DOT_U.
12222 2017-10-16  Tamar Christina  <tamar.christina@arm.com>
12224         * config/arm/arm.h (TARGET_DOTPROD): New.
12225         * config/arm/arm.c (arm_arch_dotprod): New.
12226         (arm_option_reconfigure_globals): Add arm_arch_dotprod.
12227         * config/arm/arm-c.c (__ARM_FEATURE_DOTPROD): New.
12228         * config/arm/arm-cpus.in (armv8.2-a): Enabled +dotprod.
12229         (feature dotprod, group dotprod, ALL_SIMD_INTERNAL): New.
12230         (ALL_FPU_INTERNAL): Use ALL_SIMD_INTERNAL.
12231         * config/arm/t-multilib (v8_2_a_simd_variants): Add dotprod.
12232         * doc/invoke.texi (armv8.2-a): Document dotprod
12234 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12236         * i386.c (ix86_vec_cost): New function.
12237         (ix86_rtx_costs): Handle vector operations better.
12238         * i386.h (struct processor_costs): Add sse_op, fmasd, fmass.
12239         * x86-tune-costs.h: Add new costs to all tables.
12241 2017-10-14  Jan Hubicka  <hubicka@ucw.cz>
12243         * i386.c (ix86_rtx_costs): Make difference between x87 and SSE
12244         operations.
12245         * i386.h (struct processor_costs): Add addss, mulss, mulsd, divss,
12246         divsd, sqrtss and sqrtsd
12247         * x86-tune-costs.h: Add new entries to all costs.
12248         (znver1_cost): Fix to match real instruction latencies.
12250 2017-10-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
12251             Michael Collison <michael.collison@arm.com>
12253         * compare-elim.c: Include emit-rtl.h.
12254         (can_merge_compare_into_arith): New function.
12255         (try_validate_parallel): Likewise.
12256         (try_merge_compare): Likewise.
12257         (try_eliminate_compare): Call the above when no previous clobber
12258         is available.
12259         (execute_compare_elim_after_reload): Add DF_UD_CHAIN and DF_DU_CHAIN
12260         dataflow problems.
12262 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12264         PR middle-end/62263
12265         PR middle-end/82498
12266         * tree-ssa-phiopt.c (value_replacement): Comment fix.  Handle
12267         up to 2 preparation statements for ASSIGN in MIDDLE_BB.
12269         PR middle-end/62263
12270         PR middle-end/82498
12271         * tree-ssa-forwprop.c (simplify_rotate): Allow def_arg1[N]
12272         to be any operand_equal_p operands.  For & (B - 1) require
12273         B to be power of 2.  Recognize
12274         (X << (Y & (B - 1))) | (X >> ((-Y) & (B - 1))) and similar patterns.
12276 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12278         PR bootstrap/82553
12279         * optabs.c (expand_memory_blockage): Fix call of
12280         targetm.have_memory_blockage.
12282 2017-10-14  Jakub Jelinek  <jakub@redhat.com>
12284         PR bootstrap/82548
12285         * config.gcc (*-*-solaris2*, i[34567]86-*-cygwin*,
12286         x86_64-*-cygwin*, i[34567]86-*-mingw* | x86_64-*-mingw*): Append
12287         objects to extra_objs instead of overwriting it.
12289 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12291         * config/i386/sync.md (FILD_ATOMIC/FIST_ATOMIC FP load peephole2):
12292         Use any_fp_register_operand as operand[3] predicate.  Simplify
12293         equality test for operands[2] and operands[4] memory location.
12294         (LDX_ATOMIC/STX_ATOMIC FP load peephole2): Ditto.
12295         (FILD_ATOMIC/FIST_ATOMIC FP load peephole2 with mem blockage): New.
12296         (LDX_ATOMIC/LDX_ATOMIC FP load peephole2 with mem blockage): Ditto.
12297         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2): Use
12298         any_fp_register_operand as operand[1] predicate.  Simplify
12299         equality test for operands[0] and operands[3] memory location.
12300         (LDX_ATOMIC/STX_ATOMIC FP store peephole2): Ditto.
12301         (FILD_ATOMIC/FIST_ATOMIC FP store peephole2 with mem blockage): New.
12302         (LDX_ATOMIC/LDX_ATOMIC FP storepeephole2 with mem blockage): Ditto.
12304 2017-10-14  Uros Bizjak  <ubizjak@gmail.com>
12306         * target-insns.def: Add memory_blockage.
12307         * optabs.c (expand_memory_blockage): New function.
12308         (expand_asm_memory_barrier): Rename ...
12309         (expand_asm_memory_blockage): ... to this.
12310         (expand_mem_thread_fence): Call expand_memory_blockage
12311         instead of expand_asm_memory_barrier.
12312         (expand_mem_singnal_fence): Ditto.
12313         (expand_atomic_load): Ditto.
12314         (expand_atomic_store): Ditto.
12315         * doc/md.texi (Standard Pattern Names For Generation):
12316         Document memory_blockage instruction pattern.
12318 2017-10-13  Sebastian Perta  <sebastian.perta@renesas.com>
12320         * config/rl78/rl78.c (rl78_emit_libcall): New function.
12321         * config/rl78/rl78-protos.h (rl78_emit_libcall): New function.
12322         * config/rl78/rl78.md: New define_expand "adddi3".
12324 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12326         * cfghooks.c (verify_flow_info): Disable check that all probabilities
12327         are set correctly.
12329 2017-10-13  Jeff Law  <law@redhat.com>
12331         * tree-ssa-reassoc.c (reassociate_bb): Clarify code slighly.
12333 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12335         PR target/82274
12336         * internal-fn.c (expand_mul_overflow): If both operands have
12337         the same highpart of -1 or 0 and the topmost bit of lowpart
12338         is different, overflow is if res <= 0 rather than res < 0.
12340 2017-10-13  Pat Haugen  <pthaugen@us.ibm.com>
12342         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost): Remove
12343         TARGET_P9_VECTOR code for unaligned_load case.
12345 2017-10-13  Jan Hubicka  <hubicka@ucw.cz>
12347         * cfghooks.c (verify_flow_info): Check that edge probabilities are set.
12349 2017-10-13  Nathan Sidwell  <nathan@acm.org>
12351         * tree-core.h (tree_contains_struct): Make bool.
12352         * tree.c (tree_contains_struct): Likewise.
12353         * tree.h (MARK_TS_BASE): Remove do ... while (0) idiom.
12354         (MARK_TS_TYPED, MARK_TS_COMMON, MARK_TS_TYPE_COMMON,
12355         MARK_TS_TYPE_WITH_LANG_SPECIFIC, MARK_TS_DECL_MINIMAL,
12356         MARK_TS_DECL_COMMON, MARK_TS_DECL_WRTL, MARK_TS_DECL_WITH_VIS,
12357         MARK_TS_DECL_NON_COMMON): Likewise, use comma operator.
12359 2017-10-13  Richard Biener  <rguenther@suse.de>
12361         * graphite-isl-ast-to-gimple.c
12362         (translate_isl_ast_to_gimple::get_rename_from_scev): Remove unused
12363         parameters and dominance check.
12364         (translate_isl_ast_to_gimple::graphite_copy_stmts_from_block): Adjust.
12365         (translate_isl_ast_to_gimple::copy_bb_and_scalar_dependences): Likewise.
12366         (translate_isl_ast_to_gimple::graphite_regenerate_ast_isl):
12367         Do not update SSA form here or do intermediate IL verification.
12368         * graphite.c: Include tree-ssa.h and tree-into-ssa.h.
12369         (graphite_initialize): Remove check on the number of loops in
12370         the function and inline into graphite_transform_loops.
12371         (graphite_finalize): Inline into graphite_transform_loops.
12372         (graphite_transform_loops): Perform SSA update and IL verification
12373         here.
12374         * params.def (PARAM_GRAPHITE_MIN_LOOPS_PER_FUNCTION): Remove.
12376 2017-10-13  Richard Biener  <rguenther@suse.de>
12378         * graphite-isl-ast-to-gimple.c (max_mode_int_precision,
12379         graphite_expression_type_precision): Avoid global constructor
12380         by moving ...
12381         (translate_isl_ast_to_gimple::translate_isl_ast_to_gimple): Here.
12382         (translate_isl_ast_to_gimple::graphite_expr_type): Add type member.
12383         (translate_isl_ast_to_gimple::translate_isl_ast_node_for): Use it.
12384         (translate_isl_ast_to_gimple::build_iv_mapping): Likewise.
12385         (translate_isl_ast_to_gimple::graphite_create_new_guard): Likewise.
12386         * graphite-sese-to-poly.c (build_original_schedule): Return nothing.
12388 2017-10-13  H.J. Lu  <hongjiu.lu@intel.com>
12390         PR target/82499
12391         * config/i386/i386.h (ix86_red_zone_size): New.
12392         * config/i386/i386.md (push peephole2s): Replace
12393         "!ix86_using_red_zone ()" with "ix86_red_zone_size == 0".
12395 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12396             Alan Hayward  <alan.hayward@arm.com>
12397             David Sherwood  <david.sherwood@arm.com>
12399         * combine.c (can_change_dest_mode): Reject changes in
12400         REGMODE_NATURAL_SIZE.
12402 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12403             Alan Hayward  <alan.hayward@arm.com>
12404             David Sherwood  <david.sherwood@arm.com>
12406         * cfgexpand.c (expand_debug_expr): Use GET_MODE_UNIT_BITSIZE.
12407         (expand_debug_source_expr): Likewise.
12408         * combine.c (combine_simplify_rtx): Likewise.
12409         * cse.c (fold_rtx): Likewise.
12410         * fwprop.c (canonicalize_address): Likewise.
12411         * targhooks.c (default_shift_truncation_mask): Likewise.
12413 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12414             Alan Hayward  <alan.hayward@arm.com>
12415             David Sherwood  <david.sherwood@arm.com>
12417         * optabs.c (add_equal_note): Use GET_MODE_UNIT_SIZE.
12418         (widened_mode): Likewise.
12419         (expand_unop): Likewise.
12420         * ree.c (transform_ifelse): Likewise.
12421         (merge_def_and_ext): Likewise.
12422         (combine_reaching_defs): Likewise.
12423         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
12425 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12426             Alan Hayward  <alan.hayward@arm.com>
12427             David Sherwood  <david.sherwood@arm.com>
12429         * caller-save.c (replace_reg_with_saved_mem): Use byte_lowpart_offset.
12430         * combine.c (gen_lowpart_for_combine): Likewise.
12431         * dwarf2out.c (rtl_for_decl_location): Likewise.
12432         * final.c (alter_subreg): Likewise.
12433         * rtlhooks.c (gen_lowpart_general): Likewise.
12434         (gen_lowpart_if_possible): Likewise.
12436 2017-10-13  Richard Sandiford  <richard.sandiford@linaro.org>
12437             Alan Hayward  <alan.hayward@arm.com>
12438             David Sherwood  <david.sherwood@arm.com>
12440         * calls.c (expand_call): Use subreg_lowpart_offset.
12441         * cse.c (cse_insn): Likewise.
12442         * regcprop.c (copy_value): Likewise.
12443         (copyprop_hardreg_forward_1): Likewise.
12445 2017-10-13  Jakub Jelinek  <jakub@redhat.com>
12447         PR target/82524
12448         * config/i386/i386.md (addqi_ext_1, andqi_ext_1,
12449         *andqi_ext_1_cc, *<code>qi_ext_1, *xorqi_ext_1_cc): Change
12450         =Q constraints to +Q and into insn condition add check
12451         that operands[0] and operands[1] are equal.
12452         (*addqi_ext_2, *andqi_ext_2, *<code>qi_ext_2): Change
12453         =Q constraints to +Q and into insn condition add check
12454         that operands[0] is equal to either operands[1] or operands[2].
12456         PR target/82498
12457         * fold-const.c (fold_binary_loc) <bit_rotate>: Code cleanups,
12458         instead of handling MINUS_EXPR twice (once for each argument),
12459         canonicalize operand order and handle just once, use rtype where
12460         possible.  Handle (A << B) | (A >> (-B & (Z - 1))).
12462         PR target/82498
12463         * config/i386/ia32intrin.h (__rold, __rord, __rolq, __rorq): Allow
12464         any values of __C while still being pattern recognizable as a simple
12465         rotate instruction.
12467 2017-10-13  Richard Biener  <rguenther@suse.de>
12469         PR tree-optimization/82451
12470         Revert
12471         2017-10-02  Richard Biener  <rguenther@suse.de>
12473         PR tree-optimization/82355
12474         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
12475         a mapping for the enclosing loop but avoid generating one for
12476         the loop tree root.
12477         (copy_bb_and_scalar_dependences): Remove premature codegen
12478         error on PHIs in blocks duplicated into multiple places.
12479         * graphite-scop-detection.c
12480         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
12481         in the region use it as loop and nest to analyze the DR in.
12482         (try_generate_gimple_bb): Likewise.
12483         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
12484         (add_loop_constraints): For blocks in a loop not in the region
12485         create a dimension with a single iteration.
12486         * sese.h (gbb_loop_at_index): Remove assert.
12488         * cfgloop.c (loop_preheader_edge): For the loop tree root
12489         return the single successor of the entry block.
12490         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
12491         Reset the SCEV hashtable and niters.
12492         * graphite-scop-detection.c
12493         (scop_detection::graphite_can_represent_scev): Add SCOP parameter,
12494         assert that we only have POLYNOMIAL_CHREC that vary in loops
12495         contained in the region.
12496         (scop_detection::graphite_can_represent_expr): Adjust.
12497         (scop_detection::stmt_has_simple_data_refs_p): For loops
12498         not in the region set loop to NULL.  The nest is now the
12499         entry edge to the region.
12500         (try_generate_gimple_bb): Likewise.
12501         * sese.c (scalar_evolution_in_region): Adjust for
12502         instantiate_scev change.
12503         * tree-data-ref.h (graphite_find_data_references_in_stmt):
12504         Make nest parameter the edge into the region.
12505         (create_data_ref): Likewise.
12506         * tree-data-ref.c (dr_analyze_indices): Make nest parameter an
12507         entry edge into a region and adjust instantiate_scev calls.
12508         (create_data_ref): Likewise.
12509         (graphite_find_data_references_in_stmt): Likewise.
12510         (find_data_references_in_stmt): Pass the loop preheader edge
12511         from the nest argument.
12512         * tree-scalar-evolution.h (instantiate_scev): Make instantiate_below
12513         parameter the edge into the region.
12514         (instantiate_parameters): Use the loop preheader edge as entry.
12515         * tree-scalar-evolution.c (analyze_scalar_evolution): Handle
12516         NULL loop.
12517         (get_instantiated_value_entry): Make instantiate_below parameter
12518         the edge into the region.
12519         (instantiate_scev_name): Likewise.  Adjust dominance checks,
12520         when we cannot use loop-based instantiation instantiate by
12521         walking use-def chains.
12522         (instantiate_scev_poly): Adjust.
12523         (instantiate_scev_binary): Likewise.
12524         (instantiate_scev_convert): Likewise.
12525         (instantiate_scev_not): Likewise.
12526         (instantiate_array_ref): Remove.
12527         (instantiate_scev_3): Likewise.
12528         (instantiate_scev_2): Likewise.
12529         (instantiate_scev_1): Likewise.
12530         (instantiate_scev_r): Do not blindly handle N-operand trees.
12531         Do not instantiate array-refs.  Handle all constants and invariants.
12532         (instantiate_scev): Make instantiate_below parameter
12533         the edge into the region.
12534         (resolve_mixers): Use the loop preheader edge for the region
12535         parameter to instantiate_scev_r.
12536         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Adjust.
12538 2017-10-13  Richard Biener  <rguenther@suse.de>
12540         PR tree-optimization/82525
12541         * graphite-isl-ast-to-gimple.c
12542         (translate_isl_ast_to_gimple::widest_int_from_isl_expr_int): Split
12543         out from ...
12544         (translate_isl_ast_to_gimple::gcc_expression_from_isl_expr_int): Here.
12545         Fail code generation when we cannot represent the isl integer.
12546         (binary_op_to_tree): Elide modulo operations that are no-ops
12547         in the type we code generate.  Remove now superfluous code
12548         generation errors.
12550 2017-10-13  Richard Biener  <rguenther@suse.de>
12552         * graphite-scop-detection.c (loop_ivs_can_be_represented): Remove.
12553         (scop_detection::harmful_loop_in_region): Remove premature
12554         IV type restriction.
12555         (scop_detection::graphite_can_represent_scev): We can handle
12556         pointer IVs just fine.
12558 2017-10-13  Alan Modra  <amodra@gmail.com>
12560         * doc/extend.texi (Extended Asm <Clobbers>): Rename to
12561         "Clobbers and Scratch Registers".  Add paragraph on
12562         alternative to clobbers for scratch registers and OpenBLAS
12563         example.
12565 2017-10-13  Alan Modra  <amodra@gmail.com>
12567         * doc/extend.texi (Clobbers): Correct vax example.  Delete old
12568         example of a memory input for a string of known length.  Move
12569         commentary out of table.  Add a number of new examples
12570         covering array memory inputs.
12572 2017-10-12  Martin Liska  <mliska@suse.cz>
12574         PR tree-optimization/82493
12575         * sbitmap.c (bitmap_bit_in_range_p): Fix the implementation.
12576         (test_range_functions): New function.
12577         (sbitmap_c_tests): Likewise.
12578         * selftest-run-tests.c (selftest::run_tests): Run new tests.
12579         * selftest.h (sbitmap_c_tests): New function.
12581         * tree-ssa-dse.c (live_bytes_read): Fix thinko.
12583 2017-10-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
12585         * config/rs6000/amo.h: Fix spacing issue.
12587 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12589         PR target/82498
12590         * config/i386/i386.md (*ashl<mode>3_mask_1,
12591         *<shift_insn><mode>3_mask_1, *<rotate_insn><mode>3_mask_1,
12592         *<btsc><mode>_mask_1, *btr<mode>_mask_1): New define_insn_and_split
12593         patterns.
12595 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12597         * profile-count.h (safe_scale_64bit): Fix GCC4.x path.
12598         (profile_probability): Set max_probability
12599         to (uint32_t) 1 << (n_bits - 2) and update accessors to avoid overlfows
12600         in temporaries.
12601         * profile-count.c (profile_probability::differs_from_p): Do not
12602         rely on max_probaiblity == 10000
12604 2017-10-12  Jeff Law  <law@redhat.com>
12606         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject ao_refs with
12607         negative offsets.
12609 2017-10-12  Martin Sebor  <msebor@redhat.com>
12611         PR other/82301
12612         PR c/82435
12613         * cgraphunit.c (maybe_diag_incompatible_alias): New function.
12614         (handle_alias_pairs): Call it.
12615         * common.opt (-Wattribute-alias): New option.
12616         * doc/extend.texi (ifunc attribute): Discuss C++ specifics.
12617         * doc/invoke.texi (-Wattribute-alias): Document.
12619 2017-10-12  Vladimir Makarov  <vmakarov@redhat.com>
12621         Revert
12622         2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12623         PR sanitizer/82353
12624         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12625         locations.
12626         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12627         (make_hard_regno_born, make_hard_regno_dead): Update
12628         bb_killed_pseudos and bb_gen_pseudos.
12630 2017-10-12  Jan Hubicka  <hubicka@ucw.cz>
12632         * config/i386/x86-tune-sched.c (ix86_adjust_cost): Fix Zen support.
12634 2017-10-12  Uros Bizjak  <ubizjak@gmail.com>
12636         * config/alpha/alpha.c (alpha_split_conditional_move):
12637         Use std::swap instead of manually swapping.
12638         (alpha_stdarg_optimize_hook): Ditto.
12639         (alpha_canonicalize_comparison): Ditto.
12641 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12643         * tree-loop-distribution.c (struct builtin_info): New struct.
12644         (struct partition): Refactor fields into struct builtin_info.
12645         (partition_free): Free struct builtin_info.
12646         (build_size_arg_loc, build_addr_arg_loc): Delete.
12647         (generate_memset_builtin, generate_memcpy_builtin): Get memory range
12648         information from struct builtin_info.
12649         (find_single_drs): New function refactored from classify_partition.
12650         Also moved builtin validity checks to this function.
12651         (compute_access_range, alloc_builtin): New functions.
12652         (classify_builtin_st, classify_builtin_ldst): New functions.
12653         (classify_partition): Refactor code into functions find_single_drs,
12654         classify_builtin_st and classify_builtin_ldst.
12655         (distribute_loop): Don't do runtime alias check when distributing
12656         loop nest.
12657         (find_seed_stmts_for_distribution): New function.
12658         (pass_loop_distribution::execute): Refactor code finding seed
12659         stmts into above function.  Support distribution for the innermost
12660         two-level loop nest.  Adjust dump information.
12662 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12664         * tree-loop-distribution.c: Adjust the general comment.
12665         (NUM_PARTITION_THRESHOLD): New macro.
12666         (ssa_name_has_uses_outside_loop_p): Support loop nest distribution.
12667         (classify_partition): Skip builtin pattern of loop nest's inner loop.
12668         (merge_dep_scc_partitions): New parameter ignore_alias_p and use it
12669         in call to build_partition_graph.
12670         (finalize_partitions): New parameter.  Make loop distribution more
12671         conservative by fusing more partitions.
12672         (distribute_loop): Don't do runtime alias check in case of loop nest
12673         distribution.
12674         (find_seed_stmts_for_distribution): New function.
12675         (prepare_perfect_loop_nest): New function.
12676         (pass_loop_distribution::execute): Refactor code finding seed stmts
12677         and loop nest into above functions.  Support loop nest distribution.
12678         Adjust dump information accordingly.
12680 2017-10-12  Bin Cheng  <bin.cheng@arm.com>
12682         * tree-loop-distribution.c (break_alias_scc_partitions): Add comment
12683         and set PTYPE_SEQUENTIAL for merged partition.
12685 2017-10-12  Richard Biener  <rguenther@suse.de>
12687         PR tree-optimization/69728
12688         Revert
12689         2017-09-19  Richard Biener  <rguenther@suse.de>
12691         PR tree-optimization/69728
12692         * graphite-sese-to-poly.c (schedule_error): New global.
12693         (add_loop_schedule): Handle empty domain by failing the
12694         schedule.
12695         (build_original_schedule): Handle schedule_error.
12697         * graphite-sese-to-poly.c (add_loop_schedule): Handle empty
12698         domain by returning an unchanged schedule.
12700 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12702         * genrecog.c (validate_pattern): For VEC_SELECT verify that
12703         CONST_INT selectors are 0 to GET_MODE_NUNITS (imode) - 1.
12705 2017-10-12  Aldy Hernandez  <aldyh@redhat.com>
12707         * Makefile.in (TAGS): Merge all the *.def files into one pattern.
12708         Handle params.def.
12710 2017-10-12  Jakub Jelinek  <jakub@redhat.com>
12712         PR c++/82159
12713         * expr.c (store_field): Don't optimize away bitsize == 0 store
12714         from CALL_EXPR with addressable return type.
12716 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12718         * config/rs6000/rs6000.h (TARGET_ISEL64): Delete.
12719         * config/rs6000/rs6000.md (sel): Delete mode attribute.
12720         (mov<mode>cc, isel_signed_<mode>, isel_unsigned_<mode>,
12721         *isel_reversed_signed_<mode>, *isel_reversed_unsigned_<mode>): Use
12722         TARGET_ISEL instead of TARGET_ISEL<sel>.
12724 2017-10-11  David Edelsohn  <dje.gcc@gmail.com>
12726         * config/rs6000/rs6000.c
12727         (rs6000_xcoff_asm_output_aligned_decl_common): Test for NULL decl.
12729 2017-10-11  Segher Boessenkool  <segher@kernel.crashing.org>
12731         * config/rs6000/predicates.md (zero_constant, all_ones_constant):
12732         Move up in file.
12733         (reg_or_cint_operand): Fix comment.
12734         (reg_or_zero_operand): New predicate.
12735         * config/rs6000/rs6000-protos.h (output_isel): Delete.
12736         * config/rs6000/rs6000.c (output_isel): Delete.
12737         * config/rs6000/rs6000.md (isel_signed_<mode>): Use reg_or_zero_operand
12738         instead of reg_or_cint_operand.  Output instruction directly (not via
12739         output_isel).
12740         (isel_unsigned_<mode>): Ditto.
12741         (*isel_reversed_signed_<mode>): Use reg_or_zero_operand instead of
12742         gpc_reg_operand.  Add an instruction alternative for this.  Output
12743         instruction directly.
12744         (*isel_reversed_unsigned_<mode>): Ditto.
12746 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12748         * config/i386/i386.c (ix86_canonicalize_comparison): New function.
12749         (TARGET_CANONICALIZE_COMPARISON): Define.
12751 2017-10-11  Qing Zhao  <qing.zhao@oracle.com>
12753         PR target/81422
12754         * config/aarch64/aarch64.c (aarch64_load_symref_appropriately):
12755         Check whether the dest is REG before adding REG_EQUIV note.
12757 2017-10-11  Vladimir Makarov  <vmakarov@redhat.com>
12759         PR sanitizer/82353
12760         * lra.c (collect_non_operand_hard_regs): Don't ignore operator
12761         locations.
12762         * lra-lives.c (bb_killed_pseudos, bb_gen_pseudos): Move up.
12763         (make_hard_regno_born, make_hard_regno_dead): Update
12764         bb_killed_pseudos and bb_gen_pseudos.
12766 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12768         * incpath.h (enum incpath_kind): Name enum, prefix values.
12769         (add_path, add_cpp_dir_path, get_added_cpp_dirs): Use incpath_kind.
12770         * incpath.c (heads, tails): Use INC_MAX.
12771         (add_env_var_paths, add_standard_paths): Use incpath_kind.
12772         (merge_include_chains, split_quote_chain,
12773         register_include_chains): Update incpath_kind names.
12774         (add_cpp_dir_path, add_path, get_added_cpp_dirs): Use incpath_kind.
12775         * config/darwin-c.c (add_system_framework_path): Update incpath_kind
12776         names.
12777         (add_framework_path, darwin_register_objc_includes): Likewise.
12778         * config/vms/vms-c.c (vms_c_register_includes): Likewise.
12780 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12782         * config/i386/i386.md (*cmp<X87MODEF:mode>_<SWI24:mode>_i387):
12783         Do not use float_operator operator predicate.
12784         (*cmp<X87MODEF:mode>_<SWI24:mode>_cc_i387): Ditto.
12785         * config/i386/predicates.md (float_operator): Remove predicate.
12787 2017-10-11  Uros Bizjak  <ubizjak@gmail.com>
12789         * config/i386/i386.md (*jcc<mode>_0_i387): Remove insn pattern.
12790         (*jccxf_i387): Ditto.
12791         (*jcc<mode>_i387): Ditto.
12792         (*jccu<mode>_i387): Ditto.
12793         (*jcc<X87MODEF:mode>_<SWI24:mode>_i387): Ditto.
12794         (*jcc_*_i387 splitters): Remove.
12795         * config/i386/i386-protos.h (ix86_split_fp_branch): Remove prototype.
12796         * config/i386/i386.c (ix86_split_fp_branch): Remove.
12797         * config/i386/predicates.md (ix86_swapped_fp_comparison_operator):
12798         Remove predicate.
12800 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12802         * profile-count.h (slow_safe_scale_64bit): New function.
12803         (safe_scale_64bit): New inline.
12804         (profile_count::max_safe_multiplier): Remove; use safe_scale_64bit.
12805         * profile-count.c: Include wide-int.h
12806         (slow_safe_scale_64bit): New.
12808 2017-10-11  Nathan Sidwell  <nathan@acm.org>
12810         * tree.h (DECL_ASSEMBLER_NAME_SET_P): Don't check
12811         HAS_DECL_ASSEMBLER_NAME_P.
12812         * gimple-expr.c (gimple_decl_printable_name: Check
12813         HAS_DECL_ASSEMBLER_NAME_P too.
12814         * ipa-utils.h (type_in_anonymous_namespace_p): Check
12815         DECL_ASSEMBLER_NAME_SET_P of TYPE_NAME.
12816         (odr_type_p): No need to assert TYPE_NAME is a TYPE_DECL.
12817         * passes.c (rest_of_decl_compilation): Check
12818         HAS_DECL_ASSEMBLER_NAME_P too.
12819         * recog.c (verify_changes): Likewise.
12820         * tree-pretty-print.c (dump_decl_name): Likewise.
12821         * tree-ssa-structalias.c (alias_get_name): Likewise.  Reimplement.
12823         * tree.h (DECL_ASSEMBLER_NAME_RAW): New.
12824         (SET_DECL_ASSEMBLER_NAME): Use it.
12825         (DECL_ASSEMBLER_NAME_SET_P): Likewise.
12826         (COPY_DECL_ASSEMBLER_NAME): Likewise.
12827         * tree.c (decl_assembler_name): Use DECL_ASSEMBLER_NAME_RAW.
12829 2017-10-11  Jan Hubicka  <hubicka@ucw.cz>
12831         * config.gcc (i386, x86_64): Add extra objects.
12832         * config/i386/i386-protos.h (ix86_rip_relative_addr_p): Declare.
12833         (ix86_min_insn_size): Declare.
12834         (ix86_issue_rate): Declare.
12835         (ix86_adjust_cost): Declare.
12836         (ia32_multipass_dfa_lookahead): Declare.
12837         (ix86_macro_fusion_p): Declare.
12838         (ix86_macro_fusion_pair_p): Declare.
12839         (ix86_bd_has_dispatch): Declare.
12840         (ix86_bd_do_dispatch): Declare.
12841         (ix86_core2i7_init_hooks): Declare.
12842         (ix86_atom_sched_reorder): Declare.
12843         * config/i386/i386.c Move all CPU cost tables to x86-tune-costs.h.
12844         (COSTS_N_BYTES): Move to x86-tune-costs.h.
12845         (DUMMY_STRINGOP_ALGS):Move to x86-tune-costs.h.
12846         (rip_relative_addr_p): Rename to ...
12847         (ix86_rip_relative_addr_p): ... this one; export.
12848         (memory_address_length): Update.
12849         (ix86_issue_rate): Move to x86-tune-sched.c.
12850         (ix86_flags_dependent): Move to x86-tune-sched.c.
12851         (ix86_agi_dependent): Move to x86-tune-sched.c.
12852         (exact_dependency_1): Move to x86-tune-sched.c.
12853         (exact_store_load_dependency): Move to x86-tune-sched.c.
12854         (ix86_adjust_cost): Move to x86-tune-sched.c.
12855         (ia32_multipass_dfa_lookahead): Move to x86-tune-sched.c.
12856         (ix86_macro_fusion_p): Move to x86-tune-sched.c.
12857         (ix86_macro_fusion_pair_p): Move to x86-tune-sched.c.
12858         (do_reorder_for_imul): Move to x86-tune-sched-atom.c.
12859         (swap_top_of_ready_list): Move to x86-tune-sched-atom.c.
12860         (ix86_sched_reorder): Move to x86-tune-sched-atom.c.
12861         (core2i7_first_cycle_multipass_init): Move to x86-tune-sched-core.c.
12862         (core2i7_dfa_post_advance_cycle): Move to x86-tune-sched-core.c.
12863         (min_insn_size): Rename to ...
12864         (ix86_min_insn_size): ... this one; export.
12865         (core2i7_first_cycle_multipass_begin): Move to x86-tune-sched-core.c.
12866         (core2i7_first_cycle_multipass_issue): Move to x86-tune-sched-core.c.
12867         (core2i7_first_cycle_multipass_backtrack): Move to
12868         x86-tune-sched-core.c.
12869         (core2i7_first_cycle_multipass_end): Move to x86-tune-sched-core.c.
12870         (core2i7_first_cycle_multipass_fini): Move to x86-tune-sched-core.c.
12871         (ix86_sched_init_global): Break up logic to ix86_core2i7_init_hooks.
12872         (ix86_avoid_jump_mispredicts): Update.
12873         (TARGET_SCHED_DISPATCH): Move to ix86-tune-sched-bd.c.
12874         (TARGET_SCHED_DISPATCH_DO): Move to ix86-tune-sched-bd.c.
12875         (TARGET_SCHED_REORDER): Move to ix86-tune-sched-bd.c.
12876         (DISPATCH_WINDOW_SIZE): Move to ix86-tune-sched-bd.c.
12877         (MAX_DISPATCH_WINDOWS): Move to ix86-tune-sched-bd.c.
12878         (MAX_INSN): Move to ix86-tune-sched-bd.c.
12879         (MAX_IMM): Move to ix86-tune-sched-bd.c.
12880         (MAX_IMM_SIZE): Move to ix86-tune-sched-bd.c.
12881         (MAX_IMM_32): Move to ix86-tune-sched-bd.c.
12882         (MAX_IMM_64): Move to ix86-tune-sched-bd.c.
12883         (MAX_LOAD): Move to ix86-tune-sched-bd.c.
12884         (MAX_STORE): Move to ix86-tune-sched-bd.c.
12885         (BIG): Move to ix86-tune-sched-bd.c.
12886         (enum dispatch_group): Move to ix86-tune-sched-bd.c.
12887         (enum insn_path): Move to ix86-tune-sched-bd.c.
12888         (get_mem_group): Move to ix86-tune-sched-bd.c.
12889         (is_cmp): Move to ix86-tune-sched-bd.c.
12890         (dispatch_violation): Move to ix86-tune-sched-bd.c.
12891         (is_branch): Move to ix86-tune-sched-bd.c.
12892         (is_prefetch): Move to ix86-tune-sched-bd.c.
12893         (init_window): Move to ix86-tune-sched-bd.c.
12894         (allocate_window): Move to ix86-tune-sched-bd.c.
12895         (init_dispatch_sched): Move to ix86-tune-sched-bd.c.
12896         (is_end_basic_block): Move to ix86-tune-sched-bd.c.
12897         (process_end_window): Move to ix86-tune-sched-bd.c.
12898         (allocate_next_window): Move to ix86-tune-sched-bd.c.
12899         (find_constant): Move to ix86-tune-sched-bd.c.
12900         (get_num_immediates): Move to ix86-tune-sched-bd.c.
12901         (has_immediate): Move to ix86-tune-sched-bd.c.
12902         (get_insn_path): Move to ix86-tune-sched-bd.c.
12903         (get_insn_group): Move to ix86-tune-sched-bd.c.
12904         (count_num_restricted): Move to ix86-tune-sched-bd.c.
12905         (fits_dispatch_window): Move to ix86-tune-sched-bd.c.
12906         (add_insn_window): Move to ix86-tune-sched-bd.c.
12907         (add_to_dispatch_window): Move to ix86-tune-sched-bd.c.
12908         (debug_dispatch_window_file): Move to ix86-tune-sched-bd.c.
12909         (debug_dispatch_window): Move to ix86-tune-sched-bd.c.
12910         (debug_insn_dispatch_info_file): Move to ix86-tune-sched-bd.c.
12911         (debug_ready_dispatch): Move to ix86-tune-sched-bd.c.
12912         (do_dispatch): Move to ix86-tune-sched-bd.c.
12913         (has_dispatch): Move to ix86-tune-sched-bd.c.
12914         * config/i386/t-i386: Add new object files.
12915         * config/i386/x86-tune-costs.h: New file.
12916         * config/i386/x86-tune-sched-atom.c: New file.
12917         * config/i386/x86-tune-sched-bd.c: New file.
12918         * config/i386/x86-tune-sched-core.c: New file.
12919         * config/i386/x86-tune-sched.c: New file.
12921 2017-10-11  Liu Hao  <lh_mouse@126.com>
12923         * pretty-print.c [_WIN32] (colorize_init): Remove.  Use
12924         the generic version below instead.
12925         (should_colorize): Recognize Windows consoles as terminals
12926         for MinGW targets.
12927         * pretty-print.c [__MINGW32__] (write_all): New function.
12928         [__MINGW32__] (find_esc_head): Likewise.
12929         [__MINGW32__] (find_esc_terminator): Likewise.
12930         [__MINGW32__] (eat_esc_sequence): Likewise.
12931         [__MINGW32__] (mingw_ansi_fputs): New function that handles
12932         ANSI escape codes.
12933         (pp_write_text_to_stream): Use mingw_ansi_fputs instead of fputs
12934         for MinGW targets.
12936 2017-10-11  Richard Biener  <rguenther@suse.de>
12938         * tree-ssa-loop-niter.c (infer_loop_bounds_from_pointer_arith):
12939         Properly call analyze_scalar_evolution with the loop of the stmt.
12941 2017-10-11  Richard Biener  <rguenther@suse.de>
12943         * tree.def (POLYNOMIAL_CHREC): Remove CHREC_VARIABLE tree operand.
12944         * tree-core.h (tree_base): Add chrec_var union member.
12945         * tree.h (CHREC_VAR): Remove.
12946         (CHREC_LEFT, CHREC_RIGHT, CHREC_VARIABLE): Adjust.
12947         * tree-chrec.h (build_polynomial_chrec): Adjust.
12948         * tree-chrec.c (reset_evolution_in_loop): Use build_polynomial_chrec.
12949         * tree-pretty-print.c (dump_generic_node): Use CHREC_VARIABLE.
12951 2017-10-11  Marc Glisse  <marc.glisse@inria.fr>
12953         * fold-const.c (fold_binary_loc) [X +- Y CMP X]: Move ...
12954         * match.pd: ... here.
12955         ((T) X == (T) Y): Relax condition.
12957 2017-10-11  Bin Cheng  <bin.cheng@arm.com>
12959         PR tree-optimization/82472
12960         * tree-loop-distribution.c (sort_partitions_by_post_order): Refine
12961         comment.
12962         (break_alias_scc_partitions): Update postorder number.
12964 2017-10-11  Martin Liska  <mliska@suse.cz>
12966         PR sanitizer/82490
12967         * opts.c (parse_no_sanitize_attribute): Do not use error_value
12968         variable.
12969         * opts.h (parse_no_sanitize_attribute): Remove last argument.
12971 2017-10-11  Martin Liska  <mliska@suse.cz>
12973         * print-rtl.c (print_insn): Move declaration of idbuf
12974         to same scope as name.
12976 2017-10-11  Martin Liska  <mliska@suse.cz>
12978         Revert r253637:
12980         PR sanitizer/82484
12981         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12982         volatile arguments.
12984 2017-10-11  Martin Liska  <mliska@suse.cz>
12986         PR sanitizer/82484
12987         * sanopt.c (sanitize_rewrite_addressable_params): Do not handle
12988         volatile arguments.
12990 2017-10-11  Adhemerval Zanella  <adhemerval.zanella@linaro.org>
12992         * config.gcc (default_gnu_indirect_function): Default to yes for
12993         arm*-*-linux* with glibc.
12995 2017-10-11  Richard Biener  <rguenther@suse.de>
12997         * tree-scalar-evolution.c (get_scalar_evolution): Handle
12998         default-defs and types we do not want to analyze.
12999         (interpret_loop_phi): Replace unreachable code with an assert.
13000         (compute_scalar_evolution_in_loop): Remove and inline ...
13001         (analyze_scalar_evolution_1): ... here, replacing condition with
13002         what makes the intent clearer.  Remove handling of cases
13003         get_scalar_evolution now handles.
13005 2017-10-10  Jim Wilson  <wilson@tuliptree.org>
13007         PR rtl-optimization/81434
13008         * haifa-sched.c (prune_ready_list): Init min_cost_group to 0.  Update
13009         comment for main loop.  In sched_group_found if, also add checks for
13010         pass and min_cost_group.
13012 2017-10-10  Segher Boessenkool  <segher@kernel.crashing.org>
13014         * config/rs6000/rs6000.c (TARGET_INSN_COST): New.
13015         (rs6000_insn_cost): New function.
13016         * config/rs6000/rs6000.md (cost): New attribute.
13018 2017-10-10  Jakub Jelinek  <jakub@redhat.com>
13019             H.J. Lu  <hongjiu.lu@intel.com>
13021         PR target/79565
13022         PR target/82483
13023         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Add
13024         OPTION_MASK_ISA_MMX for __builtin_ia32_maskmovq,
13025         __builtin_ia32_vec_ext_v4hi and __builtin_ia32_vec_set_v4hi.
13026         (ix86_expand_builtin): Treat OPTION_MASK_ISA_MMX similarly
13027         to OPTION_MASK_ISA_AVX512VL - builtins that have both
13028         OPTION_MASK_ISA_MMX and some other bit set require both
13029         mmx and the ISAs without the mmx bit.
13030         * config/i386/i386-builtin.def (__builtin_ia32_cvtps2pi,
13031         __builtin_ia32_cvttps2pi, __builtin_ia32_cvtpi2ps,
13032         __builtin_ia32_pavgb, __builtin_ia32_pavgw, __builtin_ia32_pmulhuw,
13033         __builtin_ia32_pmaxub, __builtin_ia32_pmaxsw, __builtin_ia32_pminub,
13034         __builtin_ia32_pminsw, __builtin_ia32_psadbw, __builtin_ia32_pmovmskb,
13035         __builtin_ia32_pshufw, __builtin_ia32_cvtpd2pi,
13036         __builtin_ia32_cvttpd2pi, __builtin_ia32_cvtpi2pd,
13037         __builtin_ia32_pmuludq, __builtin_ia32_pabsb, __builtin_ia32_pabsw,
13038         __builtin_ia32_pabsd, __builtin_ia32_phaddw, __builtin_ia32_phaddd,
13039         __builtin_ia32_phaddsw, __builtin_ia32_phsubw, __builtin_ia32_phsubd,
13040         __builtin_ia32_phsubsw, __builtin_ia32_pmaddubsw,
13041         __builtin_ia32_pmulhrsw, __builtin_ia32_pshufb, __builtin_ia32_psignb,
13042         __builtin_ia32_psignw, __builtin_ia32_psignd, __builtin_ia32_movntq,
13043         __builtin_ia32_paddq, __builtin_ia32_psubq, __builtin_ia32_palignr):
13044         Add OPTION_MASK_ISA_MMX.
13046 2017-10-10  Andreas Tobler  <andreast@gcc.gnu.org>
13048         * config.gcc (armv7*-*-freebsd*): New target.
13049         (armv6*-*-freebsd*): Remove obsolete TARGET_FREEBSD_ARMv6 define.
13051 2017-10-10  Jan Hubicka  <hubicka@ucw.cz>
13053         * x86-tune.def (X86_TUNE_AVOID_FALSE_DEP_FOR_BMI,
13054         X86_TUNE_ADJUST_UNROLL, X86_TUNE_ONE_IF_CONV_INSN): Move to right
13055         spot in the file.
13057 2017-10-10  Richard Sandiford  <richard.sandiford@linaro.org>
13059         * wide-int.h (wide_int_ref_storage): Make host_dependent_precision
13060         a template parameter.
13061         (WIDE_INT_REF_FOR): Update accordingly.
13062         * tree.h (wi::int_traits <const_tree>): Delete.
13063         (wi::tree_to_widest_ref, wi::tree_to_offset_ref): New typedefs.
13064         (wi::to_widest, wi::to_offset): Use them.  Expand commentary.
13065         (wi::tree_to_wide_ref): New typedef.
13066         (wi::to_wide): New function.
13067         * calls.c (get_size_range): Use wi::to_wide when operating on
13068         trees as wide_ints.
13069         * cgraph.c (cgraph_node::create_thunk): Likewise.
13070         * config/i386/i386.c (ix86_data_alignment): Likewise.
13071         (ix86_local_alignment): Likewise.
13072         * dbxout.c (stabstr_O): Likewise.
13073         * dwarf2out.c (add_scalar_info, gen_enumeration_type_die): Likewise.
13074         * expr.c (const_vector_from_tree): Likewise.
13075         * fold-const-call.c (host_size_t_cst_p, fold_const_call_1): Likewise.
13076         * fold-const.c (may_negate_without_overflow_p, negate_expr_p)
13077         (fold_negate_expr_1, int_const_binop_1, const_binop)
13078         (fold_convert_const_int_from_real, optimize_bit_field_compare)
13079         (all_ones_mask_p, sign_bit_p, unextend, extract_muldiv_1)
13080         (fold_div_compare, fold_single_bit_test, fold_plusminus_mult_expr)
13081         (pointer_may_wrap_p, expr_not_equal_to, fold_binary_loc)
13082         (fold_ternary_loc, multiple_of_p, fold_negate_const, fold_abs_const)
13083         (fold_not_const, round_up_loc): Likewise.
13084         * gimple-fold.c (gimple_fold_indirect_ref): Likewise.
13085         * gimple-ssa-warn-alloca.c (alloca_call_type_by_arg): Likewise.
13086         (alloca_call_type): Likewise.
13087         * gimple.c (preprocess_case_label_vec_for_gimple): Likewise.
13088         * godump.c (go_output_typedef): Likewise.
13089         * graphite-sese-to-poly.c (tree_int_to_gmp): Likewise.
13090         * internal-fn.c (get_min_precision): Likewise.
13091         * ipa-cp.c (ipcp_store_vr_results): Likewise.
13092         * ipa-polymorphic-call.c
13093         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Likewise.
13094         * ipa-prop.c (ipa_print_node_jump_functions_for_edge): Likewise.
13095         (ipa_modify_call_arguments): Likewise.
13096         * match.pd: Likewise.
13097         * omp-low.c (scan_omp_1_op, lower_omp_ordered_clauses): Likewise.
13098         * print-tree.c (print_node_brief, print_node): Likewise.
13099         * stmt.c (expand_case): Likewise.
13100         * stor-layout.c (layout_type): Likewise.
13101         * tree-affine.c (tree_to_aff_combination): Likewise.
13102         * tree-cfg.c (group_case_labels_stmt): Likewise.
13103         * tree-data-ref.c (dr_analyze_indices): Likewise.
13104         (prune_runtime_alias_test_list): Likewise.
13105         * tree-dump.c (dequeue_and_dump): Likewise.
13106         * tree-inline.c (remap_gimple_op_r, copy_tree_body_r): Likewise.
13107         * tree-predcom.c (is_inv_store_elimination_chain): Likewise.
13108         * tree-pretty-print.c (dump_generic_node): Likewise.
13109         * tree-scalar-evolution.c (iv_can_overflow_p): Likewise.
13110         (simple_iv_with_niters): Likewise.
13111         * tree-ssa-address.c (addr_for_mem_ref): Likewise.
13112         * tree-ssa-ccp.c (ccp_finalize, evaluate_stmt): Likewise.
13113         * tree-ssa-loop-ivopts.c (constant_multiple_of): Likewise.
13114         * tree-ssa-loop-niter.c (split_to_var_and_offset)
13115         (refine_value_range_using_guard, number_of_iterations_ne_max)
13116         (number_of_iterations_lt_to_ne, number_of_iterations_lt)
13117         (get_cst_init_from_scev, record_nonwrapping_iv)
13118         (scev_var_range_cant_overflow): Likewise.
13119         * tree-ssa-phiopt.c (minmax_replacement): Likewise.
13120         * tree-ssa-pre.c (compute_avail): Likewise.
13121         * tree-ssa-sccvn.c (vn_reference_fold_indirect): Likewise.
13122         (vn_reference_maybe_forwprop_address, valueized_wider_op): Likewise.
13123         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Likewise.
13124         * tree-ssa-uninit.c (is_pred_expr_subset_of): Likewise.
13125         * tree-ssanames.c (set_nonzero_bits, get_nonzero_bits): Likewise.
13126         * tree-switch-conversion.c (collect_switch_conv_info, array_value_type)
13127         (dump_case_nodes, try_switch_expansion): Likewise.
13128         * tree-vect-loop-manip.c (vect_gen_vector_loop_niters): Likewise.
13129         (vect_do_peeling): Likewise.
13130         * tree-vect-patterns.c (vect_recog_bool_pattern): Likewise.
13131         * tree-vect-stmts.c (vectorizable_load): Likewise.
13132         * tree-vrp.c (compare_values_warnv, vrp_int_const_binop): Likewise.
13133         (zero_nonzero_bits_from_vr, ranges_from_anti_range): Likewise.
13134         (extract_range_from_binary_expr_1, adjust_range_with_scev): Likewise.
13135         (overflow_comparison_p_1, register_edge_assert_for_2): Likewise.
13136         (is_masked_range_test, find_switch_asserts, maybe_set_nonzero_bits)
13137         (vrp_evaluate_conditional_warnv_with_ops, intersect_ranges): Likewise.
13138         (range_fits_type_p, two_valued_val_range_p, vrp_finalize): Likewise.
13139         (evrp_dom_walker::before_dom_children): Likewise.
13140         * tree.c (cache_integer_cst, real_value_from_int_cst, integer_zerop)
13141         (integer_all_onesp, integer_pow2p, integer_nonzerop, tree_log2)
13142         (tree_floor_log2, tree_ctz, mem_ref_offset, tree_int_cst_sign_bit)
13143         (tree_int_cst_sgn, get_unwidened, int_fits_type_p): Likewise.
13144         (get_type_static_bounds, num_ending_zeros, drop_tree_overflow)
13145         (get_range_pos_neg): Likewise.
13146         * ubsan.c (ubsan_expand_ptr_ifn): Likewise.
13147         * config/darwin.c (darwin_mergeable_constant_section): Likewise.
13148         * config/aarch64/aarch64.c (aapcs_vfp_sub_candidate): Likewise.
13149         * config/arm/arm.c (aapcs_vfp_sub_candidate): Likewise.
13150         * config/avr/avr.c (avr_fold_builtin): Likewise.
13151         * config/bfin/bfin.c (bfin_local_alignment): Likewise.
13152         * config/msp430/msp430.c (msp430_attr): Likewise.
13153         * config/nds32/nds32.c (nds32_insert_attributes): Likewise.
13154         * config/powerpcspe/powerpcspe-c.c
13155         (altivec_resolve_overloaded_builtin): Likewise.
13156         * config/powerpcspe/powerpcspe.c (rs6000_aggregate_candidate)
13157         (rs6000_expand_ternop_builtin): Likewise.
13158         * config/rs6000/rs6000-c.c
13159         (altivec_resolve_overloaded_builtin): Likewise.
13160         * config/rs6000/rs6000.c (rs6000_aggregate_candidate): Likewise.
13161         (rs6000_expand_ternop_builtin): Likewise.
13162         * config/s390/s390.c (s390_handle_hotpatch_attribute): Likewise.
13164 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13166         * tree-vect-loop-manip.c (rename_variables_in_bb): Rename PHI nodes
13167         when copying loop nest with only one inner loop.
13169 2017-10-10  Richard Biener  <rguenther@suse.de>
13171         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Avoid compacting
13172         blocks if SCEV is active.
13173         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Remove
13174         dead code.
13175         (analyze_scalar_evolution): Handle cached evolutions the obvious way.
13176         (scev_initialize): Assert we are not yet initialized.
13178 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13180         * tree-loop-distribution.c (generate_loops_for_partition): Remove
13181         inner loop's exit stmt by making it always exit the loop, otherwise
13182         we would generate an infinite empty loop.
13184 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13186         * tree-vect-loop-manip.c (slpeel_tree_duplicate_loop_to_edge_cfg): Skip
13187         renaming variables in new preheader if it's deleted.
13189 2017-10-10  Bin Cheng  <bin.cheng@arm.com>
13191         * tree-loop-distribution.c (struct partition): Remove unused field
13192         loops of the structure.
13193         (partition_alloc, partition_free): Ditto.
13194         (build_rdg_partition_for_vertex): Ditto.
13196 2017-10-09  Jeff Law  <law@redhat.com>
13198         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): Fix
13199         return type to match prototype and documentation.
13201 2010-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13203         * config/rs6000/rs6000.c (processor_costs): Move to ...
13204         * config/rs6000/rs6000.h: ... here.
13205         (rs6000_cost): Declare.
13207 2017-10-09  Eric Botcazou  <ebotcazou@adacore.com>
13209         * except.c (setjmp_fn): New global variable.
13210         (init_eh): Initialize it if DONT_USE_BUILTIN_SETJMP is defined.
13211         (sjlj_emit_function_enter): Call it instead of BUILTIN_SETJMP
13212         if DONT_USE_BUILTIN_SETJMP is defined.
13214 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13216         * target.def (insn_cost): New hook.
13217         * doc/tm.texi.in (TARGET_INSN_COST): New hook.
13218         * doc/tm.texi: Regenerate.
13219         * rtlanal.c (insn_cost): Use the new hook.
13221 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13223         * combine.c (combine_validate_cost): Compute the new insn_cost,
13224         not just pattern_cost.
13225         (try_combine): Adjust comment.
13227 2017-10-09  Segher Boessenkool  <segher@kernel.crashing.org>
13229         * cfgrtl.c (rtl_account_profile_record): Replace insn_rtx_cost with
13230         insn_cost.
13231         * combine.c (uid_insn_cost): Adjust comment.
13232         (combine_validate_cost): Adjust comment.  Use pattern_cost instead
13233         of insn_rtx_cost
13234         (combine_instructions): Use insn_cost instead of insn_rtx_cost.
13235         * dse.c (find_shift_sequence): Ditto.
13236         * ifcvt.c (cheap_bb_rtx_cost_p): Ditto.
13237         (bb_valid_for_noce_process_p): Use pattern_cost.
13238         * rtl.h (insn_rtx_cost): Delete.
13239         (pattern_cost): New prototype.
13240         (insn_cost): New prototype.
13241         * rtlanal.c (insn_rtx_cost): Rename to...
13242         (pattern_cost): ... this.
13243         (insn_cost): New.
13245 2017-10-09  Uros Bizjak  <ubizjak@gmail.com>
13247         * config/i386/i386.md (*jcc_2): Remove insn pattern.
13248         (*jcc<mode>_0_r_i387): Ditto.
13249         (*jccxf_r_i387): Ditto.
13250         (*jcc<mode>_r_i387): Ditto.
13251         (*jccu<mode>_r_i387): Ditto.
13252         (*jcc<X87MODEF:mode>_<SWI24:mode>_r_i387): Ditto.
13253         (*jcc): Rename from *jcc_1.
13255 2017-10-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13257         * config/rs6000/rs6000-p8swap.c (rs6000_analyze_swaps): Process
13258         deferred rescans after the lvx/stvx recombination pre-pass.
13260 2017-10-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
13262         * config/rs6000/amo.h: New include file to provide ISA 3.0 atomic
13263         memory operation instruction support.
13264         * config.gcc (powerpc*-*-*): Include amo.h as an extra header.
13265         (rs6000-ibm-aix[789]*): Likewise.
13266         * doc/extend.texi (PowerPC Atomic Memory Operation Functions):
13267         Document new functions.
13269 2017-10-09  Richard Biener  <rguenther@suse.de>
13271         PR tree-optimization/82397
13272         * tree-data-ref.c (data_ref_compare_tree): Make sure to return
13273         equality only for semantically equal trees.
13275 2017-10-09  Richard Biener  <rguenther@suse.de>
13277         PR tree-optimization/82449
13278         * sese.c (scev_analyzable_p): Check whether the SCEV is linear.
13279         * tree-chrec.h (evolution_function_is_constant_p): Adjust to
13280         allow constant addresses.
13281         * tree-chrec.c (scev_is_linear_expression): Constant evolutions
13282         are linear.
13284 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13286         * config/s390/s390-builtins.def (vec_nabs, vec_vfi): Fix builtin
13287         flags.
13289 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13291         PR target/82463
13292         * config/s390/vecintrin.h (vec_madd, vec_msub): Fix macro
13293         definitions.
13295 2017-10-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13297         PR target/82465
13298         * config/s390/s390-builtins.def (vec_sqrt): Fix builtin flags.
13300 2017-10-09  Jakub Jelinek  <jakub@redhat.com>
13302         PR target/82464
13303         * config/s390/s390-builtins.def (s390_vec_xor_flt_a,
13304         s390_vec_xor_flt_b, s390_vec_xor_flt_c): New.
13306 2017-10-09  Richard Sandiford  <richard.sandiford@linaro.org>
13308         * wide-int.h (WI_BINARY_OPERATOR_RESULT): New macro.
13309         (WI_BINARY_PREDICATE_RESULT): Likewise.
13310         (wi::binary_traits::operator_result): New type.
13311         (wi::binary_traits::predicate_result): Likewise.
13312         (generic_wide_int::operator~, unary generic_wide_int::operator-)
13313         (generic_wide_int::operator==, generic_wide_int::operator!=)
13314         (generic_wide_int::operator&, generic_wide_int::and_not)
13315         (generic_wide_int::operator|, generic_wide_int::or_not)
13316         (generic_wide_int::operator^, generic_wide_int::operator+
13317         (binary generic_wide_int::operator-, generic_wide_int::operator*):
13318         Delete.
13319         (operator~, unary operator-, operator==, operator!=, operator&)
13320         (operator|, operator^, operator+, binary operator-, operator*): New
13321         functions.
13322         * expr.c (get_inner_reference): Use wi::bit_and_not.
13323         * fold-const.c (fold_binary_loc): Likewise.
13324         * ipa-prop.c (ipa_compute_jump_functions_for_edge): Likewise.
13325         * tree-ssa-ccp.c (get_value_from_alignment): Likewise.
13326         (bit_value_binop): Likewise.
13327         * tree-ssa-math-opts.c (find_bswap_or_nop_load): Likewise.
13328         * tree-vrp.c (zero_nonzero_bits_from_vr): Likewise.
13329         (extract_range_from_binary_expr_1): Likewise.
13330         (masked_increment): Likewise.
13331         (simplify_bit_ops_using_ranges): Likewise.
13333 2017-10-09  Martin Jambor  <mjambor@suse.cz>
13335         PR hsa/82416
13336         * hsa-common.h (hsa_op_with_type): New method extend_int_to_32bit.
13337         * hsa-gen.c (hsa_extend_inttype_to_32bit): New function.
13338         (hsa_type_for_scalar_tree_type): Use it.  Always force min32int for
13339         COMPLEX types.
13340         (hsa_fixup_mov_insn_type): New function.
13341         (hsa_op_with_type::get_in_type): Use it.
13342         (hsa_build_append_simple_mov): Likewise.  Allow sub-32bit
13343         immediates in an assert.
13344         (hsa_op_with_type::extend_int_to_32bit): New method.
13345         (gen_hsa_insns_for_bitfield): Fixup instruction and intermediary
13346         types.  Convert to dest type if necessary.
13347         (gen_hsa_insns_for_bitfield_load): Fixup load type if necessary.
13348         (reg_for_gimple_ssa): Pass false as min32int to
13349         hsa_type_for_scalar_tree_type.
13350         (gen_hsa_addr): Fixup type when creating addresable temporary.
13351         (gen_hsa_cmp_insn_from_gimple): Extend operands if necessary.
13352         (gen_hsa_unary_operation): Extend operands and convert to dest type if
13353         necessary.  Call hsa_fixup_mov_insn_type.
13354         (gen_hsa_binary_operation): Changed operand types to hsa_op_with_type,
13355         extend operands and convert to dest type if necessary.
13356         (gen_hsa_insns_for_operation_assignment): Extend operands and convert
13357         to dest type if necessary.
13358         (set_output_in_type): Call hsa_fixup_mov_insn_type.  Just ude dest
13359         if conversion nt necessary and size matches.
13360         (gen_hsa_insns_for_load): Call hsa_fixup_mov_insn_type, convert
13361         to dest type if necessary.
13362         (gen_hsa_insns_for_store): Call hsa_fixup_mov_insn_type.
13363         (gen_hsa_insns_for_switch_stmt): Likewise. Also extend operands if
13364         necessary.
13365         (gen_hsa_clrsb): Likewise.
13366         (gen_hsa_ffs): Likewise.
13367         (gen_hsa_divmod): Extend operands and convert to dest type if
13368         necessary.
13369         (gen_hsa_atomic_for_builtin): Change type of op to hsa_op_with_type.
13371 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13373         * config/rs6000/rs6000.md (conditional branch): Clean up formatting.
13374         Remove empty default arguments.  Use a brace block as output
13375         statement.
13376         (conditional return): Ditto.
13377         (jump): Ditto.
13378         (indirect_jump): Ditto.  Use b%T0 instead of bctr/blr.
13379         (tablejump, tablejumpsi, tablejumpdi, *tablejump<mode>_internal1):
13380         Ditto.
13381         (group_ending_nop): Ditto.
13382         (doloop_end): Ditto.
13383         (ctr<mode>, ctr<mode>_internal1, ctr<mode>_internal2): Ditto.
13384         (splitters for those): Ditto.
13386 2017-10-08  Segher Boessenkool  <segher@kernel.crashing.org>
13388         * config/rs6000/rs6000-string.c (expand_strncmp_align_check): Invert
13389         a conditional jump (and the compare for it) so that pc_rtx is the
13390         last operand.
13391         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Adjust
13392         for the deleted and renamed ctr<mode>_internal[234] patterns.
13393         * config/rs6000/rs6000.md: Delete second conditional branch pattern.
13394         Delete second conditional return pattern.
13395         (ctr<mode>_internal2): Delete this second bdnz pattern.
13396         (ctr<mode>_internal3): Rename to ctr<mode>_internal2.
13397         (ctr<mode>_internal4): Delete this second bdz pattern.
13399 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13401         * tree-outof-ssa.h (ssaexpand): Add partitions_for_undefined_values.
13402         (always_initialized_rtx_for_ssa_name_p): New predicate.
13403         * tree-outof-ssa.c (remove_ssa_form): Initialize new field of SA.
13404         (finish_out_of_ssa): Free new field of SA.
13405         * tree-ssa-coalesce.h (get_undefined_value_partitions): Declare.
13406         * tree-ssa-coalesce.c: Include tree-ssa.h.
13407         (get_parm_default_def_partitions): Remove extern keyword.
13408         (get_undefined_value_partitions): New function.
13409         * expr.c (expand_expr_real_1) <expand_decl_rtl>: For a SSA_NAME, do
13410         not set SUBREG_PROMOTED_VAR_P on the sub-register if it may contain
13411         uninitialized bits.
13412         * loop-iv.c (iv_get_reaching_def): Disqualify all subregs.
13414 2017-10-08  Eric Botcazou  <ebotcazou@adacore.com>
13416         * builtins.def (BUILT_IN_SETJMP): Revert latest change.
13418 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13420         * config/i386/i386.c (ix86_expand_set_or_movmem): Disable 512bit loops
13421         for targets that preffer 128bit.
13423 2017-10-08  Jan Hubicka  <hubicka@ucw.cz>
13425         * config/i386/i386.c (has_dispatch): Disable for Ryzen.
13427 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13429         * config/arm/arm.c (arm_set_return_address): Use MEM_VOLATILE_P
13430         on the target mem instead of RTX_FRAME_RELATED_P on the insn to
13431         prevent DSE.
13432         (thumb_set_return_address): Likewise.
13434 2017-10-08  Olivier Hainque  <hainque@adacore.com>
13436         * common/config/arm/arm-common.c (arm_except_unwind_info):
13437         Handle DWARF2_UNWIND_INFO.
13439 2017-10-07  Michael Collison <michael.collison@arm.com>
13441         * config/aarch64/aarch64.md (*aarch64_reg_<optab>_minus<mode>3):
13442         New pattern.
13444 2017-10-07  Eric Botcazou  <ebotcazou@adacore.com>
13446         * builtins.def (BUILT_IN_SETJMP): Declare as library builtin instead
13447         of GCC builtin if DONT_USE_BUILTIN_SETJMP is defined.
13448         * except.c (sjlj_emit_function_enter): If DONT_USE_BUILTIN_SETJMP is
13449         defined, force the creation of a new block for a dispatch label.
13451 2017-10-07  Jan Hubicka  <hubicka@ucw.cz>
13453         * invoke.texi (Wsuggest-attribute=cold): Document.
13454         * common.opt (Wsuggest-attribute=cold): New
13455         * ipa-pure-const.c (warn_function_cold): New function.
13456         * predict.c (compute_function_frequency): Use it.
13457         * predict.h (warn_function_cold): Declare.
13459 2017-10-06  Jan Hubicka  <hubicka@ucw.cz>
13461         * tree-switch-conversion.c (do_jump_if_equal, emit_cmp_and_jump_insns):
13462         Update profile.
13464 2017-10-06  Martin Liska  <mliska@suse.cz>
13466         * sanopt.c (struct sanopt_tree_triplet_hash): Remove inline
13467         keyword for member functions.
13468         (struct sanopt_tree_couple): New struct.
13469         (struct sanopt_tree_couple_hash): New function.
13470         (struct sanopt_ctx): Add new hash_map.
13471         (has_dominating_ubsan_ptr_check): New function.
13472         (record_ubsan_ptr_check_stmt): Likewise.
13473         (maybe_optimize_ubsan_ptr_ifn): Likewise.
13474         (sanopt_optimize_walker): Handle IFN_UBSAN_PTR.
13475         (pass_sanopt::execute): Handle also SANITIZE_POINTER_OVERFLOW.
13477 2017-10-06  Sudakshina Das  <sudi.das@arm.com>
13479         PR target/82440
13480         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): Only call
13481         aarch64_simd_valid_immediate on CONST_VECTORs.
13482         (aarch64_reg_or_bic_imm): Likewise.
13484 2017-10-06  Wilco Dijkstra  <wdijkstr@arm.com>
13486         PR rtl-optimization/82396
13487         * haifa-sched.c (ready_sort_real): Disable qsort checking.
13489 2017-10-06  Sebastian Pop  <sebpop@gmail.com>
13491         * graphite-dependences.c (scop_get_reads): Move code to...
13492         (scop_get_must_writes): Move code to...
13493         (scop_get_may_writes): Move code to...
13494         (scop_get_reads_and_writes): ... here.
13495         (scop_get_dependences): Call scop_get_reads_and_writes.
13497 2017-10-06  Jakub Jelinek  <jakub@redhat.com>
13499         PR tree-optimization/82434
13500         * fold-const.h (can_native_encode_type_p,
13501         can_native_encode_string_p): Remove.
13502         * fold-const.c (native_encode_int): Formatting fixes.  If ptr is NULL,
13503         don't encode anything, just return what would be otherwise returned.
13504         (native_encode_fixed, native_encode_complex, native_encode_vector):
13505         Likewise.
13506         (native_encode_string): Likewise.  Inline by hand
13507         can_native_encode_string_p.
13508         (can_native_encode_type_p): Remove.
13509         (can_native_encode_string_p): Remove.
13510         * tree-vect-stmts.c (vectorizable_store): Instead of testing just
13511         STRING_CSTs using can_native_encode_string_p, test all
13512         CONSTANT_CLASS_P values using native_encode_expr with NULL ptr.
13513         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Remove last
13514         argument from native_encode_expr.
13515         (rhs_valid_for_store_merging_p): Use native_encode_expr with NULL ptr.
13516         (pass_store_merging::execute): Don't unnecessarily look for 3 stmts,
13517         but just 2.
13519 2017-10-06  Richard Biener  <rguenther@suse.de>
13521         PR tree-optimization/82397
13522         * tree-vect-data-refs.c (dr_group_sort_cmp): Do not use
13523         operand_equal_p but rely on data_ref_compare_tree for detecting
13524         equalities.
13525         (vect_analyze_data_ref_accesses): Use data_ref_compare_tree
13526         to match up with dr_group_sort_cmp.
13528 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13530         PR target/82322
13531         * config/s390/s390-builtins.def (s390_vfi): Define new overloaded
13532         builtin.
13533         * config/s390/s390-builtin-types.def: Regenerate.
13535 2017-10-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13537         PR target/82317
13538         * config/s390/s390-builtin-types.def: Regenerate.
13539         * config/s390/s390-builtins.def (s390_vfmaxdb_4, s390_vfmindb_4):
13540         Change flag from B_VXE to B_VX.
13541         (s390_vec_min_dbl): Remove B_VXE flag.
13543 2017-10-06  Richard Biener  <rguenther@suse.de>
13545         * graphite-isl-ast-to-gimple.c: Include ssa.h and tree-ssa.h.
13546         (translate_isl_ast_to_gimple::translate_pending_phi_nodes,
13547         translate_isl_ast_to_gimple::is_valid_rename,
13548         translate_isl_ast_to_gimple::get_rename,
13549         translate_isl_ast_to_gimple::get_def_bb_for_const,
13550         translate_isl_ast_to_gimple::get_new_name,
13551         translate_isl_ast_to_gimple::collect_all_ssa_names,
13552         translate_isl_ast_to_gimple::copy_loop_phi_args,
13553         translate_isl_ast_to_gimple::collect_all_ssa_names,
13554         translate_isl_ast_to_gimple::copy_loop_phi_args,
13555         translate_isl_ast_to_gimple::copy_loop_phi_nodes,
13556         translate_isl_ast_to_gimple::add_close_phis_to_merge_points,
13557         translate_isl_ast_to_gimple::add_close_phis_to_outer_loops,
13558         translate_isl_ast_to_gimple::copy_loop_close_phi_args,
13559         translate_isl_ast_to_gimple::copy_loop_close_phi_nodes,
13560         translate_isl_ast_to_gimple::copy_cond_phi_args,
13561         translate_isl_ast_to_gimple::copy_cond_phi_nodes,
13562         translate_isl_ast_to_gimple::edge_for_new_close_phis,
13563         translate_isl_ast_to_gimple::add_phi_arg_for_new_expr,
13564         translate_isl_ast_to_gimple::rename_uses,
13565         translate_isl_ast_to_gimple::rename_all_uses): Remove.
13566         (translate_isl_ast_to_gimple::get_rename_from_scev): Simplify.
13567         (set_rename_for_each_def): Likewise.
13568         (graphite_copy_stmts_from_block): Handle debug stmt resetting
13569         here.  Handle rewriting SCEV analyzable uses here.
13570         (copy_bb_and_scalar_dependences): Generate code for PHI
13571         copy-in/outs.
13572         (graphite_regenerate_ast_isl): Adjust.
13573         * graphite-scop-detection.c (trivially_empty_bb_p): Move to sese.[ch].
13574         (add_write, add_read): New functions.
13575         (build_cross_bb_scalars_def): Use it and simplify.
13576         (build_cross_bb_scalars_use): Likewise.
13577         (graphite_find_cross_bb_scalar_vars): Inline into...
13578         (try_generate_gimple_bb): ...here.  Add dependences for PHIs,
13579         simulating out-of-SSA.  Compute liveout and add dependencies.
13580         (build_scops): Force an empty entry block.
13581         * sese.h (sese_info_t::liveout, sese_info_t::debug_liveout): New
13582         members.
13583         (sese_build_liveouts): Declare.
13584         (sese_trivially_empty_bb_p): Likewise.
13585         * sese.c (sese_build_liveouts_bb): Properly handle PHIs,
13586         compute liveout and debug_liveout.
13587         (sese_bad_liveouts_use): Remove.
13588         (sese_reset_debug_liveouts_bb): Likewise.
13589         (sese_reset_debug_liveouts): Rewrite in terms of debug_liveout.
13590         (sese_build_liveouts): Build liveout and debug_liveout and store
13591         it in region.
13592         (new_sese_info): Adjust.
13593         (free_sese_info): Likewise.
13594         (sese_insert_phis_for_liveouts): Reset debug stmts from here,
13595         do not build liveout here.
13596         (move_sese_in_condition): Adjust region entry.
13597         (scev_analyzable_p): Match up with chrec_apply requirements.
13598         (sese_trivially_empty_bb_p): New.
13599         * tree-into-ssa.c (get_reaching_def): Properly support generating
13600         default-defs for incremental rewrite of anonymous names.
13602 2017-10-06  Richard Biener  <rguenther@suse.de>
13604         * graphite-sese-to-poly.c (extract_affine): For casts increasing
13605         precision do not perform modulo reduction.
13607 2017-10-06  Richard Biener  <rguenther@suse.de>
13609         PR tree-optimization/82436
13610         * tree-vect-slp.c (vect_supported_load_permutation_p): More
13611         conservatively choose the vectorization factor when checking
13612         whether we can perform the required load permutation.
13613         (vect_transform_slp_perm_load): Assert when we may not fail.
13615 2017-10-05  Segher Boessenkool  <segher@kernel.crashing.org>
13617         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Correct error
13618         message for incompatible -msdata=* and -mcall-* options.
13620 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13622         * config/i386/i386.c (ia32_multipass_dfa_lookahead): Default to issue
13623         rate for post-reload scheduling.
13625 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13627         * doc/sourcebuild.texi (vect_sizes_16B_8B, vect_sizes_32B_16B): New.
13629 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13631         * config/i386/i386.c (znver1_cost): Set branch_cost to 3 (instead of 2)
13632         to improve monte carlo in scimark.
13634 2017-10-05  Jan Hubicka <hubicka@ucw.cz>
13636         * config/i386/i386.c (ix86_size_cost, i386_cost, i486_cost,
13637         pentium_cost, lakemont_cost, pentiumpro_cost, geode_cost, k6_cost,
13638         athlon_cost, k8_cost, amdfam10_cost, btver1_cost, btver2_cost,
13639         pentium4_cost, nocona_cost): Set reassociation width to 1.
13640         (bdver1_cost, bdver2_cost, bdver3_cost, bdver4_cost): Set reassociation
13641         width to 2 for fp operations and 1 otherwise.
13642         (znver1_cost): Set scalar reassoc width to 4 and vector to 3 and 6
13643         for int and fp.
13644         (atom_cost): Set reassociation width to 2.
13645         (slm_cost, generic_cost): Set fp reassociation width
13646         to 2 and 1 otherwise.
13647         (intel_cost): Set fp reassociation width to 4 and 1 otherwise.
13648         (core_cost): Set fp reassociation width to 4 and vector to 2.
13649         (ix86_reassociation_width): Rewrite using cost table; special case
13650         plus/minus on Zen; honor X86_TUNE_SSE_SPLIT_REGS
13651         and TARGET_AVX128_OPTIMAL.
13652         * config/i386/i386.h (processor_costs): Add
13653         reassoc_int, reassoc_fp, reassoc_vec_int, reassoc_vec_fp.
13654         (TARGET_VECTOR_PARALLEL_EXECUTION, TARGET_REASSOC_INT_TO_PARALLEL,
13655         TARGET_REASSOC_FP_TO_PARALLEL): Remove.
13656         * x86-tune.def (X86_TUNE_REASSOC_INT_TO_PARALLEL): Remove.
13657         (X86_TUNE_REASSOC_FP_TO_PARALLEL): Remove.
13658         (X86_TUNE_VECTOR_PARALLEL_EXECUTION):  Remove.
13660 2017-10-05  Nathan Sidwell  <nathan@acm.org>
13662         * doc/invoke.texi (Wparentheses): Document C++ MVP behaviour.
13664 2017-10-05  Tamar Christina  <tamar.christina@arm.com>
13666         * config/arm/arm.c (arm_test_fpu_data): New.
13667         (arm_run_selftests): Call arm_test_fpu_data.
13669 2017-10-04  Nathan Sidwell  <nathan@acm.org>
13671         * toplev.c (toplev::main): Remove excess parens on pretty_printer
13672         decl.
13673         * caller-save.c (insert_save): Remove excess parens on TO_SAVE parm.
13675 2017-10-04  Sudakshina Das  <sudi.das@arm.com>
13677         * config/aarch64/aarch64-protos.h (enum simd_immediate_check): New
13678         check type for aarch64_simd_valid_immediate.
13679         (aarch64_output_simd_mov_immediate): Update prototype.
13680         (aarch64_simd_valid_immediate): Update prototype.
13681         * config/aarch64/aarch64-simd.md (orr<mode>3): modified pattern to add
13682         support for ORR-immediate.
13683         (and<mode>3): modified pattern to add support for BIC-immediate.
13684         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Function
13685         now checks for valid immediate for BIC and ORR based on new enum
13686         argument.
13687         (aarch64_output_simd_mov_immediate): Function now used to output
13688         BIC/ORR imm as well based on new enum argument.
13689         * config/aarch64/constraints.md (Do): New vector immediate constraint.
13690         (Db) : Likewise.
13691         * config/aarch64/predicates.md (aarch64_reg_or_orr_imm): New predicate.
13692         (aarch64_reg_or_bic_imm): Likewise.
13694 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13696         * config/s390/vx-builtins.md ("vec_mergeh<mode>")
13697         ("vec_mergel<mode>"): Change mode iterator to V_128_NOSINGLE.
13699 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13701         Revert r253399:
13703         PR rtl-optimization/82396
13704         * haifa-sched.c (autopref_multipass_init): Simplify
13705         initialization.
13706         (autopref_rank_data): Simplify sort order.
13707         * sched-int.h (autopref_multipass_data_): Remove
13708         multi_mem_insn_p, min_offset and max_offset.
13710 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13712         * doc/sourcebuild.texi: Document vect_peeling_profitable.
13714 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13716         * doc/sourcebuild.texi: Document vect_intdouble_cvt and
13717         vect_doubleint_cvt.
13719 2017-10-04  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
13721         * doc/sourcebuild.texi: Document vect_long_mult.
13723 2017-10-04  Richard Sandiford  <richard.sandiford@linaro.org>
13725         PR tree-optimization/82413
13726         * fold-const.c (build_range_check): Use widest_int when comparing
13727         the maximum ETYPE value with HIGH.
13729 2017-10-04  Wilco Dijkstra  <wdijkstr@arm.com>
13731         PR rtl-optimization/82396
13732         * haifa-sched.c (autopref_multipass_init): Simplify
13733         initialization.
13734         (autopref_rank_data): Simplify sort order.
13735         * sched-int.h (autopref_multipass_data_): Remove
13736         multi_mem_insn_p, min_offset and max_offset.
13738 2017-10-04  Jakub Jelinek  <jakub@redhat.com>
13740         PR tree-optimization/82381
13741         * tree-ssa-reassoc.c (sort_by_operand_rank): Check for different
13742         oeN->rank first.  Return 1 or -1 if one op is SSA_NAME and the other
13743         is not.
13745         PR tree-optimization/82374
13746         * omp-low.c (create_omp_child_function): Copy DECL_ATTRIBUTES,
13747         DECL_FUNCTION_SPECIFIC_OPTIMIZATION,
13748         DECL_FUNCTION_SPECIFIC_TARGET and DECL_FUNCTION_VERSIONED from
13749         current_function_decl to the new decl.
13751 2017-10-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
13753         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_2_HW): Define new
13754         helper macro for IEEE float128 hardware built-in functions.
13755         (SQRTF128_ODD): Add built-in functions with the round-to-odd
13756         semantics.
13757         (TRUNCF128_ODD): Likewise.
13758         (ADDF128_ODD): Likewise.
13759         (SUBF128_ODD): Likewise.
13760         (MULF128_ODD): Likewise.
13761         (DIVF128_ODD): Likewise.
13762         (FMAF128_ODD): Likewise.
13763         * config/rs6000/rs6000.md (UNSPEC_ROUND_TO_ODD): Rename to
13764         UNSPEC_TRUNC_ROUND_TO_ODD.
13765         (UNSPEC_TRUNC_ROUND_TO_ODD): Likewise.
13766         (UNSPEC_ADD_ROUND_TO_ODD): New unspec codes for the IEEE 128-bit
13767         floating point round to odd instructions.
13768         (UNSPEC_SUB_ROUND_TO_ODD): Likewise.
13769         (UNSPEC_MUL_ROUND_TO_ODD): Likewise.
13770         (UNSPEC_DIV_ROUND_TO_ODD): Likewise.
13771         (UNSPEC_FMA_ROUND_TO_ODD): Likewise.
13772         (UNSPEC_SQRT_ROUND_TO_ODD): Likewise.
13773         (trunc<mode>sf2_hw): Change the truncate with round to odd
13774         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.
13775         (add<mode>3_odd): Add insns for IEEE 128-bit floating point round
13776         to odd hardware instructions.
13777         (sub<mode>3_odd): Likewise.
13778         (mul<mode>3_odd): Likewise.
13779         (div<mode>3_odd): Likewise.
13780         (sqrt<mode>2_odd): Likewise.
13781         (fma<mode>4_odd): Likewise.
13782         (fms<mode>4_odd): Likewise.
13783         (nfma<mode>4_odd): Likewise.
13784         (nfms<mode>4_odd): Likewise.
13785         (trunc<mode>df2_odd): Change the truncate with round to odd
13786         expansion to use UNSPEC_TRUNC_ROUND_TO_ODD.  Add a generator
13787         function.
13788         * doc/extend.texi (PowerPC built-in functions): Update documentation
13789         for existing IEEE float128-bit built-in functions.  Add built-in
13790         functions that generate the IEEE 128-bit floating point round to
13791         odd instructions.
13793 2017-10-03  Segher Boessenkool  <segher@kernel.crashing.org>
13795         PR rtl-optimization/77729
13796         * simplify-rtx.c (simplify_binary_operation_1): Delete the (X&C1)|C2
13797         to (X&(C1&~C2))|C2 transformations.
13799 2017-10-03  Martin Jambor  <mjambor@suse.cz>
13801         PR tree-optimization/82363
13802         * tree-sra.c (propagate_subaccesses_across_link): In unrecoverable
13803         mismatch, mark lacc written regardless of racc.
13805 2017-10-03  Jakub Jelinek  <jakub@redhat.com>
13807         PR tree-optimization/82381
13808         * tree-ssa-reassoc.c (sort_by_operand_rank): Don't check
13809         stmt_to_insert nor wheather SSA_NAMEs are default defs.
13810         Return 1 or -1 if one of bba and bbb is NULL. If bb_rank is equal,
13811         fallthrough into reassoc_stmt_dominates_stmt_p.
13813         PR target/82386
13814         * combine.c (combine_instructions): Don't combine in unreachable
13815         basic blocks.
13817 2017-10-02  Peter Bergner  <bergner@vnet.ibm.com>
13819         PR target/80210
13820         * config/rs6000/rs6000.c (rs6000_option_override_internal): Rewrite
13821         function to not use the have_cpu variable.  Do not set cpu_index,
13822         rs6000_cpu_index or rs6000_tune_index if we end up using TARGET_DEFAULT
13823         or the default cpu.
13824         (rs6000_valid_attribute_p): Remove duplicate initializations of
13825         old_optimize and func_optimize.
13826         (rs6000_pragma_target_parse): Call rs6000_activate_target_options ().
13827         (rs6000_activate_target_options): Make global.
13828         * config/rs6000/rs6000-protos.h (rs6000_activate_target_options): Add
13829         prototype.
13831 2017-10-02  Jakub Jelinek  <jakub@redhat.com>
13833         * tree-dfa.c (get_ref_base_and_extent): Set *pmax_size to -1
13834         if *poffset + *pmax_size overflows in HOST_WIDE_INT.
13835         Set *poffset to 0 and *psize and *pmax_size to -1 if
13836         *poffset + *psize overflows in HOST_WIDE_INT.
13838         PR tree-optimization/82387
13839         PR tree-optimization/82388
13840         PR tree-optimization/82389
13841         * tree-ssa-dse.c (dse_classify_store): Test byte_tracking_enabled
13842         instead of live_bytes non-NULL.
13844 2017-10-02  Georg-Johann Lay  <avr@gjlay.de>
13846         PR target/41076
13847         * confg/avr/avr.md (*iorhi3.ashift8-ext.zerox): Add "r,r,0"
13848         alternative.
13850 2017-10-02  Richard Biener  <rguenther@suse.de>
13852         * graphite-isl-ast-to-gimple.c (set_codegen_error): With
13853         -fchecking and --param graphite-allow-codegen-errors=0 ICE.
13854         * params.def (PARAM_GRAPHITE_ALLOW_CODEGEN_ERRORS): New param.
13856 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13858         * tree.h (wi::int_traits <const_tree>::decompose): Assert that the
13859         requested precision matches the type's.
13860         * calls.c (alloc_max_size): Calculate the new candidate size as
13861         a widest_int and use wi::to_widest when comparing it with the
13862         current candidate size.
13863         * gimple-ssa-warn-alloca.c (pass_walloca::execute): Compare with
13864         zero rather than integer_zero_node.
13865         * match.pd: Check for a no-op conversion before using wi::add
13866         rather than after.  Use tree_to_uhwi when summing small shift
13867         counts into an unsigned int.
13869 2017-10-02  Richard Sandiford  <richard.sandiford@linaro.org>
13870             Alan Hayward  <alan.hayward@arm.com>
13871             David Sherwood  <david.sherwood@arm.com>
13873         PR target/71307
13874         * config/aarch64/aarch64.h (POINTER_AND_FP_REGS): New reg class.
13875         (REG_CLASS_NAMES, REG_CLASS_CONTENTS): Update accordingly.
13876         * config/aarch64/aarch64.c (aarch64_class_max_nregs): Handle
13877         POINTER_AND_FP_REGS.
13879 2017-10-02  Richard Biener  <rguenther@suse.de>
13881         PR tree-optimization/82355
13882         * graphite-isl-ast-to-gimple.c (build_iv_mapping): Also build
13883         a mapping for the enclosing loop but avoid generating one for
13884         the loop tree root.
13885         (copy_bb_and_scalar_dependences): Remove premature codegen
13886         error on PHIs in blocks duplicated into multiple places.
13887         * graphite-scop-detection.c
13888         (scop_detection::stmt_has_simple_data_refs_p): For a loop not
13889         in the region use it as loop and nest to analyze the DR in.
13890         (try_generate_gimple_bb): Likewise.
13891         * graphite-sese-to-poly.c (extract_affine_chrec): Adjust.
13892         (add_loop_constraints): For blocks in a loop not in the region
13893         create a dimension with a single iteration.
13894         * sese.h (gbb_loop_at_index): Remove assert.
13896 2017-10-01  Kevin Buettner  <kevinb@redhat.com>
13898         * omp-expand.c (adjust_context_scope): New function.
13899         (expand_parallel_call): Call adjust_context_scope.
13901 2017-10-01  Jeff Law  <law@redhat.com>
13903         * tree-ssa-dom.c (optimize_stmt): Make this a method within the
13904         dom_opt_dom_walker class with direct access to private members.
13905         Add comments.  Call test_for_singularity.
13906         (dom_opt_dom_walker::before_dom_children): Corresponding changes.
13907         (dom_opt_dom_walker::after_dom_children): Do not lazily initialize
13908         m_dummy_cond anymore.
13909         (class dom_opt_dom_walker): Initialize m_dummy_cond member in the
13910         class ctor.
13911         (pass_dominator:execute): Build the dummy_cond here and pass it
13912         to the dom_opt_dom_walker ctor.
13913         (test_for_singularity): New function.
13915 2017-09-30  Krister Walfridsson  <krister.walfridsson@gmail.com>
13916             Maya Rashish  <coypu@sdf.org>
13918         * config.gcc (*-*-netbsd*): New variable nbsd_tm_file containing
13919         netbsd.h, netbsd-stdint.h, and netbsd-elf.h.
13920         (alpha*-*-netbsd*) Use nbsd_tm_file.
13921         (arm*-*-netbsdelf*) Likewise.
13922         (i[34567]86-*-netbsdelf*) Likewise.
13923         (x86_64-*-netbsd*) Likewise.
13924         (mips*-*-netbsd*) Likewise.
13925         (powerpc-*-netbsd*) Likewise.
13926         (sh*-*-netbsd*) Likewise.
13927         (sparc-*-netbsdelf*) Likewise.
13928         (sparc64-*-netbsd*) Likewise.
13929         (m68k*-*-netbsdelf*) Use nbsd_tm_file and add CHAR_FAST8/SHORT_FAST16
13930         to tm_defines.
13931         (vax-*-netbsdelf*) Likewise.
13932         * config/netbsd-stdint.h (INT_FAST8_TYPE): Check CHAR_FAST8.
13933         (UINT_FAST8_TYPE) Likewise.
13934         (INT_FAST16_TYPE) Check CHAR_FAST16.
13935         (UINT_FAST16_TYPE) Likewise.
13937 2017-09-30  Jakub Jelinek  <jakub@redhat.com>
13939         PR target/82361
13940         * config/i386/i386.md
13941         (TARGET_USE_8BIT_IDIV zext divmodsi4 splitter): New define_split.
13942         (divmodsi4_zext_1, divmodsi4_zext_2, *divmodsi4_zext_1,
13943         *divmodsi4_zext_2): New define_insn_and_split.
13944         (*divmodsi4_noext_zext_1, *divmodsi4_noext_zext_2): New define_insn.
13945         (TARGET_USE_8BIT_IDIV zext udivmodsi4 splitter): New define_split.
13946         (udivmodsi4_zext_1, udivmodsi4_zext_2, *udivmodsi4_zext_1,
13947         *udivmodsi4_zext_2, *udivmodsi4_pow2_zext_1, *udivmodsi4_pow2_zext_2):
13948         New define_insn_and_split.
13949         (*udivmodsi4_noext_zext_1, *udivmodsi4_noext_zext_2): New define_insn.
13950         * config/i386/i386.c (ix86_split_idivmod): Handle operands[0] or
13951         operands[1] having DImode when mode is SImode.
13953         * config/i386/i386.c (ix86_split_idivmod): Use mode instead of
13954         always SImode for DIV and MOD in REG_EQUAL notes.
13956 2017-09-29  Yury Gribov  <tetra2005@gmail.com>
13958         PR middle-end/82319
13959         * match.pd: Fix handling of NaNs in pattern.
13961 2017-09-29  Jeff Law  <law@redhat.com>
13963         * sbitmap.c (bitmap_bit_in_range_p): New function.
13964         * sbitmap.h (bitmap_bit_in_range_p): Prototype.
13965         * tree-ssa-dse.c (live_bytes_read): New function.
13966         (dse_classify_store): Ignore reads of dead bytes.
13968         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): Fix
13969         typos and whitespace errors.
13970         * config/i386/predicates.md (address_no_seg_operand): Likewise.
13971         * config/s390/s390.c (s390_emit_prologue): Likewise.
13973 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
13975         PR target/81481
13976         * ira-costs.c (scan_one_insn): Don't take into account PIC equiv
13977         with a symbol for LRA.
13979 2017-09-29  Vladimir Makarov  <vmakarov@redhat.com>
13981         PR rtl-optimization/82338
13982         * lra-constraints.c (inherit_in_ebb): Check usage_insns check.
13984 2017-09-29  Alexander Monakov  <amonakov@ispras.ru>
13986         * genmodes.c (calc_wider_mode): Suppress qsort macro.
13987         * system.h [CHECKING_P] (qsort): Redirect to qsort_chk.
13988         (qsort_chk): Declare.
13989         * vec.c [CHECKING_P] (qsort_chk_error): New static function.
13990         (qsort_chk): New function.
13992 2017-09-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
13994         PR tree-optimization/82337
13995         * gimple-ssa-strength-reduction.c (find_phi_def): Don't record a
13996         phi definition if the PHI result appears in an abnormal PHI.
13997         (find_basis_for_base_expr): Don't record a basis if the LHS of the
13998         basis appears in an abnormal PHI.
14000 2017-09-29  Richard Biener  <rguenther@suse.de>
14002         * graphite-isl-ast-to-gimple.c
14003         (translate_isl_ast_to_gimple::set_codegen_error): New function.
14004         (binary_op_to_tree): Use it.
14005         (get_rename_from_scev): Likewise.
14006         (copy_loop_phi_nodes): Likewise.
14007         (copy_bb_and_scalar_dependences): Likewise.
14008         (translate_pending_phi_nodes): Likewise.
14010 2017-09-29  Jakub Jelinek  <jakub@redhat.com>
14012         PR target/82339
14013         * config/i386/i386.md (*movdi_internal peephole2): New -Os peephole
14014         for movabsq $(i32 << shift), r64.
14016 2017-09-28  Uros Bizjak  <ubizjak@gmail.com>
14018         * config/i386/i386.c (ix86_print_operand_address_as): Do not check
14019         index when encoding %esp as %rsp to avoid 0x67 prefix.
14021 2017-09-28  Sergey Shalnov  <Sergey.Shalnov@intel.com>
14023         * config/i386/i386.md (*movsf_internal, *movdf_internal):
14024         Return 256-bit AVX modes for TARGET_PREFER_AVX256.
14026 2017-09-28  Thomas Preud'homme  <thomas.preudhomme@arm.com>
14028         * config/arm/arm.c (arm_option_override): Forbid ARMv8-M Security
14029         Extensions with more than 16 double VFP registers.
14030         (cmse_nonsecure_entry_clear_before_return): Remove second entry of
14031         to_clear_mask and all code related to it.  Replace the remaining
14032         entry by a sbitmap and adapt code accordingly.
14034 2017-09-28  Henry Linjamäki  <henry.linjamaki@parmance.com>
14036         * brig-builtins.def: Change pure attributes to const.
14038 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14040         * config.gcc (default_gnu_indirect_function): Default to yes for
14041         sparc*-*-linux* with glibc.
14043 2017-09-28  Joseph Myers  <joseph@codesourcery.com>
14045         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor)
14046         (aarch64_elf_asm_destructor): Pass SECTION_NOTYPE to get_section
14047         when creating .init_array and .fini_array sections with priority
14048         specified.
14050 2017-09-27  Christophe Lyon  <christophe.lyon@linaro.org>
14052         PR target/71727
14053         * config/aarch64/aarch64.c
14054         (aarch64_builtin_support_vector_misalignment): Always return false
14055         when misalignment is unknown.
14057 2017-09-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
14059         * config/rs6000/rs6000-p8swap.c (const_load_sequence_p): Revise
14060         this function to return false if the definition used by the swap
14061         instruction is artificial, or if the memory address from which the
14062         constant value is loaded is not represented by a base address held
14063         in a register or if the base address register is a frame or stack
14064         pointer.  Additionally, return false if the base address of the
14065         loaded constant is a SYMBOL_REF but is not considered to be a
14066         constant.
14067         (replace_swapped_load_constant): New function.
14068         (rs6000_analyze_swaps): Add a new pass to replace a swap of a
14069         loaded constant vector with a load of a swapped constant vector.
14071 2017-09-27  Carl Love  <cel@us.ibm.com>
14073         * config/rs6000/rs6000-builtin.def (BU_FP_1MISC_1): Add define macro.
14074         (FCTID, FCTIW): Add BU_FP_MISC_1 macro expansion for builtins.
14075         * config/rs6000/rs6000.md (lrintsfsi2): Add define_insn for the
14076         fctiw instruction.
14078 2017-09-27  Alexander Monakov  <amonakov@ispras.ru>
14080         * haifa-sched.c (autopref_rank_for_schedule): Order 'irrelevant' insns
14081         first, always call autopref_rank_data otherwise.
14083 2017-09-27  Richard Biener  <rguenther@suse.de>
14085         * graphite-scop-detection.c (find_scop_parameters): Move
14086         loop bound handling ...
14087         (gather_bbs::before_dom_children): ... here, avoiding the need
14088         to build scop_info->loop_nest.
14089         (record_loop_in_sese): Remove.
14090         * sese.h (sese_info_t::loop_nest): Remove.
14091         * sese.c (new_sese_info): Do not allocate loop_nest.
14092         (free_sese_info): Do not free loop_nest.
14094 2017-09-27  Jakub Jelinek  <jakub@redhat.com>
14096         PR c++/82159
14097         * gimplify.c (gimplify_modify_expr): Don't optimize away zero sized
14098         lhs from calls if the lhs has addressable type.
14100 2017-09-27  Richard Biener  <rguenther@suse.de>
14102         * graphite.h (scop::max_alias_set): New member.
14103         * graphite-scop-detection.c: Remove references to non-existing
14104         --param in comments.
14105         (build_alias_sets): Record the maximum alias set used for drs.
14106         (build_scops): Support zero as unlimited for
14107         --param graphite-max-arrays-per-scop.
14108         * graphite-sese-to-poly.c (add_scalar_version_numbers): Remove
14109         and inline into ...
14110         (build_poly_sr_1): ... here.  Compute alias set based on the
14111         maximum alias set used for drs rather than
14112         PARAM_GRAPHITE_MAX_ARRAYS_PER_SCOP
14114 2017-09-27  Richard Biener  <rguenther@suse.de>
14116         * graphite-optimize-isl.c (get_schedule_for_node_st): Allow
14117         --param loop-block-tile-size=0 to disable tiling.
14119 2017-09-27  Richard Biener  <rguenther@suse.de>
14121         * doc/invoke.texi (graphite-max-bbs-per-function): Remove.
14122         (graphite-max-nb-scop-params): Document special value zero.
14123         * domwalk.h (dom_walker::STOP): New symbolical constant.
14124         (dom_walker::dom_walker): Add optional parameter for bb to
14125         RPO mapping.
14126         (dom_walker::~dom_walker): Declare.
14127         (dom_walker::before_dom_children): Document STOP return value.
14128         (dom_walker::m_user_bb_to_rpo): New member.
14129         (dom_walker::m_bb_to_rpo): Likewise.
14130         * domwalk.c (dom_walker::dom_walker): Compute bb to RPO
14131         mapping here if not provided by the user.
14132         (dom_walker::~dom_walker): Free bb to RPO mapping if not
14133         provided by the user.
14134         (dom_walker::STOP): Define.
14135         (dom_walker::walk): Do not compute bb to RPO mapping here.
14136         Support STOP return value from before_dom_children to stop
14137         walking.
14138         * graphite-optimize-isl.c (optimize_isl): If the schedule
14139         is the same still generate code if -fgraphite-identity
14140         or -floop-parallelize-all are given.
14141         * graphite-scop-detection.c: Include cfganal.h.
14142         (gather_bbs::gather_bbs): Get and pass through bb to RPO
14143         mapping.
14144         (gather_bbs::before_dom_children): Return STOP for BBs
14145         not in the region.
14146         (build_scops): Compute bb to RPO mapping and pass it to
14147         the domwalk.  Treat --param graphite-max-nb-scop-params=0
14148         as not limiting the number of params.
14149         * graphite.c (graphite_initialize): Remove limit on the
14150         number of basic-blocks in a function.
14151         * params.def (PARAM_GRAPHITE_MAX_BBS_PER_FUNCTION): Remove.
14152         (PARAM_GRAPHITE_MAX_NB_SCOP_PARAMS): Adjust to documented
14153         default value of 10.
14155 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14157         * config/rs6000/vsx.md (peephole for optimizing move SF to GPR):
14158         Adjust code to eliminate needing to do the shift right 32-bits
14159         operation after XSCVDPSPN.
14161 2017-09-26  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
14163         * match.pd ((X / Y) == 0 -> X < Y): New pattern.
14164         ((X / Y) != 0 -> X >= Y): Likewise.
14166 2017-09-26  Carl Love  <cel@us.ibm.com>
14168         * config/rs6000/rs6000-c.c (P9V_BUILTIN_VEC_XL_LEN_R,
14169         P9V_BUILTIN_VEC_XST_LEN_R): Add support for builtins
14170         vector unsigned char vec_xl_len_r (unsigned char *, size_t);
14171         void vec_xst_len_r (vector unsigned char, unsigned char *, size_t);
14172         * config/rs6000/altivec.h (vec_xl_len_r, vec_xst_len_r): Add defines.
14173         * config/rs6000/rs6000-builtin.def (XL_LEN_R, XST_LEN_R): Add
14174         definitions and overloading.
14175         * config/rs6000/rs6000.c (altivec_expand_builtin): Add case
14176         statement for P9V_BUILTIN_XST_LEN_R.
14177         (altivec_init_builtins): Add def_builtin for P9V_BUILTIN_STXVLL.
14178         * config/rs6000/vsx.md (lxvll, stxvll, xl_len_r, xst_len_r): Add
14179         define_expand and define_insn for the instructions and builtins.
14180         * doc/extend.texi: Update the built-in documentation file for the new
14181         built-in functions.
14182         * config/rs6000/altivec.md (altivec_lvsl_reg, altivec_lvsr_reg): Add
14183         define_insn for the instructions
14185 2017-09-26  Krister Walfridsson  <krister.walfridsson@gmail.com>
14187         PR target/39570
14188         * gcc/config/netbsd-protos.h: New file.
14189         * gcc/config/netbsd.c: New file.
14190         * gcc/config/netbsd.h (SUBTARGET_INIT_BUILTINS): Define.
14191         * gcc/config/t-netbsd: New file.
14192         * gcc/config.gcc (tm_p_file): Add netbsd-protos.h.
14193         (tmake_file) Add t-netbsd.
14194         (extra_objs) Add netbsd.o.
14196 2017-09-26  Janus Weil  <janus@gcc.gnu.org>
14198         PR fortran/82143
14199         PR fortran/82324
14200         * doc/sourcebuild.texi: Document fortran_real_10 and fortran_real_16.
14202 2017-09-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
14204         * config/rs6000/rs6000.md (extendsi<mode>2): Add a splitter to do
14205         sign extension from a vector register to a GPR by doing a 32-bit
14206         direct move and then an EXTSW.
14207         (extendsi<mode>2 splitter): Likewise.
14208         (movsi_from_sf): Adjust code to eliminate doing a 32-bit shift
14209         right or vector extract after doing XSCVDPSPN.  Use
14210         zero_extendsidi2 instead of p8_mfvsrd_4_disf to move the value to
14211         the GPRs.
14212         (movdi_from_sf_zero_ext): Likewise.
14213         (reload_gpr_from_vsxsf): Likewise.
14214         (p8_mfvsrd_4_disf): Delete, no longer used.
14215         (movsi_from_df): Optimize converting a DFmode to a SFmode, and
14216         then needing to move the SFmode to a GPR to use the XSCVDPSP
14217         instruction instead of FRSP and XSCVDPSPN.
14218         * config/rs6000/vsx.md (vsx_xscvspdp_scalar2): Move insn so that
14219         it is adjacent to the other XSCVSPDP insns.
14220         (vsx_xscvdpsp_scalar): Use "ww" constraint instead of "f" to allow
14221         SFmode to be in traditional Altivec registers.
14222         (vsx_xscvdpspn): Eliminate useless alternative constraint.
14223         (vsx_xscvspdpn): Likewise.
14224         (vsx_xscvspdpn_scalar): Likewise.
14226 2017-09-26  Martin Jambor  <mjambor@suse.cz>
14228         * tree-sra.c (compare_access_positions): Put integral types first,
14229         stabilize sorting of integral types, remove conditions putting
14230         non-full-precision integers last.
14231         (sort_and_splice_var_accesses): Disable scalarization if a
14232         non-integert would be represented by a non-full-precision integer.
14234 2017-09-26  Joseph Myers  <joseph@codesourcery.com>
14236         * config/microblaze/linux.h (TARGET_ASM_FILE_END): Likewise.
14237         * config/pa/pa.h (NEED_INDICATE_EXEC_STACK): Likewise.
14238         * config/pa/pa-linux.h (NEED_INDICATE_EXEC_STACK): Likewise.
14239         * config/pa/pa.c (pa_hpux_file_end): Rename to pa_file_end.
14240         Define unconditionally, with [ASM_OUTPUT_EXTERNAL_REAL]
14241         conditionals inside the function instead of around it.  Call
14242         file_end_indicate_exec_stack if NEED_INDICATE_EXEC_STACK.
14243         (TARGET_ASM_FILE_END): Define unconditionally to pa_file_end.
14245 2017-09-26  Richard Biener  <rguenther@suse.de>
14247         * graphite-scop-detection.c (scop_detection::build_scop_depth): Rewrite,
14248         fold in ...
14249         (scop_detection::build_scop_breadth): ... this.  Removed.
14250         (scop_detection::loop_is_valid_in_scop): Fold into single caller.
14251         (scop_detection::harmful_stmt_in_bb): Likewise.
14252         (scop_detection::graphite_can_represent_stmt): Likewise.
14253         (scop_detection::loop_body_is_valid_scop): Likewise.  Remove recursion.
14254         (scop_detection::can_represent_loop): Remove recursion, fold in ...
14255         (scop_detection::can_represent_loop_1): ... this.  Removed.
14256         (scop_detection::harmful_loop_in_region): Simplify after inlining
14257         the above and remove more quadraticness.
14258         (build_scops): Adjust.
14259         * tree-data-ref.c (loop_nest_has_data_refs): Remove pointless
14260         quadraticness.
14262 2017-09-26  Jakub Jelinek  <jakub@redhat.com>
14264         PR target/82267
14265         * config/i386/i386.c (ix86_print_operand_address_as): Only test
14266         REGNO (base) == SP_REG if base is a REG.
14268         PR middle-end/35691
14269         * tree-ssa-reassoc.c (update_range_test): Dump r->exp each time
14270         if it is different SSA_NAME.
14271         (optimize_range_tests_cmp_bitwise): New function.
14272         (optimize_range_tests): Call it.
14274 2017-09-26  Richard Biener  <rguenther@suse.de>
14276         PR tree-optimization/82321
14277         * graphite.c (canonicalize_loop_closed_ssa): Properly check
14278         for the def being inside the loop.
14280 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14282         * config/s390/vx-builtins.md ("vmslg"): Add missing operand in
14283         assembler output.
14284         * config/s390/s390-builtins.def: Fix constraint on op4.
14286 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14288         * config/s390/s390.c (s390_expand_vec_compare): Use the new mode
14289         independent expanders.
14290         * config/s390/vector.md ("vec_cmpuneq", "vec_cmpltgt")
14291         ("vec_ordered", "vec_unordered"): New expanders.
14293 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14295         * config/s390/s390.c (s390_preferred_simd_mode): Return V4SFmode
14296         for SFmode.
14298 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14300         * config/s390/vector.md ("vec_unpacks_low_v16qi"): Rename to
14301         vec_unpacks_lo_v16qi.
14302         ("vec_unpacku_low_v16qi"): Rename to vec_unpacku_lo_v16qi.
14304 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14306         * config/s390/vector.md ("vec_unpacks_lo_v4sf")
14307         ("vec_unpacks_hi_v4sf", "vec_unpacks_lo_v2df")
14308         ("vec_unpacks_hi_v2df", "vec_pack_trunc_v2df"): New expanders.
14310 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14312         * config/s390/predicates.md ("const_shift_by_byte_operand"): New
14313         predicate.
14314         * config/s390/vector.md ("*vec_srb<mode>"): Change modes to V_128
14315         and V16QI.
14316         ("*vec_slb<mode>"): New insn pattern.
14317         ("vec_shr_<mode>"): New expander.
14318         * config/s390/vx-builtins.md ("vec_slb<mode>"): Turn into expander
14319         and force the shift count operand to V16QImode.
14320         ("vec_srb<mode>"): Set shift count mode to V16QI.
14322 2017-09-26  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14324         * config/s390/vector.md ("vec_widen_umult_lo_<mode>")
14325         ("vec_widen_umult_hi_<mode>", "vec_widen_smult_lo_<mode>")
14326         ("vec_widen_smult_hi_<mode>"): New expander definitions.
14328 2017-09-26  Richard Earnshaw  <rearnsha@arm.com>
14330         PR target/82175
14331         * config/arm/arm.h (DRIVER_SELF_SPECS): Separate sub-rules with commas.
14333 2017-09-26  Richard Biener  <rguenther@suse.de>
14335         PR tree-optimization/82320
14336         * tree-ssa-sccvn.c (set_ssa_val_to): Changing undef to undef
14337         isn't a change.
14339 2017-09-25  Jeff Law  <law@redhat.com>
14341         * config/rs6000/rs6000-protos.h (output_probe_stack_range): Update
14342         prototype for new argument.
14343         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack_1): New function,
14344         mostly extracted from rs6000_emit_allocate_stack.
14345         (rs6000_emit_probe_stack_range_stack_clash): New function.
14346         (rs6000_emit_allocate_stack): Call
14347         rs6000_emit_probe_stack_range_stack_clash as needed.
14348         (rs6000_emit_probe_stack_range): Add additional argument
14349         to call to gen_probe_stack_range{si,di}.
14350         (output_probe_stack_range): New.
14351         (output_probe_stack_range_1): Renamed from output_probe_stack_range.
14352         (output_probe_stack_range_stack_clash): New.
14353         (rs6000_emit_prologue): Emit notes into dump file as requested.
14354         * rs6000.md (allocate_stack): Handle -fstack-clash-protection.
14355         (probe_stack_range<P:mode>): Operand 0 is now early-clobbered.
14356         Add additional operand and pass it to output_probe_stack_range.
14358 2017-09-25  Bin Cheng  <bin.cheng@arm.com>
14360         PR tree-optimization/82163
14361         * tree-ssa-loop-manip.h (verify_loop_closed_ssa): New parameter.
14362         (checking_verify_loop_closed_ssa): New parameter.
14363         * tree-ssa-loop-manip.c (check_loop_closed_ssa_use): Delete.
14364         (check_loop_closed_ssa_stmt): Delete.
14365         (check_loop_closed_ssa_def, check_loop_closed_ssa_bb): New functions.
14366         (verify_loop_closed_ssa): Check loop closed ssa form for LOOP.
14367         (tree_transform_and_unroll_loop): Check loop closed ssa form only for
14368         changed loops.
14370 2017-09-25  Pekka Jaaskelainen <pekka@parmance.com>
14372         * brig-builtins.def: Treat HSAIL barrier builtins as
14373         setjmp/longjump style functions.
14375 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14377         * target.def (constant_alignment): New hook.
14378         * defaults.h (CONSTANT_ALIGNMENT): Delete.
14379         * doc/tm.texi.in (CONSTANT_ALIGNMENT): Replace with...
14380         (TARGET_CONSTANT_ALIGNMENT): ...this new hook.
14381         * doc/tm.texi: Regenerate.
14382         * targhooks.h (default_constant_alignment): Declare.
14383         (constant_alignment_word_strings): Likewise.
14384         * targhooks.c (default_constant_alignment): New function.
14385         (constant_alignment_word_strings): Likewise.
14386         * builtins.c (get_object_alignment_2): Use targetm.constant_alignment
14387         instead of CONSTANT_ALIGNMENT.
14388         * varasm.c (align_variable, get_variable_align, build_constant_desc)
14389         (force_const_mem): Likewise.
14390         * config/aarch64/aarch64.h (CONSTANT_ALIGNMENT): Delete.
14391         * config/aarch64/aarch64.c (aarch64_constant_alignment): New function.
14392         (aarch64_classify_address): Call it instead of CONSTANT_ALIGNMENT.
14393         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14394         * config/alpha/alpha.h (CONSTANT_ALIGNMENT): Delete commented-out
14395         definition.
14396         * config/arc/arc.h (CONSTANT_ALIGNMENT): Delete.
14397         * config/arc/arc.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14398         constant_alignment_word_strings.
14399         * config/arm/arm.h (CONSTANT_ALIGNMENT_FACTOR): Delete.
14400         (CONSTANT_ALIGNMENT): Likewise.
14401         * config/arm/arm.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14402         (arm_constant_alignment): New function.
14403         * config/bfin/bfin.h (CONSTANT_ALIGNMENT): Delete.
14404         * config/bfin/bfin.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14405         constant_alignment_word_strings.
14406         * config/cr16/cr16.h (CONSTANT_ALIGNMENT): Delete.
14407         * config/cr16/cr16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14408         constant_alignment_word_strings.
14409         * config/cris/cris.h (CONSTANT_ALIGNMENT): Delete.
14410         * config/cris/cris.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14411         (cris_constant_alignment): New function.
14412         * config/epiphany/epiphany.h (CONSTANT_ALIGNMENT): Delete.
14413         * config/epiphany/epiphany.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14414         (epiphany_constant_alignment): New function.
14415         * config/fr30/fr30.h (CONSTANT_ALIGNMENT): Delete.
14416         * config/fr30/fr30.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14417         constant_alignment_word_strings.
14418         * config/frv/frv.h (CONSTANT_ALIGNMENT): Delete.
14419         * config/frv/frv.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14420         constant_alignment_word_strings.
14421         * config/ft32/ft32.h (CONSTANT_ALIGNMENT): Delete.
14422         * config/ft32/ft32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14423         constant_alignment_word_strings.
14424         * config/i386/i386.h (CONSTANT_ALIGNMENT): Delete.
14425         * config/i386/i386-protos.h (ix86_constant_alignment): Delete.
14426         * config/i386/i386.c (ix86_constant_alignment): Make static.
14427         Use the same interface as the target hook.
14428         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14429         * config/ia64/ia64.h (CONSTANT_ALIGNMENT): Delete.
14430         * config/ia64/ia64.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14431         constant_alignment_word_strings.
14432         * config/iq2000/iq2000.h (CONSTANT_ALIGNMENT): Delete.
14433         * config/iq2000/iq2000.c (iq2000_constant_alignment): New function.
14434         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14435         * config/lm32/lm32.h (CONSTANT_ALIGNMENT): Delete.
14436         * config/lm32/lm32.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14437         constant_alignment_word_strings.
14438         * config/m32r/m32r.h (CONSTANT_ALIGNMENT): Delete.
14439         * config/m32r/m32r.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14440         constant_alignment_word_strings.
14441         * config/mcore/mcore.h (CONSTANT_ALIGNMENT): Delete.
14442         * config/mcore/mcore.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14443         constant_alignment_word_strings.
14444         * config/microblaze/microblaze.h (CONSTANT_ALIGNMENT): Delete.
14445         * config/microblaze/microblaze.c (microblaze_constant_alignment):
14446         New function.
14447         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14448         * config/mips/mips.h (CONSTANT_ALIGNMENT): Delete.
14449         * config/mips/mips.c (mips_constant_alignment): New function.
14450         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14451         * config/mmix/mmix.h (CONSTANT_ALIGNMENT): Delete.
14452         * config/mmix/mmix-protos.h (mmix_constant_alignment): Delete.
14453         * config/mmix/mmix.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14454         (mmix_constant_alignment): Make static.  Use the same interface
14455         as the target hook.
14456         * config/moxie/moxie.h (CONSTANT_ALIGNMENT): Delete.
14457         * config/moxie/moxie.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14458         constant_alignment_word_strings.
14459         * config/nios2/nios2.h (CONSTANT_ALIGNMENT): Delete.
14460         * config/nios2/nios2.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14461         constant_alignment_word_strings.
14462         * config/pa/pa.h (CONSTANT_ALIGNMENT): Delete.
14463         * config/pa/pa.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14464         constant_alignment_word_strings.
14465         * config/powerpcspe/powerpcspe.h (CONSTANT_ALIGNMENT): Delete.
14466         * config/powerpcspe/powerpcspe.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14467         (rs6000_constant_alignment): New function.
14468         * config/riscv/riscv.h (CONSTANT_ALIGNMENT): Delete.
14469         * config/riscv/riscv.c (riscv_constant_alignment): New function.
14470         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14471         * config/rs6000/rs6000.h (CONSTANT_ALIGNMENT): Delete.
14472         * config/rs6000/rs6000.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14473         (rs6000_constant_alignment): New function.
14474         * config/s390/s390.h (CONSTANT_ALIGNMENT): Delete.
14475         * config/s390/s390.c (s390_constant_alignment): New function.
14476         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14477         * config/sh/sh.h (CONSTANT_ALIGNMENT): Delete.
14478         * config/sh/sh.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14479         constant_alignment_word_strings.
14480         * config/sparc/sparc.h (CONSTANT_ALIGNMENT): Delete.
14481         * config/sparc/sparc.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14482         (sparc_constant_alignment): New function.
14483         * config/spu/spu.h (CONSTANT_ALIGNMENT): Delete.
14484         * config/spu/spu.c (spu_constant_alignment): New function.
14485         (TARGET_CONSTANT_ALIGNMENT): Redefine.
14486         * config/stormy16/stormy16.h (CONSTANT_ALIGNMENT): Delete.
14487         * config/stormy16/stormy16.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14488         constant_alignment_word_strings.
14489         * config/tilegx/tilegx.h (CONSTANT_ALIGNMENT): Delete.
14490         * config/tilegx/tilegx.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14491         constant_alignment_word_strings.
14492         * config/tilepro/tilepro.h (CONSTANT_ALIGNMENT): Delete.
14493         * config/tilepro/tilepro.c (TARGET_CONSTANT_ALIGNMENT): Redefine to
14494         constant_alignment_word_strings.
14495         * config/visium/visium.h (CONSTANT_ALIGNMENT): Delete.
14496         * config/visium/visium.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14497         (visium_constant_alignment): New function.
14498         * config/xtensa/xtensa.h (CONSTANT_ALIGNMENT): Delete.
14499         * config/xtensa/xtensa.c (TARGET_CONSTANT_ALIGNMENT): Redefine.
14500         (xtensa_constant_alignment): New function.
14501         * system.h (CONSTANT_ALIGNMENT): Poison.
14503 2017-09-25  Will Schmidt  <will_schmidt@vnet.ibm.com>
14505         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
14506         for early folding of vector stores (ALTIVEC_BUILTIN_ST_*).
14507         (rs6000_builtin_valid_without_lhs): New helper function.
14508         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
14509         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_ST.
14511 2017-09-25  Richard Sandiford  <richard.sandiford@linaro.org>
14513         * target.h (vec_perm_indices): Use unsigned short rather than
14514         unsigned char.
14515         (auto_vec_perm_indices): Likewise.
14516         * config/aarch64/aarch64.c (aarch64_vectorize_vec_perm_const_ok):
14517         Use unsigned int rather than unsigned char.
14518         * config/arm/arm.c (arm_vectorize_vec_perm_const_ok): Likewise.
14520 2017-09-25  Richard Biener  <rguenther@suse.de>
14522         * cfgloop.h (sort_sibling_loops): Declare.
14523         * cfgloop.c (sort_sibling_loops_cmp): New helper.
14524         (sort_sibling_loops): New function sorting the sibling loop list
14525         in RPO order.
14526         * graphite.c (graphite_transform_loops): Sort sibling loops.
14528 2017-09-25  Richard Sandiford  <richard.sandifird@linaro.org>
14530         * target.def (vec_perm_const_ok): Change sel parameter to
14531         vec_perm_indices.
14532         * optabs-query.c (can_vec_perm_p): Update accordingly.
14533         * doc/tm.texi: Regenerate.
14534         * config/aarch64/aarch64.c (expand_vec_perm_d): Change perm to
14535         auto_vec_perm_indices and remove separate nelt field.
14536         (aarch64_evpc_trn, aarch64_evpc_uzp, aarch64_evpc_zip)
14537         (aarch64_evpc_ext, aarch64_evpc_rev, aarch64_evpc_dup)
14538         (aarch64_evpc_tbl, aarch64_expand_vec_perm_const_1)
14539         (aarch64_expand_vec_perm_const): Update accordingly.
14540         (aarch64_vectorize_vec_perm_const_ok): Likewise.  Change sel
14541         to vec_perm_indices.
14542         * config/arm/arm.c (expand_vec_perm_d): Change perm to
14543         auto_vec_perm_indices and remove separate nelt field.
14544         (arm_evpc_neon_vuzp, arm_evpc_neon_vzip, arm_evpc_neon_vrev)
14545         (arm_evpc_neon_vtrn, arm_evpc_neon_vext, arm_evpc_neon_vtbl)
14546         (arm_expand_vec_perm_const_1, arm_expand_vec_perm_const): Update
14547         accordingly.
14548         (arm_vectorize_vec_perm_const_ok): Likewise.  Change sel
14549         to vec_perm_indices.
14550         * config/i386/i386.c (ix86_vectorize_vec_perm_const_ok): Change
14551         sel to vec_perm_indices.
14552         * config/ia64/ia64.c (ia64_vectorize_vec_perm_const_ok): Likewise.
14553         * config/mips/mips.c (mips_vectorize_vec_perm_const_ok): Likewise.
14554         * config/powerpcspe/powerpcspe.c (rs6000_vectorize_vec_perm_const_ok):
14555         Likewise.
14556         * config/rs6000/rs6000.c (rs6000_vectorize_vec_perm_const_ok):
14557         Likewise.
14559 2017-09-25  Pierre-Marie de Rodat  <derodat@adacore.com>
14561         PR debug/82155
14562         * dwarf2out.c (dwarf2out_early_global_decl): Call dwarf2out_decl
14563         on the FUNCTION_DECL function context if it has a DIE that is a
14564         declaration.
14566 2017-09-25  Richard Biener  <rguenther@suse.de>
14568         PR tree-optimization/82285
14569         * tree-vect-patterns.c (vect_recog_bool_pattern): Also handle
14570         enumeral types.
14572 2017-09-25  Tom de Vries  <tom@codesourcery.com>
14574         PR target/80035
14575         PR target/81069
14576         * config/nvptx/nvptx.c (nvptx_output_call_insn): Add exit after call to
14577         noreturn function.
14579 2017-09-25  Richard Biener  <rguenther@suse.de>
14581         * graphite-optimize-isl.c (optimize_isl): Fail and dump if
14582         ISL errors other than isl_error_quota happen.  Dump if the
14583         schedule is the same.
14584         * graphite-sese-to-poly.c (build_poly_scop): Fail on ISL
14585         errors instead of aborting inside ISL.
14587 2017-09-25  Iain Sandoe  <iain@codesourcery.com>
14589         PR target/80556
14590         * config/i386/darwin.h (REAL_LIB_SPEC): New; put libSystem ahead
14591         of libgcc_eh for m64.
14592         * config/i386/darwin64.h: Likewise.
14594 2017-09-25  Richard Biener  <rguenther@suse.de>
14596         PR middle-end/82144
14597         * dwarf2out.c (gen_enumeration_type_die): Do not add alignment
14598         attribute for incomplete types nor twice for complete ones.
14600 2017-09-24  Uros Bizjak  <ubizjak@gmail.com>
14602         PR target/82267
14603         * config/i386/i386.c (ix86_print_operand_address_as): Encode
14604         %esp as %rsp to avoid 0x67 prefix if there is no index or base
14605         register.
14607 2017-09-23  Uros Bizjak  <ubizjak@gmail.com>
14609         PR bootstrap/82306
14610         * config/i386/i386.opt (mprefer-avx256): Use
14611         ix86_target_flags variable.
14612         * config/i386/i386.c (ix86_target_string): Move
14613         -mprefer-avx256 to flag2_opts.
14615 2017-09-22  Jakub Jelinek  <jakub@redhat.com>
14617         PR middle-end/35691
14618         * match.pd: Simplify x == -1 & y == -1 into (x & y) == -1
14619         and x != -1 | y != -1 into (x & y) != -1.
14621 2017-09-22  Steve Ellcey  <sellcey@cavium.com>
14623         * config.gcc: Add new case statement to set
14624         default_gnu_indirect_function.  Remove it from x86_64-*-linux*,
14625         i[34567]86-*, powerpc*-*-linux*spe*, powerpc*-*-linux*, s390-*-linux*,
14626         s390x-*-linux* case statements.   Added aarch64 to the list of
14627         supported architectures.
14629 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14631         PR tree-optimization/82289
14632         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Check
14633         STMT_VINFO_RELEVANT_P.
14635 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14636             Alan Hayward  <alan.hayward@arm.com>
14637             David Sherwood  <david.sherwood@arm.com>
14639         * tree-vrp.c (extract_range_from_multiplicative_op_1): Assert
14640         for VR_RANGE only; don't allow VR_ANTI_RANGE.
14641         (extract_range_from_binary_expr_1): Don't call
14642         extract_range_from_multiplicative_op_1 if !range_int_cst_p.
14644 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14645             Alan Hayward  <alan.hayward@arm.com>
14646             David Sherwood  <david.sherwood@arm.com>
14648         * target.def (preferred_vector_alignment): New hook.
14649         * doc/tm.texi.in (TARGET_VECTORIZE_PREFERRED_VECTOR_ALIGNMENT): New
14650         hook.
14651         * doc/tm.texi: Regenerate.
14652         * targhooks.h (default_preferred_vector_alignment): Declare.
14653         * targhooks.c (default_preferred_vector_alignment): New function.
14654         * tree-vectorizer.h (dataref_aux): Add a target_alignment field.
14655         Expand commentary.
14656         (DR_TARGET_ALIGNMENT): New macro.
14657         (aligned_access_p): Update commentary.
14658         (vect_known_alignment_in_bytes): New function.
14659         * tree-vect-data-refs.c (vect_calculate_required_alignment): New
14660         function.
14661         (vect_compute_data_ref_alignment): Set DR_TARGET_ALIGNMENT.
14662         Calculate the misalignment based on the target alignment rather than
14663         the vector size.
14664         (vect_update_misalignment_for_peel): Use DR_TARGET_ALIGMENT
14665         rather than TYPE_ALIGN / BITS_PER_UNIT to update the misalignment.
14666         (vect_enhance_data_refs_alignment): Mask the byte misalignment with
14667         the target alignment, rather than masking the element misalignment
14668         with the number of elements in a vector.  Also use the target
14669         alignment when calculating the maximum number of peels.
14670         (vect_find_same_alignment_drs): Use vect_calculate_required_alignment
14671         instead of TYPE_ALIGN_UNIT.
14672         (vect_duplicate_ssa_name_ptr_info): Remove stmt_info parameter.
14673         Measure DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT.
14674         (vect_create_addr_base_for_vector_ref): Update call accordingly.
14675         (vect_create_data_ref_ptr): Likewise.
14676         (vect_setup_realignment): Realign by ANDing with
14677         -DR_TARGET_MISALIGNMENT.
14678         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Calculate
14679         the number of peels based on DR_TARGET_ALIGNMENT.
14680         * tree-vect-stmts.c (get_group_load_store_type): Compare the gap
14681         with the guaranteed alignment boundary when deciding whether
14682         overrun is OK.
14683         (vectorizable_mask_load_store): Interpret DR_MISALIGNMENT
14684         relative to DR_TARGET_ALIGNMENT instead of TYPE_ALIGN_UNIT.
14685         (ensure_base_align): Remove stmt_info parameter.  Get the
14686         target base alignment from DR_TARGET_ALIGNMENT.
14687         (vectorizable_store): Update call accordingly.   Interpret
14688         DR_MISALIGNMENT relative to DR_TARGET_ALIGNMENT instead of
14689         TYPE_ALIGN_UNIT.
14690         (vectorizable_load): Likewise.
14692 2017-09-22  Richard Sandiford  <richard.sandiford@linaro.org>
14693             Alan Hayward  <alan.hayward@arm.com>
14694             David Sherwood  <david.sherwood@arm.com>
14696         * tree-vectorizer.h (vect_get_scalar_dr_size): New function.
14697         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Use it.
14698         (vect_enhance_data_refs_alignment): Likewise.
14700 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14702         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
14703         error.  Only quit immediately if parsing is complete.
14704         (BEGIN): Initialize fatal_err and parse_done.
14705         (begin fpu, end fpu): Check number of arguments.
14706         (begin arch, end arch): Likewise.
14707         (begin cpu, end cpu): Likewise.
14708         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
14709         (optalias): Likewise.
14711 2017-09-22  Richard Earnshaw  <richard.earnshaw@arm.com>
14713         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
14714         * config/arm/arm-isa.h: Delete.  Move definitions to ...
14715         * arm-cpus.in: ... here.  Use new feature and fgroup values.
14716         * config/arm/arm.c (arm_option_override): Use lower case for feature
14717         bit names.
14718         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
14719         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
14720         * config/arm/parsecpu.awk (END): Add new command 'isa'.
14721         (isa_pfx): Delete.
14722         (print_isa_bits_for): New function.
14723         (gen_isa): New function.
14724         (gen_comm_data): Use print_isa_bits_for.
14725         (define feature): New keyword.
14726         (define fgroup): New keyword.
14727         * config/arm/t-arm (TM_H): Remove.
14728         (GTM_H): Add arm-isa.h.
14729         (arm-isa.h): Add rule to generate file.
14730         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
14731         case for feature bit names.
14733 2017-09-22  Richard Biener  <rguenther@suse.de>
14735         * graphite-isl-ast-to-gimple.c (graphite_verify): Inline into
14736         single caller.
14737         (graphite_regenerate_ast_isl): Do not reset SCEV.  Move debug
14738         print of no dependency loops ...
14739         * graphite.c (graphite_transform_loops): ... here.
14740         (canonicalize_loop_closed_ssa_form): Work from inner to outer
14741         loops.
14742         (same_close_phi_node, remove_duplicate_close_phi,
14743         make_close_phi_nodes_unique, defined_in_loop_p): Fold into ...
14744         (canonicalize_loop_closed_ssa): ... here and simplify.
14745         * graphite-optimize-isl.c: Include tree-vectorizer.h.
14746         (optimize_isl): Use dump_printf_loc to tell when we stopped
14747         optimizing because of an ISL timeout.
14749 2017-09-22  Richard Biener  <rguenther@suse.de>
14751         PR tree-optimization/82291
14752         * tree-if-conv.c (predicate_mem_writes): Make sure to
14753         remove writes in blocks predicated with false.
14755 2017-09-22  Richard Biener  <rguenther@suse.de>
14757         * sese.c: Include cfganal.h.
14758         (if_region_set_false_region): Remove.
14759         (create_if_region_on_edge): Likewise.
14760         (move_sese_in_condition): Re-implement without destroying
14761         dominators.
14763 2017-09-22  Richard Biener  <rguenther@suse.de>
14765         * graphite-isl-ast-to-gimple.c (translate_pending_phi_nodes):
14766         Verify both BBs contain loop PHI nodes before dispatching to
14767         copy_loop_phi_args.
14768         (graphite_regenerate_ast_isl): Do not recompute dominators,
14769         do not verify three times.  Restructure for clarity.
14770         * graphite-scop-detection.c (same_close_phi_node,
14771         remove_duplicate_close_phi, make_close_phi_nodes_unique,
14772         defined_in_loop_p, canonicalize_loop_closed_ssa,
14773         canonicalize_loop_closed_ssa_form): Simplify, remove excess
14774         checking and SSA rewrite, move to ...
14775         * graphite.c: ... here.  Include ssa.h and tree-ssa-loop-manip.h.
14776         (graphite_initialize): Do not pass in ctx, do not reset the
14777         SCEV cache, compute only dominators.
14778         (graphite_transform_loops): Allocate ISL ctx after
14779         graphite_initialize.  Call canonicalize_loop_closed_ssa_form.
14780         Maintain post-dominators only around build_scops.
14781         * sese.c (if_region_set_false_region): Make static.  Free
14782         and recompute dominators.
14783         (move_sese_in_condition): Assert we don't get called with
14784         post-dominators computed.
14785         * sese.h (if_region_set_false_region): Remove.
14787 2017-09-22  Sergey Shalnov  <sergey.shalnov@intel.com>
14789         * config/i386/sse.md ("mov<mode>_internal"): Use <sseinsnmode>
14790         mode attribute for TARGET_AVX512VL.
14792 2017-09-21  Sergey Shalnov  <sergey.shalnov@intel.com>
14794         * config/i386/i386.opt (mprefer-avx256): New option.
14795         * config/i386/i386.c (ix86_target_string): Add -mprefer-avx256
14796         to flag_opts.
14797         (ix86_preferred_simd_mode): Return 256-bit AVX modes
14798         for TARGET_PREFER_AVX256.
14799         * doc/invoke.texi (x86 Options): Document -mprefer-avx256.
14801 2017-09-21  Jeff Law  <law@redhat.com>
14803         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash):
14804         Fix dump output if the only stack space is for pushed registers.
14806 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14808         * config/spu/spu.c (spu_sched_adjust_cost): Update after renaming
14809         of insn_cost.
14811 2017-09-21  Martin Sebor  <msebor@redhat.com>
14813         PR c/81882
14814         * doc/extend.texi (attribute ifunc): Avoid relying on ill-formed
14815         code (in C++) or code that triggers warnings.
14817 2017-09-21  Eric Botcazou  <ebotcazou@adacore.com>
14819         * stor-layout.c (bit_from_pos): Do not distribute the conversion.
14821 2017-09-21  Segher Boessenkool  <segher@kernel.crashing.org>
14823         * haifa-sched.c: Rename insn_cost to insn_sched_cost.
14824         * sched-rgn.c: Ditto.
14825         * sel-sched-ir.c: Ditto.
14827 2017-09-21  Alexander Monakov  <amonakov@ispras.ru>
14829         * toplev.h (set_random_seed): Adjust return type.
14830         * toplev.c (init_local_tick): Move eager initialization of random_seed
14831         to get_random_seed.  Adjust comment.
14832         (init_random_seed): Inline to get_random_seed, delete.
14833         (get_random_seed): Initialize random_seed lazily.
14834         (set_random_seed): Do not return previous value.
14835         (print_switch_value): Do not call get_random_seed.
14837 2017-09-21  Evgeny Kudryashov  <kudryashov@ispras.ru>
14839         * cgraph.c (delete_function_version): New, broken out from...
14840         (cgraph_node::delete_function_version): ...here.  Rename to
14841         cgraph_node::delete_function_version_by_decl.  Update all uses.
14842         (cgraph_node::remove): Call delete_function_version.
14844 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14846         PR sanitizer/81715
14847         * tree-inline.c (expand_call_inline): Emit clobber stmts for
14848         VAR_DECLs to which addressable non-volatile parameters are mapped
14849         and for id->retvar after the return value assignment.  Clear
14850         id->retval and id->retbnd after inlining.
14852 2017-09-21  Richard Biener  <rguenther@suse.de>
14854         PR tree-optimization/82276
14855         PR tree-optimization/82244
14856         * tree-vrp.c (build_assert_expr_for): Set
14857         SSA_NAME_OCCURS_IN_ABNORMAL_PHI if the variable we assert on
14858         has it set.
14859         (remove_range_assertions): Revert earlier change.
14861 2017-09-21  Wilco Dijkstra  <wdijkstr@arm.com>
14863         PR target/71951
14864         * config/aarch64/aarch64.h (LIBGCC2_UNWIND_ATTRIBUTE): Define.
14866 2017-09-21  Richard Biener  <rguenther@suse.de>
14868         * graphite-isl-ast-to-gimple.c (graphite_regenerate_ast_isl):
14869         Restore valid IL after code generation errors.
14870         * graphite.c (graphite_transform_loops): Diagnose code
14871         generation issues as MSG_MISSED_OPTIMIZATION and continue
14872         with processing SCOPs.
14874 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14875             Alan Hayward  <alan.hayward@arm.com>
14876             David Sherwood  <david.sherwood@arm.com>
14878         * calls.c (compute_argument_addresses): Use simplify_gen_binary
14879         rather than choosing between plus_constant and gen_rtx_<CODE>.
14880         * expr.c (emit_push_insn): Likewise.
14881         (expand_expr_real_2): Likewise.
14883 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14884             Alan Hayward  <alan.hayward@arm.com>
14885             David Sherwood  <david.sherwood@arm.com>
14887         * loop-unroll.c (split_iv): Call copy_rtx on the step.
14889 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14890             Alan Hayward  <alan.hayward@arm.com>
14891             David Sherwood  <david.sherwood@arm.com>
14893         * tree.c (find_atomic_core_type): Check tree_fits_uhwi_p before
14894         calling tree_to_uhwi.
14896 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14897             Alan Hayward  <alan.hayward@arm.com>
14898             David Sherwood  <david.sherwood@arm.com>
14900         * tree-ssa-ccp.c (get_value_for_expr): Use a positive test for
14901         INTEGER_CST rather than a negative test for ADDR_EXPR.
14903 2017-09-21  Richard Sandiford  <richard.sandiford@linaro.org>
14904             Alan Hayward  <alan.hayward@arm.com>
14905             David Sherwood  <david.sherwood@arm.com>
14907         * tree-vrp.c (extract_range_from_binary_expr_1): Check
14908         int_cst_rangeN before calling value_range_constant_singleton (&vrN).
14910 2017-09-21  Richard Biener  <rguenther@suse.de>
14912         PR tree-optimization/71351
14913         * graphite-isl-ast-to-gimple.c (translate_isl_ast_to_gimple::
14914         graphite_create_new_loop_guard): Remove, fold remaining parts
14915         into caller ...
14916         (translate_isl_ast_node_for): ... here and simplify.
14918 2017-09-21  Jakub Jelinek  <jakub@redhat.com>
14920         PR target/82260
14921         * config/i386/i386.md (*movqi_internal): Replace (=q,q) alternative
14922         with (=Q,Q), (=R,R) and (=r,r) alternatives, only enable the
14923         latter two for 64-bit, renumber alternatives, for -Os imov (=q,n)
14924         alternative always use QI mode, for -Os imov (=R,R) alternative
14925         always use SI mode, for imov (=Q,Q) or (=r,r) alternatives
14926         ignore -Os.
14928 2017-09-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
14929             Jeff Law  <law@redhat.com>
14931         * config/s390/s390.c (MIN_UNROLL_PROBES): Define.
14932         (allocate_stack_space): New function, partially extracted from
14933         s390_emit_prologue.
14934         (s390_emit_prologue): Track offset to most recent stack probe.
14935         Code to allocate space moved into allocate_stack_space.
14936         Dump actions when no stack is allocated.
14937         (s390_prologue_plus_offset): New function.
14938         (s390_emit_stack_probe): Likewise.
14940 2017-09-20  Alexandre Oliva  <aoliva@redhat.com>
14942         * common.opt (Wa, Wl, Wp, g, gz=): Add
14943         RejectNegative.
14944         (gno-column-info): Remove.
14945         (gcolumn-info): Drop RejectNegative.
14946         (gno-): New prefix.
14947         (gno-record-gcc-switches): Remove.
14948         (grecord-gcc-switches): Drop RejectNegative.
14949         (gno-split-dwarf): Remove.
14950         (gsplit-dwarf): Drop RejectNegative.
14951         (gno-strict-dwarf): Remove.
14952         (gstrict-dwarf): Drop RejectNegative.
14953         * config/darwin.opt (gfull, gused): Add RejectNegative.
14954         * dwarf2out.c (gen_producer_string): Drop
14955         gno-record-gcc-switches handler.
14956         * optc-gen.awk: Add g to prefixes with negative forms.
14957         * opts-common.c (remapping_prefix_p): New.
14958         (find_opt): Check it.
14959         (generate_canonical_option): Test g prefix.
14960         (option_map): Add -gno- mapping.
14961         (add_misspelling_candidates): Check remapping_prefix_p.
14963 2017-09-20  Jeff Law  <law@redhat.com>
14965         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Fix
14966         thinko in stack clash protection support.
14968         * explow.c (compute_stack_clash_protection_loop_data): Use
14969         CONST_INT_P instead of explicit test.  Verify object is a
14970         CONST_INT_P before looking at INTVAL.
14971         (anti_adjust_stack_and_probe_stack_clash): Use CONST_INT_P
14972         instead of explicit test.
14974 2017-09-20  Segher Boessenkool  <segher@kernel.crashing.org>
14976         PR target/77687
14977         * config/rs6000/rs6000.md (stack_restore_tie): Store to a scratch
14978         address instead of to r1 and r11.
14980 2017-09-20  Sebastian Peryt  <sebastian.peryt@intel.com>
14982         * config.gcc: Support "knm".
14983         * config/i386/driver-i386.c (host_detect_local_cpu): Detect "knm".
14984         * config/i386/i386-c.c (ix86_target_macros_internal): Handle
14985         PROCESSOR_KNM.
14986         * config/i386/i386.c (m_KNM): Define.
14987         (processor_target_table): Add "knm".
14988         (PTA_KNM): Define.
14989         (ix86_option_override_internal): Add "knm".
14990         (ix86_issue_rate): Add PROCESSOR_KNM.
14991         (ix86_adjust_cost): Ditto.
14992         (ia32_multipass_dfa_lookahead): Ditto.
14993         (get_builtin_code_for_version): Handle PROCESSOR_KNM.
14994         (fold_builtin_cpu): Add M_INTEL_KNM.
14995         * config/i386/i386.h (processor_costs): Define TARGET_KNM.
14996         (processor_type): Add PROCESSOR_KNM.
14997         * config/i386/x86-tune.def: Add m_KNM.
14998         * doc/invoke.texi: Add knm as x86 -march=/-mtune= CPU type.
15000 2017-09-20  Richard Biener  <rguenther@suse.de>
15002         PR tree-optimization/80213
15003         * graphite-scop-detection.c (trivially_empty_bb_p): Labels
15004         are allowed in empty BBs as well.
15005         (canonicalize_loop_closed_ssa): Also look for other complex
15006         edges.
15007         (scop_detection::get_sese): Include the loop-closed PHI block
15008         in loop SESEs.
15009         (scop_detection::merge_sese): Remove code adding extra blocks.
15010         (scop_detection::region_has_one_loop): Adjust for get_sese changes.
15011         (build_scops): Assert the final returned scop is invalid.
15013 2017-09-20  Richard Biener  <rguenther@suse.de>
15015         PR tree-optimization/82264
15016         * tree-ssa-sccvn.c (vn_phi_eq): Use safe_dyn_cast to check
15017         for GIMPLE_CONDs.
15018         (vn_phi_lookup): Likewise.
15019         (vn_phi_insert): Likewise.
15021 2017-09-20  Jakub Jelinek  <jakub@redhat.com>
15023         * dwarf2out.c (tree_add_const_value_attribute): For INTEGER_CST
15024         that fits into uhwi or shwi, add DW_AT_const_value regardless
15025         of early_dwarf without going through RTL, using add_AT_unsigned
15026         or add_AT_int.
15028         * dwarf2out.c (DEBUG_LTO_DWO_INFO_SECTION): Reorder defines.
15029         (DEBUG_LTO_ABBREV_SECTION): Likewise.
15030         (DEBUG_LTO_MACINFO_SECTION): Likewise.
15031         (DEBUG_MACRO_SECTION): Likewise.
15032         (DEBUG_LTO_MACRO_SECTION): Likewise.
15033         (DEBUG_STR_DWO_SECTION): Likewise.
15034         (DEBUG_LTO_STR_DWO_SECTION): Likewise.
15035         (DEBUG_LTO_LINE_SECTION): Drop .dwo suffix from the name.
15036         (DEBUG_LTO_DWO_LINE_SECTION): Define.
15037         (DEBUG_LTO_LINE_STR_SECTION): Define.
15038         (init_sections_and_labels): Initialize debug_line_str_section
15039         variable.  Initialize debug_loc_section for -gdwarf-5 to
15040         DEBUG_LOCLISTS_SECTION.  Formatting fixes.
15042 2017-09-20  Richard Biener  <rguenther@suse.de>
15044         * graphite-sese-to-poly.c (extract_affine): Properly handle
15045         POINTER_PLUS_EXPR, BIT_NOT_EXPR and conversion to signed.
15047 2017-09-20  Richard Biener  <rguenther@suse.de>
15049         PR tree-optimization/81373
15050         * graphite-scop-detection.c (build_cross_bb_scalars_def):
15051         Force SESE live-out defs to be handled even if they are
15052         scev_analyzable_p.
15054 2017-09-19  Jeff Law  <law@redhat.com>
15056         * combine-stack-adj.c (combine_stack_adjustments_for_block): Do
15057         nothing for stack adjustments with REG_STACK_CHECK.
15058         * sched-deps.c (parse_add_or_inc): Reject insns with
15059         REG_STACK_CHECK from dependency breaking.
15060         * config/i386/i386.c (pro_epilogue_adjust_stack): Return insn.
15061         (ix86_adjust_satck_and_probe_stack_clash): Add REG_STACK_NOTEs.
15062         * reg-notes.def (STACK_CHECK): New note.
15064         * config/i386/i386.c (ix86_adjust_stack_and_probe_stack_clash): New.
15065         (ix86_expand_prologue): Dump stack clash info as needed.
15066         Call ix86_adjust_stack_and_probe_stack_clash as needed.
15068         * function.c (dump_stack_clash_frame_info): New function.
15069         * function.h (dump_stack_clash_frame_info): Prototype.
15070         (enum stack_clash_probes): New enum.
15072         * config/alpha/alpha.c (alpha_expand_prologue): Also check
15073         flag_stack_clash_protection.
15074         * config/arm/arm.c (arm_compute_static_chain_stack_bytes): Likewise.
15075         (arm_expand_prologue, thumb1_expand_prologue): Likewise.
15076         (arm_frame_pointer_required): Likewise.
15077         * config/ia64/ia64.c (ia64_compute_frame_size): Likewise.
15078         (ia64_expand_prologue): Likewise.
15079         * config/mips/mips.c (mips_expand_prologue): Likewise.
15080         * config/powerpcspe/powerpcspe.c (rs6000_expand_prologue): Likewise.
15081         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15082         (sparc_flat_expand_prologue): Likewise.
15083         * config/spu/spu.c (spu_expand_prologue): Likewise.
15085         * explow.c: Include "params.h".
15086         (anti_adjust_stack_and_probe_stack_clash): New function.
15087         (get_stack_check_protect): Likewise.
15088         (compute_stack_clash_protection_loop_data): Likewise.
15089         (emit_stack_clash_protection_loop_start): Likewise.
15090         (emit_stack_clash_protection_loop_end): Likewise.
15091         (allocate_dynamic_stack_space): Use get_stack_check_protect.
15092         Use anti_adjust_stack_and_probe_stack_clash.
15093         * explow.h (compute_stack_clash_protection_loop_data): Prototype.
15094         (emit_stack_clash_protection_loop_start): Likewise.
15095         (emit_stack_clash_protection_loop_end): Likewise.
15096         * rtl.h (get_stack_check_protect): Prototype.
15097         * target.def (stack_clash_protection_final_dynamic_probe): New hook.
15098         * targhooks.c (default_stack_clash_protection_final_dynamic_probe): New.
15099         * targhooks.h (default_stack_clash_protection_final_dynamic_probe):
15100         Prototype.
15101         * doc/tm.texi.in (TARGET_STACK_CLASH_PROTECTION_FINAL_DYNAMIC_PROBE):
15102         Add @hook.
15103         * doc/tm.texi: Rebuilt.
15104         * config/aarch64/aarch64.c (aarch64_expand_prologue): Use
15105         get_stack_check_protect.
15106         * config/alpha/alpha.c (alpha_expand_prologue): Likewise.
15107         * config/arm/arm.c (arm_expand_prologue): Likewise.
15108         (arm_frame_pointer_required): Likewise.
15109         * config/i386/i386.c (ix86_expand_prologue): Likewise.
15110         * config/ia64/ia64.c (ia64_expand_prologue): Likewise.
15111         * config/mips/mips.c (mips_expand_prologue): Likewise.
15112         * config/powerpcspe/powerpcspe.c (rs6000_emit_prologue): Likewise.
15113         * config/rs6000/rs6000.c (rs6000_emit_prologue): Likewise.
15114         * config/sparc/sparc.c (sparc_expand_prologue): Likewise.
15115         (sparc_flat_expand_prologue): Likewise.
15117         * common.opt (-fstack-clash-protection): New option.
15118         * flag-types.h (enum stack_check_type): Note difference between
15119         -fstack-check= and -fstack-clash-protection.
15120         * params.def (PARAM_STACK_CLASH_PROTECTION_GUARD_SIZE): New PARAM.
15121         (PARAM_STACK_CLASH_PROTECTION_PROBE_INTERVAL): Likewise.
15122         * toplev.c (process_options): Issue warnings/errors for cases
15123         not handled with -fstack-clash-protection.
15124         * doc/invoke.texi (-fstack-clash-protection): Document new option.
15125         (-fstack-check): Note additional problem with -fstack-check=generic.
15126         Note that -fstack-check is primarily for Ada and refer users
15127         to -fstack-clash-protection for stack-clash-protection.
15128         Document new params for stack clash protection.
15130 2017-09-19  Uros Bizjak  <ubizjak@gmail.com>
15132         * config/i386/i386.c (ix86_split_long_move): Do not handle
15133         address used for LEA in a special way.
15135 2017-09-19  Segher Boessenkool  <segher@kernel.crashing.org>
15137         * simplify-rtx.c (simplify_binary_operation_1): Fix typo in comment.
15139 2017-09-19  Martin Sebor  <msebor@redhat.com>
15141         PR c/81854
15142         * cgraphunit.c (handle_alias_pairs): Reject aliases between functions
15143         of incompatible types.
15145 2017-09-19  Will Schmidt  <will_schmidt@vnet.ibm.com>
15147         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
15148         for early folding of vector loads (ALTIVEC_BUILTIN_LVX_*).
15149         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
15150         Remove obsoleted code for handling ALTIVEC_BUILTIN_VEC_LD.
15152 2017-09-19  Richard Biener  <rguenther@suse.de>
15154         PR tree-optimization/82244
15155         * tree-vrp.c (remove_range_assertions): Do not propagate
15156         a constant to abnormals but replace the assert with a copy.
15158 2017-09-19  Alexander Monakov  <amonakov@ispras.ru>
15160         PR rtl-optimization/57878
15161         PR rtl-optimization/68988
15162         * lra-assigns.c (reload_pseudo_compare_func): Remove fragmentation
15163         avoidance test involving non_reload_pseudos.  Move frequency test
15164         below the general fragmentation avoidance test.
15166 2017-09-19  Richard Biener  <rguenther@suse.de>
15168         PR tree-optimization/69728
15169         * graphite-sese-to-poly.c (schedule_error): New global.
15170         (add_loop_schedule): Handle empty domain by failing the
15171         schedule.
15172         (build_original_schedule): Handle schedule_error.
15174 2017-09-19  Richard Biener  <rguenther@suse.de>
15176         * graphite-scop-detection.c (scop_detection::can_represent_loop):
15177         Do not iterate to sibling loops but only to siblings of inner
15178         loops.
15180 2017-09-18  Andreas Schwab  <schwab@linux-m68k.org>
15182         PR target/81613
15183         * config/m68k/m68k.md (moveq feeding equality comparison): Check
15184         that the registers are different.
15186 2017-09-18  Uros Bizjak  <ubizjak@gmail.com>
15188         * config/i386/i386.c (fold_builtin_cpu): Add M_AMDFAM17H
15189         to processor_model and "amdfam17h" to arch_names_table.
15190         * doc/extend.texi (__builtin_cpu_is): Document amdfam17h CPU name.
15192 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15194         PR c/82234
15195         * doc/extend.texi: Add @findex entry for __builtin_shuffle.
15197 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15198             Alan Hayward  <alan.hayward@arm.com>
15199             David Sherwood  <david.sherwood@arm.com>
15201         * tree-vectorizer.h (vect_slp_analyze_operations): Replace parameters
15202         with a vec_info *.
15203         * tree-vect-loop.c (vect_analyze_loop_operations): Update call
15204         accordingly.
15205         * tree-vect-slp.c (vect_slp_analyze_node_operations): Add vec_info *
15206         parameter.  Set SLP_TREE_NUMBER_OF_VEC_STMTS here rather than in
15207         vect_schedule_slp_instance.
15208         (vect_slp_analyze_operations): Replace parameters with a vec_info *.
15209         Update call to vect_slp_analyze_node_operations.  Simplify return
15210         value.
15211         (vect_slp_analyze_bb_1): Update call accordingly.
15212         (vect_schedule_slp_instance): Remove vectorization_factor parameter.
15213         Don't calculate SLP_TREE_NUMBER_OF_VEC_STMTS here.
15214         (vect_schedule_slp): Update call accordingly.
15216 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15217             Alan Hayward  <alan.hayward@arm.com>
15218             David Sherwood  <david.sherwood@arm.com>
15220         * tree-vect-data-refs.c (vect_get_smallest_scalar_type): Cope
15221         with types that aren't in fact scalar.
15223 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15225         * tree-vect-slp.c (vect_record_max_nunits): New function,
15226         split out from...
15227         (vect_build_slp_tree_1): ...here.
15228         (vect_build_slp_tree_2): Call it for phis too.
15230 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15232         * tree-vect-stmts.c (vectorizable_mask_load_store): Pass mask_vectype
15233         to vect_get_vec_def_for_operand when getting the mask operand.
15235 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15236             Alan Hayward  <alan.hayward@arm.com>
15237             David Sherwood  <david.sherwood@arm.com>
15239         * tree-vect-loop.c (vectorizable_live_operation): Fix type of
15240         bitstart.
15242 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15243             Alan Hayward  <alan.hayward@arm.com>
15244             David Sherwood  <david.sherwood@arm.com>
15246         * tree-vect-loop.c (vectorizable_live_operation): Fix element size
15247         calculation for vector booleans.
15249 2017-09-18  Richard Sandiford  <richard.sandiford@linaro.org>
15250             Alan Hayward  <alan.hayward@arm.com>
15251             David Sherwood  <david.sherwood@arm.com>
15253         * tree-vect-stmts.c (can_vectorize_live_stmts): New function,
15254         split out from...
15255         (vect_transform_stmt): ...here.
15256         (vect_analyze_stmt): Use it instead of calling
15257         vectorizable_live_operation directly.
15259 2017-09-18  Cesar Philippidis  <cesar@codesourcery.com>
15261         * omp-offload.c (oacc_xform_loop): Enable SIMD vectorization on
15262         non-SIMT targets in acc vector loops.
15264 2017-09-18  Claudiu Zissulescu  <claziss@synopsys.com>
15266         * configure.ac: Add arc and check if assembler supports gdwarf2.
15267         * configure: Regenerate.
15269 2017-09-18  Richard Biener  <rguenther@suse.de>
15271         PR tree-optimization/82220
15272         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Exclude
15273         epilogue niters from the min_profitable_iters compute.
15275 2017-09-18  Jakub Jelinek  <jakub@redhat.com>
15277         PR target/82145
15278         * config/i386/i386.c (ix86_init_large_pic_reg): Revert 2017-09-01
15279         changes.  Turn CODE_LABEL into NOTE_INSN_DELETED_LABEL immediately.
15280         (ix86_init_pic_reg): Revert 2017-09-01 changes.
15282 2017-09-18  Eric Botcazou  <ebotcazou@adacore.com>
15284         PR target/81361
15285         * dwarf2cfi.c (add_cfis_to_fde): Do not generate DW_CFA_set_loc after
15286         switching to a new text section.
15288 2017-09-18  Richard Biener  <rguenther@suse.de>
15290         * graphite-scop-detection.c (scop_detection::stmt_has_simple_data_ref):
15291         Simplify.
15292         (build_alias_set): Reject aliases with no access function.
15294 2017-09-18  Richard Biener  <rguenther@suse.de>
15296         PR tree-optimization/79622
15297         * graphite-scop-detection.c (build_cross_bb_scalars_def): Properly
15298         handle PHIs.
15299         (build_cross_bb_scalars_use): Likewise.
15301 2017-09-18  Pierre-Marie de Rodat  <derodat@adacore.com>
15303         * cgraph.h (cgraph_thunk_info): Fix a typo in a comment.
15305 2017-09-18  Alan Modra  <amodra@gmail.com>
15307         PR target/81996
15308         * gcc/config/rs6000/rs6000.c (rs6000_return_addr): Use
15309         stack_pointer_rtx for count 0.  Update comments.  Break up
15310         large rtl expression.
15312 2017-09-17  Daniel Santos  <daniel.santos@pobox.com>
15314         * config/i386/i386.c (xlogue_layout::STUB_NAME_MAX_LEN):
15315         Increase to 20 bytes.
15316         (xlogue_layout::s_stub_names): Add an additional size-2 diminsion.
15317         (xlogue_layout::get_stub_name): Modify to select the appropairate sse
15318         or avx version of the stub.
15320 2017-09-17  H.J. Lu  <hongjiu.lu@intel.com>
15322         PR target/82166
15323         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Properly
15324         compute the minimum stack alignment.  Also update preferred stack
15325         boundary for leaf functions.
15327 2017-09-16  Richard Sandiford  <richard.sandiford@linaro.org>
15329         PR tree-optimization/82228
15330         * tree-vect-loop.c (vectorizable_live_operation): Move initialization
15331         of ncopies.
15333 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15335         * common/config/nds32/nds32-common.c
15336         (nds32_option_optimization_table): Refine formatting.
15337         (nds32_option_optimization_table): Use -fsched-pressure and
15338         -fomit-frame-pointer for specific optimization level.
15340 2017-09-16  Chung-Ju Wu  <jasonwucj@gmail.com>
15342         * config/nds32/nds32.c: Refine formatting and comments.
15343         * config/nds32/nds32.h: Likewise.
15344         * config/nds32/nds32.md: Likewise.
15345         * config/nds32/nds32-cost.c: Likewise.
15346         * config/nds32/nds32-isr.c: Likewise.
15347         * config/nds32/nds32-md-auxiliary.c: Likewise.
15348         * config/nds32/nds32-multiple.md: Likewise.
15349         * config/nds32/nds32-predicates.c: Likewise.
15351 2017-09-15  Andrew Sutton  <andrew.n.sutton@gmail.com>
15352             Jakub Jelinek  <jakub@redhat.com>
15354         Add support for -std=c++2a.
15355         * doc/cpp.texi (__cplusplus): Document value for -std=c++2a
15356         or -std=gnu+2a.
15357         * doc/invoke.texi: Document -std=c++2a and -std=gnu++2a.
15359 2017-09-15  Steve Ellcey  <sellcey@cavium.com>
15361         PR target/82066
15362         * doc/extend.texi (Common Function Attributes): Add 
15363         references to ARM, AArch64, and S/390 specific attributes.
15364         (Function Specific Option Pragmas): Add AArch64 and S/390
15365         to list of back ends that support the target pragma.
15367 2017-09-15  Nathan Sidwell  <nathan@acm.org>
15369         * doc/standards.texi: Fix C++17 description.  Update URLs for
15370         C++11 & 14.
15372 2017-09-15  Bernd Edlinger  <bernd.edlinger@hotmail.de>
15374         * common.opt (Wcast-align=strict): New warning option.
15375         * doc/invoke.texi: Document -Wcast-align=strict. 
15377 2017-09-15  Pierre-Marie de Rodat  <derodat@adacore.com>
15379         * cgraph.h (cgraph_thunk_info): Add comments.
15380         * cgraph.c (cgraph_node::create_thunk): Adjust comment, make
15381         assert for VIRTUAL_* arguments stricter.
15383 2017-09-15  Jackson Woodruff  <jackson.woodruff@arm.com>
15385         PR tree-optimization/71026
15386         * match.pd: Move RDIV patterns from fold-const.c
15387         * fold-const.c (distribute_real_division): Removed.
15388         (fold_binary_loc): Remove calls to distribute_real_divison.
15390 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15392         * doc/invoke.texi: Document -std=c++17 and -std=gnu++17 and document
15393         c++1z and gnu++1z as deprecated.  Change other references to
15394         -std=c++1z to -std=c++17 and -std=gnu++1z to -std=gnu++17.
15395         Change -Wc++1z-compat to -Wc++17-compat.
15396         * doc/cpp.texi: Document -std=c++17 defines __cplusplus 201703L.
15397         * dwarf2out.c (highest_c_language): Handle C++17.
15398         (gen_compile_unit_die): Likewise.
15400 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15402         PR rtl-optimization/82192
15403         * combine.c (make_extraction): Don't look through non-paradoxical
15404         SUBREGs or TRUNCATE if pos + len is or might be bigger than
15405         inner's mode.
15407 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15408             Alan Hayward  <alan.hayward@arm.com>
15409             David Sherwood  <david.sherwood@arm.com>
15411         * target.def (function_arg_offset): New hook.
15412         * targhooks.h (default_function_arg_offset): Declare.
15413         * targhooks.c (default_function_arg_offset): New function.
15414         * function.c (locate_and_pad_parm): Use
15415         targetm.calls.function_arg_offset instead of FUNCTION_ARG_OFFSET.
15416         * doc/tm.texi.in (FUNCTION_ARG_OFFSET): Replace with...
15417         (TARGET_FUNCTION_ARG_OFFSET): ...this.
15418         * doc/tm.texi: Regenerate.
15419         * config/spu/spu.h (FUNCTION_ARG_OFFSET): Delete.
15420         * config/spu/spu.c (spu_function_arg_offset): New function.
15421         (TARGET_FUNCTION_ARG_OFFSET): Redefine.
15422         * system.h (FUNCTION_ARG_OFFSET): Poison.
15424 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15425             Alan Hayard  <alan.hayward@arm.com>
15426             David Sherwood  <david.sherwood@arm.com>
15428         * target.def (truly_noop_truncation): New hook.
15429         (mode_rep_extended): Refer to TARGET_TRULY_NOOP_TRUNCATION rather
15430         than TRULY_NOOP_TRUNCATION.
15431         * hooks.h (hook_bool_uint_uint_true): Declare.
15432         * hooks.c (hook_bool_uint_uint_true): New function.
15433         * doc/tm.texi.in (TRULY_NOOP_TRUNCATION): Replace with...
15434         (TARGET_TRULY_NOOP_TRUNCATION): ...this.
15435         * doc/tm.texi: Regenerate.
15436         * combine.c (make_extraction): Refer to TARGET_TRULY_NOOP_TRUNCATION
15437         rather than TRULY_NOOP_TRUNCATION in comments.
15438         (simplify_comparison): Likewise.
15439         (record_truncated_value): Likewise.
15440         * expmed.c (extract_bit_field_1): Likewise.
15441         (extract_split_bit_field): Likewise.
15442         * convert.c (convert_to_integer_1): Use targetm.truly_noop_truncation
15443         instead of TRULY_NOOP_TRUNCATION.
15444         * function.c (assign_parm_setup_block): Likewise.
15445         * machmode.h (TRULY_NOOP_TRUNCATION_MODES_P): Likewise.
15446         * rtlhooks.c: Include target.h.
15447         * config/aarch64/aarch64.h (TRULY_NOOP_TRUNCATION): Delete.
15448         * config/alpha/alpha.h (TRULY_NOOP_TRUNCATION): Delete.
15449         * config/arc/arc.h (TRULY_NOOP_TRUNCATION): Delete.
15450         * config/arm/arm.h (TRULY_NOOP_TRUNCATION): Delete.
15451         * config/avr/avr.h (TRULY_NOOP_TRUNCATION): Delete.
15452         * config/bfin/bfin.h (TRULY_NOOP_TRUNCATION): Delete.
15453         * config/c6x/c6x.h (TRULY_NOOP_TRUNCATION): Delete.
15454         * config/cr16/cr16.h (TRULY_NOOP_TRUNCATION): Delete.
15455         * config/cris/cris.h (TRULY_NOOP_TRUNCATION): Delete.
15456         * config/epiphany/epiphany.h (TRULY_NOOP_TRUNCATION): Delete.
15457         * config/fr30/fr30.h (TRULY_NOOP_TRUNCATION): Delete.
15458         * config/frv/frv.h (TRULY_NOOP_TRUNCATION): Delete.
15459         * config/ft32/ft32.h (TRULY_NOOP_TRUNCATION): Delete.
15460         * config/h8300/h8300.h (TRULY_NOOP_TRUNCATION): Delete.
15461         * config/i386/i386.h (TRULY_NOOP_TRUNCATION): Delete.
15462         * config/ia64/ia64.h (TRULY_NOOP_TRUNCATION): Delete.
15463         * config/iq2000/iq2000.h (TRULY_NOOP_TRUNCATION): Delete.
15464         * config/lm32/lm32.h (TRULY_NOOP_TRUNCATION): Delete.
15465         * config/m32c/m32c.h (TRULY_NOOP_TRUNCATION): Delete.
15466         * config/m32r/m32r.h (TRULY_NOOP_TRUNCATION): Delete.
15467         * config/m68k/m68k.h (TRULY_NOOP_TRUNCATION): Delete.
15468         * config/mcore/mcore.h (TRULY_NOOP_TRUNCATION): Delete.
15469         * config/microblaze/microblaze.h (TRULY_NOOP_TRUNCATION): Delete.
15470         * config/mips/mips.h (TRULY_NOOP_TRUNCATION): Delete.
15471         * config/mips/mips.c (mips_truly_noop_truncation): New function.
15472         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15473         * config/mips/mips.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15474         rather than TRULY_NOOP_TRUNCATION in comments.
15475         * config/mmix/mmix.h (TRULY_NOOP_TRUNCATION): Delete.
15476         * config/mn10300/mn10300.h (TRULY_NOOP_TRUNCATION): Delete.
15477         * config/moxie/moxie.h (TRULY_NOOP_TRUNCATION): Delete.
15478         * config/msp430/msp430.h (TRULY_NOOP_TRUNCATION): Delete.
15479         * config/nds32/nds32.h (TRULY_NOOP_TRUNCATION): Delete.
15480         * config/nios2/nios2.h (TRULY_NOOP_TRUNCATION): Delete.
15481         * config/nvptx/nvptx.h (TRULY_NOOP_TRUNCATION): Delete.
15482         * config/pa/pa.h (TRULY_NOOP_TRUNCATION): Delete.
15483         * config/pdp11/pdp11.h (TRULY_NOOP_TRUNCATION): Delete.
15484         * config/powerpcspe/powerpcspe.h (TRULY_NOOP_TRUNCATION): Delete.
15485         * config/riscv/riscv.h (TRULY_NOOP_TRUNCATION): Delete.
15486         * config/riscv/riscv.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15487         rather than TRULY_NOOP_TRUNCATION in comments.
15488         * config/rl78/rl78.h (TRULY_NOOP_TRUNCATION): Delete.
15489         * config/rs6000/rs6000.h (TRULY_NOOP_TRUNCATION): Delete.
15490         * config/rx/rx.h (TRULY_NOOP_TRUNCATION): Delete.
15491         * config/s390/s390.h (TRULY_NOOP_TRUNCATION): Delete.
15492         * config/sh/sh.h (MAYBE_BASE_REGISTER_RTX_P): Remove
15493         TRULY_NOOP_TRUNCATION condition.
15494         (MAYBE_INDEX_REGISTER_RTX_P): Likewise.
15495         (TRULY_NOOP_TRUNCATION): Delete.
15496         * config/sparc/sparc.h (TRULY_NOOP_TRUNCATION): Delete.
15497         * config/spu/spu.h (TRULY_NOOP_TRUNCATION): Delete.
15498         * config/spu/spu.c (spu_truly_noop_truncation): New function.
15499         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15500         * config/stormy16/stormy16.h (TRULY_NOOP_TRUNCATION): Delete.
15501         * config/tilegx/tilegx.h (TRULY_NOOP_TRUNCATION): Delete.
15502         * config/tilegx/tilegx.c (tilegx_truly_noop_truncation): New fuction.
15503         (TARGET_TRULY_NOOP_TRUNCATION): Redefine.
15504         * config/tilegx/tilegx.md: Refer to TARGET_TRULY_NOOP_TRUNCATION
15505         rather than TRULY_NOOP_TRUNCATION in comments.
15506         * config/tilepro/tilepro.h (TRULY_NOOP_TRUNCATION): Delete.
15507         * config/v850/v850.h (TRULY_NOOP_TRUNCATION): Delete.
15508         * config/vax/vax.h (TRULY_NOOP_TRUNCATION): Delete.
15509         * config/visium/visium.h (TRULY_NOOP_TRUNCATION): Delete.
15510         * config/xtensa/xtensa.h (TRULY_NOOP_TRUNCATION): Delete.
15511         * system.h (TRULY_NOOP_TRUNCATION): Poison.
15513 2017-09-15  Christophe Lyon  <christophe.lyon@linaro.org>
15515         PR target/67591
15516         * config/arm/arm.md (*cmp_and): Add enabled_for_depr_it attribute.
15517         (*cmp_ior): Likewise.
15518         (*ior_scc_scc): Add alternative for enabled_for_depr_it attribute.
15519         (*ior_scc_scc_cmp): Likewise.
15520         (*and_scc_scc): Likewise.
15521         (*and_scc_scc_cmp): Likewise.
15523 2017-09-15  Richard Sandiford  <richard.sandiford@linaro.org>
15524             Alan Hayard  <alan.hayward@arm.com>
15525             David Sherwood  <david.sherwood@arm.com>
15527         * target.def (can_change_mode_class): New hook.
15528         (mode_rep_extended): Refer to it instead of CANNOT_CHANGE_MODE_CLASS.
15529         (hard_regno_nregs): Likewise.
15530         * hooks.h (hook_bool_mode_mode_reg_class_t_true): Declare.
15531         * hooks.c (hook_bool_mode_mode_reg_class_t_true): New function.
15532         * doc/tm.texi.in (CANNOT_CHANGE_MODE_CLASS): Replace with...
15533         (TARGET_CAN_CHANGE_MODE_CLASS): ...this.
15534         (LOAD_EXTEND_OP): Update accordingly.
15535         * doc/tm.texi: Regenerate.
15536         * doc/rtl.texi: Refer to TARGET_CAN_CHANGE_MODE_CLASS instead of
15537         CANNOT_CHANGE_MODE_CLASS.
15538         * hard-reg-set.h (REG_CANNOT_CHANGE_MODE_P): Replace with...
15539         (REG_CAN_CHANGE_MODE_P): ...this new macro.
15540         * combine.c (simplify_set): Update accordingly.
15541         * emit-rtl.c (validate_subreg): Likewise.
15542         * recog.c (general_operand): Likewise.
15543         * regcprop.c (mode_change_ok): Likewise.
15544         * reload1.c (choose_reload_regs): Likewise.
15545         (inherit_piecemeal_p): Likewise.
15546         * rtlanal.c (simplify_subreg_regno): Likewise.
15547         * postreload.c (reload_cse_simplify_set): Use REG_CAN_CHANGE_MODE_P
15548         instead of CANNOT_CHANGE_MODE_CLASS.
15549         (reload_cse_simplify_operands): Likewise.
15550         * reload.c (push_reload): Use targetm.can_change_mode_class
15551         instead of CANNOT_CHANGE_MODE_CLASS.
15552         (push_reload): Likewise.  Also use REG_CAN_CHANGE_MODE_P instead of
15553         REG_CANNOT_CHANGE_MODE_P.
15554         * config/alpha/alpha.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15555         * config/alpha/alpha.c (alpha_can_change_mode_class): New function.
15556         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15557         * config/arm/arm.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15558         * config/arm/arm.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15559         (arm_can_change_mode_class): New function.
15560         * config/arm/neon.md: Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15561         than CANNOT_CHANGE_MODE_CLASS in comments.
15562         * config/i386/i386.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15563         * config/i386/i386-protos.h (ix86_cannot_change_mode_class): Delete.
15564         * config/i386/i386.c (ix86_cannot_change_mode_class): Replace with...
15565         (ix86_can_change_mode_class): ...this new function, inverting the
15566         sense of the return value.
15567         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15568         * config/ia64/ia64.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15569         * config/ia64/ia64.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15570         (ia64_can_change_mode_class): New function.
15571         * config/m32c/m32c.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15572         * config/m32c/m32c-protos.h (m32c_cannot_change_mode_class): Delete.
15573         * config/m32c/m32c.c (m32c_cannot_change_mode_class): Replace with...
15574         (m32c_can_change_mode_class): ...this new function, inverting the
15575         sense of the return value.
15576         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15577         * config/mips/mips.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15578         * config/mips/mips-protos.h (mips_cannot_change_mode_class): Delete.
15579         * config/mips/mips.c (mips_cannot_change_mode_class): Replace with...
15580         (mips_can_change_mode_class): ...this new function, inverting the
15581         sense of the return value.
15582         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15583         * config/msp430/msp430.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15584         * config/msp430/msp430.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15585         (msp430_can_change_mode_class): New function.
15586         * config/nvptx/nvptx.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15587         * config/nvptx/nvptx.c (nvptx_can_change_mode_class): New function.
15588         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15589         * config/pa/pa32-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15590         * config/pa/pa64-regs.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15591         * config/pa/pa-protos.h (pa_cannot_change_mode_class): Delete.
15592         * config/pa/pa.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15593         (pa_cannot_change_mode_class): Replace with...
15594         (pa_can_change_mode_class): ...this new function, inverting the
15595         sense of the return value.
15596         (pa_modes_tieable_p): Refer to TARGET_CAN_CHANGE_MODE_CLASS rather
15597         than CANNOT_CHANGE_MODE_CLASS in comments.
15598         * config/pdp11/pdp11.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15599         * config/pdp11/pdp11-protos.h (pdp11_cannot_change_mode_class): Delete.
15600         * config/pdp11/pdp11.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15601         (pdp11_cannot_change_mode_class): Replace with...
15602         (pdp11_can_change_mode_class): ...this new function, inverting the
15603         sense of the return value.
15604         * config/powerpcspe/powerpcspe.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15605         * config/powerpcspe/powerpcspe-protos.h
15606         (rs6000_cannot_change_mode_class_ptr): Delete.
15607         * config/powerpcspe/powerpcspe.c
15608         (rs6000_cannot_change_mode_class_ptr): Delete.
15609         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15610         (rs6000_option_override_internal): Assign to
15611         targetm.can_change_mode_class instead of
15612         rs6000_cannot_change_mode_class_ptr.
15613         (rs6000_cannot_change_mode_class): Replace with...
15614         (rs6000_can_change_mode_class): ...this new function, inverting the
15615         sense of the return value.
15616         (rs6000_debug_cannot_change_mode_class): Replace with...
15617         (rs6000_debug_can_change_mode_class): ...this new function.
15618         * config/riscv/riscv.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15619         * config/riscv/riscv.c (riscv_can_change_mode_class): New function.
15620         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15621         * config/rs6000/rs6000.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15622         * config/rs6000/rs6000-protos.h (rs6000_cannot_change_mode_class_ptr):
15623         Delete.
15624         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class_ptr): Delete.
15625         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15626         (rs6000_option_override_internal): Assign to
15627         targetm.can_change_mode_class instead of
15628         rs6000_cannot_change_mode_class_ptr.
15629         (rs6000_cannot_change_mode_class): Replace with...
15630         (rs6000_can_change_mode_class): ...this new function, inverting the
15631         sense of the return value.
15632         (rs6000_debug_cannot_change_mode_class): Replace with...
15633         (rs6000_debug_can_change_mode_class): ...this new function.
15634         * config/s390/s390.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15635         * config/s390/s390-protos.h (s390_cannot_change_mode_class): Delete.
15636         * config/s390/s390.c (s390_cannot_change_mode_class): Replace with...
15637         (s390_can_change_mode_class): ...this new function, inverting the
15638         sense of the return value.
15639         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15640         * config/sh/sh.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15641         * config/sh/sh-protos.h (sh_cannot_change_mode_class): Delete.
15642         * config/sh/sh.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15643         (sh_cannot_change_mode_class): Replace with...
15644         (sh_can_change_mode_class): ...this new function, inverting the
15645         sense of the return value.
15646         * config/sparc/sparc.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15647         * config/sparc/sparc.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15648         (sparc_can_change_mode_class): New function.
15649         * config/spu/spu.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15650         * config/spu/spu.c (spu_can_change_mode_class): New function.
15651         (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15652         * config/visium/visium.h (CANNOT_CHANGE_MODE_CLASS): Delete.
15653         * config/visium/visium.c (TARGET_CAN_CHANGE_MODE_CLASS): Redefine.
15654         (visium_can_change_mode_class): New function.
15655         * system.h (CANNOT_CHANGE_MODE_CLASS): Poison.
15657 2017-09-15  Richard Biener  <rguenther@suse.de>
15659         PR tree-optimization/82217
15660         * tree-ssa-sccvn.c (visit_phi): Properly handle all VN_TOP
15661         but not undefined case.
15663 2017-09-15  Jakub Jelinek  <jakub@redhat.com>
15665         PR target/82145
15666         * postreload.c (reload_cse_simplify_operands): Skip
15667         NOTE_INSN_DELETED_LABEL similarly to skipping CODE_LABEL.
15669 2017-09-15  Richard Biener  <rguenther@suse.de>
15671         PR tree-optimization/68823
15672         * graphite-scop-detection.c (build_alias_set): If we have a
15673         possible dependence check whether we can handle them by just
15674         looking at the DRs DR_ACCESS_FNs.
15675         (build_scops): If build_alias_set fails, fail the SCOP.
15677 2017-09-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
15679         * config/rs6000/rs6000-builtin.def (BU_FLOAT128_1_HW): New macros
15680         to support float128 built-in functions that require the ISA 3.0
15681         hardware.
15682         (BU_FLOAT128_3_HW): Likewise.
15683         (SQRTF128): Add support for the IEEE 128-bit square root and fma
15684         built-in functions.
15685         (FMAF128): Likewise.
15686         (FMAQ): Likewise.
15687         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
15688         support for built-in functions that need the ISA 3.0 IEEE 128-bit
15689         floating point instructions.
15690         (rs6000_invalid_builtin): Likewise.
15691         (rs6000_builtin_mask_names): Likewise.
15692         * config/rs6000/rs6000.h (MASK_FLOAT128_HW): Likewise.
15693         (RS6000_BTM_FLOAT128_HW): Likewise.
15694         (RS6000_BTM_COMMON): Likewise.
15695         * config/rs6000/rs6000.md (fma<mode>4_hw): Add a generator
15696         function.
15697         * doc/extend.texi (RS/6000 built-in functions): Document the
15698         IEEE 128-bit floating point square root and fused multiply-add
15699         built-in functions.
15701 2017-09-14  Pat Haugen  <pthaugen@us.ibm.com>
15703         * config/rs6000/rs6000.c (rs6000_set_up_by_prologue): Make sure the TOC
15704         reg (r2) isn't in the set of registers defined in the prologue.
15706 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15707             Alan Hayward  <alan.hayward@arm.com>
15708             David Sherwood  <david.sherwood@arm.com>
15710         * tree-vectorizer.h (_loop_vec_info): Add max_vectorization_factor.
15711         (LOOP_VINFO_MAX_VECT_FACTOR): New macro.
15712         (LOOP_VINFO_ORIG_VECT_FACTOR): Replace with...
15713         (LOOP_VINFO_ORIG_MAX_VECT_FACTOR): ...this new macro.
15714         * tree-vect-data-refs.c (vect_analyze_data_ref_dependences): Update
15715         accordingly.
15716         * tree-vect-loop.c (_loop_vec_info::_loop_vec_info): Initialize
15717         max_vectorization_factor.
15718         (vect_analyze_loop_2): Set LOOP_VINFO_MAX_VECT_FACTOR.
15720 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15721             Alan Hayward  <alan.hayward@arm.com>
15722             David Sherwood  <david.sherwood@arm.com>
15724         * tree-vectorizer.h (vect_min_worthwhile_factor): Delete.
15725         (vect_worthwhile_without_simd_p): Declare.
15726         * tree-vect-loop.c (vect_worthwhile_without_simd_p): New function.
15727         (vectorizable_reduction): Use it.
15728         * tree-vect-stmts.c (vectorizable_shift): Likewise.
15729         (vectorizable_operation): Likewise.
15731 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15732             Alan Hayward  <alan.hayward@arm.com>
15733             David Sherwood  <david.sherwood@arm.com>
15735         * tree-vectorizer.h (vect_get_num_copies): New function.
15736         * tree-vect-data-refs.c (vect_get_data_access_cost): Use it.
15737         * tree-vect-loop.c (vectorizable_reduction): Likewise.
15738         (vectorizable_induction): Likewise.
15739         (vectorizable_live_operation): Likewise.
15740         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
15741         (vectorizable_bswap): Likewise.
15742         (vectorizable_call): Likewise.
15743         (vectorizable_conversion): Likewise.
15744         (vectorizable_assignment): Likewise.
15745         (vectorizable_shift): Likewise.
15746         (vectorizable_operation): Likewise.
15747         (vectorizable_store): Likewise.
15748         (vectorizable_load): Likewise.
15749         (vectorizable_condition): Likewise.
15750         (vectorizable_comparison): Likewise.
15751         (vect_analyze_stmt): Pass the slp node to vectorizable_live_operation.
15753 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15754             Alan Hayward  <alan.hayward@arm.com>
15755             David Sherwood  <david.sherwood@arm.com>
15757         * tree-vect-loop.c (vectorizable_induction): Use gimple_build instead
15758         of vect_init_vector.
15760 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15761             Alan Hayward  <alan.hayward@arm.com>
15762             David Sherwood  <david.sherwood@arm.com>
15764         * gimple-fold.h (gimple_build_vector_from_val): Declare, and provide
15765         an inline wrapper that provides a location.
15766         (gimple_build_vector): Likewise.
15767         * gimple-fold.c (gimple_build_vector_from_val): New function.
15768         (gimple_build_vector): Likewise.
15769         * tree-vect-loop.c (get_initial_def_for_reduction): Use the new
15770         functions to build the initial value.  Always return a gimple value.
15771         (get_initial_defs_for_reduction): Likewise.  Only compute
15772         neutral_vec once.
15773         (vect_create_epilog_for_reduction): Don't call force_gimple_operand or
15774         vect_init_vector on the results from get_initial_def(s)_for_reduction.
15775         (vectorizable_induction): Use gimple_build_vector rather than
15776         vect_init_vector.
15778 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15779             Alan Hayward  <alan.hayward@arm.com>
15780             David Sherwood  <david.sherwood@arm.com>
15782         * target.h (vec_perm_indices): New typedef.
15783         (auto_vec_perm_indices): Likewise.
15784         * optabs-query.h: Include target.h
15785         (can_vec_perm_p): Take a vec_perm_indices *.
15786         * optabs-query.c (can_vec_perm_p): Likewise.
15787         (can_mult_highpart_p): Update accordingly.  Use auto_vec_perm_indices.
15788         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15789         * tree-vect-generic.c (lower_vec_perm): Likewise.
15790         * tree-vect-data-refs.c (vect_grouped_store_supported): Likewise.
15791         (vect_grouped_load_supported): Likewise.
15792         (vect_shift_permute_load_chain): Likewise.
15793         (vect_permute_store_chain): Use auto_vec_perm_indices.
15794         (vect_permute_load_chain): Likewise.
15795         * fold-const.c (fold_vec_perm): Take vec_perm_indices.
15796         (fold_ternary_loc): Update accordingly.  Use auto_vec_perm_indices.
15797         Update uses of can_vec_perm_p.
15798         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Replace the
15799         mode with a number of elements.  Take a vec_perm_indices *.
15800         (vect_create_epilog_for_reduction): Update accordingly.
15801         Use auto_vec_perm_indices.
15802         (have_whole_vector_shift): Likewise.  Update call to can_vec_perm_p.
15803         * tree-vect-slp.c (vect_build_slp_tree_1): Likewise.
15804         (vect_transform_slp_perm_load): Likewise.
15805         (vect_schedule_slp_instance): Use auto_vec_perm_indices.
15806         * tree-vectorizer.h (vect_gen_perm_mask_any): Take a vec_perm_indices.
15807         (vect_gen_perm_mask_checked): Likewise.
15808         * tree-vect-stmts.c (vect_gen_perm_mask_any): Take a vec_perm_indices.
15809         (vect_gen_perm_mask_checked): Likewise.
15810         (vectorizable_mask_load_store): Use auto_vec_perm_indices.
15811         (vectorizable_store): Likewise.
15812         (vectorizable_load): Likewise.
15813         (perm_mask_for_reverse): Likewise.  Update call to can_vec_perm_p.
15814         (vectorizable_bswap): Likewise.
15816 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15817             Alan Hayward  <alan.hayward@arm.com>
15818             David Sherwood  <david.sherwood@arm.com>
15820         * tree.h (build_vector): Take a vec<tree> instead of a tree *.
15821         * tree.c (build_vector): Likewise.
15822         (build_vector_from_ctor): Update accordingly.
15823         (build_vector_from_val): Likewise.
15824         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
15825         * tree-ssa-forwprop.c (simplify_vector_constructor): Likewise.
15826         * tree-vect-generic.c (add_rshift): Likewise.
15827         (expand_vector_divmod): Likewise.
15828         (optimize_vector_constructor): Likewise.
15829         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
15830         (vect_transform_slp_perm_load): Likewise.
15831         (vect_schedule_slp_instance): Likewise.
15832         * tree-vect-stmts.c (vectorizable_bswap): Likewise.
15833         (vectorizable_call): Likewise.
15834         (vect_gen_perm_mask_any): Likewise.  Add elements in order.
15835         * expmed.c (make_tree): Likewise.
15836         * fold-const.c (fold_negate_expr_1): Use auto_vec<tree> when building
15837         a vector passed to build_vector.
15838         (fold_convert_const): Likewise.
15839         (exact_inverse): Likewise.
15840         (fold_ternary_loc): Likewise.
15841         (fold_relational_const): Likewise.
15842         (const_binop): Likewise.  Use VECTOR_CST_ELT directly when operating
15843         on VECTOR_CSTs, rather than going through vec_cst_ctor_to_array.
15844         (const_unop): Likewise.  Store the reduction accumulator in a
15845         variable rather than an array.
15846         (vec_cst_ctor_to_array): Take the number of elements as a parameter.
15847         (fold_vec_perm): Update calls accordingly.  Use auto_vec<tree> for
15848         the new vector, rather than constructing it after the input arrays.
15849         (native_interpret_vector): Use auto_vec<tree> when building
15850         a vector passed to build_vector.  Add elements in order.
15851         * tree-vect-loop.c (get_initial_defs_for_reduction): Use
15852         auto_vec<tree> when building a vector passed to build_vector.
15853         (vect_create_epilog_for_reduction): Likewise.
15854         (vectorizable_induction): Likewise.
15855         (get_initial_def_for_reduction): Likewise.  Fix indentation of
15856         case statements.
15857         * config/sparc/sparc.c (sparc_handle_vis_mul8x16): Change n_elts
15858         to a vec<tree> *.
15859         (sparc_fold_builtin): Use auto_vec<tree> when building a vector
15860         passed to build_vector.
15862 2017-09-14  Richard Sandiford  <richard.sandiford@linaro.org>
15863             Alan Hayward  <alan.hayward@arm.com>
15864             David Sherwood  <david.sherwood@arm.com>
15866         * tree-core.h (tree_base::u): Add an "nelts" field.
15867         (tree_vector): Use VECTOR_CST_NELTS as the length.
15868         * tree.c (tree_size): Likewise.
15869         (make_vector): Initialize VECTOR_CST_NELTS.
15870         * tree.h (VECTOR_CST_NELTS): Use the u.nelts field.
15871         * cfgexpand.c (expand_debug_expr): Use VECTOR_CST_NELTS instead of
15872         TYPE_VECTOR_SUBPARTS.
15873         * expr.c (const_vector_mask_from_tree): Consistently use "units"
15874         as the number of units, setting it from VECTOR_CST_NELTS.
15875         (const_vector_from_tree): Likewise.
15876         * fold-const.c (negate_expr_p): Use VECTOR_CST_NELTS instead of
15877         TYPE_VECTOR_SUBPARTS for the number of elements in a VECTOR_CST.
15878         (fold_negate_expr_1): Likewise.
15879         (fold_convert_const): Likewise.
15880         (const_binop): Likewise.  Differentiate the number of output and
15881         input elements.
15882         (const_unop): Likewise.
15883         (fold_ternary_loc): Use VECTOR_CST_NELTS for the number of elements
15884         in a VECTOR_CST, asserting that it is the same as TYPE_VECTOR_SUBPARTS
15885         in cases that did the opposite.
15887 2017-09-14  Richard Biener  <rguenther@suse.de>
15889         * tree-ssa-sccvn.c (visit_phi): Merge undefined values similar
15890         to VN_TOP.
15892 2017-09-14  Eric Botcazou  <ebotcazou@adacore.com>
15894         * dwarf2out.c (dwarf2out_source_line): Remove superfluous test.
15896 2017-09-14  Jakub Jelinek  <jakub@redhat.com>
15898         PR target/81325
15899         * cfgbuild.c (find_bb_boundaries): Ignore debug insns in decisions
15900         if and where to split a bb, except for splitting before debug insn
15901         sequences followed by non-label real insn.  Delete debug insns
15902         in between basic blocks.
15904         * combine.c (make_compound_operation_int): Formatting fixes.
15906         * config/alpha/elf.h (LINK_EH_SPEC): Add -static-pie support.
15907         * config/alpha/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15908         * config/netbsd.h (LINK_EH_SPEC): Likewise.
15909         * config/sol2.h (LINK_EH_SPEC): Likewise.
15910         * config/arm/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15911         * config/s390/linux.h (LINK_SPEC): Likewise.
15912         * config/freebsd.h (LINK_EH_SPEC): Likewise.
15913         * config/openbsd.h (LINK_EH_SPEC): Likewise.
15914         * config/lm32/uclinux-elf.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15915         * config/aarch64/aarch64-linux.h (LINUX_TARGET_LINK_SPEC): Likewise.
15916         * config/powerpcspe/sysv4.h (LINK_EH_SPEC): Likewise.
15917         * config/bfin/linux.h (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
15918         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Fix a typo.
15919         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Formatting fix.
15921 2017-09-13  Jakub Jelinek  <jakub@redhat.com>
15923         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Add -static-pie
15924         support.
15925         (ENDFILE_LINUX_SPEC): Likewise.
15926         (LINK_EH_SPEC): Likewise.
15927         * config/rs6000/linux64.h (LINK_SHLIB_SPEC): Likewise.
15928         (LINK_OS_LINUX_SPEC32): Likewise.
15929         (LINK_OS_LINUX_SPEC64): Likewise.
15930         * config/rs6000/linux.h (LINK_SHLIB_SPEC): Likewise.
15931         (LINK_OS_LINUX_SPEC): Likewise.
15933 2017-09-13  Martin Liska  <mliska@suse.cz>
15935         PR middle-end/82154
15936         * stmt.c (expand_sjlj_dispatch_table): Use CASE_LOW when
15937         CASE_HIGH is NULL_TREE.
15939 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
15940             Alan Hayward  <alan.hayward@arm.com>
15941             David Sherwood  <david.sherwood@arm.com>
15943         * target.def (secondary_memory_needed): New hook.
15944         (secondary_reload): Refer to TARGET_SECONDARY_MEMORY_NEEDED
15945         instead of SECONDARY_MEMORY_NEEDED.
15946         (secondary_memory_needed_mode): Likewise.
15947         * hooks.h (hook_bool_mode_reg_class_t_reg_class_t_false): Declare.
15948         * hooks.c (hook_bool_mode_reg_class_t_reg_class_t_false): New function.
15949         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED): Replace with...
15950         (TARGET_SECONDARY_MEMORY_NEEDED): ...this.
15951         (SECONDARY_MEMORY_NEEDED_RTX): Update reference accordingly.
15952         * doc/tm.texi: Regenerate.
15953         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED): Delete.
15954         * config/alpha/alpha.c (alpha_secondary_memory_needed): New function.
15955         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15956         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED): Delete.
15957         * config/i386/i386-protos.h (ix86_secondary_memory_needed): Delete.
15958         * config/i386/i386.c (inline_secondary_memory_needed): Put the
15959         mode argument first and change the reg_class arguments to reg_class_t.
15960         (ix86_secondary_memory_needed): Likewise.  Remove the strict parameter.
15961         Make static.  Update the call to inline_secondary_memory_needed.
15962         (ix86_register_move_cost): Update the call to
15963         inline_secondary_memory_needed.
15964         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15965         * config/ia64/ia64.h (SECONDARY_MEMORY_NEEDED): Delete commented-out
15966         definition.
15967         * config/ia64/ia64.c (spill_xfmode_rfmode_operand): Refer to
15968         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15969         in comment.
15970         * config/mips/mips.h (SECONDARY_MEMORY_NEEDED): Delete.
15971         * config/mips/mips-protos.h (mips_secondary_memory_needed): Delete.
15972         * config/mips/mips.c (mips_secondary_memory_needed): Make static
15973         and match hook interface.  Add comment from mips.h.
15974         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15975         * config/mmix/mmix.md (truncdfsf2): Refer to
15976         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
15977         in comment.
15978         * config/pa/pa-64.h (SECONDARY_MEMORY_NEEDED): Rename to...
15979         (PA_SECONDARY_MEMORY_NEEDED): ...this, and put the mode argument first.
15980         * config/pa/pa.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15981         (pa_secondary_memory_needed): New function.
15982         * config/pdp11/pdp11.h (SECONDARY_MEMORY_NEEDED): Delete.
15983         * config/pdp11/pdp11-protos.h (pdp11_secondary_memory_needed): Delete.
15984         * config/pdp11/pdp11.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15985         (pdp11_secondary_memory_needed): Make static and match hook interface.
15986         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED): Delete.
15987         * config/powerpcspe/powerpcspe-protos.h
15988         (rs6000_secondary_memory_needed_ptr): Delete.
15989         * config/powerpcspe/powerpcspe.c (rs6000_secondary_memory_needed_ptr):
15990         Delete.
15991         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
15992         (rs6000_option_override_internal): Assign to
15993         targetm.secondary_memory_needed rather than
15994         rs6000_secondary_memory_needed_ptr.
15995         (rs6000_secondary_memory_needed): Match hook interface.
15996         (rs6000_debug_secondary_memory_needed): Likewise.
15997         * config/riscv/riscv.h (SECONDARY_MEMORY_NEEDED): Delete.
15998         * config/riscv/riscv.c (riscv_secondary_memory_needed): New function.
15999         (riscv_register_move_cost): Use it instead of SECONDARY_MEMORY_NEEDED.
16000         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16001         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED): Delete.
16002         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_ptr):
16003         Delete.
16004         * config/rs6000/rs6000.c (rs6000_secondary_memory_needed_ptr): Delete.
16005         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16006         (rs6000_option_override_internal): Assign to
16007         targetm.secondary_memory_needed rather than
16008         rs6000_secondary_memory_needed_ptr.
16009         (rs6000_secondary_memory_needed): Match hook interface.
16010         (rs6000_debug_secondary_memory_needed): Likewise.
16011         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED): Delete.
16012         * config/s390/s390.c (s390_secondary_memory_needed): New function.
16013         (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16014         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED): Delete.
16015         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED): Redefine.
16016         (sparc_secondary_memory_needed): New function.
16017         * lra-constraints.c (check_and_process_move): Refer to
16018         TARGET_SECONDARY_MEMORY_NEEDED rather than SECONDARY_MEMORY_NEEDED
16019         in comment.
16020         (curr_insn_transform): Likewise.
16021         (process_alt_operands): Use targetm.secondary_memory_needed
16022         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16023         (check_secondary_memory_needed_p): Likewise.
16024         (choose_split_class): Likewise.
16025         * reload.c: Unconditionally include code that was previously
16026         conditional on SECONDARY_MEMORY_NEEDED.
16027         (push_secondary_reload): Use targetm.secondary_memory_needed
16028         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16029         (push_reload): Likewise.
16030         * reload1.c: Unconditionally include code that was previously
16031         conditional on SECONDARY_MEMORY_NEEDED.
16032         (choose_reload_regs): Use targetm.secondary_memory_needed
16033         instead of TARGET_SECONDARY_MEMORY_NEEDED.
16034         (gen_reload): Likewise.
16035         * system.h (SECONDARY_MEMORY_NEEDED): Poison.
16037 2017-09-13  Richard Sandiford  <richard.sandiford@linaro.org>
16038             Alan Hayward  <alan.hayward@arm.com>
16039             David Sherwood  <david.sherwood@arm.com>
16041         * target.def (secondary_memory_needed_mode): New hook:
16042         * targhooks.c (default_secondary_memory_needed_mode): Declare.
16043         * targhooks.h (default_secondary_memory_needed_mode): New function.
16044         * doc/tm.texi.in (SECONDARY_MEMORY_NEEDED_MODE): Replace with...
16045         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): ...this.
16046         * doc/tm.texi: Regenerate.
16047         * lra-constraints.c (check_and_process_move): Use
16048         targetm.secondary_memory_needed_mode instead of
16049         TARGET_SECONDARY_MEMORY_NEEDED_MODE.
16050         (curr_insn_transform): Likewise.
16051         * reload.c (get_secondary_mem): Likewise.
16052         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16053         * config/alpha/alpha.c (alpha_secondary_memory_needed_mode): New
16054         function.
16055         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16056         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16057         * config/i386/i386.c (ix86_secondary_memory_needed_mode): New function.
16058         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16059         * config/powerpcspe/powerpcspe.h (SECONDARY_MEMORY_NEEDED_MODE):
16060         Delete.
16061         * config/powerpcspe/powerpcspe-protos.h
16062         (rs6000_secondary_memory_needed_mode): Delete.
16063         * config/powerpcspe/powerpcspe.c
16064         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16065         (rs6000_secondary_memory_needed_mode): Make static.
16066         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16067         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_mode):
16068         Delete.
16069         * config/rs6000/rs6000.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16070         Redefine.
16071         (rs6000_secondary_memory_needed_mode): Make static.
16072         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16073         * config/s390/s390.c (s390_secondary_memory_needed_mode): New function.
16074         (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Redefine.
16075         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Delete.
16076         * config/sparc/sparc.c (TARGET_SECONDARY_MEMORY_NEEDED_MODE):
16077         Redefine.
16078         (sparc_secondary_memory_needed_mode): New function.
16079         * system.h (TARGET_SECONDARY_MEMORY_NEEDED_MODE): Poison.
16081 2017-09-13  Jackson Woodruff  <jackson.woodruff@arm.com>
16083         * config/aarch64/constraints.md (Umq): New constraint.
16084         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>):
16085         Change to use Umq.
16086         (mov<mode>): Update condition.
16088 2017-09-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
16090         * gimple-ssa-store-merging.c (sort_by_bitpos): Compare store order
16091         when bitposition is the same.
16093 2017-09-13  Richard Biener  <rguenther@suse.de>
16095         * dwarf2out.c (output_die_symbol): Remove.
16096         (output_die): Do not output a DIEs symbol.
16098 2017-09-13  Richard Biener  <rguenther@suse.de>
16100         PR middle-end/82128
16101         * gimple-fold.c (gimple_fold_call): Update SSA name in-place to
16102         default-def to avoid breaking iterator update with the weird
16103         interaction with cgraph_update_edges_for_call_stmt_node.
16105 2017-09-13  Richard Biener  <rguenther@suse.de>
16107         * tree-cfg.c (verify_gimple_assign_binary): Add verification
16108         for WIDEN_SUM_EXPR, VEC_WIDEN_MULT_{HI,LO,EVEN,ODD}_EXPR,
16109         VEC_PACK_{TRUNC,SAT,FIX_TRUNC}_EXPR.
16110         (verify_gimple_assign_ternary): Add verification for DOT_PROD_EXPR.
16112 2017-09-13  Kugan Vivekanandarajah  <kuganv@linaro.org>
16114         * config/aarch64/aarch64.c (aarch64_override_options_after_change_1):
16115         Disable pc relative literal load irrespective of
16116         TARGET_FIX_ERR_A53_84341 for default.
16118 2017-09-12  Eric Botcazou  <ebotcazou@adacore.com>
16120         * config/sparc/sparc.c (output_return): Output the source location of
16121         the insn in the delay slot, if any.
16122         (output_sibcall): Likewise.
16124 2017-09-12  H.J. Lu  <hongjiu.lu@intel.com>
16126         PR driver/81498
16127         * common.opt (-static-pie): New alias.
16128         (shared): Negate static-pie.
16129         (-no-pie): Update help text.
16130         (-pie): Likewise.
16131         (static-pie): New option.
16132         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Add
16133         -static-pie support.
16134         (GNU_USER_TARGET_ENDFILE_SPEC): Likewise.
16135         (LINK_EH_SPEC): Likewise.
16136         (LINK_GCC_C_SEQUENCE_SPEC): Likewise.
16137         * config/i386/gnu-user.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16138         * config/i386/gnu-user64.h (GNU_USER_TARGET_LINK_SPEC): Likewise.
16139         * gcc.c (LINK_COMMAND_SPEC): Likewise.
16140         (init_gcc_specs): Likewise.
16141         (init_spec): Likewise.
16142         (display_help): Update help message for -pie.
16143         * doc/invoke.texi: Update -pie, -no-pie and -static.  Document
16144         -static-pie.
16146 2017-09-12  Wilco Dijkstra  <wdijkstr@arm.com>
16148         * config/aarch64/aarch64.md (movsi_aarch64): Remove all '*'.
16149         (movdi_aarch64): Likewise.
16150         (movti_aarch64): Likewise.
16152 2017-09-12 Simon Wright <simon@pushface.org>
16154         PR target/80204
16155         * config/darwin-driver.c (darwin_find_version_from_kernel): Eliminate
16156         calculation of the minor version, always output as 0.
16158 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16160         PR target/82112
16161         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
16162         ALTIVEC_BUILTIN_VEC_LD if arg1 has array type call default_conversion
16163         on it early, rather than manual conversion late.  For
16164         ALTIVEC_BUILTIN_VEC_ST if arg2 has array type call default_conversion
16165         instead of performing manual conversion.
16167 2017-09-12  Carl Love  <cel@us.ibm.com>
16169         * config/rs6000/altivec.md (vec_widen_umult_even_v4si,
16170         vec_widen_smult_even_v4si): Add define expands for vmuleuw, vmulesw,
16171         vmulouw, vmulosw.
16172         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
16173         VMULOSW): Add definitions.
16174         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
16175         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
16176         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
16177         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
16178         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
16180 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16182         * config/aarch64/aarch64.md (movdi_aarch64): Set load/store
16183         types correctly.
16184         (movti_aarch64): Likewise.
16185         (movdf_aarch64): Likewise.
16186         (movtf_aarch64): Likewise.
16187         (load_pairdi): Likewise.
16188         (store_pairdi): Likewise.
16189         (load_pairdf): Likewise.
16190         (store_pairdf): Likewise.
16191         (loadwb_pair<GPI:mode>_<P:mode>): Likewise.
16192         (storewb_pair<GPI:mode>_<P:mode>): Likewise.
16193         (ldr_got_small_<mode>): Likewise.
16194         (ldr_got_small_28k_<mode>): Likewise.
16195         (ldr_got_tiny): Likewise.
16196         * config/aarch64/iterators.md (ldst_sz): New.
16197         (ldpstp_sz): Likewise.
16198         * config/aarch64/thunderx.md (thunderx_storepair): Split store_8
16199         to store_16.
16200         (thunderx_load): Split load_8 to load_16.
16201         * config/aarch64/thunderx2t99.md (thunderx2t99_loadpair): Split
16202         load_8 to load_16.
16203         (thunderx2t99_storepair_basic): Split store_8 to store_16.
16204         * config/arm/xgene1.md (xgene1_load_pair): Split load_8 to load_16.
16205         (xgene1_store_pair): Split store_8 to store_16.
16206         * config/aarch64/falkor.md (falkor_ld_3_ld): Split load_8 to load_16.
16207         (falkor_st_0_st_sd): Split store_8 to store_16.
16209 2017-09-12  James Greenhalgh  <james.greenhalgh@arm.com>
16211         * config/arm/types.md (type): Rename load1/2/3/4 to load_4/8/12/16
16212         and store1/2/3/4 to store_4/8/12/16.
16213         * config/aarch64/aarch64.md: Update for rename.
16214         * config/arm/arm.md: Likewise.: Likewise.
16215         * config/arm/arm.c: Likewise.
16216         * config/arm/thumb1.md: Likewise.
16217         * config/arm/thumb2.md: Likewise.
16218         * config/arm/vfp.md: Likewise.
16219         * config/arm/arm-generic.md: Likewise.
16220         * config/arm/arm1020e.md: Likewise.
16221         * config/arm/arm1026ejs.md: Likewise.
16222         * config/arm/arm1136jfs.md: Likewise.
16223         * config/arm/arm926ejs.md: Likewise.
16224         * config/arm/cortex-a15.md: Likewise.
16225         * config/arm/cortex-a17.md: Likewise.
16226         * config/arm/cortex-a5.md: Likewise.
16227         * config/arm/cortex-a53.md: Likewise.
16228         * config/arm/cortex-a57.md: Likewise.
16229         * config/arm/cortex-a7.md: Likewise.
16230         * config/arm/cortex-a8.md: Likewise.
16231         * config/arm/cortex-a9.md: Likewise.
16232         * config/arm/cortex-m4.md: Likewise.
16233         * config/arm/cortex-m7.md: Likewise.
16234         * config/arm/cortex-r4.md: Likewise.
16235         * config/arm/exynos-m1.md: Likewise.
16236         * config/arm/fa526.md: Likewise.
16237         * config/arm/fa606te.md: Likewise.
16238         * config/arm/fa626te.md: Likewise.
16239         * config/arm/fa726te.md: Likewise.
16240         * config/arm/fmp626.md: Likewise.
16241         * config/arm/iwmmxt.md: Likewise.
16242         * config/arm/ldmstm.md: Likewise.
16243         * config/arm/marvell-pj4.md: Likewise.
16244         * config/arm/xgene1.md: Likewise.
16245         * config/aarch64/thunderx.md: Likewise.
16246         * config/aarch64/thunderx2t99.md: Likewise.
16247         * config/aarch64/falkor.md: Likewise.
16249 2017-09-12  Martin Liska  <mliska@suse.cz>
16251         * attribs.c (private_lookup_attribute): New function.
16252         * attribs.h (private_lookup_attribute): Declared here.
16253         (lookup_attribute): Called from this place.
16255 2017-09-12  Richard Biener  <rguenther@suse.de>
16257         PR tree-optimization/82157
16258         * tree-ssa-pre.c (remove_dead_inserted_code): Do not remove
16259         stmts with side-effects.
16261 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16262             Alan Hayward  <alan.hayward@arm.com>
16263             David Sherwood <david.sherwood@arm.com>
16265         * target.def (hard_regno_nregs): New hook.
16266         (class_max_nregs): Refer to it instead of HARD_REGNO_NREGS.
16267         * targhooks.h (default_hard_regno_nregs): Declare.
16268         * targhooks.c (default_hard_regno_nregs): New function.
16269         * doc/tm.texi.in (HARD_REGNO_NREGS): Replace with...
16270         (TARGET_HARD_REGNO_NREGS): ...this hook.
16271         (HARD_REGNO_NREGS_HAS_PADDING): Update accordingly.
16272         (CLASS_MAX_NREGS): Likewise.
16273         * doc/tm.texi: Regenerate.
16274         * reginfo.c (init_reg_modes_target): Use targetm.hard_regno_nregs
16275         instead of HARD_REGNO_NREGS.
16276         * rtl.h (REG_NREGS): Refer to TARGET_HARD_REGNO_NREGS rather than
16277         HARD_REGNO_NREGS in the comment.
16278         * config/aarch64/aarch64.h (HARD_REGNO_NREGS): Delete.
16279         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_nregs): Delete.
16280         * config/aarch64/aarch64.c (aarch64_hard_regno_nregs): Make static.
16281         Return an unsigned int.
16282         (TARGET_HARD_REGNO_NREGS): Redefine.
16283         * config/alpha/alpha.h (HARD_REGNO_NREGS): Delete.
16284         * config/arc/arc.h (HARD_REGNO_NREGS): Delete.
16285         * config/arc/arc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16286         (arc_hard_regno_nregs): New function.
16287         * config/arm/arm.h (HARD_REGNO_NREGS): Delete.
16288         * config/arm/arm.c (TARGET_HARD_REGNO_NREGS): Redefine.
16289         (arm_hard_regno_nregs): New function.
16290         * config/avr/avr.h (HARD_REGNO_NREGS): Delete.
16291         * config/bfin/bfin.h (HARD_REGNO_NREGS): Delete.
16292         * config/bfin/bfin.c (bfin_hard_regno_nregs): New function.
16293         (TARGET_HARD_REGNO_NREGS): Redefine.
16294         * config/c6x/c6x.h (HARD_REGNO_NREGS): Delete.
16295         * config/cr16/cr16.h (LONG_REG_P): Use targetm.hard_regno_nregs.
16296         (HARD_REGNO_NREGS): Delete.
16297         * config/cr16/cr16.c (TARGET_HARD_REGNO_NREGS): Redefine.
16298         (cr16_hard_regno_nregs): New function.
16299         (cr16_memory_move_cost): Use it instead of HARD_REGNO_NREGS.
16300         * config/cris/cris.h (HARD_REGNO_NREGS): Delete.
16301         * config/cris/cris.c (TARGET_HARD_REGNO_NREGS): Redefine.
16302         (cris_hard_regno_nregs): New function.
16303         * config/epiphany/epiphany.h (HARD_REGNO_NREGS): Delete.
16304         * config/fr30/fr30.h (HARD_REGNO_NREGS): Delete.
16305         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16306         * config/frv/frv.h (HARD_REGNO_NREGS): Delete.
16307         (CLASS_MAX_NREGS): Remove outdated copy of documentation.
16308         * config/frv/frv-protos.h (frv_hard_regno_nregs): Delete.
16309         * config/frv/frv.c (TARGET_HARD_REGNO_NREGS): Redefine.
16310         (frv_hard_regno_nregs): Make static.  Take and return an
16311         unsigned int.
16312         (frv_class_max_nregs): Remove outdated copy of documentation.
16313         * config/ft32/ft32.h (HARD_REGNO_NREGS): Delete.
16314         * config/h8300/h8300.h (HARD_REGNO_NREGS): Delete.
16315         * config/h8300/h8300-protos.h (h8300_hard_regno_nregs): Delete.
16316         * config/h8300/h8300.c (h8300_hard_regno_nregs): Delete.
16317         * config/i386/i386.h (HARD_REGNO_NREGS): Delete.
16318         * config/i386/i386.c (ix86_hard_regno_nregs): New function.
16319         (TARGET_HARD_REGNO_NREGS): Redefine.
16320         * config/ia64/ia64.h (HARD_REGNO_NREGS): Delete.
16321         (CLASS_MAX_NREGS): Update comment.
16322         * config/ia64/ia64.c (TARGET_HARD_REGNO_NREGS): Redefine.
16323         (ia64_hard_regno_nregs): New function.
16324         * config/iq2000/iq2000.h (HARD_REGNO_NREGS): Delete.
16325         * config/lm32/lm32.h (HARD_REGNO_NREGS): Delete.
16326         * config/m32c/m32c.h (HARD_REGNO_NREGS): Delete.
16327         * config/m32c/m32c-protos.h (m32c_hard_regno_nregs): Delete.
16328         * config/m32c/m32c.c (m32c_hard_regno_nregs_1): Take and return
16329         an unsigned int.
16330         (m32c_hard_regno_nregs): Likewise.  Make static.
16331         (TARGET_HARD_REGNO_NREGS): Redefine.
16332         * config/m32r/m32r.h (HARD_REGNO_NREGS): Delete.
16333         * config/m68k/m68k.h (HARD_REGNO_NREGS): Delete.
16334         * config/m68k/m68k.c (TARGET_HARD_REGNO_NREGS): Redefine.
16335         (m68k_hard_regno_nregs): New function.
16336         * config/mcore/mcore.h (HARD_REGNO_NREGS): Delete.
16337         * config/microblaze/microblaze.h (HARD_REGNO_NREGS): Delete.
16338         * config/mips/mips.h (HARD_REGNO_NREGS): Delete.
16339         * config/mips/mips-protos.h (mips_hard_regno_nregs): Delete.
16340         * config/mips/mips.c (mips_hard_regno_nregs): Make static.
16341         Take and return an unsigned int.
16342         (TARGET_HARD_REGNO_NREGS): Redefine.
16343         * config/mmix/mmix.h (HARD_REGNO_NREGS): Delete.
16344         (CLASS_MAX_NREGS): Use targetm.hard_regno_nregs.
16345         * config/mn10300/mn10300.h (HARD_REGNO_NREGS): Delete.
16346         * config/moxie/moxie.h (HARD_REGNO_NREGS): Delete.
16347         * config/msp430/msp430.h (HARD_REGNO_NREGS): Delete.
16348         * config/msp430/msp430-protos.h (msp430_hard_regno_nregs): Delete.
16349         * config/msp430/msp430.c (TARGET_HARD_REGNO_NREGS): Redefine.
16350         (msp430_hard_regno_nregs): Make static.  Take and return an
16351         unsigned int.
16352         * config/nds32/nds32.h (HARD_REGNO_NREGS): Delete.
16353         * config/nds32/nds32-protos.h (nds32_hard_regno_nregs): Delete.
16354         * config/nds32/nds32.c (nds32_hard_regno_nregs): Delete.
16355         (nds32_hard_regno_mode_ok): Use targetm.hard_regno_nregs.
16356         * config/nios2/nios2.h (HARD_REGNO_NREGS): Delete.
16357         * config/nvptx/nvptx.h (HARD_REGNO_NREGS): Delete.
16358         * config/nvptx/nvptx.c (nvptx_hard_regno_nregs): New function.
16359         (TARGET_HARD_REGNO_NREGS): Redefine.
16360         * config/pa/pa32-regs.h (HARD_REGNO_NREGS): Rename to...
16361         (PA_HARD_REGNO_NREGS): ...this.
16362         * config/pa/pa64-regs.h (HARD_REGNO_NREGS): Rename to...
16363         (PA_HARD_REGNO_NREGS): ...this.
16364         * config/pa/pa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16365         (pa_hard_regno_nregs): New function.
16366         * config/pdp11/pdp11.h (HARD_REGNO_NREGS): Delete.
16367         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_NREGS): Redefine.
16368         (pdp11_hard_regno_nregs): New function.
16369         * config/powerpcspe/powerpcspe.h (HARD_REGNO_NREGS): Delete.
16370         * config/powerpcspe/powerpcspe.c (TARGET_HARD_REGNO_NREGS): Redefine.
16371         (rs6000_hard_regno_nregs_hook): New function.
16372         * config/riscv/riscv.h (HARD_REGNO_NREGS): Delete.
16373         * config/riscv/riscv-protos.h (riscv_hard_regno_nregs): Delete.
16374         * config/riscv/riscv.c (riscv_hard_regno_nregs): Make static.
16375         Take and return an unsigned int.  Move earlier in file.
16376         (TARGET_HARD_REGNO_NREGS): Redefine.
16377         * config/rl78/rl78.h (HARD_REGNO_NREGS): Delete.
16378         * config/rl78/rl78-protos.h (rl78_hard_regno_nregs): Delete.
16379         * config/rl78/rl78.c (TARGET_HARD_REGNO_NREGS): Reefine.
16380         (rl78_hard_regno_nregs): Make static.  Take and return an
16381         unsigned int.
16382         * config/rs6000/rs6000.h (HARD_REGNO_NREGS): Delete.
16383         * config/rs6000/rs6000.c (TARGET_HARD_REGNO_NREGS): Redefine.
16384         (rs6000_hard_regno_nregs_hook): New function.
16385         * config/rx/rx.h (HARD_REGNO_NREGS): Delete.
16386         * config/rx/rx.c (rx_hard_regno_nregs): New function.
16387         (TARGET_HARD_REGNO_NREGS): Redefine.
16388         * config/s390/s390.h (HARD_REGNO_NREGS): Delete.
16389         * config/s390/s390.c (REGNO_PAIR_OK): Use s390_hard_regno_nregs
16390         instead of HARD_REGNO_NREGS.
16391         (s390_hard_regno_nregs): New function.
16392         (s390_hard_regno_mode_ok): Add comment from s390.h.
16393         (TARGET_HARD_REGNO_NREGS): Redefine.
16394         * config/sh/sh.h (HARD_REGNO_NREGS): Delete.
16395         * config/sh/sh.c (TARGET_HARD_REGNO_NREGS): Redefine.
16396         (sh_hard_regno_nregs): New function.
16397         (sh_pass_in_reg_p): Use it.
16398         * config/sparc/sparc.h (HARD_REGNO_NREGS): Delete.
16399         * config/sparc/sparc.c (TARGET_HARD_REGNO_NREGS): Redefine.
16400         (sparc_hard_regno_nregs): New function.
16401         * config/spu/spu.h (HARD_REGNO_NREGS): Delete.
16402         * config/spu/spu.c (spu_hard_regno_nregs): New function.
16403         (spu_function_arg_advance): Use it, supplying a valid register number.
16404         (TARGET_HARD_REGNO_NREGS): Redefine.
16405         * config/stormy16/stormy16.h (HARD_REGNO_NREGS): Delete.
16406         * config/tilegx/tilegx.h (HARD_REGNO_NREGS): Delete.
16407         * config/tilepro/tilepro.h (HARD_REGNO_NREGS): Delete.
16408         * config/v850/v850.h (HARD_REGNO_NREGS): Delete.
16409         * config/vax/vax.h (HARD_REGNO_NREGS): Delete.
16410         * config/visium/visium.h (HARD_REGNO_NREGS): Delete.
16411         (CLASS_MAX_NREGS): Remove copy of old documentation.
16412         * config/visium/visium.c (TARGET_HARD_REGNO_NREGS): Redefine.
16413         (visium_hard_regno_nregs): New function.
16414         (visium_hard_regno_mode_ok): Use it instead of HARD_REGNO_NREGS.
16415         * config/xtensa/xtensa.h (HARD_REGNO_NREGS): Delete.
16416         * config/xtensa/xtensa.c (TARGET_HARD_REGNO_NREGS): Redefine.
16417         xtensa_hard_regno_nregs): New function.
16418         * system.h (HARD_REGNO_NREGS): Poison.
16420 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16422         * config/arm/arm.h (THUMB_SECONDARY_INPUT_RELOAD_CLASS): Use
16423         hard_regno_nregs instead of HARD_REGNO_NREGS.
16424         (THUMB_SECONDARY_OUTPUT_RELOAD_CLASS): Likewise.
16425         * config/c6x/c6x.c (c6x_expand_prologue): Likewise.
16426         (c6x_expand_epilogue): Likewise.
16427         * config/frv/frv.c (frv_alloc_temp_reg): Likewise.
16428         (frv_read_iacc_argument): Likewise.
16429         * config/sh/sh.c: Include regs.h.
16430         (sh_print_operand): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16431         (regs_used): Likewise.
16432         (output_stack_adjust): Likewise.
16433         * config/xtensa/xtensa.c (xtensa_copy_incoming_a7): Likewise.
16434         * expmed.c: Include regs.h.
16435         (store_bit_field_1): Use hard_regno_nregs instead of HARD_REGNO_NREGS.
16436         * ree.c: Include regs.h.
16437         (combine_reaching_defs): Use hard_regno_nregs instead of
16438         HARD_REGNO_NREGS.
16439         (add_removable_extension): Likewise.
16441 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16443         * regs.h (hard_regno_nregs): Turn into a function.
16444         (end_hard_regno): Update accordingly.
16445         * caller-save.c (setup_save_areas): Likewise.
16446         (save_call_clobbered_regs): Likewise.
16447         (replace_reg_with_saved_mem): Likewise.
16448         (insert_restore): Likewise.
16449         (insert_save): Likewise.
16450         * combine.c (can_change_dest_mode): Likewise.
16451         (move_deaths): Likewise.
16452         (distribute_notes): Likewise.
16453         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): Likewise.
16454         * config/powerpcspe/powerpcspe.c (rs6000_cannot_change_mode_class)
16455         (rs6000_split_multireg_move): Likewise.
16456         (rs6000_register_move_cost): Likewise.
16457         (rs6000_memory_move_cost): Likewise.
16458         * config/rs6000/rs6000.c (rs6000_cannot_change_mode_class): Likewise.
16459         (rs6000_split_multireg_move): Likewise.
16460         (rs6000_register_move_cost): Likewise.
16461         (rs6000_memory_move_cost): Likewise.
16462         * cselib.c (cselib_reset_table): Likewise.
16463         (cselib_lookup_1): Likewise.
16464         * emit-rtl.c (set_mode_and_regno): Likewise.
16465         * function.c (aggregate_value_p): Likewise.
16466         * ira-color.c (setup_profitable_hard_regs): Likewise.
16467         (check_hard_reg_p): Likewise.
16468         (calculate_saved_nregs): Likewise.
16469         (assign_hard_reg): Likewise.
16470         (improve_allocation): Likewise.
16471         (calculate_spill_cost): Likewise.
16472         * ira-emit.c (modify_move_list): Likewise.
16473         * ira-int.h (ira_hard_reg_set_intersection_p): Likewise.
16474         (ira_hard_reg_in_set_p): Likewise.
16475         * ira.c (setup_reg_mode_hard_regset): Likewise.
16476         (clarify_prohibited_class_mode_regs): Likewise.
16477         (check_allocation): Likewise.
16478         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16479         (lra_setup_reg_renumber): Likewise.
16480         (setup_try_hard_regno_pseudos): Likewise.
16481         (spill_for): Likewise.
16482         (assign_hard_regno): Likewise.
16483         (setup_live_pseudos_and_spill_after_risky_transforms): Likewise.
16484         * lra-constraints.c (in_class_p): Likewise.
16485         (lra_constraint_offset): Likewise.
16486         (simplify_operand_subreg): Likewise.
16487         (lra_constraints): Likewise.
16488         (split_reg): Likewise.
16489         (split_if_necessary): Likewise.
16490         (invariant_p): Likewise.
16491         (inherit_in_ebb): Likewise.
16492         * lra-lives.c (process_bb_lives): Likewise.
16493         * lra-remat.c (reg_overlap_for_remat_p): Likewise.
16494         (get_hard_regs): Likewise.
16495         (do_remat): Likewise.
16496         * lra-spills.c (assign_spill_hard_regs): Likewise.
16497         * mode-switching.c (create_pre_exit): Likewise.
16498         * postreload.c (reload_combine_recognize_pattern): Likewise.
16499         * recog.c (peep2_find_free_register): Likewise.
16500         * regcprop.c (kill_value_regno): Likewise.
16501         (set_value_regno): Likewise.
16502         (copy_value): Likewise.
16503         (maybe_mode_change): Likewise.
16504         (find_oldest_value_reg): Likewise.
16505         (copyprop_hardreg_forward_1): Likewise.
16506         * regrename.c (check_new_reg_p): Likewise.
16507         (regrename_do_replace): Likewise.
16508         * reload.c (push_reload): Likewise.
16509         (combine_reloads): Likewise.
16510         (find_dummy_reload): Likewise.
16511         (operands_match_p): Likewise.
16512         (find_reloads): Likewise.
16513         (find_equiv_reg): Likewise.
16514         (reload_adjust_reg_for_mode): Likewise.
16515         * reload1.c (count_pseudo): Likewise.
16516         (count_spilled_pseudo): Likewise.
16517         (find_reg): Likewise.
16518         (clear_reload_reg_in_use): Likewise.
16519         (free_for_value_p): Likewise.
16520         (allocate_reload_reg): Likewise.
16521         (choose_reload_regs): Likewise.
16522         (reload_adjust_reg_for_temp): Likewise.
16523         (emit_reload_insns): Likewise.
16524         (delete_output_reload): Likewise.
16525         * rtlanal.c (subreg_get_info): Likewise.
16526         * sched-deps.c (sched_analyze_reg): Likewise.
16527         * sel-sched.c (init_regs_for_mode): Likewise.
16528         (mark_unavailable_hard_regs): Likewise.
16529         (choose_best_reg_1): Likewise.
16530         (verify_target_availability): Likewise.
16531         * valtrack.c (dead_debug_insert_temp): Likewise.
16532         * var-tracking.c (track_loc_p): Likewise.
16533         (emit_note_insn_var_location): Likewise.
16534         * varasm.c (make_decl_rtl): Likewise.
16535         * reginfo.c (choose_hard_reg_mode): Likewise.
16536         (init_reg_modes_target): Refer directly to
16537         this_target_regs->x_hard_regno_nregs.
16539 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16541         * ira-costs.c (record_operand_costs): Use in_hard_reg_set_p
16542         instead of hard_regno_nregs.
16544 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16546         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Use
16547         end_hard_regno instead of hard_regno_nregs.
16548         * config/s390/s390.c (s390_reg_clobbered_rtx): Likewise.
16549         * config/sparc/sparc.h (ASM_DECLARE_REGISTER_GLOBAL): Likewise.
16550         * config/visium/visium.c (visium_hard_regno_mode_ok): Likewise.
16551         * ira-color.c (improve_allocation): Likewise.
16552         * lra-assigns.c (find_hard_regno_for_1): Likewise.
16553         * lra-lives.c (mark_regno_live): Likewise.
16554         (mark_regno_dead): Likewise.
16555         * lra-remat.c (operand_to_remat): Likewise.
16556         * lra.c (collect_non_operand_hard_regs): Likewise.
16557         * postreload.c (reload_combine_note_store): Likewise.
16558         (move2add_valid_value_p): Likewise.
16559         * reload.c (regno_clobbered_p): Likewise.
16561 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16563         * config/frv/frv.c (FOR_EACH_REGNO): Use END_REGNO instead of
16564         hard_regno_nregs.
16565         * config/v850/v850.c (v850_reorg): Likewise.
16566         * reload.c (refers_to_regno_for_reload_p): Likewise.
16567         (find_equiv_reg): Likewise.
16568         * reload1.c (reload_reg_reaches_end_p): Likewise.
16570 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16572         * caller-save.c (add_used_regs): Use REG_NREGS instead of
16573         hard_regno_nregs.
16574         * config/aarch64/aarch64.c (aarch64_split_combinev16qi): Likewise.
16575         * config/arm/arm.c (output_move_neon): Likewise.
16576         (arm_attr_length_move_neon): Likewise.
16577         (neon_split_vcombine): Likewise.
16578         * config/c6x/c6x.c (c6x_mark_reg_read): Likewise.
16579         (c6x_mark_reg_written): Likewise.
16580         (c6x_dwarf_register_span): Likewise.
16581         * config/i386/i386.c (ix86_save_reg): Likewise.
16582         * config/ia64/ia64.c (mark_reg_gr_used_mask): Likewise.
16583         (rws_access_reg): Likewise.
16584         * config/s390/s390.c (s390_call_saved_register_used): Likewise.
16585         * mode-switching.c (create_pre_exit): Likewise.
16586         * ree.c (combine_reaching_defs): Likewise.
16587         (add_removable_extension): Likewise.
16588         * regcprop.c (find_oldest_value_reg): Likewise.
16589         (copyprop_hardreg_forward_1): Likewise.
16590         * reload.c (reload_inner_reg_of_subreg): Likewise.
16591         (push_reload): Likewise.
16592         (combine_reloads): Likewise.
16593         (find_dummy_reload): Likewise.
16594         (reload_adjust_reg_for_mode): Likewise.
16595         * reload1.c (find_reload_regs): Likewise.
16596         (forget_old_reloads_1): Likewise.
16597         (reload_reg_free_for_value_p): Likewise.
16598         (reload_adjust_reg_for_temp): Likewise.
16599         (emit_reload_insns): Likewise.
16600         (delete_output_reload): Likewise.
16601         * sel-sched.c (choose_best_reg_1): Likewise.
16602         (choose_best_pseudo_reg): Likewise.
16604 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16605             Alan Hayward  <alan.hayward@arm.com>
16606             David Sherwood <david.sherwood@arm.com>
16608         * defaults.h (SLOW_UNALIGNED_ACCESS): Delete.
16609         * target.def (slow_unaligned_access): New hook.
16610         * targhooks.h (default_slow_unaligned_access): Declare.
16611         * targhooks.c (default_slow_unaligned_access): New function.
16612         * doc/tm.texi.in (SLOW_UNALIGNED_ACCESS): Replace with...
16613         (TARGET_SLOW_UNALIGNED_ACCESS): ...this.
16614         * doc/tm.texi: Regenerate.
16615         * config/alpha/alpha.h (SLOW_UNALIGNED_ACCESS): Delete.
16616         * config/arm/arm.h (SLOW_UNALIGNED_ACCESS): Delete.
16617         * config/i386/i386.h (SLOW_UNALIGNED_ACCESS): Delete commented-out
16618         definition.
16619         * config/powerpcspe/powerpcspe.h (SLOW_UNALIGNED_ACCESS): Delete.
16620         * config/powerpcspe/powerpcspe.c (TARGET_SLOW_UNALIGNED_ACCESS):
16621         Redefine.
16622         (rs6000_slow_unaligned_access): New function.
16623         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16624         (expand_block_compare): Likewise.
16625         (expand_strn_compare): Likewise.
16626         (rs6000_rtx_costs): Likewise.
16627         * config/riscv/riscv.h (SLOW_UNALIGNED_ACCESS): Delete.
16628         (riscv_slow_unaligned_access): Likewise.
16629         * config/riscv/riscv.c (riscv_slow_unaligned_access): Rename to...
16630         (riscv_slow_unaligned_access_p): ...this and make static.
16631         (riscv_option_override): Update accordingly.
16632         (riscv_slow_unaligned_access): New function.
16633         (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16634         * config/rs6000/rs6000.h (SLOW_UNALIGNED_ACCESS): Delete.
16635         * config/rs6000/rs6000.c (TARGET_SLOW_UNALIGNED_ACCESS): Redefine.
16636         (rs6000_slow_unaligned_access): New function.
16637         (rs6000_emit_move): Use it instead of SLOW_UNALIGNED_ACCESS.
16638         (rs6000_rtx_costs): Likewise.
16639         * config/rs6000/rs6000-string.c (expand_block_compare)
16640         (expand_strn_compare): Use targetm.slow_unaligned_access instead
16641         of SLOW_UNALIGNED_ACCESS.
16642         * config/tilegx/tilegx.h (SLOW_UNALIGNED_ACCESS): Delete.
16643         * config/tilepro/tilepro.h (SLOW_UNALIGNED_ACCESS): Delete.
16644         * calls.c (expand_call): Use targetm.slow_unaligned_access instead
16645         of SLOW_UNALIGNED_ACCESS.
16646         * expmed.c (simple_mem_bitfield_p): Likewise.
16647         * expr.c (alignment_for_piecewise_move): Likewise.
16648         (emit_group_load_1): Likewise.
16649         (emit_group_store): Likewise.
16650         (copy_blkmode_from_reg): Likewise.
16651         (emit_push_insn): Likewise.
16652         (expand_assignment): Likewise.
16653         (store_field): Likewise.
16654         (expand_expr_real_1): Likewise.
16655         * gimple-fold.c (gimple_fold_builtin_memory_op): Likewise.
16656         * lra-constraints.c (simplify_operand_subreg): Likewise.
16657         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
16658         * gimple-ssa-store-merging.c: Likewise in block comment at start
16659         of file.
16660         * tree-ssa-strlen.c: Include target.h.
16661         (handle_builtin_memcmp): Use targetm.slow_unaligned_access instead
16662         of SLOW_UNALIGNED_ACCESS.
16663         * system.h (SLOW_UNALIGNED_ACCESS): Poison.
16665 2017-09-12  Richard Sandiford  <richard.sandiford@linaro.org>
16667         PR rtl-optimization/82185
16668         * expmed.c (emit_store_flag_int): Only test tem if it has been
16669         initialized.
16671 2017-09-12  Richard Biener  <rguenther@suse.de>
16673         PR middle-end/82149
16674         * match.pd ((FTYPE) N CMP CST): Fix typo.
16676 2017-09-12  Simon Atanasyan  <simon.atanasyan@imgtec.com>
16678         * config/mips/mips.c (mips_attribute_table): Add 'short_call'
16679         attribute.
16680         (mips_near_type_p): Add 'short_call' attribute as a synonym
16681         for 'near'.
16682         * doc/extend.texi (short_call): Document new function attribute.
16684 2017-09-12  Jakub Jelinek  <jakub@redhat.com>
16686         PR target/82112
16687         * c-common.c (sync_resolve_size): Instead of c_dialect_cxx ()
16688         assertion check that in the condition.
16689         (get_atomic_generic_size): Likewise.  Before testing if parameter
16690         has pointer type, if it has array type, call for C++
16691         default_conversion to perform array-to-pointer conversion.
16693 2017-09-12  Richard Biener  <rguenther@suse.de>
16695         * tree-vect-generic.c (expand_vector_operations_1): Do nothing
16696         for operations we cannot scalarize.
16698 2017-09-12  Aldy Hernandez  <aldyh@redhat.com>
16700         * tree-ssa-threadbackward.c (fsm_find_thread_path): Make GC
16701         vectors heap vectors.  Clean up comments.
16702         Make visited_bbs a reference.
16703         (profitable_jump_thread_path): Make GC
16704         vectors heap vectors.  Clean up comments.
16705         Misc cleanups.
16706         (convert_and_register_jump_thread_path): Make GC vectors heap
16707         vectors.
16708         (check_subpath_and_update_thread_path): Same.  Clean up comments.
16709         Make visited_bbs a reference.
16710         (handle_phi): Abstract common code to to
16711         register_jump_thread_path_if_profitable.
16712         Rename VAR_BB to DEF_BB.
16713         Update comments.
16714         Make GC vectors heap vectors.
16715         Make visited_bbs a reference.
16716         (handle_assignment): Same.
16717         (register_jump_thread_path_if_profitable): New.
16718         (fsm_find_control_statement_thread_paths): Rename VAR_BB to
16719         DEF_BB.
16720         Make GC vectors heap vectors.  Clean up comments.
16721         Make visited_bbs a reference.
16722         (find_jump_threads_backwards): Make visited_bbs live in the stack.
16723         * tree-ssa-threadupdate.c (delete_jump_thread_path): Fix typo in
16724         comment.
16726 2017-09-11  Max Filippov  <jcmvbkbc@gmail.com>
16728         PR target/82181
16729         * config/xtensa/xtensa.c (xtensa_mem_offset): Check that both
16730         words of E_DImode object are reachable by xtensa_uimm8x4 access.
16732 2017-09-11  Vidya Praveen  <vidyapraveen@arm.com>
16734         Revert r251800 and r251799.
16736 2017-09-11  Martin Jambor  <mjambor@suse.cz>
16738         PR hsa/82119
16739         * hsa-gen.c (gen_hsa_phi_from_gimple_phi): Process ADDR_EXPRs in
16740         arguments in advance.
16741         * hsa-regalloc.c (naive_process_phi): New parameter predecessors,
16742         use it to find predecessor edges.
16743         (naive_outof_ssa): Collect vector of predecessors.
16745 2017-09-08  Jason Merrill  <jason@redhat.com>
16747         PR c++/70029 - ICE with ref-qualifier and -flto
16748         * langhooks.h (struct lang_hooks_for_types): Add
16749         copy_lang_qualifiers.
16750         * attribs.c (build_type_attribute_qual_variant): Use it.
16751         * langhooks-def.h (LANG_HOOKS_COPY_LANG_QUALIFIERS): Default to
16752         NULL.
16753         (LANG_HOOKS_FOR_TYPES_INITIALIZER): Use it.
16754         * tree.c (verify_type): Re-enable TYPE_CANONICAL main variant check.
16756 2017-09-08  Eric Botcazou  <ebotcazou@adacore.com>
16758         PR target/81988
16759         * config/sparc/sparc.md (mulsi3): Rename into *mulsi3_sp32.
16760         (*mulsi3_sp64): New instruction.
16761         (mulsi3): New expander.
16763 2017-09-08  Uros Bizjak  <ubizjak@gmail.com>
16765         * config/alpha/alpha.c (alpha_print_operand) <case 'S'>: Remove.
16767 2017-09-08  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
16769         * sancov.c: Include memmodel.h.
16771 2017-09-07  Eric Botcazou  <ebotcazou@adacore.com>
16773         PR target/80897
16774         * config/sparc/sparc.c (sparc_emit_set_symbolic_const64): Deal with too
16775         large offsets.
16777 2017-09-07  Carl Love  <cel@us.ibm.com>
16779         * config/rs6000/vsx.md (define_insn "*stxvl"): Add missing argument to
16780         the sldi instruction.
16782 2017-09-07  David Edelsohn  <dje.gcc@gmail.com>
16784         * sancov.c: Include tm_p.h.
16786 2017-09-07  Jakub Jelinek  <jakub@redhat.com>
16788         PR target/81979
16789         * output.h (switch_to_other_text_partition): New declaration.
16790         * varasm.c (switch_to_other_text_partition): New function.
16791         * config/rs6000/rs6000.c (uses_TOC): Return 2 if
16792         NOTE_INSN_SWITCH_TEXT_SECTIONS is seen before finding load_toc_* insn.
16793         (rs6000_elf_declare_function_name): If uses_TOC returned 2, switch
16794         to the other text partition before emitting LCL label and switch back
16795         after emitting the word after it.
16797 2017-09-07  Richard Biener  <rguenther@suse.de>
16799         * passes.def (pass_split_crit_edges): Remove instance before PRE.
16800         * tree-ssa-pre.c (pass_pre::execute): Instead manually split
16801         critical edges here, after loop init.
16802         (pass_data_pre): Remove PROP_no_crit_edges flags.
16803         * tree-ssa-sccvn.c (vn_reference_lookup_3): Use vn_valueize
16804         for valueization of call args to avoid leaking VN_TOP.
16805         (visit_use): Assert we do not visit default defs.
16806         (init_scc_vn): Use build_decl for VN_TOP to make name nicer.
16807         Use error_mark_node to more easily detect leaking VN_TOP.
16808         All default-defs are varying, not VN_TOP.  Mark them visited.
16809         (run_scc_vn): Make code match comment.
16811 2017-09-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
16813         * config/rs6000/rs6000-cpus.def (OTHER_VSX_VECTOR_MASKS): Delete
16814         OPTION_MASK_FLOAT128_KEYWORD.
16815         (POWERPC_MASKS): Likewise.
16816         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Delete
16817         support for the -mfloat128-type option, and make -mfloat128
16818         default on PowerPC Linux systems.  Define or undefine
16819         __FLOAT128__ and  __FLOAT128_HARDWARE__ for the current options.
16820         Define __float128 to be __ieee128 if IEEE 128-bit support is
16821         enabled, or undefine it.
16822         (rs6000_cpu_cpp_builtins): Delete defining __FLOAT128__ here.
16823         Delete defining __FLOAT128_TYPE__.
16824         * config/rs6000/rs6000.opt (x_TARGET_FLOAT128_TYPE): Delete the
16825         -mfloat128-type option and make -mfloat128 default on PowerPC
16826         Linux systems.
16827         (TARGET_FLOAT128_TYPE): Likewise.
16828         (-mfloat128-type): Likewise.
16829         * config/rs6000/rs6000.c (rs6000_option_override_internal):
16830         Delete the -mfloat128-type option and make -mfloat128 default on
16831         PowerPC Linux systems.  Always use __ieee128 to be the keyword for
16832         the IEEE 128-bit type, and map __float128 to __ieee128 if IEEE
16833         128-bit floating point is enabled.  Change tests from using
16834         -mfloat128-type to -mfloat128.
16835         (rs6000_mangle_type): Use the correct mangling for the __float128
16836         type even if normal long double is restricted to 64-bits.
16837         (floatn_mode): Enable the _Float128 type by default on VSX Linux
16838         systems.
16839         * config/rs6000/rs6000.h (MASK_FLOAT128_TYPE): Delete.
16840         (MASK_FLOAT128_KEYWORD): Define new shortcut macro.
16841         (RS6000BTM_FLOAT128): Define in terms of -mfloat128, not
16842         -mfloat128-type.
16843         * doc/invoke.texi (RS/6000 and PowerPC Options): Update
16844         documentation for -mfloat128.
16846 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16848         * config.gcc (powerpc-wrs-vxworksspe): Now match as vxworks*spe.
16850 2017-09-06  Wish Wu  <wishwu007@gmail.com>
16851             Jakub Jelinek  <jakub@redhat.com>
16853         * asan.c (initialize_sanitizer_builtins): Add
16854         BT_FN_VOID_UINT8_UINT8, BT_FN_VOID_UINT16_UINT16,
16855         BT_FN_VOID_UINT32_UINT32, BT_FN_VOID_UINT64_UINT64,
16856         BT_FN_VOID_FLOAT_FLOAT, BT_FN_VOID_DOUBLE_DOUBLE and
16857         BT_FN_VOID_UINT64_PTR variables.
16858         * builtin-types.def (BT_FN_VOID_UINT8_UINT8): New fn type.
16859         (BT_FN_VOID_UINT16_UINT16): Likewise.
16860         (BT_FN_VOID_UINT32_UINT32): Likewise.
16861         (BT_FN_VOID_FLOAT_FLOAT): Likewise.
16862         (BT_FN_VOID_DOUBLE_DOUBLE): Likewise.
16863         (BT_FN_VOID_UINT64_PTR): Likewise.
16864         * common.opt (flag_sanitize_coverage): New variable.
16865         (fsanitize-coverage=trace-pc): Remove.
16866         (fsanitize-coverage=): Add.
16867         * flag-types.h (enum sanitize_coverage_code): New enum.
16868         * fold-const.c (fold_range_test): Disable non-short-circuit
16869         optimization if flag_sanitize_coverage.
16870         (fold_truth_andor): Likewise.
16871         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
16872         * opts.c (COVERAGE_SANITIZER_OPT): Define.
16873         (coverage_sanitizer_opts): New array.
16874         (get_closest_sanitizer_option): Add OPTS argument, handle also
16875         OPT_fsanitize_coverage_.
16876         (parse_sanitizer_options): Adjusted to also handle
16877         OPT_fsanitize_coverage_.
16878         (common_handle_option): Add OPT_fsanitize_coverage_.
16879         * sancov.c (instrument_comparison, instrument_switch): New function.
16880         (sancov_pass): Add trace-cmp support.
16881         * sanitizer.def (BUILT_IN_SANITIZER_COV_TRACE_CMP1,
16882         BUILT_IN_SANITIZER_COV_TRACE_CMP2, BUILT_IN_SANITIZER_COV_TRACE_CMP4,
16883         BUILT_IN_SANITIZER_COV_TRACE_CMP8,
16884         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP1,
16885         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP2,
16886         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP4,
16887         BUILT_IN_SANITIZER_COV_TRACE_CONST_CMP8,
16888         BUILT_IN_SANITIZER_COV_TRACE_CMPF, BUILT_IN_SANITIZER_COV_TRACE_CMPD,
16889         BUILT_IN_SANITIZER_COV_TRACE_SWITCH): New builtins.
16890         * doc/invoke.texi: Document -fsanitize-coverage=trace-cmp.
16892 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16894         * config/arm/parsecpu.awk (fatal): Note that we've encountered an
16895         error.  Only quit immediately if parsing is complete.
16896         (BEGIN): Initialize fatal_err and parse_done.
16897         (begin fpu, end fpu): Check number of arguments.
16898         (begin arch, end arch): Likewise.
16899         (begin cpu, end cpu): Likewise.
16900         (cname, tune for, tune flags, architecture, fpu, option): Likewise.
16901         (optalias): Likewise.
16903 2017-09-06  Richard Earnshaw  <rearnsha@arm.com>
16905         * config.gcc (arm*-*-*): Don't add arm-isa.h to tm_p_file.
16906         * config/arm/arm-isa.h: Delete.  Move definitions to ...
16907         * arm-cpus.in: ... here.  Use new feature and fgroup values.
16908         * config/arm/arm.c (arm_option_override): Use lower case for feature
16909         bit names.
16910         * config/arm/arm.h (TARGET_HARD_FLOAT): Likewise.
16911         (TARGET_VFP3, TARGET_VFP5, TARGET_FMA): Likewise.
16912         * config/arm/parsecpu.awk (END): Add new command 'isa'.
16913         (isa_pfx): Delete.
16914         (print_isa_bits_for): New function.
16915         (gen_isa): New function.
16916         (gen_comm_data): Use print_isa_bits_for.
16917         (define feature): New keyword.
16918         (define fgroup): New keyword.
16919         * config/arm/t-arm (OPTIONS_H_EXTRA): Add arm-isa.h
16920         (arm-isa.h): Add rule to generate file.
16921         * common/config/arm/arm-common.c: (arm_canon_arch_option): Use lower
16922         case for feature bit names.
16924 2017-09-06  Richard Biener  <rguenther@suse.de>
16926         * tree-ssa-pre.c (NECESSARY): Remove.
16927         (create_expression_by_pieces): Do not touch pass-local flags.
16928         (insert_into_preds_of_block): Likewise.
16929         (do_pre_regular_insertion): Likewise.
16930         (eliminate_insert): Likewise.
16931         (eliminate_dom_walker::before_dom_children): Likewise.
16932         (fini_eliminate): Do not look at inserted_exprs.
16933         (mark_operand_necessary): Remove.
16934         (remove_dead_inserted_code): Replace with simple work-list
16935         algorithm based on inserted_exprs and SSA uses.
16936         (pass_pre::execute): Re-order fini_eliminate and
16937         remove_dead_inserted_code.
16939 2017-09-06  Olivier Hainque  <hainque@adacore.com>
16941         * config/powerpcspe/vxworks.h (VXCPU_FOR_8548): Correct definition
16942         for VxWorks 7.  Adjust surrounding comments.
16944 2017-09-06  Richard Biener  <rguenther@suse.de>
16946         * gimple-ssa-strength-reduction.c
16947         (find_candidates_dom_walker::before_dom_children): Also allow
16948         pointer types.
16950 2017-09-06  Richard Biener  <rguenther@suse.de>
16952         PR tree-optimization/82108
16953         * tree-vect-stmts.c (vectorizable_load): Fix pointer adjustment
16954         for gap in the non-permutation SLP case.
16956 2017-09-06  Martin Jambor  <mjambor@suse.cz>
16958         PR tree-optimization/82078
16959         * tree-sra.c (sort_and_splice_var_accesses): Move call to
16960         add_access_to_work_queue...
16961         (build_accesses_from_assign): ...here.
16962         (propagate_all_subaccesses): Make sure racc is the group
16963         representative, if there is one.
16965 2017-09-06  Jakub Jelinek  <jakub@redhat.com>
16967         PR middle-end/82095
16968         * varasm.c (categorize_decl_for_section): Use SECCAT_TBSS for TLS vars with
16969         NULL DECL_INITIAL.
16971 2017-09-06  Richard Biener  <rguenther@suse.de>
16973         * gimple-ssa-strength-reduction.c
16974         (find_candidates_dom_walker::before_doom_children): Use a
16975         type and not a mode check.
16977 2017-09-06  Bernd Edlinger  <bernd.edlinger@hotmail.de>
16979         PR target/77308
16980         * config/arm/predicates.md (arm_general_adddi_operand): Create new
16981         non-vfp predicate.
16982         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Use new predicates.
16984 2017-09-05  Jeff Law  <law@redhat.com>
16986         PR tree-optimization/64910
16987         * tree-ssa-reassoc.c (reassociate_bb): Restrict last change to
16988         cases where we have 3 or more operands.
16990 2017-09-05  Jakub Jelinek  <jakub@redhat.com>
16992         PR middle-end/81768
16993         * omp-low.c (lower_omp_for): Recompute tree invariant if
16994         gimple_omp_for_initial/final is ADDR_EXPR.
16996         PR middle-end/81768
16997         * omp-expand.c (expand_omp_simd): Force second operands of COND_EXPR
16998         into gimple val before gimplification fo the COND_EXPR.
17000 2017-09-05  Aldy Hernandez  <aldyh@redhat.com>
17002         * tree-ssa-threadupdate.c (duplicate_thread_path): Remove unused
17003         REGION_COPY argument.
17004         (thread_through_all_blocks): Remove unused argument to
17005         duplicate_thread_path.
17007 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17008             Alan Hayward  <alan.hayward@arm.com>
17009             David Sherwood  <david.sherwood@arm.com>
17011         * config/aarch64/aarch64-protos.h (aarch64_gen_adjusted_ldpstp):
17012         Take a scalar_mode rather than a machine_mode.
17013         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17014         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Likewise.
17015         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
17016         (aarch64_gen_adjusted_ldpstp): Likewise.
17017         (aarch64_expand_vector_init): Use scalar_mode instead of machine_mode.
17019 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17020             Alan Hayward  <alan.hayward@arm.com>
17021             David Sherwood  <david.sherwood@arm.com>
17023         * config/aarch64/aarch64-protos.h (aarch64_is_extend_from_extract):
17024         Take a scalar_int_mode instead of a machine_mode.
17025         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17026         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17027         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17028         (aarch64_simd_attr_length_rglist): Delete.
17029         * config/aarch64/aarch64.c (aarch64_is_extend_from_extract): Take
17030         a scalar_int_mode instead of a machine_mode.
17031         (aarch64_add_offset): Likewise.
17032         (aarch64_internal_mov_immediate): Likewise
17033         (aarch64_add_constant_internal): Likewise.
17034         (aarch64_add_constant): Likewise.
17035         (aarch64_movw_imm): Likewise.
17036         (aarch64_rtx_arith_op_extract_p): Likewise.
17037         (aarch64_mask_and_shift_for_ubfiz_p): Likewise.
17038         (aarch64_simd_scalar_immediate_valid_for_move): Likewise.
17039         Remove assert that the mode isn't a vector.
17040         (aarch64_output_scalar_simd_mov_immediate): Likewise.
17041         (aarch64_expand_mov_immediate): Update calls after above changes.
17042         (aarch64_output_casesi): Use as_a <scalar_int_mode>.
17043         (aarch64_and_bitmask_imm): Check for scalar integer modes.
17044         (aarch64_move_imm): Likewise.
17045         (aarch64_can_const_movi_rtx_p): Likewise.
17046         (aarch64_strip_extend): Likewise.
17047         (aarch64_extr_rtx_p): Likewise.
17048         (aarch64_rtx_costs): Likewise, using wode_mode as the mode of
17049         a CONST_INT when the mode parameter is VOIDmode.
17050         (aarch64_float_const_rtx_p): Use scalar_int_mode for a temporary.
17052 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17054         * machmode.h (bitwise_mode_for_mode): Return opt_mode.
17055         * stor-layout.c (bitwise_mode_for_mode): Likewise.
17056         (bitwise_type_for_mode): Update accordingly.
17058 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17060         * stor-layout.h (mode_for_size_tree): Return an opt_mode.
17061         * stor-layout.c (mode_for_size_tree): Likewise.
17062         (mode_for_array): Update accordingly.
17063         (layout_decl): Likewise.
17064         (compute_record_mode): Likewise.  Only set the mode once.
17066 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17068         * target.def (get_mask_mode): Change return type to opt_mode.
17069         Expand commentary.
17070         * doc/tm.texi: Regenerate.
17071         * targhooks.h (default_get_mask_mode): Return an opt_mode.
17072         * targhooks.c (default_get_mask_mode): Likewise.
17073         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
17074         * optabs-query.c (can_vec_mask_load_store_p): Update use of
17075         targetm.get_mask_mode.
17076         * tree.c (build_truth_vector_type): Likewise.
17078 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17080         * machmode.h (mode_for_vector): Return an opt_mode.
17081         * stor-layout.c (mode_for_vector): Likewise.
17082         (mode_for_int_vector): Update accordingly.
17083         (layout_type): Likewise.
17084         * config/i386/i386.c (emit_memmov): Likewise.
17085         (ix86_expand_set_or_movmem): Likewise.
17086         (ix86_expand_vector_init): Likewise.
17087         (ix86_get_mask_mode): Likewise.
17088         * config/powerpcspe/powerpcspe.c (rs6000_expand_vec_perm_const_1):
17089         Likewise.
17090         * config/rs6000/rs6000.c (rs6000_expand_vec_perm_const_1): Likewise.
17091         * expmed.c (extract_bit_field_1): Likewise.
17092         * expr.c (expand_expr_real_2): Likewise.
17093         * optabs-query.c (can_vec_perm_p): Likewise.
17094         (can_vec_mask_load_store_p): Likewise.
17095         * optabs.c (expand_vec_perm): Likewise.
17096         * targhooks.c (default_get_mask_mode): Likewise.
17097         * tree-vect-stmts.c (vectorizable_store): Likewise.
17098         (vectorizable_load): Likewise.
17099         (get_vectype_for_scalar_type_and_size): Likewise.
17101 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17103         * machmode.h (mode_for_int_vector): New function.
17104         * stor-layout.c (mode_for_int_vector): Likewise.
17105         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Use it.
17106         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Likewise.
17107         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
17108         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
17109         (s390_expand_vcond): Likewise.
17111 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17113         * machmode.h (opt_machine_mode): New type.
17114         (opt_mode<T>): Allow construction from anything that can be
17115         converted to a T.
17116         (is_a, as_a, dyn_cast): Add overloads for opt_mode.
17117         (mode_for_size): Return an opt_machine_mode.
17118         * stor-layout.c (mode_for_size): Likewise.
17119         (mode_for_size_tree): Update call accordingly.
17120         (bitwise_mode_for_mode): Likewise.
17121         (make_fract_type): Likewise.
17122         (make_accum_type): Likewise.
17123         * caller-save.c (replace_reg_with_saved_mem): Update call
17124         accordingly.
17125         * config/alpha/alpha.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17126         * config/i386/i386.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17127         * config/s390/s390.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17128         * config/sparc/sparc.h (SECONDARY_MEMORY_NEEDED_MODE): Likewise.
17129         * expmed.c (extract_bit_field_1): Likewise.
17130         * reload.c (get_secondary_mem): Likewise.
17131         * varasm.c (assemble_integer): Likewise.
17132         * lower-subreg.c (simplify_subreg_concatn): Likewise.  Move
17133         early-out.
17135 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17137         * machmode.h (decimal_float_mode_for_size): New function.
17138         * real.h (REAL_VALUE_TO_TARGET_LONG_DOUBLE): Use float_mode_for_size.
17139         (REAL_VALUE_TO_TARGET_DOUBLE): Likewise.
17140         (REAL_VALUE_TO_TARGET_SINGLE): Likewise.
17141         (REAL_VALUE_TO_TARGET_DECIMAL128): Use decimal_float_mode_for_size.
17142         (REAL_VALUE_TO_TARGET_DECIMAL64): Likewise.
17143         (REAL_VALUE_TO_TARGET_DECIMAL32): Likewise.
17145 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17147         * builtins.c (expand_builtin_powi): Use int_mode_for_size.
17148         (get_builtin_sync_mode): Likewise.
17149         (expand_ifn_atomic_compare_exchange): Likewise.
17150         (expand_builtin_atomic_clear): Likewise.
17151         (expand_builtin_atomic_test_and_set): Likewise.
17152         (fold_builtin_atomic_always_lock_free): Likewise.
17153         * calls.c (compute_argument_addresses): Likewise.
17154         (emit_library_call_value_1): Likewise.
17155         (store_one_arg): Likewise.
17156         * combine.c (combine_instructions): Likewise.
17157         * config/aarch64/aarch64.c (aarch64_function_value): Likewise.
17158         * config/arm/arm.c (arm_function_value): Likewise.
17159         (aapcs_allocate_return_reg): Likewise.
17160         * config/c6x/c6x.c (c6x_expand_movmem): Likewise.
17161         * config/i386/i386.c (construct_container): Likewise.
17162         (ix86_gimplify_va_arg): Likewise.
17163         (ix86_expand_sse_cmp): Likewise.
17164         (emit_memmov): Likewise.
17165         (emit_memset): Likewise.
17166         (expand_small_movmem_or_setmem): Likewise.
17167         (ix86_expand_pextr): Likewise.
17168         (ix86_expand_pinsr): Likewise.
17169         * config/lm32/lm32.c (lm32_block_move_inline): Likewise.
17170         * config/microblaze/microblaze.c (microblaze_block_move_straight):
17171         Likewise.
17172         * config/mips/mips.c (mips_function_value_1) Likewise.
17173         (mips_block_move_straight): Likewise.
17174         (mips_expand_ins_as_unaligned_store): Likewise.
17175         * config/powerpcspe/powerpcspe.c
17176         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17177         (rs6000_darwin64_record_arg_flush): Likewise.
17178         * config/rs6000/rs6000.c
17179         (rs6000_darwin64_record_arg_advance_flush): Likewise.
17180         (rs6000_darwin64_record_arg_flush): Likewise.
17181         * config/sparc/sparc.c (sparc_function_arg_1): Likewise.
17182         (sparc_function_value_1): Likewise.
17183         * config/spu/spu.c (adjust_operand): Likewise.
17184         (spu_emit_branch_or_set): Likewise.
17185         (arith_immediate_p): Likewise.
17186         * emit-rtl.c (gen_lowpart_common): Likewise.
17187         * expr.c (expand_expr_real_1): Likewise.
17188         * function.c (assign_parm_setup_block): Likewise.
17189         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): Likewise.
17190         * reload1.c (alter_reg): Likewise.
17191         * stor-layout.c (mode_for_vector): Likewise.
17192         (layout_type): Likewise.
17194 2017-09-05  Richard Sandiford  <richard.sandiford@linaro.org>
17196         * config/spu/spu.c (exp2_immediate_p): Use int_mode_for_mode.
17197         (spu_convert_move): Likewise.
17198         * lower-subreg.c (resolve_simple_move): Likewise.
17200 2017-09-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
17202         PR target/81833
17203         * config/rs6000/altivec.md (altivec_vsum2sws): Convert from a
17204         define_insn to a define_expand.
17205         (altivec_vsum2sws_direct): New define_insn.
17206         (altivec_vsumsws): Convert from a define_insn to a define_expand.
17208 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17210         * config/arm/arm.c (arm_option_params_internal): Improve setting of
17211         max_insns_skipped.
17213 2017-09-05  H.J. Lu  <hongjiu.lu@intel.com>
17215         PR target/59501
17216         PR target/81624
17217         PR target/81769
17218         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Don't
17219         realign stack if stack alignment needed is less than incoming
17220         stack boundary.
17222 2017-09-05  Marek Polacek  <polacek@redhat.com>
17224         PR sanitizer/82072
17225         * convert.c (convert_to_integer_1) <case NEGATE_EXPR>: Move the ubsan
17226         check earlier.
17228 2017-09-05  Wilco Dijkstra  <wdijkstr@arm.com>
17230         * explow.c (get_dynamic_stack_size): Improve dynamic alignment.
17232 2017-09-05  Richard Biener  <rguenther@suse.de>
17234         PR tree-optimization/82084
17235         * fold-const.c (can_native_encode_string_p): Handle wide characters.
17237 2017-09-05  Richard Biener  <rguenther@suse.de>
17239         PR tree-optimization/82102
17240         * tree-ssa-pre.c (fini_eliminate): Check if lhs is NULL.
17242 2017-09-05  Martin Liska  <mliska@suse.cz>
17244         PR tree-optimization/82032
17245         * tree-cfg.c (generate_range_test): New function.
17246         * tree-cfg.h (generate_range_test): Declared here.
17247         * tree-cfgcleanup.c (convert_single_case_switch): New function.
17248         (cleanup_control_expr_graph): Use it.
17249         * tree-switch-conversion.c (try_switch_expansion): Remove
17250         assert.
17251         (emit_case_nodes): Use generate_range_test.
17253 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17255         PR target/82098
17256         * config/i386/i386.md (*<btsc><mode>_mask): Add
17257         TARGET_USE_BT to insn constraint.
17258         (*btr<mode>_mask): Ditto.
17260 2017-09-04  Wilco Dijkstra  <wdijkstr@arm.com>
17262         * config/arm/arm.c (arm_legitimate_index_p): Add comment.
17263         (thumb2_legitimate_index_p): Use correct range for DI/DF mode.
17265 2017-09-04  Bernd Edlinger  <bernd.edlinger@hotmail.de>
17267         PR target/77308
17268         * config/arm/arm.md (*arm_adddi3, *arm_subdi3): Split early except for
17269         TARGET_NEON and TARGET_IWMMXT.
17270         (anddi3, iordi3, xordi3, one_cmpldi2): Split while expanding except for
17271         TARGET_NEON and TARGET_IWMMXT.
17272         (*one_cmpldi2_insn): Moved the body of one_cmpldi2 here.
17274 2017-09-04  Uros Bizjak  <ubizjak@gmail.com>
17276         * config/i386/i386-protos.h (ix86_tls_address_pattern_p) New prototype.
17277         (ix86_rewrite_tls_address): Ditto.
17278         * config/i386/i386.c (ix86_tls_address_pattern_p) New function.
17279         (ix86_rewrite_tls_address_1): Ditto.
17280         (ix86_rewrite_tls_address): Ditto.
17281         * config/i386/predicates.md (tls_address_pattern): New predicate.
17282         * config/i386/i386.md (TLS address splitter): New splitter.
17284 2017-09-04  Richard Biener  <rguenther@suse.de>
17286         PR tree-optimization/82084
17287         * fold-const.h (can_native_encode_string_p): Declare.
17288         * fold-const.c (can_native_encode_string_p): Factor out from ...
17289         (native_encode_string): ... here.
17290         * tree-vect-stmts.c (vectorizable_store): Call it to avoid
17291         vectorizing stores from constants we later cannot handle.
17293 2017-09-04  Marek Polacek  <polacek@redhat.com>
17295         PR c/81783
17296         * doc/invoke.texi: Update -Wtautological-compare documentation.
17298 2017-09-04  Jeff Law  <law@redhat.com>
17300         PR tree-optimization/64910
17301         * tree-ssa-reassoc.c (reassociate_bb): For bitwise binary ops,
17302         swap the first and last operand if the last is a constant.
17304 2017-09-04  Marek Polacek  <polacek@redhat.com>
17306         PR sanitizer/82072
17307         * convert.c (do_narrow): When sanitizing signed integer overflows,
17308         bail out for signed types.
17309         (convert_to_integer_1) <case NEGATE_EXPR>: Likewise.
17311 2017-09-04  Richard Biener  <rguenther@suse.de>
17313         PR tree-optimization/82060
17314         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
17315         Move devirtualization after stmt folding and before EH/AB/noreturn
17316         cleanup to get the stmt refs canonicalized.  Use a bool instead
17317         of gimple_modified_p since that doesn't work for NOPs.  Schedule
17318         NOPs generated by folding for removal.
17320 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17321             Alan Hayward  <alan.hayward@arm.com>
17322             David Sherwood  <david.sherwood@arm.com>
17324         * coretypes.h (pad_direction): New enum.
17325         * defaults.h (DEFAULT_FUNCTION_ARG_PADDING): Delete.
17326         (FUNCTION_ARG_PADDING): Likewise.
17327         * target.def (function_arg_padding): New hook.
17328         * targhooks.h (default_function_arg_padding): Declare.
17329         * targhooks.c (default_function_arg_padding): New function.
17330         * doc/tm.texi.in (FUNCTION_ARG_PADDING): Replace with...
17331         (TARGET_FUNCTION_ARG_PADDING): ...this.
17332         * doc/tm.texi: Regenerate.
17333         * calls.c (store_unaligned_arguments_into_pseudos): Use pad_direction
17334         instead of direction.
17335         (compute_argument_addresses): Likewise.
17336         (load_register_parameters): Likewise.
17337         (emit_library_call_value_1): Likewise.
17338         (store_one_arg): Use targetm.calls.function_arg_padding instead
17339         of FUNCTION_ARG_PADDING.
17340         (must_pass_in_stack_var_size_or_pad): Likewise.
17341         * expr.c (emit_group_load_1): Use pad_direction instead of direction.
17342         (emit_group_store): Likewise.
17343         (emit_single_push_insn_1): Use targetm.calls.function_arg_padding
17344         instead of FUNCTION_ARG_PADDING.
17345         (emit_push_insn): Likewise, and propagate enum change throughout
17346         function.
17347         * function.h (direction): Delete.
17348         (locate_and_pad_arg_data::where_pad): Use pad_direction instead
17349         of direction.
17350         * function.c (assign_parm_find_stack_rtl): Likewise.
17351         (assign_parm_setup_block_p): Likewise.
17352         (assign_parm_setup_block): Likewise.
17353         (gimplify_parameters): Likewise.
17354         (locate_and_pad_parm): Use targetm.calls.function_arg_padding
17355         instead of FUNCTION_ARG_PADDING, and propagate enum change throughout
17356         function.
17357         * config/aarch64/aarch64.h (FUNCTION_ARG_PADDING): Delete.
17358         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17359         * config/aarch64/aarch64-protos.h (aarch64_pad_arg_upward): Delete.
17360         * config/aarch64/aarch64.c (aarch64_pad_arg_upward): Replace with...
17361         (aarch64_function_arg_padding): ...this new function.
17362         (aarch64_gimplify_va_arg_expr): Use pad_direction instead of direction.
17363         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17364         * config/arm/arm.h (FUNCTION_ARG_PADDING): Delete.
17365         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17366         * config/arm/arm-protos.h (arm_pad_arg_upward): Delete.
17367         * config/arm/arm.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17368         (arm_pad_arg_upward): Replace with...
17369         (arm_function_arg_padding): ...this new function.
17370         * config/c6x/c6x.h (BLOCK_REG_PADDING): Use pad_direction instead
17371         of direction.
17372         * config/ia64/hpux.h (FUNCTION_ARG_PADDING): Delete.
17373         * config/ia64/ia64-protos.h (ia64_hpux_function_arg_padding): Delete.
17374         * config/ia64/ia64.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17375         (ia64_hpux_function_arg_padding): Replace with...
17376         (ia64_function_arg_padding): ...this new function.  Use pad_direction
17377         instead of direction.  Check for TARGET_HPUX.
17378         * config/iq2000/iq2000.h (FUNCTION_ARG_PADDING): Delete.
17379         * config/iq2000/iq2000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17380         (iq2000_function_arg_padding): New function.
17381         * config/mips/mips-protos.h (mips_pad_arg_upward): Delete.
17382         * config/mips/mips.c (mips_pad_arg_upward): Replace with...
17383         (mips_function_arg_padding): ...this new function.
17384         (mips_pad_reg_upward): Update accordingly.
17385         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17386         * config/mips/mips.h (PAD_VARARGS_DOWN): Use
17387         targetm.calls.function_arg_padding.
17388         (FUNCTION_ARG_PADDING): Delete.
17389         (BLOCK_REG_PADDING): Use pad_direction instead of direction.
17390         * config/nios2/nios2.h (FUNCTION_ARG_PADDING): Delete.
17391         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17392         * config/nios2/nios2-protos.h (nios2_function_arg_padding): Delete.
17393         (nios2_block_reg_padding): Return pad_direction instead of direction.
17394         * config/nios2/nios2.c (nios2_block_reg_padding): Return pad_direction
17395         instead of direction.
17396         (nios2_function_arg_padding): Likewise.  Make static.
17397         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17398         * config/pa/pa.h (FUNCTION_ARG_PADDING): Delete.
17399         (BLOCK_REG_PADDING): Use targetm.calls.function_arg_padding.
17400         * config/pa/pa-protos.h (pa_function_arg_padding): Delete.
17401         * config/pa/pa.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17402         (pa_function_arg_padding): Make static.  Return pad_direction instead
17403         of direction.
17404         * config/powerpcspe/powerpcspe.h (FUNCTION_ARG_PADDING): Delete.
17405         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17406         * config/powerpcspe/aix.h (BLOCK_REG_PADDING): Use pad_direction
17407         instead of direction.  Use targetm.calls.function_arg_padding.
17408         * config/powerpcspe/darwin.h (BLOCK_REG_PADDING): Likewise.
17409         * config/powerpcspe/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17410         * config/powerpcspe/linux64.h (BLOCK_REG_PADDING): Likewise.
17411         * config/powerpcspe/powerpcspe-protos.h (function_arg_padding): Delete.
17412         * config/powerpcspe/powerpcspe.c (TARGET_FUNCTION_ARG_PADDING):
17413         Redefine.
17414         (function_arg_padding): Rename to...
17415         (rs6000_function_arg_padding): ...this.  Make static.  Return
17416         pad_direction instead of direction.
17417         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17418         * config/rs6000/rs6000.h (FUNCTION_ARG_PADDING): Delete.
17419         (PAD_VARARGS_DOWN): Use targetm.calls.function_arg_padding.
17420         * config/rs6000/aix.h (BLOCK_REG_PADDING): Use pad_direction
17421         instead of direction.  Use targetm.calls.function_arg_padding.
17422         * config/rs6000/darwin.h (BLOCK_REG_PADDING): Likewise.
17423         * config/rs6000/freebsd64.h (BLOCK_REG_PADDING): Likewise.
17424         * config/rs6000/linux64.h (BLOCK_REG_PADDING): Likewise.
17425         * config/rs6000/rs6000-protos.h (function_arg_padding): Delete.
17426         * config/rs6000/rs6000.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17427         (function_arg_padding): Rename to...
17428         (rs6000_function_arg_padding): ...this.  Make static.  Return
17429         pad_direction instead of direction.
17430         (rs6000_return_in_msb): Use rs6000_function_arg_padding.
17431         * config/s390/s390.h (FUNCTION_ARG_PADDING): Delete.
17432         * config/s390/s390.c (s390_function_arg_padding): New function.
17433         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17434         * config/sparc/sparc.h (FUNCTION_ARG_PADDING): Delete.
17435         * config/sparc/sparc-protos.h (function_arg_padding): Delete.
17436         * config/sparc/sparc.c (TARGET_FUNCTION_ARG_PADDING): Redefine.
17437         (function_arg_padding): Rename to...
17438         (sparc_function_arg_padding): ...this.  Make static.  Return
17439         pad_direction instead of direction.
17440         * config/spu/spu.h (FUNCTION_ARG_PADDING): Delete.
17441         * config/spu/spu.c (spu_function_arg_padding): New function.
17442         (TARGET_FUNCTION_ARG_PADDING): Redefine.
17443         * system.h (FUNCTION_ARG_PADDING): Poison.
17445 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17446             Alan Hayward  <alan.hayward@arm.com>
17447             David Sherwood  <david.sherwood@arm.com>
17449         * target.def (modes_tieable_p): New hook.
17450         * doc/tm.texi (MODES_TIEABLE_P): Replace with...
17451         (TARGET_MODES_TIEABLE_P): ...this.
17452         * doc/tm.texi.in: Regenerate.
17453         * hooks.h (hook_bool_mode_mode_true): Declare.
17454         * hooks.c (hook_bool_mode_mode_true): New function.
17455         * combine.c (subst): Use targetm.modes_tieable_p instead of
17456         MODES_TIEABLE_P.
17457         * dse.c (find_shift_sequence): Likewise.
17458         * expmed.c (extract_low_bits): Likewise.
17459         * lower-subreg.c: Include target.h.
17460         (find_decomposable_subregs): Use targetm.modes_tieable_p instead of
17461         MODES_TIEABLE_P.
17462         * rtlanal.c (rtx_cost): Likewise.
17463         * config/aarch64/aarch64.h (MODES_TIEABLE_P): Delete.
17464         * config/aarch64/aarch64-protos.h (aarch64_modes_tieable_p): Delete.
17465         * config/aarch64/aarch64.c (aarch64_modes_tieable_p): Make static.
17466         (TARGET_MODES_TIEABLE_P): Redefine.
17467         * config/alpha/alpha.h (MODES_TIEABLE_P): Delete.
17468         * config/alpha/alpha.c (alpha_modes_tieable_p): New function.
17469         (TARGET_MODES_TIEABLE_P): Redefine.
17470         * config/arc/arc.h (MODES_TIEABLE_P): Delete.
17471         * config/arc/arc.c (TARGET_MODES_TIEABLE_P): Redefine.
17472         (arc_modes_tieable_p): New function.
17473         * config/arm/arm.h (MODES_TIEABLE_P): Delete.
17474         * config/arm/arm-protos.h (arm_modes_tieable_p): Delete.
17475         * config/arm/arm.c (TARGET_MODES_TIEABLE_P): Redefine.
17476         (arm_modes_tieable_p): Make static.
17477         * config/avr/avr.h (MODES_TIEABLE_P): Delete.
17478         * config/bfin/bfin.h (MODES_TIEABLE_P): Delete.
17479         * config/bfin/bfin.c (bfin_modes_tieable_p): New function.
17480         (TARGET_MODES_TIEABLE_P): Redefine.
17481         * config/c6x/c6x.h (MODES_TIEABLE_P): Delete.
17482         * config/c6x/c6x.c (c6x_modes_tieable_p): New function.
17483         (TARGET_MODES_TIEABLE_P): Redefine.
17484         * config/cr16/cr16.h (MODES_TIEABLE_P): Delete.
17485         * config/cr16/cr16.c (TARGET_MODES_TIEABLE_P): Redefine.
17486         (cr16_modes_tieable_p): New function.
17487         * config/cris/cris.h (MODES_TIEABLE_P): Delete.
17488         * config/epiphany/epiphany.h (MODES_TIEABLE_P): Delete.
17489         * config/fr30/fr30.h (MODES_TIEABLE_P): Delete.
17490         (TRULY_NOOP_TRUNCATION): Update comment.
17491         * config/frv/frv.h (MODES_TIEABLE_P): Delete.
17492         (TRULY_NOOP_TRUNCATION): Update comment.
17493         * config/frv/frv.c (TARGET_MODES_TIEABLE_P): Redefine.
17494         (frv_modes_tieable_p): New function.
17495         * config/ft32/ft32.h (MODES_TIEABLE_P): Delete.
17496         * config/h8300/h8300.h (MODES_TIEABLE_P): Delete.
17497         * config/h8300/h8300.c (h8300_modes_tieable_p): New function.
17498         (TARGET_MODES_TIEABLE_P): Redefine.
17499         * config/i386/i386.h (MODES_TIEABLE_P): Delete.
17500         * config/i386/i386-protos.h (ix86_modes_tieable_p): Delete.
17501         * config/i386/i386.c (ix86_modes_tieable_p): Make static.
17502         (TARGET_MODES_TIEABLE_P): Redefine.
17503         * config/ia64/ia64.h (MODES_TIEABLE_P): Delete.
17504         * config/ia64/ia64.c (TARGET_MODES_TIEABLE_P): Redefine.
17505         (ia64_modes_tieable_p): New function.
17506         * config/iq2000/iq2000.h (MODES_TIEABLE_P): Delete.
17507         * config/iq2000/iq2000.c (TARGET_MODES_TIEABLE_P): Redefine.
17508         (iq2000_modes_tieable_p): New function.
17509         * config/lm32/lm32.h (MODES_TIEABLE_P): Delete.
17510         * config/lm32/lm32.c (TARGET_MODES_TIEABLE_P): Redefine.
17511         (lm32_modes_tieable_p): New function.
17512         * config/m32c/m32c.h (MODES_TIEABLE_P): Delete.
17513         * config/m32c/m32c-protos.h (m32c_modes_tieable_p): Delete.
17514         * config/m32c/m32c.c (m32c_modes_tieable_p): Make static.
17515         (TARGET_MODES_TIEABLE_P): Redefine.
17516         * config/m32r/m32r.h (MODES_TIEABLE_P): Delete.
17517         * config/m32r/m32r.c (TARGET_MODES_TIEABLE_P): Redefine.
17518         (m32r_modes_tieable_p): New function.
17519         * config/m68k/m68k.h (MODES_TIEABLE_P): Delete.
17520         * config/m68k/m68k.c (TARGET_MODES_TIEABLE_P): Redefine.
17521         (m68k_modes_tieable_p): New function.
17522         * config/mcore/mcore.h (MODES_TIEABLE_P): Delete.
17523         * config/mcore/mcore.c (TARGET_MODES_TIEABLE_P): Redefine.
17524         (mcore_modes_tieable_p): New function.
17525         * config/microblaze/microblaze.h (MODES_TIEABLE_P): Delete.
17526         * config/microblaze/microblaze.c (microblaze_modes_tieable_p): New
17527         function.
17528         (TARGET_MODES_TIEABLE_P): Redefine.
17529         * config/mips/mips.h (MODES_TIEABLE_P): Delete.
17530         * config/mips/mips-protos.h (mips_modes_tieable_p): Delete.
17531         * config/mips/mips.c (mips_modes_tieable_p): Make static.
17532         (TARGET_MODES_TIEABLE_P): Redefine.
17533         * config/mmix/mmix.h (MODES_TIEABLE_P): Delete.
17534         * config/mn10300/mn10300.h (MODES_TIEABLE_P): Delete.
17535         * config/mn10300/mn10300-protos.h (mn10300_modes_tieable): Delete.
17536         * config/mn10300/mn10300.c (mn10300_modes_tieable): Rename to...
17537         (mn10300_modes_tieable_p): ...this and make static.
17538         (TARGET_MODES_TIEABLE_P): Redefine.
17539         * config/moxie/moxie.h (MODES_TIEABLE_P): Delete.
17540         * config/msp430/msp430.h (MODES_TIEABLE_P): Delete.
17541         * config/msp430/msp430-protos.h (msp430_modes_tieable_p): Delete.
17542         * config/msp430/msp430.c (TARGET_MODES_TIEABLE_P): Redefine.
17543         (msp430_modes_tieable_p): Make static.
17544         * config/nds32/nds32.h (MODES_TIEABLE_P): Delete.
17545         * config/nds32/nds32.c (nds32_modes_tieable_p): New function.
17546         (TARGET_MODES_TIEABLE_P): Redefine.
17547         * config/nios2/nios2.h (MODES_TIEABLE_P): Delete.
17548         * config/nvptx/nvptx.h (MODES_TIEABLE_P): Delete.
17549         * config/nvptx/nvptx.c (nvptx_modes_tieable_p): New function.
17550         (TARGET_MODES_TIEABLE_P): Redefine.
17551         * config/pa/pa.h (MODES_TIEABLE_P): Delete.
17552         * config/pa/pa-protos.h (pa_modes_tieable_p): Delete.
17553         * config/pa/pa.c (pa_modes_tieable_p): Make static.
17554         (TARGET_MODES_TIEABLE_P): Redefine.
17555         * config/pdp11/pdp11.h (MODES_TIEABLE_P): Delete.
17556         * config/pdp11/pdp11.c (TARGET_MODES_TIEABLE_P): Redefine.
17557         (pdp11_modes_tieable_p): New function.
17558         * config/powerpcspe/powerpcspe.h (MODES_TIEABLE_P): Delete.
17559         * config/powerpcspe/powerpcspe.c (TARGET_MODES_TIEABLE_P): Redefine.
17560         (rs6000_modes_tieable_p): New function.
17561         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17562         * config/powerpcspe/powerpcspe.md: Update comment.
17563         * config/riscv/riscv.h (MODES_TIEABLE_P): Delete.
17564         * config/riscv/riscv.c (riscv_modes_tieable_p): New function.
17565         (TARGET_MODES_TIEABLE_P): Redefine.
17566         * config/rl78/rl78.h (MODES_TIEABLE_P): Delete.
17567         * config/rl78/rl78.c (TARGET_MODES_TIEABLE_P): Redefine.
17568         (rl78_modes_tieable_p): New function.
17569         * config/rs6000/rs6000.h (MODES_TIEABLE_P): Delete.
17570         * config/rs6000/rs6000.c (TARGET_MODES_TIEABLE_P): Redefine.
17571         (rs6000_modes_tieable_p): New function.
17572         (rs6000_debug_reg_global): Use it instead of MODES_TIEABLE_P.
17573         * config/rs6000/rs6000.md: Update comment.
17574         * config/rx/rx.h (MODES_TIEABLE_P): Delete.
17575         * config/rx/rx.c (rx_modes_tieable_p): New function.
17576         (TARGET_MODES_TIEABLE_P): Redefine.
17577         * config/s390/s390.h (MODES_TIEABLE_P): Delete.
17578         * config/s390/s390.c (s390_modes_tieable_p): New function.
17579         (TARGET_MODES_TIEABLE_P): Redefine.
17580         * config/sh/sh.h (MODES_TIEABLE_P): Delete.
17581         * config/sh/sh.c (TARGET_MODES_TIEABLE_P): Redefine.
17582         (sh_modes_tieable_p): New function.
17583         * config/sparc/sparc.h (MODES_TIEABLE_P): Delete.
17584         * config/sparc/sparc-protos.h (sparc_modes_tieable_p): Delete.
17585         * config/sparc/sparc.c (TARGET_MODES_TIEABLE_P): Redefine.
17586         (sparc_modes_tieable_p): Make static.
17587         * config/spu/spu.h (MODES_TIEABLE_P): Delete.
17588         * config/spu/spu.c (spu_modes_tieable_p): New function.
17589         (TARGET_MODES_TIEABLE_P): Redefine.
17590         * config/stormy16/stormy16.h (MODES_TIEABLE_P): Delete.
17591         * config/stormy16/stormy16.c (xstormy16_modes_tieable_p): New function.
17592         (TARGET_MODES_TIEABLE_P): Redefine.
17593         * config/tilegx/tilegx.h (MODES_TIEABLE_P): Delete.
17594         * config/tilepro/tilepro.h (MODES_TIEABLE_P): Delete.
17595         * config/v850/v850.h (MODES_TIEABLE_P): Delete.
17596         * config/v850/v850.c (v850_modes_tieable_p): New function.
17597         (TARGET_MODES_TIEABLE_P): Redefine.
17598         * config/vax/vax.h (MODES_TIEABLE_P): Delete.
17599         * config/visium/visium.h (MODES_TIEABLE_P): Delete.
17600         * config/visium/visium.c (TARGET_MODES_TIEABLE_P): Redefine.
17601         (visium_modes_tieable_p): New function.
17602         * config/xtensa/xtensa.h (MODES_TIEABLE_P): Delete.
17603         * config/xtensa/xtensa.c (TARGET_MODES_TIEABLE_P): Redefine.
17604         (xtensa_modes_tieable_p): New function.
17605         * system.h (MODES_TIEABLE_P): Poison.
17607 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17608             Alan Hayward  <alan.hayward@arm.com>
17609             David Sherwood  <david.sherwood@arm.com>
17611         * target.def (hard_regno_mode_ok): New hook.
17612         * doc/tm.texi (HARD_REGNO_MODE_OK): Replace with...
17613         (TARGET_HARD_REGNO_MODE_OK): ...this.
17614         * doc/tm.texi.in: Regenerate.
17615         * hooks.h (hook_bool_uint_mode_true): Declare.
17616         * hooks.c (hook_bool_uint_mode_true): New function.
17617         * doc/md.texi: Refer to targetm.hard_regno_mode_ok instead of
17618         HARD_REGNO_MODE_OK.
17619         * genpreds.c (write_insn_preds_c): Add an include of target.h.
17620         * alias.c (init_alias_target): Use targetm.hard_regno_mode_ok
17621         instead of HARD_REGNO_MODE_OK.
17622         * caller-save.c: Include target.h.
17623         (reg_save_code): Use targetm.hard_regno_mode_ok instead of
17624         HARD_REGNO_MODE_OK.
17625         * combine.c (can_combine_p): Likewise.
17626         (combinable_i3pat): Likewise.
17627         (can_change_dest_mode): Likewise.
17628         * expr.c (init_expr_target): Likewise.
17629         (convert_move): Likewise.
17630         (convert_modes): Likewise.
17631         * ira.c (setup_prohibited_class_mode_regs): Likewise.
17632         (setup_prohibited_mode_move_regs): Likewise.
17633         * ira.h (target_ira): Likewise.
17634         * lra-assigns.c (find_hard_regno_for_1): Likewise.
17635         * lra-constraints.c (process_alt_operands): Likewise.
17636         (split_reg): Likewise.
17637         * recog.c (peep2_find_free_register): Likewise.
17638         * ree.c (combine_reaching_defs): Likewise.
17639         * regcprop.c (maybe_mode_change): Likewise.
17640         * reginfo.c (init_reg_sets_1): Likewise.
17641         (choose_hard_reg_mode): Likewise.
17642         (simplifiable_subregs): Likewise.
17643         * regrename.c (check_new_reg_p): Likewise.
17644         * reload.c (find_valid_class): Likewise.
17645         (find_valid_class_1): Likewise.
17646         (reload_inner_reg_of_subreg): Likewise.
17647         (push_reload): Likewise.
17648         (combine_reloads): Likewise.
17649         (find_dummy_reload): Likewise.
17650         (find_reloads): Likewise.
17651         * reload1.c (find_reg): Likewise.
17652         (set_reload_reg): Likewise.
17653         (allocate_reload_reg): Likewise.
17654         (choose_reload_regs): Likewise.
17655         (reload_adjust_reg_for_temp): Likewise.
17656         * rtlanal.c (subreg_size_offset_from_lsb): Likewise.
17657         (simplify_subreg_regno): Likewise.
17658         * sel-sched.c (init_regs_for_mode): Likewise.
17659         * varasm.c (make_decl_rtl): Likewise.
17660         * config/aarch64/aarch64.h (HARD_REGNO_MODE_OK): Delete.
17661         (MODES_TIEABLE_P): Use targetm.hard_regno_mode_ok instead of
17662         HARD_REGNO_MODE_OK.
17663         * config/aarch64/aarch64-protos.h (aarch64_hard_regno_mode_ok): Delete.
17664         * config/aarch64/aarch64.c (aarch64_hard_regno_mode_ok): Make static.
17665         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17666         * config/alpha/alpha.h (HARD_REGNO_MODE_OK): Delete.
17667         * config/alpha/alpha.c (alpha_hard_regno_mode_ok): New function.
17668         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17669         * config/arc/arc.h (arc_hard_regno_mode_ok): Delete.
17670         (arc_mode_class): Delete.
17671         (HARD_REGNO_MODE_OK): Delete.
17672         * config/arc/arc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17673         (arc_hard_regno_mode_ok): Rename old array to...
17674         (arc_hard_regno_mode_ok_modes): ...this.
17675         (arc_conditional_register_usage): Update accordingly.
17676         (arc_mode_class): Make static.
17677         (arc_hard_regno_mode_ok): New function.
17678         * config/arm/arm.h (HARD_REGNO_MODE_OK): Delete.
17679         * config/arm/arm-protos.h (arm_hard_regno_mode_ok): Delete.
17680         * config/arm/arm.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17681         (arm_hard_regno_mode_ok): Make static.
17682         * config/arm/arm.md (movdi): Use targetm.hard_regno_mode_ok instead of
17683         HARD_REGNO_MODE_OK.
17684         * config/avr/avr-protos.h (avr_hard_regno_mode_ok): Delete.
17685         * config/avr/avr.h (HARD_REGNO_MODE_OK): Delete.
17686         * config/avr/avr.c (avr_hard_regno_mode_ok): Make static and
17687         return a bool.
17688         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17689         * config/bfin/bfin-protos.h (hard_regno_mode_ok): Delete.
17690         * config/bfin/bfin.h (HARD_REGNO_MODE_OK): Delete.
17691         * config/bfin/bfin.c (hard_regno_mode_ok): Rename to...
17692         (bfin_hard_regno_mode_ok): ...this.  Make static and return a bool.
17693         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17694         * config/bfin/predicates.md (valid_reg_operand): Use
17695         targetm.hard_regno_mode_ok instead of HARD_REGNO_MODE_OK.
17696         * config/c6x/c6x.h (HARD_REGNO_MODE_OK): Delete.
17697         * config/c6x/c6x.c (c6x_hard_regno_mode_ok): New function.
17698         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17699         * config/cr16/cr16.h (HARD_REGNO_MODE_OK): Delete.
17700         * config/cr16/cr16-protos.h (cr16_hard_regno_mode_ok): Delete.
17701         * config/cr16/cr16.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17702         (cr16_hard_regno_mode_ok): Make static and return a bool.
17703         * config/cris/cris.h (HARD_REGNO_MODE_OK): Delete.
17704         * config/cris/cris.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17705         (cris_hard_regno_mode_ok): New function.
17706         * config/epiphany/epiphany.h (epiphany_hard_regno_mode_ok): Delete.
17707         (epiphany_mode_class): Delete.
17708         (HARD_REGNO_MODE_OK): Delete.
17709         * config/epiphany/epiphany-protos.h (hard_regno_mode_ok): Delete.
17710         * config/epiphany/epiphany.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17711         (hard_regno_mode_ok): Rename to...
17712         (epiphany_hard_regno_mode_ok): ...this.  Make static and return a bool.
17713         * config/fr30/fr30.h (HARD_REGNO_MODE_OK): Delete.
17714         * config/fr30/fr30.md: Refer to targetm.hard_regno_mode_ok instead of
17715         HARD_REGNO_MODE_OK.
17716         * config/frv/frv.h (HARD_REGNO_MODE_OK): Delete.
17717         * config/frv/frv-protos.h (frv_hard_regno_mode_ok): Delete.
17718         * config/frv/frv.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17719         (frv_hard_regno_mode_ok): Make static and return a bool.
17720         * config/frv/frv.md: Refer to targetm.hard_regno_mode_ok instead of
17721         HARD_REGNO_MODE_OK.
17722         * config/ft32/ft32.h (HARD_REGNO_MODE_OK): Delete.
17723         * config/h8300/h8300.h (HARD_REGNO_MODE_OK): Delete.
17724         * config/h8300/h8300-protos.h (h8300_hard_regno_mode_ok): Delete.
17725         * config/h8300/h8300.c (h8300_hard_regno_mode_ok): Make static
17726         and return a bool.
17727         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17728         * config/i386/i386.h (HARD_REGNO_MODE_OK): Delete.
17729         * config/i386/i386-protos.h (ix86_hard_regno_mode_ok): Delete.
17730         * config/i386/i386.c (ix86_hard_regno_mode_ok): Make static and
17731         return a bool.
17732         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17733         * config/ia64/ia64.h (HARD_REGNO_MODE_OK): Delete.
17734         * config/ia64/ia64.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17735         (ia64_hard_regno_mode_ok): New function.
17736         * config/iq2000/iq2000.h (HARD_REGNO_MODE_OK): Delete.
17737         * config/iq2000/iq2000.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17738         (iq2000_hard_regno_mode_ok): New function.
17739         * config/lm32/lm32.h (HARD_REGNO_MODE_OK): Delete.
17740         * config/lm32/lm32.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17741         (lm32_hard_regno_mode_ok): New function.
17742         * config/m32c/m32c.h (HARD_REGNO_MODE_OK): Delete.
17743         * config/m32c/m32c-protos.h (m32c_hard_regno_ok): Delete.
17744         * config/m32c/m32c.c (class_can_hold_mode): Use m32c_hard_regno_mode_ok
17745         instead of HARD_REGNO_MODE_OK.
17746         (m32c_hard_regno_ok): Rename to...
17747         (m32c_hard_regno_mode_ok): ...this.  Make static and return a bool.
17748         (m32c_cannot_change_mode_class): Update accordingly.
17749         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17750         * config/m32r/m32r.h (m32r_hard_regno_mode_ok): Delete.
17751         (m32r_mode_class): Delete.
17752         (HARD_REGNO_MODE_OK): Delete.
17753         * config/m32r/m32r.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17754         (m32r_hard_regno_mode_ok): Rename to...
17755         (m32r_hard_regno_modes): ...this.
17756         (m32r_mode_class): Make static.
17757         (m32r_hard_regno_mode_ok): New function.
17758         * config/m68k/m68k.h (HARD_REGNO_MODE_OK): Delete.
17759         * config/m68k/m68k-protos.h (m68k_regno_mode_ok): Delete.
17760         * config/m68k/m68k.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17761         (m68k_hard_regno_mode_ok): Make static.
17762         * config/mcore/mcore.h (HARD_REGNO_MODE_OK): Delete.
17763         * config/mcore/mcore.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17764         (mcore_hard_regno_mode_ok): New function.
17765         * config/microblaze/microblaze.h (microblaze_hard_regno_mode_ok)
17766         (HARD_REGNO_MODE_OK): Delete.
17767         * config/microblaze/microblaze.c (microblaze_hard_regno_mode_ok):
17768         Rename to...
17769         (microblaze_hard_regno_mode_ok_p): ...this and make static.
17770         (microblaze_hard_regno_mode_ok): New function.
17771         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17772         * config/mips/mips.h (HARD_REGNO_MODE_OK): Delete.
17773         (mips_hard_regno_mode_ok): Delete.
17774         * config/mips/mips.c (mips_hard_regno_mode_ok): Rename to...
17775         (mips_hard_regno_mode_ok_p): ...this and make static.
17776         (mips_hard_regno_mode_ok_p): Rename to...
17777         (mips_hard_regno_mode_ok_uncached): ...this.
17778         (mips_hard_regno_mode_ok): New function.
17779         (mips_class_max_nregs): Use mips_hard_regno_mode_ok instead
17780         of HARD_REGNO_MODE_OK.
17781         (mips_option_override): Update after above name changes.
17782         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17783         * config/mmix/mmix.h (HARD_REGNO_MODE_OK): Delete.
17784         * config/mn10300/mn10300.h (HARD_REGNO_MODE_OK): Delete.
17785         * config/mn10300/mn10300-protos.h (mn10300_hard_regno_mode_ok): Delete.
17786         * config/mn10300/mn10300.c (mn10300_hard_regno_mode_ok): Make static.
17787         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17788         * config/moxie/moxie.h (HARD_REGNO_MODE_OK): Delete.
17789         * config/msp430/msp430.h (HARD_REGNO_MODE_OK): Delete.
17790         * config/msp430/msp430-protos.h (msp430_hard_regno_mode_ok): Delete.
17791         * config/msp430/msp430.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17792         (msp430_hard_regno_mode_ok): Make static and return a bool.
17793         * config/nds32/nds32.h (HARD_REGNO_MODE_OK): Delete.
17794         * config/nds32/nds32-protos.h (nds32_hard_regno_mode_ok): Delete.
17795         * config/nds32/nds32.c (nds32_hard_regno_mode_ok): Make static
17796         and return a bool.
17797         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17798         * config/nios2/nios2.h (HARD_REGNO_MODE_OK): Delete.
17799         * config/nvptx/nvptx.h (HARD_REGNO_MODE_OK): Delete.
17800         * config/pa/pa.h (MODES_TIEABLE_P): Update commentary.
17801         * config/pa/pa32-regs.h (HARD_REGNO_MODE_OK): Rename to...
17802         (PA_HARD_REGNO_MODE_OK): ...this
17803         * config/pa/pa64-regs.h (HARD_REGNO_MODE_OK): Rename to...
17804         (PA_HARD_REGNO_MODE_OK): ...this.
17805         * config/pa/pa.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17806         (pa_hard_regno_mode_ok): New function.
17807         * config/pdp11/pdp11.h (HARD_REGNO_MODE_OK): Delete.
17808         * config/pdp11/pdp11.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17809         (pdp11_hard_regno_mode_ok): New function.
17810         * config/powerpcspe/powerpcspe.h (HARD_REGNO_MODE_OK): Delete.
17811         * config/powerpcspe/powerpcspe-protos.h (rs6000_hard_regno_mode_ok_p):
17812         Delete.
17813         * config/powerpcspe/powerpcspe.c (rs6000_hard_regno_mode_ok_p):
17814         Make static.
17815         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17816         (rs6000_hard_regno_mode_ok): Rename to...
17817         (rs6000_hard_regno_mode_ok_uncached): ...this.
17818         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17819         (rs6000_hard_regno_mode_ok): New function.
17820         * config/riscv/riscv.h (HARD_REGNO_MODE_OK): Delete.
17821         * config/riscv/riscv-protos.h (riscv_hard_regno_mode_ok_p): Delete.
17822         * config/riscv/riscv.c (riscv_hard_regno_mode_ok_p): Rename to...
17823         (riscv_hard_regno_mode_ok): ...this and make static.
17824         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17825         * config/rl78/rl78.h (HARD_REGNO_MODE_OK): Delete.
17826         * config/rl78/rl78-protos.h (rl78_hard_regno_mode_ok): Delete.
17827         * config/rl78/rl78.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17828         (rl78_hard_regno_mode_ok): Make static and return bool.
17829         * config/rs6000/rs6000.h (HARD_REGNO_MODE_OK): Delete.
17830         * config/rs6000/rs6000-protos.h (rs6000_hard_regno_mode_ok_p):
17831         Delete.
17832         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok_p): Make static.
17833         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17834         (rs6000_hard_regno_mode_ok): Rename to...
17835         (rs6000_hard_regno_mode_ok_uncached): ...this.
17836         (rs6000_init_hard_regno_mode_ok): Update accordingly.
17837         (rs6000_hard_regno_mode_ok): New function.
17838         * config/rx/rx.h (HARD_REGNO_MODE_OK): Delete.
17839         * config/rx/rx.c (rx_hard_regno_mode_ok): New function.
17840         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17841         * config/s390/s390.h (HARD_REGNO_MODE_OK): Delete.
17842         * config/s390/s390-protos.h (s390_hard_regno_mode_ok): Delete.
17843         * config/s390/s390.c (s390_hard_regno_mode_ok): Make static.
17844         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17845         * config/sh/sh.h (HARD_REGNO_MODE_OK): Delete.
17846         * config/sh/sh-protos.h (sh_hard_regno_mode_ok): Delete.
17847         * config/sh/sh.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17848         (sh_hard_regno_mode_ok): Make static.
17849         * config/sparc/constraints.md: Refer to targetm.hard_regno_mode_ok
17850         instead of HARD_REGNO_MODE_OK.
17851         * config/sparc/sparc.h (hard_regno_mode_classes): Delete.
17852         (sparc_mode_class): Delete.
17853         (HARD_REGNO_MODE_OK): Delete.
17854         * config/sparc/sparc.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17855         (hard_regno_mode_classes): Make static.
17856         (sparc_mode_class): Likewise.
17857         (sparc_hard_regno_mode_ok): New function.
17858         * config/spu/spu.h (HARD_REGNO_MODE_OK): Delete.
17859         * config/stormy16/stormy16.h (HARD_REGNO_MODE_OK): Delete.
17860         * config/stormy16/stormy16.c (xstormy16_hard_regno_mode_ok): New
17861         function.
17862         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17863         * config/tilegx/tilegx.h (HARD_REGNO_MODE_OK): Delete.
17864         * config/tilepro/tilepro.h (HARD_REGNO_MODE_OK): Delete.
17865         * config/v850/v850.h (HARD_REGNO_MODE_OK): Delete.
17866         * config/v850/v850.c (v850_hard_regno_mode_ok): New function.
17867         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17868         * config/vax/vax.h (HARD_REGNO_MODE_OK): Delete.
17869         * config/visium/visium.h (HARD_REGNO_MODE_OK): Delete.
17870         * config/visium/visium.c (TARGET_HARD_REGNO_MODE_OK): Redefine.
17871         (visium_hard_regno_mode_ok): New function.
17872         * config/visium/visium.md: Refer to targetm.hard_regno_mode_ok
17873         instead of HARD_REGNO_MODE_OK.
17874         * config/xtensa/xtensa.h (xtensa_hard_regno_mode_ok): Delete.
17875         (HARD_REGNO_MODE_OK): Delete.
17876         * config/xtensa/xtensa.c (xtensa_hard_regno_mode_ok): Rename to...
17877         (xtensa_hard_regno_mode_ok_p): ...this and make static.
17878         (xtensa_option_override): Update accordingly.
17879         (TARGET_HARD_REGNO_MODE_OK): Redefine.
17880         (xtensa_hard_regno_mode_ok): New function.
17881         * system.h (HARD_REGNO_MODE_OK): Poison.
17883 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17884             Alan Hayward  <alan.hayward@arm.com>
17885             David Sherwood  <david.sherwood@arm.com>
17887         * target.def (hard_regno_call_part_clobbered): New hook.
17888         * doc/tm.texi.in (HARD_REGNO_CALL_PART_CLOBBERED): Replace with...
17889         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): ...this hook.
17890         * doc/tm.texi: Regenerate.
17891         * hooks.h (hook_bool_uint_mode_false): Declare.
17892         * hooks.c (hook_bool_uint_mode_false): New function.
17893         * regs.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17894         * cselib.c (cselib_process_insn): Use
17895         targetm.hard_regno_call_part_clobbered instead of
17896         HARD_REGNO_CALL_PART_CLOBBERED.
17897         * ira-conflicts.c (ira_build_conflicts): Likewise.
17898         * ira-costs.c (ira_tune_allocno_costs): Likewise.
17899         * lra-constraints.c (need_for_call_save_p): Likewise.
17900         * lra-lives.c: Include target.h.
17901         (check_pseudos_live_through_calls): Use
17902         targetm.hard_regno_call_part_clobbered instead of
17903         HARD_REGNO_CALL_PART_CLOBBERED.
17904         * regcprop.c: Include target.h.
17905         (copyprop_hardreg_forward_1): Use
17906         targetm.hard_regno_call_part_clobbered instead of
17907         HARD_REGNO_CALL_PART_CLOBBERED.
17908         * reginfo.c (choose_hard_reg_mode): Likewise.
17909         * regrename.c (check_new_reg_p): Likewise.
17910         * reload.c (find_equiv_reg): Likewise.
17911         * reload1.c (emit_reload_insns): Likewise.
17912         * sched-deps.c (deps_analyze_insn): Likewise.
17913         * sel-sched.c (init_regs_for_mode): Likewise.
17914         (mark_unavailable_hard_regs): Likewise.
17915         * targhooks.c (default_dwarf_frame_reg_mode): Likewise.
17916         * config/aarch64/aarch64.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17917         * config/aarch64/aarch64.c (aarch64_hard_regno_call_part_clobbered):
17918         New function.
17919         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17920         * config/avr/avr.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17921         * config/avr/avr-protos.h (avr_hard_regno_call_part_clobbered):
17922         Delete.
17923         * config/avr/avr.c (avr_hard_regno_call_part_clobbered): Make static
17924         and return a bool.
17925         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17926         * config/i386/i386.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17927         * config/i386/i386.c (ix86_hard_regno_call_part_clobbered): New
17928         function.
17929         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17930         * config/mips/mips.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17931         * config/mips/mips.c (mips_hard_regno_call_part_clobbered): New
17932         function.
17933         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17934         * config/powerpcspe/powerpcspe.h (HARD_REGNO_CALL_PART_CLOBBERED):
17935         Delete.
17936         * config/powerpcspe/powerpcspe.c
17937         (rs6000_hard_regno_call_part_clobbered): New function.
17938         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17939         * config/rs6000/rs6000.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17940         * config/rs6000/rs6000.c (rs6000_hard_regno_call_part_clobbered):
17941         New function.
17942         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17943         * config/s390/s390.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17944         * config/s390/s390.c (s390_hard_regno_call_part_clobbered): New
17945         function.
17946         (TARGET_HARD_REGNO_CALL_PART_CLOBBERED): Redefine.
17947         * config/sh/sh.h (HARD_REGNO_CALL_PART_CLOBBERED): Delete.
17948         * system.h (HARD_REGNO_CALL_PART_CLOBBERED): Poison.
17950 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17951             Alan Hayward  <alan.hayward@arm.com>
17952             David Sherwood  <david.sherwood@arm.com>
17954         * rtl.h (subreg_memory_offset): Declare.
17955         * emit-rtl.c (subreg_memory_offset): New function.
17956         * expmed.c (store_bit_field_1): Use it.
17957         * expr.c (undefined_operand_subword_p): Likewise.
17958         * simplify-rtx.c (simplify_subreg): Likewise.
17960 2017-09-04  Alexander Monakov  <amonakov@ispras.ru>
17962         PR rtl-optimization/57448
17963         PR target/67458
17964         PR target/81316
17965         * optabs.c (expand_atomic_load): Place compiler memory barriers if
17966         using atomic_load pattern.
17967         (expand_atomic_store): Likewise.
17969 2017-09-04  Jakub Jelinek  <jakub@redhat.com>
17971         PR sanitizer/81981
17972         * gimple-fold.c (gimple_fold_call): Optimize away useless UBSAN_PTR
17973         and UBSAN_BOUNDS internal calls.  Clean up IFN_UBSAN_OBJECT_SIZE
17974         handling.  Use replace_call_with_value with NULL instead of
17975         gsi_replace, unlink_stmt_vdef and release_defs.
17977         * gdbhooks.py (OptMachineModePrinter.to_string): Use 8 spaces
17978         instead of tab.
17980         * lra-remat.c (reg_overlap_for_remat_p): Fix a pasto.
17982 2017-09-04  Richard Sandiford  <richard.sandiford@linaro.org>
17984         PR bootstrap/82045
17985         * rtl.h (emit_library_call_value_1): Declare.
17986         (emit_library_call): Replace declaration with a series of overloads.
17987         Remove the parameter count argument.
17988         (emit_library_call_value): Likewise.
17989         * calls.c (emit_library_call_value_1): Make global.  Replace varargs
17990         with an "rtx_mode_t *".
17991         (emit_library_call_value): Delete.
17992         (emit_library_call): Likewise.
17993         * asan.c (asan_emit_stack_protection): Update calls accordingly.
17994         (asan_emit_allocas_unpoison): Likewise.
17995         * builtins.c (expand_builtin_powi): Likewise.
17996         (expand_asan_emit_allocas_unpoison): Likewise.
17997         * cfgexpand.c (expand_main_function): Likewise.
17998         * config/aarch64/aarch64.c (aarch64_trampoline_init): Likewise.
17999         * config/aarch64/aarch64.h (PROFILE_HOOK): Likewise.
18000         * config/alpha/alpha.c (alpha_trampoline_init): Likewise.
18001         * config/arm/arm.c (arm_trampoline_init): Likewise.
18002         (arm_call_tls_get_addr): Likewise.
18003         (arm_expand_divmod_libfunc): Likewise.
18004         * config/bfin/bfin.md (umulsi3_highpart): Likewise.
18005         (smulsi3_highpart): Likewise.
18006         * config/c6x/c6x.c (c6x_initialize_trampoline): Likewise.
18007         (c6x_expand_compare): Likewise.
18008         (c6x_expand_movmem): Likewise.
18009         * config/frv/frv.c (frv_trampoline_init): Likewise.
18010         * config/i386/i386.c (ix86_trampoline_init): Likewise.
18011         (ix86_expand_divmod_libfunc): Likewise.
18012         * config/ia64/ia64.c (ia64_expand_tls_address): Likewise.
18013         (ia64_expand_compare): Likewise.
18014         (ia64_profile_hook): Likewise.
18015         * config/ia64/ia64.md (save_stack_nonlocal): Likewise.
18016         (nonlocal_goto): Likewise.
18017         (restore_stack_nonlocal): Likewise.
18018         * config/m32r/m32r.c (block_move_call): Likewise.
18019         (m32r_trampoline_init): Likewise.
18020         * config/m68k/linux.h (FINALIZE_TRAMPOLINE): Likewise.
18021         * config/m68k/m68k.c (m68k_call_tls_get_addr): Likewise.
18022         (m68k_call_m68k_read_tp): Likewise.
18023         * config/microblaze/microblaze.c (microblaze_call_tls_get_addr)
18024         (microblaze_expand_divide): Likewise.
18025         * config/mips/mips.h (mips_args): Likewise.
18026         * config/mips/sdemtk.h (mips_sync_icache): Likewise.
18027         (MIPS_ICACHE_SYNC): Likewise.
18028         * config/nios2/nios2.c (nios2_emit_expensive_div): Likewise.
18029         (nios2_trampoline_init): Likewise.
18030         * config/pa/pa.c (hppa_tls_call): Likewise.
18031         (pa_trampoline_init): Likewise.
18032         * config/pa/pa.md (canonicalize_funcptr_for_compare): Likewise.
18033         * config/powerpcspe/powerpcspe.c (rs6000_legitimize_tls_address)
18034         (expand_strn_compare): Likewise.
18035         (rs6000_generate_compare): Likewise.
18036         (rs6000_expand_float128_convert): Likewise.
18037         (output_profile_hook): Likewise.
18038         (rs6000_trampoline_init): Likewise.
18039         * config/powerpcspe/powerpcspe.md (neg<mode>2): Likewise.
18040         * config/riscv/riscv.h (PROFILE_HOOK): Likewise.
18041         * config/rs6000/rs6000-string.c (expand_strn_compare): Likewise.
18042         * config/rs6000/rs6000.c (rs6000_legitimize_tls_address): Likewise.
18043         (rs6000_generate_compare): Likewise.
18044         (rs6000_expand_float128_convert): Likewise.
18045         (output_profile_hook): Likewise.
18046         (rs6000_trampoline_init): Likewise.
18047         * config/rs6000/rs6000.md (neg<mode>2): Likewise.
18048         * config/sh/sh.c (sh_trampoline_init): Likewise.
18049         * config/sparc/sparc.c (emit_soft_tfmode_libcall): Likewise.
18050         (sparc_emit_float_lib_cmp): Likewise.
18051         (sparc32_initialize_trampoline): Likewise.
18052         (sparc64_initialize_trampoline): Likewise.
18053         (sparc_profile_hook): Likewise.
18054         * config/spu/spu.c (ea_load_store): Likewise.
18055         * config/spu/spu.md (floatunssidf2): Likewise.
18056         * config/tilegx/tilegx.c (tilegx_trampoline_init): Likewise.
18057         * config/tilepro/tilepro.c (tilepro_trampoline_init): Likewise.
18058         * config/visium/visium.c (expand_block_move_4): Likewise.
18059         (expand_block_move_2): Likewise.
18060         (expand_block_move_1): Likewise.
18061         (expand_block_set_4): Likewise.
18062         (expand_block_set_2): Likewise.
18063         (expand_block_set_1): Likewise.
18064         (visium_trampoline_init): Likewise.
18065         (visium_profile_hook): Likewise.
18066         * config/xtensa/xtensa.c (xtensa_expand_nonlocal_goto): Likewise.
18067         (xtensa_setup_frame_addresses): Likewise.
18068         (xtensa_trampoline_init): Likewise.
18069         * except.c (sjlj_emit_function_enter): Likewise.
18070         (sjlj_emit_function_exit): Likewise.
18071         * explow.c (allocate_dynamic_stack_space): Likewise.
18072         (probe_stack_range): Likewise.
18073         * expr.c (convert_mode_scalar): Likewise.
18074         * optabs.c (expand_binop): Likewise.
18075         (expand_twoval_binop_libfunc): Likewise.
18076         (expand_unop): Likewise.
18077         (prepare_cmp_insn): Likewise.
18078         (prepare_float_lib_cmp): Likewise.
18079         (expand_float): Likewise.
18080         (expand_fix): Likewise.
18081         (expand_fixed_convert): Likewise.
18082         (maybe_emit_sync_lock_test_and_set): Likewise.
18083         (expand_atomic_compare_and_swap): Likewise.
18084         (expand_mem_thread_fence): Likewise.
18085         (expand_atomic_fetch_op): Likewise.
18087 2017-09-03  Gerald Pfeifer  <gerald@pfeifer.com>
18089         * doc/generic.texi (OpenACC): Adjust URL.
18090         * doc/invoke.texi (C Dialect Options): Ditto.
18092 2017-09-03  Uros Bizjak  <ubizjak@gmail.com>
18094         * config/i386/i386.md (*bt<mode>): Use nonimmediate_operand
18095         predicate for operand 1.  Add (m,<S>) constraint.
18096         (*jcc_bt<mode>): Use nonimmediate_operand predicate for operand 1.
18097         Prevent memory operand 1 with register operand 2.
18099 2017-09-01  Segher Boessenkool  <segher@kernel.crashing.org>
18101         PR rtl-optimization/82024
18102         * combine.c (try_combine): If the combination result is a PARALLEL,
18103         and we only need to retain the SET in there that would be placed
18104         at I2, check that we can place that at I3 instead, before doing so.
18106 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18108         PR target/81766
18109         * config/i386/i386.c (ix86_init_large_pic_reg): Return label
18110         instead of void.
18111         (ix86_init_pic_reg): Remember label from ix86_init_large_pic_reg,
18112         if non-NULL and preceded by NOTE_INSN_BASIC_BLOCK, swap the note
18113         and label.
18115 2017-09-01  Joerg Sonnenberger  <joerg@bec.de>
18116             Jeff Law  <law@redhat.com>
18118         * varasm.c (bss_initializer_p): Do not put constants into .bss
18119         (categorize_decl_for_section): Handle bss_initializer_p returning
18120         false when DECL_INITIAL is NULL.
18122 2017-09-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
18124         PR target/82012
18125         * config/s390/s390.c (s390_can_inline_p): New function.
18127 2017-09-01  Jeff Law  <law@redhat.com>
18129         PR tree-optimization/82052
18130         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
18131         Always initialize the returned slot after a hash table miss
18132         when INSERT is true.
18134 2017-09-01  Alexander Monakov  <amonakov@ispras.ru>
18136         * config/s390/s390.md (mem_signal_fence): Remove.
18137         * doc/md.texi (mem_signal_fence): Remove.
18138         * optabs.c (expand_mem_signal_fence): Remove uses of mem_signal_fence.
18139         Update comments.
18140         * target-insns.def (mem_signal_fence): Remove.
18142 2017-09-01  Jakub Jelinek  <jakub@redhat.com>
18144         PR sanitizer/81902
18145         * doc/invoke.texi: Document -fsanitize=pointer-overflow.
18147         PR sanitizer/81923
18148         * asan.c (create_odr_indicator): Strip name encoding from assembler
18149         name before appending it after __odr_asan_.
18151 2017-09-01  Martin Liska  <mliska@suse.cz>
18153         PR tree-optimization/82059
18154         * gimple-ssa-isolate-paths.c (isolate_path): Add profile and
18155         frequency only when an edge is redirected.
18157 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18159         * config/arc/arc-c.c (__ARC_LPC_WIDTH__): Add builtin define.
18160         * config/arc/arc.c (ARC_MAX_LOOP_LENGTH): Define.
18161         (arc_conditional_register_usage): Remove ARC600 lp_count
18162         exception.
18163         (arc_file_start): Emit Tag_ARC_CPU_variation.
18164         (arc_can_use_doloop_p): New conditions to use ZOLs.
18165         (hwloop_fail): New function.
18166         (hwloop_optimize): Likewise.
18167         (hwloop_pattern_reg): Likewise.
18168         (arc_doloop_hooks): New struct, to be used with reorg_loops.
18169         (arc_reorg_loops): New function, calls reorg_loops.
18170         (arc_reorg): Call arc_reorg_loops.  Remove old ZOL handling.
18171         (arc600_corereg_hazard): Remove ZOL checking, case handled by
18172         hwloop_optimize.
18173         (arc_loop_hazard): Remove function, functionality moved into
18174         hwloop_optimize.
18175         (arc_hazard): Remove arc_loop_hazard call.
18176         (arc_adjust_insn_length): Remove ZOL handling, functionality moved
18177         into hwloop_optimize.
18178         (arc_label_align): Remove ZOL handling.
18179         * config/arc/arc.h (LOOP_ALIGN): Changed to 0.
18180         * config/arc/arc.md (doloop_begin): Remove pattern.
18181         (doloop_begin_i): Likewise.
18182         (doloop_end_i): Likewise.
18183         (doloop_fallback): Likewise.
18184         (doloop_fallback_m): Likewise.
18185         (doloop_end): Reimplement expand.
18186         (arc_lp): New pattern for LP instruction.
18187         (loop_end): New pattern.
18188         (loop_fail): Likewise.
18189         (decrement_and_branch_until_zero): Likewise.
18190         * config/arc/arc.opt (mlpc-width): New option.
18191         * doc/invoke.texi (mlpc-width): Document option.
18193 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18195         * config/arc/arc.c (arc_ifcvt): Remove use of merge_blocks call.
18196         (arc_ccfsm_advance): Fix checking for delay slots.
18197         (arc_reorg): Add rtl dump after each call to arc_ifcvt.
18199 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18201         * config/arc/arc.md (movqi_insn): Add stores to save constant long
18202         immediates.
18203         (movhi_insn): Update store instruction constraint which are saving
18204         6-bit short immediates.
18205         (movsi_insn): Consider also short scaled load operations.
18206         (zero_extendhisi2_i): Use Usd constraint instead of T.
18207         (extendhisi2_i): Add q constraint.
18208         (arc_clzsi2): Add type and length attributes.
18209         (arc_ctzsi2): Likewise.
18210         * config/arc/constraints.md (Usc): Update constraint, the
18211         assembler can parse two relocations for a single instruction.
18213 2017-09-01  Claudiu Zissulescu  <claziss@synopsys.com>
18215         * config/arc/arc.c (arc_use_anchors_for_symbol_p): New function.
18216         (TARGET_USE_ANCHORS_FOR_SYMBOL_P): Define.
18218 2017-08-31  Olivier Hainque  <hainque@adacore.com>
18220         * config.gcc (powerpc-wrs-vxworks|vxworksae|vxworksmils): Now
18221         match as powerpc-wrs-vxworks*.
18223 2017-08-31  James Greenhalgh  <james.greenhalgh@arm.com>
18225         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): Fix
18226         register constraint for by-element operand.
18227         (aarch64_mls_elt_merge<mode>): Likewise.
18229 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18231         * config/arc/arc.c (arc_can_follow_jump): Check for short
18232         branches.
18234 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18236         * config.gcc: Use g.opt for arc.
18237         * config/arc/arc.c (LEGITIMATE_SCALED_ADDRESS_P): Deleted,
18238         functionality moved to ...
18239         (legitimate_scaled_address_p): New function, ...here.
18240         (LEGITIMATE_SMALL_DATA_OFFSET_P): New define.
18241         (LEGITIMATE_SMALL_DATA_ADDRESS_P): Use the above define.
18242         (legitimate_offset_address_p): Delete TARGET_NO_SDATA_SET
18243         condition.
18244         (arc_override_options): Handle G option.
18245         (arc_output_pic_addr_const): Correct function definition.
18246         (arc_legitimate_address_p): Use legitimate_scaled_address_p.
18247         (arc_decl_anon_ns_mem_p): Delete.
18248         (arc_in_small_data_p): Overhaul this function to take into
18249         consideration the value given via G option.
18250         (arc_rewrite_small_data_1): Renamed and corrected old
18251         arc_rewrite_small_data function.
18252         (arc_rewrite_small_data): New function.
18253         (small_data_pattern): Don't use pic_offset_table_rtx.
18254         * config/arc/arc.h (CC1_SPEC): Recognize G option.
18255         * config/arc/simdext.md (movmisalignv2hi): Use
18256         prepare_move_operands function.
18257         (mov*): Likewise.
18258         (movmisalign*): Likewise.
18259         * doc/invoke.texi (ARC options): Document -G option.
18261 2017-08-31  Claudiu Zissulescu  <claziss@synopsys.com>
18263         * config/arc/arc-protos.h (compact_sda_memory_operand): Update
18264         prototype.
18265         * config/arc/arc.c (arc_print_operand): Output scalled address for
18266         sdata whenever is possible.
18267         (arc_in_small_data_p): Allow sdata for 64bit datum when double
18268         load/stores are available.
18269         (compact_sda_memory_operand): Check for the alignment required by
18270         code density instructions.
18271         * config/arc/arc.md (movsi_insn): Use newly introduced Us0
18272         constraint.
18273         * config/arc/constraints.md (Usd): Update constraint.
18274         (Us0): New constraint.
18275         (Usc): Update constraint.
18277 2017-08-31  Richard Biener  <rguenther@suse.de>
18279         PR middle-end/82054
18280         * dwarf2out.c (dwarf2out_early_global_decl): Process each
18281         function only once.
18283 2017-08-31  Tamar Christina  <tamar.christina@arm.com>
18285         * config/aarch64/aarch64-builtins.c (aarch64_init_simd_builtins):
18286         Resize type_signature.
18288 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18289             Alan Hayward  <alan.hayward@arm.com>
18290             David Sherwood  <david.sherwood@arm.com>
18292         * config/aarch64/aarch64.c (aarch64_base_register_rtx_p): Only allow
18293         subregs whose inner modes can be stored in GPRs.
18294         (aarch64_classify_index): Likewise.
18296 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18297             Alan Hayward  <alan.hayward@arm.com>
18298             David Sherwood  <david.sherwood@arm.com>
18300         * config/aarch64/iterators.md (V_cmp_result): Rename to...
18301         (V_INT_EQUIV): ...this.
18302         (v_cmp_result): Rename to...
18303         (v_int_equiv): ...this.
18304         * config/aarch64/aarch64.md (xorsign<mode>3): Update accordingly.
18305         * config/aarch64/aarch64-simd.md (xorsign<mode>3): Likewise.
18306         (copysign<mode>3): Likewise.
18307         (aarch64_simd_bsl<mode>_internal): Likewise.
18308         (aarch64_simd_bsl<mode>): Likewise.
18309         (vec_cmp<mode><mode>): Likewise.
18310         (vcond<mode><mode>): Likewise.
18311         (vcond<v_cmp_mixed><mode>): Likewise.
18312         (vcondu<mode><v_cmp_mixed>): Likewise.
18313         (aarch64_cm<optab><mode>): Likewise.
18314         (aarch64_cmtst<mode>): Likewise.
18315         (aarch64_fac<optab><mode>): Likewise.
18316         (vec_perm_const<mode>): Likewise.
18317         (vcond_mask_<mode><v_cmp_result>): Rename to...
18318         (vcond_mask_<mode><v_int_equiv>): ...this.
18319         (vec_cmp<mode><v_cmp_result>): Rename to...
18320         (vec_cmp<mode><v_int_equiv>): ...this.
18322 2017-08-31  Richard Sandiford  <richard.sandiford@linaro.org>
18323             Alan Hayward  <alan.hayward@arm.com>
18324             David Sherwood  <david.sherwood@arm.com>
18326         * config/aarch64/aarch64-modes.def: Remove 32-, 48- and 64-byte
18327         vector modes.
18328         * config/aarch64/iterators.md (VRL2, VRL3, VRL4): Delete.
18329         * config/aarch64/aarch64.md (UNSPEC_LD2_DREG, UNSPEC_LD3_DREG)
18330         (UNSPEC_LD4_DREG): New unspecs.
18331         * config/aarch64/aarch64-simd.md (aarch64_ld2<mode>_dreg_le)
18332         (aarch64_ld2<mode>_dreg_be): Replace with...
18333         (aarch64_ld2<mode>_dreg): ...this pattern and use the new DREG
18334         unspec.
18335         (aarch64_ld3<mode>_dreg_le)
18336         (aarch64_ld3<mode>_dreg_be): Replace with...
18337         (aarch64_ld3<mode>_dreg): ...this pattern and use the new DREG
18338         unspec.
18339         (aarch64_ld4<mode>_dreg_le)
18340         (aarch64_ld4<mode>_dreg_be): Replace with...
18341         (aarch64_ld4<mode>_dreg): ...this pattern and use the new DREG
18342         unspec.
18344 2017-08-30  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
18346         PR tree-optimization/81987
18347         * gimple-ssa-strength-reduction.c (insert_initializers): Don't
18348         insert an initializer in a location not dominated by the stride
18349         definition.
18351 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18353         * tree-eh.c (lower_try_finally_switch): Set the location of the finally
18354         on the entire header of the finally block in the fallthru case.
18356 2017-08-30  Eric Botcazou  <ebotcazou@adacore.com>
18358         * varasm.c (decode_addr_const): Deal with INDIRECT_REF <INTEGER_CST>.
18360 2017-08-30  Pat Haugen  <pthaugen@us.ibm.com>
18362         * config/rs6000/rs6000.c (rs6000_emit_prologue_move_from_cr): Rename from
18363         rs6000_emit_move_from_cr and call renamed function.
18364         (rs6000_emit_prologue): Call renamed functions.
18365         * config/rs6000/rs6000.md (prologue_movesi_from_cr): Rename from
18366         movesi_from_cr, remove volatile CRs.
18368 2017-08-30  Jon Beniston  <jon@beniston.com>
18369             Richard Biener  <rguenther@suse.de>
18371         * tree-vect-patterns.c (vect_pattern_recog_1): Use VECTOR_TYPE_P instead
18372         of VECTOR_MODE_P check.
18373         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Allow single
18374         element vector types.
18376 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18378         * df.h (df_read_modify_subreg_p): Remove in favor of...
18379         * rtl.h (read_modify_subreg_p): ...this new function.  Take a
18380         const_rtx instead of an rtx.
18381         * cprop.c (local_cprop_find_used_regs): Update accordingly.
18382         * df-problems.c (df_word_lr_mark_ref): Likewise.
18383         * ira-lives.c (mark_pseudo_reg_live): Likewise.
18384         (mark_pseudo_reg_dead): Likewise.
18385         (mark_ref_dead): Likewise.
18386         * reginfo.c (init_subregs_of_mode): Likewise.
18387         * sched-deps.c (sched_analyze_1): Likewise.
18388         * df-scan.c (df_def_record_1): Likewise.
18389         (df_uses_record): Likewise.
18390         (df_read_modify_subreg_p): Remove in favor of...
18391         * rtlanal.c (read_modify_subreg_p): ...this new function.  Take a
18392         const_rtx instead of an rtx.
18394 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18395             Alan Hayward  <alan.hayward@arm.com>
18396             David Sherwood  <david.sherwood@arm.com>
18398         * rtl.h (partial_subreg_p): New function.
18399         * caller-save.c (save_call_clobbered_regs): Use it.
18400         * calls.c (expand_call): Likewise.
18401         * combine.c (combinable_i3pat): Likewise.
18402         (simplify_set): Likewise.
18403         (make_extraction): Likewise.
18404         (make_compound_operation_int): Likewise.
18405         (gen_lowpart_or_truncate): Likewise.
18406         (force_to_mode): Likewise.
18407         (make_field_assignment): Likewise.
18408         (reg_truncated_to_mode): Likewise.
18409         (record_truncated_value): Likewise.
18410         (move_deaths): Likewise.
18411         * cse.c (record_jump_cond): Likewise.
18412         (cse_insn): Likewise.
18413         * cselib.c (cselib_lookup_1): Likewise.
18414         * expmed.c (extract_bit_field_using_extv): Likewise.
18415         * function.c (assign_parm_setup_reg): Likewise.
18416         * ifcvt.c (noce_convert_multiple_sets): Likewise.
18417         * ira-build.c (create_insn_allocnos): Likewise.
18418         * lra-coalesce.c (merge_pseudos): Likewise.
18419         * lra-constraints.c (match_reload): Likewise.
18420         (simplify_operand_subreg): Likewise.
18421         (curr_insn_transform): Likewise.
18422         * lra-lives.c (process_bb_lives): Likewise.
18423         * lra.c (new_insn_reg): Likewise.
18424         (lra_substitute_pseudo): Likewise.
18425         * regcprop.c (mode_change_ok): Likewise.
18426         (maybe_mode_change): Likewise.
18427         (copyprop_hardreg_forward_1): Likewise.
18428         * reload.c (push_reload): Likewise.
18429         (find_reloads): Likewise.
18430         (find_reloads_subreg_address): Likewise.
18431         * reload1.c (alter_reg): Likewise.
18432         (eliminate_regs_1): Likewise.
18433         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
18435 2017-08-30  David Edelsohn  <dje.gcc@gmail.com>
18437         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Revert
18438         back to if statements, including unpack.
18440 2017-08-30  Martin Liska  <mliska@suse.cz>
18442         PR inline-asm/82001
18443         * ipa-icf-gimple.c (func_checker::compare_tree_list_operand):
18444         Rename to ...
18445         (func_checker::compare_asm_inputs_outputs): ... this function.
18446         (func_checker::compare_gimple_asm): Use the function to compare
18447         also ASM constrains.
18448         * ipa-icf-gimple.h: Rename the function.
18450 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18451             Alan Hayward  <alan.hayward@arm.com>
18452             David Sherwood  <david.sherwood@arm.com>
18454         * coretypes.h (complex_mode): New type.
18455         * gdbhooks.py (build_pretty_printer): Handle it.
18456         * machmode.h (complex_mode): New class.
18457         (complex_mode::includes_p): New function.
18458         (is_complex_int_mode): Likewise.
18459         (is_complex_float_mode): Likewise.
18460         * genmodes.c (get_mode_class): Handle complex mode classes.
18461         * function.c (expand_function_end): Use is_complex_int_mode.
18463 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18464             Alan Hayward  <alan.hayward@arm.com>
18465             David Sherwood  <david.sherwood@arm.com>
18467         * coretypes.h (scalar_mode_pod): New typedef.
18468         * gdbhooks.py (build_pretty_printer): Handle it.
18469         * machmode.h (gt_ggc_mx, gt_pch_nx): New functions.
18470         * fixed-value.h (fixed_value::mode): Change type to scalar_mode_pod.
18471         * fold-const.c (fold_convert_const_int_from_fixed): Use scalar_mode.
18472         * tree-streamer-in.c (unpack_ts_fixed_cst_value_fields): Use
18473         as_a <scalar_mode>.
18475 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18476             Alan Hayward  <alan.hayward@arm.com>
18477             David Sherwood  <david.sherwood@arm.com>
18479         * machmode.h (mode_for_vector): Take a scalar_mode instead
18480         of a machine_mode.
18481         * stor-layout.c (mode_for_vector): Likewise.
18482         * explow.c (promote_mode): Use as_a <scalar_mode>.
18483         * sdbout.c (sdbout_parms): Use is_a <scalar_mode>.
18485 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18486             Alan Hayward  <alan.hayward@arm.com>
18487             David Sherwood  <david.sherwood@arm.com>
18489         * target.def (preferred_simd_mode): Take a scalar_mode
18490         instead of a machine_mode.
18491         * targhooks.h (default_preferred_simd_mode): Likewise.
18492         * targhooks.c (default_preferred_simd_mode): Likewise.
18493         * config/arc/arc.c (arc_preferred_simd_mode): Likewise.
18494         * config/arm/arm.c (arm_preferred_simd_mode): Likewise.
18495         * config/c6x/c6x.c (c6x_preferred_simd_mode): Likewise.
18496         * config/epiphany/epiphany.c (epiphany_preferred_simd_mode): Likewise.
18497         * config/i386/i386.c (ix86_preferred_simd_mode): Likewise.
18498         * config/mips/mips.c (mips_preferred_simd_mode): Likewise.
18499         * config/nvptx/nvptx.c (nvptx_preferred_simd_mode): Likewise.
18500         * config/powerpcspe/powerpcspe.c (rs6000_preferred_simd_mode):
18501         Likewise.
18502         * config/rs6000/rs6000.c (rs6000_preferred_simd_mode): Likewise.
18503         * config/s390/s390.c (s390_preferred_simd_mode): Likewise.
18504         * config/sparc/sparc.c (sparc_preferred_simd_mode): Likewise.
18505         * config/aarch64/aarch64.c (aarch64_preferred_simd_mode): Likewise.
18506         (aarch64_simd_scalar_immediate_valid_for_move): Update accordingly.
18507         * doc/tm.texi: Regenerate.
18508         * optabs-query.c (can_vec_mask_load_store_p): Return false for
18509         non-scalar modes.
18511 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18512             Alan Hayward  <alan.hayward@arm.com>
18513             David Sherwood  <david.sherwood@arm.com>
18515         * target.def (scalar_mode_supported_p): Take a scalar_mode
18516         instead of a machine_mode.
18517         * targhooks.h (default_scalar_mode_supported_p): Likewise.
18518         * targhooks.c (default_scalar_mode_supported_p): Likewise.
18519         * config/aarch64/aarch64.c (aarch64_scalar_mode_supported_p): Likewise.
18520         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
18521         * config/arm/arm.c (arm_scalar_mode_supported_p): Likewise.
18522         * config/avr/avr.c (avr_scalar_mode_supported_p): Likewise.
18523         * config/c6x/c6x.c (c6x_scalar_mode_supported_p): Likewise.
18524         * config/i386/i386.c (ix86_scalar_mode_supported_p): Likewise.
18525         * config/ia64/ia64.c (ia64_scalar_mode_supported_p): Likewise.
18526         * config/mips/mips.c (mips_scalar_mode_supported_p): Likewise.
18527         * config/msp430/msp430.c (msp430_scalar_mode_supported_p): Likewise.
18528         * config/pa/pa.c (pa_scalar_mode_supported_p): Likewise.
18529         * config/pdp11/pdp11.c (pdp11_scalar_mode_supported_p): Likewise.
18530         * config/powerpcspe/powerpcspe.c (rs6000_scalar_mode_supported_p):
18531         Likewise.
18532         * config/rs6000/rs6000.c (rs6000_scalar_mode_supported_p): Likewise.
18533         * config/s390/s390.c (s390_scalar_mode_supported_p): Likewise.
18534         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
18535         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
18536         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p):
18537         Likewise.
18538         * doc/tm.texi: Regenerate.
18540 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18541             Alan Hayward  <alan.hayward@arm.com>
18542             David Sherwood  <david.sherwood@arm.com>
18544         * coretypes.h (opt_scalar_mode): New typedef.
18545         * gdbhooks.py (build_pretty_printers): Handle it.
18546         * machmode.h (mode_iterator::get_2xwider): Add overload for
18547         opt_mode<T>.
18548         * emit-rtl.c (init_emit_once): Use opt_scalar_mode when iterating
18549         over scalar modes.
18550         * expr.c (convert_mode_scalar): Likewise.
18551         * omp-low.c (omp_clause_aligned_alignment): Likewise.
18552         * optabs.c (expand_float): Likewise.
18553         (expand_fix): Likewise.
18554         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18556 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18557             Alan Hayward  <alan.hayward@arm.com>
18558             David Sherwood  <david.sherwood@arm.com>
18560         * optabs.c (expand_float): Explicitly check for scalars before
18561         using a branching expansion.
18562         (expand_fix): Likewise.
18564 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18565             Alan Hayward  <alan.hayward@arm.com>
18566             David Sherwood  <david.sherwood@arm.com>
18568         * expr.c (convert_mode): Split scalar handling out into...
18569         (convert_mode_scalar): ...this new function.  Treat the modes
18570         as scalar_modes.
18572 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18573             Alan Hayward  <alan.hayward@arm.com>
18574             David Sherwood  <david.sherwood@arm.com>
18576         * omp-expand.c (expand_omp_atomic): Use is_int_mode, is_float_mode
18577         and scalar_mode.
18578         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Likewise.
18580 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18581             Alan Hayward  <alan.hayward@arm.com>
18582             David Sherwood  <david.sherwood@arm.com>
18584         * fixed-value.h (fixed_from_double_int): Take a scalar_mode
18585         rather than a machine_mode.
18586         (fixed_from_string): Likewise.
18587         (fixed_convert): Likewise.
18588         (fixed_convert_from_int): Likewise.
18589         (fixed_convert_from_real): Likewise.
18590         (real_convert_from_fixed): Likewise.
18591         * fixed-value.c (fixed_from_double_int): Likewise.
18592         (fixed_from_string): Likewise.
18593         (fixed_convert): Likewise.
18594         (fixed_convert_from_int): Likewise.
18595         (fixed_convert_from_real): Likewise.
18596         (real_convert_from_fixed): Likewise.
18597         * config/avr/avr.c (avr_out_round): Use as_a <scalar_mode>.
18599 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18600             Alan Hayward  <alan.hayward@arm.com>
18601             David Sherwood  <david.sherwood@arm.com>
18603         * emit-rtl.c (immed_double_const): Use is_a <scalar_mode> instead
18604         of separate mode class checks.  Do not allow vector modes here.
18605         (immed_wide_int_const): Use as_a <scalar_mode>.
18606         * explow.c (trunc_int_for_mode): Likewise.
18607         * rtl.h (wi::int_traits<rtx_mode_t>::get_precision): Likewise.
18608         (wi::shwi): Likewise.
18609         (wi::min_value): Likewise.
18610         (wi::max_value): Likewise.
18611         * dwarf2out.c (loc_descriptor): Likewise.
18612         * simplify-rtx.c (simplify_immed_subreg): Fix rtx_mode_t argument
18613         for CONST_WIDE_INT.
18615 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18616             Alan Hayward  <alan.hayward@arm.com>
18617             David Sherwood  <david.sherwood@arm.com>
18619         * tree.h (SCALAR_TYPE_MODE): New macro.
18620         * expr.c (expand_expr_addr_expr_1): Use it.
18621         (expand_expr_real_2): Likewise.
18622         * fold-const.c (fold_convert_const_fixed_from_fixed): Likeise.
18623         (fold_convert_const_fixed_from_int): Likewise.
18624         (fold_convert_const_fixed_from_real): Likewise.
18625         (native_encode_fixed): Likewise
18626         (native_encode_complex): Likewise
18627         (native_encode_vector): Likewise.
18628         (native_interpret_fixed): Likewise.
18629         (native_interpret_real): Likewise.
18630         (native_interpret_complex): Likewise.
18631         (native_interpret_vector): Likewise.
18632         * omp-simd-clone.c (simd_clone_adjust_return_type): Likewise.
18633         (simd_clone_adjust_argument_types): Likewise.
18634         (simd_clone_init_simd_arrays): Likewise.
18635         (simd_clone_adjust): Likewise.
18636         * stor-layout.c (layout_type): Likewise.
18637         * tree.c (build_minus_one_cst): Likewise.
18638         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
18639         * tree-inline.c (estimate_move_cost): Likewise.
18640         * tree-ssa-math-opts.c (convert_plusminus_to_widen): Likewise.
18641         * tree-vect-loop.c (vect_create_epilog_for_reduction): Likewise.
18642         (vectorizable_reduction): Likewise.
18643         * tree-vect-patterns.c (vect_recog_widen_mult_pattern): Likewise.
18644         (vect_recog_mixed_size_cond_pattern): Likewise.
18645         (check_bool_pattern): Likewise.
18646         (adjust_bool_pattern): Likewise.
18647         (search_type_for_mask_1): Likewise.
18648         * tree-vect-slp.c (vect_schedule_slp_instance): Likewise.
18649         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
18650         (vectorizable_load): Likewise.
18651         (vectorizable_store): Likewise.
18652         * ubsan.c (ubsan_encode_value): Likewise.
18653         * varasm.c (output_constant): Likewise.
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         * coretypes.h (scalar_mode): New class.
18660         * machmode.h (scalar_mode): Likewise.
18661         (scalar_mode::includes_p): New function.
18662         (mode_to_inner): Return a scalar_mode rather than a machine_mode.
18663         * gdbhooks.py (build_pretty_printers): Handle scalar_mode.
18664         * genmodes.c (get_mode_class): Handle remaining scalar modes.
18665         * cfgexpand.c (expand_debug_expr): Use scalar_mode.
18666         * expmed.c (store_bit_field_1): Likewise.
18667         (extract_bit_field_1): Likewise.
18668         * expr.c (write_complex_part): Likewise.
18669         (read_complex_part): Likewise.
18670         (emit_move_complex_push): Likewise.
18671         (expand_expr_real_2): Likewise.
18672         * function.c (assign_parm_setup_reg): Likewise.
18673         (assign_parms_unsplit_complex): Likewise.
18674         * optabs.c (expand_binop): Likewise.
18675         * rtlanal.c (subreg_get_info): Likewise.
18676         * simplify-rtx.c (simplify_immed_subreg): Likewise.
18677         * varasm.c (output_constant_pool_2): Likewise.
18679 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18680             Alan Hayward  <alan.hayward@arm.com>
18681             David Sherwood  <david.sherwood@arm.com>
18683         * expmed.c (extract_high_half): Use scalar_int_mode and remove
18684         assertion.
18685         (expmed_mult_highpart_optab): Likewise.
18686         (expmed_mult_highpart): Likewise.
18688 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18689             Alan Hayward  <alan.hayward@arm.com>
18690             David Sherwood  <david.sherwood@arm.com>
18692         * builtins.h (builtin_strncpy_read_str): Take a scalar_int_mode
18693         instead of a machine_mode.
18694         (builtin_memset_read_str): Likewise.
18695         * builtins.c (c_readstr): Likewise.
18696         (builtin_memcpy_read_str): Likewise.
18697         (builtin_strncpy_read_str): Likewise.
18698         (builtin_memset_read_str): Likewise.
18699         (builtin_memset_gen_str): Likewise.
18700         (expand_builtin_signbit): Use scalar_int_mode for local variables.
18701         * cfgexpand.c (convert_debug_memory_address): Take a scalar_int_mode
18702         instead of a machine_mode.
18703         * combine.c (simplify_if_then_else): Use scalar_int_mode for local
18704         variables.
18705         (make_extraction): Likewise.
18706         (try_widen_shift_mode): Take and return scalar_int_modes instead
18707         of machine_modes.
18708         * config/aarch64/aarch64.c (aarch64_libgcc_cmp_return_mode): Return
18709         a scalar_int_mode instead of a machine_mode.
18710         * config/avr/avr.c (avr_addr_space_address_mode): Likewise.
18711         (avr_addr_space_pointer_mode): Likewise.
18712         * config/cr16/cr16.c (cr16_unwind_word_mode): Likewise.
18713         * config/msp430/msp430.c (msp430_addr_space_pointer_mode): Likewise.
18714         (msp430_unwind_word_mode): Likewise.
18715         * config/spu/spu.c (spu_unwind_word_mode): Likewise.
18716         (spu_addr_space_pointer_mode): Likewise.
18717         (spu_addr_space_address_mode): Likewise.
18718         (spu_libgcc_cmp_return_mode): Likewise.
18719         (spu_libgcc_shift_count_mode): Likewise.
18720         * config/rl78/rl78.c (rl78_addr_space_address_mode): Likewise.
18721         (rl78_addr_space_pointer_mode): Likewise.
18722         (fl78_unwind_word_mode): Likewise.
18723         (rl78_valid_pointer_mode): Take a scalar_int_mode instead of a
18724         machine_mode.
18725         * config/alpha/alpha.c (vms_valid_pointer_mode): Likewise.
18726         * config/ia64/ia64.c (ia64_vms_valid_pointer_mode): Likewise.
18727         * config/mips/mips.c (mips_mode_rep_extended): Likewise.
18728         (mips_valid_pointer_mode): Likewise.
18729         * config/tilegx/tilegx.c (tilegx_mode_rep_extended): Likewise.
18730         * config/ft32/ft32.c (ft32_valid_pointer_mode): Likewise.
18731         (ft32_addr_space_pointer_mode): Return a scalar_int_mode instead
18732         of a machine_mode.
18733         (ft32_addr_space_address_mode): Likewise.
18734         * config/m32c/m32c.c (m32c_valid_pointer_mode): Take a
18735         scalar_int_mode instead of a machine_mode.
18736         (m32c_addr_space_pointer_mode): Return a scalar_int_mode instead
18737         of a machine_mode.
18738         (m32c_addr_space_address_mode): Likewise.
18739         * config/powerpcspe/powerpcspe.c (rs6000_abi_word_mode): Likewise.
18740         (rs6000_eh_return_filter_mode): Likewise.
18741         * config/rs6000/rs6000.c (rs6000_abi_word_mode): Likewise.
18742         (rs6000_eh_return_filter_mode): Likewise.
18743         * config/s390/s390.c (s390_libgcc_cmp_return_mode): Likewise.
18744         (s390_libgcc_shift_count_mode): Likewise.
18745         (s390_unwind_word_mode): Likewise.
18746         (s390_valid_pointer_mode): Take a scalar_int_mode rather than a
18747         machine_mode.
18748         * target.def (mode_rep_extended): Likewise.
18749         (valid_pointer_mode): Likewise.
18750         (addr_space.valid_pointer_mode): Likewise.
18751         (eh_return_filter_mode): Return a scalar_int_mode rather than
18752         a machine_mode.
18753         (libgcc_cmp_return_mode): Likewise.
18754         (libgcc_shift_count_mode): Likewise.
18755         (unwind_word_mode): Likewise.
18756         (addr_space.pointer_mode): Likewise.
18757         (addr_space.address_mode): Likewise.
18758         * doc/tm.texi: Regenerate.
18759         * dojump.c (prefer_and_bit_test): Take a scalar_int_mode rather than
18760         a machine_mode.
18761         (do_jump): Use scalar_int_mode for local variables.
18762         * dwarf2cfi.c (init_return_column_size): Take a scalar_int_mode
18763         rather than a machine_mode.
18764         * dwarf2out.c (convert_descriptor_to_mode): Likewise.
18765         (scompare_loc_descriptor_wide): Likewise.
18766         (scompare_loc_descriptor_narrow): Likewise.
18767         * emit-rtl.c (adjust_address_1): Use scalar_int_mode for local
18768         variables.
18769         * except.c (sjlj_emit_dispatch_table): Likewise.
18770         (expand_builtin_eh_copy_values): Likewise.
18771         * explow.c (convert_memory_address_addr_space_1): Likewise.
18772         Take a scalar_int_mode rather than a machine_mode.
18773         (convert_memory_address_addr_space): Take a scalar_int_mode rather
18774         than a machine_mode.
18775         (memory_address_addr_space): Use scalar_int_mode for local variables.
18776         * expmed.h (expand_mult_highpart_adjust): Take a scalar_int_mode
18777         rather than a machine_mode.
18778         * expmed.c (mask_rtx): Likewise.
18779         (init_expmed_one_conv): Likewise.
18780         (expand_mult_highpart_adjust): Likewise.
18781         (extract_high_half): Likewise.
18782         (expmed_mult_highpart_optab): Likewise.
18783         (expmed_mult_highpart): Likewise.
18784         (expand_smod_pow2): Likewise.
18785         (expand_sdiv_pow2): Likewise.
18786         (emit_store_flag_int): Likewise.
18787         (adjust_bit_field_mem_for_reg): Use scalar_int_mode for local
18788         variables.
18789         (extract_low_bits): Likewise.
18790         * expr.h (by_pieces_constfn): Take a scalar_int_mode rather than
18791         a machine_mode.
18792         * expr.c (pieces_addr::adjust):  Likewise.
18793         (can_store_by_pieces): Likewise.
18794         (store_by_pieces): Likewise.
18795         (clear_by_pieces_1): Likewise.
18796         (expand_expr_addr_expr_1): Likewise.
18797         (expand_expr_addr_expr): Use scalar_int_mode for local variables.
18798         (expand_expr_real_1): Likewise.
18799         (try_casesi): Likewise.
18800         * final.c (shorten_branches): Likewise.
18801         * fold-const.c (fold_convert_const_int_from_fixed): Change the
18802         type of "mode" to machine_mode.
18803         * internal-fn.c (expand_arith_overflow_result_store): Take a
18804         scalar_int_mode rather than a machine_mode.
18805         (expand_mul_overflow): Use scalar_int_mode for local variables.
18806         * loop-doloop.c (doloop_modify): Likewise.
18807         (doloop_optimize): Likewise.
18808         * optabs.c (expand_subword_shift): Take a scalar_int_mode rather
18809         than a machine_mode.
18810         (expand_doubleword_shift_condmove): Likewise.
18811         (expand_doubleword_shift): Likewise.
18812         (expand_doubleword_clz): Likewise.
18813         (expand_doubleword_popcount): Likewise.
18814         (expand_doubleword_parity): Likewise.
18815         (expand_absneg_bit): Use scalar_int_mode for local variables.
18816         (prepare_float_lib_cmp): Likewise.
18817         * rtl.h (convert_memory_address_addr_space_1): Take a scalar_int_mode
18818         rather than a machine_mode.
18819         (convert_memory_address_addr_space): Likewise.
18820         (get_mode_bounds): Likewise.
18821         (get_address_mode): Return a scalar_int_mode rather than a
18822         machine_mode.
18823         * rtlanal.c (get_address_mode): Likewise.
18824         * stor-layout.c (get_mode_bounds): Take a scalar_int_mode rather
18825         than a machine_mode.
18826         * targhooks.c (default_mode_rep_extended): Likewise.
18827         (default_valid_pointer_mode): Likewise.
18828         (default_addr_space_valid_pointer_mode): Likewise.
18829         (default_eh_return_filter_mode): Return a scalar_int_mode rather
18830         than a machine_mode.
18831         (default_libgcc_cmp_return_mode): Likewise.
18832         (default_libgcc_shift_count_mode): Likewise.
18833         (default_unwind_word_mode): Likewise.
18834         (default_addr_space_pointer_mode): Likewise.
18835         (default_addr_space_address_mode): Likewise.
18836         * targhooks.h (default_eh_return_filter_mode): Likewise.
18837         (default_libgcc_cmp_return_mode): Likewise.
18838         (default_libgcc_shift_count_mode): Likewise.
18839         (default_unwind_word_mode): Likewise.
18840         (default_addr_space_pointer_mode): Likewise.
18841         (default_addr_space_address_mode): Likewise.
18842         (default_mode_rep_extended): Take a scalar_int_mode rather than
18843         a machine_mode.
18844         (default_valid_pointer_mode): Likewise.
18845         (default_addr_space_valid_pointer_mode): Likewise.
18846         * tree-ssa-address.c (addr_for_mem_ref): Use scalar_int_mode for
18847         local variables.
18848         * tree-ssa-loop-ivopts.c (get_shiftadd_cost): Take a scalar_int_mode
18849         rather than a machine_mode.
18850         * tree-switch-conversion.c (array_value_type): Use scalar_int_mode
18851         for local variables.
18852         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
18853         * var-tracking.c (use_narrower_mode): Take a scalar_int_mode rather
18854         than a machine_mode.
18856 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18857             Alan Hayward  <alan.hayward@arm.com>
18858             David Sherwood  <david.sherwood@arm.com>
18860         * dojump.c (do_jump_by_parts_greater_rtx): Change the type of
18861         the mode argument to scalar_int_mode.
18862         (do_jump_by_parts_zero_rtx): Likewise.
18863         (do_jump_by_parts_equality_rtx): Likewise.
18864         (do_jump_by_parts_greater): Take a mode argument.
18865         (do_jump_by_parts_equality): Likewise.
18866         (do_jump_1): Update calls accordingly.
18868 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18869             Alan Hayward  <alan.hayward@arm.com>
18870             David Sherwood  <david.sherwood@arm.com>
18872         * is-a.h (safe_dyn_cast): New function.
18873         * rtl.h (rtx_jump_table_data::get_data_mode): New function.
18874         (jump_table_for_label): Likewise.
18875         * final.c (final_addr_vec_align): Take an rtx_jump_table_data *
18876         instead of an rtx_insn *.
18877         (shorten_branches): Use dyn_cast instead of LABEL_P and
18878         JUMP_TABLE_DATA_P.  Use jump_table_for_label and
18879         rtx_jump_table_data::get_data_mode.
18880         (final_scan_insn): Likewise.
18882 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18883             Alan Hayward  <alan.hayward@arm.com>
18884             David Sherwood  <david.sherwood@arm.com>
18886         * combine.c (try_combine): Use is_a <scalar_int_mode> when
18887         trying to combine a full-register integer set with a subreg
18888         integer set.
18890 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18891             Alan Hayward  <alan.hayward@arm.com>
18892             David Sherwood  <david.sherwood@arm.com>
18894         * expr.c (expand_expr_addr_expr): Add a new_tmode local variable
18895         that is always either address_mode or pointer_mode.
18897 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18898             Alan Hayward  <alan.hayward@arm.com>
18899             David Sherwood  <david.sherwood@arm.com>
18901         * expr.c (expand_expr_real_2): Use word_mode instead of innermode
18902         when the two are known to be equal.
18904 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18905             Alan Hayward  <alan.hayward@arm.com>
18906             David Sherwood  <david.sherwood@arm.com>
18908         * simplify-rtx.c (simplify_const_unary_operation): Use
18909         is_a <scalar_int_mode> instead of checking for a nonzero
18910         precision.  Forcibly convert op_mode to a scalar_int_mode
18911         in that case.  More clearly differentiate the operand and
18912         result modes and use the former when deciding what the value
18913         of a count-bits operation should be.  Use is_int_mode instead
18914         of checking for a MODE_INT.  Remove redundant check for whether
18915         this mode has a zero precision.
18917 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18918             Alan Hayward  <alan.hayward@arm.com>
18919             David Sherwood  <david.sherwood@arm.com>
18921         * optabs.c (widen_leading): Change the type of the mode argument
18922         to scalar_int_mode.  Use opt_scalar_int_mode for the mode iterator.
18923         (widen_bswap): Likewise.
18924         (expand_parity): Likewise.
18925         (expand_ctz): Change the type of the mode argument to scalar_int_mode.
18926         (expand_ffs): Likewise.
18927         (epand_unop): Check for scalar integer modes before calling the
18928         above routines.
18930 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18931             Alan Hayward  <alan.hayward@arm.com>
18932             David Sherwood  <david.sherwood@arm.com>
18934         * expr.c (const_scalar_mask_from_tree): Add a mode argument.
18935         Expand commentary.
18936         (expand_expr_real_1): Update call accordingly.
18938 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18939             Alan Hayward  <alan.hayward@arm.com>
18940             David Sherwood  <david.sherwood@arm.com>
18942         * expmed.c (store_bit_field_using_insv): Add op0_mode and
18943         value_mode arguments.  Use scalar_int_mode internally.
18944         (store_bit_field_1): Rename the new integer mode from imode
18945         to op0_mode and use it instead of GET_MODE (op0).  Update calls
18946         to store_split_bit_field, store_bit_field_using_insv and
18947         store_fixed_bit_field.
18948         (store_fixed_bit_field): Add op0_mode and value_mode arguments.
18949         Use scalar_int_mode internally.  Use a bit count rather than a mode
18950         when calculating the largest bit size for get_best_mode.
18951         Update calls to store_split_bit_field and store_fixed_bit_field_1.
18952         (store_fixed_bit_field_1): Add mode and value_mode arguments.
18953         Remove assertion that OP0 has a scalar integer mode.
18954         (store_split_bit_field): Add op0_mode and value_mode arguments.
18955         Update calls to extract_fixed_bit_field.
18956         (extract_bit_field_using_extv): Add an op0_mode argument.
18957         Use scalar_int_mode internally.
18958         (extract_bit_field_1): Rename the new integer mode from imode to
18959         op0_mode and use it instead of GET_MODE (op0).  Update calls to
18960         extract_split_bit_field, extract_bit_field_using_extv and
18961         extract_fixed_bit_field.
18962         (extract_fixed_bit_field): Add an op0_mode argument.  Update calls
18963         to extract_split_bit_field and extract_fixed_bit_field_1.
18964         (extract_fixed_bit_field_1): Add a mode argument.  Remove assertion
18965         that OP0 has a scalar integer mode.  Use as_a <scalar_int_mode>
18966         on the target mode.
18967         (extract_split_bit_field): Add an op0_mode argument.  Update call
18968         to extract_fixed_bit_field.
18970 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18971             Alan Hayward  <alan.hayward@arm.com>
18972             David Sherwood  <david.sherwood@arm.com>
18974         * cse.c (cse_insn): Use opt_scalar_int_mode for the mode iterator.
18975         * explow.c (hard_function_value): Likewise.
18976         * expmed.c (extract_fixed_bit_field_1): Likewise.  Move the
18977         convert_to_mode call outside the loop.
18978         * expr.c (alignment_for_piecewise_move): Use opt_scalar_int_mode
18979         for the mode iterator.  Require the mode specified by max_pieces
18980         to exist.
18981         (emit_block_move_via_movmem): Use opt_scalar_int_mode for the
18982         mode iterator.
18983         (copy_blkmode_to_reg): Likewise.
18984         (set_storage_via_setmem): Likewise.
18985         * optabs.c (prepare_cmp_insn): Likewise.
18986         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
18987         * stor-layout.c (finish_bitfield_representative): Likewise.
18989 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18990             Alan Hayward  <alan.hayward@arm.com>
18991             David Sherwood  <david.sherwood@arm.com>
18993         * rtl.h (subreg_unpromoted_mode, subreg_promoted_mode): New functions.
18994         * expr.c (convert_move): Use them.
18995         (convert_modes): Likewise.
18996         (store_expr_with_bounds): Likewise.
18998 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
18999             Alan Hayward  <alan.hayward@arm.com>
19000             David Sherwood  <david.sherwood@arm.com>
19002         * rtl.h (rtl_hooks::reg_nonzero_bits): Add a scalar_int_mode
19003         parameter for the mode of "x".  Remove the "known_x", "known_mode"
19004         and "known_ret" arguments.  Change the type of the mode argument
19005         to scalar_int_mode.
19006         (rtl_hooks:reg_num_sign_bit_copies): Likewise.
19007         * combine.c (reg_nonzero_bits_for_combine): Update accordingly.
19008         (reg_num_sign_bit_copies_for_combine): Likewise.
19009         * rtlanal.c (nonzero_bits1): Likewise.
19010         (num_sign_bit_copies1): Likewise.
19011         * rtlhooks-def.h (reg_nonzero_bits_general): Likewise.
19012         (reg_num_sign_bit_copies_general): Likewise.
19013         * rtlhooks.c (reg_num_sign_bit_copies_general): Likewise.
19014         (reg_nonzero_bits_general): Likewise.
19016 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19017             Alan Hayward  <alan.hayward@arm.com>
19018             David Sherwood  <david.sherwood@arm.com>
19020         * rtlanal.c (num_sign_bit_copies): Handle VOIDmode here rather
19021         than in subroutines.  Return 1 for non-integer modes.
19022         (cached_num_sign_bit_copies): Change the type of the mode parameter
19023         to scalar_int_mode.
19024         (num_sign_bit_copies1): Likewise.  Remove early exit for other mode
19025         classes.  Handle CONST_INT_P first and then check whether X also
19026         has a scalar integer mode.  Check the same thing for inner registers
19027         of a SUBREG and for values that are being extended or truncated.
19029 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19030             Alan Hayward  <alan.hayward@arm.com>
19031             David Sherwood  <david.sherwood@arm.com>
19033         * rtlanal.c (nonzero_bits): Handle VOIDmode here rather than
19034         in subroutines.  Return the mode mask for non-integer modes.
19035         (cached_nonzero_bits): Change the type of the mode parameter
19036         to scalar_int_mode.
19037         (nonzero_bits1): Likewise.  Remove early exit for other mode
19038         classes.  Handle CONST_INT_P first and then check whether X
19039         also has a scalar integer mode.
19041 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19042             Alan Hayward  <alan.hayward@arm.com>
19043             David Sherwood  <david.sherwood@arm.com>
19045         * expr.c (widest_int_mode_for_size): Make the comment match the code.
19046         Return a scalar_int_mode and assert that the size is greater than
19047         one byte.
19048         (by_pieces_ninsns): Update accordingly and remove VOIDmode handling.
19049         (op_by_pieces_d::op_by_pieces_d): Likewise.
19050         (op_by_pieces_d::run): Likewise.
19051         (can_store_by_pieces): Likewise.
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         * combine.c (extract_left_shift): Add a mode argument and update
19058         recursive calls.
19059         (make_compound_operation_int): Change the type of the mode parameter
19060         to scalar_int_mode and update the call to extract_left_shift.
19062 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19063             Alan Hayward  <alan.hayward@arm.com>
19064             David Sherwood  <david.sherwood@arm.com>
19066         * combine.c (simplify_and_const_int): Change the type of the mode
19067         parameter to scalar_int_mode.
19068         (simplify_and_const_int_1): Likewise.  Update recursive call.
19070 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19071             Alan Hayward  <alan.hayward@arm.com>
19072             David Sherwood  <david.sherwood@arm.com>
19074         * combine.c (simplify_compare_const): Check that the mode is a
19075         scalar_int_mode (rather than VOIDmode) before testing its
19076         precision.
19077         (simplify_comparison): Move COMPARISON_P handling out of the
19078         loop and restrict the latter part of the loop to scalar_int_modes.
19079         Check is_a <scalar_int_mode> before calling HWI_COMPUTABLE_MODE_P
19080         and when considering SUBREG_REGs.  Use is_int_mode instead of
19081         checking GET_MODE_CLASS against MODE_INT.
19083 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19084             Alan Hayward  <alan.hayward@arm.com>
19085             David Sherwood  <david.sherwood@arm.com>
19087         * combine.c (try_widen_shift_mode): Move check for equal modes to...
19088         (simplify_shift_const_1): ...here.  Use scalar_int_mode for
19089         shift_unit_mode and for modes involved in scalar shifts.
19091 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19092             Alan Hayward  <alan.hayward@arm.com>
19093             David Sherwood  <david.sherwood@arm.com>
19095         * combine.c (force_int_to_mode): New function, split out from...
19096         (force_to_mode): ...here.  Keep xmode up-to-date and use it
19097         instead of GET_MODE (x).
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         * optabs-query.h (extraction_insn::struct_mode): Change type to
19104         opt_scalar_int_mode and update comment.
19105         (extraction_insn::field_mode): Change type to scalar_int_mode.
19106         (extraction_insn::pos_mode): Likewise.
19107         * combine.c (make_extraction): Update accordingly.
19108         * optabs-query.c (get_traditional_extraction_insn): Likewise.
19109         (get_optab_extraction_insn): Likewise.
19110         * recog.c (simplify_while_replacing): Likewise.
19111         * expmed.c (narrow_bit_field_mem): Change the type of the mode
19112         parameter to opt_scalar_int_mode.
19114 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19115             Alan Hayward  <alan.hayward@arm.com>
19116             David Sherwood  <david.sherwood@arm.com>
19118         * machmode.h (bit_field_mode_iterator::next_mode): Take a pointer
19119         to a scalar_int_mode instead of a machine_mode.
19120         (bit_field_mode_iterator::m_mode): Change type to opt_scalar_int_mode.
19121         (get_best_mode): Return a boolean and use a pointer argument to store
19122         the selected mode.  Replace the limit mode parameter with a bit limit.
19123         * expmed.c (adjust_bit_field_mem_for_reg): Use scalar_int_mode
19124         for the values returned by bit_field_mode_iterator::next_mode.
19125         (store_bit_field): Update call to get_best_mode.
19126         (store_fixed_bit_field): Likewise.
19127         (extract_fixed_bit_field): Likewise.
19128         * expr.c (optimize_bitfield_assignment_op): Likewise.
19129         * fold-const.c (optimize_bit_field_compare): Likewise.
19130         (fold_truth_andor_1): Likewise.
19131         * stor-layout.c (bit_field_mode_iterator::next_mode): As above.
19132         Update for new type of m_mode.
19133         (get_best_mode): As above.
19135 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19136             Alan Hayward  <alan.hayward@arm.com>
19137             David Sherwood  <david.sherwood@arm.com>
19139         * expmed.c (strict_volatile_bitfield_p): Change the type of fieldmode
19140         to scalar_int_mode.  Remove check for SCALAR_INT_MODE_P.
19141         (store_bit_field): Check is_a <scalar_int_mode> before calling
19142         strict_volatile_bitfield_p.
19143         (extract_bit_field): Likewise.
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         * target.def (cstore_mode): Return a scalar_int_mode.
19150         * doc/tm.texi: Regenerate.
19151         * config/sparc/sparc.c (sparc_cstore_mode): Return a scalar_int_mode.
19152         * targhooks.h (default_cstore_mode): Likewise.
19153         * targhooks.c (default_cstore_mode): Likewise, using a forced
19154         conversion.
19155         * expmed.c (emit_cstore): Expect the target of the cstore to be
19156         a scalar_int_mode.
19158 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19159             Alan Hayward  <alan.hayward@arm.com>
19160             David Sherwood  <david.sherwood@arm.com>
19162         * cfgloop.h (rtx_iv): Change type of extend_mode and mode to
19163         scalar_int_mode.
19164         (niter_desc): Likewise mode.
19165         (iv_analyze): Add a mode parameter.
19166         (biv_p): Likewise.
19167         (iv_analyze_expr): Pass the mode paraeter before the rtx it describes
19168         and change its type to scalar_int_mode.
19169         * loop-iv.c: Update commentary at head of file.
19170         (iv_constant): Pass the mode paraeter before the rtx it describes
19171         and change its type to scalar_int_mode.  Remove VOIDmode handling.
19172         (iv_subreg): Change the type of the mode parameter to scalar_int_mode.
19173         (iv_extend): Likewise.
19174         (shorten_into_mode): Likewise.
19175         (iv_add): Use scalar_int_mode.
19176         (iv_mult): Likewise.
19177         (iv_shift): Likewise.
19178         (canonicalize_iv_subregs): Likewise.
19179         (get_biv_step_1): Pass the outer_mode parameter before the rtx
19180         it describes and change its mode to scalar_int_mode.   Also change
19181         the type of the returned inner_mode to scalar_int_mode.
19182         (get_biv_step): Likewise, turning outer_mode from a pointer
19183         into a direct parameter.  Update call to get_biv_step_1.
19184         (iv_analyze_biv): Add an outer_mode parameter.  Update calls to
19185         iv_constant and get_biv_step.
19186         (iv_analyze_expr): Pass the mode parameter before the rtx it describes
19187         and change its type to scalar_int_mode.  Don't initialise iv->mode
19188         to VOIDmode and remove later checks for its still being VOIDmode.
19189         Update calls to iv_analyze_op and iv_analyze_expr.  Check
19190         is_a <scalar_int_mode> when changing the mode under consideration.
19191         (iv_analyze_def): Ignore registers that don't have a scalar_int_mode.
19192         Update call to iv_analyze_expr.
19193         (iv_analyze_op): Add a mode parameter.  Reject subregs whose
19194         inner register is not also a scalar_int_mode.  Update call to
19195         iv_analyze_biv.
19196         (iv_analyze): Add a mode parameter.  Update call to iv_analyze_op.
19197         (biv_p): Add a mode parameter.  Update call to iv_analyze_biv.
19198         (iv_number_of_iterations): Use is_a <scalar_int_mode> instead of
19199         separate mode class checks.  Update calls to iv_analyze.  Remove
19200         fix-up of VOIDmodes after iv_analyze_biv.
19201         * loop-unroll.c (analyze_iv_to_split_insn): Reject registers that
19202         don't have a scalar_int_mode.  Update call to biv_p.
19204 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19205             Alan Hayward  <alan.hayward@arm.com>
19206             David Sherwood  <david.sherwood@arm.com>
19208         * cfgexpand.c (convert_debug_memory_address): Use
19209         as_a <scalar_int_mode>.
19210         * combine.c (expand_compound_operation): Likewise.
19211         (make_extraction): Likewise.
19212         (change_zero_ext): Likewise.
19213         (simplify_comparison): Likewise.
19214         * cse.c (cse_insn): Likewise.
19215         * dwarf2out.c (minmax_loc_descriptor): Likewise.
19216         (mem_loc_descriptor): Likewise.
19217         (loc_descriptor): Likewise.
19218         * expmed.c (init_expmed_one_mode): Likewise.
19219         (synth_mult): Likewise.
19220         (emit_store_flag_1): Likewise.
19221         (expand_divmod): Likewise.  Use HWI_COMPUTABLE_MODE_P instead
19222         of a comparison with size.
19223         * expr.c (expand_assignment): Use as_a <scalar_int_mode>.
19224         (reduce_to_bit_field_precision): Likewise.
19225         * function.c (expand_function_end): Likewise.
19226         * internal-fn.c (expand_arith_overflow_result_store): Likewise.
19227         * loop-doloop.c (doloop_modify): Likewise.
19228         * optabs.c (expand_binop): Likewise.
19229         (expand_unop): Likewise.
19230         (expand_copysign_absneg): Likewise.
19231         (prepare_cmp_insn): Likewise.
19232         (maybe_legitimize_operand): Likewise.
19233         * recog.c (const_scalar_int_operand): Likewise.
19234         * rtlanal.c (get_address_mode): Likewise.
19235         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19236         (simplify_cond_clz_ctz): Likewise.
19237         * tree-nested.c (get_nl_goto_field): Likewise.
19238         * tree.c (build_vector_type_for_mode): Likewise.
19239         * var-tracking.c (use_narrower_mode): Likewise.
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         * tree.h (SCALAR_INT_TYPE_MODE): New macro.
19246         * builtins.c (expand_builtin_signbit): Use it.
19247         * cfgexpand.c (expand_debug_expr): Likewise.
19248         * dojump.c (do_jump): Likewise.
19249         (do_compare_and_jump): Likewise.
19250         * dwarf2cfi.c (expand_builtin_init_dwarf_reg_sizes): Likewise.
19251         * expmed.c (make_tree): Likewise.
19252         * expr.c (expand_expr_real_2): Likewise.
19253         (expand_expr_real_1): Likewise.
19254         (try_casesi): Likewise.
19255         * fold-const-call.c (fold_const_call_ss): Likewise.
19256         * fold-const.c (unextend): Likewise.
19257         (extract_muldiv_1): Likewise.
19258         (fold_single_bit_test): Likewise.
19259         (native_encode_int): Likewise.
19260         (native_encode_string): Likewise.
19261         (native_interpret_int): Likewise.
19262         * gimple-fold.c (gimple_fold_builtin_memset): Likewise.
19263         * internal-fn.c (expand_addsub_overflow): Likewise.
19264         (expand_neg_overflow): Likewise.
19265         (expand_mul_overflow): Likewise.
19266         (expand_arith_overflow): Likewise.
19267         * match.pd: Likewise.
19268         * stor-layout.c (layout_type): Likewise.
19269         * tree-cfg.c (verify_gimple_assign_ternary): Likewise.
19270         * tree-ssa-math-opts.c (convert_mult_to_widen): Likewise.
19271         * tree-ssanames.c (get_range_info): Likewise.
19272         * tree-switch-conversion.c (array_value_type) Likewise.
19273         * tree-vect-patterns.c (vect_recog_rotate_pattern): Likewise.
19274         (vect_recog_divmod_pattern): Likewise.
19275         (vect_recog_mixed_size_cond_pattern): Likewise.
19276         * tree-vrp.c (extract_range_basic): Likewise.
19277         (simplify_float_conversion_using_ranges): Likewise.
19278         * tree.c (int_fits_type_p): Likewise.
19279         * ubsan.c (instrument_bool_enum_load): Likewise.
19280         * varasm.c (mergeable_string_section): Likewise.
19281         (narrowing_initializer_constant_valid_p): Likewise.
19282         (output_constant): Likewise.
19284 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19285             Alan Hayward  <alan.hayward@arm.com>
19286             David Sherwood  <david.sherwood@arm.com>
19288         * machmode.h (NARROWEST_INT_MODE): New macro.
19289         * expr.c (alignment_for_piecewise_move): Use it instead of
19290         GET_CLASS_NARROWEST_MODE (MODE_INT).
19291         (push_block): Likewise.
19292         * stor-layout.c (bit_field_mode_iterator::bit_field_mode_iterator):
19293         Likewise.
19294         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19296 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19297             Alan Hayward  <alan.hayward@arm.com>
19298             David Sherwood  <david.sherwood@arm.com>
19300         * postreload.c (move2add_valid_value_p): Change the type of the
19301         mode parameter to scalar_int_mode.
19302         (move2add_use_add2_insn): Add a mode parameter and use it instead
19303         of GET_MODE (reg).
19304         (move2add_use_add3_insn): Likewise.
19305         (reload_cse_move2add): Update accordingly.
19307 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19308             Alan Hayward  <alan.hayward@arm.com>
19309             David Sherwood  <david.sherwood@arm.com>
19311         * expr.c (expand_expr_real_2): Use scalar_int_mode for the
19312         double-word mode.
19313         * lower-subreg.c (resolve_shift_zext): Use is_a <scalar_int_mode>.
19314         * optabs.c (expand_unop): Likewise.
19316 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19317             Alan Hayward  <alan.hayward@arm.com>
19318             David Sherwood  <david.sherwood@arm.com>
19320         * dwarf2out.c (typed_binop): Change mode parameter to scalar_int_mode.
19321         (clz_loc_descriptor): Likewise.  Remove SCALAR_INT_MODE_P check.
19322         (popcount_loc_descriptor): Likewise.
19323         (bswap_loc_descriptor): Likewise.
19324         (rotate_loc_descriptor): Likewise.
19325         (mem_loc_descriptor): Add is_a <scalar_int_mode> checks before
19326         calling the functions above.
19328 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19329             Alan Hayward  <alan.hayward@arm.com>
19330             David Sherwood  <david.sherwood@arm.com>
19332         * combine.c (sign_extend_short_imm): Add is_a <scalar_int_mode>
19333         checks.
19334         (try_combine): Likewise.
19335         (simplify_if_then_else): Likewise.
19336         * cse.c (cse_insn): Likewise.
19337         * dwarf2out.c (mem_loc_descriptor): Likewise.
19338         * emit-rtl.c (gen_lowpart_common): Likewise.
19339         * simplify-rtx.c (simplify_truncation): Likewise.
19340         (simplify_binary_operation_1): Likewise.
19341         (simplify_const_relational_operation): Likewise.
19342         (simplify_ternary_operation): Likewise.
19343         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Likewise.
19345 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19346             Alan Hayward  <alan.hayward@arm.com>
19347             David Sherwood  <david.sherwood@arm.com>
19349         * cse.c (cse_insn): Add is_a <scalar_int_mode> checks.
19350         * reload.c (push_reload): Likewise.
19351         (find_reloads): Likewise.
19353 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19354             Alan Hayward  <alan.hayward@arm.com>
19355             David Sherwood  <david.sherwood@arm.com>
19357         * combine.c (find_split_point): Add is_a <scalar_int_mode> checks.
19358         (make_compound_operation_int): Likewise.
19359         (change_zero_ext): Likewise.
19360         * expr.c (convert_move): Likewise.
19361         (convert_modes): Likewise.
19362         * fwprop.c (forward_propagate_subreg): Likewise.
19363         * loop-iv.c (get_biv_step_1): Likewise.
19364         * optabs.c (widen_operand): Likewise.
19365         * postreload.c (move2add_valid_value_p): Likewise.
19366         * recog.c (simplify_while_replacing): Likewise.
19367         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19368         (simplify_binary_operation_1): Likewise.  Remove redundant
19369         mode equality check.
19371 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19372             Alan Hayward  <alan.hayward@arm.com>
19373             David Sherwood  <david.sherwood@arm.com>
19375         * combine.c (combine_simplify_rtx): Add checks for
19376         is_a <scalar_int_mode>.
19377         (simplify_if_then_else): Likewise.
19378         (make_field_assignment): Likewise.
19379         (simplify_comparison): Likewise.
19380         * ifcvt.c (noce_try_bitop): Likewise.
19381         * loop-invariant.c (canonicalize_address_mult): Likewise.
19382         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19384 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19385             Alan Hayward  <alan.hayward@arm.com>
19386             David Sherwood  <david.sherwood@arm.com>
19388         * gimple-fold.c (gimple_fold_builtin_memory_op): Use
19389         is_a <scalar_int_mode> instead of != BLKmode.
19391 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19392             Alan Hayward  <alan.hayward@arm.com>
19393             David Sherwood  <david.sherwood@arm.com>
19395         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>
19396         instead of != VOIDmode.
19397         * combine.c (if_then_else_cond): Likewise.
19398         (change_zero_ext): Likewise.
19399         * dwarf2out.c (mem_loc_descriptor): Likewise.
19400         (loc_descriptor): Likewise.
19401         * rtlanal.c (canonicalize_condition): Likewise.
19402         * simplify-rtx.c (simplify_relational_operation_1): Likewise.
19404 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19405             Alan Hayward  <alan.hayward@arm.com>
19406             David Sherwood  <david.sherwood@arm.com>
19408         * simplify-rtx.c (simplify_binary_operation_1): Use
19409         is_a <scalar_int_mode> instead of !VECTOR_MODE_P.
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         * wide-int.h (int_traits<unsigned char>) New class.
19416         (int_traits<unsigned short>) Likewise.
19417         * cfgexpand.c (expand_debug_expr): Use is_a <scalar_int_mode>.
19418         Use GET_MODE_UNIT_PRECISION and remove redundant test for
19419         SCALAR_INT_MODE_P.
19420         * combine.c (set_nonzero_bits_and_sign_copies): Use
19421         is_a <scalar_int_mode>.
19422         (find_split_point): Likewise.
19423         (combine_simplify_rtx): Likewise.
19424         (simplify_logical): Likewise.
19425         (expand_compound_operation): Likewise.
19426         (expand_field_assignment): Likewise.
19427         (make_compound_operation): Likewise.
19428         (extended_count): Likewise.
19429         (change_zero_ext): Likewise.
19430         (simplify_comparison): Likewise.
19431         * dwarf2out.c (scompare_loc_descriptor): Likewise.
19432         (ucompare_loc_descriptor): Likewise.
19433         (minmax_loc_descriptor): Likewise.
19434         (mem_loc_descriptor): Likewise.
19435         (loc_descriptor): Likewise.
19436         * expmed.c (init_expmed_one_mode): Likewise.
19437         * lra-constraints.c (lra_constraint_offset): Likewise.
19438         * optabs.c (prepare_libcall_arg): Likewise.
19439         * postreload.c (move2add_note_store): Likewise.
19440         * reload.c (operands_match_p): Likewise.
19441         * rtl.h (load_extend_op): Likewise.
19442         * rtlhooks.c (gen_lowpart_general): Likewise.
19443         * simplify-rtx.c (simplify_truncation): Likewise.
19444         (simplify_unary_operation_1): Likewise.
19445         (simplify_binary_operation_1): Likewise.
19446         (simplify_const_binary_operation): Likewise.
19447         (simplify_const_relational_operation): Likewise.
19448         (simplify_subreg): Likewise.
19449         * stor-layout.c (bitwise_mode_for_mode): Likewise.
19450         * var-tracking.c (adjust_mems): Likewise.
19451         (prepare_call_arguments): Likewise.
19453 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19454             Alan Hayward  <alan.hayward@arm.com>
19455             David Sherwood  <david.sherwood@arm.com>
19457         * machmode.h (is_int_mode): New fuction.
19458         * combine.c (find_split_point): Use it.
19459         (combine_simplify_rtx): Likewise.
19460         (simplify_if_then_else): Likewise.
19461         (simplify_set): Likewise.
19462         (simplify_shift_const_1): Likewise.
19463         (simplify_comparison): Likewise.
19464         * config/aarch64/aarch64.c (aarch64_rtx_costs): Likewise.
19465         * cse.c (notreg_cost): Likewise.
19466         (cse_insn): Likewise.
19467         * cselib.c (cselib_lookup_1): Likewise.
19468         * dojump.c (do_jump_1): Likewise.
19469         (do_compare_rtx_and_jump): Likewise.
19470         * dse.c (get_call_args): Likewise.
19471         * dwarf2out.c (rtl_for_decl_init): Likewise.
19472         (native_encode_initializer): Likewise.
19473         * expmed.c (emit_store_flag_1): Likewise.
19474         (emit_store_flag): Likewise.
19475         * expr.c (convert_modes): Likewise.
19476         (store_field): Likewise.
19477         (expand_expr_real_1): Likewise.
19478         * fold-const.c (fold_read_from_constant_string): Likewise.
19479         * gimple-ssa-sprintf.c (get_format_string): Likewise.
19480         * optabs-libfuncs.c (gen_int_libfunc): Likewise.
19481         * optabs.c (expand_binop): Likewise.
19482         (expand_unop): Likewise.
19483         (expand_abs_nojump): Likewise.
19484         (expand_one_cmpl_abs_nojump): Likewise.
19485         * simplify-rtx.c (mode_signbit_p): Likewise.
19486         (val_signbit_p): Likewise.
19487         (val_signbit_known_set_p): Likewise.
19488         (val_signbit_known_clear_p): Likewise.
19489         (simplify_relational_operation_1): Likewise.
19490         * tree.c (vector_type_mode): Likewise.
19492 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19493             Alan Hayward  <alan.hayward@arm.com>
19494             David Sherwood  <david.sherwood@arm.com>
19496         * machmode.h (smallest_mode_for_size): Fix formatting.
19497         (smallest_int_mode_for_size): New function.
19498         * cfgexpand.c (expand_debug_expr): Use smallest_int_mode_for_size
19499         instead of smallest_mode_for_size.
19500         * combine.c (make_extraction): Likewise.
19501         * config/arc/arc.c (arc_expand_movmem): Likewise.
19502         * config/arm/arm.c (arm_expand_divmod_libfunc): Likewise.
19503         * config/i386/i386.c (ix86_get_mask_mode): Likewise.
19504         * config/s390/s390.c (s390_expand_insv): Likewise.
19505         * config/sparc/sparc.c (assign_int_registers): Likewise.
19506         * config/spu/spu.c (spu_function_value): Likewise.
19507         (spu_function_arg): Likewise.
19508         * coverage.c (get_gcov_type): Likewise.
19509         (get_gcov_unsigned_t): Likewise.
19510         * dse.c (find_shift_sequence): Likewise.
19511         * expmed.c (store_bit_field_1): Likewise.
19512         * expr.c (convert_move): Likewise.
19513         (store_field): Likewise.
19514         * internal-fn.c (expand_arith_overflow): Likewise.
19515         * optabs-query.c (get_best_extraction_insn): Likewise.
19516         * optabs.c (expand_twoval_binop_libfunc): Likewise.
19517         * stor-layout.c (layout_type): Likewise.
19518         (initialize_sizetypes): Likewise.
19519         * targhooks.c (default_get_mask_mode): Likewise.
19520         * tree-ssa-loop-manip.c (canonicalize_loop_ivs): Likewise.
19522 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19523             Alan Hayward  <alan.hayward@arm.com>
19524             David Sherwood  <david.sherwood@arm.com>
19526         * machmode.h (opt_mode::else_blk): New function.
19527         (int_mode_for_mode): Declare.
19528         * stor-layout.c (int_mode_for_mode): Return an opt_scalar_int_mode.
19529         * builtins.c (expand_builtin_signbit): Adjust for new int_mode_for_mode
19530         return type.
19531         * cfgexpand.c (expand_debug_expr): Likewise.
19532         * combine.c (gen_lowpart_or_truncate): Likewise.
19533         (gen_lowpart_for_combine): Likewise.
19534         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Likewise.
19535         * config/avr/avr.c (avr_to_int_mode): Likewise.
19536         (avr_out_plus_1): Likewise.
19537         (avr_out_plus): Likewise.
19538         (avr_out_round): Likewise.
19539         * config/i386/i386.c (ix86_split_to_parts): Likewise.
19540         * config/s390/s390.c (s390_expand_vec_compare_cc): Likewise.
19541         (s390_expand_vcond): Likewise.
19542         * config/spu/spu.c (spu_split_immediate): Likewise.
19543         (spu_expand_mov): Likewise.
19544         * dse.c (get_stored_val): Likewise.
19545         * expmed.c (store_bit_field_1): Likewise.
19546         (convert_extracted_bit_field): Use int_mode_for_mode instead of
19547         int_mode_for_size.
19548         (extract_bit_field_1): Adjust for new int_mode_for_mode return type.
19549         (extract_low_bits): Likewise.
19550         * expr.c (emit_group_load_1): Likewise.  Separate out the BLKmode
19551         handling rather than repeating the check.
19552         (emit_group_store): Likewise.
19553         (emit_move_via_integer): Adjust for new int_mode_for_mode return type.
19554         * optabs.c (expand_absneg_bit): Likewise.
19555         (expand_copysign_absneg): Likewise.
19556         (expand_copysign_bit): Likewise.
19557         * tree-if-conv.c (ifcvt_can_use_mask_load_store): Likewise.
19558         * tree-vect-slp.c (vect_transform_slp_perm_load): Likewise.
19559         * tree-vect-stmts.c (vect_gen_perm_mask_any): Likewise.
19560         * var-tracking.c (prepare_call_arguments):  Likewise.
19561         * config/powerpcspe/powerpcspe.c (rs6000_do_expand_vec_perm): Use
19562         int_mode_for_mode instead of mode_for_size.
19563         * config/rs6000/rs6000.c (rs6000_do_expand_vec_perm): Likewise.
19565 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19566             Alan Hayward  <alan.hayward@arm.com>
19567             David Sherwood  <david.sherwood@arm.com>
19569         * machmode.h (int_mode_for_size): New function.
19570         * builtins.c (set_builtin_user_assembler_name): Use int_mode_for_size
19571         instead of mode_for_size.
19572         * calls.c (save_fixed_argument_area): Likewise.  Make use of BLKmode
19573         explicit.
19574         * combine.c (expand_field_assignment): Use int_mode_for_size
19575         instead of mode_for_size.
19576         (make_extraction): Likewise.
19577         (simplify_shift_const_1): Likewise.
19578         (simplify_comparison): Likewise.
19579         * dojump.c (do_jump): Likewise.
19580         * dwarf2out.c (mem_loc_descriptor): Likewise.
19581         * emit-rtl.c (init_derived_machine_modes): Likewise.
19582         * expmed.c (flip_storage_order): Likewise.
19583         (convert_extracted_bit_field): Likewise.
19584         * expr.c (copy_blkmode_from_reg): Likewise.
19585         * graphite-isl-ast-to-gimple.c (max_mode_int_precision): Likewise.
19586         * internal-fn.c (expand_mul_overflow): Likewise.
19587         * lower-subreg.c (simple_move): Likewise.
19588         * optabs-libfuncs.c (init_optabs): Likewise.
19589         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
19590         * tree.c (vector_type_mode): Likewise.
19591         * tree-ssa-strlen.c (handle_builtin_memcmp): Likewise.
19592         * tree-vect-data-refs.c (vect_lanes_optab_supported_p): Likewise.
19593         * tree-vect-generic.c (expand_vector_parallel): Likewise.
19594         * tree-vect-stmts.c (vectorizable_load): Likewise.
19595         (vectorizable_store): Likewise.
19597 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19598             Alan Hayward  <alan.hayward@arm.com>
19599             David Sherwood  <david.sherwood@arm.com>
19601         * coretypes.h (pod_mode): New type.
19602         (scalar_int_mode_pod): New typedef.
19603         * machmode.h (pod_mode): New class.
19604         (int_n_data_t::m): Change type to scalar_int_mode_pod.
19605         * genmodes.c (emit_mode_int_n): Update accordingly.
19606         * lower-subreg.h (target_lower_subreg): Change type to
19607         scalar_int_mode_pod.
19608         * gdbhooks.py (build_pretty_printer): Handle pod_mode and
19609         scalar_int_mode_pod.
19611 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19612             Alan Hayward  <alan.hayward@arm.com>
19613             David Sherwood  <david.sherwood@arm.com>
19615         * config/powerpcspe/powerpcspe.h (rs6000_pmode): Change type from
19616         machine_mode to scalar_int_mode.
19617         * config/powerpcspe/powerpcspe.c (rs6000_pmode): Likewise.
19618         (rs6000_option_override_internal): Remove cast to int.
19619         * config/rs6000/rs6000.h (rs6000_pmode): Change type from
19620         machine_mode to scalar_int_mode.
19621         * config/rs6000/rs6000.c (rs6000_pmode): Likewise.
19622         (rs6000_option_override_internal): Remove cast to int.
19623         * config/s390/s390.h (Pmode): Remove cast to machine_mode.
19624         * config/epiphany/epiphany.h (RTX_OK_FOR_OFFSET_P): Add cast
19625         to machine_mode.
19626         * config/s390/s390.c (s390_expand_builtin): Likewise.
19627         * coretypes.h (scalar_int_mode): New type.
19628         (opt_scalar_int_mode): New typedef.
19629         * machmode.h (scalar_int_mode): New class.
19630         (scalar_int_mode::includes_p): New function.
19631         (byte_mode): Change type to scalar_int_mode.
19632         (word_mode): Likewise.
19633         (ptr_mode): Likewise.
19634         * emit-rtl.c (byte_mode): Likewise.
19635         (word_mode): Likewise.
19636         (ptr_mode): Likewise.
19637         (init_derived_machine_modes): Update accordingly.
19638         * genmodes.c (get_mode_class): Return scalar_int_mode for MODE_INT
19639         and MODE_PARTIAL_INT.
19640         * gdbhooks.py (build_pretty_printer): Handle scalar_int_mode and
19641         opt_scalar_int_mode.
19643 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19644             Alan Hayward  <alan.hayward@arm.com>
19645             David Sherwood  <david.sherwood@arm.com>
19647         * target.def (libgcc_floating_mode_supported_p): Take a
19648         scalar_float_mode.
19649         * doc/tm.texi: Regenerate.
19650         * targhooks.h (default_libgcc_floating_mode_supported_p): Take a
19651         scalar_float_mode.
19652         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
19653         * config/aarch64/aarch64.c (aarch64_libgcc_floating_mode_supported_p):
19654         Likewise.
19656 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19657             Alan Hayward  <alan.hayward@arm.com>
19658             David Sherwood  <david.sherwood@arm.com>
19660         * target.def (default_floatn_mode): Return an opt_scalar_float_mode.
19661         * doc/tm.texi: Regenerate.
19662         * config/arm/arm.c (arm_floatn_mode): Return an opt_scalar_float_mode.
19663         * config/powerpcspe/powerpcspe.c (rs6000_floatn_mode): Likewise.
19664         * config/rs6000/rs6000.c (rs6000_floatn_mode): Likewise.
19665         * targhooks.h (default_floatn_mode): Likewise.
19666         * targhooks.c (default_floatn_mode): Likewise.
19667         * tree.c (build_common_tree_nodes): Update accordingly.
19669 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19670             Alan Hayward  <alan.hayward@arm.com>
19671             David Sherwood  <david.sherwood@arm.com>
19673         * machmode.h (mode_iterator::start): Provide overload for opt_modes.
19674         (mode_iterator::iterate_p): Likewise.
19675         (mode_iterator::get_wider): Likewise.
19676         * expr.c (init_expr_target): Use opt_scalar_float_mode.
19678 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19679             Alan Hayward  <alan.hayward@arm.com>
19680             David Sherwood  <david.sherwood@arm.com>
19682         * coretypes.h (opt_scalar_float_mode): New typedef.
19683         * machmode.h (float_mode_for_size): New function.
19684         * emit-rtl.c (double_mode): Delete.
19685         (init_emit_once): Use float_mode_for_size.
19686         * stor-layout.c (layout_type): Likewise.
19687         * gdbhooks.py (build_pretty_printer): Handle opt_scalar_float_mode.
19689 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19690             Alan Hayward  <alan.hayward@arm.com>
19691             David Sherwood  <david.sherwood@arm.com>
19693         * output.h (assemble_real): Take a scalar_float_mode.
19694         * config/arm/arm.c (arm_assemble_integer): Update accordingly.
19695         * config/arm/arm.md (consttable_4): Likewise.
19696         (consttable_8): Likewise.
19697         (consttable_16): Likewise.
19698         * config/mips/mips.md (consttable_float): Likewise.
19699         * config/s390/s390.c (s390_output_pool_entry): Likewise.
19700         * varasm.c (assemble_real): Take a scalar_float_mode.
19701         (output_constant_pool_2): Update accordingly.
19702         (output_constant): Likewise.
19704 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19705             Alan Hayward  <alan.hayward@arm.com>
19706             David Sherwood  <david.sherwood@arm.com>
19708         * tree.h (SCALAR_FLOAT_TYPE_MODE): New macro.
19709         * builtins.c (expand_builtin_signbit): Use it instead of TYPE_MODE.
19710         * fold-const.c (fold_convert_const_real_from_fixed): Likewise.
19711         (native_encode_real): Likewise.
19712         (native_interpret_real): Likewise.
19713         * hsa-brig.c (emit_immediate_scalar_to_buffer): Likewise.
19714         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19716 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19717             Alan Hayward  <alan.hayward@arm.com>
19718             David Sherwood  <david.sherwood@arm.com>
19720         * optabs-libfuncs.c (gen_trunc_conv_libfunc): Use is_a
19721         <scalar_float_mode>.  Simplify.
19722         (gen_extend_conv_libfunc): Likewise.
19724 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19725             Alan Hayward  <alan.hayward@arm.com>
19726             David Sherwood  <david.sherwood@arm.com>
19728         * coretypes.h (scalar_float_mode): New type.
19729         * machmode.h (mode_traits::from_int): Use machine_mode if
19730         USE_ENUM_MODES is defined.
19731         (is_a): New function.
19732         (as_a): Likewise.
19733         (dyn_cast): Likewise.
19734         (scalar_float_mode): New class.
19735         (scalar_float_mode::includes_p): New function.
19736         (is_float_mode): Likewise.
19737         * gdbhooks.py (MachineModePrinter): New class.
19738         (build_pretty_printer): Use it for scalar_float_mode.
19739         * real.h (FLOAT_MODE_FORMAT): Use as_a <scalar_float_mode>.
19740         (format_helper::format_helper): Turn into a template.
19741         * genmodes.c (get_mode_class): New function.
19742         (emit_insn_modes_h): Give modes the class returned by get_mode_class,
19743         or machine_mode if none.
19744         * config/aarch64/aarch64.c (aarch64_simd_valid_immediate): Use
19745         as_a <scalar_float_mode>.
19746         * dwarf2out.c (mem_loc_descriptor): Likewise.
19747         (insert_float): Likewise.
19748         (add_const_value_attribute): Likewise.
19749         * simplify-rtx.c (simplify_immed_subreg): Likewise.
19750         * optabs.c (expand_absneg_bit): Take a scalar_float_mode.
19751         (expand_unop): Update accordingly.
19752         (expand_abs_nojump): Likewise.
19753         (expand_copysign_absneg): Take a scalar_float_mode.
19754         (expand_copysign_bit): Likewise.
19755         (expand_copysign): Update accordingly.
19757 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19758             Alan Hayward  <alan.hayward@arm.com>
19759             David Sherwood  <david.sherwood@arm.com>
19761         * coretypes.h (opt_mode): New class.
19762         * machmode.h (opt_mode): Likewise.
19763         (opt_mode::else_void): New function.
19764         (opt_mode::require): Likewise.
19765         (opt_mode::exists): Likewise.
19766         (GET_MODE_WIDER_MODE): Turn into a function and return an opt_mode.
19767         (GET_MODE_2XWIDER_MODE): Likewise.
19768         (mode_iterator::get_wider): Update accordingly.
19769         (mode_iterator::get_2xwider): Likewise.
19770         (mode_iterator::get_known_wider): Likewise, turning into a template.
19771         * combine.c (make_extraction): Update use of GET_MODE_WIDER_MODE,
19772         forcing a wider mode to exist.
19773         * config/cr16/cr16.h (LONG_REG_P): Likewise.
19774         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19775         * config/c6x/c6x.c (c6x_rtx_costs): Update use of
19776         GET_MODE_2XWIDER_MODE, forcing a wider mode to exist.
19777         * lower-subreg.c (init_lower_subreg): Likewise.
19778         * optabs-libfuncs.c (init_sync_libfuncs_1): Likewise, but not
19779         on the final iteration.
19780         * config/i386/i386.c (ix86_expand_set_or_movmem): Check whether
19781         a wider mode exists before asking for a move pattern.
19782         (get_mode_wider_vector): Update use of GET_MODE_WIDER_MODE,
19783         forcing a wider mode to exist.
19784         (expand_vselect_vconcat): Update use of GET_MODE_2XWIDER_MODE,
19785         returning false if no such mode exists.
19786         * config/ia64/ia64.c (expand_vselect_vconcat): Likewise.
19787         * config/mips/mips.c (mips_expand_vselect_vconcat): Likewise.
19788         * expmed.c (init_expmed_one_mode): Update use of GET_MODE_WIDER_MODE.
19789         Avoid checking for a MODE_INT if we already know the mode is not a
19790         SCALAR_INT_MODE_P.
19791         (extract_high_half): Update use of GET_MODE_WIDER_MODE,
19792         forcing a wider mode to exist.
19793         (expmed_mult_highpart_optab): Likewise.
19794         (expmed_mult_highpart): Likewise.
19795         * expr.c (expand_expr_real_2): Update use of GET_MODE_WIDER_MODE,
19796         using else_void.
19797         * lto-streamer-in.c (lto_input_mode_table): Likewise.
19798         * optabs-query.c (find_widening_optab_handler_and_mode): Likewise.
19799         * stor-layout.c (bit_field_mode_iterator::next_mode): Likewise.
19800         * internal-fn.c (expand_mul_overflow): Update use of
19801         GET_MODE_2XWIDER_MODE.
19802         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19803         * tree-ssa-math-opts.c (convert_mult_to_widen): Update use of
19804         GET_MODE_WIDER_MODE.
19805         (convert_plusminus_to_widen): Likewise.
19806         * tree-switch-conversion.c (array_value_type): Likewise.
19807         * var-tracking.c (emit_note_insn_var_location): Likewise.
19808         * tree-vrp.c (simplify_float_conversion_using_ranges): Likewise.
19809         Return false inside rather than outside the loop if no wider mode
19810         exists
19811         * optabs.c (expand_binop): Update use of GET_MODE_WIDER_MODE
19812         and GET_MODE_2XWIDER_MODE
19813         (can_compare_p): Use else_void.
19814         * gdbhooks.py (OptMachineModePrinter): New class.
19815         (build_pretty_printer): Use it for opt_mode.
19817 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19818             Alan Hayward  <alan.hayward@arm.com>
19819             David Sherwood  <david.sherwood@arm.com>
19821         * tree-switch-conversion.c (array_value_type): Only read TYPE_MODE
19822         once.  Use get_narrowest_mode instead of GET_CLASS_NARROWEST_MODE.
19824 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19825             Alan Hayward  <alan.hayward@arm.com>
19826             David Sherwood  <david.sherwood@arm.com>
19828         * machmode.h (mode_traits): New structure.
19829         (get_narrowest_mode): New function.
19830         (mode_iterator::start): Likewise.
19831         (mode_iterator::iterate_p): Likewise.
19832         (mode_iterator::get_wider): Likewise.
19833         (mode_iterator::get_known_wider): Likewise.
19834         (mode_iterator::get_2xwider): Likewise.
19835         (FOR_EACH_MODE_IN_CLASS): New mode iterator.
19836         (FOR_EACH_MODE): Likewise.
19837         (FOR_EACH_MODE_FROM): Likewise.
19838         (FOR_EACH_MODE_UNTIL): Likewise.
19839         (FOR_EACH_WIDER_MODE): Likewise.
19840         (FOR_EACH_2XWIDER_MODE): Likewise.
19841         * builtins.c (expand_builtin_strlen): Use new mode iterators.
19842         * combine.c (simplify_comparison): Likewise
19843         * config/i386/i386.c (type_natural_mode): Likewise.
19844         * cse.c (cse_insn): Likewise.
19845         * dse.c (find_shift_sequence): Likewise.
19846         * emit-rtl.c (init_derived_machine_modes): Likewise.
19847         (init_emit_once): Likewise.
19848         * explow.c (hard_function_value): Likewise.
19849         * expmed.c (extract_fixed_bit_field_1): Likewise.
19850         (extract_bit_field_1): Likewise.
19851         (expand_divmod): Likewise.
19852         (emit_store_flag_1): Likewise.
19853         * expr.c (init_expr_target): Likewise.
19854         (convert_move): Likewise.
19855         (alignment_for_piecewise_move): Likewise.
19856         (widest_int_mode_for_size): Likewise.
19857         (emit_block_move_via_movmem): Likewise.
19858         (copy_blkmode_to_reg): Likewise.
19859         (set_storage_via_setmem): Likewise.
19860         (compress_float_constant): Likewise.
19861         * omp-low.c (omp_clause_aligned_alignment): Likewise.
19862         * optabs-query.c (get_best_extraction_insn): Likewise.
19863         * optabs.c (expand_binop): Likewise.
19864         (expand_twoval_unop): Likewise.
19865         (expand_twoval_binop): Likewise.
19866         (widen_leading): Likewise.
19867         (widen_bswap): Likewise.
19868         (expand_parity): Likewise.
19869         (expand_unop): Likewise.
19870         (prepare_cmp_insn): Likewise.
19871         (prepare_float_lib_cmp): Likewise.
19872         (expand_float): Likewise.
19873         (expand_fix): Likewise.
19874         (expand_sfix_optab): Likewise.
19875         * postreload.c (move2add_use_add2_insn): Likewise.
19876         * reg-stack.c (reg_to_stack): Likewise.
19877         * reginfo.c (choose_hard_reg_mode): Likewise.
19878         * rtlanal.c (init_num_sign_bit_copies_in_rep): Likewise.
19879         * stor-layout.c (mode_for_size): Likewise.
19880         (smallest_mode_for_size): Likewise.
19881         (mode_for_vector): Likewise.
19882         (finish_bitfield_representative): Likewise.
19883         * tree-ssa-math-opts.c (target_supports_divmod_p): Likewise.
19884         * tree-vect-generic.c (type_for_widest_vector_mode): Likewise.
19885         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
19886         * var-tracking.c (prepare_call_arguments): Likewise.
19888 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19889             Alan Hayward  <alan.hayward@arm.com>
19890             David Sherwood  <david.sherwood@arm.com>
19892         * genconditions.c (write_header): Add a "#define USE_ENUM_MODES".
19893         * genmodes.c (emit_insn_modes_h): Define FOOmode to E_FOOmode if
19894         USE_ENUM_MODES is defined and to ((void) 0, E_FOOmode) otherwise.
19895         * machmode.h (mode_size): Move earlier in file.
19896         (mode_precision): Likewise.
19897         (mode_inner): Likewise.
19898         (mode_nunits): Likewise.
19899         (mode_unit_size): Likewise.
19900         (unit_unit_precision): Likewise.
19901         (mode_wider): Likewise.
19902         (mode_2xwider): Likewise.
19903         (machine_mode): New class.
19904         (mode_to_bytes): New function.
19905         (mode_to_bits): Likewise.
19906         (mode_to_precision): Likewise.
19907         (mode_to_inner): Likewise.
19908         (mode_to_unit_size): Likewise.
19909         (mode_to_unit_precision): Likewise.
19910         (mode_to_nunits): Likewise.
19911         (GET_MODE_SIZE): Use mode_to_bytes.
19912         (GET_MODE_BITSIZE): Use mode_to_bits.
19913         (GET_MODE_PRECISION): Use mode_to_precision.
19914         (GET_MODE_INNER): Use mode_to_inner.
19915         (GET_MODE_UNIT_SIZE): Use mode_to_unit_size.
19916         (GET_MODE_UNIT_PRECISION): Use mode_to_unit_precision.
19917         (GET_MODE_NUNITS): Use mode_to_nunits.
19918         * system.h (ALWAYS_INLINE): New macro.
19919         * config/powerpcspe/powerpcspe-c.c
19920         (altivec_resolve_overloaded_builtin): Use machine_mode instead of
19921         int for arg1_mode and arg2_mode.
19923 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
19924             Alan Hayward  <alan.hayward@arm.com>
19925             David Sherwood  <david.sherwood@arm.com>
19927         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
19928         Prefix mode names with E_ in case statements.
19929         * config/aarch64/aarch64-elf.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19930         * config/aarch64/aarch64.c (aarch64_split_simd_combine): Likewise.
19931         (aarch64_split_simd_move): Likewise.
19932         (aarch64_gen_storewb_pair): Likewise.
19933         (aarch64_gen_loadwb_pair): Likewise.
19934         (aarch64_gen_store_pair): Likewise.
19935         (aarch64_gen_load_pair): Likewise.
19936         (aarch64_get_condition_code_1): Likewise.
19937         (aarch64_constant_pool_reload_icode): Likewise.
19938         (get_rsqrte_type): Likewise.
19939         (get_rsqrts_type): Likewise.
19940         (get_recpe_type): Likewise.
19941         (get_recps_type): Likewise.
19942         (aarch64_gimplify_va_arg_expr): Likewise.
19943         (aarch64_simd_container_mode): Likewise.
19944         (aarch64_emit_load_exclusive): Likewise.
19945         (aarch64_emit_store_exclusive): Likewise.
19946         (aarch64_expand_compare_and_swap): Likewise.
19947         (aarch64_gen_atomic_cas): Likewise.
19948         (aarch64_emit_bic): Likewise.
19949         (aarch64_emit_atomic_swap): Likewise.
19950         (aarch64_emit_atomic_load_op): Likewise.
19951         (aarch64_evpc_trn): Likewise.
19952         (aarch64_evpc_uzp): Likewise.
19953         (aarch64_evpc_zip): Likewise.
19954         (aarch64_evpc_ext): Likewise.
19955         (aarch64_evpc_rev): Likewise.
19956         (aarch64_evpc_dup): Likewise.
19957         (aarch64_gen_ccmp_first): Likewise.
19958         (aarch64_gen_ccmp_next): Likewise.
19959         * config/alpha/alpha.c (alpha_scalar_mode_supported_p): Likewise.
19960         (alpha_emit_xfloating_libcall): Likewise.
19961         (emit_insxl): Likewise.
19962         (alpha_arg_type): Likewise.
19963         * config/arc/arc.c (arc_vector_mode_supported_p): Likewise.
19964         (arc_preferred_simd_mode): Likewise.
19965         (arc_secondary_reload): Likewise.
19966         (get_arc_condition_code): Likewise.
19967         (arc_print_operand): Likewise.
19968         (arc_legitimate_constant_p): Likewise.
19969         * config/arc/arc.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19970         * config/arc/arc.md (casesi_load): Likewise.
19971         (casesi_compact_jump): Likewise.
19972         * config/arc/predicates.md (proper_comparison_operator): Likewise.
19973         (cc_use_register): Likewise.
19974         * config/arm/aout.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
19975         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
19976         (arm_init_iwmmxt_builtins): Likewise.
19977         * config/arm/arm.c (thumb1_size_rtx_costs): Likewise.
19978         (neon_expand_vector_init): Likewise.
19979         (arm_attr_length_move_neon): Likewise.
19980         (maybe_get_arm_condition_code): Likewise.
19981         (arm_emit_vector_const): Likewise.
19982         (arm_preferred_simd_mode): Likewise.
19983         (arm_output_iwmmxt_tinsr): Likewise.
19984         (thumb1_output_casesi): Likewise.
19985         (thumb2_output_casesi): Likewise.
19986         (arm_emit_load_exclusive): Likewise.
19987         (arm_emit_store_exclusive): Likewise.
19988         (arm_expand_compare_and_swap): Likewise.
19989         (arm_evpc_neon_vuzp): Likewise.
19990         (arm_evpc_neon_vzip): Likewise.
19991         (arm_evpc_neon_vrev): Likewise.
19992         (arm_evpc_neon_vtrn): Likewise.
19993         (arm_evpc_neon_vext): Likewise.
19994         (arm_validize_comparison): Likewise.
19995         * config/arm/neon.md (neon_vc<cmp_op><mode>): Likewise.
19996         * config/avr/avr-c.c (avr_resolve_overloaded_builtin): Likewise.
19997         * config/avr/avr.c (avr_rtx_costs_1): Likewise.
19998         * config/c6x/c6x.c (c6x_vector_mode_supported_p): Likewise.
19999         (c6x_preferred_simd_mode): Likewise.
20000         * config/epiphany/epiphany.c (get_epiphany_condition_code): Likewise.
20001         (epiphany_rtx_costs): Likewise.
20002         * config/epiphany/predicates.md (proper_comparison_operator):
20003         Likewise.
20004         * config/frv/frv.c (condexec_memory_operand): Likewise.
20005         (frv_emit_move): Likewise.
20006         (output_move_single): Likewise.
20007         (output_condmove_single): Likewise.
20008         (frv_hard_regno_mode_ok): Likewise.
20009         (frv_matching_accg_mode): Likewise.
20010         * config/h8300/h8300.c (split_adds_subs): Likewise.
20011         (h8300_rtx_costs): Likewise.
20012         (h8300_print_operand): Likewise.
20013         (compute_mov_length): Likewise.
20014         (output_logical_op): Likewise.
20015         (compute_logical_op_length): Likewise.
20016         (compute_logical_op_cc): Likewise.
20017         (h8300_shift_needs_scratch_p): Likewise.
20018         (output_a_shift): Likewise.
20019         (compute_a_shift_length): Likewise.
20020         (compute_a_shift_cc): Likewise.
20021         (expand_a_rotate): Likewise.
20022         (output_a_rotate): Likewise.
20023         * config/i386/i386.c (classify_argument): Likewise.
20024         (function_arg_advance_32): Likewise.
20025         (function_arg_32): Likewise.
20026         (function_arg_64): Likewise.
20027         (function_value_64): Likewise.
20028         (ix86_gimplify_va_arg): Likewise.
20029         (ix86_legitimate_constant_p): Likewise.
20030         (put_condition_code): Likewise.
20031         (split_double_mode): Likewise.
20032         (ix86_avx256_split_vector_move_misalign): Likewise.
20033         (ix86_expand_vector_logical_operator): Likewise.
20034         (ix86_split_idivmod): Likewise.
20035         (ix86_expand_adjust_ufix_to_sfix_si): Likewise.
20036         (ix86_build_const_vector): Likewise.
20037         (ix86_build_signbit_mask): Likewise.
20038         (ix86_match_ccmode): Likewise.
20039         (ix86_cc_modes_compatible): Likewise.
20040         (ix86_expand_branch): Likewise.
20041         (ix86_expand_sse_cmp): Likewise.
20042         (ix86_expand_sse_movcc): Likewise.
20043         (ix86_expand_int_sse_cmp): Likewise.
20044         (ix86_expand_vec_perm_vpermi2): Likewise.
20045         (ix86_expand_vec_perm): Likewise.
20046         (ix86_expand_sse_unpack): Likewise.
20047         (ix86_expand_int_addcc): Likewise.
20048         (ix86_split_to_parts): Likewise.
20049         (ix86_vectorize_builtin_gather): Likewise.
20050         (ix86_vectorize_builtin_scatter): Likewise.
20051         (avx_vpermilp_parallel): Likewise.
20052         (inline_memory_move_cost): Likewise.
20053         (ix86_tieable_integer_mode_p): Likewise.
20054         (x86_maybe_negate_const_int): Likewise.
20055         (ix86_expand_vector_init_duplicate): Likewise.
20056         (ix86_expand_vector_init_one_nonzero): Likewise.
20057         (ix86_expand_vector_init_one_var): Likewise.
20058         (ix86_expand_vector_init_concat): Likewise.
20059         (ix86_expand_vector_init_interleave): Likewise.
20060         (ix86_expand_vector_init_general): Likewise.
20061         (ix86_expand_vector_set): Likewise.
20062         (ix86_expand_vector_extract): Likewise.
20063         (emit_reduc_half): Likewise.
20064         (ix86_emit_i387_round): Likewise.
20065         (ix86_mangle_type): Likewise.
20066         (ix86_expand_round_sse4): Likewise.
20067         (expand_vec_perm_blend): Likewise.
20068         (canonicalize_vector_int_perm): Likewise.
20069         (ix86_expand_vec_one_operand_perm_avx512): Likewise.
20070         (expand_vec_perm_1): Likewise.
20071         (expand_vec_perm_interleave3): Likewise.
20072         (expand_vec_perm_even_odd_pack): Likewise.
20073         (expand_vec_perm_even_odd_1): Likewise.
20074         (expand_vec_perm_broadcast_1): Likewise.
20075         (ix86_vectorize_vec_perm_const_ok): Likewise.
20076         (ix86_expand_vecop_qihi): Likewise.
20077         (ix86_expand_mul_widen_hilo): Likewise.
20078         (ix86_expand_sse2_abs): Likewise.
20079         (ix86_expand_pextr): Likewise.
20080         (ix86_expand_pinsr): Likewise.
20081         (ix86_preferred_simd_mode): Likewise.
20082         (ix86_simd_clone_compute_vecsize_and_simdlen): Likewise.
20083         * config/i386/sse.md (*andnot<mode>3): Likewise.
20084         (<mask_codefor><code><mode>3<mask_name>): Likewise.
20085         (*<code><mode>3): Likewise.
20086         * config/ia64/ia64.c (ia64_expand_vecint_compare): Likewise.
20087         (ia64_expand_atomic_op): Likewise.
20088         (ia64_arg_type): Likewise.
20089         (ia64_mode_to_int): Likewise.
20090         (ia64_scalar_mode_supported_p): Likewise.
20091         (ia64_vector_mode_supported_p): Likewise.
20092         (expand_vec_perm_broadcast): Likewise.
20093         * config/iq2000/iq2000.c (iq2000_move_1word): Likewise.
20094         (iq2000_function_arg_advance): Likewise.
20095         (iq2000_function_arg): Likewise.
20096         * config/m32c/m32c.c (m32c_preferred_reload_class): Likewise.
20097         * config/m68k/m68k.c (output_dbcc_and_branch): Likewise.
20098         (m68k_libcall_value): Likewise.
20099         (m68k_function_value): Likewise.
20100         (sched_attr_op_type): Likewise.
20101         * config/mcore/mcore.c (mcore_output_move): Likewise.
20102         * config/microblaze/microblaze.c (microblaze_function_arg_advance):
20103         Likewise.
20104         (microblaze_function_arg): Likewise.
20105         * config/mips/mips.c (mips16_build_call_stub): Likewise.
20106         (mips_print_operand): Likewise.
20107         (mips_mode_ok_for_mov_fmt_p): Likewise.
20108         (mips_vector_mode_supported_p): Likewise.
20109         (mips_preferred_simd_mode): Likewise.
20110         (mips_expand_vpc_loongson_even_odd): Likewise.
20111         (mips_expand_vec_unpack): Likewise.
20112         (mips_expand_vi_broadcast): Likewise.
20113         (mips_expand_vector_init): Likewise.
20114         (mips_expand_vec_reduc): Likewise.
20115         (mips_expand_msa_cmp): Likewise.
20116         * config/mips/mips.md (casesi_internal_mips16_<mode>): Likewise.
20117         * config/mn10300/mn10300.c (mn10300_print_operand): Likewise.
20118         (cc_flags_for_mode): Likewise.
20119         * config/msp430/msp430.c (msp430_print_operand): Likewise.
20120         * config/nds32/nds32-md-auxiliary.c (nds32_mem_format): Likewise.
20121         (nds32_output_casesi_pc_relative): Likewise.
20122         * config/nds32/nds32.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20123         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Likewise.
20124         (nvptx_gen_unpack): Likewise.
20125         (nvptx_gen_pack): Likewise.
20126         (nvptx_gen_shuffle): Likewise.
20127         (nvptx_gen_wcast): Likewise.
20128         (nvptx_preferred_simd_mode): Likewise.
20129         * config/pa/pa.c (pa_secondary_reload): Likewise.
20130         * config/pa/predicates.md (base14_operand): Likewise.
20131         * config/powerpcspe/powerpcspe-c.c
20132         (altivec_resolve_overloaded_builtin): Likewise.
20133         * config/powerpcspe/powerpcspe.c (rs6000_setup_reg_addr_masks):
20134         Likewise.
20135         (rs6000_preferred_simd_mode): Likewise.
20136         (output_vec_const_move): Likewise.
20137         (rs6000_expand_vector_extract): Likewise.
20138         (rs6000_split_vec_extract_var): Likewise.
20139         (reg_offset_addressing_ok_p): Likewise.
20140         (rs6000_legitimate_offset_address_p): Likewise.
20141         (rs6000_legitimize_address): Likewise.
20142         (rs6000_emit_set_const): Likewise.
20143         (rs6000_const_vec): Likewise.
20144         (rs6000_emit_move): Likewise.
20145         (spe_build_register_parallel): Likewise.
20146         (rs6000_darwin64_record_arg_recurse): Likewise.
20147         (swap_selector_for_mode): Likewise.
20148         (spe_init_builtins): Likewise.
20149         (paired_init_builtins): Likewise.
20150         (altivec_init_builtins): Likewise.
20151         (do_load_for_compare): Likewise.
20152         (rs6000_generate_compare): Likewise.
20153         (rs6000_expand_float128_convert): Likewise.
20154         (emit_load_locked): Likewise.
20155         (emit_store_conditional): Likewise.
20156         (rs6000_output_function_epilogue): Likewise.
20157         (rs6000_handle_altivec_attribute): Likewise.
20158         (rs6000_function_value): Likewise.
20159         (emit_fusion_gpr_load): Likewise.
20160         (emit_fusion_p9_load): Likewise.
20161         (emit_fusion_p9_store): Likewise.
20162         * config/powerpcspe/predicates.md (easy_fp_constant): Likewise.
20163         (fusion_gpr_mem_load): Likewise.
20164         (fusion_addis_mem_combo_load): Likewise.
20165         (fusion_addis_mem_combo_store): Likewise.
20166         * config/rs6000/predicates.md (easy_fp_constant): Likewise.
20167         (fusion_gpr_mem_load): Likewise.
20168         (fusion_addis_mem_combo_load): Likewise.
20169         (fusion_addis_mem_combo_store): Likewise.
20170         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20171         Likewise.
20172         * config/rs6000/rs6000-string.c (do_load_for_compare): Likewise.
20173         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
20174         (rs6000_preferred_simd_mode): Likewise.
20175         (output_vec_const_move): Likewise.
20176         (rs6000_expand_vector_extract): Likewise.
20177         (rs6000_split_vec_extract_var): Likewise.
20178         (reg_offset_addressing_ok_p): Likewise.
20179         (rs6000_legitimate_offset_address_p): Likewise.
20180         (rs6000_legitimize_address): Likewise.
20181         (rs6000_emit_set_const): Likewise.
20182         (rs6000_const_vec): Likewise.
20183         (rs6000_emit_move): Likewise.
20184         (rs6000_darwin64_record_arg_recurse): Likewise.
20185         (swap_selector_for_mode): Likewise.
20186         (paired_init_builtins): Likewise.
20187         (altivec_init_builtins): Likewise.
20188         (rs6000_expand_float128_convert): Likewise.
20189         (emit_load_locked): Likewise.
20190         (emit_store_conditional): Likewise.
20191         (rs6000_output_function_epilogue): Likewise.
20192         (rs6000_handle_altivec_attribute): Likewise.
20193         (rs6000_function_value): Likewise.
20194         (emit_fusion_gpr_load): Likewise.
20195         (emit_fusion_p9_load): Likewise.
20196         (emit_fusion_p9_store): Likewise.
20197         * config/rx/rx.c (rx_gen_move_template): Likewise.
20198         (flags_from_mode): Likewise.
20199         * config/s390/predicates.md (s390_alc_comparison): Likewise.
20200         (s390_slb_comparison): Likewise.
20201         * config/s390/s390.c (s390_handle_vectorbool_attribute): Likewise.
20202         (s390_vector_mode_supported_p): Likewise.
20203         (s390_cc_modes_compatible): Likewise.
20204         (s390_match_ccmode_set): Likewise.
20205         (s390_canonicalize_comparison): Likewise.
20206         (s390_emit_compare_and_swap): Likewise.
20207         (s390_branch_condition_mask): Likewise.
20208         (s390_rtx_costs): Likewise.
20209         (s390_secondary_reload): Likewise.
20210         (__SECONDARY_RELOAD_CASE): Likewise.
20211         (s390_expand_cs): Likewise.
20212         (s390_preferred_simd_mode): Likewise.
20213         * config/s390/vx-builtins.md (vec_packsu_u<mode>): Likewise.
20214         * config/sh/sh.c (sh_print_operand): Likewise.
20215         (dump_table): Likewise.
20216         (sh_secondary_reload): Likewise.
20217         * config/sh/sh.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20218         * config/sh/sh.md (casesi_worker_1): Likewise.
20219         (casesi_worker_2): Likewise.
20220         * config/sparc/predicates.md (icc_comparison_operator): Likewise.
20221         (fcc_comparison_operator): Likewise.
20222         * config/sparc/sparc.c (sparc_expand_move): Likewise.
20223         (emit_soft_tfmode_cvt): Likewise.
20224         (sparc_preferred_simd_mode): Likewise.
20225         (output_cbranch): Likewise.
20226         (sparc_print_operand): Likewise.
20227         (sparc_expand_vec_perm_bmask): Likewise.
20228         (vector_init_bshuffle): Likewise.
20229         * config/spu/spu.c (spu_scalar_mode_supported_p): Likewise.
20230         (spu_vector_mode_supported_p): Likewise.
20231         (spu_expand_insv): Likewise.
20232         (spu_emit_branch_or_set): Likewise.
20233         (spu_handle_vector_attribute): Likewise.
20234         (spu_builtin_splats): Likewise.
20235         (spu_builtin_extract): Likewise.
20236         (spu_builtin_promote): Likewise.
20237         (spu_expand_sign_extend): Likewise.
20238         * config/tilegx/tilegx.c (tilegx_scalar_mode_supported_p): Likewise.
20239         (tilegx_simd_int): Likewise.
20240         * config/tilepro/tilepro.c (tilepro_scalar_mode_supported_p): Likewise.
20241         (tilepro_simd_int): Likewise.
20242         * config/v850/v850.c (const_double_split): Likewise.
20243         (v850_print_operand): Likewise.
20244         (ep_memory_offset): Likewise.
20245         * config/vax/vax.c (vax_rtx_costs): Likewise.
20246         (vax_output_int_move): Likewise.
20247         (vax_output_int_add): Likewise.
20248         (vax_output_int_subtract): Likewise.
20249         * config/visium/predicates.md (visium_branch_operator): Likewise.
20250         * config/visium/visium.c (rtx_ok_for_offset_p): Likewise.
20251         (visium_print_operand_address): Likewise.
20252         * config/visium/visium.h (ASM_OUTPUT_ADDR_DIFF_ELT): Likewise.
20253         * config/xtensa/xtensa.c (xtensa_mem_offset): Likewise.
20254         (xtensa_expand_conditional_branch): Likewise.
20255         (xtensa_copy_incoming_a7): Likewise.
20256         (xtensa_output_literal): Likewise.
20257         * dfp.c (decimal_real_maxval): Likewise.
20258         * targhooks.c (default_libgcc_floating_mode_supported_p): Likewise.
20260 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20261             Alan Hayward  <alan.hayward@arm.com>
20262             David Sherwood  <david.sherwood@arm.com>
20264         * genmodes.c (mode_size_inline): Add an E_ prefix to mode names.
20265         (mode_nunits_inline): Likewise.
20266         (mode_inner_inline): Likewise.
20267         (mode_unit_size_inline): Likewise.
20268         (mode_unit_precision_inline): Likewise.
20269         (emit_insn_modes_h): Likewise.  Also emit a #define of the
20270         unprefixed name.
20271         (emit_mode_wider): Add an E_ prefix to mode names.
20272         (emit_mode_complex): Likewise.
20273         (emit_mode_inner): Likewise.
20274         (emit_mode_adjustments): Likewise.
20275         (emit_mode_int_n): Likewise.
20276         * config/aarch64/aarch64-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP)
20277         (v2si_UP, v2sf_UP, v1df_UP, di_UP, df_UP, v16qi_UP, v8hi_UP, v8hf_UP)
20278         (v4si_UP, v4sf_UP, v2di_UP, v2df_UP, ti_UP, oi_UP, ci_UP, xi_UP)
20279         (si_UP, sf_UP, hi_UP, hf_UP, qi_UP): Likewise.
20280         (CRC32_BUILTIN, ENTRY): Likewise.
20281         * config/aarch64/aarch64.c (aarch64_push_regs): Likewise.
20282         (aarch64_pop_regs): Likewise.
20283         (aarch64_process_components): Likewise.
20284         * config/alpha/alpha.c (alpha_emit_conditional_move): Likewise.
20285         * config/arm/arm-builtins.c (v8qi_UP, v4hi_UP, v4hf_UP, v2si_UP)
20286         (v2sf_UP, di_UP, v16qi_UP, v8hi_UP, v8hf_UP, v4si_UP, v4sf_UP)
20287         (v2di_UP, ti_UP, ei_UP, oi_UP, hf_UP, si_UP, void_UP): Likewise.
20288         * config/arm/arm.c (arm_init_libfuncs): Likewise.
20289         * config/i386/i386-builtin-types.awk (ix86_builtin_type_vect_mode):
20290         Likewise.
20291         * config/i386/i386-builtin.def (pcmpestr): Likewise.
20292         (pcmpistr): Likewise.
20293         * config/microblaze/microblaze.c (double_memory_operand): Likewise.
20294         * config/mmix/mmix.c (mmix_output_condition): Likewise.
20295         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
20296         Likewise.
20297         * config/rl78/rl78.c (mduc_regs): Likewise.
20298         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20299         (htm_expand_builtin): Likewise.
20300         * config/sh/sh.h (REGISTER_NATURAL_MODE): Likewise.
20301         * config/sparc/sparc.c (emit_save_or_restore_regs): Likewise.
20302         * config/xtensa/xtensa.c (print_operand): Likewise.
20303         * expmed.h (NUM_MODE_PARTIAL_INT): Likewise.
20304         (NUM_MODE_VECTOR_INT): Likewise.
20305         * genoutput.c (null_operand): Likewise.
20306         (output_operand_data): Likewise.
20307         * genrecog.c (print_parameter_value): Likewise.
20308         * lra.c (debug_operand_data): Likewise.
20310 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20312         * dwarf2out.c (scompare_loc_descriptor_wide)
20313         (scompare_loc_descriptor_narrow): New functions, split out from...
20314         (scompare_loc_descriptor): ...here.
20315         * expmed.c (emit_store_flag_int): New function, split out from...
20316         (emit_store_flag): ...here.
20318 2017-08-30  Richard Biener  <rguenther@suse.de>
20320         * dwarf2out.c (dwarf2out_finish): Remove setting AT_pubnames.
20321         (dwarf2out_early_finish): Move setting of AT_pubnames from
20322         early debug output to early finish.
20324 2017-08-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20326         * gcc/config/msp430/msp430.h: Pass -mcode/data-region to the linker
20327         and -mdata-region to the assembler.
20329 2017-08-30  Richard Biener  <rguenther@suse.de>
20331         * dwarf2out.c (add_dwarf_attr): Check we don't add duplicate
20332         attributes.
20333         (gen_subprogram_die): Add DW_AT_object_pointer only early.
20334         (dwarf2out_early_global_decl): Only generate a DIE for the
20335         abstract origin if it doesn't already exist or is a declaration DIE.
20336         (resolve_addr): Do not add the linkage name twice when
20337         generating a stub DIE for the DW_TAG_GNU_call_site target.
20339 2017-08-30  Richard Sandiford  <richard.sandiford@linaro.org>
20341         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
20342         Use machine_mode rather than int for arg1_mode.
20344 2017-08-29  Michael Meissner  <meissner@linux.vnet.ibm.com>
20346         PR target/82015
20347         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Insure
20348         that the second argument of the built-in functions to unpack
20349         128-bit scalar types to 64-bit values is 0 or 1.  Change to use a
20350         switch statement instead a lot of if statements.
20351         * config/rs6000/rs6000.md (unpack<mode>, FMOVE128_VSX iterator):
20352         Allow 64-bit values to be in Altivec registers as well as
20353         traditional floating point registers.
20354         (pack<mode>, FMOVE128_VSX iterator): Likewise.
20356 2017-08-29  Alexander Monakov  <amonakov@ispras.ru>
20358         * ira-costs.c (record_address_regs): Handle both operands of PLUS for
20359         MAX_REGS_PER_ADDRESS == 1.
20361 2017-08-29  Uros Bizjak  <ubizjak@gmail.com>
20363         * config/i386/i386.opt (flag_fentry): Do not init to -1.
20364         * config/i386/i386.c (ix86_option_override_internal): Simplify
20365         setting of opts->x_flag_entry.
20367 2017-08-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20368             Jakub Jelinek  <jakub@redhat.com>
20369             Richard Biener  <rguenther@suse.de>
20371         PR tree-optimization/81503
20372         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Ensure
20373         folded constant fits in the target type; reorder tests for clarity.
20375 2017-08-29  Martin Liska  <mliska@suse.cz>
20377         * passes.def: Include pass_lower_switch.
20378         * stmt.c (dump_case_nodes): Remove and move to
20379         tree-switch-conversion.
20380         (case_values_threshold): Likewise.
20381         (expand_switch_as_decision_tree_p): Likewise.
20382         (emit_case_decision_tree): Likewise.
20383         (expand_case): Likewise.
20384         (balance_case_nodes): Likewise.
20385         (node_has_low_bound): Likewise.
20386         (node_has_high_bound): Likewise.
20387         (node_is_bounded): Likewise.
20388         (emit_case_nodes): Likewise.
20389         (struct simple_case_node): New struct.
20390         (add_case_node): Remove.
20391         (emit_case_dispatch_table): Use vector instead of case_list.
20392         (reset_out_edges_aux): Remove.
20393         (compute_cases_per_edge): Likewise.
20394         (expand_case): Build list of simple_case_node.
20395         (expand_sjlj_dispatch_table): Use it.
20396         * tree-switch-conversion.c (struct case_node): Moved from
20397         stmt.c and adjusted.
20398         (emit_case_nodes): Likewise.
20399         (node_has_low_bound): Likewise.
20400         (node_has_high_bound): Likewise.
20401         (node_is_bounded): Likewise.
20402         (case_values_threshold): Likewise.
20403         (reset_out_edges_aux): Likewise.
20404         (compute_cases_per_edge): Likewise.
20405         (add_case_node): Likewise.
20406         (dump_case_nodes): Likewise.
20407         (balance_case_nodes): Likewise.
20408         (expand_switch_as_decision_tree_p): Likewise.
20409         (emit_jump): Likewise.
20410         (emit_case_decision_tree): Likewise.
20411         (try_switch_expansion): Likewise.
20412         (do_jump_if_equal): Likewise.
20413         (emit_cmp_and_jump_insns): Likewise.
20414         (fix_phi_operands_for_edge): New function.
20415         (record_phi_operand_mapping): Likewise.
20416         (class pass_lower_switch): New pass.
20417         (pass_lower_switch::execute): New function.
20418         (make_pass_lower_switch): Likewise.
20419         (conditional_probability):
20420         * timevar.def: Add TV_TREE_SWITCH_LOWERING.
20421         * tree-pass.h: Add make_pass_lower_switch.
20423 2017-08-29  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
20425         PR target/80993
20426         * gcc/config/msp430/msp430.c (msp430_attr): Mark interrupt
20427         handlers as used.
20429 2017-08-29  Richard Biener  <rguenther@suse.de>
20431         * dwarf2out.c (add_dwarf_attr): When checking is enabled verify
20432         we do not add a DW_AT_inline attribute twice.
20433         (gen_subprogram_die): Remove code setting DW_AT_inline on
20434         DECL_ABSTRACT_P nodes.
20436 2017-08-29  Richard Sandiford  <richard.sandiford@linaro.org>
20438         * gimplify.c (gimplify_call_expr): Copy the nothrow flag to
20439         calls to internal functions.
20440         (gimplify_modify_expr): Likewise.
20441         * tree-call-cdce.c (use_internal_fn): Likewise.
20442         * tree-ssa-math-opts.c (pass_cse_reciprocals::execute): Likewise.
20443         (convert_to_divmod): Set the nothrow flag.
20444         * tree-if-conv.c (predicate_mem_writes):  Likewise.
20445         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
20446         (vectorizable_call): Likewise.
20447         (vectorizable_store): Likewise.
20448         (vectorizable_load): Likewise.
20449         * tree-vect-patterns.c (vect_recog_pow_pattern): Likewise.
20450         (vect_recog_mask_conversion_pattern): Likewise.
20452 2017-08-29  Martin Liska  <mliska@suse.cz>
20454         PR other/39851
20455         * gcc.c (driver_handle_option): Add new argument.
20456         * opts-common.c (handle_option): Pass
20457         target_option_override_hook.
20458         * opts-global.c (lang_handle_option): Add new option.
20459         (set_default_handlers):  Add new argument.
20460         (decode_options): Likewise.
20461         * opts.c (target_handle_option): Likewise.
20462         (common_handle_option): Call target_option_override_hook.
20463         * opts.h (struct cl_option_handler_func): Add hook for
20464         target option override.
20465         (struct cl_option_handlers): Likewise.
20466         (set_default_handlers): Add new argument.
20467         (decode_options): Likewise.
20468         (common_handle_option): Likewise.
20469         (target_handle_option): Likewise.
20470         * toplev.c (toplev::main): Pass targetm.target_option.override
20471         hook.
20473 2017-08-29  Richard Biener  <rguenther@suse.de>
20474         Dominik Infuehr <dominik.infuehr@theobroma-systems.com>
20476         * tree-vect-slp.c (vect_bb_slp_scalar_cost): Properly confine
20477         life to the active subtree.
20479 2017-08-28  Jeff Law  <law@redhat.com>
20481         * tree-ssa-dom.c (edge_info::record_simple_equiv): Call
20482         derive_equivalences.
20483         (derive_equivalences_from_bit_ior, record_temporary_equivalences):
20484         Code moved into....
20485         (edge_info::derive_equivalences): New private member function
20487         * tree-ssa-dom.c (class edge_info): Changed from a struct
20488         to a class.  Add ctor/dtor, methods and data members.
20489         (edge_info::edge_info): Renamed from allocate_edge_info.
20490         Initialize additional members.
20491         (edge_info::~edge_info): New.
20492         (free_dom_edge_info): Delete the edge info.
20493         (record_edge_info): Use new class & associated member functions.
20494         Tighten forms for testing for edge equivalences.
20495         (record_temporary_equivalences): Iterate over the simple
20496         equivalences rather than assuming there's only one per edge.
20497         (cprop_into_successor_phis): Iterate over the simple
20498         equivalences rather than assuming there's only one per edge.
20499         (optimize_stmt): Use operand_equal_p rather than pointer
20500         equality for mini-DSE code.
20502 2017-08-28  Nathan Sidwell  <nathan@acm.org>
20504         * gcc.c (execute): Fold SIGPIPE handling into switch
20505         statement.  Adjust internal error message.
20507 2017-08-28  Richard Biener  <rguenther@suse.de>
20509         PR debug/81993
20510         * dwarf2out.c (gen_remaining_tmpl_value_param_die_attributes):
20511         Do nothing for removed DIEs.
20513 2017-08-28  Richard Biener  <rguenther@suse.de>
20515         PR tree-optimization/81977
20516         * tree-ssa-sccvn.c (vn_reference_lookup_3): Fix look through
20517         memcpy.
20519 2017-08-28  Alexander Monakov  <amonakov@ispras.ru>
20521         PR target/80640
20522         * doc/md.texi (mem_thread_fence): Remove mention of mode.  Rewrite.
20523         * optabs.c (expand_mem_thread_fence): Emit a compiler barrier when
20524         using targetm.gen_mem_thread_fence.
20526 2017-08-27  Uros Bizjak  <ubizjak@gmail.com>
20528         PR target/81995
20529         * config/i386/i386.md (*<btsc><mode>): Change operand 2
20530         predicate to register_operand.  Reorder operands.
20531         (*btr<mode>): Ditto.
20532         (*<btsc><mode>_mask): Change operand 3 predicate to register_operand.
20533         (*btr<mode>_mask): Ditto.
20535 2017-08-25  Steven Munroe  <munroesj@gcc.gnu.org>
20537         * config.gcc (powerpc*-*-*): Add xmmintrin.h and mm_malloc.h.
20538         * config/rs6000/xmmintrin.h: New file.
20539         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include xmmintrin.h.
20541 2017-08-25  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20543         PR target/81504
20544         * config/rs6000/rs6000-p8swap.c (find_alignment_op): Add reference
20545         parameter and_insn and return it.
20546         (recombine_lvx_pattern): Insert a copy to ensure availability of
20547         the base register of the copied masking operation at the point of
20548         the instruction replacement.
20549         (recombine_stvx_pattern): Likewise.
20551 2017-08-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
20553         * config/rs6000/rs6000.opt (-mpower9-dform-scalar): Delete
20554         undocumented switches.
20555         (-mpower9-dform-vector): Likewise.
20556         (-mpower9-dform): Likewise.
20557         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Update
20558         comments to delete references to -mpower9-dform* switches.
20559         * config/rs6000/predicates.md (vsx_quad_dform_memory_operand):
20560         Delete reference to -mpower9-dform* switches, test for
20561         -mpower9-vector instead.
20562         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Likewise.
20563         (OTHER_P9_VECTOR_MASKS): Likewise.
20564         (POWERPC_MASKS): Likewise.
20565         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Change
20566         tests against -mpower9-dform* to -mpower9-vector.  Delete code
20567         that checked for -mpower9-dform* consistancy with other options.
20568         Add test for -mpower9-misc to enable other power9 switches.
20569         (rs6000_init_hard_regno_mode_ok): Likewise.
20570         (rs6000_option_override_internal): Likewise.
20571         (rs6000_emit_prologue): Likewise.
20572         (rs6000_emit_epilogue): Likewise.
20573         (rs6000_opt_masks): Delete -mpower9-dform-{scalar,vector}.
20574         (rs6000_disable_incompatiable_switches): Delete -mpower9-dform.
20575         (emit_fusion_p9_load): Change tests for -mpower9-dform-scalar
20576         -mpower9-vector.
20577         (emit_fusion_p9_store): Likewise.
20578         * config/rs6000/rs6000.h (TARGET_P9_DFORM_SCALAR): Delete
20579         resetting these macros if the assembler does not support ISA 3.0
20580         instructions.
20581         (TARGET_P9_DFORM_VECTOR): Likewise.
20582         * config/rs6000/rs6000.md (peepholes to optimize altivec memory):
20583         Change to use -mpower9-vector instead of -mpower9-dform-scalar.
20585 2017-08-25  Alan Modra  <amodra@gmail.com>
20587         PR rtl-optimization/81747
20588         * cse.c (cse_extended_basic_block): Don't attempt to record
20589         equivalences for degenerate conditional branches that branch
20590         to their fall-through.
20592 2017-08-24  Martin Sebor  <msebor@redhat.com>
20594         PR middle-end/81908
20595         * gimple-fold.c (size_must_be_zero_p): New function.
20596         (gimple_fold_builtin_memory_op): Call it.
20598 2017-08-24  Steven Munroe  <munroesj@gcc.gnu.org>
20600         * config/rs6000/mm_malloc.h: New file.
20602 2017-08-24  Bin Cheng  <bin.cheng@arm.com>
20604         PR tree-optimization/81913
20605         * tree-ssa-loop-niter.c (number_of_iterations_cond): Skip niter
20606         analysis when either IVs in condition can wrap.
20608 2017-08-24  Uros Bizjak  <ubizjak@gmail.com>
20610         * dwarf2out.c (MAX_ARTIFICIAL_LABEL_BYTES): Increase to 40.
20611         * dwarf2cfi.c (MAX_ARTIFICIAL_LABEL_BYTES): Remove.
20613 2017-08-24  Richard Biener  <rguenther@suse.de>
20615         PR target/81921
20616         * targhooks.c (default_target_can_inline_p): Properly
20617         use target_option_default_node when no DECL_FUNCTION_SPECIFIC_TARGET
20618         is present and always compare.
20619         * config/i386/i386.c (ix86_valid_target_attribute_tree): Do not
20620         infer -mfpmath=sse from TARGET_SSE_P.
20621         (ix86_can_inline_p): Properly use target_option_default_node when
20622         no DECL_FUNCTION_SPECIFIC_TARGET is present and always compare.
20624 2017-08-24  Richard Biener  <rguenther@suse.de>
20626         PR debug/81936
20627         * dwarf2out.c (output_die): Handle flag_generate_offload like
20628         flag_generate_lto.
20629         (output_comp_unit): Likewise.
20630         (gen_array_type_die): Likewise.
20631         (dwarf2out_early_finish): Likewise.
20632         (note_variable_value_in_expr): Likewise.
20633         (dwarf2out_finish): Likewise.  Adjust assert.
20634         * cgraphunit.c (symbol_table::compile): Move setting of
20635         flag_generate_offload earlier ...
20636         (symbol_table::finalize_compilation_unit): ... here, before
20637         early debug finalization.
20639 2017-08-24  Richard Biener  <rguenther@suse.de>
20641         * config/i386/i386.c: Include symbol-summary.h, ipa-prop.h
20642         and ipa-fnsummary.h.
20643         (ix86_can_inline_p): When ix86_fpmath flags do not match
20644         check whether the callee uses FP math at all.
20646 2017-08-24  Aldy Hernandez  <aldyh@redhat.com>
20648         PR middle-end/81931
20649         * tree-ssanames.c (get_nonzero_bits): Use element_precision
20650         instead of TYPE_PRECISION.
20652 2017-08-24  Richard Sandiford  <richard.sandiford@linaro.org>
20653             Alan Hayward  <alan.hayward@arm.com>
20654             David Sherwood  <david.sherwood@arm.com>
20656         * combine.c (make_extraction): Use subreg_offset_from_lsb.
20658 2017-08-23  Daniel Santos  <daniel.santos@pobox.com>
20660         * config/i386/i386.h (ix86_frame::stack_realign_allocate_offset):
20661         Remove field.
20662         (ix86_frame::stack_realign_allocate): New field.
20663         (struct machine_frame_state): Modify comments.
20664         (machine_frame_state::sp_realigned_fp_end): New field.
20665         * config/i386/i386.c (ix86_compute_frame_layout): Rework stack frame
20666         layout calculation.
20667         (sp_valid_at): Add assertion to assure no attempt to access invalid
20668         offset of a realigned stack.
20669         (fp_valid_at): Likewise.
20670         (choose_baseaddr): Modify comments.
20671         (ix86_emit_outlined_ms2sysv_save): Adjust to changes in
20672         ix86_expand_prologue.
20673         (ix86_expand_prologue): Modify stack realignment and allocation.
20674         (ix86_expand_epilogue): Modify comments.
20675         * doc/sourcebuild.texi: Add documentation for target selectors avx2,
20676         avx2_runtime, avx512f, and avx512f_runtime.
20678 2017-08-23  Uros Bizjak  <ubizjak@gmail.com>
20680         * config/i386/i386.opt: Remove unneeded Init(0) initializations.
20681         (mstackrealign): Do not init to -1.
20682         * config/i386/i386.c (ix86_option_override_internal):
20683         Check opts_set, not opts when setting default value of
20684         opts->x_ix86_force_align_arg_pointer.
20686 2017-08-23  Richard Biener  <rguenther@suse.de>
20688         * function.c (fndecl_name): Use verbosity 1 (no arguments) for
20689         lang_hooks.decl_printable_name.
20690         * print-rtl-function.c (print_rtx_function): Likewise.
20691         * tree-pretty-print.c (dump_function_header): Likewise.
20693 2017-08-23  Richard Biener  <rguenther@suse.de>
20695         PR lto/81940
20696         * dwarf2out.c (dwarf2out_abstract_function): Handle LTO with
20697         -g0 at compile-time.
20699 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20701         PR middle-end/19706
20702         * doc/sourcebuild.texi (Other hardware attributes):
20703         Document xorsign.
20705 2017-08-23  Tamar Christina  <tamar.christina@arm.com>
20707         PR middle-end/19706
20708         * tree-ssa-math-opts.c (convert_expand_mult_copysign):
20709         Fix single-use check.
20711 2017-08-23  Thomas Preud'homme  <thomas.preudhomme@arm.com>
20713         * gcc.c (execute): Only test for SIGKILL and SIGQUIT if available.
20715 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20717         * doc/install.texi: Modify to add more details on running selected
20718         tests.
20720 2017-08-22  Daniel Santos  <daniel.santos@pobox.com>
20722         * config/i386/i386.c (ix86_option_override_internal): Error when -mx32
20723         is combined with -mabi=ms.
20724         (ix86_function_type_abi): Limit errors for mixing -mx32 with attribute
20725         ms_abi.
20727 2017-08-22  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
20729         PR tree-optimization/81488
20730         * gimple-ssa-strength-reduction (struct slsr_cand_d): Add visited
20731         and cached_basis fields.
20732         (MAX_SPREAD): New constant.
20733         (alloc_cand_and_find_basis): Initialize new fields.
20734         (clear_visited): New function.
20735         (create_phi_basis_1): Rename from create_phi_basis, set visited
20736         and cached_basis fields.
20737         (create_phi_basis): New wrapper function.
20738         (phi_add_costs_1): Rename from phi_add_costs, add spread
20739         parameter, set visited field, short-circuit when limits reached.
20740         (phi_add_costs): New wrapper function.
20741         (record_phi_increments_1): Rename from record_phi_increments, set
20742         visited field.
20743         (record_phi_increments): New wrapper function.
20744         (phi_incr_cost_1): Rename from phi_incr_cost, set visited field.
20745         (phi_incr_cost): New wrapper function.
20746         (all_phi_incrs_profitable_1): Rename from
20747         all_phi_incrs_profitable, set visited field.
20748         (all_phi_incrs_profitable): New wrapper function.
20750 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20751             Alan Hayward  <alan.hayward@arm.com>
20752             David Sherwood  <david.sherwood@arm.com>
20754         * rtl.h (paradoxical_subreg_p): Define inline, and add a version
20755         that takes the outer and inner modes.
20756         * doc/rtl.texi: Use paradoxical_subreg_p instead of a GET_MODE_SIZE
20757         comparison as the canonical test for a paradoxical subreg.
20758         * combine.c (simplify_set): Use paradoxical_subreg_p.
20759         (make_extraction): Likewise.
20760         (force_to_mode): Likewise.
20761         (rtx_equal_for_field_assignment_p): Likewise.
20762         (gen_lowpart_for_combine): Likewise.
20763         (simplify_comparison): Likewise.
20764         * cse.c (equiv_constant): Likewise.
20765         * expmed.c (store_bit_field_1): Likewise.
20766         * final.c (alter_subreg): Likewise.
20767         * fwprop.c (propagate_rtx): Likewise.
20768         (forward_propagate_subreg): Likewise.
20769         * ira-conflicts.c (ira_build_conflicts): Likewise.
20770         * lower-subreg.c (simplify_gen_subreg_concatn): Likewise.
20771         * lra-constraints.c (curr_insn_transform): Likewise.
20772         (split_reg): Likewise.
20773         * lra-eliminations.c (move_plus_up): Likewise.
20774         (lra_eliminate_regs_1): Likewise.
20775         * recog.c (general_operand): Likewise.
20776         * ree.c (combine_reaching_defs): Likewise.
20777         * reload.c (push_reload): Likewise.
20778         (find_reloads): Likewise.
20779         * reload1.c (elimination_effects): Likewise.
20780         (compute_reload_subreg_offset): Likewise.
20781         (choose_reload_regs): Likewise.
20782         * rtlanal.c (subreg_lsb_1): Likewise.
20783         * simplify-rtx.c (simplify_unary_operation_1): Likewise.
20784         (simplify_subreg): Likewise.
20785         * var-tracking.c (track_loc_p): Likewise.
20786         * emit-rtl.c (byte_lowpart_offset): Likewise.
20787         (paradoxical_subreg_p): Delete out-of-line definition.
20789 2017-08-22  Jeff Law  <law@redhat.com>
20791         PR tree-optimization/81741
20792         PR tree-optimization/71947
20793         * tree-ssa-dom.c: Include tree-inline.h.
20794         (record_temporary_equivalences): Only record SSA_NAME = SSA_NAME
20795         equivalences if one is more expensive to compute than the other.
20796         * tree-ssa-scopedtables.h (class const_or_copies): Make
20797         record_const_or_copy_raw method private.
20798         (class avail_exprs_stack): New method simplify_binary_operation.
20799         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr): Call
20800         avail_exprs_stack::simplify_binary_operation as needed.
20801         (avail_exprs_stack::simplify_binary_operation): New function.
20803 2017-08-22  Sebastian Huber  <sebastian.huber@embedded-brains.de>
20805         * config.gcc (powerpc-*-rtems*): Add rs6000/linux64.opt.
20806         * config/rs6000/rtems.h (ASM_PREFERRED_EH_DATA_FORMAT): New define.
20807         (DOT_SYMBOLS): Likewise.
20808         (MINIMAL_TOC_SECTION_ASM_OP): Likewise.
20809         (RELOCATABLE_NEEDS_FIXUP): Likewise.
20810         (RS6000_ABI_NAME): Likewise.
20811         (TARGET_CMODEL): Likewise.
20812         (TOC_SECTION_ASM_OP): Likewise.
20813         (SET_CMODEL): New macro.
20814         (SUBSUBTARGET_OVERRIDE_OPTIONS): Evaluate cmodel options.
20816 2017-08-22  Richard Biener  <rguenther@suse.de>
20818         * tree-inline.c (remap_type_1): Change asserts on TYPE_SIZE[_UNIT]
20819         to allow for free-lang-data replacements similar to verify_type_variant.
20821 2017-08-22  Richard Sandiford  <richard.sandiford@linaro.org>
20822             Alan Hayward  <alan.hayward@arm.com>
20823             David Sherwood  <david.sherwood@arm.com>
20825         * config/aarch64/aarch64.md (casesi): Use DImode rather than
20826         VOIDmode for the LABEL_REF.
20828 2017-08-22  Richard Biener  <rguenther@suse.de>
20830         * tree-cfg.c (gimple_split_edge): Avoid reallocating target PHI nodes.
20832 2017-08-22  Richard Biener  <rguenther@suse.de>
20834         * common.opt (feliminate-dwarf2-dups): Ignore.
20835         * doc/invoke.texi (feliminate-dwarf2-dups): Remove documentation.
20836         * dwarf2out.c (push_new_compile_unit, pop_compile_unit,
20837         same_die_p_wrap, compute_section_prefix,
20838         is_symbol_die, assign_symbol_names, break_out_includes): Remove.
20839         (comdat_symbol_id, comdat_symbol_number): Likewise.
20840         (cu_hash_table_entry, cu_hash_table_entry_hasher, cu_hash_type):
20841         Likewise.
20842         (check_duplicate_cu, record_comdat_symbol_number): Likewise.
20843         (output_die): Mark unreachable path unreachable.
20844         (dwarf2out_start_source_file): Do not create DW_TAG_GNU_BINCL.
20845         (dwarf2out_end_source_file): Do not create DW_TAG_GNU_EINCL.
20846         (dwarf2out_init): Remove code handling flag_eliminate_dwarf2_dups.
20847         (dwarf2out_early_finish): Likewise.
20849 2017-08-22  Aldy Hernandez  <aldyh@redhat.com>
20851         * wide-int.h (hwi_with_prec::hwi_with_prec): Sign extend.
20853 2017-08-22  Georg-Johann Lay  <avr@gjlay.de>
20855         PR target/81910
20856         * config/avr/avr.c (avr_handle_addr_attribute): Early return if
20857         not VAR_P. Filter attribute warnings with OPT_Wattributes.
20858         (avr_attribute_table) <io, io_low, address>: Initialize
20859         .decl_required with true.
20861 2017-08-21  Michael Meissner  <meissner@linux.vnet.ibm.com>
20863         * config/rs6000/rs6000-cpus.def (-mvsx-scalar-float): Delete
20864         undocumented debugging options.
20865         (-mvsx-scalar-double): Likewise.
20866         (-mallow-df-permute): Likewise.
20867         (-mvectorize-builtins): Likewise.
20868         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
20869         (rs6000_builtin_vectorized_function): Likewise.
20870         (rs6000_builtin_md_vectorized_function): Likewise.
20871         (rs6000_opt_vars): Likewise.
20873 2017-08-21  Uros Bizjak  <ubizjak@gmail.com>
20875         PR target/46091
20876         * config/i386/i386.md (*btsq_imm): Rename from *btsq.
20877         (*btrq_imm): Rename from *btrq.
20878         (*btcq_imm): Rename from *btcq.
20879         (btsc): New code attribute.
20880         (*<btsc><mode>): New insn pattern.
20881         (*btr<mode>): Ditto.
20882         (*<btsc><mode>_mask): New insn_and_split pattern.
20883         (*btr<mode>_mask): Ditto.
20885 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20886             Alan Hayward  <alan.hayward@arm.com>
20887             David Sherwood  <david.sherwood@arm.com>
20889         * function.c (pad_below): Simplify padding calculation.
20891 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20892             Alan Hayward  <alan.hayward@arm.com>
20893             David Sherwood  <david.sherwood@arm.com>
20895         * target.def (function_prologue): Remove frame size argument.
20896         (function_epilogue): Likewise.
20897         * doc/tm.texi: Regenerate.
20898         * final.c (final_start_function): Update call to function_prologue.
20899         (final_end_function): Update call to function_epilogue.
20900         (default_function_pro_epilogue): Remove frame size argument.
20901         * output.h (default_function_pro_epilogue): Likewise.
20902         * config/arm/arm.c (arm_output_function_epilogue): Likewise.
20903         (arm_output_function_prologue): Likewise.
20904         * config/frv/frv.c (frv_function_prologue): Likewise.
20905         (frv_function_epilogue): Likewise.
20906         * config/i386/i386.c (ix86_output_function_epilogue): Likewise.
20907         * config/ia64/ia64.c (ia64_output_function_prologue): Likewise.
20908         (ia64_output_function_epilogue): Likewise.
20909         * config/m32r/m32r.c (m32r_output_function_prologue): Likewise.
20910         (m32r_output_function_epilogue): Likewise.
20911         * config/microblaze/microblaze.c (microblaze_function_prologue)
20912         (microblaze_function_epilogue): Likewise.
20913         * config/mips/mips.c (mips_output_function_prologue): Likewise.
20914         (mips_output_function_epilogue): Likewise.
20915         * config/mmix/mmix.c (mmix_target_asm_function_prologue): Likewise.
20916         (mmix_target_asm_function_epilogue): Likewise.
20917         * config/msp430/msp430.c (msp430_start_function): Likewise.
20918         * config/nds32/nds32.c (nds32_asm_function_prologue): Likewise.
20919         (nds32_asm_function_epilogue): Likewise.
20920         * config/nios2/nios2.c (nios2_asm_function_prologue): Likewise.
20921         * config/pa/pa.c (pa_output_function_prologue): Likewise.
20922         (pa_output_function_epilogue): Likewise.
20923         * config/powerpcspe/powerpcspe.c (rs6000_output_function_prologue)
20924         (rs6000_output_function_epilogue): Likewise.
20925         * config/rl78/rl78.c (rl78_start_function): Likewise.
20926         * config/rs6000/rs6000.c (rs6000_output_function_prologue): Likewise.
20927         (rs6000_output_function_epilogue): Likewise.
20928         * config/rx/rx.c (rx_output_function_prologue): Likewise.
20929         * config/sh/sh.c (sh_output_function_epilogue): Likewise.
20930         * config/sparc/sparc.c (sparc_asm_function_prologue): Likewise.
20931         (sparc_asm_function_epilogue): Likewise.
20933 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20935         * tree.h (type_has_mode_precision_p): New function.
20936         * convert.c (convert_to_integer_1): Use it.
20937         * expr.c (expand_expr_real_2): Likewise.
20938         (expand_expr_real_1): Likewise.
20939         * fold-const.c (fold_single_bit_test_into_sign_test): Likewise.
20940         * match.pd: Likewise.
20941         * tree-ssa-forwprop.c (simplify_rotate): Likewise.
20942         * tree-ssa-math-opts.c (convert_mult_to_fma): Likewise.
20943         * tree-tailcall.c (process_assignment): Likewise.
20944         * tree-vect-loop.c (vectorizable_reduction): Likewise.
20945         * tree-vect-patterns.c (vect_recog_vector_vector_shift_pattern)
20946         (vect_recog_mult_pattern, vect_recog_divmod_pattern): Likewise.
20947         * tree-vect-stmts.c (vectorizable_conversion): Likewise.
20948         (vectorizable_assignment): Likewise.
20949         (vectorizable_shift): Likewise.
20950         (vectorizable_operation): Likewise.
20951         * tree-vrp.c (register_edge_assert_for_2): Likewise.
20953 2017-08-21  Wilco Dijkstra  <wdijkstr@arm.com>
20955         * match.pd: Add pow (C, x) simplification.
20957 2017-08-21  Richard Biener  <rguenther@suse.de>
20959         PR tree-optimization/81900
20960         * tree-ssa-pre.c (compute_antic_aux): Properly compute changed
20961         for blocks with abnormal predecessors.
20962         (compute_antic): Do not set visited flag prematurely.
20964 2017-08-21  Georg-Johann Lay  <avr@gjlay.de>
20966         PR target/79883
20967         * config/avr/avr.c (avr_set_current_function): Typo in diagnostic.
20969 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
20971         * stor-layout.h (vector_type_mode): Move to...
20972         * tree.h (vector_type_mode): ...here.
20973         * stor-layout.c (vector_type_mode): Move to...
20974         * tree.c (vector_type_mode): ...here.  Include rtl.h and regs.h.
20976 2017-08-21  Richard Biener  <rguenther@suse.de>
20978         * debug.h (struct gcc_debug_hooks): Add die_ref_for_decl and
20979         register_external_die hooks.
20980         (debug_false_tree_charstarstar_uhwistar): Declare.
20981         (debug_nothing_tree_charstar_uhwi): Likewise.
20982         * debug.c (do_nothing_debug_hooks): Adjust.
20983         (debug_false_tree_charstarstar_uhwistar): New do nothing.
20984         (debug_nothing_tree_charstar_uhwi): Likewise.
20985         * dbxout.c (dbx_debug_hooks): Adjust.
20986         (xcoff_debug_hooks): Likewise.
20987         * sdbout.c (sdb_debug_hooks): Likewise.
20988         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
20989         * dwarf2out.c (macinfo_label_base): New global.
20990         (dwarf2out_register_external_die): New function for the
20991         register_external_die hook.
20992         (dwarf2out_die_ref_for_decl): Likewise for die_ref_for_decl.
20993         (dwarf2_debug_hooks): Use them.
20994         (dwarf2_lineno_debug_hooks): Adjust.
20995         (struct die_struct): Add with_offset flag.
20996         (DEBUG_LTO_DWO_INFO_SECTION, DEBUG_LTO_INFO_SECTION,
20997         DEBUG_LTO_DWO_ABBREV_SECTION, DEBUG_LTO_ABBREV_SECTION,
20998         DEBUG_LTO_DWO_MACINFO_SECTION, DEBUG_LTO_MACINFO_SECTION,
20999         DEBUG_LTO_DWO_MACRO_SECTION, DEBUG_LTO_MACRO_SECTION,
21000         DEBUG_LTO_LINE_SECTION, DEBUG_LTO_DWO_STR_OFFSETS_SECTION,
21001         DEBUG_LTO_STR_DWO_SECTION, DEBUG_STR_LTO_SECTION): New macros
21002         defining section names for the early LTO debug variants.
21003         (reset_indirect_string): New helper.
21004         (add_AT_external_die_ref): Helper for dwarf2out_register_external_die.
21005         (print_dw_val): Add support for offsetted symbol references.
21006         (get_ultimate_context): Split out from is_cxx.
21007         (is_cxx): Use get_ultimate_context.
21008         (is_fortran): Add decl overload.
21009         (compute_comp_unit_symbol): Split out worker from
21010         compute_section_prefix.
21011         (compute_section_prefix): Call compute_comp_unit_symbol and
21012         set comdat_type_p here.
21013         (output_die): Skip DIE symbol output for the LTO added one.
21014         Handle DIE symbol references with offset.
21015         (output_comp_unit): Guard section name mangling properly.
21016         For LTO debug sections emit a symbol at the section beginning
21017         which we use to refer to its DIEs.
21018         (add_abstract_origin_attribute): For DIEs registered via
21019         dwarf2out_register_external_die directly refer to the early
21020         DIE rather than indirectly through the shadow one we created.
21021         Remove obsolete call to dwarf2out_abstract_function for
21022         non-function/block origins.
21023         (gen_array_type_die): When generating early LTO debug do
21024         not emit DW_AT_string_length.
21025         (gen_formal_parameter_die): Do not re-create DIEs for PARM_DECLs
21026         late when in LTO.  As suggested place a gcc_unreachable for
21027         the DECL_ABSTRACT_P case.
21028         (gen_subprogram_die): Avoid another specification DIE
21029         for early built declarations/definitions for the late LTO case.
21030         (gen_variable_die): Add type references for late duplicated VLA dies
21031         when in late LTO.
21032         (gen_inlined_subroutine_die): Do not call dwarf2out_abstract_function,
21033         we have the abstract instance already.
21034         (process_scope_var): Adjust decl DIE contexts in LTO which
21035         first puts them in limbo.
21036         (gen_decl_die): Do not generate type DIEs late apart from
21037         types for VLAs or for decls we do not yet have a DIE.  Do not
21038         call dwarf2out_abstract_function late.
21039         (dwarf2out_early_global_decl): Make sure to create DIEs
21040         for abstract instances of a decl first.
21041         (dwarf2out_late_global_decl): Adjust comment.
21042         (output_macinfo_op): With multiple macro sections use
21043         macinfo_label_base to distinguish labels.
21044         (output_macinfo): Likewise.  Update macinfo_label_base.
21045         Pass in the line info label.
21046         (note_variable_value_in_expr): When generating LTO resolve
21047         all variable values here by generating DIEs as needed.
21048         (init_sections_and_labels): Add early LTO debug flag parameter
21049         and generate different sections and names if set.  Add generation
21050         counter for the labels so we can have multiple of them.
21051         (reset_dies): Helper to allow DIEs to be output multiple times.
21052         (dwarf2out_finish): When outputting DIEs to the fat part of an
21053         LTO object first reset DIEs.
21054         (dwarf2out_early_finish): Output early DIEs when generating LTO.
21055         (modified_type_die): Check for decl_ultimate_origin being self
21056         before recursing.
21057         (gen_type_die_with_usage): Likewise.
21058         (gen_typedef_die): Allow decl_ultimate_origin being self.
21059         (set_decl_abstract_flags): Remove.
21060         (set_block_abstract_flags): Likewise.
21061         (dwarf2out_abstract_function): Treat the early generated DIEs
21062         as the abstract copy and only add DW_AT_inline and
21063         DW_AT_artificial here and call set_decl_origin_self.
21064         If the DIE has an abstract origin don't do anything.
21065         * tree.c (free_lang_data): Build a dummy TRANSLATION_UNIT_DECL
21066         if we have none yet (Go fails to build one, PR78628).
21067         (variably_modified_type_p): Prevent endless recursion for Ada
21068         cyclic pointer types.
21069         * lto-streamer-in.c: Include debug.h.
21070         (dref_queue): New global.
21071         (lto_read_tree_1): Stream in DIE references.
21072         (lto_input_tree): Register DIE references.
21073         (input_function): Stream DECL_DEBUG_ARGS.
21074         * lto-streamer-out.c: Include debug.h.
21075         (lto_write_tree_1): Output DIE references.
21076         (DFS::DFS_write_tree_body): Follow DECL_ABSTRACT_ORIGIN.
21077         Force a TRANSLATION_UNIT_DECL DECL_CONTEXT for file-scope decls.
21078         (output_function): Stream DECL_DEBUG_ARGS.
21079         * tree-streamer-in.c (lto_input_ts_decl_common_tree_pointers):
21080         Stream DECL_ABSTRACT_ORIGIN.
21081         * tree-streamer-out.c (write_ts_decl_common_tree_pointers): Likewise.
21082         (write_ts_decl_minimal_tree_pointers): Force a TRANSLATION_UNIT_DECL
21083         DECL_CONTEXT for file-scope decls.
21084         * lto-streamer.h (struct dref_entry): Declare.
21085         (dref_queue): Likewise.
21086         * cfgexpand.c (pass_expand::execute): Do not call the
21087         outlining_inline_function hook here.
21088         * lto-wrapper.c (debug_obj): New global.
21089         (tool_cleanup): Unlink it if required.
21090         (debug_objcopy): New function.
21091         (run_gcc): Handle early debug sections in the IL files by
21092         extracting them to separate files, partially linkin them and
21093         feeding the result back as result to the linker.
21094         * config/darwin.h (DEBUG_LTO_INFO_SECTION, DEBUG_LTO_ABBREV_SECTION,
21095         DEBUG_LTO_MACINFO_SECTION, DEBUG_LTO_LINE_SECTION,
21096         DEBUG_STR_LTO_SECTION, DEBUG_LTO_MACRO_SECTION): Put early debug
21097         sections into a separate segment.
21098         * config/darwin.c (darwin_asm_named_section): Handle __GNU_DWARF_LTO
21099         segments.
21100         (darwin_asm_dwarf_section): Likewise.
21101         (darwin_asm_output_dwarf_offset): Likewise.
21102         * config/i386/i386.c (make_resolver_func): Set DECL_IGNORED_P.
21104 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21105             Alan Hayward  <alan.hayward@arm.com>
21106             David Sherwood  <david.sherwood@arm.com>
21108         * read-md.h (md_reader::record_potential_iterator_use): Replace
21109         pointer argument with an rtx and an index.
21110         * read-rtl.c (iterator_group::apply_iterator): Likewise.
21111         (apply_mode_iterator): Likewise.
21112         (apply_code_iterator): Likewise.
21113         (apply_int_iterator): Likewise.
21114         (apply_subst_iterator): Likewise.
21115         (record_iterator_use): Likewise.
21116         (record_attribute_use): Likewise.
21117         (md_reader::record_potential_iterator_use): Likewise.  Update calls
21118         to record_iterator_use and apply_iterator.
21119         (iterator_use): Replace ptr with x and index.
21120         (attribute_use): Likewise.
21121         (apply_attribute_uses): Update calls to apply_iterator.
21122         (apply_iterators): Likewise.  Update initialization of iterator_use.
21123         (rtx_reader::read_rtx_code): Update calls to record_iterator_use
21124         and record_potential_iterator_use.
21125         (rtx_reader::read_rtx_operand): Likewise.
21127 2017-08-21  Richard Sandiford  <richard.sandiford@linaro.org>
21128             Alan Hayward  <alan.hayward@arm.com>
21129             David Sherwood  <david.sherwood@arm.com>
21131         * varasm.c (const_rtx_hash_1): Don't hash in the mode of a
21132         CONST_WIDE_INT.
21134 2017-08-21  Richard Biener  <rguenther@suse.de>
21136         PR middle-end/81884
21137         * tree-ssa-alias.c (stmt_kills_ref_p): Handle array accesses
21138         at struct end conservatively when comparing common bases.
21140 2017-08-21  Richard Biener  <rguenther@suse.de>
21142         * tree-ssa-loop-im.c (struct lim_aux_data): Add ref index member.
21143         (mem_ref_in_stmt): Remove.
21144         (determine_max_movement): Use ref index to get at the reference.
21145         (invariantness_dom_walker::before_dom_children): Deal with
21146         lim data already initialized.
21147         (gather_mem_refs_stmt): Initialize lim data and record ref index.
21149 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21151         * config/i386/i386.h (OPTION_MASK_ISA_ROUND): Remove.
21152         (TARGET_ISA_ROUND): Ditto.
21153         (TARGET_ROUND): Ditto.
21154         * config/i386/i386.c: Substitute TARGET_ROUND with TARGET_SSE4_1.
21155         * config/i386/i386.md: Ditto.
21156         * config/i386/sse.md: Ditto.
21157         * config/i386/i386-builtin.def: Substitute OPTION_MASK_ISA_ROUND
21158         with OPTION_MASK_ISA_SSE4_1.
21160 2017-08-19  Uros Bizjak  <ubizjak@gmail.com>
21162         PR target/81894
21163         * doc/extend.texi (x86 Built-in Functions): Correct the name of
21164         __builtin_ia32_lzcnt_u16.
21166 2017-08-18  Peter Bergner  <bergner@vnet.ibm.com>
21168         PR target/80210
21169         * config/rs6000/rs6000.c (rs6000_activate_target_options): New function.
21170         (rs6000_set_current_function): Rewrite function to use it.
21172 2017-08-18  H.J. Lu  <hongjiu.lu@intel.com>
21174         PR c/53037
21175         * print-tree.c (print_node): Support DECL_WARN_IF_NOT_ALIGN
21176         and TYPE_WARN_IF_NOT_ALIGN.
21177         * stor-layout.c (do_type_align): Merge DECL_WARN_IF_NOT_ALIGN.
21178         (handle_warn_if_not_align): New.
21179         (place_union_field): Call handle_warn_if_not_align.
21180         (place_field): Call handle_warn_if_not_align.
21181         Copy TYPE_WARN_IF_NOT_ALIGN.
21182         (finish_builtin_struct): Copy TYPE_WARN_IF_NOT_ALIGN.
21183         (layout_type): Likewise.
21184         * tree-core.h (tree_type_common): Add warn_if_not_align.  Set
21185         spare to 18.
21186         (tree_decl_common): Add warn_if_not_align.
21187         * tree.c (build_range_type_1): Copy TYPE_WARN_IF_NOT_ALIGN.
21188         * tree.h (TYPE_WARN_IF_NOT_ALIGN): New.
21189         (SET_TYPE_WARN_IF_NOT_ALIGN): Likewise.
21190         (DECL_WARN_IF_NOT_ALIGN): Likewise.
21191         (SET_DECL_WARN_IF_NOT_ALIGN): Likewise.
21192         * doc/extend.texi: Document warn_if_not_aligned attribute.
21193         * doc/invoke.texi: Document -Wif-not-aligned and -Wpacked-not-aligned.
21195 2017-08-17  Martin Liska  <mliska@suse.cz>
21197         PR bootstrap/81864
21198         * tree-loop-distribution.c (ddrs_table): Change type to pointer type.
21199         (get_data_dependence): Use it as pointer type.
21200         (distribute_loop): Likewise.
21202 2017-08-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21204         * config/rs6000/altivec.md (UNSPEC_VMRGOW_DIRECT): New constant.
21205         (p8_vmrgew_v4sf_direct): Generalize to p8_vmrgew_<mode>_direct.
21206         (p8_vmrgow_<mode>_direct): New define_insn.
21207         * config/rs6000/rs6000.c (altivec_expand_vec_perm_const): Properly
21208         handle endianness for vmrgew and vmrgow permute patterns.
21210 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21212         * config/rs6000/altivec.md (VParity): Remove TARGET_VSX_TIMODE.
21213         * config/rs6000/rs6000-cpus.def: Remove comment.
21214         (ISA_2_7_MASKS_SERVER): Delete OPTION_MASK_VSX_TIMODE;
21215         (POWERPC_MASKS): Likewise.
21216         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Remove unneeded
21217         use of TARGET_VSX_TIMODE.
21218         (rs6000_setup_reg_addr_masks): Change TARGET_VSX_TIMODE to TARGET_VSX.
21219         (rs6000_init_hard_regno_mode_ok): Remove unneeded uses of
21220         TARGET_VSX_TIMODE.  Change use of TARGET_VSX_TIMODE to TARGET_VSX.
21221         (rs6000_option_override_internal): Remove dead code.
21222         (rs6000_legitimize_address): Change TARGET_VSX_TIMODE to TARGET_VSX.
21223         (rs6000_legitimize_reload_address): Likewise.
21224         (rs6000_legitimate_address_p): Likewise.
21225         (rs6000_opt_masks): Delete "vsx-timode".
21226         (rs6000_disable_incompatible_switches): Remove mention of -mvsx-timode
21227         from function comment.
21228         * config/rs6000/rs6000.h (MASK_VSX_TIMODE): Delete.
21229         * config/rs6000/rs6000.md (FMOVE128_GPR): Remove TARGET_VSX_TIMODE.
21230         (V16QI, V8HI, V4SI, V4SF, V2DI, V2DF, V1TI): Remove useless empty
21231         condition.
21232         * config/rs6000/rs6000.opt (mvsx-timode): Replace with stub.
21233         * config/rs6000/vector.md (VEC_IP): Remove TARGET_VSX_TIMODE.
21234         * config/rs6000/vsx.md (VSX_LE_128): Likewise.
21235         (VSX_TI): Likewise.
21236         (VSX_M): Likewise.
21237         (define_peephole2): Likewise.
21239 2017-08-17  Martin Sebor  <msebor@redhat.com>
21241         PR c/81859
21242         * pretty-print.c (pp_format): Use strnlen in %.*s to avoid reading
21243         past the end of an array.
21244         (test_pp_format): Add test cases.
21246 2017-08-17  Richard Sandiford  <richard.sandiford@linaro.org>
21248         * internal-fn.def (CLRSB, CLZ, CTZ, FFS, PARITY, POPCOUNT): Add
21249         missing ECF_NOTHROW flags.
21251 2017-08-17  Peter Bergner  <bergner@vnet.ibm.com>
21253         PR target/72804
21254         * config/rs6000/vsx.md (*vsx_le_permute_<mode>): Add support for
21255         operands residing in integer registers.
21256         (*vsx_le_perm_load_<mode>): Likewise.
21257         (*vsx_le_perm_store_<mode>): Likewise.
21258         (define_peephole2): Add peepholes to optimize the above.
21260 2017-08-17  Marek Polacek  <polacek@redhat.com>
21262         PR middle-end/81814
21263         * fold-const.c (operand_equal_for_comparison_p): Remove code that used
21264         to mimic what shorten_compare did.  Change the return type to bool.
21265         (fold_cond_expr_with_comparison): Update call to
21266         operand_equal_for_comparison_p.
21267         (fold_ternary_loc): Likewise.
21269 2017-08-17  Jackson Woodruff  <jackson.woodruff@arm.com>
21271         * aarch64-simd.md (mov<mode>): No longer force zero immediate into
21272         register.
21273         (*aarch64_simd_mov<mode>): Add new case for stp using zero immediate.
21275 2017-08-17  Richard Biener  <rguenther@suse.de>
21277         * tree-ssa-structalias.c (solve_graph): When propagating
21278         to successors update the graphs succ edges and avoid duplicate work.
21280 2017-08-17  Maxim Ostapenko  <m.ostapenko@samsung.com>
21282         PR target/81861
21283         * config/i386/i386.c (ix86_option_override_internal): Save target
21284         specific options after ix86_stack_protector_guard_reg was changed.
21286 2017-08-17  Richard Biener  <rguenther@suse.de>
21288         PR tree-optimization/81827
21289         * tree-ssa-structalias.c (struct variable_info): Add is_reg_var flag.
21290         (new_var_info): Initialize it conservatively.
21291         (get_call_vi): Mark register vars.
21292         (new_scalar_tmp_constraint_exp): Likewise.
21293         (handle_rhs_call): Likewise.
21294         (handle_const_call): Likewise.
21295         (create_function_info_for): Likewise.
21296         (solve_constraints): Sort varinfos to separate register from
21297         non-register vars to pack points-to solution bitmaps during
21298         iteration.
21300 2017-08-17  Marek Polacek  <polacek@redhat.com>
21302         * gimplify.c (gimplify_adjust_omp_clauses): Compare with 0 instead of 1.
21304 2017-08-17  Richard Biener  <rguenther@suse.de>
21306         * tree-vrp.c (vrp_int_const_binop): Do not set *overflow_p
21307         to true when overflow is undefined and we saturated the result.
21309 2017-08-17  Alan Modra  <amodra@gmail.com>
21311         PR target/80938
21312         * config/rs6000/rs6000.c (rs6000_savres_strategy): Revert 2017-08-09.
21313         Don't use store multiple if only one reg needs saving.
21314         (interesting_frame_related_regno): New function.
21315         (rs6000_frame_related): Don't emit frame info for regs that
21316         don't need saving.
21317         (rs6000_emit_epilogue): Likewise.
21319 2017-08-16  Nathan Sidwell  <nathan@acm.org>
21321         * tree-core.h (tree_type_non_common): Rename binfo to lang_1.
21322         * tree.h (TYPE_BINFO): Use type_non_common.maxval.
21323         (TYPE_LANG_SLOT_1): Use type_non_common.lang_1, for any type.
21324         * tree.c (free_lang_data_in_type): Use else-if chain.  Always
21325         clear TYPE_LANG_1.  Remove obsolete member-function stripping.
21326         (find_decls_types_r): Comment about TYPE_MAX_VALUES_RAW.
21327         (verify_type): Adjust for TYPE_BINFO move.
21328         * lto-streamer-out.c (DFS::DFS_write_tree_body): No need to
21329         process TYPE_BINFO directly.
21330         (hash_tree): Likewise.
21331         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
21332         Likewise.
21333         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
21334         Likewise.
21336 2017-08-16  David Malcolm  <dmalcolm@redhat.com>
21338         * diagnostic-show-locus.c (colorizer::m_caret): Remove unused field.
21340 2017-08-16  Uros Bizjak  <ubizjak@gmail.com>
21342         PR target/46091
21343         * config/i386/i386.md (*anddi_1_btr): Change predicates of
21344         operand 0 and operand 1 to nomimmediate_operand. Add "m" constraint.
21345         Add ix86_binary_operator_ok to insn constraint.
21346         (*iordi_1_bts): Ditto.
21347         (*xordi_1_btc): Ditto.
21348         (*btsq): Change predicate of operand 0 to nonimmediate_operand.
21349         Update corresponding peephole2 pattern.
21350         (*btrq): Ditto.
21351         (*btcq): Ditto.
21353 2017-08-16  Bin Cheng  <bin.cheng@arm.com>
21355         PR tree-optimization/81832
21356         * tree-ssa-loop-ch.c (should_duplicate_loop_header_p): Don't
21357         copy loop header which has IFN_LOOP_DIST_ALIAS call.
21359 2017-08-16  Marek Polacek  <polacek@redhat.com>
21361         PR middle/81695
21362         * fold-const.c (fold_indirect_ref_1): Restore original behavior
21363         regarding size_zero_node.
21365 2017-08-16  Martin Liska  <mliska@suse.cz>
21367         PR target/81753
21368         * config.gcc: Respect previously set extra_objs in case
21369         of darwin target.
21371 2017-08-16  Richard Sandiford  <richard.sandiford@linaro.org>
21373         PR tree-optimization/81835
21374         * tree-vect-loop.c (vect_is_simple_reduction): Simply checks for
21375         the phi SSA_NAME.  Check that the condition in a COND_EXPR does
21376         not depend on the phi.
21378 2017-08-16  Alan Modra  <amodra@gmail.com>
21380         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Delete
21381         dead code.
21383 2017-08-16  Alan Modra  <amodra@gmail.com>
21385         * config/rs6000/rs6000.c (rs6000_reg_live_or_pic_offset_p): Merge..
21386         (save_reg_p): ..into this.  Update all callers.
21387         (first_reg_to_save): Simplify.
21389 2017-08-16  Alan Modra  <amodra@gmail.com>
21391         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't restore
21392         fixed regs.
21394 2017-08-15  Joseph Myers  <joseph@codesourcery.com>
21396         PR target/78460
21397         PR target/67712
21398         * config/sh/sh-mem.cc (sh_expand_cmpnstr): Only unroll for
21399         constant count if that count is less than 32.
21401 2017-08-15  Nathan Sidwell  <nathan@acm.org>
21403         * gcc.c (execute): Emit friendlier message if inferior is killed
21404         by an external cause.
21406 2017-08-15  Richard Biener  <rguenther@suse.de>
21408         PR tree-optimization/81790
21409         * tree-ssa-sccvn.c (vn_lookup_simplify_result): Handle both
21410         CONSTRUCTORs from simplifying and VN.
21412 2017-08-14  Martin Sebor  <msebor@redhat.com>
21414         * builtin-attrs.def: Add comments.
21416 2017-08-14  Martin Sebor  <msebor@redhat.com>
21418         PR c/81117
21419         * doc/extend.texi (attribute nonstring): Document new attribute.
21421 2017-08-14  Martin Sebor  <msebor@redhat.com>
21423         PR c/81117
21424         * tree-diagnostic.c (default_tree_printer): Handle %G.
21425         * gimple-pretty-print.h (percent_G_format): Declare new function.
21426         * gimple-pretty-print.c (percent_G_format): Define.
21427         * tree-pretty-print.c (percent_K_format): Add argument.
21429 2017-08-14  Martin Sebor  <msebor@redhat.com>
21431         PR translation/79998
21432         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call):
21433         Remove a stray space.
21435 2017-08-14  Uros Bizjak  <ubizjak@gmail.com>
21437         PR target/46091
21438         * config/i386/i386.md (*anddi_1_btr): New insn_and_split pattern.
21439         (*iordi_1_bts): Ditto.
21440         (*xordi_1_btc): Ditto.
21442 2017-08-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21444         PR target/79845
21445         * config/rs6000/linux64.h (INVALID_64BIT): Use quoted strings.
21446         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
21447         Likewise.
21448         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Use
21449         quoted strings, and make more translator-friendly.
21450         (darwin_rs6000_override_options): Likewise.
21451         (rs6000_option_override_internal): Likewise.
21452         (rs6000_return_in_memory): Fix overlong line.
21453         (init_cmulative_args): Use quoted strings, and make more
21454         translator-friendly.
21455         (rs6000_pass_by_reference): Fix overlong line.
21456         (def_builtin): Use quoted strings.
21457         (altivec_expand_predicate_builtin): Use quoted strings, and make
21458         more translator-friendly.
21459         (htm_expand_builtin): Use quoted strings.
21460         (cpu_expand_builtin): Use quoted strings, and make more
21461         translator-friendly.
21462         (altivec_expand_builtin): Likewise.
21463         (paired_expand_predicate_builtin): Likewise.
21464         (rs6000_invalid_builtin): Likewise.
21465         (builtin_function_type): Use quoted strings.
21466         (rs6000_expand_split_stack_prologue): Use quoted strings, and make
21467         more translator-friendly.
21468         (rs6000_trampoline_init): Likewise.
21469         (rs6000_handle_altivec_attribute): Likewise.
21470         (rs6000_inner_target_options): Use quoted strings.
21471         (rs6000_disable_incompatible_switches): Likewise.
21472         * config/rs6000/sysv4.h (SUBTARGET_OVERRIDE_OPTIONS): Use quoted
21473         strings, and make more translator-friendly.
21474         (SUBSUBTARGET_OVERRIDE_OPTIONS): Use quoted strings.
21476 2017-08-14  Bin Cheng  <bin.cheng@arm.com>
21478         PR tree-optimization/81799
21479         * tree-loop-distribution.c (version_loop_by_alias_check): Force
21480         cond_expr to simple gimple operand.
21482 2017-08-14  Wilco Dijkstra  <wdijkstr@arm.com>
21484         PR middle-end/46932
21485         * doc/sourcebuild.texi (autoincdec): Add autoincdec description.
21487 2017-08-14  Georg-Johann Lay  <avr@gjlay.de>
21489         PR target/81754
21490         PR target/81268
21491         * config/avr/avr.opt (mgas-isr-prologues): New Var avr_gasisr_prologues.
21492         * config/avr/avr.md (gasisr, *gasisr): Use it instead of
21493         TARGET_GASISR_PROLOGUES.
21494         * config/avr/avr.c (avr_option_override): Same.
21495         (avr_pass_pre_proep::execute): Same.
21497 2017-08-13  H.J. Lu  <hongjiu.lu@intel.com>
21499         PR target/81820
21500         * config/i386/i386.c (ix86_finalize_stack_frame_flags): Replace
21501         frame pointer with stack pointer - UNITS_PER_WORD in debug insns.
21503 2017-08-13  Uros Bizjak  <ubizjak@gmail.com>
21505         * config/i386/i386.md (*load_tp_<mode>): Redefine as
21506         define_insn_and_split.  Split to a memory load from 0 in
21507         DEFAULT_TLS_SEG_REG address space.  Merge with *load_tp_x32
21508         using PTR mode iterator.
21509         (*load_tp_x32_zext"): Redefine as define_insn_and_split.
21510         Split to a memory load from 0 in DEFAULT_TLS_SEG_REG address space.
21511         (*add_tp_<mode>): Redefine as define_insn_and_split.
21512         Split to an add with a memory load from 0 in DEFAULT_TLS_SEG_REG
21513         address space.  Merge with *add_tp_x32 using PTR mode iterator.
21514         (*add_tp_x32_zext"): Redefine as define_insn_and_split.
21515         Split to an add with a  memory load from 0 in
21516         DEFAULT_TLS_SEG_REG address space.
21518 2017-08-12  Andrew Pinski  <apinski@cavium.com>
21520         * config/aarch64/aarch64-option-extensions.def (rdma):
21521         Fix feature string to what Linux prints out in /proc/cpuinfo.
21523 2017-08-12  Pierre-Marie de Rodat  <derodat@adacore.com>
21525         PR ada/79542
21526         * dwarf2out.c (modified_type_die): For C typedef types that have
21527         an ultimate origin, process the ultimate origin instead of the
21528         input type.
21529         (gen_typedef_die): Assert that input DECLs have no ultimate
21530         origin.
21531         (gen_type_die_with_usage): For typedef variants that have an
21532         ultimate origin, just call gen_decl_die on the original DECL.
21533         (process_scope_var): Avoid creating DIEs for local typedefs and
21534         concrete static variables.
21536 2017-08-12  Alan Modra  <amodra@gmail.com>
21538         PR target/81170
21539         PR target/81295
21540         * config/rs6000/sysv4.h (STARTFILE_LINUX_SPEC): Upgrade to
21541         match gnu-user.h startfile.
21542         (ENDFILE_LINUX_SPEC): Similarly.
21544 2017-08-11  Thomas Schwinge  <thomas@codesourcery.com>
21546         PR lto/81430
21547         * config/nvptx/nvptx.c (nvptx_override_options_after_change):
21548         Remove function.
21549         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Remove definition.
21551 2017-08-11  Tamar Christina  <tamar.christina@arm.com>
21552         * config/aarch64/aarch64.md (mov<mode>): Change.
21553         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
21554         aarch64_reg_or_fp_float into aarch64_reg_or_fp_zero.
21555         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): Removed.
21557 2017-08-11  Eric Botcazou  <ebotcazou@adacore.com>
21559         * tree-sra.c (build_access_from_expr_1): Use more precise diagnostics
21560         for storage order barriers.
21562 2017-08-11  Martin Liska  <mliska@suse.cz>
21564         PR tree-opt/79987
21565         * tree-chkp.c (chkp_get_bounds_for_decl_addr): Do not instrument
21566         variables of void type.
21568 2017-08-11  Martin Liska  <mliska@suse.cz>
21570         * asan.c (asan_protect_global): Replace ASM_OUTPUT_DEF with
21571         TARGET_SUPPORTS_ALIASES.
21572         * cgraph.c (cgraph_node::create_same_body_alias): Likewise.
21573         * ipa-visibility.c (can_replace_by_local_alias): Likewise.
21574         (optimize_weakref): Likewise.
21575         * symtab.c (symtab_node::noninterposable_alias): Likewise.
21576         * varpool.c (varpool_node::create_extra_name_alias): Likewise.
21577         * defaults.h: Introduce TARGET_SUPPORTS_ALIASES.
21579 2017-08-11  Martin Liska  <mliska@suse.cz>
21581         PR ipa/81213
21582         * config/i386/i386.c (make_resolver_func): Do complete
21583         refactoring of the function.
21585 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21587         PR target/81708
21588         * config/i386/i386.opt (mstack-protector-guard-symbol=): New option
21589         * config/i386/i386.c (ix86_stack_protect_guard): Use
21590         ix86_stack_protect_guard_symbol_str to generate varible declaration.
21591         * doc/invoke.texi (x86 Options): Document
21592         -mstack-protector-guard-symbol= option.
21594 2017-08-10  Uros Bizjak  <ubizjak@gmail.com>
21596         * config/i386/i386-protos.h (ix86_split_stack_guard): New prototype.
21597         * config/i386/i386.c (ix86_split_stack_guard): New function.
21598         (ix86_xpand_split_stack_prologue): Call ix86_split_stack_guard.
21599         (ix86_legitimate_address_p) <case UNSPEC_STACK_CHECK>: Remove.
21600         (i386_asm_output_addr_const_extra) <case UNSPEC_STACK_CHECK>: Ditto.
21601         (optput_pic_addr_const): Remove UNSPEC_STACK_CHECK handling.
21602         * config/i386/i386.md (unspec): Remove UNSPEC_STACK_CHECK.
21603         (split_stack_space_check): Call ix86_split_stack_guard.
21605 2017-08-10  Martin Sebor  <msebor@redhat.com>
21607         * print-tree.c (print_node): Print location using the established
21608         format %s:%i%i.
21609         Replace spaces with colons.
21610         (debug_raw, debug): Ditto.
21612 2017-08-10  Martin Sebor  <msebor@redhat.com>
21614         PR c++/81586
21615         * pretty-print.c (pp_format): Correct the handling of %s precision.
21617 2017-08-10  H.J. Lu  <hongjiu.lu@intel.com>
21619         PR target/81736
21620         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21621         to ...
21622         (ix86_finalize_stack_frame_flags): This.  Also clear
21623         frame_pointer_needed if -fno-omit-frame-pointer is used without
21624         stack access.
21625         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
21626         with ix86_finalize_stack_frame_flags.
21627         (ix86_expand_epilogue): Likewise.
21628         (ix86_expand_split_stack_prologue): Likewise.
21629         * doc/invoke.texi: Add a note for -fno-omit-frame-pointer.
21631 2017-08-10  Martin Liska  <mliska@suse.cz>
21633         PR c++/81355
21634         * c-attribs.c (handle_target_attribute):
21635         Report warning for an empty string argument of target attribute.
21637 2017-08-09  Jakub Jelinek  <jakub@redhat.com>
21639         PR c/81687
21640         * omp-low.c (omp_copy_decl): Don't remap FORCED_LABEL or DECL_NONLOCAL
21641         LABEL_DECLs.
21642         * tree-cfg.c (move_stmt_op): Don't adjust DECL_CONTEXT of FORCED_LABEL
21643         or DECL_NONLOCAL labels.
21644         (move_stmt_r) <case GIMPLE_LABEL>: Adjust DECL_CONTEXT of FORCED_LABEL
21645         or DECL_NONLOCAL labels here.
21647 2017-08-09  Will Schmidt  <will_schmidt@vnet.ibm.com>
21649         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add blurb
21650         to indicate when early gimple folding has been disabled.
21651         (rs6000_gimple_fold_builtin): Add debug content.
21652         (rs6000_invalid_builtin): Fix whitespace.
21653         (rs6000_expand_builtin): Fix whitespace.
21654         * config/rs6000/rs6000.opt: Add option for -mfold-gimple.
21656 2017-08-09  Segher Boessenkool  <segher@kernel.crashing.org>
21658         PR target/80938
21659         * config/rs6000/rs6000.c (rs6000_savres_strategy): Don't use
21660         SAVE_MULTIPLE if not all the registers that saves, should be saved.
21662 2017-08-09  Jim Wilson  <jim.wilson@linaro.org>
21664         * config/aarch64/aarch64-cores.def (falkor): Use falkor pipeline.
21665         (qdf24xx): Likewise.
21666         * config/aarch64/aarch64.md: Include falkor.md.
21667         * config/aarch64/falkor.md: New.
21669 2017-08-09  Marek Polacek  <polacek@redhat.com>
21671         PR c/81233
21672         * diagnostic-core.h (emit_diagnostic_valist): Add declaration.
21673         * diagnostic.c (emit_diagnostic): Add a comment.
21674         (emit_diagnostic_valist): New function.
21676 2017-08-09  Marek Polacek  <polacek@redhat.com>
21678         PR c/81417
21679         * input.c (make_location): New overload.
21680         * input.h (make_location): Declare.
21682 2017-08-08  Alan Modra  <amodra@gmail.com>
21683             H.J. Lu  <hongjiu.lu@intel.com>
21685         PR driver/81523
21686         * gcc.c (NO_PIE_SPEC): Delete.
21687         (PIE_SPEC): Define as !no-pie/pie.  Move static|shared|r
21688         exclusion..
21689         (LINK_PIE_SPEC): ..to here.
21690         (LINK_COMMAND_SPEC): Support -no-pie.
21691         * config/gnu-user.h (GNU_USER_TARGET_STARTFILE_SPEC): Correct
21692         chain of crtbegin*.o selection, update for PIE_SPEC changes and
21693         format.
21694         (GNU_USER_TARGET_ENDFILE_SPEC): Similarly.
21695         * config/sol2.h (STARTFILE_CRTBEGIN_SPEC): Similarly.
21696         (ENDFILE_CRTEND_SPEC): Similarly.
21698 2017-08-08  Uros Bizjak  <ubizjak@gmail.com>
21700         PR target/81708
21701         * config/i386/i386.opt (mstack-protector-guard-reg=): New option
21702         (mstack-protector-guard-offset=): Ditto.
21703         * config/i386/i386.c (ix86_option_override): Handle
21704         -mstack-protector-guard-reg= and -mstack-protector-guard-offset=
21705         options.
21706         (ix86_stack_protect_guard): Use ix86_stack_protect_guard_reg and
21707         ix86_stack_protect_guard_offset variables.
21708         (TARGET_STACK_PROTECT_GUARD): Always define.
21709         * doc/invoke.texi (x86 Options): Document -mstack-protector-guard-reg=
21710         and -mstack-protector-guard-offset= options.
21712 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21714         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): Handle
21715         boundary case for the last candidate.
21717 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21719         * doc/invoke.texi: Document -ftree-loop-distribution for O3.
21720         * opts.c (default_options_table): Add OPT_ftree_loop_distribution.
21722 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21724         PR middle-end/19706
21725         * config/aarch64/aarch64.md (xorsign<mode>3): New optabs.
21726         * config/aarch64/aarch64-builtins.c
21727         (aarch64_builtin_vectorized_function): Added CASE_CFN_XORSIGN.
21728         * config/aarch64/aarch64-simd-builtins.def: Added xorsign BINOP.
21729         * config/aarch64/aarch64-simd.md: Added xorsign<mode>3
21731 2017-08-08  Tamar Christina  <tamar.christina@arm.com>
21732             Andrew Pinski <pinskia@gmail.com>
21734         PR middle-end/19706
21735         * internal-fn.def (XORSIGN): New.
21736         * optabs.def (xorsign_optab): New.
21737         * tree-ssa-math-opts.c (is_copysign_call_with_1): New.
21738         (convert_expand_mult_copysign): New.
21739         (pass_optimize_widening_mul::execute): Call
21740         convert_expand_mult_copysign.
21742 2017-08-08  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
21744         PR tree-optimization/81354
21745         * gimple-ssa-strength-reduction.c (create_add_on_incoming_edge):
21746         Insert on edges rather than explicitly creating landing pads.
21747         (analyze_candidates_and_replace): Commit edge inserts.
21749 2017-08-08  Richard Biener  <rguenther@suse.de>
21751         PR middle-end/81719
21752         * tree-ssa-loop-niter.c: Include tree-dfa.h.
21753         (expand_simple_operations): Also look through ADDR_EXPRs with
21754         MEM_REF bases treating them as POINTER_PLUS_EXPR.
21756 2017-08-08  Richard Biener  <rguenther@suse.de>
21758         PR tree-optimization/81723
21759         * tree-vect-slp.c (struct bst_traits): New hash traits.
21760         (bst_fail): New global.
21761         (vect_build_slp_tree_2): New worker, split out from ...
21762         (vect_build_slp_tree): ... this now wrapping it with using
21763         bst_fail set to cache SLP tree build fails.  Properly handle
21764         max_tree_size.
21765         (vect_analyze_slp_instance): Allocate and free bst_fail.
21767 2017-08-08  Martin Liska  <mliska@suse.cz>
21769         PR tree-opt/81696
21770         * ipa-icf-gimple.c (func_checker::compare_cst_or_decl): Consider
21771         LABEL_DECLs that can be from a different function.
21773 2017-08-08  Bin Cheng  <bin.cheng@arm.com>
21775         PR tree-optimization/81744
21776         * tree-predcom.c (prepare_finalizers_chain): Deep copy expr of
21777         loop's number of iterations.
21779 2017-08-08  Martin Liska  <mliska@suse.cz>
21781         * asan.c: Include header files.
21782         * attribs.c (build_decl_attribute_variant): New function moved
21783         from tree.[ch].
21784         (build_type_attribute_qual_variant): Likewise.
21785         (cmp_attrib_identifiers): Likewise.
21786         (simple_cst_list_equal): Likewise.
21787         (omp_declare_simd_clauses_equal): Likewise.
21788         (attribute_value_equal): Likewise.
21789         (comp_type_attributes): Likewise.
21790         (build_type_attribute_variant): Likewise.
21791         (lookup_ident_attribute): Likewise.
21792         (remove_attribute): Likewise.
21793         (merge_attributes): Likewise.
21794         (merge_type_attributes): Likewise.
21795         (merge_decl_attributes): Likewise.
21796         (merge_dllimport_decl_attributes): Likewise.
21797         (handle_dll_attribute): Likewise.
21798         (attribute_list_equal): Likewise.
21799         (attribute_list_contained): Likewise.
21800         * attribs.h (lookup_attribute): New function moved from tree.[ch].
21801         (lookup_attribute_by_prefix): Likewise.
21802         * bb-reorder.c: Include header files.
21803         * builtins.c: Likewise.
21804         * calls.c: Likewise.
21805         * cfgexpand.c: Likewise.
21806         * cgraph.c: Likewise.
21807         * cgraphunit.c: Likewise.
21808         * convert.c: Likewise.
21809         * dwarf2out.c: Likewise.
21810         * final.c: Likewise.
21811         * fold-const.c: Likewise.
21812         * function.c: Likewise.
21813         * gimple-expr.c: Likewise.
21814         * gimple-fold.c: Likewise.
21815         * gimple-pretty-print.c: Likewise.
21816         * gimple.c: Likewise.
21817         * gimplify.c: Likewise.
21818         * hsa-common.c: Likewise.
21819         * hsa-gen.c: Likewise.
21820         * internal-fn.c: Likewise.
21821         * ipa-chkp.c: Likewise.
21822         * ipa-cp.c: Likewise.
21823         * ipa-devirt.c: Likewise.
21824         * ipa-fnsummary.c: Likewise.
21825         * ipa-inline.c: Likewise.
21826         * ipa-visibility.c: Likewise.
21827         * ipa.c: Likewise.
21828         * lto-cgraph.c: Likewise.
21829         * omp-expand.c: Likewise.
21830         * omp-general.c: Likewise.
21831         * omp-low.c: Likewise.
21832         * omp-offload.c: Likewise.
21833         * omp-simd-clone.c: Likewise.
21834         * opts-global.c: Likewise.
21835         * passes.c: Likewise.
21836         * predict.c: Likewise.
21837         * sancov.c: Likewise.
21838         * sanopt.c: Likewise.
21839         * symtab.c: Likewise.
21840         * toplev.c: Likewise.
21841         * trans-mem.c: Likewise.
21842         * tree-chkp.c: Likewise.
21843         * tree-eh.c: Likewise.
21844         * tree-into-ssa.c: Likewise.
21845         * tree-object-size.c: Likewise.
21846         * tree-parloops.c: Likewise.
21847         * tree-profile.c: Likewise.
21848         * tree-ssa-ccp.c: Likewise.
21849         * tree-ssa-live.c: Likewise.
21850         * tree-ssa-loop.c: Likewise.
21851         * tree-ssa-sccvn.c: Likewise.
21852         * tree-ssa-structalias.c: Likewise.
21853         * tree-ssa.c: Likewise.
21854         * tree-streamer-in.c: Likewise.
21855         * tree-vectorizer.c: Likewise.
21856         * tree-vrp.c: Likewise.
21857         * tsan.c: Likewise.
21858         * ubsan.c: Likewise.
21859         * varasm.c: Likewise.
21860         * varpool.c: Likewise.
21861         * tree.c: Remove functions moved to attribs.[ch].
21862         * tree.h: Likewise.
21863         * config/aarch64/aarch64.c: Add attrs.h header file.
21864         * config/alpha/alpha.c: Likewise.
21865         * config/arc/arc.c: Likewise.
21866         * config/arm/arm.c: Likewise.
21867         * config/avr/avr.c: Likewise.
21868         * config/bfin/bfin.c: Likewise.
21869         * config/c6x/c6x.c: Likewise.
21870         * config/cr16/cr16.c: Likewise.
21871         * config/cris/cris.c: Likewise.
21872         * config/darwin.c: Likewise.
21873         * config/epiphany/epiphany.c: Likewise.
21874         * config/fr30/fr30.c: Likewise.
21875         * config/frv/frv.c: Likewise.
21876         * config/ft32/ft32.c: Likewise.
21877         * config/h8300/h8300.c: Likewise.
21878         * config/i386/winnt.c: Likewise.
21879         * config/ia64/ia64.c: Likewise.
21880         * config/iq2000/iq2000.c: Likewise.
21881         * config/lm32/lm32.c: Likewise.
21882         * config/m32c/m32c.c: Likewise.
21883         * config/m32r/m32r.c: Likewise.
21884         * config/m68k/m68k.c: Likewise.
21885         * config/mcore/mcore.c: Likewise.
21886         * config/microblaze/microblaze.c: Likewise.
21887         * config/mips/mips.c: Likewise.
21888         * config/mmix/mmix.c: Likewise.
21889         * config/mn10300/mn10300.c: Likewise.
21890         * config/moxie/moxie.c: Likewise.
21891         * config/msp430/msp430.c: Likewise.
21892         * config/nds32/nds32-isr.c: Likewise.
21893         * config/nds32/nds32.c: Likewise.
21894         * config/nios2/nios2.c: Likewise.
21895         * config/nvptx/nvptx.c: Likewise.
21896         * config/pa/pa.c: Likewise.
21897         * config/pdp11/pdp11.c: Likewise.
21898         * config/powerpcspe/powerpcspe.c: Likewise.
21899         * config/riscv/riscv.c: Likewise.
21900         * config/rl78/rl78.c: Likewise.
21901         * config/rx/rx.c: Likewise.
21902         * config/s390/s390.c: Likewise.
21903         * config/sh/sh.c: Likewise.
21904         * config/sol2.c: Likewise.
21905         * config/sparc/sparc.c: Likewise.
21906         * config/spu/spu.c: Likewise.
21907         * config/stormy16/stormy16.c: Likewise.
21908         * config/tilegx/tilegx.c: Likewise.
21909         * config/tilepro/tilepro.c: Likewise.
21910         * config/v850/v850.c: Likewise.
21911         * config/vax/vax.c: Likewise.
21912         * config/visium/visium.c: Likewise.
21913         * config/xtensa/xtensa.c: Likewise.
21915 2017-08-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
21917         PR target/81593
21918         * config/rs6000/vsx.md (vsx_concat_<mode>, VSX_D): Cleanup
21919         constraints since the -mupper-regs-* switches have been
21920         eliminated.
21921         (vsx_concat_<mode>_1): New combiner insns to recognize inserting
21922         into a vector from a double word element that was extracted from
21923         another vector, and eliminate extra XXPERMDI instructions.
21924         (vsx_concat_<mode>_2): Likewise.
21925         (vsx_concat_<mode>_3): Likewise.
21926         (vsx_set_<mode>, VSX_D): Rewrite vector set in terms of vector
21927         concat to allow optimizing inserts from previous extracts.
21929 2017-08-07  Uros Bizjak  <ubizjak@gmail.com>
21931         * config/i386/i386.c (ix86_stack_protect_guard): Generate
21932         memory reference to a SSP offset in TLS address space.
21933         (ix86_print_operand) <case '@'>: Remove.
21934         (ix86_print_operand_punct_valid_p): Remove '@' code.
21935         * config/i386/i386.md (unspec): Remove UNSPEC_SP_TLS_SET and
21936         UNSPEC_SP_TLS_TEST.
21937         (stack_tls_protect_set_<mode>): Remove.
21938         (stack_protect_set): Do not call gen_stack_tls_protect_set_<mode>.
21939         (stack_tls_protect_test_<mode>): Remove.
21940         (stack_protect_test): Do not call gen_stack_tls_protect_test_<mode>.
21942 2017-08-07  Olivier Hainque  <hainque@adacore.com>
21944         PR target/81755
21945         * config/vxworksae.h (VXWORKS_HAVE_TLS): Define.
21947 2017-08-07  Douglas Rupp  <rupp@adacore.com>
21949         * Makefile.in (install-mkheaders): Fix typo, where the multi_dir
21950         variable was referenced as multidir in command.
21952 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
21954         PR c/69389
21955         * gimplify.c (goa_stabilize_expr): Handle BIT_INSERT_EXPR and
21956         BIT_FIELD_REF.
21958 2017-08-07  Martin Liska  <mliska@suse.cz>
21960         * config/m32c/m32c.c: Add include of stringpool.h and attribs.h.
21961         * config/rl78/rl78.c: Add include of attribs.h.
21962         * config/sh/sh.c: Likewise.
21963         * config/v850/v850.c: Likewise.
21965 2017-08-07  Tom de Vries  <tom@codesourcery.com>
21967         PR middle-end/78266
21968         * omp-expand.c (expand_oacc_for): Ensure diff_type is large enough.
21970 2017-08-07  Martin Liska  <mliska@suse.cz>
21972         * config/mips/mips.c: Include attribs.h.
21974 2017-08-07  Thomas Koenig  <tkoenig@gcc.gnu.org>
21976         PR fortran/68829
21977         * doc/invoke.texi: Document change in behvaior for -Ofast for
21978         Fortran.
21980 2017-08-07  Wilco Dijkstra  <wdijkstr@arm.com>
21982         * config/aarch64/aarch64.c (aarch64_pushwb_single_reg):
21983         Use gen_frame_mem.
21984         (aarch64_pop_regs): Likewise.
21985         (aarch64_gen_load_pair): Likewise.
21986         (aarch64_save_callee_saves): Likewise.
21987         (aarch64_restore_callee_saves): Likewise.
21989 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
21991         * config/i386/i386.c: Revert the last change.
21993 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
21995         PR target/81736
21996         * config/i386/i386.c (ix86_finalize_stack_realign_flags): Renamed
21997         to ...
21998         (ix86_finalize_stack_frame_flags): This.  Also clear
21999         frame_pointer_needed if -fno-omit-frame-pointer is used without
22000         stack access.
22001         (ix86_expand_prologue): Replace ix86_finalize_stack_realign_flags
22002         with ix86_finalize_stack_frame_flags.
22003         (ix86_expand_epilogue): Likewise.
22004         (ix86_expand_split_stack_prologue): Likewise.
22006 2017-08-07  H.J. Lu  <hongjiu.lu@intel.com>
22008         PR target/81743
22009         * config/i386/i386.c (get_builtin_code_for_version): Set priority
22010         to P_AES for Westmere.
22012 2017-08-07  Jonathan Yong  <10walls@gmail.com>
22014         * config/i386/mingw.opt (fset-stack-executable): Removed.
22015         * config/i386/cygming.opt (fset-stack-executable): Moved
22016         from mingw.opt.
22017         * config/i386/cygwin.h: Define CHECK_EXECUTE_STACK_ENABLED.
22019 2017-08-07  Segher Boessenkool  <segher@kernel.crashing.org>
22021         * print-rtl.c (print_exp): Print NOT as "~" instead of as "!".
22023 2017-08-07  Marek Polacek  <polacek@redhat.com>
22025         PR middle-end/81737
22026         * fold-const.c (fold_indirect_ref_1): Check type_domain.
22028 2017-08-07  Martin Liska  <mliska@suse.cz>
22030         * attribs.h (canonicalize_attr_name): New function.
22031         (cmp_attribs): Move from c-format.c and adjusted.
22032         (is_attribute_p): Moved from tree.h.
22033         * tree-inline.c: Add new includes.
22034         * tree.c (cmp_attrib_identifiers): Use cmp_attribs.
22035         (private_is_attribute_p): Remove.
22036         (private_lookup_attribute): Likewise.
22037         (private_lookup_attribute_by_prefix): Simplify.
22038         (remove_attribute): Use is_attribute_p.
22039         * tree.h: Remove removed declarations.
22041 2017-08-07  Jakub Jelinek  <jakub@redhat.com>
22043         PR middle-end/81698
22044         * stmt.c (emit_case_dispatch_table): Add DEFAULT_EDGE argument,
22045         instead of computing it in the function.  Formatting fix.
22046         (expand_case): Don't rely on default_edge being the first edge,
22047         clear it if removing it, pass default_edge to
22048         emit_case_dispatch_table.
22049         (expand_sjlj_dispatch_table): Pass NULL as DEFAULT_EDGE, formatting
22050         fix.
22052 2017-08-06  Uros Bizjak  <ubizjak@gmail.com>
22054         * config/alpha/alpha.c (alpha_reorg): If trap is the last active
22055         insn in the function, emit NOP after the insn.
22057 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22059         * omp-expand.c (expand_oacc_for): Add missing edge probability for tile
22060         and element loops.
22062 2017-08-06  Tom de Vries  <tom@codesourcery.com>
22064         * omp-expand.c (expand_oacc_for): Add missing edge probability for chunk
22065         loop.
22067 2017-08-04  Yury Gribov  <tetra2005@gmail.com>
22069         PR tree-optimization/57371
22070         * match.pd: New pattern.
22072 2017-08-04  Marek Polacek  <polacek@redhat.com>
22074         PR middle-end/81695
22075         * fold-const.c (fold_indirect_ref_1): For ((int *)&a + 4 -> a[1],
22076         perform the computation in offset_int.
22078 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22080         PR tree-optimization/81136
22081         * tree-vectorizer.h: Include tree-hash-traits.h.
22082         (vec_base_alignments): New typedef.
22083         (vec_info): Add a base_alignments field.
22084         (vect_record_base_alignments): Declare.
22085         * tree-data-ref.h (data_reference): Add an is_conditional_in_stmt
22086         field.
22087         (DR_IS_CONDITIONAL_IN_STMT): New macro.
22088         (create_data_ref): Add an is_conditional_in_stmt argument.
22089         * tree-data-ref.c (create_data_ref): Likewise.  Use it to initialize
22090         the is_conditional_in_stmt field.
22091         (data_ref_loc): Add an is_conditional_in_stmt field.
22092         (get_references_in_stmt): Set the is_conditional_in_stmt field.
22093         (find_data_references_in_stmt): Update call to create_data_ref.
22094         (graphite_find_data_references_in_stmt): Likewise.
22095         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Likewise.
22096         * tree-vect-data-refs.c (vect_analyze_data_refs): Likewise.
22097         (vect_record_base_alignment): New function.
22098         (vect_record_base_alignments): Likewise.
22099         (vect_compute_data_ref_alignment): Adjust base_addr and aligned_to
22100         for nested statements even if we fail to compute a misalignment.
22101         Use pooled base alignments for unconditional references.
22102         (vect_find_same_alignment_drs): Compare base addresses instead
22103         of base objects.
22104         (vect_analyze_data_refs_alignment): Call vect_record_base_alignments.
22105         * tree-vect-slp.c (vect_slp_analyze_bb_1): Likewise.
22107 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22109         * tree-vectorizer.h (vec_info): Add a constructor and destructor.
22110         Add an explicit name for the enum.  Use auto_vec for slp_instances
22111         and grouped_stores.
22112         (_loop_vec_info): Add a constructor and destructor.  Use auto_vec
22113         for all vectors.
22114         (_bb_vec_info): Add a constructor and destructor.
22115         (vinfo_for_stmt): Return NULL for uids of -1 as well.
22116         (destroy_loop_vec_info): Delete.
22117         (vect_destroy_datarefs): Likewise.
22118         * tree-vectorizer.c (vect_destroy_datarefs): Delete.
22119         (vec_info::vec_info): New function.
22120         (vec_info::~vec_info): Likewise.
22121         (vectorize_loops): Use delete instead of destroy_loop_vec_info.
22122         * tree-parloops.c (gather_scalar_reductions): Use delete instead of
22123         destroy_loop_vec_info.
22124         * tree-vect-loop.c (new_loop_vec_info): Replace with...
22125         (_loop_vec_info::_loop_vec_info): ...this.
22126         (destroy_loop_vec_info): Replace with...
22127         (_loop_vec_info::~_loop_vec_info): ...this.  Unconditionally delete
22128         the stmt_vec_infos.  Leave handling of vec_info information to its
22129         destructor.  Remove explicit vector releases.
22130         (vect_analyze_loop_form): Use new instead of new_loop_vec_info.
22131         (vect_analyze_loop): Use delete instead of destroy_loop_vec_info.
22132         * tree-vect-slp.c (new_bb_vec_info): Replace with...
22133         (_bb_vec_info::_bb_vec_info): ...this.  Don't reserve space in
22134         BB_VINFO_GROUPED_STORES or BB_VINFO_SLP_INSTANCES.
22135         (destroy_bb_vec_info): Replace with...
22136         (_bb_vec_info::~_bb_vec_info): ...this.  Leave handling of vec_info
22137         information to its destructor.
22138         (vect_slp_analyze_bb_1): Use new and delete instead of
22139         new_bb_vec_info and destroy_bb_vec_info.
22140         (vect_slp_bb): Replace 2 calls to destroy_bb_vec_info with a
22141         single delete.
22143 2017-08-04  Richard Sandiford  <richard.sandiford@linaro.org>
22145         * tree-data-ref.h (subscript): Add access_fn field.
22146         (data_dependence_relation): Add could_be_independent_p.
22147         (SUB_ACCESS_FN, DDR_COULD_BE_INDEPENDENT_P): New macros.
22148         (same_access_functions): Move to tree-data-ref.c.
22149         * tree-data-ref.c (ref_contains_union_access_p): New function.
22150         (access_fn_component_p): Likewise.
22151         (access_fn_components_comparable_p): Likewise.
22152         (dr_analyze_indices): Add a reference to access_fn_component_p.
22153         (dump_data_dependence_relation): Use SUB_ACCESS_FN instead of
22154         DR_ACCESS_FN.
22155         (constant_access_functions): Likewise.
22156         (add_other_self_distances): Likewise.
22157         (same_access_functions): Likewise.  (Moved from tree-data-ref.h.)
22158         (initialize_data_dependence_relation): Use XCNEW and remove
22159         explicit zeroing of DDR_REVERSED_P.  Look for a subsequence
22160         of access functions that have the same type.  Allow the
22161         subsequence to end with different bases in some circumstances.
22162         Record the chosen access functions in SUB_ACCESS_FN.
22163         (build_classic_dist_vector_1): Replace ddr_a and ddr_b with
22164         a_index and b_index.  Use SUB_ACCESS_FN instead of DR_ACCESS_FN.
22165         (subscript_dependence_tester_1): Likewise dra and drb.
22166         (build_classic_dist_vector): Update calls accordingly.
22167         (subscript_dependence_tester): Likewise.
22168         * tree-ssa-loop-prefetch.c (determine_loop_nest_reuse): Check
22169         DDR_COULD_BE_INDEPENDENT_P.
22170         * tree-vectorizer.h (LOOP_REQUIRES_VERSIONING_FOR_ALIAS): Test
22171         comp_alias_ddrs instead of may_alias_ddrs.
22172         * tree-vect-data-refs.c (vect_analyze_possibly_independent_ddr):
22173         New function.
22174         (vect_analyze_data_ref_dependence): Use it if
22175         DDR_COULD_BE_INDEPENDENT_P, but fall back to using the recorded
22176         distance vectors if that fails.
22177         (dependence_distance_ge_vf): New function.
22178         (vect_prune_runtime_alias_test_list): Use it.  Don't clear
22179         LOOP_VINFO_MAY_ALIAS_DDRS.
22181 2017-08-04  Richard Biener  <rguenther@suse.de>
22183         PR middle-end/81705
22184         * fold-const.c (fold_binary_loc): Properly restrict
22185         minus_var0 && minus_var1 case when associating undefined overflow
22186         entities.
22188 2017-08-04  Tom de Vries  <tom@codesourcery.com>
22190         * omp-simd-clone.c (simd_clone_adjust): Add missing edge probability.
22192 2017-08-03  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
22194         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
22195         Don't start diagnostic messages with a capital letter.
22196         * config/rs6000/rs6000.c (rs6000_option_override_internal):
22197         Likewise.
22198         (rs6000_invalid_builtin): Likewise.
22199         (rs6000_trampoline_init): Likewise.
22201 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22203         PR target/81621
22204         * bb-reorder.c (pass_partition_blocks::execute): Return TODO_df_finish
22205         after setting changeable df flags.
22207 2017-08-03  Richard Biener  <rguenther@suse.de>
22209         * tree-ssa-reassoc.c (should_break_up_subtract): Also break
22210         up if the use is in USE - X.
22212 2017-08-03  Alexander Monakov  <amonakov@ispras.ru>
22214         * toplev.c (dumpfile.h): New include.
22215         (internal_error_reentered): New static function.  Use it...
22216         (internal_error_function): ...here to handle reentered internal_error.
22218 2017-08-03  Richard Biener  <rguenther@suse.de>
22220         PR middle-end/81148
22221         * fold-const.c (split_tree): Add minus_var and minus_con
22222         arguments, remove unused loc arg.  Never generate NEGATE_EXPRs
22223         here but always use minus_*.
22224         (associate_trees): Assert we never associate with MINUS_EXPR
22225         and NULL first operand.  Do not recurse for PLUS_EXPR operands
22226         when associating as MINUS_EXPR either.
22227         (fold_binary_loc): Track minus_var and minus_con.
22229 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22231         PR lto/81430
22232         * tree-streamer-in.c (lto_input_ts_function_decl_tree_pointers): If
22233         ACCEL_COMPILER, apply finish_options on
22234         DECL_FUNCTION_SPECIFIC_OPTIMIZATION.
22236 2017-08-03  Tom de Vries  <tom@codesourcery.com>
22238         PR target/81662
22239         * config/nvptx/nvptx.c (nvptx_option_override): Emit sorry if
22240         function_entry_patch_area_size > 0.
22242 2017-08-03  Jakub Jelinek  <jakub@redhat.com>
22244         PR driver/81650
22245         * calls.c (alloc_max_size): Use HOST_WIDE_INT_UC (10??)
22246         instead of 10??LU, perform unit multiplication in wide_int,
22247         don't change alloc_object_size_limit if the limit is larger
22248         than SSIZE_MAX.
22250         PR tree-optimization/81655
22251         PR tree-optimization/81588
22252         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): Handle also
22253         the case when ranges[i].low and high are 1 for unsigned type with
22254         precision 1.
22256         PR middle-end/81052
22257         * omp-low.c (diagnose_sb_0): Handle flag_openmp_simd like flag_openmp.
22258         (pass_diagnose_omp_blocks::gate): Enable also for flag_openmp_simd.
22260 2017-08-03  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
22262         * tree-vrp.h: Add include guard.
22264 2017-08-02  Uros Bizjak  <ubizjak@gmail.com>
22266         PR target/81644
22267         * config/i386/i386.md (unspecv): Add UNSPECV_UD2.
22268         (ud2): New insn pattern.
22269         * config/i386/i386.c (ix86_expand_epilogue):
22270         For naked functions, generate ud2 instead of trap insn.
22272 2017-08-02  Marek Polacek  <polacek@redhat.com>
22274         PR other/81667
22275         * alloc-pool.h (base_pool_allocator): Initialize m_elt_size.
22277 2017-08-02  Tom de Vries  <tom@codesourcery.com>
22278             Cesar Philippidis  <cesar@codesourcery.com>
22280         * config/nvptx/nvptx.c (nvptx_lockless_update, nvptx_lockfull_update):
22281         Add missing edge probabilities.
22283 2017-08-02  Tamar Christina  <tamar.christina@arm.com>
22285         * config/aarch64/aarch64.c (aarch64_reinterpret_float_as_int):
22286         Correct endianness.
22288 2017-08-02  Jakub Jelinek  <jakub@redhat.com>
22290         PR middle-end/79499
22291         * function.c (thread_prologue_and_epilogue_insns): Determine blocks
22292         for find_many_sub_basic_blocks bitmap by looking up BLOCK_FOR_INSN
22293         of first NONDEBUG_INSN_P in each of the split_prologue_seq and
22294         prologue_seq sequences - if any.
22296 2017-08-02  Richard Biener  <rguenther@suse.de>
22298         * tree-vect-stmts.c (vectorizable_store): Perform vector extracts
22299         via vectors if supported, integer extracts via punning if supported
22300         or otherwise vector extracts.
22302 2017-08-02  Richard Biener  <rguenther@suse.de>
22304         * tree-ssa-pre.c (bitmap_insert_into_set_1): Remove and inline
22305         into ...
22306         (bitmap_insert_into_set): ... this.
22308 2017-08-02  Richard Biener  <rguenther@suse.de>
22310         PR tree-optimization/81633
22311         Revert
22312         2015-08-17  Alan Hayward  <alan.hayward@arm.com>
22314         PR tree-optimization/71752
22315         * tree-vect-slp.c (vect_get_slp_defs): Handle null operands.
22317 2017-08-01  Daniel Santos  <daniel.santos@pobox.com>
22319         * config/i386/i386.h (ix86_frame::outlined_save_offset): Remove field.
22320         (machine_function::call_ms2sysv_pad_out): Remove field.
22321         * config/i386/i386.c (xlogue_layout::get_stack_space_used): Modify.
22322         (ix86_compute_frame_layout): Likewise.
22324 2017-08-01  H.J. Lu  <hongjiu.lu@intel.com>
22326         PR target/81654
22327         * config/i386/i386.c (ix86_set_func_type): Disallow naked
22328         attribute with interrupt attribute.
22330 2017-08-01  Andrew Pinski  <apinski@cavium.com>
22332         * tree-ssa-scopedtables.c (hashable_expr_equal_p): Check
22333         BIT_INSERT_EXPR's operand 1
22334         to see if the types precision matches.
22336 2017-08-01  Martin Liska  <mliska@suse.cz>
22338         PR middle-end/70140
22339         * builtins.c (expand_builtin_memcpy_args): Remove.
22340         (expand_builtin_memcpy): Call newly added function
22341         expand_builtin_memory_copy_args.
22342         (expand_builtin_memcpy_with_bounds): Likewise.
22343         (expand_builtin_mempcpy): Remove last argument.
22344         (expand_builtin_mempcpy_with_bounds): Likewise.
22345         (expand_builtin_memory_copy_args): New function created from
22346         expand_builtin_mempcpy_args with small modifications.
22347         (expand_builtin_mempcpy_args): Remove.
22348         (expand_builtin_stpcpy): Remove unused argument.
22349         (expand_builtin): Likewise.
22350         (expand_builtin_with_bounds): Likewise.
22352 2017-08-01  Martin Liska  <mliska@suse.cz>
22354         Revert r250771
22355         Make mempcpy more optimal (PR middle-end/70140).
22357 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22359         PR target/81622
22360         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): For
22361         __builtin_vec_cmpne verify both arguments are compatible vectors
22362         before looking at TYPE_MODE on the element type.  For __builtin_vec_ld
22363         verify arg1_type is a pointer or array type.  For __builtin_vec_st,
22364         move computation of aligned to after checking the argument types.
22365         Formatting fixes.
22367         PR target/80846
22368         * config/rs6000/vsx.md (vextract_fp_from_shorth,
22369         vextract_fp_from_shortl): Add element mode after mode in gen_vec_init*
22370         calls.
22372 2017-08-01  Jerome Lambourg  <lambourg@adacore.com>
22373             Doug Rupp  <rupp@adacore.com>
22374             Olivier Hainque  <hainque@adacore.com>
22376         * config.gcc (arm-wrs-vxworks*): Rework to handle arm-wrs-vxworks7 as
22377         well as arm-wrs-vxworks. Update target_cpu_name from arm6 (arch v3) to
22378         arm8 (arch v4).
22379         * config/arm/vxworks.h (MAYBE_TARGET_BPABI_CPP_BUILTINS): New, helper
22380         for TARGET_OS_CPP_BUILTIN.
22381         (TARGET_OS_CPP_BUILTIN): Invoke MAYBE_TARGET_BPABI_CPP_BUILTINS(),
22382         refine CPU definitions for arm_arch5 and add those for arm_arch6 and
22383         arm_arch7.
22384         (MAYBE_ASM_ABI_SPEC): New, helper for SUBTARGET_EXTRA_ASM_SPEC,
22385         passing required abi options to the assembler for EABI configurations.
22386         (EXTRA_CC1_SPEC): New macro, to help prevent the implicit production
22387         of .text.hot and .text.unlikely sections for kernel modules when
22388         using ARM style exceptions.
22389         (CC1_SPEC): Remove obsolete attempt at mimicking Diab toolchain
22390         options. Add EXTRA_CC1_SPEC.
22391         (VXWORKS_ENDIAN_SPEC): Adjust comment and remove handling of Diab
22392         toolchain options.
22393         (DWARF2_UNWIND_INFO): Redefine to handle the pre/post VxWorks 7
22394         transition.
22395         (ARM_TARGET2_DWARF_FORMAT): Define.
22396         * config/arm/t-vxworks: Adjust multilib control to removal of the
22397         Diab command line options.
22399 2017-08-01  Martin Liska  <mliska@suse.cz>
22401         PR gcov-profile/81561
22402         * gcov.c (unblock): Make unblocking safe as we need to preserve
22403         index correspondence of blocks and block_lists.
22405 2017-08-01  Richard Biener  <rguenther@suse.de>
22407         PR tree-optimization/81181
22408         * tree-ssa-pre.c (compute_antic_aux): Defer clean() to ...
22409         (compute_antic): ... end of iteration here.
22411 2017-08-01  James Greenhalgh  <james.greenhalgh@arm.com>
22413         * common.opt (ftree-vectorize): No longer set flag_tree_vectorize.
22414         (ftree-loop-vectorize): Set as EnabledBy ftree-vectorize.
22415         (ftree-slp-vectorize): Likewise.
22416         * omp-expand (expand_omp_simd): Remove flag_tree_vectorize, as it
22417         can no longer be set independent of flag_tree_loop_vectorize.
22418         * omp-general.c (emp_max_vf): Likewise.
22419         * opts.c (enable_fdo_optimizations): Remove references to
22420         flag_tree_vectorize, these are now implicit.
22421         (common_handle_option): Remove handling for OPT_ftree_vectorize,
22422         and leave it for the options machinery.
22424 2017-08-01  Martin Liska  <mliska@suse.cz>
22426         PR middle-end/70140
22427         * builtins.c (expand_builtin_memcpy_args): Remove.
22428         (expand_builtin_memcpy): Call newly added function
22429         expand_builtin_memory_copy_args.
22430         (expand_builtin_memcpy_with_bounds): Likewise.
22431         (expand_builtin_mempcpy): Remove last argument.
22432         (expand_builtin_mempcpy_with_bounds): Likewise.
22433         (expand_builtin_memory_copy_args): New function created from
22434         expand_builtin_mempcpy_args with small modifications.
22435         (expand_builtin_mempcpy_args): Remove.
22436         (expand_builtin_stpcpy): Remove unused argument.
22437         (expand_builtin): Likewise.
22438         (expand_builtin_with_bounds): Likewise.
22440 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22442         PR target/81641
22443         * config/i386/i386.c (ix86_print_operand_address_as): For -masm=intel
22444         print "ds:" only for immediates in generic address space.
22446 2017-08-01  Uros Bizjak  <ubizjak@gmail.com>
22448         PR target/81639
22449         * config/i386/i386.c (ix86_funciton_naked): New prototype.
22450         (ix86_function_ok_for_sibcall): Return false for naked functions.
22452 2017-08-01  Richard Biener  <rguenther@suse.de>
22454         * tree-ssa-pre.c (print_pre_expr): Handle NULL expr.
22455         (compute_antic): Seed worklist with exit block predecessors.
22456         * cfganal.c (dfs_find_deadend): For a cycle return the source
22457         of the edge closing it.
22459 2017-08-01  Tamar Christina  <tamar.christina@arm.com>
22461         * config/aarch64/aarch64.c
22462         (aarch64_can_const_movi_rtx_p): Move 0 check.
22464 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22466         * tree.h (POINTER_TYPE_OVERFLOW_UNDEFINED): Delete.
22467         * fold-const.c (fold_comparison, fold_binary_loc): Delete use of
22468         above macro.
22469         * match.pd: Ditto in address comparison pattern.
22471 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22473         PR tree-optimization/81627
22474         * tree-predcom.c (prepare_finalizers): Always rewrite into loop
22475         closed ssa form for store-store chain.
22477 2017-08-01  Bin Cheng  <bin.cheng@arm.com>
22479         PR tree-optimization/81620
22480         * tree-predcom.c (add_ref_to_chain): Don't set has_max_use_after
22481         for store-store chain.
22483 2017-08-01  Jakub Jelinek  <jakub@redhat.com>
22485         PR tree-optimization/81588
22486         * tree-ssa-reassoc.c (optimize_range_tests_var_bound): If
22487         ranges[i].in_p, invert comparison code ccode.  For >/>=,
22488         swap rhs1 and rhs2 and comparison code unconditionally,
22489         for </<= don't do that.  Don't swap rhs1/rhs2 again if
22490         ranges[i].in_p, instead invert comparison code ccode if
22491         opcode or oe->rank is BIT_IOR_EXPR.
22493         PR target/80846
22494         * optabs.def (vec_extract_optab, vec_init_optab): Change from
22495         a direct optab to conversion optab.
22496         * optabs.c (expand_vector_broadcast): Use convert_optab_handler
22497         with GET_MODE_INNER as last argument instead of optab_handler.
22498         * expmed.c (extract_bit_field_1): Likewise.  Use vector from
22499         vector extraction if possible and optab is available.
22500         * expr.c (store_constructor): Use convert_optab_handler instead
22501         of optab_handler.  Use vector initialization from smaller
22502         vectors if possible and optab is available.
22503         * tree-vect-stmts.c (vectorizable_load): Likewise.
22504         * doc/md.texi (vec_extract, vec_init): Document that the optabs
22505         now have two modes.
22506         * config/i386/i386.c (ix86_expand_vector_init): Handle expansion
22507         of vec_init from half-sized vectors with the same element mode.
22508         * config/i386/sse.md (ssehalfvecmode): Add V4TI case.
22509         (ssehalfvecmodelower, ssescalarmodelower): New mode attributes.
22510         (reduc_plus_scal_v8df, reduc_plus_scal_v4df, reduc_plus_scal_v2df,
22511         reduc_plus_scal_v16sf, reduc_plus_scal_v8sf, reduc_plus_scal_v4sf,
22512         reduc_<code>_scal_<mode>, reduc_umin_scal_v8hi): Add element mode
22513         after mode in gen_vec_extract* calls.
22514         (vec_extract<mode>): Renamed to ...
22515         (vec_extract<mode><ssescalarmodelower>): ... this.
22516         (vec_extract<mode><ssehalfvecmodelower>): New expander.
22517         (rotl<mode>3, rotr<mode>3, <shift_insn><mode>3, ashrv2di3): Add
22518         element mode after mode in gen_vec_init* calls.
22519         (VEC_INIT_HALF_MODE): New mode iterator.
22520         (vec_init<mode>): Renamed to ...
22521         (vec_init<mode><ssescalarmodelower>): ... this.
22522         (vec_init<mode><ssehalfvecmodelower>): New expander.
22523         * config/i386/mmx.md (vec_extractv2sf): Renamed to ...
22524         (vec_extractv2sfsf): ... this.
22525         (vec_initv2sf): Renamed to ...
22526         (vec_initv2sfsf): ... this.
22527         (vec_extractv2si): Renamed to ...
22528         (vec_extractv2sisi): ... this.
22529         (vec_initv2si): Renamed to ...
22530         (vec_initv2sisi): ... this.
22531         (vec_extractv4hi): Renamed to ...
22532         (vec_extractv4hihi): ... this.
22533         (vec_initv4hi): Renamed to ...
22534         (vec_initv4hihi): ... this.
22535         (vec_extractv8qi): Renamed to ...
22536         (vec_extractv8qiqi): ... this.
22537         (vec_initv8qi): Renamed to ...
22538         (vec_initv8qiqi): ... this.
22539         * config/rs6000/vector.md (VEC_base_l): New mode attribute.
22540         (vec_init<mode>): Renamed to ...
22541         (vec_init<mode><VEC_base_l>): ... this.
22542         (vec_extract<mode>): Renamed to ...
22543         (vec_extract<mode><VEC_base_l>): ... this.
22544         * config/rs6000/paired.md (vec_initv2sf): Renamed to ...
22545         (vec_initv2sfsf): ... this.
22546         * config/rs6000/altivec.md (splitter, altivec_copysign_v4sf3,
22547         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22548         vec_unpacku_lo_v8hi, mulv16qi3, altivec_vreve<mode>2): Add
22549         element mode after mode in gen_vec_init* calls.
22550         * config/aarch64/aarch64-simd.md (vec_init<mode>): Renamed to ...
22551         (vec_init<mode><Vel>): ... this.
22552         (vec_extract<mode>): Renamed to ...
22553         (vec_extract<mode><Vel>): ... this.
22554         * config/aarch64/iterators.md (Vel): New mode attribute.
22555         * config/s390/s390.c (s390_expand_vec_strlen, s390_expand_vec_movstr):
22556         Add element mode after mode in gen_vec_extract* calls.
22557         * config/s390/vector.md (non_vec_l): New mode attribute.
22558         (vec_extract<mode>): Renamed to ...
22559         (vec_extract<mode><non_vec_l>): ... this.
22560         (vec_init<mode>): Renamed to ...
22561         (vec_init<mode><non_vec_l>): ... this.
22562         * config/s390/s390-builtins.def (s390_vlgvb, s390_vlgvh, s390_vlgvf,
22563         s390_vlgvf_flt, s390_vlgvg, s390_vlgvg_dbl): Add element mode after
22564         vec_extract mode.
22565         * config/arm/iterators.md (V_elem_l): New mode attribute.
22566         * config/arm/neon.md (vec_extract<mode>): Renamed to ...
22567         (vec_extract<mode><V_elem_l>): ... this.
22568         (vec_extractv2di): Renamed to ...
22569         (vec_extractv2didi): ... this.
22570         (vec_init<mode>): Renamed to ...
22571         (vec_init<mode><V_elem_l>): ... this.
22572         (reduc_plus_scal_<mode>, reduc_plus_scal_v2di, reduc_smin_scal_<mode>,
22573         reduc_smax_scal_<mode>, reduc_umin_scal_<mode>,
22574         reduc_umax_scal_<mode>, neon_vget_lane<mode>, neon_vget_laneu<mode>):
22575         Add element mode after gen_vec_extract* calls.
22576         * config/mips/mips-msa.md (vec_init<mode>): Renamed to ...
22577         (vec_init<mode><unitmode>): ... this.
22578         (vec_extract<mode>): Renamed to ...
22579         (vec_extract<mode><unitmode>): ... this.
22580         * config/mips/loongson.md (vec_init<mode>): Renamed to ...
22581         (vec_init<mode><unitmode>): ... this.
22582         * config/mips/mips-ps-3d.md (vec_initv2sf): Renamed to ...
22583         (vec_initv2sfsf): ... this.
22584         (vec_extractv2sf): Renamed to ...
22585         (vec_extractv2sfsf): ... this.
22586         (reduc_plus_scal_v2sf, reduc_smin_scal_v2sf, reduc_smax_scal_v2sf):
22587         Add element mode after gen_vec_extract* calls.
22588         * config/mips/mips.md (unitmode): New mode iterator.
22589         * config/spu/spu.c (spu_expand_prologue, spu_allocate_stack,
22590         spu_builtin_extract): Add element mode after gen_vec_extract* calls.
22591         * config/spu/spu.md (inner_l): New mode attribute.
22592         (vec_init<mode>): Renamed to ...
22593         (vec_init<mode><inner_l>): ... this.
22594         (vec_extract<mode>): Renamed to ...
22595         (vec_extract<mode><inner_l>): ... this.
22596         * config/sparc/sparc.md (veltmode): New mode iterator.
22597         (vec_init<VMALL:mode>): Renamed to ...
22598         (vec_init<VMALL:mode><VMALL:veltmode>): ... this.
22599         * config/ia64/vect.md (vec_initv2si): Renamed to ...
22600         (vec_initv2sisi): ... this.
22601         (vec_initv2sf): Renamed to ...
22602         (vec_initv2sfsf): ... this.
22603         (vec_extractv2sf): Renamed to ...
22604         (vec_extractv2sfsf): ... this.
22605         * config/powerpcspe/vector.md (VEC_base_l): New mode attribute.
22606         (vec_init<mode>): Renamed to ...
22607         (vec_init<mode><VEC_base_l>): ... this.
22608         (vec_extract<mode>): Renamed to ...
22609         (vec_extract<mode><VEC_base_l>): ... this.
22610         * config/powerpcspe/paired.md (vec_initv2sf): Renamed to ...
22611         (vec_initv2sfsf): ... this.
22612         * config/powerpcspe/altivec.md (splitter, altivec_copysign_v4sf3,
22613         vec_unpacku_hi_v16qi, vec_unpacku_hi_v8hi, vec_unpacku_lo_v16qi,
22614         vec_unpacku_lo_v8hi, mulv16qi3): Add element mode after mode in
22615         gen_vec_init* calls.
22617 2017-08-01  Richard Biener  <rguenther@suse.de>
22619         PR tree-optimization/81297
22620         * tree-vrp.c (get_single_symbol): Remove assert, instead drop
22621         TREE_OVERFLOW from INTEGER_CSTs.
22623 2017-07-31  Segher Boessenkool  <segher@kernel.crashing.org>
22625         * config/rs6000/rs6000.c (enum rs6000_reg_type): Delete trailing comma.
22627 2017-07-31  Carl Love  <cel@us.ibm.com>
22629         * config/rs6000/rs6000-c: Add support for built-in functions
22630         vector signed char vec_xl_be (signed long long, signed char *);
22631         vector unsigned char vec_xl_be (signed long long, unsigned char *);
22632         vector signed int vec_xl_be (signed long long, signed int *);
22633         vector unsigned int vec_xl_be (signed long long, unsigned int *);
22634         vector signed long long vec_xl_be (signed long long, signed long long *);
22635         vector unsigned long long vec_xl_be (signed long long, unsigned long long *);
22636         vector signed short vec_xl_be (signed long long, signed short *);
22637         vector unsigned short vec_xl_be (signed long long, unsigned short *);
22638         vector double vec_xl_be (signed long long, double *);
22639         vector float vec_xl_be (signed long long, float *);
22640         * config/rs6000/altivec.h (vec_xl_be): Add #define.
22641         * config/rs6000/rs6000-builtin.def (XL_BE_V16QI, XL_BE_V8HI, XL_BE_V4SI)
22642         XL_BE_V2DI, XL_BE_V4SF, XL_BE_V2DF, XL_BE): Add definitions
22643         for the builtins.
22644         * config/rs6000/rs6000.c (altivec_expand_xl_be_builtin): Add function.
22645         (altivec_expand_builtin): Add switch statement to call
22646         altivec_expand_xl_be for each builtin.
22647         (altivec_init_builtins): Add def_builtin for _builtin_vsx_le_be_v8hi,
22648         __builtin_vsx_le_be_v4si, __builtin_vsx_le_be_v2di,
22649         __builtin_vsx_le_be_v4sf, __builtin_vsx_le_be_v2df,
22650         __builtin_vsx_le_be_v16qi.
22651         * doc/extend.texi: Update the built-in documentation file for the
22652         new built-in functions.
22654 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22656         PR target/25967
22657         * config/i386/i386.c (ix86_allocate_stack_slots_for_args):
22658         New function.
22659         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
22661 2017-07-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
22663         * config.gcc: Add z14.
22664         * config/s390/driver-native.c (s390_host_detect_local_cpu): Add
22665         CPU model numbers for z13s and z14.
22666         * config/s390/s390-c.c (s390_resolve_overloaded_builtin): Replace
22667         arch12 with z14.
22668         * config/s390/s390-opts.h (enum processor_type): Rename
22669         PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22670         * config/s390/s390.c (processor_table): Add field for CPU name to
22671         be passed to Binutils.
22672         (s390_asm_output_machine_for_arch): Use the new field in
22673         processor_table for Binutils.
22674         (s390_expand_builtin): Replace arch12 with z14.
22675         (s390_issue_rate): Rename PROCESSOR_ARCH12 to PROCESSOR_3906_Z14.
22676         (s390_get_sched_attrmask): Likewise.
22677         (s390_get_unit_mask): Likewise.
22678         * config/s390/s390.opt: Add z14 to processor_type enum.
22680 2017-07-31  Martin Jambor  <mjambor@suse.cz>
22682         PR hsa/81477
22683         * ipa-fnsummary.c (ipa_fn_summary_generate): Set versionable
22684         regardless of optimization level.
22686 2017-07-31  Jan Hubicka <hubicka@ucw.cz>
22687             Martin Liska  <mliska@suse.cz>
22689         * predict.def: Remove old comment and adjust probability.
22690         * gimplify.c (should_warn_for_implicit_fallthrough): Ignore
22691         PREDICT statements.
22693 2017-07-31  Uros Bizjak  <ubizjak@gmail.com>
22695         PR target/25967
22696         * config/i386/i386.c (ix86_function_naked): New function.
22697         (ix86_can_use_return_insn_p): Return false for naked functions.
22698         (ix86_expand_prologue): Skip prologue for naked functions.
22699         (ix86_expand_epilogue): Skip epilogue for naked functions
22700         and emit trap instruction.
22701         (ix86_warn_func_return): New function.
22702         (ix86_attribute_table): Add "naked" attribute specification.
22703         (TARGET_WARN_FUNC_RETURN): Define.
22704         * doc/extend.texi (x86 Function Attributes) <naked>: Document it.
22706 2017-07-31  Martin Liska  <mliska@suse.cz>
22708         * gimple-pretty-print.c (dump_gimple_label): Never dump BB info.
22709         (dump_gimple_bb_header): Always dump BB info.
22710         (pp_cfg_jump): Do not append info about BB when dumping a jump.
22712 2017-07-31  Martin Liska  <mliska@suse.cz>
22714         PR sanitize/81530
22715         * convert.c (convert_to_integer_1): Guard condition with flag_sanitize_p
22716         also with current_function_decl non-null equality.
22718 2017-07-31  Jakub Jelinek  <jakub@redhat.com>
22720         PR sanitizer/81604
22721         * ubsan.c (ubsan_type_descriptor): For UBSAN_PRINT_ARRAY don't
22722         change type to the element type, instead add eltype variable and
22723         use it where we are interested in the element type.
22725         PR tree-optimization/81603
22726         * ipa-polymorphic-call.c
22727         (ipa_polymorphic_call_context::ipa_polymorphic_call_context): Perform
22728         offset arithmetic in offset_int, bail out if the resulting bit offset
22729         doesn't fit into shwi.
22731 2017-07-31  Martin Liska  <mliska@suse.cz>
22733         * gimplify.c (mostly_copy_tree_r): Remove Java specific hunk.
22734         (gimplify_save_expr): Fix comment.
22736 2017-07-30  H.J. Lu  <hongjiu.lu@intel.com>
22738         PR target/79793
22739         * config/i386/i386.c (ix86_function_arg): Update arguments for
22740         exception handler.
22741         (ix86_compute_frame_layout): Set the initial stack offset to
22742         INCOMING_FRAME_SP_OFFSET.  Update red-zone offset with
22743         INCOMING_FRAME_SP_OFFSET.
22744         (ix86_expand_epilogue): Don't pop the 'ERROR_CODE' off the
22745         stack before exception handler returns.
22746         * config/i386/i386.h (INCOMING_FRAME_SP_OFFSET): Add the
22747         the 'ERROR_CODE' for exception handler.
22749 2017-07-30  Uros Bizjak  <ubizjak@gmail.com>
22751         * config/i386/i386.h (ASM_PRINTF_EXTENSIONS): New macro.
22752         (ASM_OUTPUT_REG_PUSH): Rewrite with new operand modifiers.
22753         (ASM_OUTPUT_REG_POP): Ditto.
22754         * config/i386/i386.c (ix86_asm_output_function_label): Use fputs
22755         instead of asm_fprintf to output pure string.
22757 2017-07-29  Jakub Jelinek  <jakub@redhat.com>
22759         * debug.h (struct gcc_debug_hooks): Add IMPLICIT argument
22760         to imported_module_or_decl hook.
22761         (debug_nothing_tree_tree_tree_bool): Remove.
22762         (debug_nothing_tree_tree_tree_bool_bool): New declaration.
22763         * debug.c (do_nothing_debug_hooks): Use
22764         debug_nothing_tree_tree_tree_bool_bool instead of
22765         debug_nothing_tree_tree_tree_bool.
22766         * vmsdbgout.c (vmsdbg_debug_hooks): Likewise.
22767         * dbxout.c (dbx_debug_hooks, xcoff_debug_hooks): Likewise.
22768         * sdbout.c (sdb_debug_hooks): Likewise.
22769         * dwarf2out.c (dwarf2_lineno_debug_hooks): Likewise.
22770         (gen_namespace_die): Add DW_AT_export_symbols attribute if
22771         langhook wants it.
22772         (dwarf2out_imported_module_or_decl): Add IMPLICIT argument,
22773         if true, -gdwarf-5 and decl will have DW_AT_export_symbols
22774         attribute, don't add anything.
22776 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22778         * fold-const.c (fold_build1_stat_loc): Remove _stat from name.
22779         (fold_build2_stat_loc): Likewise.
22780         (fold_build3_stat_loc): Likewise.
22781         * fold-const.h (fold_build1, fold_build2, fold_build3): Adjust.
22782         (fold_build1_loc): Remove macro.
22783         (fold_build2_loc): Likewise.
22784         (fold_build3_loc): Likewise.
22786 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22788         * gimple.c (gimple_build_debug_bind_stat): Remove _stat from name.
22789         (gimple_build_debug_bind_source_stat): Likewise.
22790         * gimple.h (gimple_build_debug_bind): Remove macro.
22791         (gimple_build_debug_bind_source): Likewise.
22793 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22795         * bitmap.c (bitmap_alloc): Adjust.
22796         (bitmap_gc_alloc): Likewise.
22797         * bitmap.h (bitmap_initialize_stat): Remove _stat from name.
22799 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22801         * bitmap.c (bitmap_obstack_alloc_stat): Rename to bitmap_alloc.
22802         (bitmap_gc_alloc_stat): Rename to bitmap_gc_alloc.
22803         * bitmap.h (bitmap_obstack_alloc_stat): Adjust prototype.
22804         (bitmap_gc_alloc_stat): Likewise.
22805         (BITMAP_ALLOC, BITMAP_GGC_ALLOC): Adjust.
22807 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22809         * rtl.c (shallow_copy_rtx_stat): Remove _stat from name.
22810         * rtl.h (shallow_copy_rtx): Remove macro.
22812 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22814         * emit-rtl.c (gen_raw_REG): Adjust.
22815         * gengenrtl.c (gendef): Likewise.
22816         * rtl.c (rtx_alloc_stat): Remove _stat from name.
22817         * rtl.h (rtx_alloc): Remove macro.
22819 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22821         * tree.c (build_tree_list_vec_stat): Remove _stat from name.
22822         (build_tree_list_stat): Likewise.
22823         * tree.h (build_tree_list): Remove macro.
22824         (build_tree_list_vec): Likewise.
22826 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22828         * tree.c (make_vector_stat): Remove _stat from name.
22829         (build_vector_stat): Likewise.
22830         * tree.h (make_vector_stat): Remove macro.
22831         (build_vector_stat): Likewise.
22833 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22835         * tree.h (build_var_debug_value): Remove prototype.
22837 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22839         * tree.c (tree_cons_stat): Remove _stat from name.
22840         * tree.h (tree_cons): Remove macro.
22842 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22844         * tree.c (build_vl_exp_stat): Remove _stat from name.
22845         * tree.h (build_vl_exp): Remove macro.
22847 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22849         * tree.c (build_decl_stat): Remove _stat from name.
22850         * tree.h (build_decl): Remove macro.
22852 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22854         * gimple.c (gimple_build_with_ops_stat): Adjust.
22855         (gimple_alloc_stat): Remove _stat from name.
22856         * gimple.h (gimple_alloc): Remove macro.
22858 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22860         * tree.c (make_tree_vec_stat): Remove _stat from name.
22861         (grow_tree_vec_stat): Likewise.
22862         * tree.h (make_tree_vec_stat): Adjust prototype.
22863         (grow_tree_vec_stat): Likewise.
22864         (make_tree_vec): Remove macro.
22865         (grow_tree_vec): Likewise.
22867 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22869         * fold-const.c (fold_build1_stat_loc): Adjust.
22870         (fold_build2_stat_loc): Likewise.
22871         (fold_build3_stat_loc): Likewise.
22872         * tree.c (build0_stat): Remove _stat from name.
22873         (build1_stat): Likewise.
22874         (build2_stat): Likewise.
22875         (build3_stat): Likewise.
22876         (build4_stat): Likewise.
22877         (build5_stat): Likewise.
22878         * tree.h (build1_loc): Remove macro, and rename _stat function
22879         to this.
22880         (build2_loc): Likewise.
22881         (build3_loc): Likewise.
22882         (build4_loc): Likewise.
22883         (build5_loc): Likewise.
22885 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22887         * tree.c (make_int_cst_stat): Remove _stat from name.
22888         * tree.h (make_int_cst_stat): Adjust prototype.
22889         (make_int_cst): Remove macro.
22891 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22893         * tree.c (make_tre_binfo_stat): Remove _stat from name.
22894         * tree.h (make_tree_binfo_stat): Adjust prototype.
22895         (make_tree_binfo): Remove.
22897 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22899         * tree.c (copy_node_stat): Rename to copy_node.
22900         (build_distinct_type_copy): Adjust.
22901         * tree.h (copy_node_stat): Adjust prototype.
22902         (copy_node): Remove macro.
22904 2017-07-28  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
22906         * tree.c (make_node_stat): rename to make_node.
22907         (build_tree_list_stat): Adjust.
22908         (build0_stat): Likewise.
22909         (build2_stat): Likewise.
22910         (build3_stat): Likewise.
22911         (build4_stat): Likewise.
22912         (build5_stat): Likewise.
22913         (build_decl_stat): Likewise.
22914         * tree.h (make_node_stat): Adjust prototype.
22915         (make_node): remove macro.
22917 2017-07-28  Peter Bergner  <bergner@vnet.ibm.com>
22919         * config/rs6000/ppc-auxv.h (PPC_FEATURE2_DARN): New define.
22920         (PPC_FEATURE2_SCV): Likewise.
22921         * config/rs6000/rs6000.c (cpu_supports_info): Use them.
22923 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22925         * config/aarch64/aarch64.c
22926         (aarch64_internal_mov_immediate): Add new special pattern.
22927         * config/aarch64/aarch64.md (*movdi_aarch64):
22928         Add reg/32bit const mov case.
22930 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22931             Richard Sandiford <richard.sandiford@linaro.org>
22933         * config/aarch64/aarch64.md (mov<mode>): Generalize.
22934         (*movhf_aarch64, *movsf_aarch64, *movdf_aarch64):
22935         Add integer and movi cases.
22936         (movi-split-hf-df-sf split, fp16): New.
22937         (enabled): Added TARGET_FP_F16INST.
22938         * config/aarch64/iterators.md (GPF_HF): New.
22939         * config/aarch64/predicates.md (aarch64_reg_or_fp_float): New.
22941 2017-07-28  Tamar Christina  <tamar.christina@arm.com>
22943         * config/aarch64/aarch64.c
22944         (aarch64_simd_container_mode): Add prototype.
22945         (aarch64_expand_mov_immediate): Add HI support.
22946         (aarch64_reinterpret_float_as_int, aarch64_float_const_rtx_p: New.
22947         (aarch64_can_const_movi_rtx_p): New.
22948         (aarch64_preferred_reload_class):
22949         Remove restrictions of using FP registers for certain SIMD operations.
22950         (aarch64_rtx_costs): Added new cost for CONST_DOUBLE moves.
22951         (aarch64_valid_floating_const): Add integer move validation.
22952         (aarch64_simd_imm_scalar_p): Remove.
22953         (aarch64_output_scalar_simd_mov_immediate): Generalize function.
22954         (aarch64_legitimate_constant_p): Expand list of supported cases.
22955         * config/aarch64/aarch64-protos.h
22956         (aarch64_float_const_rtx_p, aarch64_can_const_movi_rtx_p): New.
22957         (aarch64_reinterpret_float_as_int): New.
22958         (aarch64_simd_imm_scalar_p): Remove.
22959         * config/aarch64/constraints.md (Uvi): New.
22960         (Dd): Split into Ds and new Dd.
22961         * config/aarch64/aarch64.md (*movsi_aarch64):
22962         Add SIMD mov case.
22963         (*movdi_aarch64): Add SIMD mov case.
22965 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22967         * tree-predcom.c: (struct chain): Handle store-store chain in which
22968         stores for elimination only store loop invariant values.
22969         (execute_pred_commoning_chain): Ditto.
22970         (prepare_initializers_chain_store_elim): Ditto.
22971         (prepare_finalizers): Ditto.
22972         (is_inv_store_elimination_chain): New function.
22973         (initialize_root_vars_store_elim_1): New function.
22975 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
22977         * tree-predcom.c: Revise general description of the pass.
22978         (enum chain_type): New enum type for store elimination.
22979         (struct chain): New field supporting store elimination.
22980         (struct component): Ditto.
22981         (dump_chain): Dump store-stores chain.
22982         (release_chain): Release resources.
22983         (split_data_refs_to_components): Compute and create component
22984         contains only stores for elimination.
22985         (get_chain_last_ref_at): New function.
22986         (make_invariant_chain): Initialization.
22987         (make_rooted_chain): Specify chain type in parameter and record it.
22988         (add_looparound_copies): Skip for store-stores chain.
22989         (determine_roots_comp): Compute type of chain and pass it to
22990         make_rooted_chain.
22991         (initialize_root_vars_store_elim_2): New function.
22992         (finalize_eliminated_stores): New function.
22993         (remove_stmt): Handle store for elimination.
22994         (execute_pred_commoning_chain): Execute predictive commoning on
22995         store-store chains.
22996         (determine_unroll_factor): Skip unroll for store-stores chain.
22997         (prepare_initializers_chain_store_elim): New function.
22998         (prepare_initializers_chain): Hanlde store-store chain.
22999         (prepare_finalizers_chain, prepare_finalizers): New function.
23000         (tree_predictive_commoning_loop): Return integer value indicating
23001         if loop is unrolled or lcssa form is corrupted.
23002         (tree_predictive_commoning): Rewrite for lcssa form if necessary.
23004 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23006         * tree-predcom.c (initialize_root): Delete.
23007         (execute_pred_commoning_chain): Initialize root vars and replace
23008         reference of non-combined chain directly, rather than call above
23009         function.
23011 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23013         * tree-predcom.c (ref_at_iteration): Add parameter NITERS.  Compute
23014         memory reference to DR at (NITERS + ITERS)-th iteration of loop.
23016 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23018         * tree-predcom.c (struct chain): New field init_seq.
23019         (release_chain): Release init_seq.
23020         (prepare_initializers_chain): Record intialization stmts in above
23021         field.
23022         (insert_init_seqs): New function.
23023         (tree_predictive_commoning_loop): Call insert_init_seqs.
23025 2017-07-28  Bin Cheng  <bin.cheng@arm.com>
23027         * tree-predcom.c (determine_roots_comp): Skip trivial components.
23029 2017-07-28  Richard Biener  <rguenther@suse.de>
23031         * match.pd: Remove superfluous :c.
23032         * genmatch.c (simplify::id): Add member.
23033         (lower_commutative, lower_opt_convert, lower_cond, lower_for):
23034         Copy id.
23035         (current_id): New global.
23036         (dt_node::parent): Move from ...
23037         (dt_operand::parent): ... here.  Add for_id member.
23038         (is_a_helper <dt_operand *>::test): DT_TRUE is also a dt_operand.
23039         (decision_tree::find_node): Relax order requirement when
23040         merging DT_TRUE nodes to ones inbetween the current simplify
23041         and the one we try to merge with.  Add diagnostic whenever
23042         we need to enforce pattern order by not merging.
23043         (decision_tree::insert): Set current_id.
23044         (decision_tree::print_node): Dump parent node and for_id.
23045         (parser::last_id): Add member.
23046         (parser::push_simplify): Assign unique id.
23047         (parser::parser): Initialize last_id.
23049 2017-07-28  Martin Liska  <mliska@suse.cz>
23051         PR sanitizer/81340
23052         * sanopt.c (sanitize_rewrite_addressable_params): Set VALUE_EXPR after
23053         gimple_build_debug_bind.
23055 2017-07-28  Richard Biener  <rguenther@suse.de>
23057         PR tree-optimization/81502
23058         * match.pd: Add pattern combining BIT_INSERT_EXPR with
23059         BIT_FIELD_REF.
23060         * tree-cfg.c (verify_expr): Verify types of BIT_FIELD_REF
23061         size/pos operands.
23062         (verify_gimple_assign_ternary): Likewise for BIT_INSERT_EXPR pos.
23063         * gimple-fold.c (maybe_canonicalize_mem_ref_addr): Use bitsizetype
23064         for BIT_FIELD_REF args.
23065         * fold-const.c (make_bit_field_ref): Likewise.
23066         * tree-vect-stmts.c (vectorizable_simd_clone_call): Likewise.
23068 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23070         PR sanitizer/80998
23071         * sanopt.c (pass_sanopt::execute): Handle IFN_UBSAN_PTR.
23072         * tree-ssa-alias.c (call_may_clobber_ref_p_1): Likewise.
23073         * flag-types.h (enum sanitize_code): Add SANITIZER_POINTER_OVERFLOW.
23074         Or it into SANITIZER_UNDEFINED.
23075         * ubsan.c: Include gimple-fold.h and varasm.h.
23076         (ubsan_expand_ptr_ifn): New function.
23077         (instrument_pointer_overflow): New function.
23078         (maybe_instrument_pointer_overflow): New function.
23079         (instrument_object_size): Formatting fix.
23080         (pass_ubsan::execute): Call instrument_pointer_overflow
23081         and maybe_instrument_pointer_overflow.
23082         * internal-fn.c (expand_UBSAN_PTR): New function.
23083         * ubsan.h (ubsan_expand_ptr_ifn): Declare.
23084         * sanitizer.def (__ubsan_handle_pointer_overflow,
23085         __ubsan_handle_pointer_overflow_abort): New builtins.
23086         * tree-ssa-tail-merge.c (merge_stmts_p): Handle IFN_UBSAN_PTR.
23087         * internal-fn.def (UBSAN_PTR): New internal function.
23088         * opts.c (sanitizer_opts): Add pointer-overflow.
23089         * lto-streamer-in.c (input_function): Handle IFN_UBSAN_PTR.
23090         * fold-const.c (build_range_check): Compute pointer range check in
23091         integral type if pointer arithmetics would be needed.  Formatting
23092         fixes.
23094 2017-07-28  Martin Liska  <mliska@suse.cz>
23096         PR sanitizer/81460
23097         * sanopt.c (sanitize_rewrite_addressable_params): Do not rewrite
23098         parameters that are of a variable-length.
23100 2017-07-28  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23102         * config.gcc (powerpc-*-rtems*): Remove rs6000/eabi.h.  Add
23103         rs6000/biarch64.h.
23104         * config/rs6000/rtems.h (ASM_DECLARE_FUNCTION_SIZE): New macro.
23105         (ASM_OUTPUT_SPECIAL_POOL_ENTRY_P): Likewise.
23106         (CRT_CALL_STATIC_FUNCTION): Likewise.
23107         (ASM_DEFAULT_SPEC): New define.
23108         (ASM_SPEC32): Likewise.
23109         (ASM_SPEC64): Likewise.
23110         (ASM_SPEC_COMMON): Likewise.
23111         (ASM_SPEC): Likewise.
23112         (INVALID_64BIT): Likewise.
23113         (LINK_OS_DEFAULT_SPEC): Likewise.
23114         (LINK_OS_SPEC32): Likewise.
23115         (LINK_OS_SPEC64): Likewise.
23116         (POWERPC_LINUX): Likewise.
23117         (PTRDIFF_TYPE): Likewise.
23118         (RESTORE_FP_PREFIX): Likewise.
23119         (RESTORE_FP_SUFFIX): Likewise.
23120         (SAVE_FP_PREFIX): Likewise.
23121         (SAVE_FP_SUFFIX): Likewise.
23122         (SIZE_TYPE): Likewise.
23123         (SUBSUBTARGET_OVERRIDE_OPTIONS): Likewise.
23124         (TARGET_64BIT): Likewise.
23125         (TARGET_64BIT): Likewise.
23126         (TARGET_AIX): Likewise.
23127         (WCHAR_TYPE_SIZE): Likewise.
23128         (WCHAR_TYPE): Undefine.
23129         (TARGET_OS_CPP_BUILTINS): Add 64-bit PowerPC defines.
23130         (CPP_OS_DEFAULT_SPEC): Use previous CPP_OS_RTEMS_SPEC.
23131         (CPP_OS_RTEMS_SPEC): Delete.
23132         (SUBSUBTARGET_EXTRA_SPECS): Remove cpp_os_rtems.  Add
23133         asm_spec_common, asm_spec32, asm_spec64, link_os_spec32, and
23134         link_os_spec64.
23135         * config/rs6000/t-rtems: Add mcpu=e6500/m64 multilibs.
23137 2017-07-28  Jakub Jelinek  <jakub@redhat.com>
23139         PR tree-optimization/81578
23140         * tree-parloops.c (build_new_reduction): Bail out if
23141         reduction_code isn't one of the standard OpenMP reductions.
23142         Move the details printing after that decision.
23144 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23146         * config/rs6000/predicates.md (volatile_mem_operand): Remove code
23147         related to reload_in_progress.
23148         (splat_input_operand): Likewise.
23149         * config/rs6000/rs6000-protos.h (rs6000_secondary_memory_needed_rtx):
23150         Delete prototype.
23151         * config/rs6000/rs6000.c (machine_function): Remove sdmode_stack_slot
23152         field.
23153         (TARGET_EXPAND_TO_RTL_HOOK): Delete.
23154         (TARGET_INSTANTIATE_DECLS): Likewise.
23155         (legitimate_indexed_address_p): Delete reload_in_progress code.
23156         (rs6000_debug_legitimate_address_p): Likewise.
23157         (rs6000_eliminate_indexed_memrefs): Likewise.
23158         (rs6000_emit_le_vsx_store): Likewise.
23159         (rs6000_emit_move_si_sf_subreg): Likewise.
23160         (rs6000_emit_move): Likewise.
23161         (register_to_reg_type): Likewise.
23162         (rs6000_pre_atomic_barrier): Likewise.
23163         (rs6000_machopic_legitimize_pic_address): Likewise.
23164         (rs6000_allocate_stack_temp): Likewise.
23165         (rs6000_address_for_fpconvert): Likewise.
23166         (rs6000_address_for_altivec): Likewise.
23167         (rs6000_secondary_memory_needed_rtx): Delete function.
23168         (rs6000_check_sdmode): Likewise.
23169         (rs6000_alloc_sdmode_stack_slot): Likewise.
23170         (rs6000_instantiate_decls): Likewise.
23171         * config/rs6000/rs6000.h (SECONDARY_MEMORY_NEEDED_RTX): Delete.
23172         * config/rs6000/rs6000.md (splitter for *movsi_got_internal):
23173         Delete reload_in_progress.
23174         (*vec_reload_and_plus_<mptrsize>): Likewise.
23175         * config/rs6000/vsx.md (vsx_mul_v2di): Likewise.
23176         (vsx_div_v2di): Likewise.
23177         (vsx_udiv_v2di): Likewise.
23179 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23181         * config/rs6000/rs6000.opt (mlra): Replace with stub.
23182         * config/rs6000/rs6000-cpus.def (POWERPC_MASKS): Delete OPTION_MASK_LRA.
23183         * config/rs6000/rs6000.c (TARGET_LRA_P): Delete.
23184         (rs6000_debug_reg_global): Delete print of LRA status.
23185         (rs6000_option_override_internal): Delete dead LRA related code.
23186         (rs6000_lra_p): Delete function.
23187         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete -mlra.
23189 2017-07-27  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23191         * config.gcc (riscv*-*-elf*): Add (riscv*-*-rtems*).
23192         * config/riscv/rtems.h: New file.
23194 2017-07-27  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
23195             Sudakshina Das  <sudi.das@arm.com>
23197         * config/aarch64/aarch64.md
23198         (define_split for and<mode>3nr_compare): Move
23199         non aarch64_logical_operand to a register.
23200         (define_split for and_<SHIFT:optab><mode>3nr_compare0): Move non
23201         register immediate operand to a register.
23202         * config/aarch64/predicates.md (aarch64_mov_imm_operand): New.
23204 2017-07-27  Peter Bergner  <bergner@vnet.ibm.com>
23206         PR middle-end/81564
23207         * tree-cfg.c (group_case_labels_stmt): Handle already deleted blocks.
23209 2017-07-27  Richard Biener  <rguenther@suse.de>
23211         PR tree-optimization/81573
23212         PR tree-optimization/81494
23213         * tree-vect-loop.c (vect_create_epilog_for_reduction): Handle
23214         multi defuse cycle case.
23216 2017-07-27  Richard Biener  <rguenther@suse.de>
23218         PR tree-optimization/81571
23219         * tree-vect-slp.c (vect_build_slp_tree): Properly verify reduction
23220         PHIs.
23222 2017-07-27  Eric Botcazou  <ebotcazou@adacore.com>
23224         * config/sparc/sparc.c (sparc_option_override): Set MASK_FSMULD flag
23225         earlier and only if MASK_FPU is set.  Adjust formatting.
23227 2017-07-27  Martin Liska  <mliska@suse.cz>
23229         * opt-functions.awk: Add validation of value of Init.
23230         * optc-gen.awk: Pass new argument.
23232 2017-07-27  Martin Liska  <mliska@suse.cz>
23234         * auto-profile.c (autofdo_source_profile::update_inlined_ind_target):
23235         Fix wrong condition.
23237 2017-07-27  Martin Liska  <mliska@suse.cz>
23239         * auto-profile.c (afdo_annotate_cfg): Assign zero counts to
23240         BBs and edges seen by autoFDO.
23242 2017-07-27  Richard Biener  <rguenther@suse.de>
23244         PR tree-optimization/81502
23245         * tree-ssa.c (non_rewritable_lvalue_p): Handle BIT_INSERT_EXPR
23246         with incompatible but same sized type.
23247         (execute_update_addresses_taken): Likewise.
23249 2017-07-27  James Greenhalgh  <james.greenhalgh@arm.com>
23251         * tree-ssa-loop-ch.c (pass_ch::process_loop_p): Guard on
23252         flag_tree_loop_vectorize rather than flag_tree_vectorize.
23254 2017-07-27  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
23256         PR target/81534
23257         * config/s390/s390.md ("*atomic_compare_and_swap<mode>_1")
23258         ("*atomic_compare_and_swapdi_2", "*atomic_compare_and_swapsi_3"):
23259         Change s_operand to memory_operand.
23261 2017-07-27  Richard Sandiford  <richard.sandiford@linaro.org>
23263         * config/rs6000/rs6000-protos.h (rs6000_emit_le_vsx_permute): Declare.
23264         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Replace with...
23265         (rs6000_emit_le_vsx_permute): ...this.  Take the destination as input.
23266         Emit instructions rather than returning an expression.  Handle TFmode
23267         and KFmode by casting to TImode.
23268         (rs6000_emit_le_vsx_load): Update to use rs6000_emit_le_vsx_permute.
23269         (rs6000_emit_le_vsx_store): Likewise.
23270         * config/rs6000/vsx.md (VSX_TI): New iterator.
23271         (*vsx_le_permute_<mode>): Use it instead of VSX_LE_128.
23272         (*vsx_le_undo_permute_<mode>): Likewise.
23273         (*vsx_le_perm_load_<mode>): Use rs6000_emit_le_vsx_permute to
23274         emit the split sequence.
23275         (*vsx_le_perm_store_<mode>): Likewise.
23277 2017-07-27  Jakub Jelinek  <jakub@redhat.com>
23279         PR tree-optimization/81555
23280         PR tree-optimization/81556
23281         * tree-ssa-reassoc.c (rewrite_expr_tree): Add NEXT_CHANGED argument,
23282         if true, force CHANGED for the recursive invocation.
23283         (reassociate_bb): Remember original length of ops array, pass
23284         len != orig_len as NEXT_CHANGED in rewrite_expr_tree call.
23286         * attribs.c (decl_attributes): Imply noinline, noclone and no_icf
23287         attributes for noipa attribute.  For naked attribute use
23288         lookup_attribute first before lookup_attribute_spec.
23289         * final.c (rest_of_handle_final): Disable IPA RA for functions with
23290         noipa attribute.
23291         * ipa-visibility.c (non_local_p): Fix comment typos.  Return true
23292         for functions with noipa attribute.
23293         (cgraph_externally_visible_p): Return true for functions with noipa
23294         attribute.
23295         * cgraph.c (cgraph_node::get_availability): Return AVAIL_INTERPOSABLE
23296         for functions with noipa attribute.
23297         * doc/extend.texi: Document noipa function attribute.
23298         * tree-ssa-structalias.c (refered_from_nonlocal_fn): Set *nonlocal_p
23299         also for functions with noipa attribute.
23300         (ipa_pta_execute): Set nonlocal_p also for nodes with noipa attribute.
23302 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23304         * config/aarch64/aarch64.c (thunderx_vector_cost): Decrease cost of
23305         vec_unalign_load_cost and vec_unalign_store_cost.
23307 2017-07-26  Michael Meissner  <meissner@linux.vnet.ibm.com>
23309         * config/rs6000/rs6000-cpus.def (ISA_2_7_MASKS_SERVER): Delete
23310         -mvsx-small-integer option.
23311         (ISA_3_0_MASKS_IEEE): Likewise.
23312         (OTHER_VSX_VECTOR_MASKS): Likewise.
23313         (POWERPC_MASKS): Likewise.
23314         * config/rs6000/rs6000.opt (-mvsx-small-integer): Likewise.
23315         * config/rs6000/rs6000.c (rs6000_hard_regno_mode_ok): Simplify
23316         code, only testing for DImode being allowed in non-VSX floating
23317         point registers.
23318         (rs6000_init_hard_regno_mode_ok): Change TARGET_VSX_SMALL_INTEGER
23319         to TARGET_P8_VECTOR test.  Remove redundant VSX test inside of
23320         another VSX test.
23321         (rs6000_option_override_internal): Delete -mvsx-small-integer.
23322         (rs6000_expand_vector_set): Change TARGET_VSX_SMALL_INTEGER to
23323         TARGET_P8_VECTOR test.
23324         (rs6000_secondary_reload_simple_move): Likewise.
23325         (rs6000_preferred_reload_class): Delete TARGET_VSX_SMALL_INTEGER,
23326         since TARGET_P9_VECTOR was already tested.
23327         (rs6000_opt_masks): Remove -mvsx-small-integer.
23328         * config/rs6000/vsx.md (vsx_extract_<mode>): Delete
23329         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23330         used.
23331         (vsx_extract_<mode>_p9): Delete TARGET_VSX_SMALL_INTEGER, since a
23332         test for TARGET_VEXTRACTUB was used, and that uses
23333         TARGET_P9_VECTOR.
23334         (p9 extract splitter): Likewise.
23335         (vsx_extract_<mode>_di_p9): Likewise.
23336         (vsx_extract_<mode>_store_p9): Likewise.
23337         (vsx_extract_si): Delete TARGET_VSX_SMALL_INTEGER, since a test
23338         for TARGET_P9_VECTOR was used.  Delete code that is now dead with
23339         the elimination of TARGET_VSX_SMALL_INTEGER.
23340         (vsx_extract_<mode>_p8): Likewise.
23341         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_fl_<FL_CONV:mode>): Likewise.
23342         (vsx_ext_<VSX_EXTRACT_I:VS_scalar>_ufl_<FL_CONV:mode>): Likewise.
23343         (vsx_set_<mode>_p9): Likewise.
23344         (vsx_set_v4sf_p9): Likewise.
23345         (vsx_set_v4sf_p9_zero): Likewise.
23346         (vsx_insert_extract_v4sf_p9): Likewise.
23347         (vsx_insert_extract_v4sf_p9_2): Likewise.
23348         * config/rs6000/rs6000.md (sign extend splitter): Change
23349         TARGET_VSX_SMALL_INTEGER to TARGET_P8_VECTOR test.
23350         (floatsi<mode>2_lfiwax_mem): Likewise.
23351         (floatunssi<mode>2_lfiwzx_mem): Likewise.
23352         (float<QHI:mode><FP_ISA3:mode>2): Delete TARGET_VSX_SMALL_INTEGER,
23353         since a test for TARGET_P9_VECTOR was used.
23354         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23355         (floatuns<QHI:mode><FP_ISA3:mode>2): Likewise.
23356         (floatuns<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23357         (fix_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23358         TARGET_P8_VECTOR test.
23359         (fix_trunc<mode>si2_stfiwx): Likewise.
23360         (fix_trunc<mode>si2_internal): Likewise.
23361         (fix_trunc<SFDF:mode><QHI:mode>2): Delete
23362         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23363         used.
23364         (fix_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23365         (fixuns_trunc<mode>si2): Change TARGET_VSX_SMALL_INTEGER to
23366         TARGET_P8_VECTOR test.
23367         (fixuns_trunc<mode>si2_stfiwx): Likewise.
23368         (fixuns_trunc<SFDF:mode><QHI:mode>2): Delete
23369         TARGET_VSX_SMALL_INTEGER, since a test for TARGET_P9_VECTOR was
23370         used.
23371         (fixuns_trunc<SFDF:mode><QHI:mode>2_internal): Likewise.
23372         (fctiw<u>z_<mode>_smallint): Delete TARGET_VSX_SMALL_INTEGER,
23373         since a test for TARGET_P9_VECTOR was used.
23374         (splitter for loading small constants): Likewise.
23376 2017-07-26  Andrew Pinski  <apinski@cavium.com>
23378         * config/aarch64/aarch64.c (thunderx_vector_cost): Fix
23379         vec_fp_stmt_cost.
23381 2017-07-26  H.J. Lu  <hongjiu.lu@intel.com>
23383         PR target/81563
23384         * config/i386/i386.c (sp_valid_at): Properly check CFA offset.
23385         (fp_valid_at): Likewise.
23387 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23389         * config/aarch64/aarch64.c (cortexa57_addrcost_table): Remove.
23390         (qdf24xx_addrcost_table): Likewise.
23391         (cortexa57_tunings): Update to use generic_branch_cost.
23392         (cortexa72_tunings): Likewise.
23393         (cortexa73_tunings): Likewise.
23394         (qdf24xx_tunings): Likewise.
23396 2017-07-26  James Greenhalgh  <james.greenhalgh@arm.com>
23398         * config/aarch64/aarch64.c (cortexa57_branch_cost): Remove.
23399         (thunderx2t99_branch_cost): Likewise.
23400         (cortexa35_tunings): Update to use generic_branch_cost.
23401         (cortexa53_tunings): Likewise.
23402         (cortexa57_tunings): Likewise.
23403         (cortexa72_tunings): Likewise.
23404         (cortexa73_tunings): Likewise.
23405         (thunderx2t99_tunings): Likewise.
23407 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23409         * config/sparc/sparc.c (dump_target_flag_bits): Dump MASK_FSMULD.
23410         (sparc_option_override): Honour MASK_FSMULD.
23411         * config/sparc/sparc.h (MASK_FEATURES): Add MASK_FSMULD.
23412         * config/sparc/sparc.md (muldf3_extend): Use TARGET_FSMULD.
23413         * config/sparc/sparc.opt (mfsmuld): New option.
23414         * doc/invoke.texi (mfsmuld): Document option.
23416 2017-07-26  Marek Polacek  <polacek@redhat.com>
23418         PR middle-end/70992
23419         * tree.c (build2_stat): Don't set TREE_CONSTANT on divisions by zero.
23421 2017-07-26  Richard Biener  <rguenther@suse.de>
23423         * gimple-match-head.c (do_valueize): Return OP if valueize
23424         returns NULL_TREE.
23425         (get_def): New helper to get at the def stmt of a SSA name
23426         if valueize allows.
23427         * genmatch.c (dt_node::gen_kids_1): Use get_def instead of
23428         do_valueize to get at the def stmt.
23429         (dt_operand::gen_gimple_expr): Simplify do_valueize calls.
23431 2017-07-26  Wilco Dijkstra  <wdijkstr@arm.com>
23433         PR middle-end/46932
23434         * auto-inc-dec.c (parse_add_or_inc): Block autoinc on sfp.
23436 2017-07-26  Martin Liska  <mliska@suse.cz>
23438         PR sanitize/81186
23439         * function.c (expand_function_start): Make expansion of
23440         nonlocal_goto_save_area after parm_birth_insn.
23442 2017-07-26  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23444         * config/sparc/sparc.c (sparc_option_override): Remove MASK_FPU
23445         from all CPU target flags enable members.
23447 2017-07-26  Richard Biener  <rguenther@suse.de>
23449         * genmatch.c (dt_simplify::gen): Make iterator vars const.
23450         (decision_tree::gen): Make 'type' const.
23451         (write_predicate): Likewise.
23453 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23455         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok):
23456         Eliminate TARGET_UPPER_REGS_{DF,DI,SF} usage.
23457         (rs6000_option_override_internal): Likewise.
23458         (rs6000_expand_vector_set): Likewise.
23459         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Delete.
23460         (TARGET_UPPER_REGS_SF): Likewise.
23461         (TARGET_UPPER_REGS_DI): Likewise.
23462         (TARGET_VEXTRACTUB): Eliminate TARGET_UPPER_REGS_{DF,DI,SF}.
23463         (TARGET_DIRECT_MOVE_64BIT): Likewise.
23464         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Likewise.
23465         (float<QHI:mode><FP_ISA3:mode>2_internal): Likewise.
23466         (Splitters for DI constants in Altivec registers): Likewise.
23467         * config/rs6000/vsx.md (vsx_set_<mode>_p9): Likewise.
23468         (vsx_set_v4sf_p9): Likewise.
23469         (vsx_set_v4sf_p9_zero): Likewise.
23470         (vsx_insert_extract_v4sf_p9): Likewise.
23471         (vsx_insert_extract_v4sf_p9_2): Likewise.
23473 2017-07-25  Carl Love  <cel@us.ibm.com>
23475         * doc/extend.texi: Update the built-in documentation file for the
23476         existing built-in functions
23477         vector signed char vec_cnttz (vector signed char);
23478         vector unsigned char vec_cnttz (vector unsigned char);
23479         vector signed short vec_cnttz (vector signed short);
23480         vector unsigned short vec_cnttz (vector unsigned short);
23481         vector signed int vec_cnttz (vector signed int);
23482         vector unsigned int vec_cnttz (vector unsigned int);
23483         vector signed long long vec_cnttz (vector signed long long);
23484         vector unsigned long long vec_cnttz (vector unsigned long long);
23486 2017-07-25  Andrew Pinski  <apinski@cavium.com>
23488         * tree-ssa-uninit.c (warn_uninitialized_vars): Don't warn about memory
23489         accesses where the use is for the first operand of a BIT_INSERT.
23491 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23493         PR bootstrap/81521
23494         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23495         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23497 2017-07-25  Jim Wilson  <jim.wilson@linaro.org>
23499         * config/i386/gstabs.h: Delete.
23500         * config/i386/openbsd.h, config/i386/t-openbsd: Likewise.
23502 2017-07-25  Uros Bizjak  <ubizjak@gmail.com>
23504         * config/i386/i386.c (ix86_decompose_address): Do not check for
23505         register RTX when looking at index_reg or base_reg.
23506         * config/i386/i386.h (INCOMING_RETURN_ADDR_RTX): Use stack_pointer_rtx.
23508 2017-07-25  Eric Botcazou  <ebotcazou@adacore.com>
23510         * gimple.c (gimple_assign_set_rhs_with_ops): Do not ask gsi_replace
23511         to update EH info here.
23513 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23515         * match.pd ((X * CST1) * CST2): Simplify to X * (CST1 * CST2).
23517 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23519         * match.pd ((X * CST) * Y): Reassociate to (X * Y) * CST.
23521 2017-07-25  Torsten Duwe  <duwe@suse.de>
23523         * common.opt: Introduce -fpatchable-function-entry
23524         command line option, and its variables function_entry_patch_area_size
23525         and function_entry_patch_area_start.
23526         * opts.c (common_handle_option): Add -fpatchable_function_entry_ case,
23527         including a two-value parser.
23528         * target.def (print_patchable_function_entry): New target hook.
23529         * targhooks.h (default_print_patchable_function_entry): New function.
23530         * targhooks.c (default_print_patchable_function_entry): Likewise.
23531         * toplev.c (process_options): Switch off IPA-RA if
23532         patchable function entries are being generated.
23533         * varasm.c (assemble_start_function): Look at the
23534         patchable-function-entry command line switch and current
23535         function attributes and maybe generate NOP instructions by
23536         calling the print_patchable_function_entry hook.
23537         * doc/extend.texi: Document patchable_function_entry attribute.
23538         * doc/invoke.texi: Document -fpatchable_function_entry
23539         command line option.
23540         * doc/tm.texi.in (TARGET_ASM_PRINT_PATCHABLE_FUNCTION_ENTRY):
23541         New target hook.
23542         * doc/tm.texi: Re-generate.
23544 2017-07-25  Jakub Jelinek  <jakub@redhat.com>
23546         PR target/81532
23547         * config/i386/constraints.md (Yd, Ye): Use ALL_SSE_REGS for
23548         TARGET_AVX512DQ rather than TARGET_AVX512BW.
23550 2017-07-25  Tamar Christina  <tamar.christina@arm.com>
23552         * config/arm/parsecpu.awk (all_cores): Remove duplicates.
23554 2017-07-25  Richard Biener  <rguenther@suse.de>
23556         PR tree-optimization/81455
23557         * tree-ssa-loop-unswitch.c (find_loop_guard): Make sure to
23558         not walk in cycles when looking for guards.
23560 2017-07-25  Richard Biener  <rguenther@suse.de>
23562         PR tree-optimization/81529
23563         * tree-vect-stmts.c (process_use): Disregard live induction PHIs
23564         when optimizing backedge uses.
23566 2017-07-25  David Edelsohn  <dje.gcc@gmail.com>
23568         * dwarf2asm.c (dw2_asm_output_nstring): Encode double quote
23569         character for AIX.
23570         * dwarf2out.c (output_macinfo): Copy debug_line_section_label
23571         to dl_section_ref.  On AIX, append an expression to subtract
23572         the size of the section length to dl_section_ref.
23574 2017-07-25  Segher Boessenkool  <segher@kernel.crashing.org>
23576         * configure.ac: If any of the config.* scripts fail, exit 1.
23577         * configure: Regenerate.
23579 2017-07-25  Richard Biener  <rguenther@suse.de>
23581         PR middle-end/81546
23582         * tree-ssa-operands.c (verify_imm_links): Remove cap on number
23583         of immediate uses, be more verbose on errors.
23585 2017-07-25  Richard Biener  <rguenther@suse.de>
23587         PR tree-optimization/81510
23588         * tree-vect-loop.c (vect_is_simple_reduction): When the
23589         reduction stmt is not inside the loop bail out.
23591 2017-07-25  Richard Biener  <rguenther@suse.de>
23593         PR tree-optimization/81303
23594         * tree-vect-loop-manip.c (vect_loop_versioning): Build
23595         profitability check against LOOP_VINFO_NITERSM1.
23597 2017-07-25  Alexander Monakov  <amonakov@ispras.ru>
23599         * domwalk.c (cmp_bb_postorder): Simplify.
23600         (sort_bbs_postorder): New function.  Use it...
23601         (dom_walker::walk): ...here to optimize common cases.
23603 2017-07-25  Martin Liska  <mliska@suse.cz>
23605         PR ipa/81520
23606         * ipa-visibility.c (function_and_variable_visibility): Make the
23607         redirection just on target that supports aliasing.
23608         Fix GNU coding style.
23610 2017-07-25  Sebastian Huber  <sebastian.huber@embedded-brains.de>
23612         PR libgcc/61152
23613         * config/aarch64/rtems.h: Add GCC Runtime Library Exception.
23614         Format changes.
23615         * config/arm/rtems.h: Likewise.
23616         * config/bfin/rtems.h: Likewise.
23617         * config/i386/rtemself.h: Likewise.
23618         * config/lm32/rtems.h: Likewise.
23619         * config/m32c/rtems.h: Likewise.
23620         * config/m68k/rtemself.h: Likewise.
23621         * config/microblaze/rtems.h: Likewise.
23622         * config/mips/rtems.h: Likewise.
23623         * config/moxie/rtems.h: Likewise.
23624         * config/nios2/rtems.h: Likewise.
23625         * config/powerpcspe/rtems.h: Likewise.
23626         * config/rs6000/rtems.h: Likewise.
23627         * config/rtems.h: Likewise.
23628         * config/sh/rtems.h: Likewise.
23629         * config/sh/rtemself.h: Likewise.
23630         * config/sparc/rtemself.h: Likewise.
23632 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23634         PR 81487
23635         * hsa-brig.c (brig_init): Use xasprintf instead of asprintf.
23636         * gimple-pretty-print.c (dump_profile, dump_probability): Same.
23637         * tree-ssa-structalias.c (alias_get_name): Same.
23639 2017-07-25  Bin Cheng  <bin.cheng@arm.com>
23641         PR target/81414
23642         * config/aarch64/cortex-a57-fma-steering.c (analyze): Skip fmul/fmac
23643         instructions if no du chain is found.
23645 2017-07-25  Georg-Johann Lay  <avr@gjlay.de>
23647         * config/avr/avr-log.c (avr_log_vadump) ['T']: Print NULL-TREE.
23649 2017-07-25  Richard Biener  <rguenther@suse.de>
23651         PR middle-end/81505
23652         * fold-const.c (fold_negate_const): TREE_OVERFLOW should be
23653         sticky.
23655 2017-07-24  Michael Meissner  <meissner@linux.vnet.ibm.com>
23657         * config/rs6000/rs6000-cpus.def (ISA_2_6_MASKS_SERVER): Delete
23658         upper-regs options.
23659         (ISA_2_7_MASKS_SERVER): Likewise.
23660         (ISA_3_0_MASKS_IEEE): Likewise.
23661         (OTHER_P8_VECTOR_MASKS): Likewise.
23662         (OTHER_VSX_VECTOR_MASKS): Likewise.
23663         (POWERPC_MASKS): Likewise.
23664         (power7 cpu): Use ISA_2_6_MASKS_SERVER instead of using a
23665         duplicate list of options.
23666         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Remove
23667         explicit -mupper-regs options.
23668         * config/rs6000/rs6000.opt (-mvsx-scalar-memory): Delete
23669         -mupper-regs* options.  Delete -mvsx-scalar-memory, which was an
23670         alias for -mupper-regs-df.
23671         * config/rs6000/rs6000.c (rs6000_setup_reg_addr_masks): Likewise.
23672         (rs6000_init_hard_regno_mode_ok): Likewise.
23673         (rs6000_option_override_internal): Likewise.
23674         (rs6000_opt_masks): Likewise.
23675         * config/rs6000/rs6000.h (TARGET_UPPER_REGS_DF): Define upper regs
23676         options in terms of whether -mvsx or -mpower8-vector was used.
23677         (TARGET_UPPER_REGS_DI): Likewise.
23678         (TARGET_UPPER_REGS_SF): Likewise.
23679         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete the
23680         -mupper-regs-* options.
23682 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23684         * passes.c (emergency_dump_function): Print some empty lines and a
23685         header before the RTL dump.
23687 2017-07-24  Segher Boessenkool  <segher@kernel.crashing.org>
23689         * cfgrtl.c (rtl_dump_bb): Don't call NEXT_INSN on NULL.
23691 2017-07-24  Wilco Dijkstra  <wdijkstr@arm.com>
23693         PR target/79041
23694         * config/aarch64/aarch64.c (aarch64_classify_symbol):
23695         Avoid SYMBOL_SMALL_ABSOLUTE for literals with pc-relative literals.
23697 2017-07-24  Carl Love  <cel@us.ibm.com>
23699         * config/rs6000/rs6000-c.c: Add support for built-in functions
23700         vector float vec_extract_fp32_from_shorth (vector unsigned short);
23701         vector float vec_extract_fp32_from_shortl (vector unsigned short);
23702         * config/rs6000/altivec.h (vec_extract_fp_from_shorth,
23703         vec_extract_fp_from_shortl): Add defines for the two builtins.
23704         * config/rs6000/rs6000-builtin.def (VEXTRACT_FP_FROM_SHORTH,
23705         VEXTRACT_FP_FROM_SHORTL): Add BU_P9V_OVERLOAD_1 and BU_P9V_VSX_1
23706         new builtins.
23707         * config/rs6000/vsx.md vsx_xvcvhpsp): Add define_insn.
23708         (vextract_fp_from_shorth, vextract_fp_from_shortl): Add define_expands.
23709         * doc/extend.texi: Update the built-in documentation file for the
23710         new built-in function.
23712 2017-07-24  Jakub Jelinek  <jakub@redhat.com>
23714         PR bootstrap/81521
23715         * tree.def: Remove TYPE_METHODS documentation, adjust TYPE_FIELDS
23716         documentation.
23717         * doc/generic.texi: Likewise.
23718         * config/i386/winnt-cxx.c (i386_pe_adjust_class_at_definition): Look
23719         for FUNCTION_DECLs in TYPE_FIELDS rather than TYPE_METHODS.
23721 2017-07-24  Jackson Woodruff  <jackson.woodruff@arm.com>
23723         * config/aarch64/aarch64-simd.md (aarch64_mla_elt_merge<mode>): New.
23724         (aarch64_mls_elt_merge<mode>): Likewise.
23726 2017-07-23  Krister Walfridsson  <krister.walfridsson@gmail.com>
23728         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
23729         having __cxa_atexit.
23731 2017-07-23  Michael Collison  <michael.collison@arm.com>
23733         * config/arm/arm.c (arm_option_override): Deprecate
23734         use of -mstructure-size-boundary.
23735         * config/arm/arm.opt: Deprecate -mstructure-size-boundary.
23736         * doc/invoke.texi: Deprecate -mstructure-size-boundary.
23738 2017-07-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
23740         PR target/80695
23741         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
23742         Reduce cost estimate for direct moves.
23744 2017-07-23  Uros Bizjak  <ubizjak@gmail.com>
23746         PR target/80569
23747         * config/i386/i386.c (ix86_option_override_internal): Disable
23748         BMI, BMI2 and TBM instructions for -m16.
23750 2017-07-21  Carl Love  <cel@us.ibm.com>
23752         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
23753         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
23754         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
23755         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
23756         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
23757         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
23758         VMULOSW): New enum "unspec" values.
23759         (altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
23760         altivec_vmulosw): New patterns.
23761         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
23762         VMULOSW): Add definitions.
23764 2017-07-21  Jim Wilson  <jim.wilson@linaro.org>
23766         * config/aarch64/aarch64-cores.def (falkor): Add AARCH64_FL_RDMA.
23767         (qdf24xx): Likewise.
23768         * config/aarch64/aarch64-options-extensions.def (rdma); New.
23769         * config/aarch64/aarch64.h (AARCH64_FL_RDMA): New.
23770         (AARCH64_FL_V8_1): Renumber.
23771         (AARCH64_FL_FOR_ARCH8_1): Add AARCH64_FL_RDMA.
23772         (AARCH64_ISA_RDMA): Use AARCH64_FL_RDMA.
23773         * config/aarch64/arm_neon.h: Use +rdma instead of arch=armv8.1-a.
23774         * doc/invoke.texi (AArch64 Options): Mention +rmda in -march docs.  Add
23775         rdma to feature modifiers list.
23777 2017-07-21  Yury Gribov  <tetra2005@gmail.com>
23779         PR middle-end/56727
23780         * ipa-visibility (function_and_variable_visibility): Convert
23781         recursive PLT call to direct call if appropriate.
23783 2017-07-21  Andrew Pinski  <apinski@cavium.com>
23785         * tree-ssa-sccvn.c (vn_nary_op_eq): Check BIT_INSERT_EXPR's
23786         operand 1 to see if the types precision matches.
23787         * fold-const.c (operand_equal_p): Likewise.
23789 2017-07-21  Richard Biener  <rguenther@suse.de>
23791         PR tree-optimization/81303
23792         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs): Pass
23793         in datarefs vector.  Allow NULL dr0 for no peeling cost estimate.
23794         (vect_peeling_hash_get_lowest_cost): Adjust.
23795         (vect_enhance_data_refs_alignment): Likewise.  Use
23796         vect_get_peeling_costs_all_drs to compute the penalty for no
23797         peeling to match up costs.
23799 2017-07-21  Richard Biener  <rguenther@suse.de>
23801         PR tree-optimization/81500
23802         * tree-vect-loop.c (vect_is_simple_reduction): Properly fail if
23803         we didn't identify a reduction path.
23805 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23806             Cesar Philippidis  <cesar@codesourcery.com>
23808         PR gcov-profile/81442
23809         * config/nvptx/nvptx.c (nvptx_goacc_reduction_init): Add missing edge
23810         probabilities.
23812 2017-07-21  Tom de Vries  <tom@codesourcery.com>
23814         PR lto/81430
23815         * config/nvptx/nvptx.c (nvptx_override_options_after_change): New
23816         function.
23817         (TARGET_OVERRIDE_OPTIONS_AFTER_CHANGE): Define to
23818         nvptx_override_options_after_change.
23820 2017-07-21  Ulrich Drepper  <drepper@redhat.com>
23822         * dwarf2out.c (output_file_names): Avoid double testing for
23823         dwarf_version >= 5.
23825 2017-07-21  Georg-Johann Lay  <avr@gjlay.de>
23827         * doc/invoke.texi (AVR Built-in Functions): Re-layout section.
23829 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23831         * cfgcleanup.c (flow_find_cross_jump): Do not crossjump across
23832         hot/cold regions.
23833         (try_crossjump_to_edge): Do not punt on partitioned functions.
23835 2016-07-21  Jan Hubicka  <hubicka@ucw.cz>
23837         * bb-reorder.c (find_rarely_executed_basic_blocks_and_crossing_edges):
23838         Put all BBs reachable only via paths crossing cold region to cold
23839         region.
23840         * cfgrtl.c (find_bbs_reachable_by_hot_paths): New function.
23842 2016-07-21  Richard Biener  <rguenther@suse.de>
23844         PR tree-optimization/81303
23845         * tree-vect-loop.c (vect_estimate_min_profitable_iters): Take
23846         into account prologue and epilogue iterations when raising
23847         min_profitable_iters to sth at least covering one vector iteration.
23849 2017-07-21  Tamar Christina  <tamar.christina@arm.com>
23851         * config/arm/arm.c (arm_test_cpu_arch_dat):
23852         Check for overlap.
23854 2017-07-20  Nathan Sidwell  <nathan@acm.org>
23856         Remove TYPE_METHODS.
23857         * tree.h (TYPE_METHODS): Delete.
23858         * dwarf2out.c (gen_member_die): Member fns are on TYPE_FIELDS.
23859         * dbxout.c (dbxout_type_fields): Ignore FUNCTION_DECLs.
23860         (dbxout_type_methods): Scan TYPE_FIELDS.
23861         (dbxout_type): Don't check TYPE_METHODS here.
23862         * function.c (use_register_for_decl): Always ignore register for
23863         class types when not optimizing.
23864         * ipa-devirt.c (odr_types_equivalent_p): Delete TYPE_METHODS scan.
23865         * tree.c (free_lang_data_in_type): Stitch out member functions and
23866         templates from TYPE_FIELDS.
23867         (build_distinct_type_copy, verify_type_variant,
23868         verify_type): Member fns are on TYPE_FIELDS.
23869         * tree-dump.c (dequeue_and_dump): No TYPE_METHODS.
23870         * tree-pretty-print.c (dump_generic_node): Likewise.
23872 2017-07-20  Jakub Jelinek  <jakub@redhat.com>
23874         PR target/80846
23875         * config/i386/i386.c (ix86_expand_vector_init_general): Handle
23876         V2TImode and V4TImode.
23877         (ix86_expand_vector_extract): Likewise.
23878         * config/i386/sse.md (VMOVE): Enable V4TImode even for just
23879         TARGET_AVX512F, instead of only for TARGET_AVX512BW.
23880         (ssescalarmode): Handle V4TImode and V2TImode.
23881         (VEC_EXTRACT_MODE): Add V4TImode and V2TImode.
23882         (*vec_extractv2ti, *vec_extractv4ti): New insns.
23883         (VEXTRACTI128_MODE): New mode iterator.
23884         (splitter for *vec_extractv?ti first element): New.
23885         (VEC_INIT_MODE): New mode iterator.
23886         (vec_init<mode>): Consolidate 3 expanders into one using
23887         VEC_INIT_MODE mode iterator.
23889 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23891         * lra-assigns.c (pseudo_compare_func): Fix comparison step based on
23892         non_spilled_static_chain_regno_p.
23894 2017-07-20  Alexander Monakov  <amonakov@ispras.ru>
23896         * gimple-ssa-store-merging.c (sort_by_bitpos): Return 0 on equal bitpos.
23898 2017-07-20  Jan Hubicka  <hubicka@ucw.cz>
23900         * bb-reorder.c (connect_traces): Allow copying of blocks within
23901         single partition.
23903 2017-07-20  Richard Biener  <rguenther@suse.de>
23905         * gimple.h (gimple_phi_result): Add gphi * overload.
23906         (gimple_phi_result_ptr): Likewise.
23907         (gimple_phi_arg): Likewise.  Adjust index assert to only
23908         allow actual argument accesses rather than all slots available
23909         by capacity.
23910         (gimple_phi_arg_def): Add gphi * overload.
23911         * tree-phinodes.c (make_phi_node): Initialize only actual
23912         arguments.
23913         (resize_phi_node): Clear memory not covered by old node,
23914         do not initialize excess argument slots.
23915         (reserve_phi_args_for_new_edge): Initialize new argument slot
23916         completely.
23918 2017-07-20  Bin Cheng  <bin.cheng@arm.com>
23920         PR tree-optimization/81388
23921         Revert r238585:
23922         2016-07-21  Bin Cheng  <bin.cheng@arm.com>
23924         * tree-ssa-loop-niter.c (number_of_iterations_lt_to_ne): Clean up
23925         by removing computation of may_be_zero.
23927 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
23928             Tom de Vries  <tom@codesourcery.com>
23930         PR middle-end/81030
23931         * cfgbuild.c (find_many_sub_basic_blocks): Update REG_BR_PROB note
23932         when gimple level profile disagrees with what RTL expander did.
23934 2017-07-20  Richard Biener  <rguenther@suse.de>
23936         PR tree-optimization/61171
23937         * tree-vectorizer.h (slp_instance): Add reduc_phis member.
23938         (vect_analyze_stmt): Add slp instance parameter.
23939         (vectorizable_reduction): Likewise.
23940         * tree-vect-loop.c (vect_analyze_loop_operations): Adjust.
23941         (vect_is_simple_reduction): Deal with chains not detected
23942         as SLP reduction chain, specifically not properly associated
23943         chains containing a mix of plus/minus.
23944         (get_reduction_op): Remove.
23945         (get_initial_defs_for_reduction): Simplify, pass in whether
23946         this is a reduction chain, pass in the SLP node for the PHIs.
23947         (vect_create_epilog_for_reduction): Get the SLP instance as
23948         arg and adjust.
23949         (vectorizable_reduction): Get the SLP instance as arg.
23950         During analysis remember the SLP node with the PHIs in the
23951         instance.  Simplify getting at the vectorized reduction PHIs.
23952         * tree-vect-slp.c (vect_slp_analyze_node_operations): Pass
23953         through SLP instance.
23954         (vect_slp_analyze_operations): Likewise.
23955         * tree-vect-stms.c (vect_analyze_stmt): Likewise.
23956         (vect_transform_stmt): Likewise.
23958 2017-07-20  Tom de Vries  <tom@codesourcery.com>
23960         PR tree-optimization/81489
23961         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Move
23962         read of phi arg location to before loop that modifies phi.
23964 2017-07-20  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
23966         * match.pd (((m1 >/</>=/<= m2) * d -> (m1 >/</>=/<= m2) ? d : 0):
23967         New pattern.
23969 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23971         PR middle-end/81331
23972         * except.c (execute): Fix ordering issue.
23974 2018-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
23976         PR rtl-optimization/81423
23977         * combine.c (make_compound_operation_int): Don't try to optimize
23978         the AND of a SUBREG of an LSHIFTRT if that SUBREG is paradoxical.
23980 2017-07-19  Segher Boessenkool  <segher@kernel.crashing.org>
23982         PR rtl-optimization/81423
23983         * simplify-rtx.c (simplify_truncation): Handle truncating an IOR
23984         with a constant that is -1 in the truncated to mode.
23986 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23988         * predict.c (propagate_unlikely_bbs_forward): Break out from ...
23989         (determine_unlikely_bbs): ... here.
23990         * predict.h (propagate_unlikely_bbs_forward): Declare.
23991         * cfgexpand.c (pass_expand::execute): Use it.
23992         * bb-reorder.c (sanitize_hot_paths): Do not consider known to be
23993         unlikely edges.
23994         (find_rarely_executed_basic_blocks_and_crossing_edges): Use
23995         propagate_unlikely_bbs_forward.
23997 2017-07-19  Jan Hubicka  <hubicka@ucw.cz>
23999         PR middle-end/81331
24000         * except.c (maybe_add_nop_after_section_switch): New function.
24001         (execute): Use it.
24003 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24005         * gimple.h (gimple_phi_set_arg): Make assert more strict.
24007 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24009         * gimple.h (gimple_phi_arg): Make assert more strict.
24011 2017-07-19  Steven Munroe  <munroesj@gcc.gnu.org>
24013         * config.gcc (powerpc*-*-*): Add mmintrin.h.
24014         * config/rs6000/mmintrin.h: New file.
24015         * config/rs6000/x86intrin.h [__ALTIVEC__]: Include mmintrin.h.
24017 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24019         PR tree-optimization/81346
24020         * match.pd: Optimize (X - 1U) <= INT_MAX-1U into (int) X > 0.
24022 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24024         * config/nvptx/nvptx.md (VECIM): Add V2DI.
24026 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24028         * config/nvptx/nvptx-modes.def: Add V2DImode.
24029         * config/nvptx/nvptx-protos.h (nvptx_data_alignment): Declare.
24030         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2DImode.
24031         (nvptx_output_mov_insn): Handle lack of mov.b128.
24032         (nvptx_print_operand): Handle 'H' and 'L' codes.
24033         (nvptx_vector_mode_supported): Allow V2DImode.
24034         (nvptx_preferred_simd_mode): New function.
24035         (nvptx_data_alignment): New function.
24036         (TARGET_VECTORIZE_PREFERRED_SIMD_MODE): Redefine to
24037         nvptx_preferred_simd_mode.
24038         * config/nvptx/nvptx.h (STACK_BOUNDARY, BIGGEST_ALIGNMENT): Change from
24039         64 to 128 bits.
24040         (DATA_ALIGNMENT): Define.  Set to nvptx_data_alignment.
24042 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24044         * config/nvptx/nvptx-modes.def: New file.  Add V2SImode.
24045         * config/nvptx/nvptx.c (nvptx_ptx_type_from_mode): Handle V2SImode.
24046         (nvptx_vector_mode_supported): New function.  Allow V2SImode.
24047         (TARGET_VECTOR_MODE_SUPPORTED_P): Redefine to nvptx_vector_mode_supported.
24048         * config/nvptx/nvptx.md (VECIM): New mode iterator. Add V2SI.
24049         (mov<VECIM>_insn): New define_insn.
24050         (define_expand "mov<VECIM>): New define_expand.
24052 2017-07-19  Tom de Vries  <tom@codesourcery.com>
24054         * config/nvptx/nvptx.c (nvptx_print_operand): Handle v2 vector mode.
24056 2017-07-19  Jakub Jelinek  <jakub@redhat.com>
24058         PR tree-optimization/81346
24059         * fold-const.h (fold_div_compare, range_check_type): Declare.
24060         * fold-const.c (range_check_type): New function.
24061         (build_range_check): Use range_check_type.
24062         (fold_div_compare): No longer static, rewritten into
24063         a match.pd helper function.
24064         (fold_comparison): Don't call fold_div_compare here.
24065         * match.pd (X / C1 op C2): New optimization using fold_div_compare
24066         as helper function.
24068 2017-07-19  Nathan Sidwell  <nathan@acm.org>
24070         * tree.h (TYPE_MINVAL, TYPE_MAXVAL): Rename to ...
24071         (TYPE_MIN_VALUE_RAW, TYPE_MAX_VALUE_RAW): ... these.
24072         * tree.c (find_decls_types_r, verify_type): Use
24073         TYPE_{MIN,MAX}_VALUE_RAW.
24074         * lto-streamer-out.c (DFS::DFS_write_tree_body): Likewise.
24075         (hash_tree): Likewise.
24076         * tree-streamer-in.c (lto_input_ts_type_non_common_tree_pointers):
24077         Likewise.
24078         * tree-streamer-out.c (write_ts_type_non_common_tree_pointers):
24079         Likewise.
24081 2017-07-18  Tom de Vries  <tom@codesourcery.com>
24083         PR middle-end/81464
24084         * omp-expand.c (expand_omp_for_static_chunk): Handle
24085         equal-argument loop exit phi.
24087 2017-07-18  Uros Bizjak  <ubizjak@gmail.com>
24089         PR target/81471
24090         * config/i386/i386.md (rorx_immediate_operand): New mode attribute.
24091         (*bmi2_rorx<mode>3_1): Use rorx_immediate_operand as
24092         operand 2 predicate.
24093         (*bmi2_rorxsi3_1_zext): Use const_0_to_31_operand as
24094         operand 2 predicate.
24095         (ror,rol -> rorx splitters): Use const_int_operand as
24096         operand 2 predicate.
24098 2017-06-18  Richard Biener  <rguenther@suse.de>
24100         PR tree-optimization/81410
24101         * tree-vect-stmts.c (vectorizable_load): Properly adjust for
24102         the gap in the ! slp_perm SLP case after each group.
24104 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24106         PR middle-end/81463
24107         * cfgloopmanip.c (scale_loop_profile): Watch out for zero frequency
24108         again.
24110 2017-07-18  Jan Hubicka  <hubicka@ucw.cz>
24112         PR middle-end/81462
24113         * predict.c (set_even_probabilities): Cleanup; do not affect
24114         probabilities that are already known.
24115         (combine_predictions_for_bb): Call even when count is set.
24117 2017-07-18  Nathan Sidwell  <nathan@acm.org>
24119         * tree-parloops.c (try_transform_to_exit_first_loop_alt): Use
24120         TYPE_MAX_VALUE.
24122 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24124         PR target/81408
24125         * tree-ssa-loop-niter.c (number_of_iterations_exit): Dump missed
24126         optimization for loop niter analysis.
24128 2017-07-18  Georg-Johann Lay  <avr@gjlay.de>
24130         PR target/81473
24131         * config/avr/avr.c (avr_optimize_casesi): Don't use
24132         INT8_MIN, INT8_MAX, UINT8_MAX, INT16_MIN, INT16_MAX, UINT16_MAX.
24134 2017-07-18  Robin Dapp  <rdapp@linux.vnet.ibm.com>
24136         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Remove
24137         body_cost_vec from _vect_peel_extended_info.
24138         (vect_peeling_hash_get_lowest_cost): Do not set body_cost_vec.
24139         (vect_peeling_hash_choose_best_peeling): Remove body_cost_vec and
24140         npeel.
24142 2017-07-18  Bin Cheng  <bin.cheng@arm.com>
24144         * config/arm/arm.c (emit_unlikely_jump): Remove unused var.
24146 2017-07-18  Richard Biener  <rguenther@suse.de>
24148         PR tree-optimization/80620
24149         PR tree-optimization/81403
24150         * tree-ssa-pre.c (phi_translate_1): Clear range and points-to
24151         info when re-using a VN table entry.
24153 2017-07-18  Richard Biener  <rguenther@suse.de>
24155         PR tree-optimization/81418
24156         * tree-vect-loop.c (vectorizable_reduction): Properly compute
24157         vectype_in.  Verify that with lane-reducing reduction operations
24158         we have a single def-use cycle.
24160 2017-07-17  Carl Love  <cel@us.ibm.com>
24162         Revert commit r249424   2017-06-20  Carl Love  <cel@us.ibm.com>
24164         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24165         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
24166         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
24167         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
24168         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
24169         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
24170         VMULOSW): New enum "unspec" values.
24171         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
24172         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
24173         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
24174         altivec_vmulosw): New patterns.
24175         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
24176         VMULOSW): Add definitions.
24178 2017-07-17  Uros Bizjak  <ubizjak@gmail.com>
24180         * config/alpha/alpha.c: Include predict.h.
24182 2017-07-17  Yury Gribov  <tetra2005@gmail.com>
24184         * tree-vrp.c (compare_assert_loc): Fix comparison function
24185         to return predictable results.
24187 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24189         * config/arc/arc.md (adddi3): Remove support for mexpand-adddi
24190         option.
24191         (subdi3): Likewise.
24192         * config/arc/arc.opt (mexpand-adddi): Deprecate it.
24193         * doc/invoke.texi (mexpand-adddi): Update text.
24195 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24197         * config/arc/arc.md (clzsi2): Expand to an arc_clzsi2 instruction
24198         that also clobbers the CC register. The old expand code is moved
24199         to ...
24200         (*arc_clzsi2): ... here.
24201         (ctzsi2): Expand to an arc_ctzsi2 instruction that also clobbers
24202         the CC register. The old expand code is moved to ...
24203         (arc_ctzsi2): ... here.
24205 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24207         * config/arc/arc.opt (mindexed-loads): Use initial value
24208         TARGET_INDEXED_LOADS_DEFAULT.
24209         (mauto-modify-reg): Use initial value
24210         TARGET_AUTO_MODIFY_REG_DEFAULT.
24211         * config/arc/elf.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24212         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24213         * config/arc/linux.h (TARGET_INDEXED_LOADS_DEFAULT): Define.
24214         (TARGET_AUTO_MODIFY_REG_DEFAULT): Likewise.
24216 2017-07-17  Martin Liska  <mliska@suse.cz>
24218         PR sanitizer/81302
24219         * opts.c (finish_options): Do not allow -fgnu-tm
24220         w/ -fsanitize={kernel-,}address.  Say sorry.
24222 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24224         PR target/81369
24225         * tree-loop-distribution.c (classify_partition): Only assert on
24226         numer of iterations.
24227         (merge_dep_scc_partitions): Delete prameter.  Update function call.
24228         (distribute_loop): Remove code handling loop with unknown niters.
24229         (pass_loop_distribution::execute): Skip loop with unknown niters.
24231 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24233         PR target/81369
24234         * tree-loop-distribution.c (merge_dep_scc_partitions): Sink call to
24235         function sort_partitions_by_post_order.
24237 2017-07-17  Bin Cheng  <bin.cheng@arm.com>
24239         PR tree-optimization/81374
24240         * tree-loop-distribution.c (pass_loop_distribution::execute): Record
24241         the max index of basic blocks, rather than number of basic blocks.
24243 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24245         * config/arc/arc-protos.h (arc_legitimate_pc_offset_p): Remove
24246         proto.
24247         (arc_legitimate_pic_operand_p): Likewise.
24248         * config/arc/arc.c (arc_legitimate_pic_operand_p): Remove
24249         function.
24250         (arc_needs_pcl_p): Likewise.
24251         (arc_legitimate_pc_offset_p): Likewise.
24252         (arc_legitimate_pic_addr_p): Remove LABEL_REF case, as this
24253         function is also used in constrains.md.
24254         (arc_legitimate_constant_p): Use arc_legitimate_pic_addr_p to
24255         validate pic constants. Handle CONST_INT, CONST_DOUBLE, MINUS and
24256         PLUS.  Only return true/false in known cases, otherwise assert.
24257         (arc_legitimate_address_p): Remove arc_legitimate_pic_addr_p as it
24258         is already called in arc_legitimate_constant_p.
24259         * config/arc/arc.h (CONSTANT_ADDRESS_P): Consider also LABEL for
24260         pic addresses.
24261         (LEGITIMATE_PIC_OPERAND_P): Use
24262         arc_raw_symbolic_reference_mentioned_p function.
24263         * config/arc/constraints.md (Cpc): Use arc_legitimate_pic_addr_p
24264         function.
24265         (Cal): Likewise.
24266         (C32): Likewise.
24268 2017-07-17  Claudiu Zissulescu  <claziss@synopsys.com>
24269         Andrew Burgess  <andrew.burgess@embecosm.com>
24271         * config/arc/arc-protos.h (arc_compute_function_type): Change prototype.
24272         (arc_return_address_register): New function.
24273         * config/arc/arc.c (arc_handle_fndecl_attribute): New function.
24274         (arc_handle_fndecl_attribute): Add naked attribute.
24275         (TARGET_ALLOCATE_STACK_SLOTS_FOR_ARGS): Define.
24276         (TARGET_WARN_FUNC_RETURN): Likewise.
24277         (arc_allocate_stack_slots_for_args): New function.
24278         (arc_warn_func_return): Likewise.
24279         (machine_function): Change type fn_type.
24280         (arc_compute_function_type): Consider new naked function type,
24281         change function return type.
24282         (arc_must_save_register): Adapt to handle new
24283         arc_compute_function_type's return type.
24284         (arc_expand_prologue): Likewise.
24285         (arc_expand_epilogue): Likewise.
24286         (arc_return_address_regs): Delete.
24287         (arc_return_address_register): New function.
24288         (arc_epilogue_uses): Use above function.
24289         * config/arc/arc.h (arc_return_address_regs): Delete prototype.
24290         (arc_function_type): Change encoding, add naked type.
24291         (ARC_INTERRUPT_P): Change to handle the new encoding.
24292         (ARC_FAST_INTERRUPT_P): Likewise.
24293         (ARC_NORMAL_P): Define.
24294         (ARC_NAKED_P): Likewise.
24295         (arc_compute_function_type): Delete prototype.
24296         * config/arc/arc.md (in_ret_delay_slot): Use
24297         arc_return_address_register function.
24298         (simple_return): Likewise.
24299         (p_return_i): Likewise.
24301 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24303         PR tree-optimization/81428
24304         * match.pd (X / X -> one): Don't optimize _Fract divisions, as 1
24305         can't be built for those types.
24307 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24309         Remove stuff dead since r239246.
24311         * config/avr/avr-arch.h (avr_inform_devices): Remove dead proto.
24312         * config/avr/avr-devices.c (mcu_name, comparator, avr_mcus_str)
24313         (avr_inform_devices): Remove dead stuff.
24315 2017-07-17  Tamar Christina  <tamar.christina@arm.com>
24317         * config/arm/arm_neon.h: Fix softp typo.
24319 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24321         PR tree-optimization/81365
24322         * tree-ssa-phiprop.c (propagate_with_phi): When considering hoisting
24323         aggregate moves onto bb predecessor edges, make sure there are no
24324         loads that could alias the lhs in between the start of bb and the
24325         loads from *phi.
24327 2017-07-17  Georg-Johann Lay  <avr@gjlay.de>
24329         PR 80929
24330         * config/avr/avr.c (avr_mul_highpart_cost): New static function.
24331         (avr_rtx_costs_1) [TRUNCATE]: Use it to compute mul_highpart cost.
24332         [LSHIFTRT, outer_code = TRUNCATE]: Same.
24334 2017-07-17  Jakub Jelinek  <jakub@redhat.com>
24336         PR tree-optimization/81396
24337         * tree-ssa-math-opts.c (struct symbolic_number): Add n_ops field.
24338         (init_symbolic_number): Initialize it to 1.
24339         (perform_symbolic_merge): Add n_ops from both operands into the new
24340         n_ops.
24341         (find_bswap_or_nop): Don't consider n->n == cmpnop computations
24342         without base_addr as useless if they need more than one operation.
24343         (bswap_replace): Handle !bswap case for NULL base_addr.
24345 2017-07-17  Tom de Vries  <tom@codesourcery.com>
24347         PR target/81069
24348         * config/nvptx/nvptx.c (nvptx_single): Insert diverging branch as late
24349         as possible.
24351 2017-07-17  Sebastian Huber  <sebastian.huber@embedded-brains.de>
24353         * config/sparc/rtemself.h (TARGET_OS_CPP_BUILTINS): Add
24354         conditional builtin define __FIX_LEON3FT_B2BST.
24356 2017-07-17  Daniel Cederman  <cederman@gaisler.com>
24358         * config/sparc/t-rtems: Add mfix-gr712rc multilibs. Replace
24359         MULTILIB_EXCEPTIONS with MULTILIB_REQUIRED. Match -mfix-gr712rc
24360         with -mfix-ut700.
24362 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24364         PR rtl-optimization/81424
24365         * optabs.c (prepare_cmp_insn): Use copy_to_reg instead of force_reg
24366         to remove potential trapping from operands if -fnon-call-exceptions.
24368 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24370         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Use
24371         profile_proability for scalling.
24372         * scale_profile_for_vect_loop.c (scale_profile_for_vect_loop): Likewise.
24374 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24376         * cgraph.c (cgraph_edge::redirect_call_stmt_to_caller): Cleanup.
24378 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24380         * cfgloopmanip.c (scale_loop_profile): Avoid use of REG_BR_PROB_BASE
24381         fixpoint arithmetics.
24383 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24385         * tree-ssa-loop-unswitch.c (hoist_guard): Avoid use of REG_BR_PROB_BASE
24386         fixpoint arithmetics.
24388 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24390         * asan.c (create_cond_insert_point): Avoid use of REG_BR_PROB_BASE
24391         fixpoint arithmetics.
24393 2017-07-16  Jan Hubicka  <hubicka@ucw.cz>
24395         * profile-count.h (profile_probability::from_reg_br_prob_note,
24396         profile_probability::to_reg_br_prob_note): New functions.
24397         * doc/rtl.texi (REG_BR_PROB_NOTE): Update documentation.
24398         * reg-notes.h (REG_BR_PROB, REG_BR_PRED): Update docs.
24399         * predict.c (probability_reliable_p): Update.
24400         (edge_probability_reliable_p): Update.
24401         (br_prob_note_reliable_p): Update.
24402         (invert_br_probabilities): Update.
24403         (add_reg_br_prob_note): New function.
24404         (combine_predictions_for_insn): Update.
24405         * asan.c (asan_clear_shadow): Update.
24406         * cfgbuild.c (compute_outgoing_frequencies): Update.
24407         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
24408         (update_br_prob_note): Update.
24409         (rtl_verify_edges): Update.
24410         (purge_dead_edges): Update.
24411         (fixup_reorder_chain): Update.
24412         * emit-rtl.c (try_split): Update.
24413         * ifcvt.c (cond_exec_process_insns): Update.
24414         (cond_exec_process_if_block): Update.
24415         (dead_or_predicable): Update.
24416         * internal-fn.c (expand_addsub_overflow): Update.
24417         (expand_neg_overflow): Update.
24418         (expand_mul_overflow): Update.
24419         * loop-doloop.c (doloop_modify): Update.
24420         * loop-unroll.c (compare_and_jump_seq): Update.
24421         * optabs.c (emit_cmp_and_jump_insn_1): Update.
24422         * predict.h: Update.
24423         * reorg.c (mostly_true_jump): Update.
24424         * rtl.h: Update.
24425         * config/aarch64/aarch64.c (aarch64_emit_unlikely_jump): Update.
24426         * config/alpha/alpha.c (emit_unlikely_jump): Update.
24427         * config/arc/arc.c: (emit_unlikely_jump): Update.
24428         * config/arm/arm.c: (emit_unlikely_jump): Update.
24429         * config/bfin/bfin.c (cbranch_predicted_taken_p): Update.
24430         * config/frv/frv.c (frv_print_operand_jump_hint): Update.
24431         * config/i386/i386.c (ix86_expand_split_stack_prologue): Update.
24432         (ix86_print_operand): Update.
24433         (ix86_split_fp_branch): Update.
24434         (predict_jump): Update.
24435         * config/ia64/ia64.c (ia64_print_operand): Update.
24436         * config/mmix/mmix.c (mmix_print_operand): Update.
24437         * config/powerpcspe/powerpcspe.c (output_cbranch): Update.
24438         (rs6000_expand_split_stack_prologue): Update.
24439         * config/rs6000/rs6000.c: Update.
24440         * config/s390/s390.c (s390_expand_vec_strlen): Update.
24441         (s390_expand_vec_movstr): Update.
24442         (s390_expand_cs_tdsi): Update.
24443         (s390_expand_split_stack_prologue): Update.
24444         * config/sh/sh.c (sh_print_operand): Update.
24445         (expand_cbranchsi4): Update.
24446         (expand_cbranchdi4): Update.
24447         * config/sparc/sparc.c (output_v9branch): Update.
24448         * config/spu/spu.c (get_branch_target): Update.
24449         (ea_load_store_inline): Update.
24450         * config/tilegx/tilegx.c (cbranch_predicted_p): Update.
24451         * config/tilepro/tilepro.c: Update.
24453 2017-07-16  Eric Botcazou  <ebotcazou@adacore.com>
24455         * gimplify.c (mostly_copy_tree_r): Revert latest change.
24456         (gimplify_save_expr): Likewise.
24458 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24460         * ipa-visibility.c (function_and_variable_visibility): Fix pasto.
24462 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24464         * ipa-fnsummary.c (pass_data_ipa_fn_summary): Use
24465         TV_IPA_FNSUMMARY.
24466         * timevar.def (TV_IPA_FNSUMMARY): Define.
24468 2017-07-16  Daniel Cederman  <cederman@gaisler.com>
24470         * config/sparc/sparc.md (divdf3_fix): Add NOP to prevent back
24471         to back store errata sensitive sequence from being generated.
24472         (sqrtdf2_fix): Likewise.
24474 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24476         * tree-ssa-threadupdate.c (compute_path_counts,
24477         update_joiner_offpath_counts): Use profile_probability.
24479 2017-07-15  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24481         Revert:
24482         2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24484         * config/arm/arm-c.c (arm_cpu_builtins): Define
24485         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24487 2017-07-14  Kelvin Nilsen  <kelvin@gcc.gnu.org>
24489         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
24490         array entries to represent __ieee128 versions of the
24491         scalar_test_data_class, scalar_test_neg, scalar_extract_exp,
24492         scalar_extract_sig, and scalar_insert_exp built-in functions.
24493         (altivec_resolve_overloaded_builtin): Add special case handling
24494         for the __builtin_scalar_insert_exp function, as represented by
24495         the P9V_BUILTIN_VEC_VSIEDP constant.
24496         * config/rs6000/rs6000-builtin.def (VSEEQP): Add scalar extract
24497         exponent support for __ieee128 argument.
24498         (VSESQP): Add scalar extract signature support for __ieee128
24499         argument.
24500         (VSTDCNQP): Add scalar test negative support for __ieee128
24501         argument.
24502         (VSIEQP): Add scalar insert exponent support for __int128 argument
24503         with __ieee128 result.
24504         (VSIEQPF): Add scalar insert exponent support for __ieee128
24505         argument with __ieee128 result.
24506         (VSTDCQP): Add scalar test data class support for __ieee128
24507         argument.
24508         (VSTDCNQP): Add overload support for scalar test negative with
24509         __ieee128 argument.
24510         (VSTDCQP): Add overload support for scalar test data class
24511         __ieee128 argument.
24512         * config/rs6000/vsx.md (UNSPEC_VSX_SXSIG) Replace
24513         UNSPEC_VSX_SXSIGDP.
24514         (UNSPEC_VSX_SIEXPQP): New constant.
24515         (xsxexpqp): New insn for VSX scalar extract exponent quad
24516         precision.
24517         (xsxsigqp): New insn for VSX scalar extract significand quad
24518         precision.
24519         (xsiexpqpf): New insn for VSX scalar insert exponent quad
24520         precision with floating point argument.
24521         (xststdcqp): New expand for VSX scalar test data class quad
24522         precision.
24523         (xststdcnegqp): New expand for VSX scalar test negative quad
24524         precision.
24525         (xststdcqp): New insn to match expansions for VSX scalar test data
24526         class quad precision and VSX scalar test negative quad precision.
24527         * config/rs6000/rs6000.c (rs6000_expand_binop_builtin): Add
24528         special case operand checking to enforce that second operand of
24529         VSX scalar test data class with quad precision argument is a 7-bit
24530         unsigned literal.
24531         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
24532         prototypes and descriptions of __ieee128 versions of
24533         scalar_extract_exp, scalar_extract_sig, scalar_insert_exp,
24534         scalar_test_data_class, and scalar_test_neg built-in functions.
24536 2016-07-14  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
24538         PR tree-optimization/81162
24539         * gimple-ssa-strength-reduction.c (replace_mult_candidate): Don't
24540         replace a negate with an add.
24542 2017-07-14  James Greenhalgh  <james.greenhalgh@arm.com>
24544         * doc/invoke.texi (arm/-mcpu): Document +crypto.
24546 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24548         * config/arm/arm-c.c (arm_cpu_builtins): Define
24549         __ARM_FEATURE_NUMERIC_MAXMIN solely based on TARGET_VFP5.
24551 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24553         * config/arm/arm-cpus.in (cortex-r52): Add new entry.
24554         (armv8-r): Set ARM Cortex-R52 as default CPU.
24555         * config/arm/arm-tables.opt: Regenerate.
24556         * config/arm/arm-tune.md: Regenerate.
24557         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
24558         Cortex-R52.
24559         * doc/invoke.texi: Mention -mtune=cortex-r52 and availability of fp.dp
24560         extension for -mcpu=cortex-r52.
24562 2017-07-14  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24564         * config/arm/arm-isa.h (isa_bit_FP_ARMv8): Delete enumerator.
24565         (ISA_FP_ARMv8): Define as ISA_FPv5 and ISA_FP_D32.
24566         * config/arm/arm-cpus.in (armv8-r): Define fp.sp as enabling FPv5.
24567         (fp-armv8): Define it as FP_ARMv8 only.
24568         config/arm/arm.h (TARGET_FPU_ARMV8): Delete.
24569         (TARGET_VFP_FP16INST): Define using TARGET_VFP5 rather than
24570         TARGET_FPU_ARMV8.
24571         config/arm/arm.c (arm_rtx_costs_internal): Replace checks against
24572         TARGET_FPU_ARMV8 by checks against TARGET_VFP5.
24573         * config/arm/arm-builtins.c (arm_builtin_vectorized_function): Define
24574         first ARM_CHECK_BUILTIN_MODE definition using TARGET_VFP5 rather
24575         than TARGET_FPU_ARMV8.
24576         * config/arm/arm-c.c (arm_cpu_builtins): Likewise for
24577         __ARM_FEATURE_NUMERIC_MAXMIN macro definition.
24578         * config/arm/arm.md (cmov<mode>): Condition on TARGET_VFP5 rather than
24579         TARGET_FPU_ARMV8.
24580         * config/arm/neon.md (neon_vrint): Likewise.
24581         (neon_vcvt): Likewise.
24582         (neon_<fmaxmin_op><mode>): Likewise.
24583         (<fmaxmin><mode>3): Likewise.
24584         * config/arm/vfp.md (l<vrint_pattern><su_optab><mode>si2): Likewise.
24585         * config/arm/predicates.md (arm_cond_move_operator): Check against
24586         TARGET_VFP5 rather than TARGET_FPU_ARMV8 and fix spacing.
24588 2017-07-14  Jackson Woodruff  <jackson.woodruff@arm.com>
24590         * config/aarch64/aarch64.c (aarch64_print_operand): Move comments
24591         to top of function.
24593 2017-07-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
24595         * gimple-ssa-store-merging.c (clear_bit_region): Replace reference to
24596         loop in comment with memset.
24598 2017-07-14  Martin Liska  <mliska@suse.cz>
24600         * cfgexpand.c (expand_gimple_basic_block): Remove dead comment.
24601         * dwarf2out.c (is_java): Remove the function.
24602         (output_pubname): Remove usage of the function.
24603         (lower_bound_default): Remove usage of DW_LANG_Java.
24604         (gen_compile_unit_die): Likewise.
24605         * gcc.c: Remove compiler defaults for .java and .zip files.
24606         * gimple-expr.c (remove_suffix): Change as there's no longer
24607         extension than 4-letter one.
24608         * gimplify.c (mostly_copy_tree_r): Remove Java-special part.
24609         (gimplify_save_expr): Likewise.
24610         * ipa-utils.h (polymorphic_type_binfo_p): Remove the comment
24611         as it's possible even for other languages than Java.
24612         * langhooks.h (struct lang_hooks): Remove Java from a comment.
24613         * lto-opts.c (lto_write_options): Remove reference to Java.
24614         * opts.c (strip_off_ending): Update file extension handling.
24615         * tree-cfg.c (verify_gimple_call): Remove comment with Java.
24616         * tree-eh.c (lower_resx): Likewise.
24617         * tree.c (free_lang_data_in_type): Remove dead code.
24618         (find_decls_types_r): Likewise.
24619         (build_common_builtin_nodes): Remove Java from a comment.
24620         (verify_type): Remove dead code.
24621         * varasm.c (assemble_external): Remove Java from a comment.
24623 2017-07-14  Martin Liska  <mliska@suse.cz>
24625         * opts.c (finish_options): Add quotes.
24626         (common_handle_option): Likewise.
24628 2017-07-14  Martin Liska  <mliska@suse.cz>
24630         * dbxout.c (get_lang_number): Do not handle GNU Pascal.
24631         * dbxout.h (extern void dbxout_stab_value_internal_label_diff):
24632         Remove N_SO_PASCAL.
24633         * dwarf2out.c (lower_bound_default): Do not handle
24634         DW_LANG_Pascal83.
24635         (gen_compile_unit_die): Likewise.
24636         * gcc.c: Remove default extension binding for GNU Pascal.
24637         * stmt.c: Remove Pascal language from a comment.
24638         * xcoffout.c: Likewise.
24640 2017-07-13  David Malcolm  <dmalcolm@redhat.com>
24642         PR c/81405
24643         * diagnostic-show-locus.c (fixit_cmp): New function.
24644         (layout::layout): Sort m_fixit_hints.
24645         (column_range::column_range): Assert that the values are valid.
24646         (struct char_span): New struct.
24647         (correction::overwrite): New method.
24648         (struct source_line): New struct.
24649         (line_corrections::add_hint): Add assertions.  Reimplement memcpy
24650         calls in terms of classes source_line and char_span, and
24651         correction::overwrite.
24652         (selftest::test_overlapped_fixit_printing_2): New function.
24653         (selftest::diagnostic_show_locus_c_tests): Call it.
24655 2017-07-13  Will Schmidt  <will_schmidt@vnet.ibm.com>
24657         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Return
24658         early if there is no lhs.
24660 2017-07-13  Martin Liska  <mliska@suse.cz>
24662         * dwarf2out.c (gen_pointer_type_die): Remove dead code.
24663         (gen_reference_type_die): Likewise.
24664         * stor-layout.c: Remove Pascal-related comment.
24666 2017-07-13  Martin Liska  <mliska@suse.cz>
24668         * opts.c (finish_options): Add quotes to error messages.
24669         (parse_sanitizer_options): Likewise.
24671 2017-07-13  Thomas Preud'homme  <thomas.preudhomme@arm.com>
24673         * doc/invoke.texi (armv8-r): Document +fp.sp ARMv8-R extension.
24675 2017-07-13  Richard Earnshaw  <rearnsha@arm.com>
24677         * config/arm/vxworks.h (TARGET_ENDIAN_DEFAULT): Define.
24679 2017-07-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
24681         * asan.c (asan_emit_allocas_unpoison): Use ptr_mode for arguments
24682         during expansion.
24683         * builtins.c (expand_asan_emit_allocas_unpoison): Likewise.
24685 2017-07-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
24687         PR target/81193
24688         * config/rs6000/rs6000-c.c (rs6000_cpu_cpp_builtins): If GLIBC
24689         provides the hardware capability bits, define the macro
24690         __BUILTIN_CPU_SUPPORTS__.
24691         * config/rs6000/rs6000.c (cpu_expand_builtin): Generate a warning
24692         if GLIBC does not provide the hardware capability bits.  Add a
24693         gcc_unreachable call if the built-in cpu function is neither
24694         __builtin_cpu_is nor __builtin_cpu_supports.
24695         (rs6000_get_function_versions_dispatcher): Change the warning
24696         that an old GLIBC is used which does not export the capability
24697         bits to be an error.
24698         * doc/extend.texi (target_clones attribute): Document the
24699         restriction that GLIBC 2.23 or newer is needed on the PowerPC.
24700         (PowerPC built-in functions): Document that GLIBC 2.23 or newer is
24701         needed by __builtin_cpu_is and __builtin_cpu_supports.  Document
24702         the macros defined by GCC if the newer GLIBC is available.
24704 2017-07-12  Jeff Law  <law@redhat.com>
24706         * config/riscv/riscv.c: Remove unnecessary includes.  Reorder
24707         remaining includes slightly.
24708         * config/riscv/riscv-builtins.c: Include profile-count.h.
24710 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24712         PR target/79883
24713         * config/avr/avr.c (avr_set_current_function): In diagnostic
24714         messages: Quote keywords and (parts of) identifiers.
24715         [WITH_AVRLIBC]: Warn for functions named "ISR", "SIGNAL" or
24716         "INTERUPT".
24718 2017-07-12  Carl Love  <cel@us.ibm.com>
24720         * config/rs6000/rs6000-c.c: Add support for built-in functions
24721         vector bool char vec_revb (vector bool char);
24722         vector bool short vec_revb (vector short char);
24723         vector bool int vec_revb (vector bool int);
24724         vector bool long long vec_revb (vector bool long long);
24725         * doc/extend.texi: Update the built-in documentation file for the
24726         new built-in functions.
24728 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24730         * config/s390/s390.md: Remove movcc splitter.
24732 2017-07-12  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
24734         * config/s390/s390.c (s390_rtx_costs): Return proper costs for
24735         load/store on condition.
24737 2017-07-12  Georg-Johann Lay  <avr@gjlay.de>
24739         PR target/81407
24740         * config/avr/avr.c (avr_encode_section_info)
24741         [progmem && !TREE_READONLY]: Error if progmem object needs
24742         constructing.
24744 2017-07-11  Michael Collison  <michael.collison@arm.com>
24746         * config/aarch64/aarch64-simd.md (aarch64_sub<mode>_compare0):
24747         New pattern.
24749 2017-07-11  Carl Love  <cel@us.ibm.com>
24751         * config/rs6000/rs6000-c.c: Add support for builtins
24752         vector unsigned int vec_parity_lsbb (vector signed int);
24753         vector unsigned int vec_parity_lsbb (vector unsigned int);
24754         vector unsigned __int128 vec_parity_lsbb (vector signed __int128);
24755         vector unsigned __int128 vec_parity_lsbb (vector unsigned __int128);
24756         vector unsigned long long vec_parity_lsbb (vector signed long long);
24757         vector unsigned long long vec_parity_lsbb (vector unsigned long long);
24758         * config/rs6000/rs6000-builtin.def (VPARITY_LSBB): Add BU_P9V_OVERLOAD1.
24759         * config/rs6000/altivec.h (vec_parity_lsbb): Add define.
24760         * doc/extend.texi: Update the built-in documentation file for the
24761         new built-in functions.
24763 2017-07-11  David Malcolm  <dmalcolm@redhat.com>
24765         * diagnostic-show-locus.c: Include "gcc-rich-location.h".
24766         (layout::m_primary_loc): New field.
24767         (layout::layout): Initialize new field.  Move location filtering
24768         logic from here to...
24769         (layout::maybe_add_location_range): ...this new method.  Add
24770         support for filtering to just the lines already specified by other
24771         locations.
24772         (layout::will_show_line_p): New method.
24773         (gcc_rich_location::add_location_if_nearby): New method.
24774         (selftest::test_add_location_if_nearby): New test function.
24775         (selftest::diagnostic_show_locus_c_tests): Call it.
24776         * gcc-rich-location.h (gcc_rich_location::add_location_if_nearby):
24777         New method.
24779 2017-07-11  Tom de Vries  <tom@codesourcery.com>
24781         * config/nvptx/nvptx.c (WORKAROUND_PTXJIT_BUG): New macro.
24782         (bb_first_real_insn): New function.
24783         (nvptx_single): Add extra initialization of broadcasted condition
24784         variables.
24786 2017-07-11  Nathan Sidwell  <nathan@acm.org>
24788         * dwarf2out.c (gen_member_die): Remove useless check for anon ctors.
24790 2017-07-11  Georg-Johann Lay  <avr@gjlay.de>
24792         * doc/extend.texi (AVR Function Attributes): Remove weblink to
24793         Binutils doc as TEXI will mess them up.
24794         * doc/invoke.texi (AVR Options): Same here.
24796 2017-07-11  Daniel Cederman  <cederman@gaisler.com>
24798         * config/sparc/sparc.opt (mfix-ut700): New option.
24799         (mfix-gr712rc): Likewise.
24800         (sparc_fix_b2bst): New variable.
24801         * doc/invoke.texi (SPARC options): Document them.
24802         (ARM options): Fix warnings.
24803         * config/sparc/sparc.c (sparc_do_work_around_errata): Insert NOP
24804         instructions to prevent sequences that can trigger the store-store
24805         errata for certain LEON3FT processors.
24806         (pass_work_around_errata::gate): Also test sparc_fix_b2bst.
24807         (sparc_option_override): Set sparc_fix_b2bst appropriately.
24808         * config/sparc/sparc.md (fix_b2bst): New attribute.
24809         (in_branch_delay): Prevent stores in delay slot if fix_b2bst.
24811 2017-07-10  Uros Bizjak  <ubizjak@gmail.com>
24813         PR target/81375
24814         * config/i386/i386.md (divsf3): Add TARGET_SSE to TARGET_SSE_MATH.
24815         (rcpps): Ditto.
24816         (*rsqrtsf2_sse): Ditto.
24817         (rsqrtsf2): Ditto.
24818         (div<mode>3): Macroize insn from divdf3 and divsf3
24819         using MODEF mode iterator.
24821 2017-07-10  Martin Sebor  <msebor@redhat.com>
24823         PR tree-optimization/80397
24824         * gimple-ssa-sprintf.c (format_integer): Use INTEGRAL_TYPE_P()
24825         instead of testing for equality to INTEGER_TYPE.
24827 2017-07-10  Vineet Gupta <vgupta@synopsys.com>
24829         * config.gcc: Remove uclibc from arc target spec.
24831 2017-07-10  Claudiu Zissulescu  <claziss@synopsys.com>
24833         * config/arc/arc.h (ADDITIONAL_REGISTER_NAMES): Define.
24835 2017-07-07  Jan Hubicka  <hubicka@ucw.cz>
24837         PR lto/80838
24838         * lto-wrapper.c (remove_option): New function.
24839         (merge_and_complain): Merge PIC/PIE options more realistically.
24841 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24843         Better ISR prologues by supporting GASes __gcc_isr pseudo insn.
24845         PR target/20296
24846         PR target/81268
24847         * configure.ac [target=avr]: Add GAS check for -mgcc-isr.
24848         (HAVE_AS_AVR_MGCCISR_OPTION):  If so, AC_DEFINE it.
24849         * config.in: Regenerate.
24850         * configure: Regenerate.
24851         * doc/extend.texi (AVR Function Attributes) <no_gccisr>: Document it.
24852         * doc/invoke.texi (AVR Options) <-mgas-isr-prologues>: Document it.
24853         * config/avr/avr.opt (-mgas-isr-prologues): New option and...
24854         (TARGET_GASISR_PROLOGUES): ...target mask.
24855         * common/config/avr/avr-common.c
24856         (avr_option_optimization_table) [OPT_LEVELS_1_PLUS_NOT_DEBUG]:
24857         Set -mgas-isr-prologues.
24858         * config/avr/avr-passes.def (avr_pass_pre_proep): Add
24859         INSERT_PASS_BEFORE for it.
24860         * config/avr/avr-protos.h (make_avr_pass_pre_proep): New proto.
24861         * config/avr/avr.c (avr_option_override)
24862         [!HAVE_AS_AVR_MGCCISR_OPTION]: Unset TARGET_GASISR_PROLOGUES.
24863         (avr_no_gccisr_function_p, avr_hregs_split_reg): New static functions.
24864         (avr_attribute_table) <no_gccisr>: Add new function attribute.
24865         (avr_set_current_function) <is_no_gccisr>: Init machine field.
24866         (avr_pass_data_pre_proep, avr_pass_pre_proep): New pass data
24867         and rtl_opt_pass.
24868         (make_avr_pass_pre_proep): New function.
24869         (emit_push_sfr) <treg>: Add argument to function and use it
24870         instead of TMP_REG.
24871         (avr_expand_prologue) [machine->gasisr.maybe]: Emit gasisr insn
24872         and set machine->gasisr.yes.
24873         (avr_expand_epilogue) [machine->gasisr.yes]: Similar.
24874         (avr_asm_function_end_prologue) [machine->gasisr.yes]: Add
24875         __gcc_isr.n_pushed to .L__stack_usage.
24876         (TARGET_ASM_FINAL_POSTSCAN_INSN): Define to...
24877         (avr_asm_final_postscan_insn): ...this new static function.
24878         * config/avr/avr.h (machine_function)
24879         <is_no_gccisr, use_L__stack_usage>: New fields.
24880         <gasisr, gasisr.yes, gasisr.maybe, gasisr.regno>: New fields.
24881         * config/avr/avr.md (UNSPECV_GASISR): Add unspecv enum.
24882         (GASISR_Prologue, GASISR_Epilogue, GASISR_Done): New define_constants.
24883         (gasisr, *gasisr): New expander and insn.
24884         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
24885         [HAVE_AS_AVR_MGCCISR_OPTION]: Print asm_gccisr spec.
24886         * config/avr/specs.h (ASM_SPEC) <asm_gccisr>: Add sub spec.
24888 2017-07-10  Richard Earnshaw  <rearnsha@arm.com>
24890         * config/arm/parsecpu.awk (gen_comm_data): Do not escape single quotes
24891         in quoted strings.
24893 2017-07-10  Georg-Johann Lay  <avr@gjlay.de>
24895         Move jump-tables out of .text again.
24897         PR target/81075
24898         * config/avr/avr.c (ASM_OUTPUT_ADDR_VEC_ELT): Remove function.
24899         (ASM_OUTPUT_ADDR_VEC): New function.
24900         (avr_adjust_insn_length) [JUMP_TABLE_DATA_P]: Return 0.
24901         (avr_final_prescan_insn) [avr_log.insn_addresses]: Dump
24902         INSN_ADDRESSes as asm comment.
24903         * config/avr/avr.h (JUMP_TABLES_IN_TEXT_SECTION): Adjust comment.
24904         (ASM_OUTPUT_ADDR_VEC_ELT): Remove define.
24905         (ASM_OUTPUT_ADDR_VEC): Define to avr_output_addr_vec.
24906         * config/avr/avr.md (*tablejump): Adjust comment.
24907         * config/avr/elf.h (ASM_OUTPUT_BEFORE_CASE_LABEL): Remove.
24908         * config/avr/avr-log.c (avr_log_set_avr_log) <insn_addresses>:
24909         New detail.
24910         * config/avr/avr-protos.h (avr_output_addr_vec_elt): Remove proto.
24911         (avr_output_addr_vec): New proto.
24912         (avr_log_t) <insn_addresses>: New field.
24914 2017-07-09  H.J. Lu  <hongjiu.lu@intel.com>
24916         PR target/81313
24917         * config/i386/i386.c (ix86_function_arg_advance): Set
24918         outgoing_args_on_stack to true if there are outgoing arguments
24919         on stack.
24920         (ix86_function_arg): Likewise.
24921         (ix86_get_drap_rtx): Use DRAP only if there are outgoing
24922         arguments on stack and ACCUMULATE_OUTGOING_ARGS is false.
24923         * config/i386/i386.h (machine_function): Add
24924         outgoing_args_on_stack.
24926 2017-07-09  Krister Walfridsson  <krister.walfridsson@gmail.com>
24928         * config.gcc (*-*-netbsd*): Remove check for NetBSD versions not
24929         supporting pthreds.
24930         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always enable pthreads.
24932 2017-07-08  Richard Sandiford  <richard.sandiford@linaro.org>
24934         * Makefile.in (HOOKS_H, RTL_BASE_H, FUNCTION_H, EXPR_H, REGS_H)
24935         (REAL_H): Remove $(MACHMODE_H).
24936         (FIXED_VALUE_H, TREE_CORE_H, CFGLOOP_H): Remove $(MACHMODE_H) and
24937         double-int.h.
24938         (CORETYPES_H): Add signop.h, wide-int.h, wide-int-print.h,
24939         $(MACHMODE_H) and double-int.h.
24940         (build/min-insn-modes.o): Depend on $(CORETYPES_H) rather than
24941         $(MACHMODE_H).
24942         (gengtype-state.o, gengtype.o, build/gengtype.o): Don't depend on
24943         double-int.h.
24945 2017-07-07  Andrew Pinski  <apinski@cavium.com>
24947         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Check
24948         prev_set and curr_set for AARCH64_FUSE_ALU_BRANCH.
24950 2017-07-07  Michael Meissner  <meissner@linux.vnet.ibm.com>
24952         * config/rs6000/rs6000.c (rs6000_get_function_versions_dispatcher):
24953         Add warning if GCC was not configured to link against a GLIBC that
24954         exports the hardware capability bits.
24955         (make_resolver_func): Make resolver function private and not a
24956         COMDAT function.  Create the name with clone_function_name instead
24957         of make_unique_name.
24959         PR target/81348
24960         * config/rs6000/rs6000.md (HI sign_extend splitter): Use the
24961         correct operand in doing the split.
24963 2017-07-07 Carl Love  <cel@us.ibm.com>
24965         * config/rs6000/rs6000-c: Add support for built-in function
24966         vector unsigned short vec_pack_to_short_fp32 (vector float,
24967                                                       vector float).
24968         * config/rs6000/rs6000-builtin.def (CONVERT_4F32_8I16): Add
24969         BU_P9V_AV_2 and BU_P9V_OVERLOAD_2 definitions.
24970         * config/rs6000/altivec.h (vec_pack_to_short_fp32): Add define.
24971         * config/rs6000/altivec.md(UNSPEC_CONVERT_4F32_8I16): Add UNSPEC.
24972         (convert_4f32_8i16): Add define_expand.
24973         * doc/extend.texi: Update the built-in documentation file for the
24974         new built-in function.
24976 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24978         * config/sparc/m8.md: New file.
24979         * config/sparc/sparc.md: Include m8.md.
24981 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
24983         * config/sparc/sparc.opt: New option -mvis4b.
24984         * config/sparc/sparc.c (dump_target_flag_bits): Handle MASK_VIS4B.
24985         (sparc_option_override): Handle VIS4B.
24986         (enum sparc_builtins): Define
24987         SPARC_BUILTIN_DICTUNPACK{8,16,32},
24988         SPARC_BUILTIN_FPCMP{LE,GT,EQ,NE}{8,16,32}SHL,
24989         SPARC_BUILTIN_FPCMPU{LE,GT}{8,16,32}SHL,
24990         SPARC_BUILTIN_FPCMPDE{8,16,32}SHL and
24991         SPARC_BUILTIN_FPCMPUR{8,16,32}SHL.
24992         (check_constant_argument): New function.
24993         (sparc_vis_init_builtins): Define builtins
24994         __builtin_vis_dictunpack{8,16,32},
24995         __builtin_vis_fpcmp{le,gt,eq,ne}{8,16,32}shl,
24996         __builtin_vis_fpcmpu{le,gt}{8,16,32}shl,
24997         __builtin_vis_fpcmpde{8,16,32}shl and
24998         __builtin_vis_fpcmpur{8,16,32}shl.
24999         (sparc_expand_builtin): Check that the constant operands to
25000         __builtin_vis_fpcmp*shl and _builtin_vis_dictunpack* are indeed
25001         constant and in range.
25002         * config/sparc/sparc-c.c (sparc_target_macros): Handle
25003         TARGET_VIS4B.
25004         * config/sparc/sparc.h (SPARC_IMM2_P): Define.
25005         (SPARC_IMM5_P): Likewise.
25006         * config/sparc/sparc.md (cpu_feature): Add new feagure "vis4b".
25007         (enabled): Handle vis4b.
25008         (UNSPEC_DICTUNPACK): New unspec.
25009         (UNSPEC_FPCMPSHL): Likewise.
25010         (UNSPEC_FPUCMPSHL): Likewise.
25011         (UNSPEC_FPCMPDESHL): Likewise.
25012         (UNSPEC_FPCMPURSHL): Likewise.
25013         (cpu_feature): New CPU feature `vis4b'.
25014         (dictunpack{8,16,32}): New insns.
25015         (FPCSMODE): New mode iterator.
25016         (fpcscond): New code iterator.
25017         (fpcsucond): Likewise.
25018         (fpcmp{le,gt,eq,ne}{8,16,32}{si,di}shl): New insns.
25019         (fpcmpu{le,gt}{8,16,32}{si,di}shl): Likewise.
25020         (fpcmpde{8,16,32}{si,di}shl): Likewise.
25021         (fpcmpur{8,16,32}{si,di}shl): Likewise.
25022         * config/sparc/constraints.md: Define constraints `q' for unsigned
25023         2-bit integer constants and `t' for unsigned 5-bit integer
25024         constants.
25025         * config/sparc/predicates.md (imm5_operand_dictunpack8): New
25026         predicate.
25027         (imm5_operand_dictunpack16): Likewise.
25028         (imm5_operand_dictunpack32): Likewise.
25029         (imm2_operand): Likewise.
25030         * doc/invoke.texi (SPARC Options): Document -mvis4b.
25031         * doc/extend.texi (SPARC VIS Built-in Functions): Document the
25032         ditunpack* and fpcmp*shl builtins.
25034 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25036         * config.gcc: Handle m8 in --with-{cpu,tune} options.
25037         * config.in: Add HAVE_AS_SPARC6 define.
25038         * config/sparc/driver-sparc.c (cpu_names): Add entry for the SPARC
25039         M8.
25040         * config/sparc/sol2.h (CPP_CPU64_DEFAULT_SPEC): Define for
25041         TARGET_CPU_m8.
25042         (ASM_CPU32_DEFAUILT_SPEC): Likewise.
25043         (CPP_CPU_SPEC): Handle m8.
25044         (ASM_CPU_SPEC): Likewise.
25045         * config/sparc/sparc-opts.h (enum processor_type): Add
25046         PROCESSOR_M8.
25047         * config/sparc/sparc.c (m8_costs): New struct.
25048         (sparc_option_override): Handle TARGET_CPU_m8.
25049         (sparc32_initialize_trampoline): Likewise.
25050         (sparc64_initialize_trampoline): Likewise.
25051         (sparc_issue_rate): Likewise.
25052         (sparc_register_move_cost): Likewise.
25053         * config/sparc/sparc.h (TARGET_CPU_m8): Define.
25054         (CPP_CPU64_DEFAULT_SPEC): Define for M8.
25055         (ASM_CPU64_DEFAULT_SPEC): Likewise.
25056         (CPP_CPU_SPEC): Handle M8.
25057         (ASM_CPU_SPEC): Likewise.
25058         (AS_M8_FLAG): Define.
25059         * config/sparc/sparc.md: Add m8 to the cpu attribute.
25060         * config/sparc/sparc.opt: New option -mcpu=m8 for sparc targets.
25061         * configure.ac (HAVE_AS_SPARC6): Check for assembler support for
25062         M8 instructions.
25063         * configure: Regenerate.
25064         * doc/invoke.texi (SPARC Options): Document -mcpu=m8 and
25065         -mtune=m8.
25067 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25069         * config/sparc/niagara7.md: Rework the DFA scheduler to use insn
25070         subtypes.
25071         * config/sparc/sparc.md: Remove the `v3pipe' insn attribute.
25072         ("*movdi_insn_sp32"): Do not set v3pipe.
25073         ("*movsi_insn"): Likewise.
25074         ("*movdi_insn_sp64"): Likewise.
25075         ("*movsf_insn"): Likewise.
25076         ("*movdf_insn_sp32"): Likewise.
25077         ("*movdf_insn_sp64"): Likewise.
25078         ("*zero_extendsidi2_insn_sp64"): Likewise.
25079         ("*sign_extendsidi2_insn"): Likewise.
25080         ("*mov<VM32:mode>_insn"): Likewise.
25081         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25082         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25083         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25084         ("<vlop:code><VL:mode>3"): Likewise.
25085         ("*not_<vlop:code><VL:mode>3"): Likewise.
25086         ("*nand<VL:mode>_vis"): Likewise.
25087         ("*<vlnotop:code>_not1<VL:mode>_vis"): Likewise.
25088         ("*<vlnotop:code>_not2<VL:mode>_vis"): Likewise.
25089         ("one_cmpl<VL:mode>2"): Likewise.
25090         ("faligndata<VM64:mode>_vis"): Likewise.
25091         ("alignaddrsi_vis"): Likewise.
25092         ("alignaddrdi_vis"): Likweise.
25093         ("alignaddrlsi_vis"): Likewise.
25094         ("alignaddrldi_vis"): Likewise.
25095         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25096         ("bmaskdi_vis"): Likewise.
25097         ("bmasksi_vis"): Likewise.
25098         ("bshuffle<VM64:mode>_vis"): Likewise.
25099         ("cmask8<P:mode>_vis"): Likewise.
25100         ("cmask16<P:mode>_vis"): Likewise.
25101         ("cmask32<P:mode>_vis"): Likewise.
25102         ("pdistn<P:mode>_vis"): Likewise.
25103         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25105 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25107         * config/sparc/sparc.md ("subtype"): New insn attribute.
25108         ("*wrgsr_sp64"): Set insn subtype.
25109         ("*rdgsr_sp64"): Likewise.
25110         ("alignaddrsi_vis"): Likewise.
25111         ("alignaddrdi_vis"): Likewise.
25112         ("alignaddrlsi_vis"): Likewise.
25113         ("alignaddrldi_vis"): Likewise.
25114         ("<plusminus_insn><VADDSUB:mode>3"): Likewise.
25115         ("fexpand_vis"): Likewise.
25116         ("fpmerge_vis"): Likewise.
25117         ("faligndata<VM64:mode>_vis"): Likewise.
25118         ("bshuffle<VM64:mode>_vis"): Likewise.
25119         ("cmask8<P:mode>_vis"): Likewise.
25120         ("cmask16<P:mode>_vis"): Likewise.
25121         ("cmask32<P:mode>_vis"): Likewise.
25122         ("fchksm16_vis"): Likewise.
25123         ("v<vis3_shift_patname><GCM:mode>3"): Likewise.
25124         ("fmean16_vis"): Likewise.
25125         ("fp<plusminus_insn>64_vis"): Likewise.
25126         ("<plusminus_insn>v8qi3"): Likewise.
25127         ("<vis3_addsub_ss_patname><VASS:mode>3"): Likewise.
25128         ("<vis4_minmax_patname><VMMAX:mode>3"): Likewise.
25129         ("<vis4_uminmax_patname><VMMAX:mode>3"): Likewise.
25130         ("<vis3_addsub_ss_patname>v8qi3"): Likewise.
25131         ("<vis4_addsub_us_patname><VAUS:mode>3"): Likewise.
25132         ("*movqi_insn"): Likewise.
25133         ("*movhi_insn"): Likewise.
25134         ("*movsi_insn"): Likewise.
25135         ("movsi_pic_gotdata_op"): Likewise.
25136         ("*movdi_insn_sp32"): Likewise.
25137         ("*movdi_insn_sp64"): Likewise.
25138         ("movdi_pic_gotdata_op"): Likewise.
25139         ("*movsf_insn"): Likewise.
25140         ("*movdf_insn_sp32"): Likewise.
25141         ("*movdf_insn_sp64"): Likewise.
25142         ("*zero_extendhisi2_insn"): Likewise.
25143         ("*zero_extendqihi2_insn"): Likewise.
25144         ("*zero_extendqisi2_insn"): Likewise.
25145         ("*zero_extendqidi2_insn"): Likewise.
25146         ("*zero_extendhidi2_insn"): Likewise.
25147         ("*zero_extendsidi2_insn_sp64"): Likewise.
25148         ("ldfsr"): Likewise.
25149         ("prefetch_64"): Likewise.
25150         ("prefetch_32"): Likewise.
25151         ("tie_ld32"): Likewise.
25152         ("tie_ld64"): Likewise.
25153         ("*tldo_ldub_sp32"): Likewise.
25154         ("*tldo_ldub1_sp32"): Likewise.
25155         ("*tldo_ldub2_sp32"): Likewise.
25156         ("*tldo_ldub_sp64"): Likewise.
25157         ("*tldo_ldub1_sp64"): Likewise.
25158         ("*tldo_ldub2_sp64"): Likewise.
25159         ("*tldo_ldub3_sp64"): Likewise.
25160         ("*tldo_lduh_sp32"): Likewise.
25161         ("*tldo_lduh1_sp32"): Likewise.
25162         ("*tldo_lduh_sp64"): Likewise.
25163         ("*tldo_lduh1_sp64"): Likewise.
25164         ("*tldo_lduh2_sp64"): Likewise.
25165         ("*tldo_lduw_sp32"): Likewise.
25166         ("*tldo_lduw_sp64"): Likewise.
25167         ("*tldo_lduw1_sp64"): Likewise.
25168         ("*tldo_ldx_sp64"): Likewise.
25169         ("*mov<VM32:mode>_insn"): Likewise.
25170         ("*mov<VM64:mode>_insn_sp64"): Likewise.
25171         ("*mov<VM64:mode>_insn_sp32"): Likewise.
25173 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25175         * config/sparc/sparc.md ("type"): New insn type viscmp.
25176         ("fcmp<gcond:code><GCM:gcm_name><P:mode>_vis"): Set insn type to
25177         viscmp.
25178         ("fpcmp<gcond:code>8<P:mode>_vis"): Likewise.
25179         ("fucmp<gcond:code>8<P:mode>_vis"): Likewise.
25180         ("fpcmpu<gcond:code><GCM:gcm_name><P:mode>_vis"): Likewise.
25181         * config/sparc/niagara7.md ("n7_vis_logical_v3pipe"): Handle
25182         viscmp.
25183         ("n7_vis_logical_11cycle"): Likewise.
25184         * config/sparc/niagara4.md ("n4_vis_logical"): Likewise.
25185         * config/sparc/niagara2.md ("niag3_vis": Likewise.
25186         * config/sparc/niagara.md ("niag_vis"): Likewise.
25187         * config/sparc/ultra3.md ("us3_fga"): Likewise.
25188         * config/sparc/ultra1_2.md ("us1_fga_double"): Likewise.
25190 2017-07-07  Jose E. Marchesi  <jose.marchesi@oracle.com>
25192         * config/sparc/sparc.md: New instruction type `bmask'.
25193         (bmaskdi_vis): Use the `bmask' type.
25194         (bmasksi_vis): Likewise.
25195         * config/sparc/ultra3.md (us3_array): Likewise.
25196         * config/sparc/niagara7.md (n7_array): Likewise.
25197         * config/sparc/niagara4.md (n4_array): Likewise.
25198         * config/sparc/niagara2.md (niag2_vis): Likewise.
25199         (niag3_vis): Likewise.
25200         * config/sparc/niagara.md (niag_vis): Likewise.
25202 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25204         * ipa-comdats.c: Remove optimize check from gate.
25205         * ipa-fnsummary.c (ipa_fn_summary_generate): do not generate summary
25206         for functions not optimized.
25207         (ipa_fn_summary_read): Skip optimize check.
25208         (ipa_fn_summary_write): Likewise.
25209         * ipa-inline-analysis.c (do_estimate_growth_1): Check that caller
25210         is optimized.
25211         * ipa-inline.c (can_inline_edge_p): Not optimized functions are
25212         uninlinable.
25213         (can_inline_edge_p): Check flag_pcc_struct_return for match.
25214         (check_callers): Give up on caller which is not optimized.
25215         (inline_small_functions): Likewise.
25216         (ipa_inline): Do not give up when not optimizing.
25217         * ipa-visbility.c (function_and_variable_visibility): Do not optimize
25218         away unoptimizes cdtors.
25219         (whole_program_function_and_variable_visibility): Do
25220         ipa_discover_readonly_nonaddressable_vars in LTO mode.
25221         * ipa.c (process_references): Do not check optimize.
25222         (symbol_table::remove_unreachable_nodes): Update optimize check.
25223         (set_writeonly_bit): Update optimize check.
25224         (pass_ipa_cdtor_merge::gate): Do not check optimize.
25225         (pass_ipa_single_use::gate): Remove.
25227 2017-07-06  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
25229         * config/rs6000/rs6000.c (union_defs, union_uses, insn_is_load_p,
25230         insn_is_store_p, insn_is_swap_p, const_load_sequence_p, v2df_reduction_p,
25231         rtx_is_swappable_p, insn_is_swappable_p, chain_contains_only_swaps,
25232         mark_swaps_for_removal, swap_const_vector_halves, adjust_subreg_index,
25233         permute_load, permute_store, adjust_extract, adjust_splat,
25234         adjust_xxpermdi, adjust_concat, adjust_vperm, handle_special_swappables,
25235         replace_swap_with_copy, dump_swap_insn_table,
25236         alignment_with_canonical_addr, alignment_mask, find_alignment_op,
25237         recombine_lvx_pattern, recombine_stvx_pattern,
25238         recombine_lvx_stvx_patterns, rs6000_analyze_swaps,
25239         make_pass_analyze_swaps): Move all code related to p8 swap optimizations
25240         to file rs6000-p8swap.c.
25241         * config/rs6000/rs6000-p8swap.c: New file.
25242         * config/rs6000/t-rs6000: Add rule to build rs6000-p8swap.o.
25243         * config.gcc: Add rs6000-p8swap.o to extra_objs for powerpc*-*-*
25244         and rs6000*-*-* targets.
25246 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25248         * Makefile.in (selftest): Remove dependency on s-selftest-c++.
25250 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25252         * lto-wrapper.c (merge_and_complain): Do not merge
25253         fexceptions, fnon_call_exceptions, ftrapv, ffp_contract_, fmath_errno,
25254         fsigned_zeros, ftrapping_math, fwrapv.
25255         (append_compiler_options): Do not track these options.
25256         (append_linker_options): Likewie
25258 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25260         * cgraphunit.c (cgraph_node::finalize_function): When
25261         !flag_toplevel_reorde set no_reorder flag.
25262         (varpool_node::finalize_decl): Likewise.
25263         (symbol_table::compile): Drop no toplevel reorder path.
25265 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25267         * bb-reorder.c (better_edge_p): Do not build traces across abnormal/eh
25268         edges; zero probability is not better than uninitialized.
25270 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25272         * asan.h (asan_sanitize_allocas_p): Declare.
25273         * asan.c (asan_sanitize_allocas_p): New function.
25274         (handle_builtin_stack_restore): Bail out if !asan_sanitize_allocas_p.
25275         (handle_builtin_alloca): Likewise.
25276         * cfgexpand.c (expand_used_vars): Do not add allocas unpoisoning stuff
25277         if !asan_sanitize_allocas_p.
25278         * params.def (asan-instrument-allocas): Add new option.
25279         * params.h (ASAN_PROTECT_ALLOCAS): Define.
25280         * opts.c (common_handle_option): Disable allocas sanitization for
25281         KASan by default.
25283 2017-07-06  Maxim Ostapenko  <m.ostapenko@samsung.com>
25285         * asan.c: Include gimple-fold.h.
25286         (get_last_alloca_addr): New function.
25287         (handle_builtin_stackrestore): Likewise.
25288         (handle_builtin_alloca): Likewise.
25289         (asan_emit_allocas_unpoison): Likewise.
25290         (get_mem_refs_of_builtin_call): Add new parameter, remove const
25291         quallifier from first paramerer. Handle BUILT_IN_ALLOCA,
25292         BUILT_IN_ALLOCA_WITH_ALIGN and BUILT_IN_STACK_RESTORE builtins.
25293         (instrument_builtin_call): Pass gimple iterator to
25294         get_mem_refs_of_builtin_call.
25295         (last_alloca_addr): New global.
25296         * asan.h (asan_emit_allocas_unpoison): Declare.
25297         * builtins.c (expand_asan_emit_allocas_unpoison): New function.
25298         (expand_builtin): Handle BUILT_IN_ASAN_ALLOCAS_UNPOISON.
25299         * cfgexpand.c (expand_used_vars): Call asan_emit_allocas_unpoison
25300         if function calls alloca.
25301         * gimple-fold.c (replace_call_with_value): Remove static keyword.
25302         * gimple-fold.h (replace_call_with_value): Declare.
25303         * internal-fn.c: Include asan.h.
25304         * sanitizer.def (BUILT_IN_ASAN_ALLOCA_POISON,
25305         BUILT_IN_ASAN_ALLOCAS_UNPOISON): New builtins.
25307 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25309         * Makefile.in (SELFTEST_FLAGS): Drop "-x c", moving it to...
25310         (C_SELFTEST_FLAGS): New.
25311         (CPP_SELFTEST_FLAGS): New.
25312         (SELFTEST_DEPS): New, from deps of s-selftest.
25313         (C_SELFTEST_DEPS): New, from deps of s-selftest.
25314         (CPP_SELFTEST_DEPS): New.
25315         (selftest): Add dependency on s-selftest-c++.
25316         (s-selftest): Rename to...
25317         (s-selftest-c): ...this, moving deps to SELFTEST_DEPS
25318         and C_SELFTEST_DEPS, and using C_SELFTEST_FLAGS rather
25319         than SELFTEST_FLAGS.
25320         (selftest-gdb): Rename to...
25321         (selftest-c-gdb): ...this, using C_SELFTEST_DEPS and
25322         C_SELFTEST_FLAGS.
25323         (selftest-gdb): Reintroduce as an alias for selftest-c-gdb.
25324         (selftest-valgrind): Rename to...
25325         (selftest-c-valgrind): ...this, using C_SELFTEST_DEPS and
25326         C_SELFTEST_FLAGS.
25327         (selftest-valgrind): Reintroduce as an alias for
25328         selftest-c-valgrind.
25329         (s-selftest-c++): New.
25330         (selftest-c++-gdb): New.
25331         (selftest-c++-valgrind): New.
25333 2017-07-06  Olivier Hainque  <hainque@adacore.com>
25335         * gcc.c (process_command): When deciding if undefined variables
25336         should be ignored when processing specs, accept "gcc -v" as well.
25338 2017-07-06  Jan Hubicka  <hubicka@ucw.cz>
25340         * auto-profile.c (afdo_set_bb_count, afdo_propagate_edge,
25341         afdo_annotate_cfg): Set counts/probabilities as determined by afdo.
25343 2017-07-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25345         * config/arm/arm-cpus.in (armv8-r): Add new entry.
25346         * config/arm/arm-isa.h (ISA_ARMv8r): Define macro.
25347         * config/arm/arm-tables.opt: Regenerate.
25348         * config/arm/arm.h (enum base_architecture): Add BASE_ARCH_8R
25349         enumerator.
25350         * doc/invoke.texi: Mention -march=armv8-r and its extensions.
25352 2017-07-06  Carl Love  <cel@us.ibm.com>
25354         * ChangeLog: Clean up from mid air collision
25356 2017-07-06  Carl Love  <cel@us.ibm.com>
25358         * config/rs6000/rs6000-c.c: Add support for built-in functions
25359         vector signed int vec_subc (vector signed int, vector signed int);
25360         vector signed __int128 vec_subc (vector signed __int128,
25361                                          vector signed __int128);
25362         vector unsigned __int128 vec_subc (vector unsigned __int128,
25363                                            vector unsigned __int128);
25364         vector signed int vec_sube (vector signed int, vector signed int,
25365                                     vector signed int);
25366         vector unsigned int vec_sube (vector unsigned int,
25367                                       vector unsigned int,
25368                                       vector unsigned int);
25369         vector signed __int128 vec_sube (vector signed __int128,
25370                                          vector signed __int128,
25371                                          vector signed__int128);
25372         vector unsigned __int128 vec_sube (vector unsigned __int128,
25373                                            vector unsigned __int128,
25374                                            vector unsigned __int128);
25375         vector signed int vec_subec (vector signed int, vector signed int,
25376                                      vector signed int);
25377         vector unsigned int vec_subec (vector unsigned int,
25378                                        vector unsigned int,
25379                                        vector unsigned int);
25380         vector signed __int128 vec_subec (vector signed __int128,
25381                                           vector signed __int128,
25382                                           vector signed__int128);
25383         vector unsigned __int128 vec_subec (vector unsigned __int128,
25384                                             vector unsigned __int128,
25385                                             vector unsigned __int128);
25386         * config/rs6000/rs6000.c (ALTIVEC_BUILTIN_VEC_SUBE,
25387         ALTIVEC_BUILTIN_VEC_SUBEC): Add ef_builtins.
25388         * config/rs6000/rs6000-builtin.def (SUBE, SUBEC): Add
25389         BU_ALTIVEC_OVERLOAD_X definitions.
25390         * config/rs6000/altivec.h (vec_sube, vec_subec): Add builtin defines.
25391         * doc/extend.texi: Update the built-in documentation file for the new
25392         built-in functions.
25394 2017-07-06  David Malcolm  <dmalcolm@redhat.com>
25396         PR c++/79300
25397         * diagnostic-show-locus.c (layout::layout): Use start and finish
25398         spelling location for the start and finish of each range.
25399         * genmatch.c (linemap_client_expand_location_to_spelling_point):
25400         Add unused aspect param.
25401         * input.c (expand_location_1): Add "aspect" param, and use it
25402         to access the correct part of the location.
25403         (expand_location): Pass LOCATION_ASPECT_CARET to new param of
25404         expand_location_1.
25405         (expand_location_to_spelling_point): Likewise.
25406         (linemap_client_expand_location_to_spelling_point): Add "aspect"
25407         param, and pass it to expand_location_1.
25409 2017-07-06  Sebastian Peryt  <sebastian.peryt@intel.com>
25411         * config/i386/avx512fintrin.h (_mm_mask_getexp_round_ss,
25412         _mm_maskz_getexp_round_ss, _mm_mask_getexp_round_sd,
25413         _mm_maskz_getexp_round_sd, _mm_mask_getmant_round_sd,
25414         _mm_maskz_getmant_round_sd, _mm_mask_getmant_round_ss,
25415         _mm_maskz_getmant_round_ss, _mm_mask_getexp_ss, _mm_maskz_getexp_ss,
25416         _mm_mask_getexp_sd, _mm_maskz_getexp_sd, _mm_mask_getmant_sd,
25417         _mm_maskz_getmant_sd, _mm_mask_getmant_ss,
25418         _mm_maskz_getmant_ss): New intrinsics.
25419         (__builtin_ia32_getexpss128_mask): Changed to ...
25420         __builtin_ia32_getexpss128_round ... this.
25421         (__builtin_ia32_getexpsd128_mask): Changed to ...
25422         __builtin_ia32_getexpsd128_round ... this.
25423         * config/i386/i386-builtin-types.def
25424         ((V2DF, V2DF, V2DF, INT, V2DF, UQI, INT),
25425         (V4SF, V4SF, V4SF, INT, V4SF, UQI, INT)): New function type aliases.
25426         * config/i386/i386-builtin.def (__builtin_ia32_getexpsd_mask_round,
25427         __builtin_ia32_getexpss_mask_round,     __builtin_ia32_getmantsd_mask_round,
25428         __builtin_ia32_getmantss_mask_round): New builtins.
25429         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_INT_V2DF_UQI_INT,
25430         V4SF_FTYPE_V4SF_V4SF_INT_V4SF_UQI_INT): Handle new types.
25431         (CODE_FOR_avx512f_vgetmantv2df_mask_round,
25432         CODE_FOR_avx512f_vgetmantv4sf_mask_round): New cases.
25433         * config/i386/sse.md
25434         (avx512f_sgetexp<mode><round_saeonly_name>): Changed to ...
25435         avx512f_sgetexp<mode><mask_scalar_name>
25436         <round_saeonly_scalar_name> ... this.
25437         (vgetexp<ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|
25438         %0, %1, %2<round_saeonly_op3>}): Changed to ...
25439         vgetexp<ssescalarmodesuffix>
25440         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25441         %0<mask_scalar_operand3>, %1, %2<round_saeonly_scalar_mask_op3>} ... this.
25442         (avx512f_vgetmant<mode><round_saeonly_name>): Changed to ...
25443         avx512f_vgetmant<mode><mask_scalar_name>
25444         <round_saeonly_scalar_name> ... this.
25445         (vgetmant<ssescalarmodesuffix>\t{%3, <round_saeonly_op4>%2, %1, %0|
25446         %0, %1, %2<round_saeonly_op4>, %3}): Changed to ...
25447         vgetmant<ssescalarmodesuffix>
25448         \t{%3, <round_saeonly_scalar_mask_op4>%2, %1, %0<mask_scalar_operand4>|
25449         %0<mask_scalar_operand4>, %1, %2
25450         <round_saeonly_scalar_mask_op4>, %3} ... this.
25451         * config/i386/subst.md (mask_scalar_operand4,
25452         round_saeonly_scalar_mask_operand4,     round_saeonly_scalar_mask_op4,
25453         round_saeonly_scalar_nimm_predicate): New subst attributes.
25455 2017-07-06  Julia Koval  <julia.koval@intel.com>
25457         * config/i386/i386.c (ix86_erase_embedded_rounding):
25458         Remove code for old rounding pattern.
25460 2017-07-06  Richard Earnshaw  <rearnsha@arm.com>
25462         * config/arm/t-arm (GTM_H): Add arm-cpu.h.
25464 2017-07-06  Christophe Lyon  <christophe.lyon@linaro.org>
25466         * doc/sourcebuild.texi (Test Directives, Variants of
25467         dg-require-support): Add documentation for dg-require-stack-check.
25469 2017-07-05  Sebastian Peryt  <sebastian.peryt@intel.com>
25471         * config/i386/subst.md (mask_scalar, round_scalar,
25472         round_saeonly_scalar): New meta-templates.
25473         (mask_scalar_name, mask_scalar_operand3, round_scalar_name,
25474         round_scalar_mask_operand3, round_scalar_mask_op3,
25475         round_scalar_constraint, round_scalar_prefix, round_saeonly_scalar_name,
25476         round_saeonly_scalar_mask_operand3, round_saeonly_scalar_mask_op3,
25477         round_saeonly_scalar_constraint,
25478         round_saeonly_scalar_prefix): New subst attribute.
25479         * config/i386/sse.md
25480         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): Renamed to ...
25481         <sse>_vm<plusminus_insn><mode>3<mask_scalar_name>
25482         <round_scalar_name> ... this.
25483         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): Renamed to ...
25484         <sse>_vm<multdiv_mnemonic><mode>3<mask_scalar_name>
25485         <round_scalar_name> ... this.
25486         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): Renamed to ...
25487         <sse>_vm<code><mode>3<mask_scalar_name>
25488         <round_saeonly_scalar_name> ... this.
25489         (v<plusminus_mnemonic><ssescalarmodesuffix>
25490         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25491         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25492         v<plusminus_mnemonic><ssescalarmodesuffix>
25493         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25494         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25495         (v<multdiv_mnemonic><ssescalarmodesuffix>
25496         \t{<round_mask_op3>%2, %1, %0<mask_operand3>|
25497         %0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}): Changed to ...
25498         v<multdiv_mnemonic><ssescalarmodesuffix>
25499         \t{<round_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25500         %0<mask_scalar_operand3>, %1, %<iptr>2<round_scalar_mask_op3>} ... this.
25501         (v<maxmin_float><ssescalarmodesuffix>
25502         \t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|
25503         %0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}): Changed to ...
25504         v<maxmin_float><ssescalarmodesuffix>
25505         \t{<round_saeonly_scalar_mask_op3>%2, %1, %0<mask_scalar_operand3>|
25506         %0<mask_scalar_operand3>, %1, %<iptr>2
25507         <round_saeonly_scalar_mask_op3>} ... this.
25509 2017-07-05  Richard Earnshaw  <rearnsha@arm.com>
25511         * config/arm/arm.c (arm_fixed_condition_code_regs): New function.
25512         (TARGET_FIXED_CONDITION_CODE_REGS): Redefine.
25514 2017-07-05  Richard Sandiford  <richard.sandiford@linaro.org>
25515             Alan Hayward  <alan.hayward@arm.com>
25516             David Sherwood  <david.sherwood@arm.com>
25518         * combine.c (simplify_if_then_else): Remove "enum" before
25519         "machine_mode".
25520         * compare-elim.c (can_eliminate_compare): Likewise.
25521         * config/aarch64/aarch64-builtins.c (aarch64_simd_builtin_std_type):
25522         Likewise.
25523         (aarch64_lookup_simd_builtin_type): Likewise.
25524         (aarch64_simd_builtin_type): Likewise.
25525         (aarch64_init_simd_builtin_types): Likewise.
25526         (aarch64_simd_expand_args): Likewise.
25527         * config/aarch64/aarch64-protos.h (aarch64_simd_attr_length_rglist):
25528         Likewise.
25529         (aarch64_reverse_mask): Likewise.
25530         (aarch64_simd_emit_reg_reg_move): Likewise.
25531         (aarch64_gen_adjusted_ldpstp): Likewise.
25532         (aarch64_ccmp_mode_to_code): Likewise.
25533         (aarch64_operands_ok_for_ldpstp): Likewise.
25534         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25535         * config/aarch64/aarch64.c (aarch64_ira_change_pseudo_allocno_class):
25536         Likewise.
25537         (aarch64_min_divisions_for_recip_mul): Likewise.
25538         (aarch64_reassociation_width): Likewise.
25539         (aarch64_get_condition_code_1): Likewise.
25540         (aarch64_simd_emit_reg_reg_move): Likewise.
25541         (aarch64_simd_attr_length_rglist): Likewise.
25542         (aarch64_reverse_mask): Likewise.
25543         (aarch64_operands_ok_for_ldpstp): Likewise.
25544         (aarch64_operands_adjust_ok_for_ldpstp): Likewise.
25545         (aarch64_gen_adjusted_ldpstp): Likewise.
25546         * config/aarch64/cortex-a57-fma-steering.c (fma_node::rename):
25547         Likewise.
25548         * config/arc/arc.c (legitimate_offset_address_p): Likewise.
25549         * config/arm/arm-builtins.c (arm_simd_builtin_std_type): Likewise.
25550         (arm_lookup_simd_builtin_type): Likewise.
25551         (arm_simd_builtin_type): Likewise.
25552         (arm_init_simd_builtin_types): Likewise.
25553         (arm_expand_builtin_args): Likewise.
25554         * config/arm/arm-protos.h (arm_expand_builtin): Likewise.
25555         * config/ft32/ft32.c (ft32_libcall_value): Likewise.
25556         (ft32_setup_incoming_varargs): Likewise.
25557         (ft32_function_arg): Likewise.
25558         (ft32_function_arg_advance): Likewise.
25559         (ft32_pass_by_reference): Likewise.
25560         (ft32_arg_partial_bytes): Likewise.
25561         (ft32_valid_pointer_mode): Likewise.
25562         (ft32_addr_space_pointer_mode): Likewise.
25563         (ft32_addr_space_legitimate_address_p): Likewise.
25564         * config/i386/i386-protos.h (ix86_operands_ok_for_move_multiple):
25565         Likewise.
25566         * config/i386/i386.c (ix86_setup_incoming_vararg_bounds): Likewise.
25567         (ix86_emit_outlined_ms2sysv_restore): Likewise.
25568         (iamcu_alignment): Likewise.
25569         (canonicalize_vector_int_perm): Likewise.
25570         (ix86_noce_conversion_profitable_p): Likewise.
25571         (ix86_mpx_bound_mode): Likewise.
25572         (ix86_operands_ok_for_move_multiple): Likewise.
25573         * config/microblaze/microblaze-protos.h
25574         (microblaze_expand_conditional_branch_reg): Likewise.
25575         * config/microblaze/microblaze.c
25576         (microblaze_expand_conditional_branch_reg): Likewise.
25577         * config/powerpcspe/powerpcspe.c (rs6000_init_hard_regno_mode_ok):
25578         Likewise.
25579         (rs6000_reassociation_width): Likewise.
25580         (rs6000_invalid_binary_op): Likewise.
25581         (fusion_p9_p): Likewise.
25582         (emit_fusion_p9_load): Likewise.
25583         (emit_fusion_p9_store): Likewise.
25584         * config/riscv/riscv-protos.h (riscv_regno_mode_ok_for_base_p):
25585         Likewise.
25586         (riscv_hard_regno_mode_ok_p): Likewise.
25587         (riscv_address_insns): Likewise.
25588         (riscv_split_symbol): Likewise.
25589         (riscv_legitimize_move): Likewise.
25590         (riscv_function_value): Likewise.
25591         (riscv_hard_regno_nregs): Likewise.
25592         (riscv_expand_builtin): Likewise.
25593         * config/riscv/riscv.c (riscv_build_integer_1): Likewise.
25594         (riscv_build_integer): Likewise.
25595         (riscv_split_integer): Likewise.
25596         (riscv_legitimate_constant_p): Likewise.
25597         (riscv_cannot_force_const_mem): Likewise.
25598         (riscv_regno_mode_ok_for_base_p): Likewise.
25599         (riscv_valid_base_register_p): Likewise.
25600         (riscv_valid_offset_p): Likewise.
25601         (riscv_valid_lo_sum_p): Likewise.
25602         (riscv_classify_address): Likewise.
25603         (riscv_legitimate_address_p): Likewise.
25604         (riscv_address_insns): Likewise.
25605         (riscv_load_store_insns): Likewise.
25606         (riscv_force_binary): Likewise.
25607         (riscv_split_symbol): Likewise.
25608         (riscv_force_address): Likewise.
25609         (riscv_legitimize_address): Likewise.
25610         (riscv_move_integer): Likewise.
25611         (riscv_legitimize_const_move): Likewise.
25612         (riscv_legitimize_move): Likewise.
25613         (riscv_address_cost): Likewise.
25614         (riscv_subword): Likewise.
25615         (riscv_output_move): Likewise.
25616         (riscv_canonicalize_int_order_test): Likewise.
25617         (riscv_emit_int_order_test): Likewise.
25618         (riscv_function_arg_boundary): Likewise.
25619         (riscv_pass_mode_in_fpr_p): Likewise.
25620         (riscv_pass_fpr_single): Likewise.
25621         (riscv_pass_fpr_pair): Likewise.
25622         (riscv_get_arg_info): Likewise.
25623         (riscv_function_arg): Likewise.
25624         (riscv_function_arg_advance): Likewise.
25625         (riscv_arg_partial_bytes): Likewise.
25626         (riscv_function_value): Likewise.
25627         (riscv_pass_by_reference): Likewise.
25628         (riscv_setup_incoming_varargs): Likewise.
25629         (riscv_print_operand): Likewise.
25630         (riscv_elf_select_rtx_section): Likewise.
25631         (riscv_save_restore_reg): Likewise.
25632         (riscv_for_each_saved_reg): Likewise.
25633         (riscv_register_move_cost): Likewise.
25634         (riscv_hard_regno_mode_ok_p): Likewise.
25635         (riscv_hard_regno_nregs): Likewise.
25636         (riscv_class_max_nregs): Likewise.
25637         (riscv_memory_move_cost): Likewise.
25638         * config/rl78/rl78-protos.h (rl78_split_movsi): Likewise.
25639         * config/rl78/rl78.c (rl78_split_movsi): Likewise.
25640         (rl78_addr_space_address_mode): Likewise.
25641         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin):
25642         Likewise.
25643         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Likewise.
25644         (rs6000_reassociation_width): Likewise.
25645         (rs6000_invalid_binary_op): Likewise.
25646         (fusion_p9_p): Likewise.
25647         (emit_fusion_p9_load): Likewise.
25648         (emit_fusion_p9_store): Likewise.
25649         * config/visium/visium-protos.h (prepare_move_operands): Likewise.
25650         (ok_for_simple_move_operands): Likewise.
25651         (ok_for_simple_move_strict_operands): Likewise.
25652         (ok_for_simple_arith_logic_operands): Likewise.
25653         (visium_legitimize_reload_address): Likewise.
25654         (visium_select_cc_mode): Likewise.
25655         (output_cbranch): Likewise.
25656         (visium_split_double_move): Likewise.
25657         (visium_expand_copysign): Likewise.
25658         (visium_expand_int_cstore): Likewise.
25659         (visium_expand_fp_cstore): Likewise.
25660         * config/visium/visium.c (visium_pass_by_reference): Likewise.
25661         (visium_function_arg): Likewise.
25662         (visium_function_arg_advance): Likewise.
25663         (visium_libcall_value): Likewise.
25664         (visium_setup_incoming_varargs): Likewise.
25665         (visium_legitimate_constant_p): Likewise.
25666         (visium_legitimate_address_p): Likewise.
25667         (visium_legitimize_address): Likewise.
25668         (visium_secondary_reload): Likewise.
25669         (visium_register_move_cost): Likewise.
25670         (visium_memory_move_cost): Likewise.
25671         (prepare_move_operands): Likewise.
25672         (ok_for_simple_move_operands): Likewise.
25673         (ok_for_simple_move_strict_operands): Likewise.
25674         (ok_for_simple_arith_logic_operands): Likewise.
25675         (visium_function_value_1): Likewise.
25676         (rtx_ok_for_offset_p): Likewise.
25677         (visium_legitimize_reload_address): Likewise.
25678         (visium_split_double_move): Likewise.
25679         (visium_expand_copysign): Likewise.
25680         (visium_expand_int_cstore): Likewise.
25681         (visium_expand_fp_cstore): Likewise.
25682         (visium_split_cstore): Likewise.
25683         (visium_select_cc_mode): Likewise.
25684         (visium_split_cbranch): Likewise.
25685         (output_cbranch): Likewise.
25686         (visium_print_operand_address): Likewise.
25687         * expmed.c (flip_storage_order): Likewise.
25688         * expmed.h (emit_cstore): Likewise.
25689         (flip_storage_order): Likewise.
25690         * genrecog.c (validate_pattern): Likewise.
25691         * hsa-gen.c (gen_hsa_addr): Likewise.
25692         * internal-fn.c (expand_arith_overflow): Likewise.
25693         * ira-color.c (allocno_copy_cost_saving): Likewise.
25694         * lra-assigns.c (find_hard_regno_for_1): Likewise.
25695         * lra-constraints.c (prohibited_class_reg_set_mode_p): Likewise.
25696         (process_invariant_for_inheritance): Likewise.
25697         * lra-eliminations.c (move_plus_up): Likewise.
25698         * omp-low.c (lower_oacc_reductions): Likewise.
25699         * simplify-rtx.c (simplify_subreg): Likewise.
25700         * target.def (TARGET_SETUP_INCOMING_VARARG_BOUNDS): Likewise.
25701         (TARGET_CHKP_BOUND_MODE): Likewise..
25702         * targhooks.c (default_chkp_bound_mode): Likewise.
25703         (default_setup_incoming_vararg_bounds): Likewise.
25704         * targhooks.h (default_chkp_bound_mode): Likewise.
25705         (default_setup_incoming_vararg_bounds): Likewise.
25706         * tree-ssa-math-opts.c (divmod_candidate_p): Likewise.
25707         * tree-vect-loop.c (calc_vec_perm_mask_for_shift): Likewise.
25708         (have_whole_vector_shift): Likewise.
25709         * tree-vect-stmts.c (vectorizable_load): Likewise.
25710         * doc/tm.texi: Regenerate.
25712 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25714         Graceful degrade if Binutils PR21472 is not available.
25716         PR target/81072
25717         * configure.ac [target=avr]: WARN instead of ERROR if avrxmega3
25718         .rodata in flash test fails.
25719         (HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH): Define it if test passes.
25720         * confgure: Regenerate.
25721         * config.in: Regenerate.
25722         * config/avr/avr.c (avr_asm_named_section)
25723         [HAVE_LD_AVR_AVRXMEGA3_RODATA_IN_FLASH]: Only trigger
25724         __do_copy_data for stuff in .rodata if flash_pm_offset = 0.
25725         (avr_asm_init_sections): Same.
25727 2017-07-05  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
25729         * config/arm/neon.md (fma<VCVTF:mode>4): Remove %?.
25730         (fma<VH:mode>4_intrinsic): Likewise.
25731         (*fmsub<VCVTF:mode>4): Likewise.
25732         (*fmsub<VH:mode>4_intrinsic): Likewise.
25734 2017-07-05  Georg-Johann Lay  <avr@gjlay.de>
25736         PR target/81305
25737         * config/avr/avr.c (avr_out_movhi_mr_r_xmega) [CONSTANT_ADDRESS_P]:
25738         Don't depend on "optimize > 0".
25739         (out_movhi_r_mr, out_movqi_mr_r): Same.
25740         (out_movhi_mr_r, out_movqi_r_mr): Same.
25741         (avr_address_cost) [CONSTANT_ADDRESS_P]: Don't depend cost for
25742         io_address_operand on "optimize > 0".
25744 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25746         * tree-loop-distribution.c: Add general explanantion on the pass.
25747         (generate_loops_for_partition): Mark distributed loop.
25748         (pg_add_dependence_edges): New parameter.  Handle alias data
25749         dependence specially and record it in the parameter if asked.
25750         (struct pg_vdata, pg_edata, pg_edge_callback_data): New structs.
25751         (init_partition_graph_vertices, add_partition_graph_edge): New.
25752         (pg_skip_alias_edge, free_partition_graph_edata_cb): New.
25753         (free_partition_graph_vdata, build_partition_graph): New.
25754         (sort_partitions_by_post_order, merge_dep_scc_partitions): New.
25755         (pg_collect_alias_ddrs, break_alias_scc_partitions): New.
25756         (data_ref_segment_size, latch_dominated_by_data_ref): New.
25757         (compute_alias_check_pairs, version_loop_by_alias_check): New.
25758         (version_for_distribution_p, finalize_partitions): New.
25759         (distribute_loop): Handle alias data dependence specially.  Factor
25760         out loop fusion code as functions and call these functions.
25762 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25764         * tree-loop-distribution.c (classify_partition): New parameter and
25765         better handle reduction statement.
25766         (rdg_build_partitions): Revise comment.
25767         (distribute_loop): Compute statements in all partitions and pass it
25768         to classify_partition.
25770 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25772         * tree-loop-distribution.c (enum partition_type): New.
25773         (struct partition): New field type.
25774         (partition_merge_into): Add parameter.  Update partition type.
25775         (data_dep_in_cycle_p, update_type_for_merge): New functions.
25776         (build_rdg_partition_for_vertex): Compute partition type.
25777         (rdg_build_partitions): Dump partition type.
25778         (distribute_loop): Update calls to partition_merge_into.
25780 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25782         * tree-loop-distribution.c (struct ddr_hasher): New.
25783         (ddr_hasher::hash, ::equal, get_data_dependence): New function.
25784         (ddrs_table): New.
25785         (classify_partition): Call get_data_dependence.
25786         (pg_add_dependence_edges): Ditto.
25787         (distribute_loop): Release data dependence hash table.
25789 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25791         * tree-loop-distribution.c (ref_base_address): Delete.
25792         (similar_memory_accesses): Rename ...
25793         (share_memory_accesses): ... to this.  Check if partitions access
25794         the same memory reference.
25795         (distribute_loop): Call share_memory_accesses.
25797 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25799         * tree-loop-distribution.c (struct partition): New field recording
25800         its data reference.
25801         (partition_alloc, partition_free): Init and release data refs.
25802         (partition_merge_into): Merge data refs.
25803         (build_rdg_partition_for_vertex): Collect data refs for partition.
25804         (pg_add_dependence_edges): Change parameters from vector to bitmap.
25805         Update uses.
25806         (distribute_loop): Remve data refs from vertice data of partition
25807         graph.
25809 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25811         * tree-loop-distribution.c (params.h): Include header file.
25812         (MAX_DATAREFS_NUM, DR_INDEX): New macro.
25813         (datarefs_vec): New global var.
25814         (create_rdg_vertices): Use datarefs_vec directly.
25815         (free_rdg): Don't free data references.
25816         (build_rdg): Update use.  Don't free data references.
25817         (distribute_loop): Compute global variable for data references.
25818         Bail out if there are too many data references.
25820 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25822         * tree-loop-distribution.c (loop_nest): New global var.
25823         (build_rdg): Use loop directly, rather than loop nest.
25824         (pg_add_dependence_edges): Remove loop nest parameter.  Use global
25825         variable directly.
25826         (distribute_loop): Compute global variable loop nest.  Update use.
25828 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25830         * tree-loop-distribution.c (enum fuse_type, fuse_message): New.
25831         (partition_merge_into): New parameter.  Dump reason for fusion.
25832         (distribute_loop): Update use of partition_merge_into.
25834 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25836         * tree-loop-distribution.c (bb_top_order_index): New.
25837         (bb_top_order_index_size, bb_top_order_cmp): New.
25838         (stmts_from_loop): Use topological order.
25839         (pass_loop_distribution::execute): Compute and release topological
25840         order for basic blocks.
25842 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25844         * tree-loop-distribution.c (pass_loop_distribution::execute): Skip
25845         if no loops.
25847 2017-07-05  Bin Cheng  <bin.cheng@arm.com>
25849         * cfgloop.h (struct loop): Add comment.  New field orig_loop_num.
25850         * cfgloopmanip.c (lv_adjust_loop_entry_edge): Comment change.
25851         * internal-fn.c (expand_LOOP_DIST_ALIAS): New function.
25852         * internal-fn.def (LOOP_DIST_ALIAS): New.
25853         * tree-vectorizer.c (fold_loop_vectorized_call): Rename to ...
25854         (fold_loop_internal_call): ... this.
25855         (vect_loop_dist_alias_call): New function.
25856         (set_uid_loop_bbs): Call fold_loop_internal_call.
25857         (vectorize_loops): Fold IFN_LOOP_VECTORIZED and IFN_LOOP_DIST_ALIAS
25858         internal calls.
25860 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25862         PR target/81300
25863         * config/i386/i386.md (setcc + movzbl/and to xor + setcc peepholes):
25864         Require dead FLAGS_REG at the beginning of a peephole.
25866 2017-07-04  Uros Bizjak  <ubizjak@gmail.com>
25868         PR target/81294
25869         * config/i386/adxintrin.h (_subborrow_u32): Swap _X and _Y
25870         arguments in the call to __builtin_ia32_sbb_u32.
25871         (_subborrow_u64): Swap _X and _Y arguments in the call to
25872         __builtin_ia32_sbb_u64.
25874 2017-07-04  Jakub Jelinek  <jakub@redhat.com>
25876         PR debug/81278
25877         * tree-vrp.c (compare_assert_loc): Turn into a function template
25878         with stable template parameter.  Only test if a->e is NULL,
25879         !a->e == !b->e has been verified already.  Use e == NULL or
25880         e != NULL instead of e or ! e tests.  If stable is true, don't use
25881         iterative_hash_expr, on the other side allow a or b or both NULL
25882         and sort the NULLs last.
25883         (process_assert_insertions): Sort using compare_assert_loc<false>
25884         instead of compare_assert_loc, later sort using
25885         compare_assert_loc<true> before calling process_assert_insertions_for
25886         in a loop.  Use break instead of continue once seen NULL pointer.
25888 2017-07-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
25890         * config/arm/driver-arm.c (arm_cpu_table): Add entry for ARM
25891         Cortex-R7 and Cortex-R8 processors.
25893 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25895         * ipa-utils.c (ipa_merge_profiles): Fix merging when dst is
25896         uninitialized while src is not.
25898 2017-07-04  Richard Earnshaw  <rearnsha@arm.com>
25900         * common/config/arm/arm-common.c: Adjust include path for
25901         arm-cpu-cdata.h
25902         * config/arm/t-arm (TM_H): Adjust path for arm-cpu.h.
25903         (arm-cpu.h): Create in build directory.  Adjust dependency rules.
25904         (arm-cpu-data.h): Likewise.
25905         (arm-cpu-cdata.h): Likewise.
25906         * config/arm/arm-cpu.h: Delete.
25907         * config/arm/arm-cpu-cdata.h: Delete.
25908         * config/arm/arm-cpu-data.h: Delete.
25910 2017-07-04  James Greenhalgh  <james.greenhalgh@arm.com>
25912         * config/arm/arm-cpus.in (cortex-a55): New.
25913         (cortex-a75): Likewise.
25914         (cortex-a75.cortex-a55): Likewise.
25915         * config/arm/driver-arm.c (arm_cpu_table): Add cortex-a55 and
25916         cortex-a75.
25917         * doc/invoke.texi (-mcpu): Document cortex-a55 and cortex-a75.
25918         * config/arm/arm-cpu-cdata.h: Regenerate.
25919         * config/arm/arm-cpu-data.h: Regenerate.
25920         * config/arm/arm-cpu.h: Regenerate.
25921         * config/arm/arm-tables.opt: Regenerate.
25922         * config/arm/arm-tune.md: Regenerate.
25924 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25926         * haifa-sched.c (sched_create_recovery_edges): Update profile.
25928 2017-07-04  Jan Hubicka  <hubicka@ucw.cz>
25930         * bb-reorder.c (better_edge_p): Fix handling of uninitialized
25931         probability.
25933 2017-07-04  Richard Sandiford  <richard.sandiford@linaro.org>
25935         PR tree-optimization/81292
25936         * tree-ssa-strlen.c (handle_builtin_strlen): When setting
25937         full_string_p, also call adjust_related_strinfos if the adjustment
25938         is simple, otherwise invalidate related strinfos.
25940 2017-07-04  Martin Liska  <mliska@suse.cz>
25942         PR sanitizer/81040
25943         * sanopt.c (sanitize_rewrite_addressable_params): Mark the
25944         newly created variable as DECL_IGNORED_P.
25946 2017-07-04  Martin Liska  <mliska@suse.cz>
25948         PR ipa/81293
25949         * ipa-inline.c (inline_small_functions):
25950         Use xstrdup_for_dump.
25952 2017-07-04  Tom de Vries  <tom@codesourcery.com>
25954         * graph.c (draw_cfg_edges): Save and restore EDGE_DFS_BACK.
25956 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
25958         PR target/81033
25959         * config/darwin.c (darwin_function_switched_text_sections):
25960         Fix spaces.
25962 2017-07-03  Jan Hubicka  <hubicka@ucw.cz>
25964         * tree-vect-loop-manip.c (vect_do_peeling): Fix scaling up.
25966 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
25968         * doc/invoke.texi (ARM Options): Add -mbe8 and -mbe32 to option summary.
25970 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
25972         * tree-vect-loop.c (vect_analyze_loop_2): Treat min_scalar_loop_bound,
25973         min_profitable_iters, and th as inclusive lower bounds.
25974         Fix LOOP_VINFO_PEELING_FOR_GAPS condition.
25975         (vect_estimate_min_profitable_iters): Return inclusive lower bounds
25976         for min_profitable_iters and min_profitable_estimate.
25977         (vect_transform_loop): Treat th as an inclusive lower bound.
25978         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
25980 2017-07-03  Dominique d'Humieres  <dominiq@lps.ens.fr>
25982         PR target/81033
25983         * config/darwin.c (darwin_function_switched_text_sections):
25984         Replace DECL_NAME with DECL_ASSEMBLER_NAME, split assemble_name_raw
25985         in two pieces, and suppress the use of buf.
25987 2017-07-03  Nathan Sidwell  <nathan@acm.org>
25989         * hash-table.h (hash_table_mod1): Fix indentation.
25991 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
25993         PR middle-end/81290
25994         * predict.c (force_edge_cold): Be more careful about propagation
25995         backward.
25996         * profile-count.h (profile_probability::guessed,
25997         profile_probability::fdo, profile_count::guessed, profile_count::fdo):
25998         New.
25999         * tree-ssa-threadupdate.c (recompute_probabilities): Result is guessed.
26001 2017-07-03  James Greenhalgh  <james.greenhalgh@arm.com>
26003         * doc/invoke.texi (rcpc architecture extension): Document it.
26005 2017-07-03  Richard Biener  <rguenther@suse.de>
26007         PR tree-optimization/60510
26008         * tree-vect-loop.c (vect_create_epilog_for_reduction): Pass in
26009         the scalar reduction PHI and use it.
26010         (vectorizable_reduction): Properly guard the single_defuse_cycle
26011         path for non-SLP reduction chains where we cannot use it.
26012         Rework reduc_def/index and vector type deduction.  Rework
26013         vector operand gathering during reduction op code-gen.
26014         * tree-vect-slp.c (vect_analyze_slp): For failed SLP reduction
26015         chains dissolve the chain and leave it to non-SLP reduction
26016         handling.
26018 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26020         * tree-data-ref.h (dr_alignment): Declare.
26021         * tree-data-ref.c (dr_alignment): New function.
26022         * tree-vectorizer.h (dataref_aux): Remove base_element_aligned.
26023         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Don't
26024         set it.
26025         * tree-vect-stmts.c (vectorizable_store): Use dr_alignment.
26027 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26029         * tree-data-ref.h (innermost_loop_behavior): Add base_alignment
26030         and base_misalignment fields.
26031         (DR_BASE_ALIGNMENT, DR_BASE_MISALIGNMENT): New macros.
26032         * tree-data-ref.c: Include builtins.h.
26033         (dr_analyze_innermost): Set up the new innmost_loop_behavior fields.
26034         * tree-vectorizer.h (STMT_VINFO_DR_BASE_ALIGNMENT): New macro.
26035         (STMT_VINFO_DR_BASE_MISALIGNMENT): Likewise.
26036         * tree-vect-data-refs.c: Include tree-cfg.h.
26037         (vect_compute_data_ref_alignment): Use the new innermost_loop_behavior
26038         fields instead of calculating an alignment here.
26039         (vect_analyze_data_refs): Use dr_analyze_innermost.  Dump the new
26040         innermost_loop_behavior fields.
26042 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26044         * tree-data-ref.h (innermost_loop_behavior): Add a step_alignment
26045         field.
26046         (DR_STEP_ALIGNMENT): New macro.
26047         * tree-vectorizer.h (STMT_VINFO_DR_STEP_ALIGNMENT): Likewise.
26048         * tree-data-ref.c (dr_analyze_innermost): Initalize step_alignment.
26049         (create_data_ref): Print it.
26050         * tree-vect-stmts.c (vectorizable_load): Use the step alignment
26051         to tell whether the step preserves vector (mis)alignment.
26052         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26053         Move the check for an integer step and generalise to all INTEGER_CST.
26054         (vect_analyze_data_refs): Set DR_STEP_ALIGNMENT when setting DR_STEP.
26055         Print the outer step alignment.
26057 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26059         * tree-data-ref.h (innermost_loop_behavior): Replace aligned_to
26060         with offset_alignment.
26061         (DR_ALIGNED_TO): Delete.
26062         (DR_OFFSET_ALIGNMENT): New macro.
26063         * tree-vectorizer.h (STMT_VINFO_DR_ALIGNED_TO): Delete.
26064         (STMT_VINFO_DR_OFFSET_ALIGNMENT): New macro.
26065         * tree-data-ref.c (dr_analyze_innermost): Update after above changes.
26066         (create_data_ref): Likewise.
26067         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Likewise.
26068         (vect_analyze_data_refs): Likewise.
26069         * tree-if-conv.c (if_convertible_loop_p_1): Use memset before
26070         creating dummy innermost behavior.
26072 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26074         * tree-data-ref.h (dr_analyze_innermost): Replace the dr argument
26075         with a "innermost_loop_behavior *" and refeence tree.
26076         * tree-data-ref.c (dr_analyze_innermost): Likewise.
26077         (create_data_ref): Update call accordingly.
26078         * tree-predcom.c (find_looparound_phi): Likewise.
26080 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26082         * tree-vectorizer.h (_stmt_vec_info): Replace individual dr_*
26083         fields with dr_wrt_vec_loop.
26084         (STMT_VINFO_DR_BASE_ADDRESS, STMT_VINFO_DR_INIT, STMT_VINFO_DR_OFFSET)
26085         (STMT_VINFO_DR_STEP, STMT_VINFO_DR_ALIGNED_TO): Update accordingly.
26086         (STMT_VINFO_DR_WRT_VEC_LOOP): New macro.
26087         (vect_dr_behavior): New function.
26088         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26089         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Use
26090         vect_dr_behavior.  Use a step_preserves_misalignment_p boolean to
26091         track whether the step preserves the misalignment.
26092         (vect_create_addr_base_for_vector_ref): Remove loop parameter.
26093         Use vect_dr_behavior.
26094         (vect_setup_realignment): Update call accordingly.
26095         (vect_create_data_ref_ptr): Likewise.  Use vect_dr_behavior.
26096         * tree-vect-loop-manip.c (vect_gen_prolog_loop_niters): Update
26097         call to vect_create_addr_base_for_vector_ref.
26098         (vect_create_cond_for_align_checks): Likewise.
26099         * tree-vect-patterns.c (vect_recog_bool_pattern): Copy
26100         STMT_VINFO_DR_WRT_VEC_LOOP as a block.
26101         (vect_recog_mask_conversion_pattern): Likewise.
26102         * tree-vect-stmts.c (compare_step_with_zero): Use vect_dr_behavior.
26103         (new_stmt_vec_info): Remove redundant zeroing.
26105 2017-07-03  Richard Earnshaw  <rearnsha@arm.com>
26107         * common/config/arm/arm-common.c (arm_be8_option): New function.
26108         * config/arm/arm-isa.h (isa_feature): Add new feature bit isa_bit_be8.
26109         (ISA_ARMv6): Add isa_bit_be8.
26110         * config/arm/arm.h (arm_be8_option): Add prototype.
26111         (BE8_SPEC_FUNCTION): New define.
26112         (EXTRA_SPEC_FUNCTIONS): Add BE8_SPEC_FUNCTION.
26113         * config/arm/arm.opt (mbig-endian): Mark as Negative of mlittle-endian.
26114         (mlittle-endian): Similarly.
26115         (mbe8, mbe32): New options.
26116         * config/arm/bpabi.h (BE8_LINK_SPEC): Call arm_be8_option.
26117         * doc/invoke.texi (ARM Options): Document -mbe8 and -mbe32.
26119 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26121         * tree-cfg.c (gimple_find_sub_bbs): Fix profile updating.
26123 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26125         * tree-cfgcleanup.c (want_merge_blocks_p): New function.
26126         (cleanup_tree_cfg_bb): Use it.
26127         * profile-count.h (profile_count::of_for_merging, profile_count::merge):
26128         New functions.
26129         * tree-cfg.c (gimple_merge_blocks): Use profile_count::merge.
26131 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26133         PR bootstrap/81285
26134         * loop-doloop.c (add_test): Update profile.
26136 2017-07-03  Martin Liska  <mliska@suse.cz>
26138         PR sanitize/81040
26139         * sanopt.c (rewrite_usage_of_param): New function.
26140         (sanitize_rewrite_addressable_params): Likewise.
26141         (pass_sanopt::execute): Call rewrite_usage_of_param.
26143 2017-07-03  Richard Biener  <rguenther@suse.de>
26145         * tree-vect-loop.c (vect_create_epilog_for_reduction): Revert
26146         back to using VIEW_CONVERT_EXPR.
26148 2017-07-03  Martin Liska  <mliska@suse.cz>
26150         PR other/78366
26151         * doc/extend.texi: Document when a resolver function is
26152         generated for target_clones.
26154 2017-07-03  Martin Liska  <mliska@suse.cz>
26156         * asan.c (asan_emit_stack_protection): Unpoison just red zones
26157         and shadow memory of auto variables which are subject of
26158         use-after-scope sanitization.
26159         (asan_expand_mark_ifn): Add do set only when is_poison.
26161 2016-07-03  Richard Biener  <rguenther@suse.de>
26163         * tree-vect-loop.c (vect_analyze_loop_operations): Also analyze
26164         reduction PHIs.
26165         (vect_force_simple_reduction): Record reduction def -> phi mapping.
26166         (vectorizable_reduction): Perform reduction PHI creation when
26167         visiting a reduction PHI and adjust and simplify code generation
26168         phase of the reduction op.  Cache dts, use fold_binary, not fold_build2.
26169         (vect_transform_loop): Visit reduction PHIs.
26170         * tree-vect-slp.c (vect_get_and_check_slp_defs): Record reduction
26171         defs into the SLP tree.
26172         (vect_build_slp_tree): Reduction defs terminate the recursion.
26173         * tree-vect-stmts.c (vect_get_vec_def_for_operand_1): Allow lookup
26174         of reduction defs.
26175         (vect_get_vec_defs_for_stmt_copy): Export.
26176         (vect_get_vec_defs): Likewise.
26177         * tree-vectorizer.h (struct _stmt_vec_info): Amend reduc_def
26178         purpose.
26179         (vect_get_vec_defs_for_stmt_copy): Declare.
26180         (vect_get_vec_defs): Likewise.
26182 2017-07-03  Richard Sandiford  <richard.sandiford@linaro.org>
26184         * tree-data-ref.c (dr_analyze_innermost): Replace the "nest"
26185         parameter with a "loop" parameter and use it instead of the
26186         loop containing DR_STMT.  Don't check simple_iv when doing
26187         BB analysis.  Describe the two analysis modes in the comment.
26189 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26191         PR tree-optimization/69468
26192         * tree-ssa-tail-merge.c (ignore_edge_flags): New constant.
26193         (find_same_succ_bb): Handle ignore_edge_flags.
26195 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26197         PR tree-optimization/81192
26198         * tree-ssa-tail-merge.c (same_succ_hash): Use bb->loop_father->num in
26199         hash.
26200         (same_succ::equal): Don't find bbs to be equal if bb->loop_father
26201         differs.
26202         (find_same_succ_bb): Remove obsolete test on bb->loop_father->latch.
26204 2017-07-03  Tom de Vries  <tom@codesourcery.com>
26206         PR tree-optimization/81192
26207         * tree-ssa-tail-merge.c (same_succ_flush_bb): Handle
26208         BB_SAME_SUCC (bb) == NULL.
26210 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26212         * cfgrtl.c (rtl_verify_edges): Enable checking of profile_probability
26213         consistency.
26215 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26217         * dumpfile.c: Include profile-count.h
26218         * tree-cfg.c (gimple_duplicate_sese_tail): Drop UNUSED attributes;
26219         update profile.
26220         (insert_cond_bb): Update profile.
26221         * tree-cfg.h (insert_cond_bb): Update prototype.
26222         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Update.
26223         * tree-dump.c: Do not include tree-cfg.
26225 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26227         * bb-reorder.c (fix_up_crossing_landing_pad): Update profile.
26229 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26231         * expect.c (dw2_build_landing_pads): Update profile of the landing pad
26232         bb.
26234 2017-07-02  Jan Hubicka  <hubicka@ucw.cz>
26236         * tree-complex.c (expand_complex_div_wide): update profile.
26238 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26239             Alan Hayward  <alan.hayward@arm.com>
26240             David Sherwood  <david.sherwood@arm.com>
26242         * Makefile.in (MACHMODE_H): Remove insn-modes.h
26243         (CORETYPES_H): New define.
26244         (MOSTLYCLEANFILES): Add insn-modes-inline.h.
26245         (insn-modes-inline.h, s-modes-inline-h): New rules.
26246         (generated_files): Add insn-modes-inline.h.
26247         (RTL_BASE_H, TREE_CORE_H): Use CORETYPES_H instead of coretypes.h.
26248         (build/gensupport.o, build/ggc-none.o, build/print-rtl.o): Likewise.
26249         (build/read-md.o, build/read-rtl.o, build/rtl.o): Likewise.
26250         (build/vec.o, build/hash-table.o, build/inchash.o): Likewise.
26251         (build/gencondmd.o, build/genattr.o, build/genattr-common.o): Likewise.
26252         (build/genattrtab.o, build/genautomata.o, build/gencheck.o): Likewise.
26253         (build/gencodes.o, build/genconditions.o): Likewise.
26254         (build/genconfig.o, build/genconstants.o, build/genemit.o): Likewise.
26255         (build/genenums.o, build/genextract.o, build/genflags.o): Likewise.
26256         (build/gentarget-def.o, build/genmddeps.o, build/genopinit.o)
26257         (build/genoutput.o, build/genpeep.o, build/genpreds.o): Likewise.
26258         (build/genrecog.o, build/genmddump.o, build/genmatch.o): Likewise.
26259         (build/gencfn-macros.o, build/gcov-iov.o): Likewise.
26260         * coretypes.h: Include everything up to real.h for generators.
26261         Include insn-modes.h first.  Include wide-int-print.h after
26262         wide-int.h.  Include insn-modes-inline.h and then machmode.h.
26263         * machmode.h: Don't include insn-modes.h here.
26264         * function-tests.c: Remove includes of signop.h, machmode.h,
26265         double-int.h and wide-int.h.
26266         * rtl.h: Likewise.
26267         * gcc-rich-location.c: Remove includes of machmode.h, double-int.h
26268         and wide-int.h.
26269         * optc-save-gen.awk: Likewise.
26270         * gencheck.c (BITS_PER_UNIT): Delete dummy definition.
26271         * godump.c: Remove include of wide-int-print.h.
26272         * pretty-print.h: Likewise.
26273         * wide-int-print.cc: Likewise.
26274         * wide-int.cc: Likewise.
26275         * hash-map-tests.c: Remove include of signop.h.
26276         * hash-set-tests.c: Likewise.
26277         * rtl-tests.c: Likewise.
26278         * mkconfig.sh: Remove include of machmode.h.
26279         * genmodes.c (emit_insn_modes_h): Split emission of inline functions
26280         into...
26281         (emit_insn_modes_inline_h): ...this new function.  Emit the code
26282         into an insn-modes-inline.h header file, adding appropriate
26283         include guards and end comments.
26284         (emit_insn_modes_c_header): Remove include of machmode.h.
26285         (emit_min_insn_modes_c_header): Include coretypes.h rather than
26286         machmode.h.
26287         (main): Handle -i flag and call emit_insn_modes_inline_h when
26288         it is passed.
26290 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26292         * tree-ssa-strlen.c (strinfo): Rename the length field to
26293         nonzero_chars.  Add a full_string_p field.
26294         (compare_nonzero_chars, zero_length_string_p): New functions.
26295         (get_addr_stridx): Add an offset_out parameter.
26296         Use compare_nonzero_chars.
26297         (get_stridx): Update accordingly.  Use compare_nonzero_chars.
26298         (new_strinfo): Update after above changes to strinfo.
26299         (set_endptr_and_length): Set full_string_p.
26300         (get_string_length): Update after above changes to strinfo.
26301         (unshare_strinfo): Update call to new_strinfo.
26302         (maybe_invalidate): Likewise.
26303         (get_stridx_plus_constant): Change off to unsigned HOST_WIDE_INT.
26304         Use compare_nonzero_chars and zero_string_p.  Treat nonzero_chars
26305         as a uhwi instead of an shwi.  Update after above changes to
26306         strinfo and new_strinfo.
26307         (zero_length_string): Assert that chainsi contains full strings.
26308         Use zero_length_string_p.  Update call to new_strinfo.
26309         (adjust_related_strinfos): Update after above changes to strinfo.
26310         Copy full_string_p from origsi.
26311         (adjust_last_stmt): Use zero_length_string_p.
26312         (handle_builtin_strlen): Update after above changes to strinfo and
26313         new_strinfo.  Install the lhs as the string length if the previous
26314         entry didn't describe a full string.
26315         (handle_builtin_strchr): Update after above changes to strinfo
26316         and new_strinfo.
26317         (handle_builtin_strcpy): Likewise.
26318         (handle_builtin_strcat): Likewise.
26319         (handle_builtin_malloc): Likewise.
26320         (handle_pointer_plus): Likewise.
26321         (handle_builtin_memcpy): Likewise.  Track nonzero characters
26322         that aren't necessarily followed by a nul terminator.
26323         (handle_char_store): Likewise.
26325 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26327         PR tree-optimization/80769
26328         * tree-ssa-strlen.c (strinfo): Document that "stmt" is also used
26329         for malloc and calloc.  Document the new invariant that all related
26330         strinfos have delayed lengths or none do.
26331         (verify_related_strinfos): Move earlier in file.
26332         (set_endptr_and_length): New function, split out from...
26333         (get_string_length): ...here.  Also set the lengths of related
26334         strinfos.
26335         (zero_length_string): Assert that chainsi has known (rather than
26336         delayed) lengths.
26337         (adjust_related_strinfos): Likewise.
26339 2017-07-02  Richard Sandiford  <richard.sandiford@linaro.org>
26341         PR tree-optimization/81136
26342         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Only
26343         assert that two references with the same misalignment have the same
26344         compile-time misalignment if those compile-time misalignments
26345         are known.
26347 2017-07-01  Andi Kleen  <ak@linux.intel.com>
26349         * print-tree.c (print_node): Print all attributes.
26351 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26353         * cfg.c (scale_bbs_frequencies): New function.
26354         * cfg.h (scale_bbs_frequencies): Declare it.
26355         * cfgloopanal.c (single_likely_exit): Cleanup.
26356         * cfgloopmanip.c (scale_loop_frequencies): Take profile_probability
26357         as parameter.
26358         (scale_loop_profile): Likewise.
26359         (loop_version): Likewise.
26360         (create_empty_loop_on_edge): Update.
26361         * cfgloopmanip.h (scale_loop_frequencies, scale_loop_profile,
26362         scale_loop_frequencies, scale_loop_profile, loopify,
26363         loop_version): Update prototypes.
26364         * modulo-sched.c (sms_schedule): Update.
26365         * predict.c (unlikely_executed_edge_p): Also check probability.
26366         (probably_never_executed_edge_p): Fix typo.
26367         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26368         * tree-parloops.c (gen_parallel_loop): Update.
26369         * tree-ssa-loop-ivcanon.c (try_peel_loop): Update.
26370         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26371         * tree-ssa-loop-split.c (split_loop): Update.
26372         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26373         * tree-vect-loop-manip.c (vect_do_peeling): Update.
26374         (vect_loop_versioning): Update.
26375         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26377 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26379         * trans-mem.c (split_bb_make_tm_edge): Update profile.
26381 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26383         * tree-if-conv.c (combine_blocks): Use make_single_succ_edge
26384         to keep profile consistent.
26386 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26388         * cfgrtl.c (rtl_flow_call_edges_add): Update profile.
26389         * tree-cfg.c (gimple_flow_call_edges_add): Likewise.
26390         * profile-count.h (max_safe_multiplier): Make unsigned.
26391         (profile_count::guessed_zero): New.
26393 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26395         * bb-reorder.c (fix_up_crossing_landing_pad,
26396         fix_crossing_conditional_branches): Use make_single_succ_edge
26397         to keep profile consistent.
26399 2017-07-01  Jan Hubicka  <hubicka@ucw.cz>
26401         * tree-vect-loop.c (optimize_mask_stores): Use make_single_succ_edge
26402         to update profile.
26404 2017-07-01  Jakub Jelinek  <jakub@redhat.com>
26406         PR sanitizer/81262
26407         * bb-reorder.c (fix_up_fall_thru_edges): Move variable declarations to
26408         the right scopes, make sure cond_jump isn't preserved between multiple
26409         iterations.  Search for fallthru edge whenever there are 3+ edges and
26410         use find_fallthru_edge for it.
26412 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26414         Patch by Alexander Monakov <amonakov@ispras.ru>
26415         * sel-sched-ir.c (compute_succs_info): Handle uninitialized
26416         probabilities consistently.
26418 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26420         * pa.c (pa_expand_compare_and_swap_loop): Update call of
26421         emit_cmp_and_jump_insns.
26423 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26425         PR ipa/81261
26426         * tree-inline.c (expand_call_inline): Combine profile statuses.
26428 2017-06-30  Andrew Pinski  <apinski@cavium.com>
26430         * tree-if-conv.c (predicate_scalar_phi): Update new_stmt if
26431         fold_stmt returned true.
26433 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26435         * ggc.h (empty_string): Delete.
26436         * cfgexpand.c (expand_asm_stmt): Use plain "".
26437         * optabs.c (expand_asm_memory_barrier): Likewise.
26438         * stringpool.c (empty_string): Delete.
26439         (digit_vector, digit_string): Delete.
26440         (ggc_alloc_string): Use plain "", don't optimize single digit
26441         strings.  Use ggc_alloc_atomic.
26443 2017-06-30  Richard Earnshaw  <rearnsha@arm.com>
26445         * rtlanal.c (insn_rtx_cost): If a parallel contains exactly one
26446         comparison set and one other set, use the cost of the non-comparison
26447         set.
26449 2017-06-30  Nathan Sidwell  <nathan@acm.org>
26451         * ggc.h: Replace all 'static inline' with plain 'inline'.  Fix
26452         some formatting.
26454 2017-06-30  Peter Bergner  <bergner@vnet.ibm.com>
26456         * tree-cfg.c (group_case_labels_stmt): Merge scanning and compressing
26457         loops.  Remove now unneeded calls to gimple_switch_set_label() that
26458         just set removed labels to NULL_TREE.
26460 2017-06-30  Aldy Hernandez  <aldyh@redhat.com>
26462         * tree-ssanames.c (set_range_info_raw): Abstract from ...
26463         (set_range_info): ...here.  Only call set_range_info_raw if domain
26464         is useful.
26465         (set_nonzero_bits): Call set_range_info_raw.
26466         * tree-ssanames.h (set_range_info_raw): New.
26468 2017-06-30  Jakub Jelinek  <jakub@redhat.com>
26470         PR target/81225
26471         * config/i386/sse.md (vec_extract_lo_<mode><mask_name>): For V8FI,
26472         V16FI and VI8F_256 iterators, use <store_mask_predicate> instead
26473         of nonimmediate_operand and <store_mask_constraint> instead of m
26474         for the input operand.  For V8FI iterator, always split if input
26475         is a MEM.  For V16FI and V8SF_256 iterators, don't test if both
26476         operands are MEM if <mask_applied>.  For VI4F_256 iterator, use
26477         <store_mask_predicate> instead of register_operand and
26478         <store_mask_constraint> instead of v for the input operand.  Make
26479         sure both operands aren't MEMs for if not <mask_applied>.
26481 2017-06-30  Sylvestre Ledru  <sylvestre@debian.org>
26483         * lto-wrapper.c (copy_file) Close both file descriptors before
26484         exiting normally.
26486 2017-06-30  Martin Liska  <mliska@suse.cz>
26488         PR ipa/81214
26489         * multiple_target.c (create_dispatcher_calls): Make ifunc
26490         also for function that don't have calls or are not referenced.
26492 2017-06-30  Richard Biener  <rguenther@suse.de>
26494         * tree-vect-slp.c (vect_slp_analyze_node_operations): Only
26495         analyze the first scalar stmt.  Move vector type computation
26496         for the BB case here from ...
26497         * tree-vect-stmts.c (vect_analyze_stmt): ... here.  Guard
26498         live operation processing in the SLP case properly.
26500 2017-06-30  Richard Biener  <rguenther@suse.de>
26502         * graph.c (draw_cfg_node_succ_edges): Fix broken dot syntax.
26504 2017-06-30  Martin Liska  <mliska@suse.cz>
26506         PR sanitizer/81021
26507         * tree-eh.c (lower_resx): Call BUILT_IN_ASAN_HANDLE_NO_RETURN
26508         before BUILT_IN_UNWIND_RESUME when ASAN is used.
26510 2017-06-30  Yvan Roux  <yvan.roux@linaro.org>
26512         * doc/invoke.texi (AArch64): Add missing options and remove redundant
26513         ones.
26515 2017-06-30  Richard Biener  <rguenther@suse.de>
26517         PR tree-optimization/81249
26518         * tree-vect-loop.c (vect_create_epilog_for_reduction): Convert
26519         condition reduction result to original scalar type.
26521 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26523         * profile-count.h (enum profile_quality): Fix typos and whitespace
26524         issues.
26526 2017-06-30  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
26528         * config/s390/s390.c (s390_expand_setmem): Adjust to the new data
26529         type for branch probabilities.
26531 2017-06-29  Julian Brown  <julian@codesourcery.com>
26532             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26534         * config/aarch64/aarch64-fusion-pairs.def: Add ALU_BRANCH entry.
26535         * config/aarch64/aarch64.c (AARCH64_FUSE_ALU_BRANCH): New fusion type.
26536         (thunderx2t99_tunings): Set AARCH64_FUSE_ALU_BRANCH flag.
26537         (aarch_macro_fusion_pair_p): Add support for AARCH64_FUSE_ALU_BRANCH.
26539 2017-06-29  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
26541         * config/aarch64/aarch64.c (aarch_macro_fusion_pair_p): Push the
26542         check for CC usage into AARCH64_FUSE_CMP_BRANCH.
26543         * config/i386/i386.c (ix86_macro_fusion_pair_p): Push the check for
26544         CC usage from generic code to here.
26545         * sched-deps.c (sched_macro_fuse_insns): Move the condition for
26546         CC usage into the target macros.
26548 2017-06-29  Maya Rashish  <coypu@sdf.org>
26550         * config/netbsd.h (NETBSD_LIB_SPEC): Add -lc when creating shared
26551         objects.
26553 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26555         * arm/arm-builtins.c: Include profile-count.h
26556         * except.c (sjlj_emit_function_enter): Use
26557         profile_probability::unlikely.
26559 2017-06-29  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
26561         * config/rs6000/rs6000.c (toc_relative_expr_p): Make tocrel_base
26562         and tocrel_offset be pointer args rather than implicitly using
26563         static versions.
26564         (legitimate_constant_pool_address_p, rs6000_emit_move,
26565         const_load_sequence_p, adjust_vperm): Add local tocrel_base and
26566         tocrel_offset and use in toc_relative_expr_p call.
26567         (print_operand, print_operand_address): Use static tocrel_base_oac
26568         and tocrel_offset_oac.
26569         (rs6000_output_addr_const_extra): Use static tocrel_base_oac and
26570         tocrel_offset_oac.
26572 2017-06-29  Maya Rashish  <coypu@sdf.org>
26574         * config/vax/builtins.md (ffssi2_internal): Correct constraint.
26576 2017-06-29  Eric Botcazou  <ebotcazou@adacore.com>
26578         * expr.c (expand_expr) <normal_inner_ref>: When testing for unaligned
26579         objects, take into account only the alignment of 'op0' and 'mode1' if
26580         'op0' is a MEM.
26582 2017-06-29  Steve Ellcey  <sellcey@cavium.com>
26584         * ccmp.c (ccmp_tree_comparison_p): New function.
26585         (ccmp_candidate_p): Update to use above function.
26586         (get_compare_parts): New function.
26587         (expand_ccmp_next): Update to use new functions.
26588         (expand_ccmp_expr_1): Take tree arg instead of gimple, update to use
26589         new functions.
26590         (expand_ccmp_expr): Pass tree instead of gimple to expand_ccmp_expr_1,
26591         take mode as argument.
26592         * ccmp.h (expand_ccmp_expr): Add mode as argument.
26593         * expr.c (expand_expr_real_1): Pass mode as argument.
26595 2017-06-29  Segher Boessenkool  <segher@kernel.crashing.org>
26597         * combine.c (combine_instructions): Print insns to dump_file, together
26598         with their costs.
26600 2017-06-29  Jan Hubicka  <hubicka@ucw.cz>
26602         * asan.c (asan_emit_stack_protection): Update.
26603         (create_cond_insert_point): Update.
26604         * auto-profile.c (afdo_propagate_circuit): Update.
26605         * basic-block.h (struct edge_def): Turn probability to
26606         profile_probability.
26607         (EDGE_FREQUENCY): Update.
26608         * bb-reorder.c (find_traces_1_round): Update.
26609         (better_edge_p): Update.
26610         (sanitize_hot_paths): Update.
26611         * cfg.c (unchecked_make_edge): Initialize probability to uninitialized.
26612         (make_single_succ_edge): Update.
26613         (check_bb_profile): Update.
26614         (dump_edge_info): Update.
26615         (update_bb_profile_for_threading): Update.
26616         * cfganal.c (connect_infinite_loops_to_exit): Initialize new edge
26617         probabilitycount to 0.
26618         * cfgbuild.c (compute_outgoing_frequencies): Update.
26619         * cfgcleanup.c (try_forward_edges): Update.
26620         (outgoing_edges_match): Update.
26621         (try_crossjump_to_edge): Update.
26622         * cfgexpand.c (expand_gimple_cond): Update make_single_succ_edge.
26623         (expand_gimple_tailcall): Update.
26624         (construct_init_block): Use make_single_succ_edge.
26625         (construct_exit_block): Use make_single_succ_edge.
26626         * cfghooks.c (verify_flow_info): Update.
26627         (redirect_edge_succ_nodup): Update.
26628         (split_edge): Update.
26629         (account_profile_record): Update.
26630         * cfgloopanal.c (single_likely_exit): Update.
26631         * cfgloopmanip.c (scale_loop_profile): Update.
26632         (set_zero_probability): Remove.
26633         (duplicate_loop_to_header_edge): Update.
26634         * cfgloopmanip.h (loop_version): Update prototype.
26635         * cfgrtl.c (try_redirect_by_replacing_jump): Update.
26636         (force_nonfallthru_and_redirect): Update.
26637         (update_br_prob_note): Update.
26638         (rtl_verify_edges): Update.
26639         (purge_dead_edges): Update.
26640         (rtl_lv_add_condition_to_bb): Update.
26641         * cgraph.c: (cgraph_edge::redirect_call_stmt_to_calle): Update.
26642         * cgraphunit.c (init_lowered_empty_function): Update.
26643         (cgraph_node::expand_thunk): Update.
26644         * cilk-common.c: Include profile-count.h
26645         * dojump.c (inv): Remove.
26646         (jumpifnot): Update.
26647         (jumpifnot_1): Update.
26648         (do_jump_1): Update.
26649         (do_jump): Update.
26650         (do_jump_by_parts_greater_rtx): Update.
26651         (do_compare_rtx_and_jump): Update.
26652         * dojump.h (jumpifnot, jumpifnot_1, jumpif_1, jumpif, do_jump,
26653         do_jump_1. do_compare_rtx_and_jump): Update prototype.
26654         * dwarf2cfi.c: Include profile-count.h
26655         * except.c (dw2_build_landing_pads): Use make_single_succ_edge.
26656         (sjlj_emit_dispatch_table): Likewise.
26657         * explow.c: Include profile-count.h
26658         * expmed.c (emit_store_flag_force): Update.
26659         (do_cmp_and_jump): Update.
26660         * expr.c (compare_by_pieces_d::generate): Update.
26661         (compare_by_pieces_d::finish_mode): Update.
26662         (emit_block_move_via_loop): Update.
26663         (store_expr_with_bounds): Update.
26664         (store_constructor): Update.
26665         (expand_expr_real_2): Update.
26666         (expand_expr_real_1): Update.
26667         * expr.h (try_casesi, try_tablejump): Update prototypes.
26668         * gimple-pretty-print.c (dump_probability): Update.
26669         (dump_profile): New.
26670         (dump_gimple_label): Update.
26671         (dump_gimple_bb_header): Update.
26672         * graph.c (draw_cfg_node_succ_edges): Update.
26673         * hsa-gen.c (convert_switch_statements): Update.
26674         * ifcvt.c (cheap_bb_rtx_cost_p): Update.
26675         (find_if_case_1): Update.
26676         (find_if_case_2): Update.
26677         * internal-fn.c (expand_arith_overflow_result_store): Update.
26678         (expand_addsub_overflow): Update.
26679         (expand_neg_overflow): Update.
26680         (expand_mul_overflow): Update.
26681         (expand_vector_ubsan_overflow): Update.
26682         * ipa-cp.c (good_cloning_opportunity_p): Update.
26683         * ipa-split.c (split_function): Use make_single_succ_edge.
26684         * ipa-utils.c (ipa_merge_profiles): Update.
26685         * loop-doloop.c (add_test): Update.
26686         (doloop_modify): Update.
26687         * loop-unroll.c (compare_and_jump_seq): Update.
26688         (unroll_loop_runtime_iterations): Update.
26689         * lra-constraints.c (lra_inheritance): Update.
26690         * lto-streamer-in.c (input_cfg): Update.
26691         * lto-streamer-out.c (output_cfg): Update.
26692         * mcf.c (adjust_cfg_counts): Update.
26693         * modulo-sched.c (sms_schedule): Update.
26694         * omp-expand.c (expand_omp_for_init_counts): Update.
26695         (extract_omp_for_update_vars): Update.
26696         (expand_omp_ordered_sink): Update.
26697         (expand_omp_for_ordered_loops): Update.
26698         (expand_omp_for_generic): Update.
26699         (expand_omp_for_static_nochunk): Update.
26700         (expand_omp_for_static_chunk): Update.
26701         (expand_cilk_for): Update.
26702         (expand_omp_simd): Update.
26703         (expand_omp_taskloop_for_outer): Update.
26704         (expand_omp_taskloop_for_inner): Update.
26705         * omp-simd-clone.c (simd_clone_adjust): Update.
26706         * optabs.c (expand_doubleword_shift): Update.
26707         (expand_abs): Update.
26708         (emit_cmp_and_jump_insn_1): Update.
26709         (expand_compare_and_swap_loop): Update.
26710         * optabs.h (emit_cmp_and_jump_insns): Update prototype.
26711         * predict.c (predictable_edge_p): Update.
26712         (edge_probability_reliable_p): Update.
26713         (set_even_probabilities): Update.
26714         (combine_predictions_for_insn): Update.
26715         (combine_predictions_for_bb): Update.
26716         (propagate_freq): Update.
26717         (estimate_bb_frequencies): Update.
26718         (force_edge_cold): Update.
26719         * profile-count.c (profile_count::dump): Add missing space into dump.
26720         (profile_count::debug): Add newline.
26721         (profile_count::differs_from_p): Explicitly convert to unsigned.
26722         (profile_count::stream_in): Update.
26723         (profile_probability::dump): New member function.
26724         (profile_probability::debug): New member function.
26725         (profile_probability::differs_from_p): New member function.
26726         (profile_probability::differs_lot_from_p): New member function.
26727         (profile_probability::stream_in): New member function.
26728         (profile_probability::stream_out): New member function.
26729         * profile-count.h (profile_count_quality): Rename to ...
26730         (profile_quality): ... this one.
26731         (profile_probability): New.
26732         (profile_count): Update.
26733         * profile.c (compute_branch_probabilities): Update.
26734         * recog.c (peep2_attempt): Update.
26735         * sched-ebb.c (schedule_ebbs): Update.
26736         * sched-rgn.c (find_single_block_region): Update.
26737         (compute_dom_prob_ps): Update.
26738         (schedule_region): Update.
26739         * sel-sched-ir.c (compute_succs_info): Update.
26740         * stmt.c (struct case_node): Update.
26741         (do_jump_if_equal): Update.
26742         (get_outgoing_edge_probs): Update.
26743         (conditional_probability): Update.
26744         (emit_case_dispatch_table): Update.
26745         (expand_case): Update.
26746         (expand_sjlj_dispatch_table): Update.
26747         (emit_case_nodes): Update.
26748         * targhooks.c: Update.
26749         * tracer.c (better_p): Update.
26750         (find_best_successor): Update.
26751         * trans-mem.c (expand_transaction): Update.
26752         * tree-call-cdce.c: Update.
26753         * tree-cfg.c (gimple_split_edge): Upate.
26754         (move_sese_region_to_fn): Upate.
26755         * tree-cfgcleanup.c (cleanup_control_expr_graph): Upate.
26756         * tree-eh.c (lower_resx): Upate.
26757         (cleanup_empty_eh_move_lp): Upate.
26758         * tree-if-conv.c (version_loop_for_if_conversion): Update.
26759         * tree-inline.c (copy_edges_for_bb): Update.
26760         (copy_cfg_body): Update.
26761         * tree-parloops.c (gen_parallel_loop): Update.
26762         * tree-profile.c (gimple_gen_ic_func_profiler): Update.
26763         (gimple_gen_time_profiler): Update.
26764         * tree-ssa-dce.c (remove_dead_stmt): Update.
26765         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
26766         * tree-ssa-loop-im.c (execute_sm_if_changed): Update.
26767         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update.
26768         (unloop_loops): Update.
26769         (try_peel_loop): Update.
26770         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
26771         * tree-ssa-loop-split.c (connect_loops): Update.
26772         (split_loop): Update.
26773         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Update.
26774         (hoist_guard): Update.
26775         * tree-ssa-phionlycprop.c (propagate_rhs_into_lhs): Update.
26776         * tree-ssa-phiopt.c (replace_phi_edge_with_variable): Update.
26777         (value_replacement): Update.
26778         * tree-ssa-reassoc.c (branch_fixup): Update.
26779         * tree-ssa-tail-merge.c (replace_block_by): Update.
26780         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges): Update.
26781         (create_edge_and_update_destination_phis): Update.
26782         (compute_path_counts): Update.
26783         (recompute_probabilities): Update.
26784         (update_joiner_offpath_counts): Update.
26785         (freqs_to_counts_path): Update.
26786         (duplicate_thread_path): Update.
26787         * tree-switch-conversion.c (hoist_edge_and_branch_if_true): Update.
26788         (struct switch_conv_info): Update.
26789         (gen_inbound_check): Update.
26790         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
26791         (vect_do_peeling): Update.
26792         (vect_loop_versioning): Update.
26793         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
26794         (optimize_mask_stores): Update.
26795         * ubsan.c (ubsan_expand_null_ifn): Update.
26796         * value-prof.c (gimple_divmod_fixed_value): Update.
26797         (gimple_divmod_fixed_value_transform): Update.
26798         (gimple_mod_pow2): Update.
26799         (gimple_mod_pow2_value_transform): Update.
26800         (gimple_mod_subtract): Update.
26801         (gimple_mod_subtract_transform): Update.
26802         (gimple_ic): Update.
26803         (gimple_stringop_fixed_value): Update.
26804         (gimple_stringops_transform): Update.
26805         * value-prof.h: Update.
26807 2017-06-29  Carl Love  <cel@us.ibm.com>
26809         * config/rs6000/rs6000-c.c: Add support for built-in functions
26810         vector signed int vec_signed (vector float);
26811         vector signed long long vec_signed (vector double);
26812         vector signed int vec_signed2 (vector double, vector double);
26813         vector signed int vec_signede (vector double);
26814         vector signed int vec_signedo (vector double);
26815         * config/rs6000/rs6000.c (rs6000_generate_vsigned2_code): Add
26816         instruction generator.
26817         * config/rs6000/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26818         UNSPEC_VSX_VSIGNED2): Add UNSPECS.
26819         (vsx_xvcvspsxws, vsx_xvcvdpuxds_scale, vsx_xvcvspuxws, vsigned2_v2df):
26820         Add define_insn.
26821         (vsignedo_v2df, vsignede_v2df, vunsigned2_v2df, vunsignedo_v2df,
26822         vunsignede_v2df): Add define_expands.
26823         * config/rs6000/rs6000-builtin.def (VEC_SIGNED, VEC_UNSIGNED,
26824         VEC_SIGNED2, VEC_UNSIGNED2, VEC_SIGNEDE, VEC_UNSIGNEDE, VEC_SIGNEDO,
26825         VEC_UNSIGNEDO): Add definitions.
26826         * config/vsx.md (UNSPEC_VSX_XVCVSPSXWS, UNSPEC_VSX_XVCVSPSXDS,
26827         UNSPEC_VSX_VSIGNED2): Add UNSPECs.
26828         (vsx_xvcvspsxws, vsx_xvcvspuxws): Add define_insn.
26829         (vsigned2_v2df, vsigendo_v2df, vsignede_v2df,
26830         vunsigned2_v2df, vunsignedo_v2df, vunsignede_v2df): Add define_expands.
26831         * config/rs6000/altivec.h (vec_signed, vec_signed2,
26832         vec_signede and vec_signedo, vec_unsigned, vec_unsigned2,
26833         vec_unsignede, vec_unsignedo): Add builtin defines.
26834         * config/rs6000-protos.h (rs6000_generate_vsigned2_code): Add extern
26835         declaration.
26836         * doc/extend.texi: Update the built-in documentation file for the
26837         new built-in functions.
26839 2017-06-29  Richard Biener  <rguenther@suse.de>
26841         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Do not add
26842         reduction chains to LOOP_VINFO_REDUCTIONS.
26843         * tree-vect-slp.c (vect_analyze_slp): Continue looking for
26844         SLP reductions after processing reduction chains.
26846 2017-06-29  Nathan Sidwell  <nathan@acm.org>
26848         * builtins.c (fold_builtin_FUNCTION): Use
26849         lang_hooks.decl_printable_name.
26851 2017-06-29  Peter Bergner  <bergner@vnet.ibm.com>
26853         PR middle-end/81194
26854         * cfgexpand.c (expand_gimple_stmt_1): Handle switch statements
26855         with only one label.
26856         * stmt.c (expand_case): Assert NCASES is greater than one.
26858 2017-06-29  Richard Biener  <rguenther@suse.de>
26860         * tree-cfg.c (group_case_labels_stmt): Return whether we changed
26861         anything.
26862         (group_case_labels): Likewise.
26863         (find_taken_edge): Push sanity checking on val to workers...
26864         (find_taken_edge_cond_expr): ... here
26865         (find_taken_edge_switch_expr): ... and here, handle cases
26866         with just a default label.
26867         * tree-cfg.h (group_case_labels_stmt): Adjust prototype.
26868         (group_case_labels): Likewise.
26869         * tree-cfgcleanup.c (execute_cleanup_cfg_post_optimizing): When
26870         group_case_labels does anything cleanup the CFG again.
26872 2017-06-29  Bin Cheng  <bin.cheng@arm.com>
26874         PR tree-optimization/81196
26875         * tree-ssa-loop-niter.c (number_of_iterations_cond): Handle loop
26876         exit condition comparing two IVs.
26878 2017-06-29  Richard Earnshaw  <rearnsha@arm.com>
26880         * config/arm/parsecpu.awk (gen_comm_data): Add initializer for
26881         profile to the dummy entry at the end of the list of architectures.
26882         * config/arm/arm-cpu-cdata.h: Regenerated.
26884 2017-06-29  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
26885             Michael Collison <michael.collison@arm.com>
26887         PR target/70119
26888         * config/aarch64/aarch64.md (*aarch64_<optab>_reg_<mode>3_mask1):
26889         New pattern.
26890         (*aarch64_reg_<mode>3_neg_mask2): New pattern.
26891         (*aarch64_reg_<mode>3_minus_mask): New pattern.
26892         (*aarch64_<optab>_reg_di3_mask2): New pattern.
26893         * config/aarch64/aarch64.c (aarch64_rtx_costs): Account for cost
26894         of shift when the shift amount is masked with constant equal to
26895         the size of the mode.
26896         * config/aarch64/predicates.md (subreg_lowpart_operator): New
26897         predicate.
26899 2017-06-29  Martin Liska  <mliska@suse.cz>
26901         * config/i386/i386.opt: Change range from [1,5] to [0,5].
26903 2017-06-29  Yury Gribov  <tetra2005@gmail.com>
26905         PR bootstrap/80565
26906         * ipa-cp.c (allocate_and_init_ipcp_value): Add initialization
26907         code.
26908         * ipa-inline.h
26909         (edge_growth_cache_entry::edge_growth_cache_entry): New
26910         function.
26911         (reset_edge_growth_cache): Update to use constructor.
26913 2017-06-28  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
26915         * config/aarch64/aarch64.h (AARCH64_EXPAND_ALIGNMENT): New.
26916         (DATA_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26917         (LOCAL_ALIGNMENT): Update to use AARCH64_EXPAND_ALIGNMENT.
26919 2017-06-28  Sebastian Peryt  <sebastian.peryt@intel.com>
26921         * config/i386/avx512vlintrin.h (_mm256_permutexvar_epi64)
26922         (_mm256_permutexvar_epi32, _mm256_permutex_epi64): New intrinsics.
26924 2017-06-28  Szabolcs Nagy  <szabolcs.nagy@arm.com>
26926         * config.gcc (*-linux-musl*): Add t-musl tmake_file.
26927         (*-linux-uclibc*): Add t-uclibc tmake_file.
26928         * config/t-musl: New.
26929         * config/t-uclibc: New.
26931 2017-06-28  Richard Earnshaw  <rearnsha@arm.com>
26933         * config/arm/parsecpu.awk (profile): Parse new keyword in an arch
26934         context.
26935         (gen_comm_data): Emit architectural setting of arch_prof.
26936         * config/arm/arm-cpus.in (armv6-m, armv6s-m, armv7-a, armv7ve): Set the
26937         profile.
26938         (armv7-r, armv7-m, armv7e-m, armv8-a, armv8.1-a, armv8.2-a): Likewise.
26939         (armv8-m.base, armv8-m.main): Likewise.
26940         * arm-protos.h (arm_build_target): Add profile field.
26941         (arch_option): Likewise.
26942         * config/arm/arm.c (arm_configure_build_target): Copy the profile to
26943         the active target.
26944         * config/arm/arm.h (TARGET_ARM_ARCH_PROFILE): Use
26945         arm_active_target.profile.
26947 2017-06-28  Richard Biener  <rguenther@suse.de>
26949         PR middle-end/81227
26950         * fold-const.c (negate_expr_p): Use TYPE_UNSIGNED, not
26951         TYPE_OVERFLOW_WRAPS.
26952         * match.pd (negate_expr_p): Likewise.
26953         * tree-ssa-reassoc.c (optimize_range_tests_diff): Use
26954         fold_build2, not fold_binary.
26956 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26958         * config/aarch64/aarch64 (aarch64_expand_mov_immediate):
26959         Convert memory address to Pmode.
26960         (aarch64_print_operand): Assert MEM operands are always Pmode.
26962 2017-06-28  Wilco Dijkstra  <wdijkstr@arm.com>
26964         PR target/79665
26965         * config/arm/aarch-common.c (arm_no_early_alu_shift_dep):
26966         Remove redundant if.
26967         (aarch_forward_to_shift_is_not_shifted_reg): Remove.
26968         * config/arm/aarch-common-protos.h
26969         (aarch_forward_to_shift_is_not_shifted_re): Remove.
26970         * config/arm/cortex-a53.md: Use arm_no_early_alu_shift_dep in bypass.
26972 2017-06-28  Michael Meissner  <meissner@linux.vnet.ibm.com>
26974         PR ipa/81238
26975         * multiple_target.c (create_dispatcher_calls): Set the default
26976         clone to be static, not public.
26978 2017-06-28  Richard Biener  <rguenther@suse.de>
26980         * tree-vect-loop.c (vectorizable_reduction): Move special
26981         cond reduction IV var creation ...
26982         (vect_create_epilog_for_reduction): ... here.  Remove induction_index
26983         parameter.  Use STMT_VINFO_VECTYPE.
26984         * tree-vect-slp.c (vect_get_constant_vectors): Properly reset
26985         constant_p.
26987 2017-06-28  Martin Liska  <mliska@suse.cz>
26989         PR ipa/81128
26990         * ipa-visibility.c (non_local_p): Handle visibility.
26992 2017-06-28  Martin Liska  <mliska@suse.cz>
26994         PR driver/79659
26995         * common.opt: Add IntegerRange to various options.
26996         * opt-functions.awk (integer_range_info): New function.
26997         * optc-gen.awk: Add integer_range_info to cl_options struct.
26998         * opts-common.c (decode_cmdline_option): Handle
26999         CL_ERR_INT_RANGE_ARG.
27000         (cmdline_handle_error): Likewise.
27001         * opts.c (print_filtered_help): Show valid interval in
27002         when --help is provided.
27003         * opts.h (struct cl_option): Add range_min and range_max fields.
27004         * config/i386/i386.opt: Add IntegerRange for -mbranch-cost.
27006 2017-06-28  Marc Glisse  <marc.glisse@inria.fr>
27008         * match.pd ((X & ~Y) | (~X & Y)): Generalize to + and ^.
27009         (x * C EQ/NE y * C): New transformation.
27011 2017-06-28  Christophe Lyon  <christophe.lyon@linaro.org>
27013         * genmultilib (combination_space): Accept '+' in option names.
27015 2017-06-28  Martin Liska  <mliska@suse.cz>
27017         PR sanitizer/81224
27018         * asan.c (instrument_derefs): Bail out inner references
27019         that are hard register variables.
27021 2017-06-28  Jakub Jelinek  <jakub@redhat.com>
27023         PR target/81175
27024         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Use def_builtin
27025         rather than def_builtin_pure for __builtin_ia32_gatherpf*.
27027 2017-06-28  Richard Biener  <rguenther@suse.de>
27029         * tree-vectorizer.h (vect_get_vec_defs): Remove.
27030         (vect_get_slp_defs): Adjust.
27031         * tree-vect-loop.c (get_initial_defs_for_reduction): Split
27032         out from ...
27033         * tree-vect-slp.c (vect_get_constant_vectors): ... here and
27034         simplify.
27035         * tree-vect-loop.c (vect_create_epilog_for_reduction): Use
27036         get_initial_defs_for_reduction instead of vect_get_vec_defs.
27037         (vectorizable_reduction): Adjust.
27038         * tree-vect-slp.c (vect_get_constant_vectors): Remove reduction
27039         handling.
27040         (vect_get_slp_defs): Likewise.
27041         * tree-vect-stmts.c (vect_get_vec_defs): Make static and adjust.
27042         (vectorizable_bswap): Adjust.
27043         (vectorizable_call): Likewise.
27044         (vectorizable_conversion): Likewise.
27045         (vectorizable_assignment): Likewise.
27046         (vectorizable_shift): Likewise.
27047         (vectorizable_operation): Likewise.
27048         (vectorizable_store): Likewise.
27049         (vectorizable_condition): Likewise.
27050         (vectorizable_comparison): Likewise.
27052 2017-06-28  Michael Collison  <michael.collison@arm.com>
27054         PR target/68535
27055         * config/arm/arm.c (gen_ldm_seq): Remove last unnecessary
27056         set of base_reg
27057         (arm_gen_movmemqi): Removed unused variable 'i'.
27058         Convert 'for' loop into 'while' loop.
27059         (arm_expand_prologue): Remove last unnecessary set of insn.
27060         (thumb_pop): Remove unused variable 'pushed_words'.
27061         (thumb_exit): Remove last unnecessary set of regs_to_pop.
27063 2017-06-28  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
27065         * config/s390/predicates.md: Use s390_rel_address_ok_p.
27066         * config/s390/s390-protos.h: Add prototype of
27067         s390_rel_address_ok_p.
27068         * config/s390/s390.c (s390_got_symbol): New function.
27069         (s390_rel_address_ok_p): New function.
27070         (legitimize_pic_address): Use s390_rel_address_ok_p.
27071         (s390_load_got): Use s390_got_symbol.
27072         (s390_option_override): Issue error if
27073         -mno-pic-data-is-text-relative is used without -fpic/-fPIC.
27074         * config/s390/s390.h (TARGET_DEFAULT_PIC_DATA_IS_TEXT_RELATIVE):
27075         New macro.
27076         * config/s390/s390.opt: New option mpic-data-is-text-relative.
27078 2017-06-27  Andrew Pinski  <apinski@cavium.com>
27080         * match.pd (X >/>=/</<= 0 ? 1.0 : -1.0): New patterns.
27081         (X * copysign (1.0, X)): New pattern.
27082         (X * copysign (1.0, -X)): New pattern.
27083         (copysign (-1.0, CST)): New pattern.
27085 2017-06-27  Joseph Myers  <joseph@codesourcery.com>
27087         * genmultilib (combination_space): Remove variable.
27088         Validate reuse rules against regular expression for any sequence
27089         of multilib options in any order.
27091 2017-06-27  Michael Collison  <michael.collison@arm.com>
27093         * config/aarch64/aarch64-simd.md (aarch64_combine<mode>): Directly
27094         call aarch64_split_simd_combine.
27095         * (aarch64_combine_internal<mode>): Delete pattern.
27096         * config/aarch64/aarch64.c (aarch64_split_simd_combine):
27097         Allow register and subreg operands.
27099 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27101         * config/i386/vxworks.h (ASM_SPEC): Remove definition. No target
27102         specific need, just fallback on defaults.
27103         (ASM_OUTPUT_ALIGNED_BSS): Add #undef before #define.
27105 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27106             Olivier Hainque  <hainque@adacore.com>
27108         * config/i386/vxworks.h (DBX_REGISTER_NUMBER): Pick distinct
27109         map for 64bits.
27110         (TARGET_OS_CPP_BUILTINS): builtin_define CPU to X86_64 for 64bit
27111         targets. Pick a default if no particular attempt applied.
27112         (STACK_CHECK_PROTECT): Double for 64bit targets, which have
27113         larger contexts.
27115 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27117         * config.gcc (i*86-wrs-vxworks7): Handle new acceptable triplet.
27118         (x86_64-wrs-vxworks7): Likewise.
27120 2017-06-27  Marek Polacek  <polacek@redhat.com>
27122         PR sanitizer/81223
27123         * ubsan.c (instrument_null): Check get_base_address's result for null.
27125 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27127         * match.pd ((A+-B)+(C-A), (A+B)-(A-C)): New transformations.
27129 2017-06-27  Marc Glisse  <marc.glisse@inria.fr>
27131         * builtin-types.def (BT_FENV_T_PTR, BT_CONST_FENV_T_PTR,
27132         BT_FEXCEPT_T_PTR, BT_CONST_FEXCEPT_T_PTR): New primitive types.
27133         (BT_FN_INT_FENV_T_PTR, BT_FN_INT_CONST_FENV_T_PTR,
27134         BT_FN_INT_FEXCEPT_T_PTR_INT, BT_FN_INT_CONST_FEXCEPT_T_PTR_INT):
27135         New function types.
27136         * builtins.def (BUILT_IN_FECLEAREXCEPT, BUILT_IN_FEGETENV,
27137         BUILT_IN_FEGETEXCEPTFLAG, BUILT_IN_FEGETROUND,
27138         BUILT_IN_FEHOLDEXCEPT, BUILT_IN_FERAISEEXCEPT,
27139         BUILT_IN_FESETENV, BUILT_IN_FESETEXCEPTFLAG,
27140         BUILT_IN_FESETROUND, BUILT_IN_FETESTEXCEPT,
27141         BUILT_IN_FEUPDATEENV): New builtins.
27142         * tree-core.h (TI_FENV_T_PTR_TYPE, TI_CONST_FENV_T_PTR_TYPE,
27143         TI_FEXCEPT_T_PTR_TYPE, TI_CONST_FEXCEPT_T_PTR_TYPE): New entries.
27144         * tree.h (fenv_t_ptr_type_node, const_fenv_t_ptr_type_node,
27145         fexcept_t_ptr_type_node, const_fexcept_t_ptr_type_node): New
27146         macros.
27147         (builtin_structptr_types): Adjust size.
27148         * tree.c (builtin_structptr_types): Add four entries.
27150 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27151             Olivier Hainque  <hainque@adacore.com>
27153         * config/vxworks.h (VXWORKS_LIB_SPEC): Incorporate ...
27154         (TLS_SYM): New local macro, forcing reference to __tls__ on
27155         link command lines for VxWorks 7 RTPs, triggering initialization
27156         of tlsLib.
27157         (VXWORKS_HAVE_TLS): New macro. State whether the target VxWorks
27158         OS features TLS support, true for RTPs on VxWorks 7.
27159         * config/vxworks.c (vxworks_override_options): Setup emutls
27160         accordingly.
27162 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27164         * predict.c (test_prediction_value_range): Use -1U instead of -1
27165         to avoid narrowing conversion warning.
27166         * dumpfile.c (dump_options): Wrap all value into dump_flags_t cast
27167         to avoid narrowing conversion warning.
27168         * opt-functions.awk (var_ref): Return (unsigned short) -1 instead of
27169         -1.
27170         * optc-gen.awk (END): Expect (unsigned short) -1 instead of -1.
27172 2017-06-27  Jerome Lambourg  <lambourg@adacore.com>
27174         * config/vxworks.h (VXWORKS_LIBS_RTP): Alternative definition for
27175         64bit configurations.
27176         (PTR_DIFF_TYPE): Alternative definition for TARGET_LP64.
27177         (SIZE_TYPE): Likewise.
27178         * config/vxworks.c (vxworks_emutls_var_fields): Use
27179         long_unsigned_type_node instead of unsigned_type_node as the offset
27180         field type, which is "pointer" mode in emutls.c.
27182 2017-06-27  Jakub Jelinek  <jakub@redhat.com>
27184         PR sanitizer/81209
27185         * ubsan.c (ubsan_encode_value): Initialize DECL_CONTEXT on var.
27187         PR middle-end/81207
27188         * gimple-fold.c (replace_call_with_call_and_fold): Handle
27189         gimple_vuse copying separately from gimple_vdef copying.
27191 2017-06-27  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
27193         * value-prof.c (free_hist): Remove call to memset and the enclosing if
27194         condition.
27196 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27197             Olivier Hainque  <hainque@adacore.com>
27199         * config.gcc (*-*-vxworks*): Add TARGET_VXWORKS7=1 to tm_defines
27200         for all vxworks7 targets.
27201         * config/vxworks.h (TARGET_VXWORKS7): If not defined, define to 0.
27202         (VXWORKS_ADDITIONAL_CPP_SPEC): Alternative definition for VXWORKS7.
27203         (VXWORKS_LIBS_RTP, VXWORKS_LIBS_RTP_DIR): New macros, allowing
27204         variations for VX6/VX7 and 32/64bits later on in ...
27205         (VXWORKS_LIB_SPEC): Leverage new macros.
27206         (VXWORKS_OS_CPP_BUILTINS): Define _VSB_CONFIG_FILE for VXWORKS7,
27207         as well as _ALLOW_KEYWORD_MACROS when "inline" is not a keyword.
27209 2017-06-26  Jerome Lambourg  <lambourg@adacore.com>
27211         * config/vxworks.h (VXWORKS_OS_CPP_BUILTINS): builtin_define
27212         _VX_TOOL_FAMILY and _VX_TOOL to gnu.
27214 2017-06-26  Carl Love  <cel@us.ibm.com>
27216         * config/rs6000/rs6000-c.c: Add support for built-in functions
27217         vector bool char vec_reve (vector bool char);
27218         vector signed char vec_reve (vector signed char);
27219         vector unsigned char vec_reve (vector unsigned char);
27220         vector bool int vec_reve (vector bool int);
27221         vector signed int vec_reve (vector signed int);
27222         vector unsigned int vec_reve (vector unsigned int);
27223         vector bool long long vec_reve (vector bool long long);
27224         vector signed long long vec_reve (vector signed long long);
27225         vector unsigned long long vec_reve (vector unsigned long long);
27226         vector bool short vec_reve (vector bool short);
27227         vector signed short vec_reve (vector signed short);
27228         vector double vec_reve (vector double);
27229         vector float vec_reve (vector float);
27230         * config/rs6000/rs6000-builtin.def (VREVE_V2DI, VREVE_V4SI,
27231         VREVE_V8HI, VREVE_V16QI, VREVE_V2DF, VREVE_V4SF, VREVE): New builtin.
27232         * config/rs6000/altivec.md (UNSPEC_VREVEV): New UNSPEC.
27233         (altivec_vreve): New pattern.
27234         * config/rs6000/altivec.h (vec_reve): New define.
27235         * doc/extend.texi (vec_rev): Update the built-in documentation file
27236         for the new built-in functions.
27238 2016-06-26  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
27240         PR tree-optimization/71815
27241         * gimple-ssa-strength-reduction.c (uses_consumed_by_stmt): New
27242         function.
27243         (find_basis_for_candidate): Call uses_consumed_by_stmt rather than
27244         has_single_use.
27245         (slsr_process_phi): Likewise.
27246         (replace_uncond_cands_and_profitable_phis): Don't replace a
27247         multiply candidate with a stride of 1 (copy or cast).
27248         (phi_incr_cost): Call uses_consumed_by_stmt rather than
27249         has_single_use.
27250         (lowest_cost_path): Likewise.
27251         (total_savings): Likewise.
27253 2017-06-26  Richard Biener  <rguenther@suse.de>
27255         PR target/81175
27256         * config/i386/i386.c (ix86_init_mmx_sse_builtins):
27257         Use def_builtin_pure for all gather builtins.
27259 2017-06-26  Richard Biener  <rguenther@suse.de>
27261         PR tree-optimization/81203
27262         * tree-tailcall.c (find_tail_calls): Do not move stmts into
27263         non-dominating BBs.
27265 2017-06-26  Marek Polacek  <polacek@redhat.com>
27267         PR c/80116
27268         * doc/invoke.texi: Document -Wmultistatement-macros.
27270 2017-06-26  Christophe Lyon  <christophe.lyon@linaro.org>
27272         * doc/sourcebuild.texi (ARM-specific attributes): Document new
27273         arm_neon_ok_no_float_abi effective target.
27275 2017-06-26  Richard Biener  <rguenther@suse.de>
27277         PR tree-optimization/80928
27278         * cfghooks.c (duplicate_block): Do not copy BB_DUPLICATED flag.
27279         (copy_bbs): Set BB_DUPLICATED flag early.
27280         (execute_on_growing_pred): Do not execute for BB_DUPLICATED
27281         marked blocks.
27282         (execute_on_shrinking_pred): Likewise.
27283         * tree-ssa.c (ssa_redirect_edge): Do not look for PHI args in
27284         BB_DUPLICATED blocks.
27285         * tree-ssa-phionlycoprop.c (eliminate_degenerate_phis_1): Properly
27286         iterate over all PHIs considering removal of *gsi.
27288 2017-06-23  Jim Wilson  <jim.wilson@linaro.org>
27290         * doc/invoke.texi (AArch64 Options, -mtune): Re-add falkor and
27291         qdf24xx.
27293 2017-06-23  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
27295         * config/rs6000/rs6000-string.c: (expand_block_clear,
27296         do_load_for_compare, select_block_compare_mode,
27297         compute_current_alignment, expand_block_compare,
27298         expand_strncmp_align_check, expand_strn_compare,
27299         expand_block_move, rs6000_output_load_multiple)
27300         Move functions related to string/block move/compare
27301         to a separate file.
27302         * config/rs6000/rs6000.c: Move above functions to rs6000-string.c.
27303         * config/rs6000/rs6000-protos.h (rs6000_emit_dot_insn): Add prototype
27304         for this function which is now used in two files.
27305         * config/rs6000/t-rs6000: Add rule to compile rs6000-string.o.
27306         * config.gcc: Add rs6000-string.o to extra_objs for
27307         targets powerpc*-*-* and rs6000*-*-*.
27309 2017-06-23  Michael Meissner  <meissner@linux.vnet.ibm.com>
27311         PR target/80510
27312         * config/rs6000/rs6000.md (ALTIVEC_DFORM): Do not allow DImode in
27313         32-bit, since indexed is not valid for DImode.
27314         (mov<mode>_hardfloat32): Reorder ISA 2.07 load/stores before ISA
27315         3.0 d-form load/stores to be the same as mov<mode>_hardfloat64.
27316         (define_peephole2 for Altivec d-form load): Add 32-bit support.
27317         (define_peephole2 for Altivec d-form store): Likewise.
27319         PR ipa/81185
27320         * multiple_target.c (create_dispatcher_calls): Only create the
27321         dispatcher call if the function is the default clone of a
27322         versioned function.
27324 2017-06-23  Segher Boessenkool  <segher@kernel.crashing.org>
27326         PR middle-end/80902
27327         * builtins.c (expand_builtin_atomic_fetch_op): If emitting code after
27328         a call, force the call to not be a tail call.
27330 2017-06-23  Jeff Law  <law@redhat.com>
27332         * doc/contrib.texi: Add entry for Steven Pemberton's work on
27333         enquire.
27335 2017-06-23  Will Schmidt  <will_schmidt@vnet.ibm.com>
27337         * config/rs6000/rs6000.c: Add include of ssa-propagate.h for
27338         update_call_from_tree().  (rs6000_gimple_fold_builtin): Add
27339         handling for early expansion of vector shifts (sl,sr,sra,rl).
27340         (builtin_function_type): Add vector shift right instructions
27341         to the unsigned argument list.
27343 2017-06-23  Bernd Edlinger  <bernd.edlinger@hotmail.de>
27345         rtl-optimizatoin/79286
27346         * ira.c (update_equiv_regs): Revert to using may_trap_or_fault_p again.
27347         * rtlanal.c (rtx_addr_can_trap_p_1): SYMBOL_REF_FUNCTION_P can never
27348         trap.  PIC register plus a const unspec without offset can never trap.
27350 2017-06-23  Marc Glisse  <marc.glisse@inria.fr>
27352         * tree.h (builtin_structptr_type): New type.
27353         (builtin_structptr_types): Declare new array.
27354         * tree.c (builtin_structptr_types): New array.
27355         (free_lang_data, build_common_tree_nodes): Use it.
27357 2017-06-23  Jonathan Wakely  <jwakely@redhat.com>
27359         PR c++/81187
27360         * doc/invoke.texi (-Wnoexcept-type): Fix name of option, from
27361         -Wnoexcept.
27363 2017-06-22  Matt Turner  <mattst88@gmail.com>
27365         * config/i386/driver-i386.c (host_detect_local_cpu): Add Kaby
27366         Lake models to skylake case.  Assume skylake for unknown
27367         models with clflushopt.
27369 2017-06-22  Jeff Law  <law@redhat.com>
27371         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Handle
27372         frame sizes that do not satisfy aarch64_uimm12_shift.
27374 2017-06-22  Jan Hubicka <hubicka@ucw.cz>
27376         * profile-count.h (apply_probability,
27377         apply_scale, probability_in): Fix checks for zero.
27379 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27381         * incpath.c (add_sysroot_to_chain): Allow for $SYSROOT prefix.
27382         * doc/cppdiropts.texi (-I @var{dir}): Document it.
27384 2016-06-22  Richard Biener  <rguenther@suse.de>
27386         * tree-vect-loop.c (vect_model_reduction_cost): Handle
27387         COND_REDUCTION and INTEGER_INDUC_COND_REDUCTION without
27388         REDUC_MAX_EXPR support.
27389         (vectorizable_reduction): Likewise.
27390         (vect_create_epilog_for_reduction): Likewise.
27392 2017-06-22  James Greenhalgh  <james.greenhalgh@arm.com>
27394         * match.pd (A / (1 << B) -> A >> B): New.
27395         * generic-match-head.c: Include optabs-tree.h.
27396         * gimple-match-head.c: Likewise.
27397         * optabs-tree.h (target_supports_op_p): New.
27398         * optabs-tree.c (target_supports_op_p): New.
27400 2017-06-22  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
27402         * configure.ac (gcc_cv_ld_static_dynamic): Also check stderr for
27403         $gcc_cv_ld --help output.
27404         (gcc_cv_ld_demangle): Likewise.
27405         (gcc_cv_ld_eh_frame_hdr): Likewise.
27406         (gcc_cv_ld_pie): Likewise.
27407         (gcc_cv_ld_as_needed): Likewise.  Prefer native forms unless $gnu_ld.
27408         (gcc_cv_ld_buildid): Likewise.
27409         (gcc_cv_ld_sysroot): Likewise.
27410         (ld_bndplt_support): Likewise.
27411         (ld_pushpopstate_support): Likewise.
27412         * configure: Regenerate.
27413         * config/sol2.h [!USE_GLD] (SYSROOT_SPEC): Define.
27415 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27417         PR target/81151
27418         * config/i386/sse.md (round<mode>2): Renumber match_dup and
27419         operands indexes to avoid gap between operands and match_dups.
27421 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27423         * config/aarch64/aarch64-cost-tables.h (thunderx_extra_costs):
27424         Increment Arith_shift and Arith_shift_reg by 1.
27425         * config/aarch64/aarch64-tuning-flags.def (cheap_shift_extend):
27426         New tuning flag.
27427         * config/aarch64/aarch64.c (thunderx_tunings): Enable
27428         AARCH64_EXTRA_TUNE_CHEAP_SHIFT_EXTEND.
27429         (aarch64_strip_extend): Add new argument and test for it.
27430         (aarch64_cheap_mult_shift_p): New function.
27431         (aarch64_rtx_mult_cost): Call aarch64_cheap_mult_shift_p and don't
27432         add a cost if it is true.
27433         Update calls to aarch64_strip_extend.
27434         (aarch64_rtx_costs): Update calls to aarch64_strip_extend.
27436 2017-06-21  Andrew Pinski  <apinski@cavium.com>
27438         * config/aarch64/aarch64-cores.def (thunderxt88p1): Use thunderxt88
27439         tunings.
27440         (thunderxt88): Likewise.
27441         * config/aarch64/aarch64.c (thunderxt88_prefetch_tune): New variable.
27442         (thunderx_prefetch_tune): New variable.
27443         (thunderx2t99_prefetch_tune): Update for the correct values.
27444         (thunderxt88_tunings): New variable.
27445         (thunderx_tunings): Use thunderx_prefetch_tune instead of
27446         generic_prefetch_tune.
27447         (thunderx2t99_tunings): Use AUTOPREFETCHER_WEAK.
27449 2017-06-21  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
27451         * config/aarch64/atomics.md (aarch64_compare_and_swap<mode>_lse,
27452         SHORT): Relax operand 3 to aarch64_reg_or_zero and constraint to Z.
27453         (aarch64_compare_and_swap<mode>_lse, GPI): Likewise.
27454         (aarch64_atomic_cas<mode>, SHORT): Likewise for operand 2.
27455         (aarch64_atomic_cas<mode>, GPI): Likewise.
27457 2017-06-21  Martin Liska  <mliska@suse.cz>
27459         * gimplify.c (gimplify_label_expr): Insert GIMPLE_PREDICT
27460         statements on cold and hot labels.
27461         * predict.c (tree_estimate_probability_bb): Remove the
27462         prediction from this place.
27464 2017-06-21  Martin Liska  <mliska@suse.cz>
27466         PR tree-optimization/79489
27467         * gimplify.c (maybe_add_early_return_predict_stmt): New
27468         function.
27469         (gimplify_return_expr): Call the function.
27470         * predict.c (tree_estimate_probability_bb): Remove handling
27471         of early return.
27472         * predict.def: Update comment about early return predictor.
27473         * gimple-predict.h (is_gimple_predict): New function.
27474         * predict.def: Change default value of early return to 66.
27475         * tree-tailcall.c (find_tail_calls): Skip GIMPLE_PREDICT
27476         statements.
27477         * passes.def: Put pass_strip_predict_hints to the beginning of
27478         IPA passes.
27480 2017-06-21  Pierre-Marie de Rodat  <derodat@adacore.com>
27482         * dwarf2out.c (gen_decl_die): Remove the guard to skip file-scope
27483         FUNCTION_DECL declarations.
27484         (dwarf2out_early_global_decl): Remove the guard to skip FUNCTION_DECL
27485         declarations.
27486         (dwaf2out_decl): Likewise.
27487         * godump.c (go_early_global_decl): Skip call to the real debug hook
27488         for FUNCTION_DECL declarations.
27489         * passes.c (rest_of_decl_compilation): Skip call to the
27490         early_global_decl debug hook for FUNCTION_DECL declarations, unless
27491         -fdump-go-spec is passed.
27493 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27495         * config/i386/i386.c (struct builtin_isa): New field pure_p.
27496         Reorder for compactness.
27497         (def_builtin, def_builtin2, ix86_add_new_builtins): Handle pure_p.
27498         (def_builtin_pure, def_builtin_pure2): New functions.
27499         (ix86_init_mmx_sse_builtins) [__builtin_ia32_stmxcsr]: Mark as pure.
27501 2017-06-21  Marc Glisse  <marc.glisse@inria.fr>
27503         * match.pd (nop_convert): New predicate.
27504         ((A +- CST1) +- CST2): Allow some NOP conversions.
27506 2017-06-21  Jakub Jelinek  <jakub@redhat.com>
27508         PR c++/81130
27509         * gimplify.c (omp_add_variable): Don't force GOVD_SEEN for types
27510         with ctors/dtors if GOVD_SHARED is set.
27512 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27514         * config/aarch64/aarch64.md (movti_aarch64):
27515         Emit mov rather than orr.
27516         (movtf_aarch64): Likewise.
27517         * config/aarch64/aarch64-simd.md (aarch64_simd_mov):
27518         Emit mov rather than orr.
27520 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27522         * config/aarch64/aarch64-simd.md (aarch64_simd_dup):
27523         Swap alternatives, make integer dup more expensive.
27525 2017-06-21  Wilco Dijkstra  <wdijkstr@arm.com>
27527         * config/aarch64/aarch64.c (aarch64_legitimate_constant_p):
27528         Return true for non-tls symbols.
27530 2017-06-21  James Greenhalgh  <james.greenhalgh@arm.com>
27532         * config/aarch64/aarch64-cores.def (cortex-a55): New.
27533         (cortex-a75): Likewise.
27534         (cortex-a75.cortex-a55): Likewise.
27535         * config/aarch64/aarch64-tune.md: Regenerate.
27536         * doc/invoke.texi (-mtune): Document new values for -mtune.
27538 2017-06-21  Tom de Vries  <tom@codesourcery.com>
27540         * doc/sourcebuild.texi (Add Options, Features for dg-add-options): Add
27541         stack_size feature.
27542         (Effective-Target Keywords, Other attributes): Suggest using
27543         dg-add-options stack_size feature to get stack limit in stack_size
27544         effective target documentation.
27546 2017-06-21  Julian Brown  <julian@codesourcery.com>
27547             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
27549         * config/aarch64/aarch64-simd.md (aarch64_crypto_pmulldi)
27550         (aarch64_crypto_pmullv2di): Change type attribute to crypto_pmull.
27551         * config/aarch64/thunderx2t99.md (thunderx2t99_pmull): New
27552         reservation.
27553         * config/arm/cortex-a53.md (cortex_a53_advsimd_type): Add crypto_pmull to
27554         attribute type list for neon_multiply.
27555         * config/arm/cortex-a57.md (cortex_a57_neon_type): Add crypto_pmull to
27556         attribute type list for neon_multiply.
27557         * config/arm/crypto.md (crypto_vmullp64): Change type to crypto_pmull.
27558         * config/arm/exynos-m1.md (exynos_m1_neon_type): Add crypto_pmull to
27559         attribute type list for neon_multiply.
27560         * config/arm/types.md (crypto_pmull): Add.
27561         * config/arm/xgene1.md (xgene1_neon_pmull): Add crypto_pmull to
27562         attribute type list.
27564 2017-06-20  Andreas Tobler  <andreast@gcc.gnu.org>
27566         * config.gcc (armv6*-*-freebsd*): Change the target_cpu_cname to
27567         arm1176jzf-s.
27569 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27571         * ira-costs.c (find_costs_and_classes): Initialize cost_classes later
27572         to make sure not to dereference a NULL cost_classes_ptr pointer.
27574 2017-06-20  Carl Love  <cel@us.ibm.com>
27576         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27577         ALTIVEC_BUILTIN_VMULESW, ALTIVEC_BUILTIN_VMULEUW,
27578         ALTIVEC_BUILTIN_VMULOSW, ALTIVEC_BUILTIN_VMULOUW entries.
27579         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin,
27580         builtin_function_type): Add ALTIVEC_BUILTIN_* case statements.
27581         * config/rs6000/altivec.md (MVULEUW, VMULESW, VMULOUW,
27582         VMULOSW): New enum "unspec" values.
27583         (vec_widen_umult_even_v4si, vec_widen_smult_even_v4si,
27584         vec_widen_umult_odd_v4si, vec_widen_smult_odd_v4si,
27585         altivec_vmuleuw, altivec_vmulesw, altivec_vmulouw,
27586         altivec_vmulosw): New patterns.
27587         * config/rs6000/rs6000-builtin.def (VMLEUW, VMULESW, VMULOUW,
27588         VMULOSW): Add definitions.
27590 2017-06-20  Julia Koval  <julia.koval@intel.com>
27592         * config/i386/i386.c: Fix rounding expand for new pattern.
27593         * config/i386/subst.md: Fix pattern (parallel -> unspec).
27595 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27597         * config/aarch64/aarch64-option-extensions.def (rcpc): New.
27598         * config/aarch64/aarch64.h (AARCH64_FL_RCPC): New.
27600 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27602         * config/aarch64/aarch64-option-extensions.def (fp16): Fix expected
27603         feature string.
27605 2017-06-20  James Greenhalgh  <james.greenhalgh@arm.com>
27607         * config/aarch64/aarch64-cores.def: Rearrange to sort by
27608         architecture, then by implementer ID.
27609         * config/aarch64/aarch64-tune.md: Regenerate.
27611 2017-06-20  Richard Biener  <rguenther@suse.de>
27613         PR middle-end/81097
27614         * fold-const.c (split_tree): Fold to type before negating.
27616 2017-06-20  David Malcolm  <dmalcolm@redhat.com>
27618         * diagnostic-show-locus.c
27619         (selftest::test_fixit_deletion_affecting_newline): New function.
27620         (selftest::diagnostic_show_locus_c_tests): Call it.
27622 2017-06-20  Andreas Schwab  <schwab@suse.de>
27624         PR target/80970
27625         * config/m68k/m68k.md (bsetdreg, bchgdreg, bclrdreg): Use "=d"
27626         instead of "+d".
27628 2017-06-20  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
27630         * config/arm/arm-c.c (arm_cpu_builtins): New block to define
27631         __ARM_FEATURE_COPROC according to support.
27633 2017-06-20  Jakub Jelinek  <jakub@redhat.com>
27635         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
27636         Rewritten to avoid overflow for > 32-bit pointers.
27638         PR sanitizer/81125
27639         * ubsan.h (ubsan_encode_value): Workaround buggy clang++ parser
27640         by removing enum keyword.
27641         (ubsan_type_descriptor): Likewise.  Formatting fix.
27643         PR target/81121
27644         * config/i386/i386.md (TARGET_USE_VECTOR_CONVERTS float si->{sf,df}
27645         splitter): Require TARGET_SSE2 in the condition.
27647 2017-06-20  Michael Meissner  <meissner@linux.vnet.ibm.com>
27649         PR target/79799
27650         * config/rs6000/rs6000.c (rs6000_expand_vector_init): Add support
27651         for doing vector set of SFmode on ISA 3.0.
27652         * config/rs6000/vsx.md (vsx_set_v4sf_p9): Likewise.
27653         (vsx_set_v4sf_p9_zero): Special case setting 0.0f to a V4SF
27654         element.
27655         (vsx_insert_extract_v4sf_p9): Add an optimization for inserting a
27656         SFmode value into a V4SF variable that was extracted from another
27657         V4SF variable without converting the element to double precision
27658         and back to single precision vector format.
27659         (vsx_insert_extract_v4sf_p9_2): Likewise.
27661 2017-06-19  Jakub Jelinek  <jakub@redhat.com>
27663         * tree-ssa-structalias.c (get_constraint_for_ptr_offset): Multiply
27664         in UWHI to avoid undefined overflow.
27666         PR sanitizer/81125
27667         * ubsan.h (enum ubsan_encode_value_phase): New.
27668         (ubsan_encode_value): Change second argument to
27669         enum ubsan_encode_value_phase with default value of
27670         UBSAN_ENCODE_VALUE_GENERIC.
27671         * ubsan.c (ubsan_encode_value): Change second argument to
27672         enum ubsan_encode_value_phase PHASE from bool IN_EXPAND_P,
27673         adjust uses, for UBSAN_ENCODE_VALUE_GENERIC use just
27674         create_tmp_var_raw instead of create_tmp_var and use a
27675         TARGET_EXPR.
27676         (ubsan_expand_bounds_ifn, ubsan_build_overflow_builtin,
27677         instrument_bool_enum_load, ubsan_instrument_float_cast): Adjust
27678         ubsan_encode_value callers.
27680         PR sanitizer/81111
27681         * ubsan.c (ubsan_encode_value): If current_function_decl is NULL,
27682         use create_tmp_var_raw instead of create_tmp_var, mark it addressable
27683         just by setting TREE_ADDRESSABLE on the result and use a TARGET_EXPR.
27685 2017-06-19  Richard Biener  <rguenther@suse.de>
27687         PR middle-end/81118
27688         * tree-cfgcleanup.c (cleanup_tree_cfg_noloop): Clear niter
27689         estimates if we changed anything.
27691 2017-06-19  Richard Biener  <rguenther@suse.de>
27693         PR tree-optimization/80887
27694         * tree-ssa-sccvn.c (mprts_hook_cnt): New global.
27695         (vn_lookup_simplify_result): Allow only mprts_hook_cnt succesful
27696         simplified lookups, then reset mprts_hook.
27697         (vn_nary_build_or_lookup_1): Set mprts_hook_cnt to 9 before
27698         simplifying.
27699         (try_to_simplify): Likewise.
27701 2017-06-19  Martin Liska  <mliska@suse.cz>
27703         PR sanitizer/80879
27704         * gimplify.c (gimplify_switch_expr):
27705         Initialize live_switch_vars for SWITCH_BODY == STATEMENT_LIST.
27707 2017-06-19  Martin Liska  <mliska@suse.cz>
27709         * doc/install.texi: Document that PGO runs in 4 stages.
27711 2017-06-19  Martin Liska  <mliska@suse.cz>
27713         PR ipa/80732
27714         * attribs.c (make_dispatcher_decl): Do not append '.ifunc'
27715         to dispatcher function name.
27716         * multiple_target.c (replace_function_decl): New function.
27717         (create_dispatcher_calls): Redirect both edges and references.
27719 2017-06-19  Jan Hubicka <hubicka@ucw.cz>
27721         * profile-count.c (profile_count::dump): Dump quality.
27722         (profile_count::differs_from_p): Update for unsigned val.
27723         * profile-count.h (profile_count_quality): New enum.
27724         (profile_count): Turn m_val to 62bit unsigned, add quality tracking.
27726 2017-06-19  Richard Biener  <rguenther@suse.de>
27728         * tree-ssa-loop-niter.h (estimate_numbers_of_iterations): Take
27729         struct function as arg.
27730         (estimate_numbers_of_iterations): Export overload with loop arg.
27731         (free_numbers_of_iterations_estimates_loop): Use an overload of
27732         free_numbers_of_iterations_estimates instead.
27733         * tree-cfg.c (remove_bb): Adjust.
27734         * tree-cfgcleanup.c (remove_forwarder_block_with_phi): Likewise.
27735         * tree-parloops.c (gen_parallel_loop): Likewise.
27736         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables):
27737         Likewise.
27738         (tree_unroll_loops_completely): Likewise.
27739         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop):
27740         Use an overload instead and export.
27741         (estimated_loop_iterations): Adjust.
27742         (max_loop_iterations): Likewise.
27743         (likely_max_loop_iterations): Likewise.
27744         (estimate_numbers_of_iterations): Take struct function as arg
27745         and adjust.
27746         (loop_exits_before_overflow): Adjust.
27747         (free_numbers_of_iterations_estimates_loop): Use an overload.
27748         * tree-vect-loop.c (vect_analyze_loop_form): Adjust.
27749         * tree-vectorizer.c (vect_free_loop_info_assumptions): Likewise.
27751 2017-06-19  Richard Biener  <rguenther@suse.de>
27753         PR ipa/81112
27754         * ipa-prop.c (find_constructor_constant_at_offset): Handle
27755         RANGE_EXPR conservatively.
27757 2017-06-16  Carl Love  <cel@us.ibm.com>
27759         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
27760         definitions for vec_float, vec_float2, vec_floato,
27761         vec_floate built-ins.
27762         * config/rs6000/vsx.md (define_c_enum "unspec"): Add RTL code
27763         for instructions vsx_xvcvsxws vsx_xvcvuxwsp, float2, floato and
27764         floate.
27765         * config/rs6000/rs6000-builtin.def (FLOAT2_V2DI, FLOATE_V2DF,
27766         FLOATE_2DI, FLOATO_V2DF, FLOATEE_V2DI, XVCVSXWSP_V4SF,
27767         UNS_FLOATO_V2DI, UNS_FLOATE_V2DI): Add definitions.
27768         * config/altivec.md (define_insn "p8_vmrgew_<mode>",
27769         define_mode_attr VF_sxddp): Add V4SF type to p8_vmrgew.
27770         * config/rs6000/altivec.h (vec_float, vec_float2, vec_floate,
27771         vec_floato): Add builtin defines.
27772         * doc/extend.texi (vec_float, vec_float2, vec_floate, vec_floato):
27773         Update the built-in documentation file for the new built-in
27774         functions.
27776 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27778         * config/arm/arm.opt (marm): Mark as the negative of of -mthumb.
27779         (mthumb): Mark as the negative of -marm.
27781 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27783         * doc/invoke.texi (ARM Options, -mcpu): Document supported
27784         extension options.
27785         (ARM Options, -mtune): Document that this accepts the same
27786         extension options as -mcpu.
27787         (ARM Options, -mfpu): Document addition of -mfpu=auto.
27789 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27791         * doc/invoke.texi (ARM Options, -march=): Document new syntax and
27792         permitted extensions.
27794 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27796         * config/arm/arm-cpus.in (armv7): Add extension +nofp.
27797         (armv7-r): Add aliases vfpv3xd and vfpv3-d16.
27798         (armv8-m.main): Add option +nodsp.
27799         * config/arm/arm-cpu-cdata.h: Regenerated.
27801 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27803         * config/arm/t-fuchsia: New file.
27804         * config.gcc (arm*-*-fuchsia*): Use it.
27806 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27808         * config/arm/t-symbian: Rewrite for new option infrastructure.
27810 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27812         * config/arm/t-phoenix (MULTILIB_REUSE): Clear variable.
27813         (MULTILIB_REQUIRED): Likewise.
27815 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27817         * config/arm/t-linux-eabi (MULTILIB_EXCEPTIONS): Set to empty.
27818         (MULTILIB_RESUE): Likewise.
27819         (MULTILIB_MATCHES): Likewise.
27820         (MULTLIB_REQUIRED): Likewise.
27822 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27824         * config/arm/t-rtems: Rewrite for new option framework.
27826 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27828         * config/arm/t-aprofile (v7_a_nosimd_variants, v7_a_simd_variants)
27829         (v7ve_nosimd_variatns, v7ve_vfpv3_simd_variants)
27830         (v7ve_vfpv4_simd_variants, v8_a_nosimd_variants, v8_a_simd_variants)
27831         (v8_1_a_simd_variants, v8_2_a_simd_variants): Move to ...
27832         * config/arm/t-multilib: ... here.
27833         (MULTILIB_OPTIONS): Add armv7 and armv7+fp architectures.
27834         (MULTILIB_MATCHES): Use armv7 libraries for armv7-r.  Also use for
27835         armv7-a and armv8*-a when A-profile libraries have not been built.
27836         * config/arm/t-rmprofile: Rewrite.
27838 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27840         * genmultilib (multilib_reuse): Allow an explicit period to be escaped
27841         with a backslash.  Remove the backslash after substituting unescaped
27842         periods.
27843         * doc/fragments.texi (MULTILIB_REUSE): Document it.
27845 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27847         * config.gcc: (arm*-*-*): When building a-profile libraries, force
27848         the driver to pass through the default setting of -mfloat-abi.
27849         * common/config/arm/arm-common.c (arm_target_thumb_only): Return -marm
27850         rather than NULL.
27851         * config/arm/t-multilib (MULTILIB_REUSE): Initialize to empty.
27852         (all_feat_combs): New rule.
27853         (MULTILIB_OPTIONS): Use explicit ARM and Thumb directories.  Rework
27854         default libraries.
27855         * config/arm/t-aprofile: Rewrite.
27857 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27859         * config/arm/arm.h (FPUTYPE_AUTO): Define.
27860         * config/arm/arm.c (arm_option_override): Use FPUTYPE_AUTO if the
27861         fpu is not specified by the user/command-line.
27862         * config/arm/bpabi.h (FPUTYPE_DEFAULT): Delete.
27863         * config/arm/netbsd-elf.h (FPUTYPE_DEFAULT): Delete.
27864         * config/arm/linux-elf.h (FPUTYPE_DEFAULT): Delete.
27865         * config/arm/vxworks.h (FPUTYPE_DEFAULT): Delete.
27866         * common/config/arm/arm-common.c (arm_canon_arch_option): Use
27867         FPUTYPE_AUTO insted of FPUTYPE_DEFAULT.
27869 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27871         * config/arm/elf.h (MULTILIB_DEFAULTS): Delete.
27872         * config/arm/t-arm-elf: Rewritten.
27874 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27876         * config/arm/arm.h (TARGET_HARD_FLOAT): Also check that we
27877         have some floating-point instructions.
27878         (TARGET_SOFT_FLOAT): Define as inverse of TARGET_HARD_FLOAT.
27879         (TARGET_MAYBE_HARD_FLOAT): New macro.
27880         * config/arm/arm-builtins.c (arm_init_builtins): Use
27881         TARGET_MAYBE_HARD_FLOAT.
27882         * config/arm/arm.c (arm_option_override): Use TARGET_HARD_FLOAT_ABI.
27884 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27886         * common/config/arm/arm-common.c: Define INCLUDE_LIST.
27887         (configargs.h): Include it.
27888         (arm_print_hint_for_fpu_option): New function.
27889         (arm_parse_fpu_option): New function.
27890         (candidate_extension): New class.
27891         (arm_canon_for_multilib): New function.
27892         * config/arm/arm.h (CANON_ARCH_SPEC_FUNCTION): New macro.
27893         (EXTRA_SPEC_FUNCTIONS): Add CANON_ARCH_SPEC_FUNCTION.
27894         (ARCH_CANONICAL_SPECS): New macro.
27895         (DRIVER_SELF_SPECS): Add ARCH_CANONICAL_SPECS.
27897 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27899         * config.gcc (arm*-*-*): Ensure both target_cpu_cname and with_cpu
27900         are set after handling multilib fragments.  Set target_cpu_default2
27901         from with_cpu.
27903 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27905         * config.gcc (arm*-*-fucshia*): Set target_cpu_cname to the real
27906         cpu name.
27907         (arm*-*-*): Set target_cpu_default2 to a quoted string.
27908         * config/arm/parsecpu.awk (check_cpu): Validate any extension
27909         options.
27910         (check_arch): Likewise.
27911         * config/arm/arm.c (arm_configure_build_target): Handle
27912         TARGET_CPU_DEFAULT being a string constant.  Scan any feature
27913         options in the default.
27915 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27917         * config/arm/arm-protos.h (cpu_arch_extension): Add field to record
27918         when an option is an alias of another.
27919         * config/arm/parsecpu.awk (optalias): New parser token.
27920         (gen_comm_data): Mark non-alias options as such.  Emit entries
27921         for extension aliases.
27922         * config/arm/arm-cpus.in (armv5e): Make vfpv2 an alias.
27923         (armv5te, armv5tej, armv6, armv6j, armv6k, armv6z): Likewise.
27924         (armv6kz, armv6zk, armv6t2): Likewise.
27925         (armv7): Make vfpv3-d16 an alias.
27926         (armv7-a): Make vfpv3-d16, neon and neon-vfpv3 aliases.  Sort in
27927         canonical order.
27928         (armv7ve): Make vfpv4-d16, neon-vfpv3 and neon-vfpv4 aliases.
27929         Sort in canonical order.
27930         (armv8-a): Sort in canonical order.
27931         (armv8.1-a, armv8.2-a):  Likewise.
27932         (generic-armv7-a): Make neon and neon-vfpv3 aliases.  Sort in
27933         canonical order.
27934         (cortex-a9): Sort in canonical order.
27935         * config/arm/arm.c (selftests.h): Include it.
27936         (arm_test_cpu_arch_data): New function.
27937         (arm_run_self_tests): New function.
27938         (TARGET_RUN_TARGET_SELFTESTS): Redefine.
27939         (targetm): Move declaration to the end of the file.
27940         * arm-cpu-cdata.h: Regenerated.
27942 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27944         * config/arm/arm.h (TARGET_MODE_SPECS): Add additional parameter to
27945         call to target_mode_check describing the type of option passed.
27946         * common/config/arm/arm-common.c (arm_arch_core_flag): Delete.
27947         (arm_target_thumb_only): Use arm_parse_arch_option_name or
27948         arm_parse_cpu_option_name to match parameters against list of
27949         available targets.
27950         * config/arm/parsecpu.awk (gen_comm_data): Don't generate
27951         arm_arch_core_flags data structure.
27952         * config/arm/arm-cpu_cdata.h: Regenerated.
27954 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27956         * common/config/arm/arm-common.c (arm_initialize_isa): Moved here from
27957         config/arm/arm.c.
27958         (arm_print_hint_for_cpu_option): Likewise.
27959         (arm_print_hint_for_arch_option): Likewise.
27960         (arm_parse_cpu_option_name): Likewise.
27961         (arm_parse_arch_option_name): Likewise.
27962         * config/arm/arm.c (arm_identify_fpu_from_isa): Use the computed number
27963         of entries in the all_fpus list.
27964         * config/arm/arm-protos.h (all_architectures, all_cores): Declare.
27965         (arm_parse_cpu_option_name): Declare.
27966         (arm_parse_arch_option_name): Declare.
27967         (arm_parse_option_features): Declare.
27968         (arm_intialize_isa): Declare.
27969         * config/arm/parsecpu.awk (gen_data): Move CPU and architecture
27970         data tables to ...
27971         (gen_comm_data): ... here.  Make definitions non-static.
27972         * config/arm/arm-cpu-data.h: Regenerated.
27973         * config/arm/arm-cpu-cdata.h: Regenerated.
27975 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
27977         * config/arm/arm-protos.h (arm_build_target): Remove arch_core.
27978         (cpu_arch_extension): New structure.
27979         (cpu_arch_option, arch_option, cpu_option): New structures.
27980         * config/arm/parsecpu.awk (gen_headers): Build an enumeration of
27981         architecture types.
27982         (gen_data): Generate new format data tables.
27983         * config/arm/arm.c (cpu_tune): New structure.
27984         (cpu_option, processors): Delete.
27985         (arm_print_hint_for_core_or_arch): Delete.  Replace with ...
27986         (arm_print_hint_for_cpu_option): ... this and ...
27987         (arm_print_hint_for_arch_option): ... this.
27988         (arm_parse_arch_cpu_name): Delete.  Replace with ...
27989         (arm_parse_cpu_option_name): ... this and ...
27990         (arm_parse_arch_option_name): ... this.
27991         (arm_unrecognized_feature): Change type of target parameter to
27992         cpu_arch_option.
27993         (arm_parse_arch_cpu_features): Delete.  Replace with ...
27994         (arm_parse_option_features): ... this.
27995         (arm_configure_build_target): Rework to use new configuration data
27996         tables.
27997         (arm_print_tune_info): Rework for new configuration data tables.
27998         * config/arm/arm-cpu-data.h: Regenerated.
27999         * config/arm/arm-cpu.h: Regenerated.
28001 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28003         * Makefile.in (OBJS): Move sbitmap.o from here ...
28004         (OBJS-libcommon): ... to here.
28006 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28008         * config/arm/arm-isa.h (ISA_ALL_FPU_INTERNAL): Renamed from ISA_ALL_FPU.
28009         (ISA_ALL_CRYPTO): New macro.
28010         (ISA_ALL_SIMD): New macro
28011         (ISA_ALL_FP): New macro.
28012         * config/arm/arm.c (fpu_bitlist): Update initializer.
28013         * config/arm/arm-cpus.in: Use new ISA_ALL macros to disable crypto,
28014         simd or fp.
28015         (arm9e): Add fpu.  Add option for nofp
28016         (arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e): Likewise.
28017         (arm926ej-s, arm1026ej-s): Likewise.
28018         (generic-armv7-a): Add fpu.  Add options for simd, vfpv3, vfpv3-d16,
28019         vfpv3-fp16, vfpv3-d16-fp16, vfpv4, vfpv4-d16, neon, neon-vfp3,
28020         neon-fp16, neon-vfpv4, nofp and nosimd.
28021         (cortex-a5, cortex-a7): Add fpu.  Add options for nosimd and nofp.
28022         (cortex-a8): Add fpu.  Add option for nofp.
28023         (cortex-a9): Add fpu.  Add options for nosimd and nofp.
28024         (cortex-a12, cortex-a15, cortex-a17): Add fpu.  Add option for nofp.
28025         (cortex-r4f): Add fpu.
28026         (cortex-r5): Add fpu.  Add options for nofp.dp and nofp.
28027         (cortex-r7): Use idiv option from architecture.  Add fpu.  Add option
28028         for nofp.
28029         (cortex-r8): Likewise.
28030         (cortex-m4): Add fpu.  Add option for nofp.
28031         (cortex-a15.cortex-a7): Add fpu.  Add option for nofp.
28032         (cortex-a17.cortex-a7): Likewise.
28033         (cortex-a32): Add fpu.  Add options for crypto and nofp.
28034         (cortex-a35, cortex-a53): Likewise.
28035         (cortex-a57): Add fpu.  Add option for crypto.
28036         (cortex-a72, cortex-a73): Likewise.
28037         (exynos-m1): Likewise.
28038         (cortex-a57.cortex-a53, cortex-a72.cortex-a53): Likewise.
28039         (cortex-a73.cortex-a35, cortex-a73.cortex-a53): Likewise.
28040         (cortex-m33): Add fpu.  Add option for nofp.
28041         * config/arm/arm-cpu-cdata.h: Regenerated
28042         * config/arm/arm-cpu-data.h: Regenerated.
28044 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28046         * arm-cpus.in (armv5e): Add options fp, vfpv2 and nofp.
28047         (armv5te, armv5tej): Likewise.
28048         (armv6, armv6j, armv6k, armv6z, armv6kz, armv6zk, armv6t2): Likewise.
28049         (armv7): Add options fp and vfpv3-d16.
28050         (armv7-a): Add options fp, simd, vfpv3, vfpv3-d16, vfpv3-d16-fp16,
28051         vfpv3-fp16, vfpv4, vfpv4-d16, neon, neon-vfpv3, neon-fp16, neon-vfpv4,
28052         nofp and nosimd.
28053         (armv7ve): Likewise.
28054         (armv7-r): Add options fp, fp.sp, idiv, nofp and noidiv.
28055         (armv7e-m): Add options fp, fpv5, fp.dp and nofp.
28056         (armv8-a): Add nocrypto option.
28057         (armv8.1-a, armv8.2-a): Likewise.
28058         (armv8-m.main): add options fp, fp.dp and nofp.
28060 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28062         * config/arm/arm-cpus.in (armv8-a): Add options crc, simd crypto and
28063         nofp.
28064         (armv8-a+crc): Delete.
28065         (armv8.1-a): Add options simd, crypto and nofp.
28066         (armv8.2-a): Add options fp16, simd, crypto and nofp.
28067         (armv8.2-a+fp16): Delete.
28068         (armv8-m.main): Add option dsp.
28069         (armv8-m.main+dsp): Delete.
28070         (cortex-a8): Add fpu.  Add nofp option.
28071         (cortex-a9): Add fpu.  Add nofp and nosimd options.
28072         * config/arm/parsecpu.awk (gen_data): Generate option tables and
28073         link to main cpu and architecture data structures.
28074         (gen_comm_data): Only put isa attributes from the main architecture
28075         in common tables.
28076         (option): New statement for architecture and CPU entries.
28077         * arm.c (struct cpu_option): New structure.
28078         (struct processors): Add entry for options.
28079         (arm_unrecognized_feature): New function.
28080         (arm_parse_arch_cpu_name): Ignore any characters after the first
28081         '+' character.
28082         (arm_parse_arch_cpu_feature): New function.
28083         (arm_configure_build_target): Separate out any CPU and architecture
28084         features and parse separately.  Don't error out if -mfpu=auto is
28085         used with only an architecture string.
28086         (arm_print_asm_arch_directives): New function.
28087         (arm_file_start): Call it.
28088         * config/arm/arm-cpu-cdata.h: Regenerated.
28089         * config/arm/arm-cpu-data.h: Likewise.
28090         * config/arm/arm-tables.opt: Likewise.
28092 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28094         * config/arm/elf.h (ASM_SPEC): Only pass -mfpu through to the
28095         assembler when it is not -mfpu=auto.
28097 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28099         * config/arm/arm.h (BIG_LITTLE_SPEC): Delete macro.
28100         (ASM_REWRITE_SPEC_FUNCTIONS): New macro.
28101         (BIG_LITTLE_CPU_SPEC_FUNCTIONS): Delete macro.
28102         (ASM_CPU_SPEC): Rewrite.
28103         (MCPU_MTUNE_NATIVE_FUNCTIONS): New macro.
28104         (EXTRA_SPEC_FUNCTIONS): Move outside of ifdef.  Use
28105         MCPU_MTUNE_NATIVE_FUNCTIONS and ASM_REWRITE_SPEC_FUNCTIONS.  Remove
28106         reference to BIG_LITTLE_CPU_SPEC_FUNCTIONS.
28107         * common/config/arm/arm-common.c (arm_rewrite_selected_cpu): Ensure
28108         copied string is NUL-terminated.  Also strip any characters prefixed
28109         by '+'.
28110         (arm_rewrite_selected_arch): New function.
28111         (arm_rewrite_march): New function.
28113 2017-06-16  Richard Earnshaw  <rearnsha@arm.com>
28115         * config/arm/arm.opt (x_arm_arch_string): New TargetSave option.
28116         (x_arm_cpu_string, x_arm_tune_string): Likewise.
28117         (march, mcpu, mtune): Convert to string-based options.
28118         * config/arm/arm.c (arm_print_hint_for_core_or_arch): New function.
28119         (arm_parse_arch_cpu_name): New function.
28120         (arm_configure_build_target): Use arm_parse_arch_cpu_name to
28121         identify selected architecture or CPU.
28122         (arm_option_save): New function.
28123         (TARGET_OPTION_SAVE): Redefine.
28124         (arm_option_restore): Restore string options.
28125         (arm_option_print): Print string options.
28127 2017-06-16  Martin Sebor  <msebor@redhat.com>
28129         PR tree-optimization/80933
28130         PR tree-optimization/80934
28131         * builtins.c (fold_builtin_3): Do not handle bcmp here.
28132         * gimple-fold.c (gimple_fold_builtin_bcmp): New function.
28133         (gimple_fold_builtin_bcopy, gimple_fold_builtin_bzero): Likewise.
28134         (gimple_fold_builtin): Call them.
28136 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28138         * gimple-ssa-isolate-paths.c (isolate_path): Set edge leading to path
28139         as unlikely; update profile.
28141 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28143         * predict.c (force_edge_cold): Handle declaring edges impossible
28144         more aggresively.
28146 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28148         * tree-ssa-loop-ivcanon.c (remove_exits_and_undefined_stmts): Update
28149         profile.
28150         (try_unroll_loop_completely): Fix reporting.
28152 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28154         * tree-ssa-tail-merge.c (replace_block_by): Fix profile updating.
28156 2017-06-16  James Greenhalgh  <james.greenhalgh@arm.com>
28158         PR target/71778
28159         * config/arm/arm-builtins.c (arm_expand_builtin_args): Return TARGET
28160         if given a non-constant argument for an intrinsic which requires a
28161         constant.
28163 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28165         * profile.c (compare_freqs): New function.
28166         (branch_prob): Sort edge list.
28167         (find_spanning_tree): Assume that the list is priority sorted.
28169 2017-06-16  Richard Biener  <rguenther@suse.de>
28171         PR tree-optimization/81090
28172         * passes.def (pass_record_bounds): Remove.
28173         * tree-pass.h (make_pass_record_bounds): Likewise.
28174         * tree-ssa-loop.c (pass_data_record_bounds, pass_record_bounds,
28175         make_pass_record_bounds): Likewise.
28176         * tree-ssa-loop-ivcanon.c (canonicalize_induction_variables): Do
28177         not free niter estimates at the beginning but at the end.
28178         * tree-scalar-evolution.c (scev_finalize): Free niter estimates.
28180 2017-06-16  Richard Biener  <rguenther@suse.de>
28182         * tree-switch-conversion.c (emit_case_bit_tests): Adjust
28183         initializer to workaround ICE in host GCC 4.8.
28185 2017-06-16  Jan Hubicka  <hubicka@ucw.cz>
28187         * ipa-inline-transform.c (update_noncloned_frequencies): Update also
28188         counts.
28189         (clone_inlined_nodes): Update.
28191 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28193         * config/aarch64/aarch64.c (qdf24xx_prefetch_tune): Update
28194         prefetch settings, and enable prefetching by default at -O3.
28196 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28198         * config/aarch64/aarch64.c (aarch64_override_options_internal):
28199         Set flag_prefetch_loop_arrays according to tuning data.
28201 2017-06-16  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
28203         * config/aarch64/aarch64-protos.h (struct cpu_prefetch_tune):
28204         New tune structure.
28205         (struct tune_params): Use cpu_prefetch_tune instead of cache_line_size.
28206         [Unrelated to main purpose of the patch] Place the pointer field last
28207         to enable type checking errors when tune structure are wrongly merged.
28208         * config/aarch64/aarch64.c (generic_prefetch_tune,)
28209         (exynosm1_prefetch_tune, qdf24xx_prefetch_tune,)
28210         (thunderx2t99_prefetch_tune): New tune constants.
28211         (tune_params *_tunings): Update all tunings (no functional change).
28212         (aarch64_override_options_internal): Set PARAM_SIMULTANEOUS_PREFETCHES,
28213         PARAM_L1_CACHE_SIZE, PARAM_L1_CACHE_LINE_SIZE, and PARAM_L2_CACHE_SIZE
28214         from tunings structures.
28216 2017-06-16  Jakub Jelinek  <jakub@redhat.com>
28218         PR sanitizer/81094
28219         * ubsan.c (instrument_null): Add T argument, use it instead
28220         of computing it based on IS_LHS.
28221         (instrument_object_size): Likewise.
28222         (pass_ubsan::execute): Adjust instrument_null and
28223         instrument_object_size callers to pass gimple_get_lhs or
28224         gimple_assign_rhs1 result to it.  Use instrument_null instead of
28225         calling get_base_address and instrument_mem_ref.  Handle
28226         aggregate call arguments for object-size sanitization.
28228 2017-06-16  Yury Gribov  <tetra2005@gmail.com>
28230         PR tree-optimization/81089
28231         * tree-vrp.c (is_masked_range_test): Validate operands of
28232         subexpression.
28234 2017-06-15  Martin Sebor  <msebor@redhat.com>
28236         PR c++/80560
28237         * dumpfile.c (dump_register): Avoid calling memset to initialize
28238         a class with a default ctor.
28239         * gcc.c (struct compiler): Remove const qualification.
28240         * genattrtab.c (gen_insn_reserv): Replace memset with initialization.
28241         * hash-table.h: Ditto.
28242         * ipa-cp.c (allocate_and_init_ipcp_value): Replace memset with
28243           assignment.
28244         * ipa-prop.c (ipa_free_edge_args_substructures): Ditto.
28245         * omp-low.c (lower_omp_ordered_clauses): Replace memset with
28246         default ctor.
28247         * params.h (struct param_info): Make struct members non-const.
28248         * tree-switch-conversion.c (emit_case_bit_tests): Replace memset
28249         with default initialization.
28250         * vec.h (vec_copy_construct, vec_default_construct): New helper
28251         functions.
28252         (vec<T>::copy, vec<T>::splice, vec<T>::reserve): Replace memcpy
28253         with vec_copy_construct.
28254         (vect<T>::quick_grow_cleared): Replace memset with default ctor.
28255         (vect<T>::vec_safe_grow_cleared, vec_safe_grow_cleared): Same.
28256         * doc/invoke.texi (-Wclass-memaccess): Document.
28258 2017-06-15  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
28260         * emit-rtl.h (is_leaf): Update comment about local
28261         register allocator.
28263 2017-06-15  Jozef Lawrynowicz   <jozef.l@somniumtech.com>
28265         PR target/78818
28266         * config/msp430/msp430.c (msp430_data_attr): Check that it's possible
28267         for a variable to have a section before checking if the section has a
28268         name.
28269         Set section to.persistent if persistent attribute is set.
28270         Warn if .persistent attribute is used on an automatic variable.
28272 2017-06-15  Eric Botcazou  <ebotcazou@adacore.com>
28274         PR rtl-optimization/80474
28275         * reorg.c (update_block): Do not ignore instructions in a delay slot.
28277 2017-06-15  Segher Boessenkool  <segher@kernel.crashing.org>
28279         * config/rs6000/rs6000.md (add<mode>3): Use reg_or_subregno instead
28280         of REGNO.
28282 2017-06-14  Maciej W. Rozycki  <macro@imgtec.com>
28284         * config/mips/mips.md (MIPS16_T_REGNUM): Remove constant.
28285         (casesi): Emit bounds checking as RTL.
28286         (casesi_internal_mips16_<mode>): Remove bounds checking.
28288 2017-06-14  Max Filippov  <jcmvbkbc@gmail.com>
28290         * config/xtensa/xtensa.c (xtensa_option_override): Append
28291         MASK_CONST16 to target_flags in the absence of TARGET_L32R.
28292         (hwloop_optimize, hwloop_fail, hwloop_pattern_reg,
28293          xtensa_doloop_hooks): Define unconditionally.
28294         (xtensa_reorg_loops): Only call reorg_loops in the presence of
28295         TARGET_LOOPS.
28296         * config/xtensa/xtensa.h (TARGET_L32R): New definition.
28297         (TARGET_DEFAULT): Remove XCHAL_HAVE_L32R condition and account
28298         for it in xtensa_option_override.
28299         (HARD_FRAME_POINTER_IS_FRAME_POINTER,
28300          HARD_FRAME_POINTER_IS_ARG_POINTER): New definitions.
28302 2017-06-14  Boris Kolpackov  <boris@codesynthesis.com>
28304         * doc/cppopts.texi: Document '-' special value to -MF.
28306 2017-06-14  Wilco Dijkstra  <wdijkstr@arm.com>
28308         * config/arm/cortex-a53.md (cortex_a53_fpalu) Adjust latency.
28309         (cortex_a53_fconst): Likewise.
28310         (cortex_a53_fpmul): Likewise.
28311         (cortex_a53_f_load_64): Likewise.
28312         (cortex_a53_f_load_many): Likewise.
28313         (cortex_a53_advsimd_alu): Likewise.
28314         (cortex_a53_advsimd_alu_q): Likewise.
28315         (cortex_a53_advsimd_mul): Likewise.
28316         (cortex_a53_advsimd_mul_q): Likewise.
28317         (fpmac bypass): Add new bypass for fpmac-fpmac case.
28318         Add missing fmul, r2f_cvt and fconst cases.
28320 2017-06-14  Richard Biener  <rguenther@suse.de>
28322         PR middle-end/81088
28323         * fold-const.c (split_tree): Drop TREE_OVERFLOW flag from
28324         literal constants.
28325         (fold_binary_loc): When associating do not treat pre-existing
28326         TREE_OVERFLOW on literal constants as a reason to allow
28327         TREE_OVERFLOW on associated literal constants.
28329 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28331         * config/sparc/sparc.h (MASK_ISA): Add MASK_LEON and MASK_LEON3.
28332         (MASK_FEATURES): New macro.
28333         * config/sparc/sparc.c (sparc_option_override): Remove the special
28334         handling of -mfpu and generalize it to all MASK_FEATURES switches.
28336 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28338         * simplify-rtx.c (simplify_binary_operation_1) <UDIV>: Do not simplify
28339         a division of 0 if non-call exceptions are enabled.
28341 2017-06-14  Andrew Pinski  <apinski@cavium.com>
28342             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
28344         PR target/71663
28345         * config/aarch64/aarch64.c (aarch64_expand_vector_init):
28346         Improve vector initialization code gen for only variable case.
28348 2017-06-14  Eric Botcazou  <ebotcazou@adacore.com>
28350         * config/sparc/driver-sparc.c (cpu_names): Add SPARC-T5 entry.
28352 2017-06-14  Richard Biener  <rguenther@suse.de>
28354         PR tree-optimization/81083
28355         * tree-ssa-sccvn.c (vn_reference_lookup_3): Do not use abnormals
28356         as values.
28358 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28360         * config/rs6000/rs6000.c: Update all comments that mentioned SPE.
28361         (rs6000_expand_builtin): Remove RS6000_BTC_EVSEL.
28362         * config/rs6000/rs6000.h (RS6000_BTC_EVSEL): Delete.
28363         * config/rs6000/vxworks.h (VXCPU_FOR_8548): Delete.  Adjust former use.
28364         * config/rs6000/vxworksae.h (VXCPU_FOR_8548): Delete.
28365         * config/rs6000/vxworksmils.h (VXCPU_FOR_8548): Delete.
28367 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28369         * config/rs6000/rs6000-opts.h (enum rs6000_vector): Delete VECTOR_SPE.
28370         * config/rs6000/rs6000.c (rs6000_debug_vector_unit): Delete VECTOR_SPE.
28372 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28374         * config/rs6000/rs6000.h (FIXED_SCRATCH): Delete.
28376 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28378         * config/rs6000/t-rtems: Don't handle SPE.
28380 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28382         * config/rs6000/t-linux: Don't handle SPE.
28384 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28386         * config/rs6000/eabispe.h: Delete file.
28388 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28390         * config/rs6000/t-spe: Delete file.
28392 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28394         * config/rs6000/rs6000.c (SPE_CONST_OFFSET_OK): Delete.
28395         (rs6000_legitimate_offset_address_p): Return false for anything in
28396         V2SImode or V2SFmode.
28398 2017-06-13  Segher Boessenkool  <segher@kernel.crashing.org>
28400         * config/rs6000/rs6000-modes.def: Remove all 8-byte vector modes
28401         except V2SF and V2SI.  Rearrange the vector modes, and add comments.
28402         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Remove V8QImode
28403         and V4HImode.
28404         (reg_offset_addressing_ok_p): Remove V4HImode and V1DImode.
28405         (rs6000_legitimate_offset_address_p): Ditto.
28406         (rs6000_emit_move): Ditto.
28407         (rs6000_init_builtins): Remove V4HI_type_node.
28409 2017-06-13  Martin Liska  <mliska@suse.cz>
28411         PR sanitize/78204
28412         * asan.c (asan_sanitize_stack_p): Use sanitize_flags_p.
28413         (gate_asan): Likewise.
28414         * asan.h (asan_no_sanitize_address_p): Remove the function.
28415         (sanitize_flags_p): New function.
28416         * builtins.def: Fix coding style.
28417         * common.opt: Use renamed enum value.
28418         * convert.c (convert_to_integer_1): Use sanitize_flags_p.
28419         * doc/extend.texi: Document no_sanitize attribute.
28420         * flag-types.h (enum sanitize_code): Rename SANITIZE_NONDEFAULT
28421         to SANITIZE_UNDEFINED_NONDEFAULT.
28422         * gcc.c (sanitize_spec_function): Use the renamed enum value.
28423         * gimple-fold.c (optimize_atomic_compare_exchange_p):
28424         Use sanitize_flags_p.
28425         * gimplify.c (gimplify_function_tree): Likewise.
28426         * ipa-inline.c (sanitize_attrs_match_for_inline_p): Likewise.
28427         * opts.c (parse_no_sanitize_attribute): New function.
28428         (common_handle_option): Use renamed enum value.
28429         * opts.h (parse_no_sanitize_attribute): Declare.
28430         * tree.c (sanitize_flags_p): New function.
28431         * tree.h: Declared here.
28432         * tsan.c: Use sanitize_flags_p.
28433         * ubsan.c (ubsan_expand_null_ifn): Likewise.
28434         (instrument_mem_ref): Likewise.
28435         (instrument_bool_enum_load): Likewise.
28436         (do_ubsan_in_current_function): Remove the function.
28437         (pass_ubsan::execute): Use sanitize_flags_p.
28438         * ubsan.h: Remove do_ubsan_in_current_function
28439         * tree-cfg.c (print_no_sanitize_attr_value): New function.
28440         (dump_function_to_file): Use it here.
28442 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28444         PR tree-optimization/80803
28445         PR tree-optimization/81063
28446         * tree-sra.c (subtree_mark_written_and_enqueue): Move up in the file.
28447         (propagate_subaccesses_across_link): Enqueue subtree whenever
28448         necessary instead of relying on the caller.
28450 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28452         * tree-sra.c (add_access_to_work_queue): Only enqueue accesses
28453         that have a first_link.
28454         (sort_and_splice_var_accesses): Do not check first_link before
28455         enquing.
28456         (subtree_mark_written_and_enqueue): Likewise.
28457         (propagate_all_subaccesses): Likewise and do not stop at first
28458         parent with a first_link.
28460 2017-06-13  Martin Jambor  <mjambor@suse.cz>
28462         * tree-sra.c (dump_access_tree_1): Fix accidental dumping to stderr
28463         instead of f.
28465 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28467         * match.pd: New pattern.
28469 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28471         * tree-vrp.c (is_masked_range_test): New function.
28472         (register_edge_assert_for): Determine ranges for
28473         some bit tests.
28475 2017-06-13  Yury Gribov  <tetra2005@gmail.com>
28477         PR tree-optimization/67328
28478         * fold-const.c (maskable_range_p): New function.
28479         (build_range_check): Generate bittests if possible.
28481 2017-06-13  Martin Liska  <mliska@suse.cz>
28483         * gimple-pretty-print.c (dump_probability): Add new argument.
28484         (dump_edge_probability): Dump both probability and count.
28485         (dump_gimple_label): Likewise.
28486         (dump_gimple_bb_header): Likewise.
28488 2017-06-13  Georg-Johann Lay  <avr@gjlay.de>
28490         PR target/81072
28491         * config/avr/avr-devices.c: Fix indentation.
28492         * config/avr/gen-avr-mmcu-specs.c: Dito.
28494 2017-06-13  Richard Biener  <rguenther@suse.de>
28496         * tree-vect-loop.c (vect_model_reduction_cost): Do not fail,
28497         instead get vector type from stmt_info.
28498         (vectorizable_reduction): Adjust.  Remove dead code.
28500 2017-06-13  Richard Biener  <rguenther@suse.de>
28502         PR middle-end/81065
28503         * fold-const.c (extract_muldiv_1): Remove bogus distribution
28504         case of C * (x * C2 + C3).
28505         (fold_addr_of_array_ref_difference): Properly fold index difference.
28507 2017-06-12  David S. Miller  <davem@davemloft.net>
28509         PR target/80968
28510         * config/sparc/sparc.md (return expander): Emit frame blockage if
28511         function uses alloca.
28513 2017-06-12  Richard Sandiford  <richard.sandiford@linaro.org>
28515         * combine.c (make_field_assignment): Check len rather than the mode
28516         precision when calling force_to_mode.
28518 2017-06-12  Georg-Johann Lay  <avr@gjlay.de>
28520         Support multilibs and devices that see flash in RAM address range.
28522         PR target/81072
28523         * config/avr/avr-arch.h (avr_arch_id) <ARCH_AVRXMEGA3>: New enum.
28524         (avr_mcu_t) <flash_pm_offset>: New field.
28525         (avr_device_specific_features) <AVR_ISA_RCALL>: New enum.
28526         * config/avr/avr.h (AVR_SHORT_CALLS): New define.
28527         (AVR_HAVE_JMP_CALL): Don't set if AVR_SHORT_CALLS.
28528         (AVR_TINY_PM_OFFSET): Remove macro.
28529         * config/avr/avr.opt (-mshort-calls): New option.
28530         * config/avr/gen-avr-mmcu-specs.c (print_mcu)
28531         [*self_spec]: Add / remove -mshort-calls depending on AVR_ISA_RCALL.
28532         * config/avr/avr-c.c (avr_cpu_cpp_builtins)
28533         <__AVR_SHORT_CALLS__>: Built-in define if AVR_SHORT_CALLS.
28534         <__AVR_HAVE_JMP_CALL__>: Use AVR_HAVE_JMP_CALL as condition
28535         instead of avr_arch->have_jmp_call.
28536         <__AVR_PM_BASE_ADDRESS__>: Built-in define if avr_arch->flash_pm_offset.
28537         [AVR_TINY] <__AVR_TINY_PM_BASE_ADDRESS__>: Use
28538         avr_arch->flash_pm_offset to define.
28539         * config/avr/avr-devices.c (avr_arch_types): Add initializers for
28540         new field flash_pm_offset.  Add entry for avrxmega3.
28541         (avr_texinfo): Add entry for avrxmega3.
28542         * config/avr/avr-mcus.def: Add entries for: avrxmega3,
28543         attiny212, attiny214,
28544         attiny412, attiny414, attiny416, attiny417,
28545         attiny814, attiny816, attiny817,
28546         attiny1614, attiny1616, attiny1617,
28547         attiny3214, attiny3216, attiny3217.
28548         * config/avr/avr.c (avr_assemble_integer)[AVR_TINY]: Use
28549         avr_arch->flash_pm_offset instead of AVR_TINY_PM_OFFSET.
28550         (avr_print_operand_address) [AVR_TINY]: Same.
28551         (avr_asm_init_sections) <readonly_data_section>: Only patch
28552         callback if avr_arch->flash_pm_offset = 0.
28553         (avr_asm_named_section) <avr_need_copy_data_p>: Skip setting it
28554         for rodata if avr_arch->flash_pm_offset != 0.
28555         (avr_encode_section_info) [AVR_TINY]: Adjust comment.
28556         * config/avr/genmultilib.awk (dir_rcall, opt_rcall): New vars.
28557         (opts) [AVR_ISA_RCALL]: Append opt_rcall.
28558         (m_options): Append opt_rcall.
28559         (m_dirnames): Append dir_rcall.
28560         * config/avr/t-multilib: Regenerate.
28562         * configure.ac [target=avr]: Check whether avrxmega3 default
28563         linker description file works as needed.
28564         * configure: Regenerate.
28565         * doc/avr-mmcu.texi: Regenerate.
28566         * doc/invoke.texi (AVR Options) <-mshort-calls>: Document it.
28567         <__AVR_ARCH__>: Document avrxmega3 and 103.
28568         <__AVR_HAVE_JMP_CALL__>: Adjust documentation.
28569         <__AVR_SHORT_CALLS__>: Document it.
28570         <__AVR_PM_BASE_ADDRESS__>: Document it.
28571         * doc/extend.texi (AVR Options) <-mshort-calls>: Document it.
28572         (AVR Variable Attributes) <progmem>: Document this is
28573         not needed for avrxmega3.
28574         (AVR Named Address Spaces) <__flash>: Dito.
28576 2017-06-12  Jan Hubicka  <hubicka@ucw.cz>
28578         * cgraph.c (cgraph_node::dump): Complain about profile insanities.
28580 2017-06-12  Doug Rupp  <rupp@adacore.com>
28582         * config.gcc (*-*-vxworks*): Set use_gcc_stdint to "provide".
28583         Append vxworks-stdint.h to the tm_file list.
28584         * config/vxworks-stdint.h: New file.
28586 2017-06-12  Martin Liska  <mliska@suse.cz>
28588         PR tree-optimization/81041
28589         * tree-profile.c (gimple_gen_ic_func_profiler):
28590         Create an extra BB in profile-generate
28591         (gimple_gen_time_profiler): Likewise.
28593 2017-06-12  Jakub Jelinek  <jakub@redhat.com>
28595         PR tree-optimization/81003
28596         * tree-ssa-reassoc.c (force_into_ssa_name): New function.
28597         (update_range_test): Use it instead of force_gimple_operand_gsi.
28599 2017-06-12  Richard Biener  <rguenther@suse.de>
28601         PR tree-optimization/81053
28602         * tree-vect-loop.c (vect_is_simple_reduction): Handle PHI
28603         with backedge value not defined in loop.  Simplify def stmt
28604         compute.
28606 2017-06-11  Tom de Vries  <tom@codesourcery.com>
28608         PR target/79939
28609         * config/nvptx/nvptx.c (nvptx_cannot_force_const_mem): New function.
28610         Return true.
28611         (TARGET_CANNOT_FORCE_CONST_MEM): Redefine to
28612         nvptx_cannot_force_const_mem.
28614 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28616         * opts.c (finish_options): Move test for flag_split_stack after
28617         it has been initialized.
28619 2017-06-11  Jason Merrill  <jason@redhat.com>
28621         * tree.h (id_equal): New.
28622         * dwarf2out.c, hsa-gen.c, ipa-devirt.c, omp-expand.c,
28623         omp-simd-clone.c, read-rtl-function.c, tree-chkp.c, tree.c: Use it
28624         instead of strcmp of IDENTIFIER_POINTER.
28626 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28628         * ipa-inline-transform.c: Include function.h, cfg.h and basic-block.h
28629         (mark_all_inlined_calls_cdtor): Fix formating.
28630         (inline_transform): Rescale profile before inlining.
28632 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28634         * cgraph.h (cgraph_edge::clone): Update prototype.
28635         * cgraphclones.c (cgraph_edge::clone): Update profile scaling.
28636         (cgraph_node::create_clone): Update.
28637         (cgraph_node::create_version_clone): Update.
28638         * tree-inline.c (copy_bb): Update.
28639         (expand_call_inline): Update.
28641 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28643         * config/rs6000/rs6000.c (emit_vrsave_prologue): New function,
28644         factored out from ...
28645         (rs6000_emit_prologue): ... here.
28647 2017-06-10  Segher Boessenkool  <segher@kernel.crashing.org>
28649         * config/rs6000/rs6000.c (emit_split_stack_prologue): New function,
28650         factored out from ...
28651         (rs6000_emit_prologue): ... here.
28653 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28655         * predict.c (drop_profile): Also drop individual bb/edge and cgraph
28656         edge counts.
28657         (handle_missing_profiles): Fix computation of tp_first_run.
28658         (counts_to_freqs): Do not touch freqs when count is 0.
28660 2017-06-10  Jan Hubicka  <hubicka@ucw.cz>
28662         * cgraphbuild.c (cgraph_edge::rebuild_references): Do not touch
28663         profile.
28665 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28667         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
28668         attributes): Document signal effective target.
28670 2017-06-10  Tom de Vries  <tom@codesourcery.com>
28672         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
28673         Document effective target stack_size.
28675 2017-06-09  David Malcolm  <dmalcolm@redhat.com>
28677         * diagnostic.c (diagnostic_report_diagnostic): Only add fixits
28678         to the edit_context if they can be auto-applied.
28680 2017-06-9  Ian Lance Taylor  <iant@golang.org>
28682         * opts.c (finish_options): If -fsplit-stack, disable implicit
28683         -forder-blocks-and-partition.
28684         * doc/invoke.texi (Optimize Options): Document that when using
28685         -fsplit-stack -forder-blocks-and-partition is not implicitly
28686         enabled.
28688 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28690         * builtin-attrs.def (ATTR_NORETURN_NOTHROW_LEAF_COLD_LIST,
28691         ATTR_CONST_NORETURN_NOTHROW_LEAF_COLD_LIST,
28692         ATTR_TMPURE_NORETURN_NOTHROW_LEAF_COLD_LIST): New.
28693         * builtins.def (abort, trap, unreachable): Declare cold.
28694         * calls.c (flags_from_decl_or_type): Lookup ECF_COLD.
28695         * tree-core.h (ECF_COLD): New.
28696         * tree.c (set_call_expr_flags): Handle ECF_COLD.
28697         (build_common_builtin_nodes): Mark unreachable and abort as cold.
28699 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28701         * predict.c (unlikely_executed_stmt_p): Cleanup.
28703 2017-06-09  Richard Biener  <rguenther@suse.de>
28705         * tree-ssa-loop-im.c (execute_sm): Do not force multi-threaded
28706         model if the ref is always written to.
28708 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28710         * config/aarch64/aarch64.md (lrint<GPF:mode><GPI:mode>2): New.
28712 2017-06-09  Tamar Christina  <tamar.christina@arm.com>
28714         * config/arm/arm.c (arm_rtx_costs_internal): Make sdiv more expensive
28715         than udiv.
28717 2017-06-09  Tom de Vries  <tom@codesourcery.com>
28719         PR target/80855
28720         * config/nvptx/nvptx.md (define_expand "mov<QHSDISDFM>"): Error out with
28721         "target cannot support label values" when encountering LABEL_REF.
28723 2017-06-09  Martin Liska  <mliska@suse.cz>
28725         * tree-profile.c (gimple_gen_ic_profiler): Update comment.
28726         (gimple_gen_ic_func_profiler): Emit direct comparison
28727         of __gcov_indirect_call_callee with NULL.
28728         (gimple_gen_time_profiler): Change probability from
28729         PROB_VERY_UNLIKELY to PROB_UNLIKELY.
28731 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28733         * profile.c (edge_gcov_counts): Turn to pointer.
28734         (compute_branch_probabilities, compute_branch_probabilities): Update.
28735         (branch_prob): Do not clear edge_gcov_count.
28736         * profile.h (edge_gcov_counts): Turn to pointer.
28737         (edge_gcov_count): Update.
28739 2017-06-09  Jan Hubicka  <hubicka@ucw.cz>
28741         * gimple.h (gimple_check_failed): Mark cold.
28743 2017-06-09  Richard Biener  <rguenther@suse.de>
28745         PR tree-optimization/66623
28746         * tree-vect-loop.c (vect_is_simple_reduction): Cleanup,
28747         refactor check_reduction into two parts, properly computing
28748         whether we have to check reduction validity for outer loop
28749         vectorization.
28751 2017-06-09  Richard Biener  <rguenther@suse.de>
28753         PR tree-optimization/79483
28754         * graphite-scop-detection.c (order): New global.
28755         (get_order): Compute bb to order mapping that satisfies code
28756         generation constraints.
28757         (cmp_pbbs): New helper.
28758         (build_scops): Start domwalk at entry block, sort generated
28759         pbbs.
28761 2017-06-09  Richard Biener  <rguenther@suse.de>
28763         PR middle-end/81007
28764         * ipa-polymorphic-call.c
28765         (ipa_polymorphic_call_context::restrict_to_inner_class):
28766         Skip FIELD_DECLs with error_mark_node type.
28767         * passes.def (all_lowering_passes): Run pass_build_cgraph_edges
28768         last again.
28770 2017-06-09  Martin Liska  <mliska@suse.cz>
28772         * predict.c (struct branch_predictor): New struct.
28773         (test_prediction_value_range): New test.
28774         (predict_c_tests): New function.
28775         * selftest-run-tests.c (selftest::run_tests): Run the function.
28776         * selftest.h: Declare new tests.
28778 2017-06-09  Segher Boessenkool  <segher@kernel.crashing.org>
28780         PR target/80966
28781         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Assert that
28782         gen_add3_insn did not fail.
28783         * config/rs6000/rs6000.md (add<mode>3): If asked to add a constant to
28784         r0, construct that number in a temporary reg and add that reg to r0.
28785         If asked to put the result in r0 as well, fail.
28787 2017-06-08  Will Schmidt  <will_schmidt@vnet.ibm.com>
28789         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
28790         for early expansion of vec_eqv.
28792 2017-06-08  Jakub Jelinek  <jakub@redhat.com>
28794         PR middle-end/81005
28795         * ubsan.c (instrument_null): Avoid pointless code temporary.
28796         (pass_ubsan::execute): Instrument aggregate arguments of calls.
28798 2017-06-08  Uros Bizjak  <ubizjak@gmail.com>
28800         PR target/81015
28801         Revert:
28802         2016-12-14  Uros Bizjak  <ubizjak@gmail.com>
28804         PR target/59874
28805         * config/i386/i386.md (*ctzhi2): New insn_and_split pattern.
28806         (*clzhi2): Ditto.
28808 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28810         * predict.c (unlikely_executed_edge_p): Move ahead.
28811         (probably_never_executed_edge_p): Use it.
28813 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28815         PR middle-end/79988
28816         * tree-chkp.c (chkp_gimple_call_builtin_p): Remove
28817         gimple_call_builtin_p call.
28819 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28821         * system.h (fancy_abort): Annotate by ATTRIBUTE_COLD.
28822         * rtl.h (rtl_check_failed_bounds, rtl_check_failed_type1,
28823         rtl_check_failed_type2, rtl_check_failed_code1,
28824         rtl_check_failed_code2, rtl_check_failed_code_mode,
28825         rtl_check_failed_block_symbol, cwi_check_failed_bounds,
28826         rtvec_check_failed_bounds, rtl_check_failed_flag,
28827         _fatal_insn_not_found, _fatal_insn): Likewise.
28828         * tree.h (tree_contains_struct_check_failed,
28829         tree_check_failed, tree_not_check_failed,
28830         tree_class_check_failed, tree_range_check_failed,
28831         tree_not_class_check_failed, tree_int_cst_elt_check_failed,
28832         tree_vec_elt_check_failed, phi_node_elt_check_failed,
28833         tree_operand_check_failed, omp_clause_check_failed,
28834         omp_clause_operand_check_failed, omp_clause_range_check_failed):
28835         Likewise.
28837 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28839         * cgraph.c (cgraph_edge::maybe_hot_p): Do not check
28840         flag_branch_probabilities.
28841         * ipa-inline.c (edge_badness): Likewise.
28842         * ipa-profile.c (ipa_propagate_frequency_1): Likewise.
28843         * postreload-gcse.c (eliminate_partially_redundant_load): Likewise.
28844         * predict.c (maybe_hot_frequency_p): Likewise.
28845         (probably_never_executed): Likewise.
28846         * sched-ebb.c (schedule_ebbs): Likewise.
28847         * sched-rgn.c (find_single_block_region): Likewise.
28848         * tracer.c (tail_duplicate): Likewise.
28850 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28852         * opts.c (finish_options): x_flag_reorder_blocks_and_partition no
28853         longer requires x_flag_profile_use.
28855 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28857         * cfgrtl.c (cfg_layout_initialize): Check crtl->has_bb_partition
28858         instead of flag_reorder_blocks_and_partition.
28859         * dbxout.c (dbxout_function_end): Likewise.
28860         * dwarf2out.c (gen_subprogram_die): Likewise.
28861         * haifa-sched.c (sched_create_recovery_edges): Likewise.
28862         * hw-doloop.c (reorg_loops): Likewise.
28863         * varasm.c (assemble_start_function,
28864         assemble_end_function): Likewise.
28865         (decide_function_section): Do not check for
28866         flag_reorder_blocks_and_partition.
28868 2017-06-08  Alexander Ivchenko  <alexander.ivchenko@intel.com>
28870         * tree-chkp.c (chkp_get_hard_register_var_fake_base_address):
28871         New function.
28872         (chkp_get_hard_register_fake_addr_expr): Ditto.
28873         (chkp_build_addr_expr): Add check for hard reg case.
28874         (chkp_parse_array_and_component_ref): Ditto.
28875         (chkp_find_bounds_1): Ditto.
28876         (chkp_process_stmt): Don't generate bounds store for
28877         hard reg case.
28879 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28881         * predict.c (maybe_hot_bb_p): Do not check profile status.
28882         (maybe_hot_edge_p): Likewise.
28883         (probably_never_executed): Check for zero counts even if profile
28884         is not read.
28885         (unlikely_executed_edge_p): New function.
28886         (unlikely_executed_stmt_p): New function.
28887         (unlikely_executed_bb_p): New function.
28888         (set_even_probabilities): Use unlikely predicates.
28889         (combine_predictions_for_bb): Likewise.
28890         (predict_paths_for_bb): Likewise.
28891         (predict_paths_leading_to_edge): Likewise.
28892         (determine_unlikely_bbs): New function.
28893         (estimate_bb_frequencies): Use it.
28894         (compute_function_frequency): Use zero counts even if profile is
28895         not read.
28896         * profile-count.h: Fix typo.
28898 2017-08-08  Julia Koval  <julia.koval@intel.com>
28900         * config/i386/avx512bwintrin.h (_mm512_mask_cvtepi16_storeu_epi8,
28901         _mm512_mask_cvtsepi16_storeu_epi8,
28902         _mm512_mask_cvtusepi16_storeu_epi8): New intrinsics.
28903         * config/i386/avx512vlbwintrin.h (_mm256_mask_cvtepi16_storeu_epi8,
28904         _mm_mask_cvtsepi16_storeu_epi8, _mm256_mask_cvtsepi16_storeu_epi8,
28905         _mm_mask_cvtusepi16_storeu_epi8, _mm256_mask_cvtusepi16_storeu_epi8,
28906         _mm_mask_cvtepi16_storeu_epi8): New intrinsics.
28907         * config/i386/i386-builtin-types.def (PV8Q, V8QI): New pointer type.
28908         (VOID_FTYPE_PV32QI_V32HI_USI, VOID_FTYPE_PV8QI_V8HI_UQI,
28909         VOID_FTYPE_PV16QI_V16HI_UHI): New function types.
28910         * config/i386/i386-builtin.def (__builtin_ia32_pmovwb128mem_mask,
28911         __builtin_ia32_pmovwb256mem_mask, __builtin_ia32_pmovswb128mem_mask,
28912         __builtin_ia32_pmovswb256mem_mask, __builtin_ia32_pmovuswb128mem_mask,
28913         __builtin_ia32_pmovuswb256mem_mask,
28914         __builtin_ia32_pmovuswb512mem_mask, __builtin_ia32_pmovswb512mem_mask)
28915         __builtin_ia32_pmovwb512mem_mask): New builtins.
28917 2017-08-08  Julia Koval  <julia.koval@intel.com>
28919         PR target/73350,80862
28920         * config/i386/subst.md (round): Fix round pattern.
28921         * config/i386/i386.c (ix86_erase_embedded_rounding):
28922         Fix erasing rounding for the fixed pattern.
28924 2017-06-08  Jan Hubicka  <hubicka@ucw.cz>
28926         * cfgbuild.c (find_many_sub_basic_blocks): Fix thinko.
28928 2017-06-08  Martin Liska  <mliska@suse.cz>
28930         PR gcov-profile/80911
28931         * gcov.c (block_info::block_info): New constructor.
28933 2017-06-07  Carl Love  <cel@us.ibm.com>
28935         * config/rs6000/rs6000-c: The return type of the following
28936         built-in functions was implemented as int not long long.  Fix sign
28937         of return value for the unsigned version of vec_mulo and vec_mule.
28938         vector unsigned long long vec_bperm (vector unsigned long long,
28939                                              vector unsigned char)
28940         vector signed long long vec_mule (vector signed int,
28941                                           vector signed int)
28942         vector unsigned long long vec_mule (vector unsigned int,
28943                                             vector unsigned int)
28944         vector signed long long vec_mulo (vector signed int,
28945                                           vector signed int)
28946         vector unsigned long long vec_mulo (vector unsigned int,
28947                                             vector unsigned int)
28948         * doc/extend.texi: Fix the documentation for the built-in
28949         functions.
28951 2017-06-07  Carl Love  <cel@us.ibm.com>
28953         PR target/80982
28954         * config/rs6000/altivec.md (double<mode>2): Fix the implementation of
28955         for BE.
28957 2017-06-07  Carl Love  <cel@us.ibm.com>
28959         * config/rs6000/altivec.md: Fix argument swizzle in vec_doublel
28960         support, Generate       doublehv for signed int/float for BE case only.
28962 2017-06-07  Alexander Monakov  <amonakov@ispras.ru>
28964         * doc/invoke.texi (mcx16): Rewrite.
28966 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28968         * config/rs6000/predicates.md (rs6000_nonimmediate_operand): Delete.
28969         * config/rs6000/rs6000.md (*movsi_internal1, movsi_from_sf,
28970         *mov<mode>_softfloat, and an anonymous splitter): Use
28971         nonimmediate_operand instead of rs6000_nonimmediate_operand.
28973 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28975         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE_ACC and
28976         SPEFSCR registers.
28977         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
28978         (enum rs6000_reg_type): Delete SPE_ACC_TYPE and SPEFSCR_REG_TYPE.
28979         (rs6000_debug_reg_global): Adjust.
28980         (rs6000_init_hard_regno_mode_ok): Adjust.
28981         (rs6000_dbx_register_number): Adjust.
28982         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change to 115.
28983         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
28984         Remove SPE_ACC and SPEFSCR.
28985         (REG_ALLOC_ORDER): Ditto.
28986         (FRAME_POINTER_REGNUM): Change to 111.
28987         (enum reg_class): Remove the SPE_ACC and SPEFSCR registers.
28988         (REG_CLASS_NAMES): Ditto.
28989         (REG_CLASS_CONTENTS): Delete the SPE_ACC and SPEFSCR registers.
28990         (REGISTER_NAMES): Ditto.
28991         (ADDITIONAL_REG_NAMES): Ditto.
28992         (rs6000_reg_names): Ditto.
28993         * config/rs6000/rs6000.md: Renumber some register number
28994         define_constants.
28996 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
28998         * config/rs6000/darwin.h (REGISTER_NAMES): Delete the SPE high
28999         registers.
29000         * config/rs6000/rs6000.c (rs6000_reg_names, alt_reg_names): Ditto.
29001         * config/rs6000/rs6000.h (FIRST_PSEUDO_REGISTER): Change from 149
29002         to 117.
29003         (DWARF_REG_TO_UNWIND_COLUMN): Do not define.
29004         (FIXED_REGISTERS, CALL_USED_REGISTERS, CALL_REALLY_USED_REGISTERS):
29005         Delete the SPE high registers.
29006         (REG_ALLOC_ORDER): Ditto.
29007         (enum reg_class): Remove SPE_HIGH_REGS.
29008         (REG_CLASS_NAMES): Ditto.
29009         (REG_CLASS_CONTENTS): Delete the SPE high registers.
29010         (REGISTER_NAMES): Ditto.
29011         (rs6000_reg_names): Ditto.
29012         * doc/tm.texi.in: Remove SPE as example.
29013         * doc/tm.texi: Regenerate.
29015 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29017         * config/rs6000/8540.md (ppc8540_brinc): Delete.
29018         * config/rs6000/e500mc.md (e500mc_brinc): Delete.
29019         * config/rs6000/e500mc64.md (e500mc64_brinc): Delete.
29020         * config/rs6000/rs6000.md (type): Remove "brinc".
29022 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29024         * config.gcc (powerpc*-*-*): Don't add spe.h to extra_headers.
29025         (powerpc*-linux*spe*): Use ${cpu_type} instead of rs6000.
29026         * config/rs6000/linuxspe.h: Delete file.
29027         * config/rs6000/rs6000.md: Don't include spe.md.
29028         * config/rs6000/spe.h: Delete file.
29029         * config/rs6000/spe.md: Delete file.
29030         * config/rs6000/t-rs6000: Remove spe.md.
29032 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29034         * config/rs6000/predicates.md (reg_or_mem_operand): Reformat.
29035         (reg_or_none500mem_operand): Delete.
29036         * config/rs6000/rs6000.md (extendsfdf2): Use reg_or_mem_operand
29037         instead of reg_or_none500mem_operand.
29039 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29041         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
29042         handling of SPE flags.
29043         * config/rs6000/rs6000.opt (-mspe, -mspe=no, -mspe=yes): Delete.
29045 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29047         * config/rs6000/rs6000-common.c (rs6000_handle_option): Remove
29048         SPE ABI handling.
29049         * config/rs6000/paired.md (paired_negv2sf2): Rename to negv2sf2.
29050         (paired_absv2sf2, paired_addv2sf3, paired_subv2sf3, paired_mulv2sf3,
29051         paired_divv2sf3): Similar.
29052         * config/rs6000/predicates.md: Replace TARGET_SPE, TARGET_SPE_ABI,
29053         SPE_VECTOR_MODE and SPE_HIGH_REGNO_P by 0; simplify.
29054         * config/rs6000/rs6000-builtin.def: Delete RS6000_BUILTIN_E and
29055         RS6000_BUILTIN_S.
29056         Delete BU_SPE_1, BU_SPE_2, BU_SPE_3, BU_SPE_E, BU_SPE_P, and BU_SPE_X.
29057         Rename the paired_* instruction patterns.
29058         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Do not
29059         define __SPE__.
29060         * config/rs6000/rs6000-protos.h (invalid_e500_subreg): Delete.
29061         * config/rs6000/rs6000.c: Delete RS6000_BUILTIN_E and RS6000_BUILTIN_S.
29062         (struct rs6000_stack): Delete fields spe_gp_save_offset, spe_gp_size,
29063         spe_padding_size, and spe_64bit_regs_used.  Replace TARGET_SPE and
29064         TARGET_SPE_ABI with 0, simplify.  Replace SPE_VECTOR_MODE with
29065         PAIRED_VECTOR_MODE.
29066         (struct machine_function): Delete field spe_insn_chain_scanned_p.
29067         (spe_func_has_64bit_regs_p): Delete.
29068         (spe_expand_predicate_builtin): Delete.
29069         (spe_expand_evsel_builtin): Delete.
29070         (TARGET_DWARF_REGISTER_SPAN): Do not define.
29071         (TARGET_MEMBER_TYPE_FORCES_BLK): Do not define.
29072         (invalid_e500_subreg): Delete.
29073         (rs6000_legitimize_address): Always force_reg op2 as well, for
29074         paired single memory accesses.
29075         (rs6000_member_type_forces_blk): Delete.
29076         (rs6000_spe_function_arg): Delete.
29077         (rs6000_expand_unop_builtin): Delete SPE handling.
29078         (rs6000_expand_binop_builtin): Ditto.
29079         (spe_expand_stv_builtin): Delete.
29080         (bdesc_2arg_spe): Delete.
29081         (spe_expand_builtin): Delete.
29082         (spe_expand_predicate_builtin): Delete.
29083         (spe_expand_evsel_builtin): Delete.
29084         (rs6000_invalid_builtin): Remove RS6000_BTM_SPE handling.
29085         (spe_init_builtins): Delete.
29086         (spe_func_has_64bit_regs_p): Delete.
29087         (savres_routine_name): Delete "info" parameter.  Adjust callers.
29088         (rs6000_emit_stack_reset): Ditto.
29089         (rs6000_dwarf_register_span): Delete.
29090         * config/rs6000/rs6000.h (TARGET_SPE_ABI, TARGET_SPE,
29091         UNITS_PER_SPE_WORD, SPE_HIGH_REGNO_P, SPE_SIMD_REGNO_P,
29092         SPE_VECTOR_MODE, RS6000_BTM_SPE, RS6000_BUILTIN_E, RS6000_BUILTIN_S):
29093         Delete.
29094         * config/rs6000/rs6000.md (FIRST_SPE_HIGH_REGNO, LAST_SPE_HIGH_REGNO):
29095         Delete.
29096         * config/rs6000/rs6000.opt (-mabi=spe, -mabi=no-spe): Delete.
29097         * config/rs6000/spe.md: Delete every pattern that uses TARGET_SPE.
29098         * config/rs6000/vector.md (absv2sf2, negv2sf2, addv2sf3, subv2sf3,
29099         mulv2sf3, divv2sf3): Delete expanders.
29101 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29103         config/rs6000/rs6000.md (UNSPEC_MV_CR_GT): Delete.
29105 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29107         * config/rs6000/rs6000-protos.h (output_e500_flip_gt_bit): Delete.
29108         * config/rs6000/rs6000.c: Ditto.
29110 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29112         * config/rs6000/predicated.md (rs6000_cbranch_operator): Delete.
29113         * config/rs6000/rs6000.md: Replace rs6000_cbranch_operator by
29114         comparison_operator.
29116 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29118         * config/rs6000/rs6000.c: Remove everything related to -mfloat-gprs.
29119         * config/rs6000/rs6000.opt: Ditto.
29120         * config/rs6000/t-rtems: Ditto.
29122 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29124         * config/rs6000/predicates.md: Replace TARGET_E500_DOUBLE and
29125         TARGET_E500_SINGLE by 0, simplify.
29126         * config/rs6000/rs6000.c: Ditto.
29127         (rs6000_option_override_internal): Delete CHECK_E500_OPTIONS.
29128         (spe_build_register_parallel): Delete.
29129         * config/rs6000/rs6000.h: Delete TARGET_E500_SINGLE,
29130         TARGET_E500_DOUBLE, and CHECK_E500_OPTIONS.
29131         * config/rs6000/rs6000.md: Replace TARGET_E500_DOUBLE,
29132         TARGET_E500_SINGLE, and <E500_CONVERT> by 0, simplify.
29133         (E500_CONVERT): Delete.
29134         * config/rs6000/spe.md: Remove many patterns and all define_constants.
29136 2017-06-07  Segher Boessenkool  <segher@kernel.crashing.org>
29138         * config/rs6000/darwin.md: Replace TARGET_FPRS by 1 and simplify.
29139         * config/rs6000/dfp.md: Ditto.
29140         (negdd2, *negdd2_fpr): Merge.
29141         (absdd2, *absdd2_fpr): Merge.
29142         (negtd2, *negtd2_fpr): Merge.
29143         (abstd2, *abstd2_fpr): Merge.
29144         * config/rs6000/e500.h: Delete file.
29145         * config/rs6000/predicates.md (rs6000_cbranch_operator): Replace
29146         TARGET_FPRS by 1 and simplify.
29147         * config/rs6000/rs6000-c.c: Ditto.
29148         * config/rs6000/rs6000.c: Ditto.  Also replace TARGET_SF_SPE and
29149         TARGET_DF_SPE by 0.
29150         * config/rs6000/rs6000.h: Ditto.  Delete TARGET_SF_SPE and
29151         TARGET_DF_SPE.
29152         * config/rs6000/rs6000.md: Ditto.
29153         (floatdidf2, *floatdidf2_fpr): Merge.
29154         (move_from_CR_gt_bit): Delete.
29155         * config/rs6000/spe.md: Replace TARGET_FPRS by 1 and simplify.
29156         (E500_CR_IOR_COMPARE): Delete.
29157         (All patterns that require !TARGET_FPRS): Delete.
29158         * config/rs6000/vsx.md: Replace TARGET_FPRS by 1 and simplify.
29160 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29162         * passes.def (pass_iv_canon): Move before pass_loop_distribution.
29164 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29166         * graphds.c (add_edge): Intitialize edge's attached data.
29167         (foll_in_subgraph, dfs_fst_edge, dfs_next_edge): New function
29168         pointer parameter.  Call pointed function on each edge during
29169         graph traversing.  Skip traversing the edge when the function
29170         returns true.
29171         (graphds_dfs, graphds_scc): Ditto.
29172         (for_each_edge): New parameter.  Pass the new parameter to callback
29173         function.
29174         * graphds.h (skip_edge_callback): New function pointer type.
29175         (graphds_dfs, graphds_scc): New function pointer parameter.
29176         (graphds_edge_callback, for_each_edge): New parameter.
29178 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29180         * tree-vect-data-refs.c (vect_mark_for_runtime_alias_test): Factor
29181         out code checking if runtime alias check is possible to below ...
29182         Call the new function.
29183         * tree-data-ref.c (runtime_alias_check_p): ... to new function.
29184         * tree-data-ref.h (runtime_alias_check_p): New decalaration.
29186 2017-06-07  Marek Polacek  <polacek@redhat.com>
29188         PR sanitizer/80932
29189         * fold-const.c (extract_muldiv_1) <case MINUS_EXPR>: Add
29190         TYPE_OVERFLOW_WRAPS check.
29192 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29194         * tree-vect-loop-manip.c (vect_do_peeling): Don't skip vector loop
29195         if versioning is required.
29196         * tree-vect-loop.c (vect_analyze_loop_2): Merge niter check for loop
29197         peeling with the check for versioning.
29199 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29201         * tree-vectorizer.h (vect_build_loop_niters): New parameter.
29202         * tree-vect-loop-manip.c (vect_build_loop_niters): New parameter.
29203         Set true to new parameter if new ssa variable is defined.
29204         (vect_gen_vector_loop_niters): Refactor.  Set range information
29205         for the new vector loop bound variable.
29206         (vect_do_peeling): Ditto.
29208 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29210         * tree-affine.c (ssa.h): Include header file.
29211         (tree_to_aff_combination): Handle (T1)(X - CST) when inner type
29212         has wrapping overflow behavior.
29214 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29216         * tree-affine.c (tree_to_aff_combination): Handle (T1)(X + X).
29218 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29220         (aff_combination_expand): Move (T1)(X *+- CST) simplification to ...
29221         (tree_to_aff_combination): ... here.
29223 2017-06-07  Bin Cheng  <bin.cheng@arm.com>
29225         * tree-ssa-loop-ivopts.c (ivopts_estimate_reg_pressure): New
29226         reg_pressure model function.
29227         (ivopts_global_cost_for_size): Delete.
29228         (determine_set_costs, iv_ca_recount_cost): Call new model function
29229         ivopts_estimate_reg_pressure.
29231 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29233         * config/aarch64/aarch64.c (aarch64_rtx_costs): Make sdiv more
29234         expensive than udiv.  Remove floating point cases from mod.
29236 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29238         * config/arm/aarch-cost-tables.h (cortexa53_extra_cost):
29239         Increase idiv cost.
29241 2017-06-07  Tamar Christina  <tamar.christina@arm.com>
29243         * config/aarch64/aarch64.md
29244         (copysignsf3): Fix mask generation.
29246 2017-06-07  Jakub Jelinek  <jakub@redhat.com>
29248         * dumpfile.h (enum tree_dump_index): Rename TDI_generic to
29249         TDI_gimple.
29250         (class dump_manager): Add register_dumps method.
29251         * dumpfile.c: Include langhooks.h.
29252         (dump_files): Use 0 instead of 3/4/5 for TDI_{original,gimple,nested}.
29253         (FIRST_AUTO_NUMBERED_DUMP): Decrease to 1.
29254         (FIRST_ME_AUTO_NUMBERED_DUMP): Define.
29255         (dump_manager::dump_register): Start with 512 entries instead of 32.
29256         (dump_manager::register_dumps): New method.
29257         * toplev.c (general_init): Instead of invoking register_dumps
29258         langhook, invoke register_dumps method on the dump manager.
29259         * gimplify.c (gimplify_function_tree): Use TDI_gimple instead of
29260         TDI_generic.
29262 2017-06-07  Richard Sandiford  <richard.sandiford@linaro.org>
29264         * doc/md.texi: Clarify the restrictions on a define_insn condition.
29265         Say that # requires an associated define_split to exist, and that
29266         the define_split must be suitable for use after register allocation.
29268 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29270         * cfgbuild.c (find_bb_boundaries): Initialize profile of split blocks.
29271         (compute_outgoing_frequencies): Also initialize zero counts.
29272         (find_many_sub_basic_blocks): Do not produce uninitialized profile
29273         around loops; preserve more of profile when nothing changes.
29275 2017-06-06  Jim Wilson  <jim.wilson@linaro.org>
29277         * config/aarch64/aarch64-cost-tables.h (qdf24xx_extra_costs): Move to
29278         here.
29279         * config/arm/aarch-cost-tables.h (qdf24xx_extra_costs): From here.
29280         * config/arm/arm-cpu-cdata.h: Regenerate.
29281         * config/arm/arm-cpu-data.h, config/arm/arm-cpu.h: Likewise.
29282         * config/arm/arm-tables.opt, config/arm/arm-tune.md: Likewise.
29283         * config/arm/arm-cpus.in: Delete falkor and qdf24xx entries.
29284         * config/arm/arm.c (arm_qdf24xx_tune): Delete.
29285         * config/arm/bpabi.h (BE8_LINK_SPEC): Delete falkor and qdf24xx
29286         support.
29287         * config/arm/t-aprofile (MULTILIB_MATCHES): Delete falkor and qdf24xx
29288         support.
29289         * config/arm/t-rmprofile: Likewise.
29290         * doc/invoke.texi (ARM Options): Drop falkor and qdf24xx support.
29292 2017-06-06  David S. Miller  <davem@davemloft.net>
29294         PR target/80968
29295         * config/sparc/sparc.c (sparc_expand_prologue): Emit frame
29296         blockage if function uses alloca.
29298 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29300         * tree-ssa-loop-prefetch.c (struct mem_ref_group, struct mem_ref):
29301         New "uid" fields to hold pretty-print IDs of group and ref.
29302         Memory references are now identified as <group_id>:<ref_id>
29303         instead of using [random] addresses.
29304         (dump_mem_details): Simplify, no functional change.
29305         (dump_mem_ref): Simplify and make output more concise.
29306         Replace couple of fprintf's throughout code with calls to dump_mem_ref.
29307         (find_or_create_group): Initialize group uid.
29308         (record_ref): Initialize ref uid.  Improve debug output.
29309         (prune_group_by_reuse, should_issue_prefetch_p,)
29310         (should_issue_prefetch_p, schedule_prefetches, issue_prefetch_ref,)
29311         (mark_nontemporal_store, determine_loop_nest_reuse):
29312         Improve debug output.
29314 2017-06-06  Maxim Kuvyrkov  <maxim.kuvyrkov@linaro.org>
29316         * dbgcnt.def (prefetch): New debug counter.
29317         * tree-ssa-loop-prefetch.c (dbgcnt.h): New include.
29318         (schedule_prefetches): Stop issueing prefetches if debug counter
29319         tripped.
29321 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29323         * doc/sourcebuild.texi (Testsuites, C Language Testsuites,
29324         gcc.c-torture/compile): Remove mention of NO_LABEL_VALUES in fixme.
29326 2017-06-06  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29328         * config/aarch64/atomics.md (atomic_compare_and_swap<mode> expander):
29329         Use aarch64_reg_or_zero predicate for operand 4.
29330         (aarch64_compare_and_swap<mode> define_insn_and_split):
29331         Use aarch64_reg_or_zero predicate for operand 3.  Add 'Z' constraint.
29332         (aarch64_store_exclusive<mode>): Likewise for operand 2.
29334 2017-06-06  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29336         * config/arm/arm.c (arm_compute_save_reg_mask): Rename into ...
29337         (arm_compute_save_core_reg_mask): This.
29338         (thumb1_compute_save_reg_mask): Rename into ...
29339         (thumb1_compute_save_core_reg_mask): This.
29340         (arm_compute_save_reg0_reg12_mask): Adapt comment.
29341         (arm_compute_frame_layout): Likewise.
29343 2017-06-06  Richard Biener  <rguenther@suse.de>
29345         PR tree-optimization/80974
29346         * tree-ssa-sccvn.c (set_ssa_val_to): Do not change but only
29347         keep or clear leaders SSA info.
29349 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29351         * config/nvptx/nvptx.c (split_mode_p): New function.
29352         (nvptx_declare_function_name, nvptx_print_operand): Use split_mode_p.
29354 2017-06-06  Tom de Vries  <tom@codesourcery.com>
29356         * config/nvptx/nvptx.c (nvptx_print_operand): Use maybe_split_mode.
29358 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29360         PR bootstrap/80978
29361         * tree-cfg.c (execute_fixup_cfg): Fix condition on when to rescale
29362         profile.
29364 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29366         * shrink-wrap.c (handle_simple_exit): Update profile.
29367         (try_shrink_wrapping): Upate profile.
29369 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29371         * predict.c (tree_estimate_probability_bb): Add LOCAL_ONLY.
29372         (tree_guess_outgoing_edge_probabilities): New.
29373         * predict.h (tree_guess_outgoing_edge_probabilities): Declare.
29374         * tree-cfg.c (gimple_find_sub_bbs): Propagate profile.
29376 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29378         * ipa-split.c (split_function): Initialize return bb profile.
29380 2017-06-06  Jan Hubicka  <hubicka@ucw.cz>
29382         * profile.c (compute_branch_probabilities): Also initialize
29383         EXIT_BLOCK profile.
29385 2017-06-06  Richard Biener  <rguenther@suse.de>
29387         PR tree-optimization/80928
29388         * tree-vect-loop.c (vect_update_vf_for_slp): Amend dumps.
29389         (vect_analyze_loop_operations): Properly guard analysis for
29390         pure SLP case.
29391         (vect_transform_loop): Likewise.
29392         (vect_analyze_loop_2): Also reset SLP type on PHIs.
29393         (vect_model_induction_cost): Do not cost for pure SLP.
29394         (vectorizable_induction): Pass in SLP node, implement SLP vectorization
29395         of induction in inner loop vectorization.
29396         * tree-vect-slp.c (vect_create_new_slp_node): Handle PHIs.
29397         (vect_get_and_check_slp_defs): Handle vect_induction_def.
29398         (vect_build_slp_tree): Likewise.  Handle PHIs as terminating the
29399         recursion.
29400         (vect_analyze_slp_cost_1): Cost induction.
29401         (vect_detect_hybrid_slp_stmts): Handle PHIs.
29402         (vect_get_slp_vect_defs): Likewise.
29403         * tree-vect-stmts.c (vect_analyze_stmt): Handle induction.
29404         (vect_transform_stmt): Handle SLP reductions.
29405         * tree-vectorizer.h (vectorizable_induction): Adjust.
29407 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29409         * config/rs6000/rs6000.c (make_resolver_func): Update
29410         init_lowered_empty_function call.
29412 2017-06-05  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29414         * doc/invoke.texi: Document the -fprofile-abs-path option.
29415         * common.opt (fprofile-abs-path): New option.
29416         * gcov-io.h (gcov_write_filename): Declare.
29417         * gcov-io.c (gcov_write_filename): New function.
29418         * coverage.c (coverage_begin_function): Use gcov_write_filename.
29419         * profile.c (output_location): Likewise.
29421 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29423         * shring-wrap.c: Revert accidental commit.
29425 2017-06-05  Volker Reichelt  <v.reichelt@netcologne.de>
29427         * doc/invoke.texi (-Wduplicated-branches): Add to warning list.
29429 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29431         * cfgexpand.c (expand_gimple_tailcall): Initialize profile of
29432         new edge.
29433         * ipa-inline.c (want_inline_self_recursive_call_p): Watch for missing
29434         profile in callgraph edge.
29435         * profile-count.h (apply_probability): If THIS is 0, then result is 0
29436         (apply_scale): Likewise.
29437         * tree-inline.c (copy_bb, copy_edges_for_bb, copy_cfg_body):
29438         Also scale profile when inlining function with zero profile.
29439         (initialize_cfun): Update exit block profile even when it is zero.
29440         * tree-ssa-threadupdate.c (clear_counts_path): Handle correctly case
29441         when profile is read.
29443 2017-06-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
29445         * config/rs6000/rs6000.c (toplevel): Include attribs.h.
29446         (CLONE_*): New constants to define the processors we can generate
29447         code for with the target_clone attribute.
29448         (rs6000_clone_map): New array to identify which clone processors
29449         the current program is running on.
29450         (TARGET_COMPARE_VERSION_PRIORITY): Define to enable the
29451         target_clone attribute.
29452         (TARGET_GENERATE_VERSION_DISPATCHER_BODY): Likewise.
29453         (TARGET_GET_FUNCTION_VERSIONS_DISPATCHER): Likewise.
29454         (TARGET_OPTION_FUNCTION_VERSIONS): Likewise.
29455         (cpu_expand_builtin): Add support for target_clone attribute.
29456         (rs6000_valid_attribute_p): Allow "default" attribute.
29457         (get_decl_name): New debug function to simplify printing the
29458         current function name in debugging statements.
29459         (rs6000_clone_priority): New functions to support the target_clone
29460         attribute, and be able to generate code to switch between ISA 2.05
29461         through ISA 3.0 (power6 through power9).
29462         (rs6000_compare_version_priority): Likewise.
29463         (rs6000_get_function_versions_dispatcher): Likewise.
29464         (make_resolver_func): Likewise.
29465         (add_condition_to_bb): Likewise.
29466         (dispatch_function_versions): Likewise.
29467         (rs6000_generate_version_dispatcher_body): Likewise.
29468         (rs6000_can_inline_p): Call get_decl_name for debugging usage.
29469         (fusion_gpr_load_p): Fix a spacing issue.
29470         * doc/extend.texi (Common Function Attributes): Document that the
29471         PowerPC supports the target_clone attribute.
29473 2017-06-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
29475         * config/arm/arm.h: explain F symbol found in description of ARM
29476         register allocation in its legend.
29478 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29480         * config/mips/frame-header-opt.c: Include profile-count.h.
29481         * config/riscv/riscv.c: Include profile-count.h
29483 2017-06-05  Jan Hubicka  <hubicka@ucw.cz>
29485         * tree-ssa-loop-im.c (execute_sm_if_changed): Add FLAG_BBS parameter;
29486         update profile.
29487         (sm_set_flag_if_changed): Add bbs field.
29488         (execute_sm_if_changed_flag_set): Pass BBS.
29489         (execute_sm): Update.
29491 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29493         * config/aarch64/aarch64-simd.md (aarch64_store_lane0<mode>):
29494         New pattern.
29496 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29498         * config/aarch64/aarch64.md (sub<mode>3_compare1_imm): New define_insn.
29499         (peephole2): New peephole2 to emit the above.
29500         * config/aarch64/predicates.md (aarch64_sub_immediate): New predicate.
29502 2017-06-05  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29504         * config/aarch64/aarch64.c (define_peephole2 above
29505         *sub_<shift>_<mode>): New peephole.
29507 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
29509         * config/i386/i386.c (make_resolver_func): Update.
29510         * Makefile.in: Add profile-count.h and profile-count.o
29511         * auto-profile.c (afdo_indirect_call): Update to new API.
29512         (afdo_set_bb_count): Update.
29513         (afdo_propagate_edge): Update.
29514         (afdo_propagate_circuit): Update.
29515         (afdo_calculate_branch_prob): Update.
29516         (afdo_annotate_cfg): Update.
29517         * basic-block.h: Include profile-count.h
29518         (struct edge_def): Turn count to profile_count.
29519         (struct basic_block_def): Likewie.
29520         (REG_BR_PROB_BASE): Move to profile-count.h
29521         (RDIV): Move to profile-count.h
29522         * bb-reorder.c (max_entry_count): Turn to profile_count.
29523         (find_traces): Update.
29524         (rotate_loop):Update.
29525         (connect_traces):Update.
29526         (sanitize_hot_paths):Update.
29527         * bt-load.c (migrate_btr_defs): Update.
29528         * cfg.c (RDIV): Remove.
29529         (init_flow): Use alloc_block.
29530         (alloc_block): Uninitialize count.
29531         (unchecked_make_edge): Uninitialize count.
29532         (check_bb_profile): Update.
29533         (dump_edge_info): Update.
29534         (dump_bb_info): Update.
29535         (update_bb_profile_for_threading): Update.
29536         (scale_bbs_frequencies_int): Update.
29537         (scale_bbs_frequencies_gcov_type): Update.
29538         (scale_bbs_frequencies_profile_count): New.
29539         * cfg.h (update_bb_profile_for_threading): Update.
29540         (scale_bbs_frequencies_profile_count): Declare.
29541         * cfgbuild.c (compute_outgoing_frequencies): Update.
29542         (find_many_sub_basic_blocks): Update.
29543         * cfgcleanup.c (try_forward_edges): Update.
29544         (try_crossjump_to_edge): Update.
29545         * cfgexpand.c (expand_gimple_tailcall): Update.
29546         (construct_exit_block): Update.
29547         * cfghooks.c (verify_flow_info): Update.
29548         (dump_bb_for_graph): Update.
29549         (split_edge): Update.
29550         (make_forwarder_block): Update.
29551         (duplicate_block): Update.
29552         (account_profile_record): Update.
29553         * cfgloop.c (find_subloop_latch_edge_by_profile): Update.
29554         (get_estimated_loop_iterations): Update.
29555         * cfgloopanal.c (expected_loop_iterations_unbounded): Update.
29556         (single_likely_exit): Update.
29557         * cfgloopmanip.c (scale_loop_profile): Update.
29558         (loopify): Update.
29559         (set_zero_probability): Update.
29560         (lv_adjust_loop_entry_edge): Update.
29561         * cfgrtl.c (force_nonfallthru_and_redirect): Update.
29562         (purge_dead_edges): Update.
29563         (rtl_account_profile_record): Update.
29564         * cgraph.c (cgraph_node::create): Uninitialize count.
29565         (symbol_table::create_edge): Uninitialize count.
29566         (cgraph_update_edges_for_call_stmt_node): Update.
29567         (cgraph_edge::dump_edge_flags): Update.
29568         (cgraph_node::dump): Update.
29569         (cgraph_edge::maybe_hot_p): Update.
29570         * cgraph.h: Include profile-count.h
29571         (create_clone), create_edge, create_indirect_edge): Update.
29572         (cgraph_node): Turn count to profile_count.
29573         (cgraph_edge0: Likewise.
29574         (make_speculative, clone): Update.
29575         (create_edge): Update.
29576         (init_lowered_empty_function): Update.
29577         * cgraphclones.c (cgraph_edge::clone): Update.
29578         (duplicate_thunk_for_node): Update.
29579         (cgraph_node::create_clone): Update.
29580         * cgraphunit.c (cgraph_node::analyze): Update.
29581         (cgraph_node::expand_thunk): Update.
29582         * final.c (dump_basic_block_info): Update.
29583         * gimple-streamer-in.c (input_bb): Update.
29584         * gimple-streamer-out.c (output_bb): Update.
29585         * graphite.c (print_global_statistics): Update.
29586         (print_graphite_scop_statistics): Update.
29587         * hsa-brig.c: Include basic-block.h.
29588         * hsa-dump.c: Include basic-block.h.
29589         * hsa-gen.c (T sum_slice): Update.
29590         (convert_switch_statements):Update.
29591         * hsa-regalloc.c: Include basic-block.h.
29592         * ipa-chkp.c (chkp_produce_thunks): Update.
29593         * ipa-cp.c (struct caller_statistics): Update.
29594         (init_caller_stats): Update.
29595         (gather_caller_stats): Update.
29596         (ipcp_cloning_candidate_p): Update.
29597         (good_cloning_opportunity_p): Update.
29598         (get_info_about_necessary_edges): Update.
29599         (dump_profile_updates): Update.
29600         (update_profiling_info): Update.
29601         (update_specialized_profile): Update.
29602         (perhaps_add_new_callers): Update.
29603         (decide_about_value): Update.
29604         (ipa_cp_c_finalize): Update.
29605         * ipa-devirt.c (struct odr_type_warn_count): Update.
29606         (struct decl_warn_count): Update.
29607         (struct final_warning_record): Update.
29608         (possible_polymorphic_call_targets): Update.
29609         (ipa_devirt): Update.
29610         * ipa-fnsummary.c (redirect_to_unreachable): Update.
29611         * ipa-icf.c (sem_function::merge): Update.
29612         * ipa-inline-analysis.c (do_estimate_edge_time): Update.
29613         * ipa-inline.c (compute_uninlined_call_time): Update.
29614         (compute_inlined_call_time): Update.
29615         (want_inline_small_function_p): Update.
29616         (want_inline_self_recursive_call_p): Update.
29617         (edge_badness): Update.
29618         (lookup_recursive_calls): Update.
29619         (recursive_inlining): Update.
29620         (inline_small_functions): Update.
29621         (dump_overall_stats): Update.
29622         (dump_inline_stats): Update.
29623         * ipa-profile.c (ipa_profile_generate_summary): Update.
29624         (ipa_propagate_frequency): Update.
29625         (ipa_profile): Update.
29626         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
29627         * ipa-utils.c (ipa_merge_profiles): Update.
29628         * loop-doloop.c (doloop_modify): Update.
29629         * loop-unroll.c (report_unroll): Update.
29630         (unroll_loop_runtime_iterations): Update.
29631         * lto-cgraph.c (lto_output_edge): Update.
29632         (lto_output_node): Update.
29633         (input_node): Update.
29634         (input_edge): Update.
29635         (merge_profile_summaries): Update.
29636         * lto-streamer-in.c (input_cfg): Update.
29637         * lto-streamer-out.c (output_cfg): Update.
29638         * mcf.c (create_fixup_graph): Update.
29639         (adjust_cfg_counts): Update.
29640         (sum_edge_counts): Update.
29641         * modulo-sched.c (sms_schedule): Update.
29642         * postreload-gcse.c (eliminate_partially_redundant_load): Update.
29643         * predict.c (maybe_hot_count_p): Update.
29644         (probably_never_executed): Update.
29645         (dump_prediction): Update.
29646         (combine_predictions_for_bb): Update.
29647         (propagate_freq): Update.
29648         (handle_missing_profiles): Update.
29649         (counts_to_freqs): Update.
29650         (rebuild_frequencies): Update.
29651         (force_edge_cold): Update.
29652         * predict.h: Include profile-count.h
29653         (maybe_hot_count_p, counts_to_freqs): UPdate.
29654         * print-rtl-function.c: Do not include cfg.h
29655         * print-rtl.c: Include basic-block.h
29656         * profile-count.c: New file.
29657         * profile-count.h: New file.
29658         * profile.c (is_edge_inconsistent): Update.
29659         (correct_negative_edge_counts): Update.
29660         (is_inconsistent): Update.
29661         (set_bb_counts): Update.
29662         (read_profile_edge_counts): Update.
29663         (compute_frequency_overlap): Update.
29664         (compute_branch_probabilities): Update; Initialize and deinitialize
29665         gcov_count tables.
29666         (branch_prob): Update.
29667         * profile.h (bb_gcov_counts, edge_gcov_counts): New.
29668         (edge_gcov_count): New.
29669         (bb_gcov_count): New.
29670         * shrink-wrap.c (try_shrink_wrapping): Update.
29671         * tracer.c (better_p): Update.
29672         * trans-mem.c (expand_transaction): Update.
29673         (ipa_tm_insert_irr_call): Update.
29674         (ipa_tm_insert_gettmclone_call): Update.
29675         * tree-call-cdce.c: Update.
29676         * tree-cfg.c (gimple_duplicate_sese_region): Update.
29677         (gimple_duplicate_sese_tail): Update.
29678         (gimple_account_profile_record): Update.
29679         (execute_fixup_cfg): Update.
29680         * tree-inline.c (copy_bb): Update.
29681         (copy_edges_for_bb): Update.
29682         (initialize_cfun): Update.
29683         (freqs_to_counts): Update.
29684         (copy_cfg_body): Update.
29685         (expand_call_inline): Update.
29686         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): Update.
29687         * tree-ssa-loop-ivcanon.c (unloop_loops): Update.
29688         (try_unroll_loop_completely): Update.
29689         (try_peel_loop): Update.
29690         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Update.
29691         * tree-ssa-loop-niter.c (estimate_numbers_of_iterations_loop): Update.
29692         * tree-ssa-loop-split.c (connect_loops): Update.
29693         * tree-ssa-loop-unswitch.c (hoist_guard): Update.
29694         * tree-ssa-reassoc.c (branch_fixup): Update.
29695         * tree-ssa-tail-merge.c (replace_block_by): Update.
29696         * tree-ssa-threadupdate.c (create_block_for_threading): Update.
29697         (compute_path_counts): Update.
29698         (update_profile): Update.
29699         (recompute_probabilities): Update.
29700         (update_joiner_offpath_counts): Update.
29701         (estimated_freqs_path): Update.
29702         (freqs_to_counts_path): Update.
29703         (clear_counts_path): Update.
29704         (ssa_fix_duplicate_block_edges): Update.
29705         (duplicate_thread_path): Update.
29706         * tree-switch-conversion.c (case_bit_test_cmp): Update.
29707         (struct switch_conv_info): Update.
29708         * tree-tailcall.c (decrease_profile): Update.
29709         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Update.
29710         * tree-vect-loop.c (scale_profile_for_vect_loop): Update.
29711         * value-prof.c (check_counter): Update.
29712         (gimple_divmod_fixed_value): Update.
29713         (gimple_mod_pow2): Update.
29714         (gimple_mod_subtract): Update.
29715         (gimple_ic_transform): Update.
29716         (gimple_stringop_fixed_value): Update.
29717         * value-prof.h (gimple_ic): Update.
29719 2017-06-02  Carl Love  <cel@us.ibm.com>
29721         * config/rs6000/rs6000-c: Add support for built-in functions
29722         vector double vec_doublee (vector signed int);
29723         vector double vec_doublee (vector unsigned int);
29724         vector double vec_doublee (vector float);
29725         vector double vec_doubleh (vector signed int);
29726         vector double vec_doubleh (vector unsigned int);
29727         vector double vec_doubleh (vector float);
29728         vector double vec_doublel (vector signed int);
29729         vector double vec_doublel (vector unsigned int);
29730         vector double vec_doublel (vector float);
29731         vector double vec_doubleo (vector signed int);
29732         vector double vec_doubleo (vector unsigned int);
29733         vector double vec_doubleo (vector float);.
29734         * config/rs6000/rs6000-builtin.def: Add definitions for DOUBLEE,
29735         DOUBLEO, DOUBLEH, DOUBLEL, UNS_DOUBLEO, UNS_DOUBLEE, UNS_DOUBLEH,
29736         UNS_DOUBLEL.
29737         * config/rs6000/altivec.md: Add code generator for doublee<mode>2,
29738         unsdoubleev4si2, doubleo<mode>2, unsdoubleov4si2, doubleh<mode>2,
29739         unsdoublehv4si2, doublel<mode>2, unsdoublelv4si2, add mode attribute
29740         VS_sxwsp.
29741         * config/rs6000/altivec.h: Add define for vec_doublee, vec_doubleo,
29742         vec_doublel, vec_doubleh.
29743         * doc/extend.texi: Update the built-in documentation file for the
29744         new built-in functions.
29746 2017-06-02  David Malcolm  <dmalcolm@redhat.com>
29748         PR jit/80954
29749         * ipa-inline-analysis.c (free_growth_caches): Set
29750         edge_removal_hook_holder to NULL after removing it.
29752 2017-06-02  Sudakshina Das  <sudi.das@arm.com>
29754         * config/aarch64/aarch64.c (aarch64_select_cc_mode): Return CC_SWP for
29755         comparision with zero.
29757 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29758         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29759         for early expansion of vec_min and vec_max builtins.
29760         (builtin_function_type): Add min/max unsigned variants to those
29761         identified as having unsigned arguments.
29763 2017-06-02  Olivier Hainque  <hainque@adacore.com>
29765         * config/vx-common.h (DWARF_UNWIND_INFO): Switch #define to 1.
29767 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29769         * config/aarch64/aarch64-simd.md (*aarch64_simd_vec_copy_lane<mode>):
29770         Use VALL_F16 iterator rather than VALL.
29772 2017-06-02  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
29774         * config/aarch64/aarch64.c (aarch64_split_compare_and_swap):
29775         Emit CBNZ inside loop when doing a strong exchange and comparing
29776         against zero.  Generate the CC flags after the loop.
29778 2017-06-02  David Edelsohn  <dje.gcc@gmail.com>
29780         * dwarf2out.c (DWARF_INITIAL_LENGTH_SIZE_STR): New.
29781         (dl_section_ref): New.
29782         (dwarf2out_finish): Copy debug_line_section_label to dl_section_ref.
29783         On AIX, append an expression to subtract the size of the
29784         section length to dl_section_ref.
29786 2017-06-02  Will Schmidt  <will_schmidt@vnet.ibm.com>
29788         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling
29789         for early expansion of vector absolute builtins.
29791 2017-06-02  Richard Biener  <rguenther@suse.de>
29793         * tree-vect-slp.c (vect_detect_hybrid_slp_2): Match up
29794         what we consider a relevant use stmt with vect_detect_hybrid_slp_stmts.
29796 2017-06-02  Richard Biener  <rguenther@suse.de>
29798         PR tree-optimization/80948
29799         * tree-tailcall.c (find_tail_calls): Track stmts to move in
29800         stmt order as well.
29802 2017-06-02  Richard Biener  <rguenther@suse.de>
29804         * tree-vect-loop.c (vect_analyze_loop_operations): Not relevant
29805         PHIs are ok.
29806         * tree-vect-stmts.c (process_use): Do not mark backedge defs
29807         for inductions as relevant.
29809 2017-06-02  Richard Biener  <rguenther@suse.de>
29811         * tree-vect-loop.c (get_initial_def_for_induction): Inline into ...
29812         (vectorizable_induction): ... this.  Remove dead code.
29814 2017-06-02  Eric Botcazou  <ebotcazou@adacore.com>
29816         * builtins. (expand_builtin_alloca): Remove second parameter and
29817         infer its value from the first parameter instead.
29818         (expand_builtin) <BUILT_IN_ALLOCA>: Adjust call to above.
29820 2017-06-02  Jakub Jelinek  <jakub@redhat.com>
29822         PR rtl-optimization/80903
29823         * loop-doloop.c (add_test): Unshare sequence.
29825 2017-06-02  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29827         * doc/invoke.texi: Document the -Wsizeof-pointer-div warning.
29829 2017-06-01  Bernd Edlinger  <bernd.edlinger@hotmail.de>
29831         * config/i386/i386.c (x86_64_ms_sysv_extra_clobbered_registers): Make
29832         static.
29833         (xlogue_layout::get_stack_space_used, xlogue_layout::s_instances,
29834         xlogue_layout::get_instance, logue_layout::xlogue_layout,
29835         sp_valid_at, fp_valid_at, choose_basereg): Formatting.
29836         (xlogue_layout::get_stub_rtx): Make static.
29837         (xlogue_layout::get_stub_name): Avoid const-cast, make static.
29838         (xlogue_layout::compute_stub_managed_regs): Rename to...
29839         (xlogue_layout::count_stub_managed_regs): ...this.
29840         (xlogue_layout::is_stub_managed_reg): New function.
29841         (xlogue_layout::m_stub_names): Rename to...
29842         (xlogue_layout::s_stub_names): ...this, make static.
29843         (xlogue_layout::STUB_INDEX_OFFSET, xlogue_layout::MIN_REGS,
29844         xlogue_layout::MAX_REGS, xlogue_layout::MAX_EXTRA_REGS,
29845         xlogue_layout::VARIANT_COUNT, xlogue_layout::STUB_NAME_MAX_LEN,
29846         xlogue_layout::s_stub_names): Instantiate statics.
29847         (stub_managed_regs): Remove.
29848         (ix86_save_reg): Use xlogue_layout::compute_stub_managed_regs.
29849         (disable_call_ms2sysv_xlogues): Rename to...
29850         (warn_once_call_ms2sysv_xlogues): ...this, and warn only once.
29851         (ix86_initial_elimination_offset, ix86_expand_call): Fix call_ms2sysv
29852         warning logic.
29853         (ix86_static_chain): Make sure that ix86_static_chain_on_stack can't
29854         change after reload_completed.
29855         (ix86_can_use_return_insn_p): Use the ix86_frame data structure
29856         directly.
29857         (ix86_expand_prologue): Likewise.
29858         (ix86_expand_epilogue): Likewise.
29859         (ix86_expand_split_stack_prologue): Likewise.
29860         (ix86_compute_frame_layout): Remove frame parameter ...
29861         (TARGET_COMPUTE_FRAME_LAYOUT): ... and export it as a target hook.
29862         (ix86_finalize_stack_realign_flags): Call ix86_compute_frame_layout
29863         only if necessary.
29864         (ix86_init_machine_status): Don't set use_fast_prologue_epilogue_nregs.
29865         (ix86_frame): Move from here ...
29866         * config/i386/i386.h (ix86_frame): ... to here.
29867         (machine_function): Remove use_fast_prologue_epilogue_nregs, cache the
29868         complete ix86_frame data structure instead.  Remove some_ld_name.
29870 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29872         * dwarf2out.c (dwarf2out_late_global_decl): Add locations for
29873         symbols that hold a DECL_VALUE_EXPR.
29875 2017-06-01  Martin Jambor  <mjambor@suse.cz>
29877         PR tree-optimization/80898
29878         * tree-sra.c (process_subtree_disqualification): Removed.
29879         (disqualify_candidate): Do not acll
29880         process_subtree_disqualification.
29881         (subtree_mark_written_and_enqueue): New function.
29882         (propagate_all_subaccesses): Set grp_write of LHS subtree if the
29883         RHS has been disqualified and re-queue LHS if necessary.  Apart
29884         from that, ignore disqualified RHS.
29886 2017-06-01  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
29888         * config/s390/s390.c (s390_emit_epilogue): Disable early return
29889         address fetch for z10 or later.
29891 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29893         * config/arc/arc.md (tst_movb): Add guard when splitting.
29895 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29897         * config/arc/arc.c (arc_can_eliminate): Test against
29898         arc_frame_pointer_needed.
29900 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29902         * config/arc/arc.c (arc_expand_prologue): Emit a special barrier
29903         to prevent store reordering.
29904         * config/arc/arc.md (UNSPEC_ARC_STKTIE): Define.
29905         (type): Add block type.
29906         (stack_tie): Define special instruction to be used in
29907         expand_prologue.
29909 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29911         * config/arc/arc.md (commutative_binary_comparison): Remove 'I'
29912         constraint. It is not valid for the pattern.
29913         (noncommutative_binary_comparison): Likewise.
29915 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29917         * config/arc/simdext.md (movv2hi_insn): Change predicate to avoid
29918         scaled addresses.
29920 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29922         * config/arc/arc.c (arc_conditional_register_usage): Allow r30 to
29923         be used by the reg-alloc.
29925 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29927         * config/arc/arc.md (mulsi3): Avoid use of hard registers before
29928         reg-alloc when having mul64 or mul32x16 instructions.
29929         (mulsidi3): Likewise.
29930         (umulsidi3): Likewise.
29931         (mulsi32x16): New pattern.
29932         (mulsi64): Likewise.
29933         (mulsidi64): Likewise.
29934         (umulsidi64): Likewise.
29935         (MUL32x16_REG): Define.
29936         (mul64_600): Use MUL32x16_REG.
29937         (mac64_600): Likewise.
29938         (umul64_600): Likewise.
29939         (umac64_600): Likewise.
29941 2017-06-01  Claudiu Zissulescu  <claziss@synopsys.com>
29943         * config/arc/arc.md (mulsi3_700): Make it commutative.
29945 2017-06-01  Jose E. Marchesi  <jose.marchesi@oracle.com>
29947         * config/sparc/sparc.md (*zero_extendsidi2_insn_sp64): Set insn
29948         type for movstouw.
29949         (*sign_extendsidi2_insn): Likewise for movstosw.
29951 2017-06-01  Pierre-Marie de Rodat  <derodat@adacore.com>
29953         * dwarf2out.c (get_discr_value): Call the get_debug_type hook on
29954         the type of the input discriminant value.  Convert the
29955         discriminant value of signedness vary.
29957 2017-06-01  Volker Reichelt  <v.reichelt@netcologne.de>
29959         * doc/invoke.texi (-Wcatch-value): Document new shortcut.
29960         Add to -Wall section.
29962 2017-06-01  Richard Biener  <rguenther@suse.de>
29964         PR middle-end/66313
29965         * fold-const.c (fold_plusminus_mult_expr): If the factored
29966         factor may be zero use a wrapping type for the inner operation.
29967         * tree-tailcall.c (independent_of_stmt_p): Pass in to_move bitmap
29968         and handle moved defs.
29969         (process_assignment): Properly guard the unary op case.  Return a
29970         tri-state indicating that moving the stmt before the call may allow
29971         to continue.  Pass through to_move.
29972         (find_tail_calls): Handle moving unrelated defs before
29973         the call.
29975 2017-05-31  Segher Boessenkool  <segher@kernel.crashing.org>
29977         PR target/80618
29978         * config/rs6000/vector.md (*vector_uneq<mode>): Write the nor in the
29979         splitter result in the canonical way.
29981 2017-05-31  Uros Bizjak  <ubizjak@gmail.com>
29983         * config/i386/i386.md (*zero_extendsidi2): Enable alternative (?r, *Yj)
29984         also for 32bit target.  Update insn attributes.
29985         (zero-extendsidi2 splitter): Allow all registers for operand 1.
29987 2017-05-31  Sebastian Peryt  <sebastian.peryt@intel.com>
29989         * config/i386/avx512fintrin.h (_mm_mask_max_sd)
29990         (_mm_maskz_max_sd, _mm_mask_max_ss, _mm_maskz_max_ss)
29991         (_mm_mask_min_sd, _mm_maskz_min_sd, _mm_mask_min_ss)
29992         (_mm_maskz_min_ss): New intrinsics.
29994 2017-05-31  Martin Liska  <mliska@suse.cz>
29996         * tree-vect-loop.c (vect_create_epilog_for_reduction):
29997         Change comment style to one we normally use.
29998         (vectorizable_reduction): Likewise.
29999         (vectorizable_induction): Likewise.
30000         * tree-vect-stmts.c (vectorizable_mask_load_store): Likewise.
30001         (vectorizable_call): Likewise.
30002         (vectorizable_simd_clone_call): Likewise.
30003         (vectorizable_conversion): Likewise.
30004         (vectorizable_assignment): Likewise.
30005         (vectorizable_shift): Likewise.
30006         (vectorizable_operation): Likewise.
30007         (vectorizable_store): Likewise.
30008         (vectorizable_load): Likewise.
30009         * tree-vectorizer.h: Likewise.
30011 2017-05-31  Alexander Monakov  <amonakov@ispras.ru>
30013         * passes.c (emergency_dump_function): New.
30014         * tree-pass.h (emergency_dump_function): Declare.
30015         * plugin.c (plugins_internal_error_function): Remove.
30016         * plugin.h (plugins_internal_error_function): Remove declaration.
30017         * toplev.c (internal_error_function): New static function.  Use it...
30018         (general_init): ...here.
30020 2017-05-31  Graham Markall  <graham.markall@embecosm.com>
30022         * config/arc/arc.c (arc_print_operand): Handle constant operands.
30023         (arc_rtx_costs): Add costs for new patterns.
30024         * config/arc/arc.md: Additional *add_n and *sub_n patterns.
30025         * config/arc/predicates.md: Add _1_2_3_operand predicate.
30027 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30029         * tree-ssa-strlen.c (get_next_strinfo): New function.
30030         (get_stridx_plus_constant): Use it.
30031         (zero_length_string): Likewise.
30032         (adjust_related_strinfos): Likewise.
30033         (adjust_last_stmt): Likewise.
30035 2017-05-31  Richard Biener  <rguenther@suse.de>
30037         PR target/80880
30038         * config/i386/i386.c (ix86_expand_builtin): Remove assert
30039         for arg being an SSA name when expanding IX86_BUILTIN_BNDRET.
30041 2017-05-31  Richard Sandiford  <richard.sandiford@linaro.org>
30043         * tree-vect-data-refs.c (vect_find_same_alignment_drs): Remove
30044         loop_vinfo argument and use of dependence distance vectors.
30045         Check instead whether the two references differ only in their
30046         initial value and assume that they have the same alignment if the
30047         difference is a multiple of the vector alignment.
30048         (vect_analyze_data_refs_alignment): Update call accordingly.
30050 2017-05-31  Martin Liska  <mliska@suse.cz>
30052         PR target/79155
30053         * config/i386/cpuid.h: Fix typo in a comment in cpuid.h.
30055 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30057         * tree-vect-loop-manip.c (create_intersect_range_checks_index)
30058         (create_intersect_range_checks): Move from ...
30059         * tree-data-ref.c (create_intersect_range_checks_index)
30060         (create_intersect_range_checks): ... to here.
30061         (create_runtime_alias_checks): New function factored from ...
30062         * tree-vect-loop-manip.c (vect_create_cond_for_alias_checks): ...
30063         here.  Call above function.
30064         * tree-data-ref.h (create_runtime_alias_checks): New function.
30066 2017-05-31  Bin Cheng  <bin.cheng@arm.com>
30068         * tree-data-ref.c (prune_runtime_alias_test_list): Relax minimal
30069         segment length for dr_b and compute it in wide_int.
30071 2017-05-31  Richard Biener  <rguenther@suse.de>
30073         PR tree-optimization/80906
30074         * graphite-isl-ast-to-gimple.c (copy_loop_close_phi_nodes): Get
30075         and pass through iv_map.
30076         (copy_bb_and_scalar_dependences): Adjust.
30077         (translate_pending_phi_nodes): Likewise.
30078         (copy_loop_close_phi_args): Handle code-generating IVs instead
30079         of ICEing.
30081 2017-05-30  David Malcolm  <dmalcolm@redhat.com>
30083         * diagnostic-color.c (color_dict): Add "type-diff".
30084         (parse_gcc_colors): Update comment.
30085         * doc/invoke.texi (Diagnostic Message Formatting Options): Add
30086         -fdiagnostics-show-template-tree and -fno-elide-type.
30087         (GCC_COLORS): Add type-diff to example.
30088         (type-diff=): New.
30089         (-fdiagnostics-show-template-tree): New.
30090         (-fno-elide-type): New.
30091         * pretty-print.c (pp_format): Pass quote and formatters[argno] to
30092         the pp_format_decoder callback.  Call any m_format_postprocessor's
30093         "handle" method.
30094         (pretty_printer::pretty_printer): Initialize
30095         m_format_postprocessor.
30096         (pretty_printer::~pretty_printer): Delete any
30097         m_format_postprocessor.
30098         * pretty-print.h (printer_fn): Add bool and const char ** parameters.
30099         (class format_postprocessor): New class.
30100         (struct pretty_printer::format_decoder): Document the new parameters.
30101         (struct pretty_printer::m_format_postprocessor): New field.
30102         * tree-diagnostic.c (default_tree_printer): Update for new
30103         bool and const char ** params.
30104         * tree-diagnostic.h (default_tree_printer): Likewise.
30106 2017-05-30  Segher Boessenkool  <segher@kernel.crashing.org>
30108         * config/rs6000/predicates.md (cc_reg_not_micro_cr0_operand): Delete.
30109         (lwa_operand): Delete rs6000_gen_cell_microcode test.
30110         * config/rs6000/rs6000.c (rs6000_option_override_internal): Delete
30111         rs6000_gen_cell_microcode code.
30112         (rs6000_final_prescan_insn): Delete.
30113         (rs6000_opt_vars): Delete the "gen-cell-microcode" and
30114         "warn-cell-microcode" entries.
30115         * config/rs6000/rs6000.h (FINAL_PRESCAN_INSN): Delete.
30116         * config/rs6000/rs6000.md: Delete rs6000_gen_cell_microcode tests
30117         throughout.  Change cc_reg_not_micro_cr0_operand to
30118         cc_reg_not_cr0_operand throughout.
30119         (*extendhi<mode>2_noload): Delete.
30120         * config/rs6000/rs6000.opt (mgen-cell-microcode): Replace by stub.
30121         (mwarn-cell-microcode): Delete.
30122         * doc/invoke.texi (RS/6000 and PowerPC Options): Delete
30123         -mgen-cell-microcode and -mwarn-cell-microcode.
30125 2017-05-30  Uros Bizjak  <ubizjak@gmail.com>
30127         PR target/80833
30128         * config/i386/constraints.md (Yd): New constraint.
30129         (Ye): Ditto.
30130         * config/i386/i386.md (*movti_internal): Add (?r, Ye)
30131         and (?Yd, r) alternatives.  Update insn attributes.
30132         * config/i386/i386.md (*movti_internal): Add (?r, *Ye)
30133         and (?*Yd, r) alternatives.  Update insn attributes.
30134         (double-mode inter-unit splitters): Add new GR<->XMM splitters.
30136 2017-05-30  Pierre-Marie de Rodat  <derodat@adacore.com>
30138         * gimplify.c (gimplify_modify_expr): Don't create a
30139         DECL_DEBUG_EXPR link if *FROM_P does not belong to the current
30140         function.
30142 2017-05-30  Wilco Dijkstra  <wdijkstr@arm.com>
30144         * config/arm/arm-builtins.c (arm_expand_builtin): Remove const.
30146 2017-05-30  Richard Biener  <rguenther@suse.de>
30148         * tree-vectorizer.h (struct _stmt_vec_info): Add reduc_type
30149         and reduc_def fields.
30150         (STMT_VINFO_REDUC_TYPE): New define.
30151         (STMT_VINFO_REDUC_DEF): Likewise.
30152         (vect_force_simple_reduction): Adjust prototype.
30153         * tree-vect-loop.c (vect_analyze_scalar_cycles_1): Adjust.
30154         (vect_is_simple_reduction): Remove check_reduction argument.
30155         (vect_force_simple_reduction): Adjust and set
30156         STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30157         (vectorizable_reduction): Do not re-do reduction analysis
30158         but use STMT_VINFO_REDUC_TYPE and STMT_VINFO_REDUC_DEF.
30159         * tree-parloops.c (gather_scalar_reductions): Adjust.
30161 2017-05-30  Richard Biener  <rguenther@suse.de>
30163         PR middle-end/80901
30164         * cfgexpand.c (expand_gimple_cond): Match up loop fixup with
30165         split_edge code.
30167 2017-05-24  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30169         * tree-vect-data-refs.c (vect_get_peeling_costs_all_drs):
30170         Introduce unknown_misalignment parameter and remove vf.
30171         (vect_peeling_hash_get_lowest_cost):
30172         Pass unknown_misalignment parameter.
30173         (vect_enhance_data_refs_alignment):
30174         Fix unsupportable data ref treatment.
30176 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30178         * tree-vect-data-refs.c (vect_get_data_access_cost):
30179         Workaround for SLP handling.
30180         (vect_enhance_data_refs_alignment):
30181         Compute costs for doing no peeling at all, compare to the best
30182         peeling costs so far and avoid peeling if cheaper.
30184 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30186         * tree-vect-data-refs.c (vect_peeling_hash_choose_best_peeling):
30187         Return peeling info and set costs to zero for unlimited cost
30188         model.
30189         (vect_enhance_data_refs_alignment): Also inspect all datarefs
30190         with unknown misalignment. Compute and costs for unknown
30191         misalignment, compare them to the costs for known misalignment
30192         and choose the cheapest for peeling.
30194 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30196         * tree-vect-data-refs.c (vect_update_misalignment_for_peel): Rename.
30197         (vect_get_peeling_costs_all_drs): Create function.
30198         (vect_peeling_hash_get_lowest_cost):
30199         Use vect_get_peeling_costs_all_drs.
30200         (vect_peeling_supportable): Create function.
30201         (vect_enhance_data_refs_alignment): Use vect_peeling_supportable.
30203 2017-05-30  Robin Dapp  <rdapp@linux.vnet.ibm.com>
30205         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Create
30206         DR_HAS_NEGATIVE_STEP.
30207         (vect_update_misalignment_for_peel): Define DR_MISALIGNMENT.
30208         (vect_enhance_data_refs_alignment): Use.
30209         (vect_duplicate_ssa_name_ptr_info): Use.
30210         * tree-vectorizer.h (dr_misalignment): Use.
30211         (known_alignment_for_access_p): Use.
30213 2017-05-30  Jozef Lawrynowicz  <jozef.l@somniumtech.com>
30215         PR target/78838
30216         * config/msp430/msp430.c (gen_prefix): Return NULL when section name is
30217         .lowtext.
30218         (has_section_name): New function.
30220 2017-05-30  Martin Liska  <mliska@suse.cz>
30222         PR other/80909
30223         * auto-profile.c (get_function_decl_from_block): Fix
30224         parenthesis.
30226 2017-05-30  Richard Biener  <rguenther@suse.de>
30228         PR middle-end/80876
30229         * cfgexpand.c (expand_gimple_cond): Fixup preserving loops again.
30231 2017-05-30  Martin Liska  <mliska@suse.cz>
30233         * dumpfile.c: Use newly added macro DUMP_FILE_INFO.
30234         * dumpfile.h (struct dump_file_info): Remove ctors.
30236 2017-05-30  Martin Liska  <mliska@suse.cz>
30238         * predict.def: Fix GNU coding style.
30240 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30242         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
30243         Mark 'to' argument with ATTRIBUTE_UNUSED.
30245 2017-05-29  Max Filippov  <jcmvbkbc@gmail.com>
30247         * config/xtensa/xtensa.c (xtensa_emit_call): Use
30248         HOST_WIDE_INT_PRINT_HEX instead of 0x%lx format string.
30249         (print_operand): Use HOST_WIDE_INT_PRINT_DEC instead of %ld
30250         format string.
30252 2017-05-29  Eric Botcazou  <ebotcazou@adacore.com>
30254         * doc/install.texi (Options specification): Restore entry of
30255         --enable-sjlj-exceptions.
30257 2017-05-27  Michael Eager  <eager@eagercon.com>
30259         Revert:
30260         2016-01-21  Ajit Agarwal  <ajitkum@xilinx.com>
30262         See https://gcc.gnu.org/ml/gcc/2017-05/msg00221.html.
30264         * config/microblaze/microblaze.h
30265         (FIXED_REGISTERS): Update in macro.
30266         (CALL_USED_REGISTERS): Update in macro.
30268 2017-05-27  François-Xavier Coudett  <fxcoudert@gcc.gnu.org>
30270         * doc/install.texi: Add links to macOS binary distributions.
30272 2017-05-27  Jakub Jelinek  <jakub@redhat.com>
30274         PR bootstrap/80887
30275         Revert:
30276         2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30278         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30280 2017-05-26  Martin Liska  <mliska@suse.cz>
30282         * dumpfile.h (enum dump_kind): Renumber TDF_* flags to be contiguous.
30284 2017-05-26  Martin Liska  <mliska@suse.cz>
30286         * cfg.c (check_bb_profile): Do not use TDF_COMMENT and print
30287         always leading ';; '.
30288         (dump_bb_info): Likewise.
30289         (brief_dump_cfg): Likewise.
30290         * cfgrtl.c (print_rtl_with_bb): Do not use TDF_COMMENT.
30291         * dumpfile.c: Remove usage of TDF_VERBOSE.
30292         * dumpfile.h (enum dump_kind): Likewise.
30293         (dump_gimple_bb_header): Do not use TDF_COMMENT.
30294         * print-tree.c (debug_verbose): Remove.
30295         * tree-cfg.c (gimple_dump_cfg): Do not use TDF_COMMENT.
30296         (dump_function_to_file): Remove dumps guarded with TDF_VERBOSE.
30297         * tree-diagnostic.c (default_tree_printer): Replace
30298         TDF_DIAGNOSTIC with TDF_SLIM.
30300 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30302         * tree-vect-loop-manip.c (create_intersect_range_checks_index): Pass
30303         in parameter loop, rather than loop_vinfo.
30304         (create_intersect_range_checks): Ditto.
30305         (vect_create_cond_for_alias_checks): Update call to above functions.
30307 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30309         PR tree-optimization/80815
30310         * tree-data-ref.c (prune_runtime_alias_test_list): Simplify condition
30311         for merging runtime alias checks.  Handle negative DR_STEPs.
30313 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30315         * tree-vect-data-refs.c (Operator==, comp_dr_with_seg_len_pair):
30316         Move from ...
30317         * tree-data-ref.c (Operator==, comp_dr_with_seg_len_pair): To here.
30318         * tree-vect-data-refs.c (vect_prune_runtime_alias_test_list): Factor
30319         out code pruning runtime alias checks.
30320         * tree-data-ref.c (prune_runtime_alias_test_list): New function
30321         factored out from above.
30322         * tree-vectorizer.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30323         Move from ...
30324         * tree-data-ref.h (struct dr_with_seg_len, dr_with_seg_len_pair_t):
30325         ... to here.
30326         (prune_runtime_alias_test_list): New decalaration.
30328 2017-05-26  Bin Cheng  <bin.cheng@arm.com>
30330         * tree-vect-data-refs.c (compare_tree): Rename and move ...
30331         * tree-data-ref.c (data_ref_compare_tree): ... to here.
30332         * tree-data-ref.h (data_ref_compare_tree): New decalaration.
30333         * tree-vect-data-refs.c (dr_group_sort_cmp): Update uses.
30334         (operator==, comp_dr_with_seg_len_pair): Ditto.
30335         (vect_prune_runtime_alias_test_list): Ditto.
30337 2017-05-26  Martin Liska  <mliska@suse.cz>
30339         PR ipa/80663
30340         * params.def: Bound partial-inlining-entry-probability param.
30342 2017-05-26  Marek Polacek  <polacek@redhat.com>
30344         PR sanitizer/80875
30345         * fold-const.c (fold_binary_loc) <case MULT_EXPR>: Check if OP1
30346         can be negated.
30348 2017-05-26  Richard Biener  <rguenther@suse.de>
30350         PR tree-optimization/80842
30351         * tree-ssa-ccp.c (set_lattice_value): Always meet with the old
30352         value.
30354 2017-05-26  Richard Biener  <rguenther@suse.de>
30356         PR tree-optimization/80844
30357         * tree-vectorizer.c (adjust_simduid_builtins): Propagate results.
30359 2017-05-25  Sebastian Peryt  <sebastian.peryt@intel.com>
30361         * doc/md.texi (Machine Constraints): Update x86 family
30362         machine constraints section to match 'config/i386/constraints.md'.
30364 2017-05-25  Volker Reichelt  <v.reichelt@netcologne.de>
30366         * doc/invoke.texi (-Wcatch-value=): Document new warning option.
30368 2017-05-25  Nathan Sidwell  <nathan@acm.org>
30370         * doc/invoke.texi (--enable-languages): Update documentation.
30372 2017-05-25  Martin Liska  <mliska@suse.cz>
30374         * dumpfile.c: Add TDF_FOLDING.
30375         * dumpfile.h (enum dump_kind): Likewise.
30376         * genmatch.c (dt_simplify::gen_1): Use it.
30378 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30380         * match.pd (view_convert (convert@0 @1)): Handle zero-extension.
30382 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30384         * match.pd ((A +- CST1) +- CST2): Allow some conversions.
30385         * tree.c (drop_tree_overflow): Handle COMPLEX_CST and VECTOR_CST.
30387 2017-05-25  Marc Glisse  <marc.glisse@inria.fr>
30389         * fold-const.c (fold_binary_loc) [(A & C) == D]: Remove transformation.
30390         * match.pd (X == C): Rewrite it here.
30391         (with_possible_nonzero_bits, with_possible_nonzero_bits2,
30392         with_certain_nonzero_bits2): New predicates.
30393         * tree-ssanames.c (get_nonzero_bits): Handle INTEGER_CST.
30395 2017-05-24  Nathan Sidwell  <nathan@acm.org>
30397         * lto-streamer-in.c (lto_input_data_block): Adjust T const cast to
30398         avoid warning.
30400         * auto-profile.c (afdo_propagate): Adjust T const cast to avoid
30401         warning.
30403 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30405         * config/powerpcspe: New port.  Files are copied from the rs6000
30406         port, with "rs6000" in filenames replaced by "powerpcspe".
30408 2017-05-24  Wilco Dijkstra  <wdijkstr@arm.com>
30410         PR rtl-optimization/80754
30411         * lra-remat.c (do_remat): Add overlap checks for dst_regno.
30413 2017-05-24  Sheldon Lobo  <smlobo@sheldon.us.oracle.com>
30415         * config/sparc/sparc.md (length): Return the correct value for -mflat
30416         sibcalls to match output_sibcall.
30418 2017-05-24  Segher Boessenkool  <segher@kernel.crashing.org>
30420         PR bootstrap/80860
30421         PR bootstrap/80843
30422         * config/rs6000/rs6000.c (struct machine_function): Add new field
30423         n_components.
30424         (rs6000_get_separate_components): Init that field, use it.
30425         (rs6000_components_for_bb): Use the field.
30427 2017-05-24  Thomas Preud'homme  <thomas.preudhomme@arm.com>
30429         * config/arm/arm.c (arm_expand_prologue): Fix typo in comment.
30431 2017-05-24  Peter Bergner  <bergner@vnet.ibm.com>
30433         PR middle-end/80823
30434         * tree-cfg.c (group_case_labels_stmt): Delete increment of "i";
30436 2017-05-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
30438         PR target/80725
30439         * config/s390/s390.c (s390_check_qrst_address): Check incoming
30440         address against address_operand predicate.
30441         * config/s390/s390.md ("*indirect_jump"): Swap alternatives.
30443 2017-05-24  Eric Botcazou  <ebotcazou@adacore.com>
30445         * var-tracking.c (track_expr_p): Do not return 0 for tracked record
30446         parameters passed indirectly.
30448 2017-05-23  Uros Bizjak  <ubizjak@gmail.com>
30450         * config/i386/i386.md (*movdi_internal): Remove SSE4
30451         alternative 18 (?r, *v).  Update insn attributes.
30452         (*movsi_internal): Remove SSE4 alternative 13 (?r, *v).
30453         Update insn attributes.
30454         (*zero_extendsidi2): Remove SSE4 alternative (?r, *x).
30455         Update insn attributes.
30456         * config/i386/sse.md (vec_extract<ssevecmodelower>_0): Remove SSE4
30457         alternative 1 (r, v). Remove isa attribute.
30458         * config/i386/i386.c (dimode_scalar_chain::make_vector_copies):
30459         Always move value through stack for !TARGET_INTER_UNIT_MOVES_TO_VEC
30460         and !TARGET_INTER_UNIT_MOVES_TO_VEC targets.
30462 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30464         * doc/sourcebuild.texi (Directives, Verify compiler message): Document
30465         dg-line directive.
30467 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30469         * cgraphunit.c (symbol_table::process_new_functions): Update.
30470         * ipa-fnsummary.c (pass_data_inline_parameters): Remove.
30471         (inline_generate_summary): Rename to ...
30472         (ipa_fn_summary_generate): ... this one.
30473         (inline_read_summary): Rename to ...
30474         (ipa_fn_summary_read): ... this one.
30475         (inline_write_summary): Rename to ...
30476         (ipa_fn_summary_write): ... this one.
30477         (inline_free_summary): Rename to ...
30478         (ipa_free_fn_summary): ... this one.
30479         (pass_data_local_fn_summary, pass_local_fn_summary,
30480         make_pass_local_fn_summary, pass_data_ipa_free_fn_summary,
30481         pass_ipa_free_fn_summary, make_pass_ipa_free_fn_summary,
30482         pass_data_ipa_fn_summary, pass_ipa_fn_summary,
30483         make_pass_ipa_fn_summary): New.
30484         * ipa-fnsummary.h (inline_generate_summary, inline_read_summary,
30485         inline_write_summary, inline_free_summary): Remove.
30486         (ipa_free_fn_summary) : New.
30487         * ipa-inline.c (ipa_inline): Update.
30488         (pass_ipa_inline): Do not generate summaries.
30489         * ipa.c (pass_data_ipa_free_fn_summary, pass_ipa_free_fn_summary):
30490         Remove.
30491         * passes.def: Replace pass_inline_parameters by pass_local_fn_summary
30492         and add pass_ipa_fn_summary.
30493         * tree-pass.h (make_pass_ipa_fn_summary, make_pass_local_fn_summary):
30494         New.
30495         (make_pass_inline_parameters): Remove.
30497 2017-05-23  Thomas Schwinge  <thomas@codesourcery.com>
30499         * omp-low.c (struct omp_context): Remove "default_kind" member.
30500         Adjust all users.
30502         * omp-offload.c (execute_oacc_device_lower): Remove the
30503         parallelism dimensions function attributes for unparallelized
30504         OpenACC kernels constructs.
30506 2017-05-23  Martin Liska  <mliska@suse.cz>
30508         * cgraph.c (cgraph_node::get_create): Use symtab_node::dump_{asm_,}name
30509         functions.
30510         (cgraph_edge::make_speculative): Likewise.
30511         (cgraph_edge::resolve_speculation): Likewise.
30512         (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
30513         (cgraph_node::dump): Likewise.
30514         * cgraph.h: Likewise.
30515         * cgraphunit.c (analyze_functions): Likewise.
30516         (symbol_table::compile): Likewise.
30517         * ipa-cp.c (print_all_lattices): Likewise.
30518         (determine_versionability): Likewise.
30519         (initialize_node_lattices): Likewise.
30520         (ipcp_verify_propagated_values): Likewise.
30521         (estimate_local_effects): Likewise.
30522         (update_profiling_info): Likewise.
30523         (create_specialized_node): Likewise.
30524         (perhaps_add_new_callers): Likewise.
30525         (decide_about_value): Likewise.
30526         (decide_whether_version_node): Likewise.
30527         (identify_dead_nodes): Likewise.
30528         (ipcp_store_bits_results): Likewise.
30529         * ipa-devirt.c (dump_targets): Likewise.
30530         (ipa_devirt): Likewise.
30531         * ipa-icf.c (sem_item::dump): Likewise.
30532         (sem_function::equals): Likewise.
30533         (sem_variable::equals): Likewise.
30534         (sem_item_optimizer::read_section): Likewise.
30535         (sem_item_optimizer::execute): Likewise.
30536         (congruence_class::dump): Likewise.
30537         * ipa-inline-analysis.c (dump_ipa_call_summary): Likewise.
30538         (dump_inline_summary): Likewise.
30539         (estimate_node_size_and_time): Likewise.
30540         (inline_analyze_function): Likewise.
30541         * ipa-inline-transform.c (inline_call): Likewise.
30542         * ipa-inline.c (report_inline_failed_reason): Likewise.
30543         (want_early_inline_function_p): Likewise.
30544         (edge_badness): Likewise.
30545         (update_edge_key): Likewise.
30546         (inline_small_functions): Likewise.
30547         * ipa-profile.c (ipa_profile): Likewise.
30548         * ipa-prop.c (ipa_print_node_jump_functions): Likewise.
30549         (ipa_make_edge_direct_to_target): Likewise.
30550         (remove_described_reference): Likewise.
30551         (ipa_impossible_devirt_target): Likewise.
30552         (propagate_controlled_uses): Likewise.
30553         (ipa_print_node_params): Likewise.
30554         (ipcp_transform_function): Likewise.
30555         * ipa-pure-const.c (pure_const_read_summary): Likewise.
30556         (propagate_pure_const): Likewise.
30557         * ipa-reference.c (generate_summary): Likewise.
30558         (read_write_all_from_decl): Likewise.
30559         (propagate): Likewise.
30560         (ipa_reference_read_optimization_summary): Likewise.
30561         * ipa-utils.c (ipa_merge_profiles): Likewise.
30562         * ipa.c (walk_polymorphic_call_targets): Likewise.
30563         (symbol_table::remove_unreachable_nodes): Likewise.
30564         (ipa_single_use): Likewise.
30565         * passes.c (execute_todo): Likewise.
30566         * predict.c (drop_profile): Likewise.
30567         * symtab.c (symtab_node::get_dump_name): New function.
30568         (symtab_node::dump_name): Likewise.
30569         (symtab_node::dump_asm_name): Likewise.
30570         (symtab_node::dump_references): Likewise.
30571         (symtab_node::dump_referring): Likewise.
30572         (symtab_node::dump_base): Likewise.
30573         (symtab_node::debug_symtab): Likewise.
30574         * tree-sra.c (convert_callers_for_node): Likewise.
30575         * tree-ssa-structalias.c (ipa_pta_execute): Likewise.
30576         * value-prof.c (init_node_map): Likewise.
30578 2017-05-23  Martin Liska  <mliska@suse.cz>
30580         * cgraph.h: Move symtab_node::dump_table to symbol_table::dump
30581         and symtab_node::debug_symtab to symbol_table::debug.
30582         * cgraphunit.c (analyze_functions): Use the renamed function.
30583         (symbol_table::compile): Likewise.
30584         * ipa-cp.c (ipcp_verify_propagated_values): Likewise.
30585         * ipa-icf.c (sem_item_optimizer::execute): Likewise.
30586         * passes.c (execute_todo): Likewise.
30587         * symtab.c (symbol_table::dump): New function.
30588         * tree-ssa-structalias.c (ipa_pta_execute): Use the renamed function.
30590 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30592         * ipa-fnsummary.c (estimate_node_size_and_time): Do not sanity check
30593         that nonconst implies exec.
30595 2017-05-23  Jan Hubicka  <hubicka@ucw.cz>
30597         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30598         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30599         (inline_edge_summary_vec): Turn into ...
30600         (ipa_call_summaries): ... this one.
30601         (redirect_to_unreachable, edge_set_predicate,
30602         evaluate_properties_for_edge, inline_summary_alloc,
30603         reset_ipa_call_summary, reset_inline_summary,
30604         inline_summary_t::duplicate): Update.
30605         (inline_edge_duplication_hook): Turn to ...
30606         (ipa_call_summary_t::duplicate): ... this one.
30607         (inline_edge_removal_hook): Turn to ...
30608         (ipa_call_summary_t::remove): ... this one.
30609         (dump_inline_edge_summary): Turn to ...
30610         (dump_ipa_call_summary): ... this one.
30611         (estimate_function_body_sizes): Update.
30612         (inline_update_callee_summaries): Update.
30613         (remap_edge_change_prob): Update.
30614         (remap_edge_summaries): Update.
30615         (inline_merge_summary): Update.
30616         (do_estimate_edge_time): Update.
30617         (inline_generate_summary): Update.
30618         (inline_read_section): Update.
30619         (inline_read_summary): Update.
30620         (inline_free_summary): Update.
30621         * ipa-inline.c (can_inline_edge_p): Update.
30622         (compute_inlined_call_time): Update.
30623         (want_inline_small_function_p): Update.
30624         (edge_badness): Update.
30625         (early_inliner): Update.
30626         * ipa-inline.h (inline_edge_summary): Turn to ...
30627         (ipa_call_summary): ... this one.
30628         (ipa_call_summary_t): New class.
30629         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30630         (ipa_call_summaries): New.
30631         (inline_edge_summary): Remove.
30632         (estimate_edge_growth): Update.
30633         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30634         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30635         * ipa-split.c (execute_split_functions): Update.
30636         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30638 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30640         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30641         attributes): Document rdrand effective target.
30643 2017-05-23  Tom de Vries  <tom@codesourcery.com>
30645         * doc/sourcebuild.texi (Effective-Target Keywords, Other hardware
30646         attributes): Sort alphabetically.
30648 2017-05-23  Georg-Johann Lay  <avr@gjlay.de>
30650         * config/avr/genmultilib.awk: Use gsub instead of gensub.
30652 2017-05-22  Michael Meissner  <meissner@linux.vnet.ibm.com>
30654         PR target/80718
30655         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Split
30656         V2DF/V2DI splat into two separate patterns, one that handles
30657         registers, and the other that only handles memory.  Drop support
30658         for splatting from a GPR on ISA 2.07 and then splitting the
30659         splat into direct move and splat.
30660         (vsx_splat_<mode>_reg): Likewise.
30661         (vsx_splat_<mode>_mem): Likewise.
30663 2017-05-22  Segher Boessenkool  <segher@kernel.crashing.org>
30665         * cfgcleanup.c (bb_is_just_return): Allow CLOBBERs.
30667 2017-05-22  Jakub Jelinek  <jakub@redhat.com>
30669         PR middle-end/80809
30670         * omp-low.c (finish_taskreg_remap): New function.
30671         (finish_taskreg_scan): If unit size of ctx->record_type
30672         is non-constant, unshare the size expression and replace
30673         decls in it with possible outer var refs.
30675         PR middle-end/80809
30676         * gimplify.c (omp_add_variable): For GOVD_DEBUG_PRIVATE use
30677         GOVD_SHARED rather than GOVD_PRIVATE with it.
30678         (gimplify_adjust_omp_clauses_1, gimplify_adjust_omp_clauses): Expect
30679         GOVD_SHARED rather than GOVD_PRIVATE with GOVD_DEBUG_PRIVATE.
30681         PR middle-end/80853
30682         * omp-low.c (lower_reduction_clauses): Pass OMP_CLAUSE_PRIVATE
30683         as last argument to build_outer_var_ref for pointer bases of array
30684         section reductions.
30686 2017-05-19  Martin Sebor  <msebor@redhat.com>
30688         * print-tree.c (print_node): Print DECL_READ_P flag.
30690 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30692         * Makefile.in: Add ipa-fnsummary.o and ipa-fnsummary.h
30693         * auto-profile.c: Replace ipa-inline.h by ipa-fnsummary.h
30694         * cgraph.c: Likewise.
30695         * cgraphunit.c: Likewise.
30696         * gengtype.c: Likewise.
30697         * ipa-cp.c: Likewise.
30698         * ipa-devirt.c: Likewise.
30699         * ipa-icf.c: Likewise.
30700         * ipa-predicate.c: Likewise.
30701         * ipa-profile.c: Likewise.
30702         * ipa-prop.c: Likewise.
30703         * ipa-split.c: Likewise.
30704         * ipa.c: Likewise.
30705         * ipa-inline-analysis.c (inline_summaries, ipa_call_summaries,
30706         edge_predicate_pool, dump_inline_hints,
30707         inline_summary::account_size_time, redirect_to_unreachable,
30708         edge_set_predicate, set_hint_predicate,
30709         evaluate_conditions_for_known_args, evaluate_properties_for_edge,
30710         inline_summary_alloc, ipa_call_summary::reset, inline_summary::reset,
30711         inline_summary_t::remove, remap_hint_predicate_after_duplication,
30712         inline_summary_t::duplicate, ipa_call_summary_t::duplicate,
30713         ipa_call_summary_t::remove, initialize_growth_caches,
30714         free_growth_caches, dump_ipa_call_summary, dump_inline_summary,
30715         debug_inline_summary, dump_inline_summaries, initialize_inline_failed,
30716         mark_modified, unmodified_parm_1, unmodified_parm,
30717         unmodified_parm_or_parm_agg_item, eliminated_by_inlining_prob,
30718         set_cond_stmt_execution_predicate, set_switch_stmt_execution_predicate,
30719         compute_bb_predicates, will_be_nonconstant_expr_predicate,
30720         will_be_nonconstant_predicate, record_modified_bb_info,
30721         get_minimal_bb, record_modified, param_change_prob,
30722         phi_result_unknown_predicate, predicate_for_phi_result,
30723         array_index_predicate, clobber_only_eh_bb_p, fp_expression_p,
30724         estimate_function_body_sizes, compute_inline_parameters,
30725         compute_inline_parameters_for_curren, pass_data_inline_parameters,
30726         estimate_node_size_and_time, estimate_ipcp_clone_size_and_time,
30727         inline_update_callee_summaries, remap_edge_change_prob,
30728         remap_edge_summaries, remap_hint_predicate, inline_merge_summary,
30729         inline_update_overall_summary, inline_indirect_intraprocedural_analysis,
30730         inline_analyze_function, inline_summary_t::insert,
30731         inline_generate_summary, read_ipa_call_summary, inline_read_section,
30732         inline_read_summary, write_ipa_call_summary, inline_write_summary,
30733         inline_free_summary): Move to ipa-fnsummary.h
30734         (predicate_t): Remove.
30735         * ipa-fnsummary.c: New file.
30736         * ipa-inline.h:  Do not include sreal.h and ipa-predicate.h
30737         (enum inline_hints_vals, inline_hints, agg_position_info,
30738         INLINE_SIZE_SCALE, size_time_entry, inline_summary, inline_summary_t,
30739         inline_summaries, ipa_call_summary, ipa_call_summary_t,
30740         ipa_call_summaries, debug_inline_summary, dump_inline_summaries,
30741         dump_inline_summary, dump_inline_hints, inline_generate_summary,
30742         inline_read_summary, inline_write_summary, inline_free_summary,
30743         inline_analyze_function, initialize_inline_failed,
30744         inline_merge_summary, inline_update_overall_summary,
30745         compute_inline_parameters): Move to ipa-fnsummary.h
30746         * ipa-fnsummary.h: New file.
30747         * ipa-inline-transform.h: Include ipa-inline.h.
30748         * ipa-inline.c: LIkewise.
30750 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30752         * ipa-inline.c (edge_badness): Use inlined_time instead of
30753         inline_summaries->get.
30755 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30757         * ipa-inline.c (edge_badness): Use estimate_size_after_inlining.
30759 2017-05-22  Nathan Sidwell  <nathan@acm.org>
30761         * doc/invoke.texi (fdump-translation-unit): Delete documentation.
30762         (fdump-lang): Document 'raw' option.
30763         * dumpfile.h (TDI_tu): Delete.
30764         * dumpfile.c (dump_files): Remove translation-unit.
30765         (FIRST_AUTO_NUMBERED_DUMP): Decrement.
30767 2017-05-22  Georg-Johann Lay  <avr@gjlay.de>
30769         * config/avr/t-avr (AWK) [t-multilib]: Remove "-v FORMAT=Makefile"
30770         command option from $(AWK) call.
30771         * config/avr/genmultilib.awk: Simplify and rewrite so that it
30772         generates MULTILIB_REQUIRED instead of MULTILIB_EXCEPTIONS.
30773         [FORMAT]: Remove handling of variable.
30774         * config/avr/t-multilib: Regenerate.
30776 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30778         * ipa-inline-analysis.c (inline_summary::reset): Do not reset
30779         self_time.
30780         (dump_inline_summary): Do not print self_time.
30781         (estimate_function_body_sizes): Do not set self_time.
30782         (compute_inline_parameters): Likewise.
30783         (inline_read_section, inline_write_summary): Do not stream self_time.
30784         * ipa-inline.h (inline_summary): Drop self_time.
30786 2017-05-22  Jan Hubicka  <hubicka@ucw.cz>
30788         * ipa-inline-analysis.c (account_size_time): Rename to ...
30789         (inline_summary::account_size_time): ... this one.
30790         (reset_ipa_call_summary): Turn to ...
30791         (ipa_call_summary::reset): ... this one.
30792         (reset_inline_summary): Turn to ...
30793         (inline_summary::reset): ... this one.
30794         (inline_summary_t::remove): Update.
30795         (inline_summary_t::duplicate): Update.
30796         (ipa_call_summary_t::remove): Update.
30797         (dump_inline_summary): Update.
30798         (estimate_function_body_sizes): Update.
30799         (compute_inline_parameters): Update.
30800         (estimate_node_size_and_time): Update.
30801         (inline_merge_summary): Update.
30802         (inline_update_overall_summary): Update.
30803         (inline_read_section): Update.
30804         (inline_write_summary): Update.
30805         * ipa-inline.h (inline_summary): Rename entry to size_time_table;
30806         add account_size_time and reset member functions.
30807         (ipa_call_summary): Add reset function.
30808         * ipa-predicate.h (predicate::operator &): Constify.
30810 2017-05-22  Richard Biener  <rguenther@suse.de>
30812         * df-scan.c (df_insn_refs_verify): Speedup when not verifying.
30814 2017-05-19  Jason Merrill  <jason@redhat.com>
30816         * tree.c (make_tree_vec_stat, grow_tree_vec_stat): Use size_t.
30818 2017-05-19  Marek Polacek  <polacek@redhat.com>
30820         PR sanitizer/80800
30821         * fold-const.c (extract_muldiv_1) <case TRUNC_DIV_EXPR>: Add
30822         TYPE_OVERFLOW_WRAPS checks.
30824 2017-05-19  Thomas Schwinge  <thomas@codesourcery.com>
30826         * tree-core.h (enum omp_clause_default_kind): Add
30827         "OMP_CLAUSE_DEFAULT_PRESENT".
30828         * tree-pretty-print.c (dump_omp_clause): Handle it.
30829         * gimplify.c (enum gimplify_omp_var_data): Add
30830         "GOVD_MAP_FORCE_PRESENT".
30831         (gimplify_adjust_omp_clauses_1): Map it to
30832         "GOMP_MAP_FORCE_PRESENT".
30833         (oacc_default_clause): Handle "OMP_CLAUSE_DEFAULT_PRESENT".
30835         * gimplify.c (oacc_default_clause): Clarify.
30837 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30839         LANG_HOOK_REGISTER_DUMPS
30840         * toplev.c (general_init): Call register dump lang hook.
30841         * doc/invoke.texi: Document -fdump-lang option family.
30842         * dumpfile.c (dump_files): Remove class dump here.
30843         (FIRST_AUTO_NUMBERED_DUMP): Adjust.
30844         * dumpfile.h (tree_dump_index): Remove TDI_class.
30845         * langhooks-def.h (lhd_register_dumps): Declare.
30846         (LANG_HOOKS_REGISTER_DUMPS): Define.
30847         (LANG_HOOKS_INITIALIZER): Add it.
30848         * langhooks.c (lhd_register_dumps): Define.
30849         * langhooks.h (struct lang_hooks): Add register_dumps.
30851 2017-05-19  Nathan Sidwell  <nathan@acm.org>
30853         * context.h (context::set_passes): New.
30854         * context.c (context::context): Do not create pass manager.
30855         * toplev.c (general_init): Create pass manager here.
30857 2017-05-19  Segher Boessenkool  <segher@kernel.crashing.org>
30859         * config/rs6000/rs6000.md (splitter to load of -1 and mask): Don't
30860         use this splitter if two add or or instructions would also work for
30861         the constant we want to generate.
30863 2017-05-19  Richard Biener  <rguenther@suse.de>
30865         PR build/80821
30866         * genmatch.c (dt_node::gen_kids_1): Add missing scope around
30867         predicate evaluation.
30869 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30871         * ipa-inline.h (ipa_call_summary): Turn sizes into signed;
30872         add ctor.
30873         * ipa-inline.c (want_inline_small_function_p): Do not cast to
30874         unsigned.
30876 2017-05-19  Jan Hubicka  <hubicka@ucw.cz>
30878         * ipa-inline-analysis.c (cgraph_2edge_hook_list, cgraph_edge_hook_list,
30879         inline_edge_removal_hook, inline_edge_duplication_hook): Remove.
30880         (inline_edge_summary_vec): Turn into ...
30881         (ipa_call_summaries): ... this one.
30882         (redirect_to_unreachable, edge_set_predicate,
30883         evaluate_properties_for_edge, inline_summary_alloc,
30884         reset_ipa_call_summary, reset_inline_summary,
30885         inline_summary_t::duplicate): Update.
30886         (inline_edge_duplication_hook): Turn to ...
30887         (ipa_call_summary_t::duplicate): ... this one.
30888         (inline_edge_removal_hook): Turn to ...
30889         (ipa_call_summary_t::remove): ... this one.
30890         (dump_inline_edge_summary): Turn to ...
30891         (dump_ipa_call_summary): ... this one.
30892         (estimate_function_body_sizes): Update.
30893         (inline_update_callee_summaries): Update.
30894         (remap_edge_change_prob): Update.
30895         (remap_edge_summaries): Update.
30896         (inline_merge_summary): Update.
30897         (do_estimate_edge_time): Update.
30898         (inline_generate_summary): Update.
30899         (inline_read_section): Update.
30900         (inline_read_summary): Update.
30901         (inline_free_summary): Update.
30902         * ipa-inline.c (can_inline_edge_p): Update.
30903         (compute_inlined_call_time): Update.
30904         (want_inline_small_function_p): Update.
30905         (edge_badness): Update.
30906         (early_inliner): Update.
30907         * ipa-inline.h (inline_edge_summary): Turn to ...
30908         (ipa_call_summary): ... this one.
30909         (ipa_call_summary_t): New class.
30910         (inline_edge_summary_t, inline_edge_summary_vec): Remove.
30911         (ipa_call_summaries): New.
30912         (inline_edge_summary): Remove.
30913         (estimate_edge_growth): Update.
30914         * ipa-profile.c (ipa_propagate_frequency_1): Update.
30915         * ipa-prop.c (ipa_make_edge_direct_to_target): Update.
30916         * ipa-split.c (execute_split_functions): Update.
30917         * ipa.c (symbol_table::remove_unreachable_nodes): Update.
30919 2017-05-19  Richard Biener  <rguenther@suse.de>
30921         PR middle-end/80764
30922         * cfgexpand.c (expand_gimple_cond): Fix loop fixup.
30924 2017-05-18  Segher Boessenkool  <segher@kernel.crashing.org>
30926         * config/rs6000/rs6000.c (struct machine_function): Add field
30927         fpr_is_wrapped_separately.
30928         (rs6000_get_separate_components): Use 64 components.  Handle the
30929         new FPR components.
30930         (rs6000_components_for_bb): Handle the FPR components.
30931         (rs6000_emit_prologue_components): Handle the FPR components.
30932         (rs6000_emit_epilogue_components): Handle the FPR components.
30933         (rs6000_set_handled_components): Handle the FPR components.
30934         (rs6000_emit_prologue): Don't output prologue code for those FPRs
30935         that are already separately shrink-wrapped.
30936         (rs6000_emit_epilogue): Don't output epilogue code for those FPRs
30937         that are already separately shrink-wrapped.
30939 2017-05-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
30941         PR target/80510
30942         * config/rs6000/predicates.md (simple_offsettable_mem_operand):
30943         New predicate.
30945         * config/rs6000/rs6000.md (ALTIVEC_DFORM): New iterator.
30946         (define_peephole2 for Altivec d-form load): Add peepholes to catch
30947         cases where the register allocator uses a move and an offsettable
30948         memory operation to/from a FPR register on ISA 2.06/2.07.
30949         (define_peephole2 for Altivec d-form store): Likewise.
30951 2017-05-18  Uros Bizjak  <ubizjak@gmail.com>
30953         PR target/80799
30954         * config/i386/mmx.md (*mov<mode>_internal): Enable
30955         alternatives 11, 12, 13 and 14 also for 32bit targets.
30956         Remove alternatives 15, 16, 17 and 18.
30957         * config/i386/sse.md (vec_concatv2di): Change
30958         alternative (!x, *y) to (x, ?!*Yn).
30960 2017-05-18  Paolo Carlini  <paolo.carlini@oracle.com>
30962         * dumpfile.h (enum dump_kind): Remove stray comma.
30964 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
30966         * Makefile.in: Add ipa-predicate.o and ipa-predicate.h
30967         * ipa-inline-analysis.c (NUM_CONDITIONS): turn into
30968         predicate::num_conditions
30969         (IS_NOT_CONSTANT): turn into predicate::is_not_constant.
30970         (CHANGED): turn into predicate::changed.
30971         (agg_position_info): Move to ipa-predicate.h
30972         (add_condition, predicate::add_clause, predicate::operator &=,
30973         predicate::or_with, predicate::evaluate, predicate::probability,
30974         dump_condition, dump_clause, predicate::dump,
30975         predicate::remap_after_duplication, predicate::remap_after_inlining,
30976         predicate::stream_in, predicate::stream_out): Move to ipa-predicate.c
30977         (evaluate_conditions_for_known_args): Update.
30978         (set_cond_stmt_execution_predicate): Update.
30979         * ipa-inline.h: Include ipa-predicate.h
30980         (condition, inline_param_summary, conditions, agg_position_info,
30981         predicate): Move to ipa-predicate.h
30982         * ipa-predicate.c: New file.
30983         * ipa-predicate.h: New file.
30985 2017-05-18  Wilco Dijkstra  <wdijkstr@arm.com>
30987         * final.c (leaf_function_p): Check we are not in a sequence.
30989 2017-05-18  Martin Liska  <mliska@suse.cz>
30991         * cfgrtl.c (rtl_verify_edges): Remove usage of TDF_RTL.
30992         * dumpfile.c (dump_register): Use new enum dump_kind.
30993         (get_dump_file_name): Likewise.
30994         (dump_enable_all): Likewise.
30995         (dump_switch_p_1): Likewise.
30996         (enable_rtl_dump_file): Remove usage of TDF_RTL.
30997         * dumpfile.h (enum dump_kind): New enum type.
30998         (struct dump_file_info): Create constructor and
30999         format fields and comments.
31000         * passes.c (pass_manager::register_one_dump_file):
31001         Use num dump_kind.
31002         * statistics.c (statistics_early_init): Likewise.
31003         * tree-ssa-loop-prefetch.c (dump_mem_details): Replace
31004         TDF_TREE with TDF_SLIM.
31005         (gather_memory_references_ref): Likewise.
31007 2017-05-18  Martin Liska  <mliska@suse.cz>
31009         * vec.h (struct vnull): Use it.
31011 2017-05-18  Jan Hubicka  <hubicka@ucw.cz>
31013         * ipa-inline-analysis.c (predicate_conditions): Move to ipa-inline.h
31014         (true_predicate, false_predicate, true_predicate_p,
31015         false_predicate_p): Remove.
31016         (single_cond_predicate, not_inlined_predicate): Turn to member function
31017         in ipa-inline.h
31018         (add_condition): Update.
31019         (add_clause): Turn to...
31020         (predicate::add_clause): ... this one; update; allow passing NULL
31021         as parameter.
31022         (and_predicates): Turn to ...
31023         (predicate::operator &=): ... this one.
31024         (predicates_equal_p): Move to predicate::operator == in ipa-inline.h
31025         (or_predicates): Turn to ...
31026         (predicate::or_with): ... this one.
31027         (evaluate_predicate): Turn to ...
31028         (predicate::evaluate): ... this one.
31029         (predicate_probability): Turn to ...
31030         (predicate::probability): ... this one.
31031         (dump_condition): Update.
31032         (dump_predicate): Turn to ...
31033         (predicate::dump): ... this one.
31034         (account_size_time): Update.
31035         (edge_set_predicate): Update.
31036         (set_hint_predicate): UPdate.
31037         (evaluate_conditions_for_known_args): Update.
31038         (evaluate_properties_for_edge): Update.
31039         (remap_predicate_after_duplication): Turn to...
31040         (predicate::remap_after_duplication): ... this one.
31041         (remap_hint_predicate_after_duplication): Update.
31042         (inline_summary_t::duplicate): UPdate.
31043         (dump_inline_edge_summary): Update.
31044         (dump_inline_summary): Update.
31045         (set_cond_stmt_execution_predicate): Update.
31046         (set_switch_stmt_execution_predicate): Update.
31047         (compute_bb_predicates): Update.
31048         (will_be_nonconstant_expr_predicate): Update.
31049         (will_be_nonconstant_predicate): Update.
31050         (phi_result_unknown_predicate): Update.
31051         (predicate_for_phi_result): Update.
31052         (array_index_predicate): Update.
31053         (estimate_function_body_sizes): Update.
31054         (estimate_node_size_and_time): Update.
31055         (estimate_ipcp_clone_size_and_time): Update.
31056         (remap_predicate): Rename to ...
31057         (predicate::remap_after_inlining): ... this one.
31058         (remap_hint_predicate): Update.
31059         (inline_merge_summary): Update.
31060         (inline_update_overall_summary): Update.
31061         (estimate_size_after_inlining): Update.
31062         (read_predicate): Rename to ...
31063         (predicate::stream_in): ... this one.
31064         (read_inline_edge_summary): Update.
31065         (write_predicate): Rename to ...
31066         (predicate::stream_out): ... this one.
31067         (write_inline_edge_summary): Update.
31068         * ipa-inline.h (MAX_CLAUSES): Turn to predicate::max_clauses.
31069         (clause_t): Turn to uint32_t
31070         (predicate): Turn to class; implement constructor and operators
31071         ==, !=, &
31072         (size_time_entry): Update.
31073         (inline_summary): Update.
31074         (inline_edge_summary): Update.
31076 2017-05-18  Marc Glisse  <marc.glisse@inria.fr>
31078         * fold-const.c (fold_binary_loc): Move transformation...
31079         * match.pd (C - X CMP X): ... here.
31081 2017-05-18  Sheldon Lobo  <sheldon.lobo@oracle.com>
31083         * config/sparc/sparc.c (sparc_option_override): Set function
31084         alignment for -mcpu=niagara7 to 64 to match the I$ line.
31085         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC M7 branch
31086         latency to 1.
31087         * config/sparc/sparc.h (BRANCH_COST): Set the SPARC T4 branch
31088         latency to 2.
31089         * config/sparc/sol2.h: Fix a ASM_CPU32_DEFAULT_SPEC typo.
31091 2017-05-18  Marek Polacek  <polacek@redhat.com>
31093         PR sanitizer/80797
31094         * ubsan.c (instrument_null): Unwrap ADDR_EXPRs.
31095         (pass_ubsan::execute): Call gimple_assign_single_p instead of
31096         gimple_assign_load_p.
31098 2017-05-17  Segher Boessenkool  <segher@kernel.crashing.org>
31100         PR middle-end/80692
31101         * real.c (do_compare): Give decimal_do_compare preference over
31102         comparing just the signs.
31104 2017-05-17  Uros Bizjak  <ubizjak@gmail.com>
31106         * doc/md.texi (Canonicalization of Instructions): Describe the
31107         canonical form of instructions that inherently set a condition
31108         code register.
31110 2017-05-17  Peter Bergner  <bergner@vnet.ibm.com>
31112         PR middle-end/80775
31113         * tree-cfg.c: Move deletion of unreachable case statements to after
31114         the merging of consecutive case labels.
31116 2017-05-17  Thomas Preud'homme  <thomas.preudhomme@arm.com>
31118         * config/arm/arm.c (cmse_nonsecure_call_clear_caller_saved): Refer
31119         readers to __gnu_cmse_nonsecure_call libcall for saving, clearing and
31120         restoring of callee-saved registers.
31122 2017-05-17  Eric Botcazou  <ebotcazou@adacore.com>
31124         * compare-elim.c (equivalent_reg_at_start): Adjust outdated comment.
31125         * config/visium/visium.c (single_set_and_flags): Likewise.
31126         * config/visium/visium.md (Substitutions): Likewise.
31128 2017-05-17  Martin Liska  <mliska@suse.cz>
31130         * cfg.c: Introduce dump_flags_t type and
31131         use it instead of int type.
31132         * cfg.h: Likewise.
31133         * cfghooks.c: Likewise.
31134         * cfghooks.h (struct cfg_hooks): Likewise.
31135         * cfgrtl.c: Likewise.
31136         * cfgrtl.h: Likewise.
31137         * cgraph.c (cgraph_node::get_body): Likewise.
31138         * coretypes.h: Likewise.
31139         * domwalk.c: Likewise.
31140         * domwalk.h: Likewise.
31141         * dumpfile.c (struct dump_option_value_info): Likewise.
31142         (dump_enable_all): Likewise.
31143         (dump_switch_p_1): Likewise.
31144         (opt_info_switch_p): Likewise.
31145         * dumpfile.h (enum tree_dump_index): Likewise.
31146         (struct dump_file_info): Likewise.
31147         * genemit.c: Likewise.
31148         * generic-match-head.c: Likewise.
31149         * gengtype.c (open_base_files): Likewise.
31150         * gimple-pretty-print.c: Likewise.
31151         * gimple-pretty-print.h: Likewise.
31152         * graph.c (print_graph_cfg): Likewise.
31153         * graphite-scop-detection.c (dot_all_sese): Likewise.
31154         * ipa-devirt.c (build_type_inheritance_graph): Likewise.
31155         * loop-unroll.c (report_unroll): Likewise.
31156         * passes.c (pass_manager::register_one_dump_file): Likewise.
31157         * print-tree.c: Likewise.
31158         * statistics.c: Likewise.
31159         * tree-cfg.c: Likewise.
31160         * tree-cfg.h: Likewise.
31161         * tree-dfa.c: Likewise.
31162         * tree-dfa.h: Likewise.
31163         * tree-dump.c (dump_function): Likewise.
31164         * tree-dump.h (struct dump_info): Likewise.
31165         * tree-pretty-print.c: Likewise.
31166         * tree-pretty-print.h: Likewise.
31167         * tree-ssa-live.c: Likewise.
31168         * tree-ssa-live.h: Likewise.
31169         * tree-ssa-loop-ivcanon.c (try_unroll_loop_completely): Likewise.
31170         * tree-vect-loop.c: Likewise.
31171         * tree-vect-slp.c: Likewise.
31173 2017-05-16  James Greenhalgh  <james.greenhalgh@arm.com>
31174             Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
31176         PR tree-optimization/80457
31177         * tree-vect-stmts.c (vect_model_simple_cost): Model the cost
31178         of all arguments to a statement as scalar_to_vec operations.
31179         (vectorizable_call): Adjust call to vect_model_simple_cost for
31180         new parameter.
31181         (vectorizable_conversion): Likewise.
31182         (vectorizable_assignment): Likewise.
31183         (vectorizable_shift): Likewise.
31184         (vectorizable_operation): Likewise.
31185         (vectorizable_comparison): Likewise.
31186         (vect_is_simple_cond): Record the def types for operands.
31187         (vectorizable_condition): Likewise, call vect_model_simple_cost.
31188         * tree-vectorizer.h (vect_model_simple_cost): Add new parameter
31189         for statement argument count.
31191 2017-05-16  Carl Love  <cel@us.ibm.com>
31193         * config/rs6000/rs6000-c: Add support for built-in functions
31194         vector unsigned long long vec_bperm (vector unsigned long long,
31195                                              vector unsigned char)
31196         vector signed long long vec_mule (vector signed int,
31197                                           vector signed int)
31198         vector unsigned long long vec_mule (vector unsigned int,
31199                                             vector unsigned int)
31200         vector signed long long vec_mulo (vector signed int,
31201                                           vector signed int)
31202         vector unsigned long long vec_mulo (vector unsigned int,
31203                                             vector unsigned int)
31204         vector signed char vec_sldw (vector signed char,
31205                                      vector signed char,
31206                                      const int)
31207         vector unsigned char vec_sldw (vector unsigned char,
31208                                        vector unsigned char,
31209                                        const int)
31210         vector signed short vec_sldw (vector signed short,
31211                                       vector signed short,
31212                                       const int)
31213         vector unsigned short vec_sldw (vector unsigned short,
31214                                         vector unsigned short,
31215                                         const int)
31216         vector signed int vec_sldw (vector signed int,
31217                                     vector signed int,
31218                                     const int)
31219         vector unsigned int vec_sldw (vector unsigned int,
31220                                       vector unsigned int,
31221                                       const int)
31222         vector signed long long vec_sldw (vector signed long long,
31223                                           vector signed long long,
31224                                           const int)
31225         vector unsigned long long vec_sldw (vector unsigned long long,
31226                                             vector unsigned long long,
31227                                             const int)
31228         * config/rs6000/rs6000-c: Add support for built-in functions
31229         * config/rs6000/rs6000-builtin.def: Add definition for SLDW.
31230         * config/rs6000/altivec.h: Add defintion for vec_sldw.
31231         * doc/extend.texi: Update the built-in documentation for the
31232         new built-in functions.
31234 2017-05-16  Marek Polacek  <polacek@redhat.com>
31236         PR sanitizer/80536
31237         PR sanitizer/80386
31238         * tree.c (save_expr): Don't fold the expression.
31240 2017-05-16  Uros Bizjak  <ubizjak@gmail.com>
31242         * config/i386/i386.md (*movsi_internal): Split (?rm,*y) alternative
31243         to (?r,*Yn) and (?m,*y) alternatives, and (?*y,rm) to (?*Ym,r)
31244         and (?*y,m).  Update insn attributes.
31246 2017-05-16  Martin Liska  <mliska@suse.cz>
31248         * cgraph.c (cgraph_edge::resolve_speculation): Add default value for
31249         flags argument of print_gimple_stmt, print_gimple_expr,
31250         print_generic_stmt and print_generic_expr.
31251         * cgraphclones.c (symbol_table::materialize_all_clones): Likewise.
31252         * coretypes.h: Likewise.
31253         * except.c (dump_eh_tree): Likewise.
31254         * gimple-fold.c (gimple_fold_stmt_to_constant_1): Likewise.
31255         * gimple-pretty-print.h: Likewise.
31256         * gimple-ssa-backprop.c (dump_usage_prefix): Likewise.
31257         (backprop::push_to_worklist): Likewise.
31258         (backprop::pop_from_worklist): Likewise.
31259         (backprop::process_use): Likewise.
31260         (backprop::intersect_uses): Likewise.
31261         (note_replacement): Likewise.
31262         * gimple-ssa-store-merging.c
31263         (pass_store_merging::terminate_all_aliasing_chains): Likewise.
31264         (imm_store_chain_info::coalesce_immediate_stores): Likewise.
31265         (pass_store_merging::execute): Likewise.
31266         * gimple-ssa-strength-reduction.c (dump_candidate): Likewise.
31267         (ssa_base_cand_dump_callback): Likewise.
31268         (dump_incr_vec): Likewise.
31269         (replace_refs): Likewise.
31270         (replace_mult_candidate): Likewise.
31271         (create_add_on_incoming_edge): Likewise.
31272         (create_phi_basis): Likewise.
31273         (insert_initializers): Likewise.
31274         (all_phi_incrs_profitable): Likewise.
31275         (introduce_cast_before_cand): Likewise.
31276         (replace_one_candidate): Likewise.
31277         * gimplify.c (gimplify_expr): Likewise.
31278         * graphite-isl-ast-to-gimple.c (is_valid_rename): Likewise.
31279         (set_rename): Likewise.
31280         (rename_uses): Likewise.
31281         (copy_loop_phi_nodes): Likewise.
31282         (add_close_phis_to_merge_points): Likewise.
31283         (copy_loop_close_phi_args): Likewise.
31284         (copy_cond_phi_args): Likewise.
31285         (graphite_copy_stmts_from_block): Likewise.
31286         (translate_pending_phi_nodes): Likewise.
31287         * graphite-poly.c (print_pdr): Likewise.
31288         (dump_gbb_cases): Likewise.
31289         (dump_gbb_conditions): Likewise.
31290         (print_scop_params): Likewise.
31291         * graphite-scop-detection.c (build_cross_bb_scalars_def): Likewise.
31292         (build_cross_bb_scalars_use): Likewise.
31293         (gather_bbs::before_dom_children): Likewise.
31294         * hsa-dump.c (dump_hsa_immed): Likewise.
31295         * ipa-cp.c (print_ipcp_constant_value): Likewise.
31296         (get_replacement_map): Likewise.
31297         * ipa-inline-analysis.c (dump_condition): Likewise.
31298         (estimate_function_body_sizes): Likewise.
31299         * ipa-polymorphic-call.c (check_stmt_for_type_change): Likewise.
31300         (ipa_polymorphic_call_context::get_dynamic_type): Likewise.
31301         * ipa-prop.c (ipa_dump_param): Likewise.
31302         (ipa_print_node_jump_functions_for_edge): Likewise.
31303         (ipa_modify_call_arguments): Likewise.
31304         (ipa_modify_expr): Likewise.
31305         (ipa_dump_param_adjustments): Likewise.
31306         (ipa_dump_agg_replacement_values): Likewise.
31307         (ipcp_modif_dom_walker::before_dom_children): Likewise.
31308         * ipa-pure-const.c (check_stmt): Likewise.
31309         (pass_nothrow::execute): Likewise.
31310         * ipa-split.c (execute_split_functions): Likewise.
31311         * omp-offload.c (dump_oacc_loop_part): Likewise.
31312         (dump_oacc_loop): Likewise.
31313         * trans-mem.c (tm_log_emit): Likewise.
31314         (tm_memopt_accumulate_memops): Likewise.
31315         (dump_tm_memopt_set): Likewise.
31316         (dump_tm_memopt_transform): Likewise.
31317         * tree-cfg.c (gimple_verify_flow_info): Likewise.
31318         (print_loop): Likewise.
31319         * tree-chkp-opt.c (chkp_print_addr): Likewise.
31320         (chkp_gather_checks_info): Likewise.
31321         (chkp_get_check_result): Likewise.
31322         (chkp_remove_check_if_pass): Likewise.
31323         (chkp_use_outer_bounds_if_possible): Likewise.
31324         (chkp_reduce_bounds_lifetime): Likewise.
31325         * tree-chkp.c (chkp_register_addr_bounds): Likewise.
31326         (chkp_mark_completed_bounds): Likewise.
31327         (chkp_register_incomplete_bounds): Likewise.
31328         (chkp_mark_invalid_bounds): Likewise.
31329         (chkp_maybe_copy_and_register_bounds): Likewise.
31330         (chkp_build_returned_bound): Likewise.
31331         (chkp_get_bound_for_parm): Likewise.
31332         (chkp_build_bndldx): Likewise.
31333         (chkp_get_bounds_by_definition): Likewise.
31334         (chkp_generate_extern_var_bounds): Likewise.
31335         (chkp_get_bounds_for_decl_addr): Likewise.
31336         * tree-chrec.c (chrec_apply): Likewise.
31337         * tree-data-ref.c (dump_data_reference): Likewise.
31338         (dump_subscript): Likewise.
31339         (dump_data_dependence_relation): Likewise.
31340         (analyze_overlapping_iterations): Likewise.
31341         * tree-inline.c (expand_call_inline): Likewise.
31342         (tree_function_versioning): Likewise.
31343         * tree-into-ssa.c (dump_defs_stack): Likewise.
31344         (dump_currdefs): Likewise.
31345         (dump_names_replaced_by): Likewise.
31346         (dump_update_ssa): Likewise.
31347         (update_ssa): Likewise.
31348         * tree-object-size.c (pass_object_sizes::execute): Likewise.
31349         * tree-parloops.c (build_new_reduction): Likewise.
31350         (try_create_reduction_list): Likewise.
31351         (ref_conflicts_with_region): Likewise.
31352         (oacc_entry_exit_ok_1): Likewise.
31353         (oacc_entry_exit_single_gang): Likewise.
31354         * tree-pretty-print.h: Likewise.
31355         * tree-scalar-evolution.c (set_scalar_evolution): Likewise.
31356         (get_scalar_evolution): Likewise.
31357         (add_to_evolution): Likewise.
31358         (get_loop_exit_condition): Likewise.
31359         (analyze_evolution_in_loop): Likewise.
31360         (analyze_initial_condition): Likewise.
31361         (analyze_scalar_evolution): Likewise.
31362         (instantiate_scev): Likewise.
31363         (number_of_latch_executions): Likewise.
31364         (gather_chrec_stats): Likewise.
31365         (final_value_replacement_loop): Likewise.
31366         (scev_const_prop): Likewise.
31367         * tree-sra.c (dump_access): Likewise.
31368         (disqualify_candidate): Likewise.
31369         (create_access): Likewise.
31370         (reject): Likewise.
31371         (maybe_add_sra_candidate): Likewise.
31372         (create_access_replacement): Likewise.
31373         (analyze_access_subtree): Likewise.
31374         (analyze_all_variable_accesses): Likewise.
31375         (sra_modify_assign): Likewise.
31376         (initialize_constant_pool_replacements): Likewise.
31377         (find_param_candidates): Likewise.
31378         (decide_one_param_reduction): Likewise.
31379         (replace_removed_params_ssa_names): Likewise.
31380         * tree-ssa-ccp.c (ccp_fold_stmt): Likewise.
31381         * tree-ssa-copy.c (dump_copy_of): Likewise.
31382         (copy_prop_visit_cond_stmt): Likewise.
31383         * tree-ssa-dce.c (mark_operand_necessary): Likewise.
31384         * tree-ssa-dom.c (pass_dominator::execute): Likewise.
31385         (record_equivalences_from_stmt): Likewise.
31386         * tree-ssa-dse.c (compute_trims): Likewise.
31387         (delete_dead_call): Likewise.
31388         (delete_dead_assignment): Likewise.
31389         * tree-ssa-forwprop.c (forward_propagate_into_gimple_cond): Likewise.
31390         (forward_propagate_into_cond): Likewise.
31391         (pass_forwprop::execute): Likewise.
31392         * tree-ssa-ifcombine.c (ifcombine_ifandif): Likewise.
31393         * tree-ssa-loop-im.c (invariantness_dom_walker::before_dom_children):
31394         Likewise.
31395         (move_computations_worker): Likewise.
31396         (execute_sm): Likewise.
31397         * tree-ssa-loop-ivcanon.c (tree_estimate_loop_size): Likewise.
31398         (remove_exits_and_undefined_stmts): Likewise.
31399         (remove_redundant_iv_tests): Likewise.
31400         * tree-ssa-loop-ivopts.c (dump_use): Likewise.
31401         (adjust_iv_update_pos): Likewise.
31402         * tree-ssa-math-opts.c (bswap_replace): Likewise.
31403         * tree-ssa-phiopt.c (factor_out_conditional_conversion): Likewise.
31404         (value_replacement): Likewise.
31405         * tree-ssa-phiprop.c (phiprop_insert_phi): Likewise.
31406         * tree-ssa-pre.c (print_pre_expr): Likewise.
31407         (get_representative_for): Likewise.
31408         (create_expression_by_pieces): Likewise.
31409         (insert_into_preds_of_block): Likewise.
31410         (eliminate_insert): Likewise.
31411         (eliminate_dom_walker::before_dom_children): Likewise.
31412         (eliminate): Likewise.
31413         (remove_dead_inserted_code): Likewise.
31414         * tree-ssa-propagate.c (substitute_and_fold): Likewise.
31415         * tree-ssa-reassoc.c (get_rank): Likewise.
31416         (eliminate_duplicate_pair): Likewise.
31417         (eliminate_plus_minus_pair): Likewise.
31418         (eliminate_not_pairs): Likewise.
31419         (undistribute_ops_list): Likewise.
31420         (eliminate_redundant_comparison): Likewise.
31421         (update_range_test): Likewise.
31422         (optimize_range_tests_var_bound): Likewise.
31423         (optimize_vec_cond_expr): Likewise.
31424         (rewrite_expr_tree): Likewise.
31425         (rewrite_expr_tree_parallel): Likewise.
31426         (linearize_expr): Likewise.
31427         (break_up_subtract): Likewise.
31428         (linearize_expr_tree): Likewise.
31429         (attempt_builtin_powi): Likewise.
31430         (attempt_builtin_copysign): Likewise.
31431         (transform_stmt_to_copy): Likewise.
31432         (transform_stmt_to_multiply): Likewise.
31433         (dump_ops_vector): Likewise.
31434         * tree-ssa-sccvn.c (vn_nary_build_or_lookup_1): Likewise.
31435         (print_scc): Likewise.
31436         (set_ssa_val_to): Likewise.
31437         (visit_reference_op_store): Likewise.
31438         (visit_use): Likewise.
31439         (sccvn_dom_walker::before_dom_children): Likewise.
31440         (run_scc_vn): Likewise.
31441         * tree-ssa-scopedtables.c (avail_exprs_stack::lookup_avail_expr):
31442         Likewise.
31443         (expr_hash_elt::print): Likewise.
31444         (const_and_copies::pop_to_marker): Likewise.
31445         (const_and_copies::record_const_or_copy_raw): Likewise.
31446         * tree-ssa-structalias.c (compute_dependence_clique): Likewise.
31447         * tree-ssa-uninit.c (collect_phi_def_edges): Likewise.
31448         (dump_predicates): Likewise.
31449         (find_uninit_use): Likewise.
31450         (warn_uninitialized_phi): Likewise.
31451         (pass_late_warn_uninitialized::execute): Likewise.
31452         * tree-ssa.c (verify_vssa): Likewise.
31453         (verify_ssa): Likewise.
31454         (maybe_optimize_var): Likewise.
31455         * tree-vrp.c (dump_value_range): Likewise.
31456         (dump_all_value_ranges): Likewise.
31457         (dump_asserts_for): Likewise.
31458         (register_edge_assert_for_2): Likewise.
31459         (vrp_visit_cond_stmt): Likewise.
31460         (vrp_visit_switch_stmt): Likewise.
31461         (vrp_visit_stmt): Likewise.
31462         (vrp_visit_phi_node): Likewise.
31463         (simplify_cond_using_ranges_1): Likewise.
31464         (fold_predicate_in): Likewise.
31465         (evrp_dom_walker::before_dom_children): Likewise.
31466         (evrp_dom_walker::push_value_range): Likewise.
31467         (evrp_dom_walker::pop_value_range): Likewise.
31468         (execute_early_vrp): Likewise.
31470 2017-05-16  Richard Biener  <rguenther@suse.de>
31472         * dwarf2out.c (loc_list_from_tree_1): Do not create
31473         DW_OP_GNU_variable_value for DECL_IGNORED_P decls.
31475 2017-05-16  Richard Biener  <rguenther@suse.de>
31477         * dwarf2out.c (resolve_variable_value_in_expr): Lookup DIE
31478         just generated.
31479         (note_variable_value_in_expr): If we resolved the decl ref
31480         do not push to the stack.
31482 2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
31484         * config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
31485         operations in fast-math mode.
31486         (vaddq_f16): Likewise.
31487         (vmul_f16): Likewise.
31488         (vmulq_f16): Likewise.
31489         (vsub_f16): Likewise.
31490         (vsubq_f16): Likewise.
31491         * config/arm/neon.md (add<mode>3): New.
31492         (sub<mode>3): New.
31493         (fma:<VH:mode>3): New.  Also remove outdated comment.
31494         (mul<mode>3): New.
31496 2017-05-16  Martin Liska  <mliska@suse.cz>
31498         PR ipa/79849.
31499         PR ipa/79850.
31500         * ipa-devirt.c (warn_types_mismatch): Fix typo.
31501         (odr_types_equivalent_p): Likewise.
31503 2017-05-15  Sylvestre Ledru  <sylvestre@debian.org>
31505         * plugin.c (try_init_one_plugin): Fix ressource leaks (CID 726637).
31507 2017-05-15  Uros Bizjak  <ubizjak@gmail.com>
31509         PR target/80425
31510         * config/i386.i386.md (*zero_extendsidi2): Do not penalize
31511         non-interunit SSE move alternatives with '?'.
31512         (zero-extendsidi peephole2): New peephole to skip intermediate
31513         general register in SSE zero-extend sequence.
31515 2017-05-15  Jeff Law  <law@redhat.com>
31517         * reorg.c (relax_delay_slots): Create a new variable to hold
31518         the temporary target rather than clobbering TARGET_LABEL.
31520         * config/tilegx/tilegx.c (tilegx_expand_unaligned_load): Add
31521         missing argument to extract_bit_field call.
31522         * config/tilepro/tilepro.c (tilepro_expand_unaligned_load): Likewise.
31524 2017-05-15  Martin Liska  <mliska@suse.cz>
31526         PR driver/31468
31527         * gcc.c (process_command): Do not allow empty argument of -o option.
31529 2017-05-15  Renlin Li  <renlin.li@arm.com>
31531         * config/aarch64/aarch64-protos.h (aarch64_expand_call): Declare.
31532         * config/aarch64/aarch64.c (aarch64_expand_call): Define.
31533         * config/aarch64/constraints.md (Usf): Add long call check.
31534         * config/aarch64/aarch64.md (call): Use aarch64_expand_call.
31535         (call_value): Likewise.
31536         (sibcall): Likewise.
31537         (sibcall_value): Likewise.
31538         (call_insn): New.
31539         (call_value_insn): New.
31540         (sibcall_insn): Update rtx pattern.
31541         (sibcall_value_insn): Likewise.
31542         (call_internal): Remove.
31543         (call_value_internal): Likewise.
31544         (sibcall_internal): Likewise.
31545         (sibcall_value_internal): Likewise.
31546         (call_reg): Likewise.
31547         (call_symbol): Likewise.
31548         (call_value_reg): Likewise.
31549         (call_value_symbol): Likewise.
31551 2017-05-14  Krister Walfridsson  <krister.walfridsson@gmail.com>
31553         PR target/80600
31554         * config/netbsd.h (NETBSD_LIBGCC_SPEC): Always add -lgcc.
31556 2017-05-14  Uros Bizjak  <ubizjak@gmail.com>
31558         * config/i386.i386.c (ix86_cc_modes_compatible): CCNOmode is
31559         compatible with CCGOCmode and with CCZmode.
31561 2017-05-14  Martin Sebor  <msebor@redhat.com>
31563         PR middle-end/77671
31564         * gimple-fold.c (gimple_fold_builtin_sprintf): Make extern.
31565         (gimple_fold_builtin_snprintf): Same.
31566         * gimple-fold.h (gimple_fold_builtin_sprintf): Declare.
31567         (gimple_fold_builtin_snprintf): Same.
31568         * gimple-ssa-sprintf.c (get_format_string): Correct the detection
31569         of character types.
31570         (is_call_safe): New function.
31571         (try_substitute_return_value): Call it.
31572         (try_simplify_call): New function.
31573         (pass_sprintf_length::handle_gimple_call): Call it.
31575 2017-05-14  Martin Sebor  <msebor@redhat.com>
31577         PR middle-end/80669
31578         * builtins.c (expand_builtin_stpncpy): Simplify.
31580 2017-05-14  Daniel Santos  <daniel.santos@pobox.com>
31582         * config/i386/i386.opt: Add option -mcall-ms2sysv-xlogues.
31583         * config/i386/i386.h
31584         (x86_64_ms_sysv_extra_clobbered_registers): Change type to unsigned.
31585         (NUM_X86_64_MS_CLOBBERED_REGS): New macro.
31586         (struct machine_function): Add new members call_ms2sysv,
31587         call_ms2sysv_pad_in, call_ms2sysv_pad_out and call_ms2sysv_extra_regs.
31588         (struct machine_frame_state): New fields sp_realigned and
31589         sp_realigned_offset.
31590         * config/i386/i386.c
31591         (enum xlogue_stub): New enum.
31592         (enum xlogue_stub_sets): New enum.
31593         (class xlogue_layout): New class.
31594         (struct ix86_frame): New fields stack_realign_allocate_offset,
31595         stack_realign_offset and outlined_save_offset.  Modify comments to
31596         detail stack layout when using out-of-line stubs.
31597         (ix86_target_string): Add -mcall-ms2sysv-xlogues option.
31598         (ix86_option_override_internal): Add sorry() for TARGET_SEH and
31599         -mcall-ms2sysv-xlogues.
31600         (stub_managed_regs): New static variable.
31601         (ix86_save_reg): Add new parameter ignore_outlined to optionally omit
31602         registers managed by out-of-line stub.
31603         (disable_call_ms2sysv_xlogues): New function.
31604         (ix86_compute_frame_layout): Modify re-alignment calculations, disable
31605         m->call_ms2sysv when appropriate and compute frame layout for
31606         out-of-line stubs.
31607         (sp_valid_at, fp_valid_at): New inline functions.
31608         (choose_basereg): New function.
31609         (choose_baseaddr): Add align parameter, use choose_basereg and modify
31610         all callers.
31611         (ix86_emit_save_reg_using_mov, ix86_emit_restore_sse_regs_using_mov):
31612         Use align parameter of choose_baseaddr to generated aligned SSE movs
31613         when possible.
31614         (pro_epilogue_adjust_stack): Modify to track
31615         machine_frame_state::sp_realigned.
31616         (ix86_nsaved_regs): Modify to accommodate changes to ix86_save_reg.
31617         (ix86_nsaved_sseregs): Likewise.
31618         (ix86_emit_save_regs): Likewise.
31619         (ix86_emit_save_regs_using_mov): Likewise.
31620         (ix86_emit_save_sse_regs_using_mov): Likewise.
31621         (get_scratch_register_on_entry): Likewise.
31622         (gen_frame_set): New function.
31623         (gen_frame_load): Likewise.
31624         (gen_frame_store): Likewise.
31625         (emit_outlined_ms2sysv_save): Likewise.
31626         (emit_outlined_ms2sysv_restore): Likewise.
31627         (ix86_expand_prologue): Modify stack re-alignment code and call
31628         emit_outlined_ms2sysv_save when appropriate.
31629         (ix86_emit_leave): Clear machine_frame_state::sp_realigned.  Add
31630         parameter rtx_insn *insn, which allows the function to be used to only
31631         generate the notes.
31632         (ix86_expand_epilogue): Modify validity checks of frame and stack
31633         pointers, and call emit_outlined_ms2sysv_restore when appropriate.
31634         (ix86_expand_call): Modify to enable m->call_ms2sysv when appropriate.
31635         * config/i386/predicates.md
31636         (save_multiple): New predicate.
31637         (restore_multiple): Likewise.
31638         * config/i386/sse.md
31639         (save_multiple<mode>): New pattern.
31640         (save_multiple_realign<mode>): Likewise.
31641         (restore_multiple<mode>): Likewise.
31642         (restore_multiple_and_return<mode>): Likewise.
31643         (restore_multiple_leave_return<mode>): Likewise.
31644         * Makefile.in: Export HOSTCXX and HOSTCXXFLAGS to site.exp
31646 2017-05-14  Julia Koval  <julia.koval@intel.com>
31648         * config/i386/i386-builtin-types.def (VOID_FTYPE_INT_INT64): New type.
31649         * config/i386/i386-builtin.def (__builtin_ia32_xgetbv)
31650         (__builtin_ia32_xsetbv): New builtins.
31651         * config/i386/i386.c (ix86_expand_special_args_builtin):
31652         Process new types.
31653         (ix86_expand_builtin): Special expand for new intrinsics.
31654         * config/i386/i386.md (UNSPECV_XGETBV, UNSPECV_XSETBV): New.
31655         (xsetbv, xsetbv_rex64, xgetbv, xgetbv_rex64): New insn patterns.
31656         * config/i386/xsaveintrin.h (_xsetbv, _getbv): New intrinsics.
31658 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31660         * cfganal.c (inverted_post_order_compute): Change argument type
31661         to vec *.
31662         * cfganal.h (inverted_post_order_compute): Adjust prototype.
31663         * df-core.c (rest_of_handle_df_initialize): Adjust.
31664         (rest_of_handle_df_finish): Likewise.
31665         (df_analyze_1): Likewise.
31666         (df_analyze): Likewise.
31667         (loop_inverted_post_order_compute): Change argument to be a vec *.
31668         (df_analyze_loop): Adjust.
31669         (df_get_n_blocks): Likewise.
31670         (df_get_postorder): Likewise.
31671         * df.h (struct df_d): Change field to be a vec.
31672         * lcm.c (compute_laterin): Adjust.
31673         (compute_available): Likewise.
31674         * lra-lives.c (lra_create_live_ranges_1): Likewise.
31675         * tree-ssa-dce.c (remove_dead_stmt): Likewise.
31676         * tree-ssa-pre.c (compute_antic): Likewise.
31678 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31680         * cfganal.c (connect_infinite_loops_to_exit): Adjust.
31681         (depth_first_search::depth_first_search): Change structure init
31682         function to this constructor.
31683         (depth_first_search::add_bb): Rename function to this member.
31684         (depth_first_search::execute): Likewise.
31685         (flow_dfs_compute_reverse_finish): Adjust.
31687 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31689         * ddg.c (find_nodes_on_paths): Use auto_sbitmap.
31690         (longest_simple_path): Likewise.
31691         * shrink-wrap.c (spread_components): Likewise.
31692         (disqualify_problematic_components): Likewise.
31693         (emit_common_heads_for_components): Likewise.
31694         (emit_common_tails_for_components): Likewise.
31695         (insert_prologue_epilogue_for_components): Likewise.
31697 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31699         * tree-ssa-dse.c (dse_dom_walker): Make m_live_byes a
31700         auto_sbitmap.
31702 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31704         * df-core.c (df_set_blocks): Start using auto_bitmap.
31705         (df_compact_blocks): Likewise.
31706         * df-problems.c (df_rd_confluence_n): Likewise.
31707         * df-scan.c (df_insn_rescan_all): Likewise.
31708         (df_process_deferred_rescans): Likewise.
31709         (df_update_entry_block_defs): Likewise.
31710         (df_update_exit_block_uses): Likewise.
31711         (df_entry_block_bitmap_verify): Likewise.
31712         (df_exit_block_bitmap_verify): Likewise.
31713         (df_scan_verify): Likewise.
31714         * lra-constraints.c (lra_constraints): Likewise.
31715         (undo_optional_reloads): Likewise.
31716         (lra_undo_inheritance): Likewise.
31717         * lra-remat.c (calculate_gen_cands): Likewise.
31718         (do_remat): Likewise.
31719         * lra-spills.c (assign_spill_hard_regs): Likewise.
31720         (spill_pseudos): Likewise.
31721         * tree-ssa-pre.c (bitmap_set_and): Likewise.
31722         (bitmap_set_subtract_values): Likewise.
31724 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31726         * haifa-sched.c (estimate_shadow_tick): Replace manual bitmap
31727         management with auto_bitmap.
31728         (fix_inter_tick): Likewise.
31729         (fix_recovery_deps): Likewise.
31730         * ira.c (add_store_equivs): Likewise.
31731         (find_moveable_pseudos): Likewise.
31732         (split_live_ranges_for_shrink_wrap): Likewise.
31733         * print-rtl.c (rtx_reuse_manager::rtx_reuse_manager): Likewise.
31734         (rtx_reuse_manager::seen_def_p): Likewise.
31735         (rtx_reuse_manager::set_seen_def): Likewise.
31736         * print-rtl.h (class rtx_reuse_manager): Likewise.
31738 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31740         * bt-load.c (combine_btr_defs): Use auto_bitmap to manage bitmap
31741         lifetime.
31742         (migrate_btr_def): Likewise.
31743         * cfgloop.c (get_loop_body_in_bfs_order): Likewise.
31744         * df-core.c (loop_post_order_compute): Likewise.
31745         (loop_inverted_post_order_compute): Likewise.
31746         * hsa-common.h: Likewise.
31747         * hsa-gen.c (hsa_bb::~hsa_bb): Likewise.
31748         * init-regs.c (initialize_uninitialized_regs): Likewise.
31749         * ipa-inline.c (resolve_noninline_speculation): Likewise.
31750         (inline_small_functions): Likewise.
31751         * ipa-reference.c (ipa_reference_write_optimization_summary): Likewise.
31752         * ira.c (combine_and_move_insns): Likewise.
31753         (build_insn_chain): Likewise.
31754         * loop-invariant.c (find_invariants): Likewise.
31755         * lower-subreg.c (propagate_pseudo_copies): Likewise.
31756         * predict.c (tree_predict_by_opcode): Likewise.
31757         (predict_paths_leading_to): Likewise.
31758         (predict_paths_leading_to_edge): Likewise.
31759         (estimate_loops_at_level): Likewise.
31760         (estimate_loops): Likewise.
31761         * shrink-wrap.c (try_shrink_wrapping): Likewise.
31762         (spread_components): Likewise.
31763         * tree-cfg.c (remove_edge_and_dominated_blocks): Likewise.
31764         * tree-loop-distribution.c (rdg_build_partitions): Likewise.
31765         * tree-predcom.c (tree_predictive_commoning_loop): Likewise.
31766         * tree-ssa-coalesce.c (coalesce_ssa_name): Likewise.
31767         * tree-ssa-phionlycprop.c (pass_phi_only_cprop::execute): Likewise.
31768         * tree-ssa-pre.c (remove_dead_inserted_code): Likewise.
31769         * tree-ssa-sink.c (nearest_common_dominator_of_uses): Likewise.
31770         * tree-ssa-threadupdate.c (compute_path_counts): Likewise.
31771         (mark_threaded_blocks): Likewise.
31772         (thread_through_all_blocks): Likewise.
31773         * tree-ssa.c (verify_ssa): Likewise.
31774         (execute_update_addresses_taken): Likewise.
31775         * tree-ssanames.c (verify_ssaname_freelists): Likewise.
31777 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31779         * cfganal.c (mark_dfs_back_edges): Replace manual stack with
31780         auto_vec.
31781         (post_order_compute): Likewise.
31782         (inverted_post_order_compute): Likewise.
31783         (pre_and_rev_post_order_compute_fn): Likewise.
31785 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31787         * genrecog.c (int_set::int_set): Explicitly construct our
31788         auto_vec base class.
31789         * vec.h (auto_vec::auto_vec): New constructor.
31791 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31793         * bitmap.h (class auto_bitmap): New constructor taking
31794         bitmap_obstack * argument.
31796 2017-05-13  Trevor Saunders  <tbsaunde+gcc@tbsaunde.org>
31798         * bitmap.h (class auto_bitmap): Change type of m_bits to
31799         bitmap_head, and adjust ctor / dtor and member operators.
31801 2017-05-13  Uros Bizjak  <ubizjak@gmail.com>
31803         * compare-elim.c (equivalent_reg_at_start): Return NULL_RTX
31804         when returned register mode doesn't match original mode.
31806 2017-05-12  Jeff Law  <law@redhat.com>
31807             Jakub Jelinek  <jakub@redhat.com>
31809         * config/mn10300/mn10300.c (mn10300_match_ccmode): Fix where
31810         we look for cc setter after the compare-elim changes.
31811         * config/mn10300/mn10300.md (addsi3_flags): Fix order of patterns
31812         within the vector to match what compare-elim now expects.
31813         (subsi3_flags, andsi3_flags, iorsi3_flags): Likewise.
31814         (xorsi3_flags, one_cmplsi2_flags): Likewise.
31816         * config/rx/rx.c (rx_match_ccmode): Fix where we look cc setter
31817         after the compare-elim changes.
31818         * config/rx/rx.md (abssi2_flags): Fix order of patterns within
31819         the vector to match what compare-elim now expects.
31820         (addsi3_flags, adc_flags, addsi3_flags peepholes): Likewise.
31821         (andsi3_flags, negsi2_flags, one_cmplsi2_flags): Likewise.
31822         (iorsi3_flags, rotlsi3_flags, rotrsi3_flags): Likewise.
31823         (ashrsi3_flags, lshrsi3_flags, ashlsi3_flags): Likewise.
31824         (ssaddsi3, subsi3_flags, sbb_flags, xorsi3_flags): Likewise.
31826         * config/visium/visium.c (single_set_and_flags): Fix where
31827         we look for cc setter after the compare-elim changes.
31828         * config/visium/visium.md (flags_subst_logic): Fix order of patterns
31829         with the vector to match what compare-elim now expects.
31830         (flags_subst_arith, add<mode>3_insn_set_carry): Likewise.
31831         (add<mode>3_insn_set_overflow, addsi3_insn_set_carry): Likewise.
31832         (addsi3_insn_set_overflow, sub<mode>3_insn_set_carry): Likewise.
31833         (sub<mode>3_insn_set_overflow, subsi3_insn_set_carry): Likewise.
31834         (subsi3_insn_set_overflow, negsi2_insn_set_carry): Likewise.
31835         (neg<mode>2_insn_set_overflow): Likewise.
31837 2017-05-12  Jim Wilson  <jim.wilson@linaro.org>
31839         PR middle-end/79794
31840         * expmed.c (extract_bit_field_1): Add alt_rtl argument.  Before
31841         maybe_expand_insn call, set ops[0].target.  If still set after call,
31842         set alt_rtl.  Add extra arg to recursive calls.
31843         (extract_bit_field): Add alt_rtl argument.  Pass to
31844         extract_bit_field.
31845         * expmed.h (extract_bit_field): Fix prototype.
31846         * expr.c (emit_group_load_1, copy_blkmode_from_reg)
31847         (copy_blkmode_to_reg, read_complex_part, store_field): Pass extra NULL
31848         to extract_bit_field_calls.
31849         (expand_expr_real_1): Pass alt_rtl to expand_expr_real instead of 0.
31850         Pass alt_rtl to extract_bit_field calls.
31851         * calls.c (store_unaligned_arguments_into_psuedos)
31852         load_register_parameters): Pass extra NULL to extract_bit_field calls.
31853         * optabs.c (maybe_legitimize_operand): Clear op->target when call
31854         gen_reg_rtx.
31855         * optabs.h (struct expand_operand): Add target bitfield.
31857 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31859         * compare-elim.c (try_eliminate_compare): Canonicalize
31860         operation with embedded compare to
31861         [(set (reg:CCM) (compare:CCM (operation) (immediate)))
31862          (set (reg) (operation)].
31864         * config/i386/i386.c (TARGET_FLAGS_REGNUM): New define.
31866 2017-05-12  Uros Bizjak  <ubizjak@gmail.com>
31868         PR target/80723
31869         * config/i386/i386.c (ix86_rtx_cost) [case PLUS]: Ignore the
31870         cost of adding a carry flag for ADC instruction.
31871         [case MINUS]: Ignore the cost of subtracting a carry flag
31872         for SBB instruction.
31874 2017-05-12  Steven Munroe  <munroesj@gcc.gnu.org>
31876         * config.gcc (powerpc*-*-*): Add bmi2intrin.h, bmiintrin.h,
31877         and x86intrin.h
31878         * config/rs6000/bmiintrin.h: New file.
31879         * config/rs6000/bmi2intrin.h: New file.
31880         * config/rs6000/x86intrin.h: New file.
31882 2017-05-12  Jeff Law  <law@redhat.com>
31884         * tree-vrp.c (vrp_dom_walker::before_dom_children): Push unwinding
31885         markers.
31887 2017-05-12  Peter Bergner  <bergner@vnet.ibm.com>
31889         PR middle-end/80707
31890         * tree-cfg.c: Remove cfg edges of unreachable case statements.
31892 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31894         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31895         early expansion of vector divide builtins.
31896         (builtin_function_type): Add VSX_BUILTIN_UDIV_V2DI to the list of
31897         builtins identified as having unsigned arguments.
31899 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31901         * config/rs6000/rs6000.c (gimple-fold.h): New #include.
31902         (rs6000_gimple_fold_builtin): Add handling for early GIMPLE
31903         expansion of vector logical operations (and, andc, or, xor,
31904         nor, orc, nand).
31906 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31908         * gimple-fold.c (create_tmp_reg_or_ssa_name): Remove static declaration.
31909         * gimple-fold.h (create_tmp_reg_or_ssa_name): New prototype.
31911 2017-05-12  Will Schmidt  <will_schmidt@vnet.ibm.com>
31913         * config/rs6000/rs6000.c (rs6000_gimple_fold_builtin): Add handling for
31914         early GIMPLE expansion of vector multiplies.
31916 2017-05-12  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
31918         * config/arm/arm.md (movsi): Add TARGET_32BIT in addition to the
31919         TARGET_HAVE_MOVT conditional.
31920         (movt splitter): Likewise.
31922 2017-05-12  Richard Biener  <rguenther@suse.de>
31924         * tree-ssa-sccvn.h (has_VN_INFO): Declare.
31925         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
31926         Fold all stmts not inplace.
31928 2017-05-12  Richard Biener  <rguenther@suse.de>
31930         PR tree-optimization/80713
31931         * tree-ssa-pre.c (remove_dead_inserted_code): Clear
31932         inserted_exprs bit for not removed stmts.
31934 2017-05-12  Thomas Schwinge  <thomas@codesourcery.com>
31936         PR middle-end/69921
31937         * tree-parloops.c (create_parallel_loop): Set "oacc kernels
31938         parallelized" attribute for parallelized OpenACC kernels.
31939         * omp-offload.c (execute_oacc_device_lower): Use it.
31941         * omp-expand.c (expand_omp_target) <GF_OMP_TARGET_KIND_OACC_KERNELS>:
31942         Set "oacc kernels" attribute.
31943         * omp-general.c (oacc_set_fn_attrib): Remove is_kernel formal
31944         parameter.  Adjust all users.
31945         (oacc_fn_attrib_kernels_p): Remove function.
31946         * omp-offload.c (execute_oacc_device_lower): Look for "oacc
31947         kernels" attribute instead of calling oacc_fn_attrib_kernels_p.
31948         * tree-ssa-loop.c (gate_oacc_kernels): Likewise.
31949         * tree-parloops.c (create_parallel_loop): If oacc_kernels_p,
31950         assert "oacc kernels" attribute is set.
31952 2017-05-11  Carl Love  <cel@us.ibm.com>
31954         * config/rs6000/rs6000-c: Add support for built-in functions
31955         vector unsigned char vec_popcnt (vector signed char)
31956         vector unsigned char vec_popcnt (vector unsigned char)
31957         vector unsigned short vec_popcnt (vector signed short)
31958         vector unsigned short vec_popcnt (vector unsigned short)
31959         vector unsigned int vec_popcnt (vector signed int)
31960         vector unsigned int vec_popcnt (vector unsigned int)
31961         vector unsigned long long vec_popcnt (vector signed long long)
31962         vector unsigned long long vec_popcnt (vector unsigned long long)
31963         vector signed long long vec_slo (vector signed long long,
31964                                          vector signed char)
31965         vector signed long long vec_slo (vector signed long long,
31966                                          vector unsigned char)
31967         vector unsigned long long vec_slo (vector unsigned long long,
31968                                            vector signed char)
31969         vector unsigned long long vec_slo (vector unsigned long long,
31970                                            vector unsigned char)
31971         * config/rs6000/rs6000-builtin.def: Add definitions for VPOPCNTUB,
31972         VPOPCNTUH, VPOPCNTUW, and VPOPCNTUD overloads.
31973         * config/rs6000/altivec.h: Add define for vec_popcnt, vec_popcntb,
31974         vec_popcnth, vec_popcntw and vec_popcntd built-in functions.
31975         * doc/extend.texi: Update the built-in documentation file for the
31976         new built-in functions.
31978 2017-05-11  Michael Meissner  <meissner@linux.vnet.ibm.com>
31980         * attribs.h (sorted_attr_string): Move machine independent
31981         functions for target clone support from the i386 port to common
31982         code.  Rename ix86_function_versions to common_function_versions.
31983         Rename make_name to make_unique_name.
31984         (common_function_versions): Likewise.
31985         (make_unique_name): Likewise.
31986         (make_dispatcher_decl): Likewise.
31987         (is_function_default_version): Likewise.
31988         * attribs.c (attr_strcmp): Likewise.
31989         (sorted_attr_string): Likewise.
31990         (common_function_versions): Likewise.
31991         (make_unique_name): Likewise.
31992         (make_dispatcher_decl): Likewise.
31993         (is_function_default_version): Likewise.
31994         * config/i386/i386.c (attr_strcmp): Likewise.
31995         (sorted_attr_string): Likewise.
31996         (ix86_function_versions): Likewise.
31997         (make_name): Likewise.
31998         (make_dispatcher_decl): Likewise.
31999         (is_function_default_version): Likewise.
32000         (TARGET_OPTION_FUNCTION_VERSIONS): Update target function hook.
32002 2017-05-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32004         PR target/80695
32005         * config/rs6000/rs6000.c (rs6000_builtin_vectorization_cost):
32006         Account for direct move costs for vec_construct of integer
32007         vectors.
32009 2017-05-11  Uros Bizjak  <ubizjak@gmail.com>
32011         PR target/80706
32012         * config/i386/sync.md (UNSPEC_LDX_ATOMIC): New unspec.
32013         (UNSPEC_STX_ATOMIC): Ditto.
32014         (loaddi_via_sse): New insn.
32015         (storedi_via_sse): Ditto.
32016         (atomic_loaddi_fpu): Emit loaddi_via_sse and storedi_via_sse.
32017         Update corresponding peephole2 patterns.
32018         (atomic_storedi_fpu): Ditto.
32020 2017-05-11  Julia Koval  <julia.koval@intel.com>
32022         * config/i386/avx512fintrin.h (_mm_mask_rsqrt14_sd)
32023         (_mm_maskz_rsqrt14_sd, _mm_mask_rsqrt14_ss, _mm_maskz_rsqrt14_ss):
32024         New intrinsics.
32025         * config/i386/i386-builtin.def (__builtin_ia32_rsqrt14sd_mask)
32026         (__builtin_ia32_rsqrt14ss_mask): New builtins.
32027         * config/i386/sse.md (rsqrt14_<mode>_mask): New pattern.
32029 2017-05-11  Nathan Sidwell  <nathan@acm.org>
32031         * graphite-poly.c: Include dumpfile.h.
32033         * dumpfle.h (dump_function): Declare here ...
32034         * tree-dump.h (dump_function): ... not here.
32035         * dumpfile.c: #include tree-cfg.h.
32036         (dump_function): Move here from ...
32037         * tree-dump.c (dump_function): ... here.
32038         * gimplify.c: #include splay-tree.h, not tree-dump.h.
32039         * graphite-poly.c: Don't include tree-dump.h.
32040         * cgraphclones.c: Include dumpfile.h not tree-dump.h.
32041         * print-tree.c: Likewise.
32042         * stor-layout.c: Likewise.
32043         * tree-nested.c: Likewise.
32045         * dumpfile.c (dump_start): Use TDF_FLAGS.
32046         (dump_enable_all): Fix TDF_KIND check thinko.
32048 2017-05-11  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32050         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
32051         array entries to represent two legal parameterizations of the
32052         overloaded __builtin_cmpb function, as represented by the
32053         P6_OV_BUILTIN_CMPB constant.
32054         (altivec_resolve_overloaded_builtin): Add special case handling
32055         for the __builtin_cmpb function, as represented by the
32056         P6_OV_BUILTIN_CMPB constant.
32057         * config/rs6000/rs6000-builtin.def (BU_P6_2): New macro.
32058         (BU_P6_64BIT_2): New macro.
32059         (BU_P6_OVERLOAD_2): New macro
32060         (CMPB_32): Add 32-bit compare-bytes support for 32-bit only targets.
32061         (CMPB): Add 64-bit compare-bytes support for 32-bit and 64-bit targets.
32062         (CMPB): Add overload support to represent both 32-bit and 64-bit
32063         compare-bytes function.
32064         * config/rs6000/rs6000.c (rs6000_builtin_mask_calculate): Add
32065         support for TARGET_CMPB.
32066         * config/rs6000/rs6000.h: Add support for RS6000_BTM_CMPB.
32067         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
32068         documentation of the __builtin_cmpb overloaded built-in function.
32070 2017-05-11  Richard Biener  <rguenther@suse.de>
32072         PR tree-optimization/80705
32073         * tree-vect-data-refs.c (vect_analyze_data_refs): DECL_NONALIASED
32074         bases are not vectorizable.
32076 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32078         * tree-ssa-loop-ivopts.c (determine_set_costs): Skip non-interger
32079         when counting register pressure.
32081 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32083         * tree-ssa-loop-ivopts.c (compare_cost_pair): New.
32084         (iv_ca_more_deps): Renamed to ...
32085         (iv_ca_compare_deps): ... this.
32086         (iv_ca_extend): Extend iv_ca if NEW_CP is cheaper than OLD_CP.
32088 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32090         * tree-ssa-loop-ivopts.c (find_interesting_uses): Move inv vars dump
32091         to ...
32092         (determine_group_iv_costs): ... here.
32093         (find_inv_vars_cb): Record inv var if it's not recorded before.
32095 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32097         * tree-ssa-loop-ivopts.c (allow_ip_end_pos_p): Refine comments.
32098         (get_shiftadd_cost): Ditto.
32100 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32102         * tree-ssa-address.c: Include header file.
32103         (move_hint_to_base): Return TRUE if BASE_HINT is moved to memory
32104         address.
32105         (add_to_parts): Refactor.
32106         (addr_to_parts): New parameter.  Update use of move_hint_to_base.
32107         (create_mem_ref): Update use of addr_to_parts.  Re-associate addr
32108         in new order.
32110 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32112         PR tree-optimization/53090
32113         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New enum value
32114         COMP_IV_EXPR_2.
32115         (extract_cond_operands): Detect condition with IV on both sides
32116         and return COMP_IV_EXPR_2.
32117         (find_interesting_uses_cond): Add iv_use for both IVs in condition.
32118         (rewrite_use_compare): Simplify by removing call to function
32119         extract_cond_operands.
32121 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32123         * tree-ssa-loop-ivopts.c (enum comp_iv_rewrite): New.
32124         (extract_cond_operands): Detect condition comparing against non-
32125         invariant bound and return appropriate enum value.
32126         (find_interesting_uses_cond): Update use of extract_cond_operands.
32127         Handle its return value accordingly.
32128         (determine_group_iv_cost_cond, rewrite_use_compare): Ditto.
32130 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32132         * tree-ssa-loop-ivopts.c (rewrite_use_nonlinear_expr): Re-associate
32133         nonlinear iv_use computation in loop invariant sensitive way.
32135 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32137         * tree-ssa-loop-ivopts.c (relate_compare_use_with_all_cands): New.
32138         (find_iv_candidates): Call relate_compare_use_with_all_cands.
32140 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32142         * tree-ssa-loop-ivopts.c (struct iv_cand): New field inv_exprs.
32143         (dump_cand): Support iv_cand.inv_exprs.
32144         (add_candidate_1): Record invariant exprs in iv_cand.inv_exprs
32145         for candidates.
32146         (iv_ca_set_no_cp, iv_ca_set_cp, free_loop_data): Support
32147         iv_cand.inv_exprs.
32149 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32151         * tree-ssa-loop-ivopts.c (multiplier_allowed_in_address_p): Move
32152         from ...
32153         * tree-ssa-address.c (multiplier_allowed_in_address_p): ... to here
32154         as local function.  Include necessary header files.
32155         * tree-ssa-loop-ivopts.h (multiplier_allowed_in_address_p): Delete.
32157 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32159         * tree-ssa-loop-ivopts.c (autoinc_possible_for_pair): Simplify.
32161 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32163         * tree-ssa-loop-ivopts.c (force_expr_to_var_cost): Handle more
32164         operators: TRUNC_DIV_EXPR, BIT_AND_EXPR, BIT_IOR_EXPR, LSHIFT_EXPR,
32165         RSHIFT_EXPR and BIT_NOT_EXPR.
32167 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32169         * tree-ssa-loop-ivopts.c (get_loop_invariant_expr): Simplify.
32170         (adjust_setup_cost): New parameter supporting round up adjustment.
32171         (struct address_cost_data): Delete.
32172         (force_expr_to_var_cost): Don't bound cost with spill_cost.
32173         (split_address_cost, ptr_difference_cost): Delete.
32174         (difference_cost, compare_aff_trees, record_inv_expr): Delete.
32175         (struct ainc_cost_data): New struct.
32176         (get_address_cost_ainc): New function.
32177         (get_address_cost, get_computation_cost): Reimplement.
32178         (determine_group_iv_cost_address): Record inv_expr for all uses of
32179         a group.
32180         (determine_group_iv_cost_cond): Call get_loop_invariant_expr.
32181         (iv_ca_has_deps): Reimplemented to ...
32182         (iv_ca_more_deps): ... this.  Check if NEW_CP introduces more deps
32183         than OLD_CP.
32184         (iv_ca_extend): Call iv_ca_more_deps.
32186 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32188         * tree-ssa-address.c (struct mem_address): Move to header file.
32189         (valid_mem_ref_p, move_fixed_address_to_symbol): Make it global.
32190         * tree-ssa-address.h (struct mem_address): Move from C file.
32191         (valid_mem_ref_p, move_fixed_address_to_symbol): Declare.
32193 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32195         * tree-affine.h (aff_combination_type): New interface.
32196         (aff_combination_zero_p): Remove static.
32197         (aff_combination_const_p): New interface.
32198         (aff_combination_singleton_var_p): New interfaces.
32200 2017-05-11  Richard Biener  <rguenther@suse.de>
32202         * tree-ssa-pre.c (eliminate_dom_walker::before_dom_children):
32203         Skip unreachable blocks and destinations.
32204         (eliminate): Move stmt removal and fixup ...
32205         (fini_eliminate): ... here.  Skip inserted exprs.
32206         (pass_pre::execute): Move fini_pre after fini_eliminate.
32207         * tree-ssa-tailmerge.c: Include tree-cfgcleanup.h.
32208         (tail_merge_optimize): Run cleanup_tree_cfg if requested by
32209         PRE to get rid of dead code that has invalid SSA form and
32210         split critical edges again.
32212 2017-05-11  Bin Cheng  <bin.cheng@arm.com>
32214         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
32216 2017-05-11  Richard Biener  <rguenther@suse.de>
32218         * passes.c (execute_function_todo): Verify loops if they are
32219         said to be up-to-date.
32220         * cfgexpand.c (pass_expand::execute): Discard loops for -dx.
32221         * trans-mem.c (pass_tm_edges::execute): Mark loops for fixup.
32223 2017-05-10  John David Anglin  <danglin@gcc.gnu.org>
32225         PR target/80090
32226         * config/pa/pa.c (pa_assemble_integer): When outputting a SYMBOL_REF,
32227         handle calling assemble_external ourself.
32229         PR target/79027
32230         * config/pa/pa.c (pa_cannot_change_mode_class): Reject changes to/from
32231         modes with zero size.  Enhance comment.
32233 2017-05-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32235         * config/rs6000/rs6000.c (altivec_init_builtins): Define POWER8
32236         built-ins for vec_xl and vec_xst with short and char pointer
32237         arguments.
32239 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32241         * config/i386/avx512fintrin.h (_mm_mask_max_round_sd)
32242         (_mm_maskz_max_round_sd, _mm_mask_max_round_ss)
32243         (_mm_maskz_max_round_ss, _mm_mask_min_round_sd)
32244         (_mm_maskz_min_round_sd, _mm_mask_min_round_ss)
32245         (_mm_maskz_min_round_ss): New intrinsics.
32246         * config/i386/i386-builtin-types.def (V2DF, V2DF, V2DF, V2DF, UQI, INT)
32247         (V4SF, V4SF, V4SF, V4SF, UQI, INT): New function type aliases.
32248         * config/i386/i386-builtin.def (__builtin_ia32_maxsd_mask_round)
32249         (__builtin_ia32_maxss_mask_round, __builtin_ia32_minsd_mask_round)
32250         (__builtin_ia32_minss_mask_round): New builtins.
32251         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32252         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32253         * config/i386/sse.md (<sse>_vm<code><mode>3<round_saeonly_name>):
32254         Rename to ...
32255         (<sse>_vm<code><mode>3<mask_name><round_saeonly_name>): ... this.
32256         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_op3>%2, %1, %0|%0, %1, %<iptr>2<round_saeonly_op3>}):
32257         Change to ...
32258         (v<maxmin_float><ssescalarmodesuffix>\t{<round_saeonly_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_saeonly_mask_op3>}):
32259         ... this.
32261 2017-05-10  Sebastian Peryt  <sebastian.peryt@intel.com>
32263         * config/i386/avx512fintrin.h (_mm_mask_mul_round_sd)
32264         (_mm_maskz_mul_round_sd, _mm_mask_mul_round_ss)
32265         (_mm_maskz_mul_round_ss, _mm_mask_div_round_sd)
32266         (_mm_maskz_div_round_sd, _mm_mask_div_round_ss)
32267         (_mm_maskz_div_round_ss, _mm_mask_mul_sd, _mm_maskz_mul_sd)
32268         (_mm_mask_mul_ss, _mm_maskz_mul_ss, _mm_mask_div_sd)
32269         (_mm_maskz_div_sd, _mm_mask_div_ss, _mm_maskz_div_ss): New intrinsics.
32270         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32271         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
32272         * config/i386/i386-builtin.def (__builtin_ia32_divsd_mask_round)
32273         (__builtin_ia32_divss_mask_round, __builtin_ia32_mulsd_mask_round)
32274         (__builtin_ia32_mulss_mask_round): New builtins.
32275         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
32276         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
32277         * config/i386/sse.md (<sse>_vm<multdiv_mnemonic><mode>3<round_name>):
32278         Rename to ...
32279         (<sse>_vm<multdiv_mnemonic><mode>3<mask_name><round_name>): ... this.
32280         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
32281         Change to ...
32282         (v<multdiv_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
32283         ... this.
32285 2017-05-10  Julia Koval  <julia.koval@intel.com>
32287         * config/i386/avxintrin.h (_mm256_set_m128, _mm256_set_m128d)
32288         (_mm256_set_m128i, _mm256_setr_m128, _mm256_setr_m128d)
32289         (_mm256_setr_m128i): New intrinsics.
32291 2017-05-10  Julia Koval  <julia.koval@intel.com>
32293         * config/i386/avx512fintrin.h (_mm_mask_rcp14_sd)
32294         (_mm_maskz_rcp14_sd, _mm_mask_rcp14_ss)
32295         (_mm_maskz_rcp14_ss): New intrinsics.
32296         * config/i386/i386-builtin.def (__builtin_ia32_rcp14sd_mask)
32297         (__builtin_ia32_rcp14ss_mask): New builtins.
32298         * config/i386/sse.md (srcp14<mode>_mask): New pattern.
32300 2017-05-10  Peter Bergner  <bergner@vnet.ibm.com>
32302         PR tree-optimization/51513
32303         * tree-cfg.c (gimple_seq_unreachable_p): New function.
32304         (assert_unreachable_fallthru_edge_p): Use it.
32305         (group_case_labels_stmt): Likewise.
32306         * tree-cfg.h: Prototype it.
32307         * stmt.c: Include cfghooks.h and tree-cfg.h.
32308         (emit_case_dispatch_table) <gap_label>: New local variable.
32309         Use it to fill dispatch table gaps.
32310         Test for default_label before updating probabilities.
32311         (expand_case) <default_label>: Remove unneeded initialization.
32312         Test for unreachable default case statement and remove its edge.
32313         Set default_label accordingly.
32314         * tree-ssa-ccp.c (optimize_unreachable): Update comment.
32316 2017-05-10  Carl Love  <cel@us.ibm.com>
32318         * config/rs6000/rs6000-c: Add support for built-in functions
32319         vector signed char      vec_neg (vector signed char)
32320         vector signed short int vec_neg (vector short int)
32321         vector signed int       vec_neg (vector signed int)
32322         vector signed long long vec_neg (vector signed long long)
32323         vector float            vec_neg (vector float)
32324         vector double           vec_neg (vector double)
32325         * config/rs6000/rs6000-builtin.def: Add definitions for NEG function
32326         overload.
32327         * config/rs6000/altivec.h: Add define for vec_neg
32328         * doc/extend.texi: Update the built-in documentation for the
32329         new built-in functions.
32331 2017-05-10  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32333         PR tree-optimization/77644
32334         * match.pd (sqrt(x) cmp sqrt(y) -> x cmp y): New pattern.
32336 2017-05-10  Nathan Sidwell  <nathan@acm.org>
32338         * dumpfile.h (TDI_lang_all): New.
32339         (TDF_KIND): New. Renumber others
32340         (TDF_LANG, TDF_TREE, TDF_RTL, TDF_IPA): Enumerate value, rather
32341         than bits.
32342         * dumpfile.c (dump_files): Mark language dumps as TDF_LANG.  add
32343         lang-all.
32344         (get_dump_file_name): Adjust suffix generation.
32345         (dump_enable_all): Use TDF_KIND.
32346         * doc/invoke.texi (-fdump-lang-all): Document.
32348         * dumpfile.h: Tabify.
32350 2017-05-10  Wilco Dijkstra  <wdijkstr@arm.com>
32352         PR target/80671
32353         * config/aarch64/cortex-a57-fma-steering.c (merge_forest):
32354         Move member access before delete.
32356 2017-05-10  Alexandre Oliva  <aoliva@redhat.com>
32358         * tree-inline.c (expand_call_inline): Split block at stmt
32359         before the call.
32361 2017-05-09  Michael Meissner  <meissner@linux.vnet.ibm.com>
32363         PR target/68163
32364         * config/rs6000/rs6000.md (f32_lr): Delete mode attributes that
32365         are now unused after splitting mov{sf,sd}_hardfloat.
32366         (f32_lr2): Likewise.
32367         (f32_lm): Likewise.
32368         (f32_lm2): Likewise.
32369         (f32_li): Likewise.
32370         (f32_li2): Likewise.
32371         (f32_lv): Likewise.
32372         (f32_sr): Likewise.
32373         (f32_sr2): Likewise.
32374         (f32_sm): Likewise.
32375         (f32_sm2): Likewise.
32376         (f32_si): Likewise.
32377         (f32_si2): Likewise.
32378         (f32_sv): Likewise.
32379         (f32_dm): Likewise.
32380         (f32_vsx): Likewise.
32381         (f32_av): Likewise.
32382         (mov<mode>_hardfloat): Split into separate movsf and movsd pieces.
32383         For movsf, order stores so the VSX stores occur before the GPR
32384         store which encourages the register allocator to use a traditional
32385         FPR instead of a GPR.  For movsd, order the stores so that the GPR
32386         store comes before the VSX stores to allow the power6 to work.
32387         This is due to the power6 not having a 32-bit integer store
32388         instruction from a FPR.
32389         (movsf_hardfloat): Likewise.
32390         (movsd_hardfloat): Likewise.
32392 2017-05-09  Martin Sebor  <msebor@redhat.com>
32394         PR translation/80280
32395         * config/sol2-c.c (cmn_err_flag_specs): Initialize new data member
32396         added in r247778.
32398         PR translation/80280
32399         * config/i386/msformat-c.c (ms_printf_flag_specs): Initialize new
32400         data member added in r247778.
32401         (ms_scanf_flag_specs, ms_strftime_flag_specs): Same.
32403 2017-05-09  Nathan Sidwell  <nathan@acm.org>
32405         * tree.h (tree_fits_shwi_p, tree_fits_uhwi_p): Unconditionally pure.
32407         * ipa-devirt.c (default_hash_traits<type_pair>): Skip struct-scope
32408         typedefs.
32410 2017-05-09  Marek Polacek  <polacek@redhat.com>
32412         * doc/invoke.texi: Fix typo.
32414 2017-05-09  Richard Biener  <rguenther@suse.de>
32416         * tree-vrp.c (vrp_val_is_max): Adjust comment.
32417         (vrp_val_is_min): Likewise.
32418         (set_value_range_to_value): Likewise.
32419         (set_value_range_to_nonnegative): Likewise.
32420         (gimple_assign_nonzero_p): Likewise.
32421         (gimple_stmt_nonzero_p): Likewise.
32422         (vrp_int_const_binop): Likewise.  Remove unreachable case.
32423         (adjust_range_with_scev): Adjust comments.
32424         (compare_range_with_value): Likewise.
32425         (extract_range_from_phi_node): Likewise.
32426         (test_for_singularity): Likewise.
32428 2017-05-09  Richard Biener  <rguenther@suse.de>
32430         * tree-vrp.c (get_single_symbol): Add assert that we don't
32431         get overflowed constants as invariant part.
32432         (compare_values_warnv): Add comment before the TREE_NO_WARNING
32433         checks.  Use wi::cmp instead of recursing for integer constants.
32434         (compare_values): Just ignore whether we assumed undefined
32435         overflow instead of failing the compare.
32436         (extract_range_for_var_from_comparison_expr): Add comment before the
32437         TREE_NO_WARNING sets.
32438         (test_for_singularity): Likewise.
32439         (extract_range_from_comparison): Do not disable optimization
32440         when we assumed undefined overflow.
32441         (extract_range_basic): Remove init of unused var.
32443 2017-05-09  Richard Biener  <rguenther@suse.de>
32445         * tree-vrp.c (vrp_int_const_binop): Use wide-ints and simplify.
32446         (extract_range_from_multiplicative_op_1): Adjust.
32447         (extract_range_from_binary_expr_1): Use int_const_binop.
32449 2017-05-08  Kelvin Nilsen  <kelvin@gcc.gnu.org>
32451         PR target/80101
32452         * config/rs6000/power6.md: Replace store_data_bypass_p calls with
32453         rs6000_store_data_bypass_p in seven define_bypass directives and
32454         in several comments.
32455         * config/rs6000/rs6000-protos.h: Add prototype for
32456         rs6000_store_data_bypass_p function.
32457         * config/rs6000/rs6000.c (rs6000_store_data_bypass_p): New
32458         function implements slightly different (rs6000-specific) semantics
32459         than store_data_bypass_p, returning false rather than aborting
32460         with assertion error when arguments do not satisfy the
32461         requirements of store data bypass.
32462         (rs6000_adjust_cost): Replace six calls of store_data_bypass_p with
32463         rs6000_store_data_bypass_p.
32465 2017-05-08  Max Filippov  <jcmvbkbc@gmail.com>
32467         * config/xtensa/xtensa-protos.h
32468         (xtensa_initial_elimination_offset): New declaration.
32469         * config/xtensa/xtensa.c (xtensa_initial_elimination_offset):
32470         New function. Move its body from the INITIAL_ELIMINATION_OFFSET
32471         macro definition, add case for FRAME_POINTER_REGNUM when
32472         FRAME_GROWS_DOWNWARD.
32473         * config/xtensa/xtensa.h (FRAME_GROWS_DOWNWARD): New macro definition.
32474         (INITIAL_ELIMINATION_OFFSET): Replace body with call to
32475         xtensa_initial_elimination_offset.
32477 2017-05-08  Nathan Sidwell  <nathan@acm.org>
32479         * doc/invoke.texi: Alphabetize -fdump options.
32481 2017-05-08  Martin Sebor  <msebor@redhat.com>
32483         PR translation/80280
32484         * config/sol2-c.c (solaris_pragma_align): Correct quoting.
32486 2017-05-08  Bernd Edlinger  <bernd.edlinger@hotmail.de>
32488         * target.def (compute_frame_layout): New optional target hook.
32489         * doc/tm.texi.in (TARGET_COMPUTE_FRAME_LAYOUT): Add hook.
32490         * doc/tm.texi (TARGET_COMPUTE_FRAME_LAYOUT): Add documentation.
32491         * lra-eliminations.c (update_reg_eliminate): Call compute_frame_layout
32492         target hook.
32493         * reload1.c (verify_initial_elim_offsets): Likewise.
32494         * config/arm/arm.c (TARGET_COMPUTE_FRAME_LAYOUT): Define.
32495         (use_simple_return_p): Call arm_compute_frame_layout if needed.
32496         (arm_get_frame_offsets): Split up into this ...
32497         (arm_compute_frame_layout): ... and this function.
32499 2017-05-08  Richard Sandiford  <richard.sandiford@arm.com>
32501         * config/aarch64/constraints.md (Usa): New constraint.
32502         * config/aarch64/aarch64.md (*movsi_aarch64, *movdi_aarch64): Use it.
32504 2017-05-08  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32506         * config.gcc (arm*-*-*): Set TM_MULTILIB_CONFIG from
32507         with_multilib_list after it has been checked.
32509 2017-05-08  Richard Biener  <rguenther@suse.de>
32511         * tree-ssa-pre.c (bitmap_set_and): Avoid bitmap copy.
32512         (bitmap_set_subtract_values): Likewise.
32514 2017-05-08  Richard Biener  <rguenther@suse.de>
32516         * tree-vrp.c (gimple_assign_nonzero_warnv_p): Rename to ...
32517         (gimple_assign_nonzero): ... this and remove strict_overflow_p
32518         argument.
32519         (gimple_stmt_nonzero_warnv_p): Rename to ...
32520         (gimple_stmt_nonzero_p): ... this and remove strict_overflow_p
32521         argument.
32522         (vrp_stmt_computes_nonzero): Remove strict_overflow_p argument.
32523         (extract_range_basic): Adjust, do not disable propagation on
32524         strict overflow sensitive simplification.
32525         (vrp_visit_cond_stmt): Likewise.
32527 2017-05-05  Jan Hubicka  <hubicka@ucw.cz>
32529         * ipa-inline-analysis.c (estimate_function_body_sizes): Recompute
32530         body size unconditionally.
32532 2017-05-07  Jeff Law  <law@redhat.com>
32534         Revert:
32535         2017-05-06  Jeff Law  <law@redhat.com>
32536         PR tree-optimization/78496
32537         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32538         code.
32540         PR tree-optimization/78496
32541         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32542         (simplify_stmt_using_ranges): Call it.
32543         (vrp_dom_walker::before_dom_children): Extract equivalences
32544         from an ASSERT_EXPR with an equality comparison against a
32545         constant.
32547 2017-05-06  Jeff Law  <law@redhat.com>
32549         PR tree-optimization/78496
32550         * tree-vrp.c (simplify_assert_expr_using_ranges): Remove debugging
32551         code.
32553         PR tree-optimization/78496
32554         * tree-vrp.c (simplify_assert_expr_using_ranges): New function.
32555         (simplify_stmt_using_ranges): Call it.
32556         (vrp_dom_walker::before_dom_children): Extract equivalences
32557         from an ASSERT_EXPR with an equality comparison against a
32558         constant.
32560 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32562         * lra-constraints.c (lra_copy_reg_equiv): New function.
32563         (split_reg): Use it to copy equivalence information from the
32564         original register to the spill register.
32566 2017-05-06  Richard Sandiford  <richard.sandiford@linaro.org>
32568         PR rtl-optimization/75964
32569         * simplify-rtx.c (simplify_const_relational_operation): Remove
32570         invalid handling of comparisons of integer ABS.
32572 2017-05-06  Uros Bizjak  <ubizjak@gmail.com>
32574         * config/i386/i386.c (ext_80387_constant_init): Do not explicitly
32575         initialize to zero.
32576         (init_regs): Remove declaration.
32577         (function_arg_advance_32): Initialize error_p as boolean variable.
32579 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32581         * store-motion.c (remove_reachable_equiv_notes): Reformat long
32582         lines.  Use for (;;).
32584 2017-05-05  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
32586         * config/rs6000/rs6000.c (rs6000_vect_nonmem): New static var.
32587         (rs6000_init_cost): Initialize rs6000_vect_nonmem.
32588         (rs6000_add_stmt_cost): Update rs6000_vect_nonmem.
32589         (rs6000_finish_cost): Avoid vectorizing simple copy loops with
32590         VF=2 that require versioning.
32592 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32594         * diagnostic.h (CARET_LINE_MARGIN): Convert from macro to const
32595         int.
32597 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32599         * diagnostic.h (diagnostic_override_option_index): Convert from
32600         macro to inline function.
32602 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32604         * diagnostic.c (last_module_changed_p): New function.
32605         (set_last_module): New function.
32606         (diagnostic_report_current_module): Convert macro usage to
32607         the above functions.
32608         * diagnostic.h (diagnostic_context::last_module): Strengthen
32609         from const line_map * to const line_map_ordinary *.
32610         (diagnostic_last_module_changed): Delete macro.
32611         (diagnostic_set_last_module): Delete macro.
32613 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32615         * diagnostic.c (diagnostic_impl): Replace report_diagnostic
32616         with diagnostic_report_diagnostic.
32617         (diagnostic_n_impl_richloc): Likewise.
32618         * diagnostic.h (report_diagnostic): Delete macro.
32619         * rtl-error.c (diagnostic_for_asm): Replace report_diagnostic
32620         with diagnostic_report_diagnostic.
32621         * substring-locations.c (format_warning_va): Likewise.
32623 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32625         * diagnostic.c (diagnostic_report_diagnostic): Eliminate
32626         save/restor of format_spec.  Move option-printing code to...
32627         (print_option_information): ...this new function, and
32628         reimplement by simply printing to the pretty_printer,
32629         rather than appending to the format string.
32631 2017-05-05  David Malcolm  <dmalcolm@redhat.com>
32633         * diagnostic.c (diagnostic_report_diagnostic): Split out pragma
32634         handling logic into...
32635         (update_effective_level_from_pragmas): ...this new function.
32637 2017-05-04  Andrew Waterman  <andrew@sifive.com>
32639         * config/riscv/riscv.opt (mstrict-align): New option.
32640         * config/riscv/riscv.h (STRICT_ALIGNMENT): Use it.  Update comment.
32641         (SLOW_UNALIGNED_ACCESS): Define.
32642         (riscv_slow_unaligned_access): Declare.
32643         * config/riscv/riscv.c (riscv_tune_info): Add slow_unaligned_access
32644         field.
32645         (riscv_slow_unaligned_access): New variable.
32646         (rocket_tune_info): Set slow_unaligned_access to true.
32647         (optimize_size_tune_info): Set slow_unaligned_access to false.
32648         (riscv_cpu_info_table): Add entry for optimize_size_tune_info.
32649         (riscv_valid_lo_sum_p): Use TARGET_STRICT_ALIGN.
32650         (riscv_option_override): Set riscv_slow_unaligned_access.
32651         * doc/invoke.texi: Add -mstrict-align to RISC-V.
32653 2017-05-04  Kito Cheng  <kito.cheng@gmail.com>
32655         * config/riscv/riscv.md: Unify indentation.
32657 2017-05-05  Michael Meissner  <meissner@linux.vnet.ibm.com>
32659         PR target/79038
32660         PR target/79202
32661         PR target/79203
32662         * config/rs6000/rs6000.md (u code attribute): Add FIX and
32663         UNSIGNED_FIX.
32664         (extendsi<mode>2): Add support for doing sign extension via
32665         VUPKHSW and XXPERMDI if the value is in Altivec registers and we
32666         don't have ISA 3.0 instructions.
32667         (extendsi<mode>2 splitter): Likewise.
32668         (fix_trunc<mode>si2): If we are at ISA 2.07 (VSX small integer),
32669         generate the normal insns since SImode can now go in vector
32670         registers.  Disallow the special UNSPECs needed for previous
32671         machines to hide SImode being used.  Add new insns
32672         fctiw{,w}_<mode>_smallint if SImode can go in vector registers.
32673         (fix_trunc<mode>si2_stfiwx): Likewise.
32674         (fix_trunc<mode>si2_internal): Likewise.
32675         (fixuns_trunc<mode>si2): Likewise.
32676         (fixuns_trunc<mode>si2_stfiwx): Likewise.
32677         (fctiw<u>z_<mode>_smallint): Likewise.
32678         (fctiw<u>z_<mode>_mem): New combiner pattern to prevent conversion
32679         of floating point to 32-bit integer from doing a direct move to
32680         the GPR registers to do a store.
32681         (fctiwz_<mode>): Break long line.
32683 2017-05-05  Bin Cheng  <bin.cheng@arm.com>
32685         * Makefile.in (GTFILES): Add tree-ssa-loop-ivopts.c.
32686         * tree-ssa-loop-ivopts.c (compute_max_addr_offset): Delete.
32687         (addr_list, addr_offset_valid_p): New.
32688         (split_address_groups): Check offset validity with above function.
32689         (gt-tree-ssa-loop-ivopts.h): Include header file.
32691 2017-05-05  Nathan Sidwell  <nathan@acm.org>
32693         * config.gcc (arm*-*-*): Add missing 'fi'.
32695 2017-05-05  Steve Ellcey  <sellcey@cavium.com>
32697         * doc/invoke.texi (-fopt-info): Explicitly say order of options
32698         included in -fopt-info does not matter.
32699         * doc/optinfo.texi (-fopt-info): Fix description of default
32700         behavour. Explicitly say order of options included in -fopt-info
32701         does not matter.
32703 2017-05-05  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32705         * config.gcc: Allow combinations of aprofile and rmprofile values for
32706         --with-multilib-list.
32707         * config/arm/t-multilib: New file.
32708         * config/arm/t-aprofile: Remove initialization of MULTILIB_*
32709         variables.  Remove setting of ISA and floating-point ABI in
32710         MULTILIB_OPTIONS and MULTILIB_DIRNAMES.  Set architecture and FPU in
32711         MULTI_ARCH_OPTS_A and MULTI_ARCH_DIRS_A rather than MULTILIB_OPTIONS
32712         and MULTILIB_DIRNAMES respectively.  Add comment to introduce all
32713         matches.  Add architecture matches for marvel-pj4 and generic-armv7-a
32714         CPU options.
32715         * config/arm/t-rmprofile: Likewise except for the matches changes.
32716         * doc/install.texi (--with-multilib-list): Document the combination of
32717         aprofile and rmprofile values and warn about pitfalls in doing that.
32719 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32721         * config/aarch64/aarch64.md (movsi_aarch64): Remove '*' from r=w.
32722         (movdi_aarch64): Likewise.
32724 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32726         PR tree-optimization/80632
32727         * tree-switch-conversion.c (struct switch_conv_info): Add target_vop
32728         field.
32729         (build_arrays): Initialize it for virtual phis.
32730         (fix_phi_nodes): Use it for virtual phis.
32732         PR tree-optimization/80558
32733         * tree-vrp.c (extract_range_from_binary_expr_1): Optimize
32734         [x, y] op z into [x op, y op z] for op & or | if conditions
32735         are met.
32737 2017-05-05  Andre Vieira  <andre.simoesdiasvieira@arm.com>
32738             Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32740         PR target/71607
32741         * config/arm/arm.md (use_literal_pool): Remove.
32742         (64-bit immediate split): No longer takes cost into consideration
32743         if arm_disable_literal_pool is enabled.
32744         * config/arm/arm.c (arm_tls_referenced_p): Add diagnostic if TLS is
32745         used when arm_disable_literal_pool is enabled.
32746         (arm_max_const_double_inline_cost): Remove use of
32747         arm_disable_literal_pool.
32748         (push_minipool_fix): Add assert.
32749         (arm_reorg): Add return if arm_disable_literal_pool is enabled.
32750         * config/arm/vfp.md (no_literal_pool_df_immediate): New.
32751         (no_literal_pool_sf_immediate): New.
32753 2017-05-05  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
32755         PR tree-optimization/80613
32756         * tree-ssa-dce.c (propagate_necessity): Remove cases for
32757         BUILT_IN_STRDUP and BUILT_IN_STRNDUP.
32759 2017-05-05  Richard Biener  <rguenther@suse.de>
32761         * tree-ssa-pre.c (get_or_alloc_expr_for): Simplify.
32763 2017-05-05  Georg-Johann Lay  <avr@gjlay.de>
32765         * config/avr/avr.md [flag_strict_overflow]: Remove any occurence
32766         of this flag from insn conditions due to removal from r247495.
32768 2017-05-05  Wilco Dijkstra  <wdijkstr@arm.com>
32770         * config/arm/aarch-common.c (arm_early_load_addr_dep_ptr):
32771         New function.
32772         (arm_early_store_addr_dep_ptr): Likewise.
32773         * config/arm/aarch-common-protos.h
32774         (arm_early_load_addr_dep_ptr): Add prototype.
32775         (arm_early_store_addr_dep_ptr): Likewise.
32776         * config/arm/cortex-a53.md: Add new bypasses.
32778 2017-05-05  Jakub Jelinek  <jakub@redhat.com>
32780         * tree.c (next_type_uid): Change type to unsigned.
32781         (type_hash_canon): Decrement back next_type_uid if
32782         freeing a type node with the highest TYPE_UID.  For INTEGER_TYPEs
32783         also ggc_free TYPE_MIN_VALUE, TYPE_MAX_VALUE and TYPE_CACHED_VALUES
32784         if possible.
32786 2017-05-04  Martin Sebor  <msebor@redhat.com>
32788         * builtins.c: Fix a trivial typo in a comment.
32790         PR middle-end/79234
32791         * builtins.c (check_sizes): Adjust to handle reading past the end.
32792         Avoid printing excessive upper bound of ranges.  Use %E to print
32793         tree nodes instead of converting them to %wu.
32794         (expand_builtin_memchr): New function.
32795         (compute_dest_size): Rename...
32796         (compute_objsize): ...to this.
32797         (expand_builtin_memcpy): Adjust.
32798         (expand_builtin_mempcpy): Adjust.
32799         (expand_builtin_strcat): Adjust.
32800         (expand_builtin_strcpy): Adjust.
32801         (check_strncat_sizes): Adjust.
32802         (expand_builtin_strncat): Adjust.
32803         (expand_builtin_strncpy): Adjust and simplify.
32804         (expand_builtin_memset): Adjust.
32805         (expand_builtin_bzero): Adjust.
32806         (expand_builtin_memcmp): Adjust.
32807         (expand_builtin): Handle memcmp.
32808         (maybe_emit_chk_warning): Check strncat just once.
32810 2017-05-04  Martin Sebor  <msebor@redhat.com>
32812         PR preprocessor/79214
32813         PR middle-end/79222
32814         PR middle-end/79223
32815         * builtins.c (check_sizes): Add inlining context and issue
32816         warnings even when -Wno-system-headers is set.
32817         (check_strncat_sizes): Same.
32818         (expand_builtin_strncat): Same.
32819         (expand_builtin_memmove): New function.
32820         (expand_builtin_stpncpy): Same.
32821         (expand_builtin): Handle memmove and stpncpy.
32823 2017-05-04  Bin Cheng  <bin.cheng@arm.com>
32825         * tree-ssa-loop-ivopts.c (struct cost_pair): Remove field inv_expr
32826         which is not used any more.
32828 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32830         * config/aarch64/aarch64.c (generic_tunings): Update prefetch model.
32832 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32834         * config/aarch64/aarch64.c (cortexa35_tunings): Set jump alignment to 4.
32835         (cortexa53_tunings): Likewise.
32836         (cortexa57_tunings): Likewise.
32837         (cortexa72_tunings): Likewise.
32838         (cortexa73_tunings): Likewise.
32840 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32842         * config/aarch64/aarch64.c (generic_tunings): Set jump alignment to 4.
32843         Set loop alignment to 8.
32845 2017-05-04  Martin Sebor  <msebor@redhat.com>
32847         PR translation/80280
32848         * builtins.c (expand_builtin_object_size): Add missing quoting to
32849         %D and like directives.
32850         * hsa-gen.c (hsa_type_for_scalar_tree_type): Same.
32851         (hsa_type_for_tree_type): Same.
32852         (verify_function_arguments): Same.
32853         * symtab.c (symbol_table::change_decl_assembler_name): Same.
32854         * varasm.c (get_section): Same.
32855         (mark_weak): Same.
32857 2017-05-04  Martin Sebor  <msebor@redhat.com>
32859         PR translation/80280
32860         * config/i386/i386.c (ix86_function_versions): Quote a %D directive.
32862 2017-05-04  Wilco Dijkstra  <wdijkstr@arm.com>
32864         * config/aarch64/aarch64.c (generic_addrcost_table):
32865         Change HI/TI mode setting.
32867 2017-05-04  Martin Jambor  <mjambor@suse.cz>
32869         PR tree-optimization/80622
32870         * tree-sra.c (comes_initialized_p): New function.
32871         (build_accesses_from_assign): Only set write lazily when
32872         comes_initialized_p is false.
32873         (analyze_access_subtree): Use comes_initialized_p.
32874         (propagate_subaccesses_across_link): Assert !comes_initialized_p
32875         instead of testing for PARM_DECL.
32877 2017-05-04  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
32879         * config/aarch64/aarch64.md (prefetch); Adjust predicate and
32880         constraint on operand 0 to allow more general addressing modes.
32881         Adjust output template.
32882         * config/aarch64/aarch64.c (aarch64_address_valid_for_prefetch_p):
32883         New function.
32884         * config/aarch64/aarch64-protos.h
32885         (aarch64_address_valid_for_prefetch_p): Declare prototype.
32886         * config/aarch64/constraints.md (Dp): New address constraint.
32887         * config/aarch64/predicates.md (aarch64_prefetch_operand): New
32888         predicate.
32890 2017-05-04  Jan Hubicka  <hubicka@ucw.cz>
32892         * ipa-cp.c (perform_estimation_of_a_value): Drop base_time parameter;
32893         update use of estimate_ipcp_clone_size_and_time.
32894         (estimate_local_effects): Update use of
32895         estimate_ipcp_clone_size_and_time and perform_estimation_of_a_value.
32896         * ipa-inline.h (estimate_ipcp_clone_size_and_time): Update prototype.
32897         * ipa-inline-analysis.c (estimate_ipcp_clone_size_and_time):
32898         Return nonspecialized time.
32900 2017-05-04  Richard Biener  <rguenther@suse.de>
32902         * tree-ssa-alias.c (get_continuation_for_phi): Improve looking
32903         for the last VUSE which def dominates the PHI.  Directly call
32904         maybe_skip_until.
32905         (get_continuation_for_phi_1): Remove.
32907 2017-05-04  Richard Sandiford  <richard.sandiford@linaro.org>
32909         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): Add commentary
32910         to explain the use of truncating division.  Cap the number of
32911         iterations to the maximum given by nb_iterations_upper_bound,
32912         if defined.
32914 2017-05-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
32916         * configure.ac (--enable-mingw-wildcard): Add new configurable feature.
32917         * configure: Regenerate.
32918         * config.in: Regenerate.
32919         * config/i386/driver-mingw32.c: new file.
32920         * config/i386/x-mingw32: Add rule to build driver-mingw32.o.
32921         * config.host: Link driver-mingw32.o on MinGW host.
32922         * doc/install.texi: Document new --enable-mingw-wildcard configure
32923         option.
32925 2017-05-04  Marek Polacek  <polacek@redhat.com>
32927         PR tree-optimization/80612
32928         * calls.c (get_size_range): Check for INTEGRAL_TYPE_P.
32930 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32931             Andre Simoes Dias Vieira  <andre.simoesdiasvieira@arm.com>
32933         * config/arm/arm.md (movsi): Change TARGET_32BIT to TARGET_HAVE_MOVT.
32934         (movt splitter): Likewise.
32935         * config/arm/arm.c (arm_option_check_internal): Change arm_arch_thumb2
32936         to TARGET_HAVE_MOVT, and merge with -mslow-flash-data check.
32937         (const_ok_for_arm): Change else to else if (TARGET_THUMB2) and add else
32938         block for Thumb-1 with MOVT.
32939         (thumb2_legitimate_address_p): Move code block ...
32940         (can_avoid_literal_pool_for_label_p): ... into this new function.
32941         (thumb1_legitimate_address_p): Add check for TARGET_HAVE_MOVT and
32942         literal pool.
32943         (thumb_legitimate_constant_p): Add conditional on TARGET_HAVE_MOVT
32944         * doc/invoke.texi (-mpure-code): Change "ARMv7-M targets" for
32945         "M-profile targets with the MOVT instruction".
32947 2017-05-04  Prakhar Bahuguna  <prakhar.bahuguna@arm.com>
32949         * config/arm/arm-builtins.c (arm_init_builtins): Rename
32950         __builtin_arm_ldfscr to __builtin_arm_get_fpscr, and rename
32951         __builtin_arm_stfscr to __builtin_arm_set_fpscr.
32953 2017-05-04  Martin Liska  <mliska@suse.cz>
32955         * tree-vrp.c (simplify_cond_using_ranges_2): Remove unused
32956         variable cond_code.
32958 2017-05-04  Richard Biener  <rguenther@suse.de>
32960         * tree.c (array_at_struct_end_p): Handle arrays at struct
32961         end with flexarrays more conservatively.  Refactor and treat
32962         arrays of arrays or aggregates more strict.  Fix
32963         VIEW_CONVERT_EXPR handling.  Remove allow_compref argument.
32964         * tree.c (array_at_struct_end_p): Adjust prototype.
32965         * emit-rtl.c (set_mem_attributes_minus_bitpos): Adjust.
32966         * gimple-fold.c (get_range_strlen): Likewise.
32967         * tree-chkp.c (chkp_may_narrow_to_field): Likewise.
32969 2017-05-04  Richard Biener  <rguenther@suse.de>
32971         PR tree-optimization/31130
32972         * tree-vrp.c (needs_overflow_infinity): Remove as always returning
32973         false.
32974         (supports_overflow_infinity): Likewise.
32975         (is_negative_overflow_infinity): Likewise.
32976         (is_positive_overflow_infinity): Likewise.
32977         (is_overflow_infinity): Likewise.
32978         (stmt_overflow_infinity): Likewise.
32979         (overflow_infinity_range_p): Likewise.
32980         (usable_range_p): Remove as always returning true.
32981         (make_overflow_infinity): Remove.
32982         (negative_overflow_infinity): Likewise.
32983         (positive_overflow_infinity): Likewise.
32984         (avoid_overflow_infinity): Likewise.
32985         (set_value_range): Adjust accordingly.
32986         (set_value_range_to_nonnegative): Likewise, remove now unused
32987         overflow_infinity arg.
32988         (vrp_operand_equal_p): Adjust.
32989         (update_value_range): Likewise.
32990         (range_int_cst_singleton_p): Likewise.
32991         (operand_less_p): Likewise.
32992         (compare_values_warnv): Likewise.
32993         (extract_range_for_var_from_comparison_expr): Likewise.
32994         (vrp_int_const_binop): Likewise.
32995         (zero_nonzero_bits_from_vr): Likewise.
32996         (extract_range_from_multiplicative_op_1): Likewise.
32997         (extract_range_from_binary_expr_1): Likewise.
32998         (extract_range_from_unary_expr): Likewise.
32999         (extract_range_from_comparison): Likewise.
33000         (extract_range_basic): Likewise.
33001         (adjust_range_with_scev): Likewise.
33002         (compare_ranges): Likewise.
33003         (compare_range_with_value): Likewise.
33004         (dump_value_range): Likewise.
33005         (test_for_singularity): Likewise, remove strict_overflow_p parameter
33006         never used.
33007         (simplify_cond_using_ranges): Adjust.
33009 2017-05-04  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
33011         * brig-builtins.def: Added a builtin for class_f64.
33012         * builtin-types.def: Added a builtin type needed by class_f64.
33014 2017-05-03  Jason Merrill  <jason@redhat.com>
33016         * timevar.def: Add TV_CONSTEXPR.
33018 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33020         * common.opt (fdiagnostics-parseable-fixits): Fix typo.
33022 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33024         * ipa-prop.c (ipa_update_after_lto_read): Removed.
33025         * ipa-prop.h (ipa_update_after_lto_read): Remove declaration.
33026         * ipa-cp.c (ipcp_propagate_stage): Do not call
33027         ipa_update_after_lto_read.
33028         * ipa-inline.c (ipa_inline): Likewise.
33030 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33032         * ipa-prop.h (ipa_edge_args): Make a class.  Mark with for_user GTY
33033         tag.  Added a default constructor and a destructor.
33034         (ipa_edge_args_sum_t): New class;
33035         (ipa_edge_args_sum): Declare.
33036         (ipa_edge_args_vector): Remove declaration.
33037         (IPA_EDGE_REF): Use ipa_edge_args_sum.
33038         (ipa_free_edge_args_substructures): Remove declaration.
33039         (ipa_check_create_edge_args): Use ipa_edge_args_sum.
33040         (ipa_edge_args_info_available_for_edge_p): Likewise.
33041         * ipa-prop.c (ipa_edge_args_vector): Removed.
33042         (edge_removal_hook_holder): Likewise.
33043         (edge_duplication_hook_holder): Likewise.
33044         (ipa_edge_args_sum): New variable.
33045         (ipa_propagate_indirect_call_infos): Test ipa_edge_args_sum instead of
33046         ipa_edge_args_vector.
33047         (ipa_free_edge_args_substructures): Likewise.
33048         (ipa_free_all_edge_args): Free ipa_edge_args_sum instead of
33049         ipa_edge_args_vector.
33050         (ipa_edge_removal_hook): Turned into method
33051         ipa_edge_args_sum_t::remove.
33052         (ipa_edge_duplication_hook): Turned into method
33053         ipa_edge_args_sum_t::duplicate.
33054         (ipa_register_cgraph_hooks): Create ipa_edge_args_sum instead of
33055         registering edge hooks.
33056         (ipa_unregister_cgraph_hooks): Do not unregister edge hooks.
33057         * ipa-inline-analysis.c (estimate_function_body_sizes): Test
33058         ipa_edge_args_sum instead of ipa_edge_args_vector.
33059         * ipa-profile.c (ipa_profile): Likewise.
33061 2017-05-03  Martin Jambor  <mjambor@suse.cz>
33063         * symbol-summary.h (function_summary): New method exists.
33064         (function_summary::symtab_removal): Deallocate through release.
33065         (call_summary): New class.
33066         (gt_ggc_mx): New overload.
33067         (gt_pch_nx): Likewise.
33068         (gt_pch_nx): Likewise.
33070 2017-05-03  Jeff Law  <law@redhat.com>
33072         PR tree-optimization/78496
33073         * tree-vrp.c (simplify_cond_using_ranges_1): Renamed
33074         from simplify_cond_using_ranges.  Split off code to walk
33075         backwards through casts into ...
33076         (simplify_cond_using_ranges_2): New function.
33077         (simplify_stmt_using_ranges): Call simplify_cond_using_ranges_1.
33078         (execute_vrp): After identifying jump threads, call
33079         simplify_cond_using_ranges_2.
33081 2017-05-03  Jan Hubicka  <hubicka@ucw.cz>
33083         PR bootstrap/80609
33084         * ipa-inline.h (inline_summary): Add ctor.
33085         (create_ggc): Do not use ggc_cleared_alloc.
33087 2017-05-03  Jeff Downs  <heydowns@somuchpressure.net>
33088             Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
33090         * gcc.c (handle_braces): Support escaping in switch matching
33091         text.
33092         * doc/invoke.texi (Spec Files): Document it.
33093         Remove superfluous @code markup in items.
33095 2017-05-03  David Malcolm  <dmalcolm@redhat.com>
33097         * diagnostic-show-locus.c (struct column_range): New struct.
33098         (get_affected_columns): New function.
33099         (get_printed_columns): New function.
33100         (struct correction): New struct.
33101         (correction::ensure_capacity): New function.
33102         (correction::ensure_terminated): New function.
33103         (struct line_corrections): New struct.
33104         (line_corrections::~line_corrections): New dtor.
33105         (line_corrections::add_hint): New function.
33106         (layout::print_trailing_fixits): Reimplement in terms of the new
33107         classes.
33108         (selftest::test_overlapped_fixit_printing): New function.
33109         (selftest::diagnostic_show_locus_c_tests): Call it.
33111 2017-05-03  Nathan Sidwell  <nathan@acm.org>
33113         Canonicalize canonical type hashing
33114         * tree.h (type_hash_canon_hash): Declare.
33115         * tree.c (type_hash_list, attribute_hash_list): Move into
33116         type_hash_canon_hash.
33117         (build_type_attribute_qual_variant): Break out hash code calc into
33118         type_hash_canon_hash.
33119         (type_hash_canon_hash): New.  Generic type hash computation.
33120         (build_range_type_1, build_array_type_1, build_function_type,
33121         build_method_type_directly, build_offset_type, build_complex_type,
33122         make_vector_type): Call it.
33124 2017-05-03  Richard Biener  <rguenther@suse.de>
33126         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment):
33127         When all DRs have unknown misaligned do not always peel
33128         when there is a store but apply the same costing model as if
33129         there were only loads.
33131 2017-05-03  Richard Biener  <rguenther@suse.de>
33133         Revert
33134         PR tree-optimization/80492
33135         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33136         compare_base_decls returning dont-know properly.
33138 2017-05-03  Thomas Preud'homme  <thomas.preudhomme@arm.com>
33140         * config/arm/iterators.md (CCSI): New mode iterator.
33141         (arch): New mode attribute.
33142         * config/arm/sync.md (atomic_compare_and_swap<mode>_1): Rename into ...
33143         (atomic_compare_and_swap<CCSI:arch><NARROW:mode>_1): This and ...
33144         (atomic_compare_and_swap<CCSI:arch><SIDI:mode>_1): This.  Use CCSI
33145         code iterator for success result mode.
33146         * config/arm/arm.c (arm_expand_compare_and_swap): Adapt code to use
33147         the corresponding new insn generators.
33149 2017-05-03  Bin Cheng  <bin.cheng@arm.com>
33151         Revert r247509
33152         2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33153         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33155 2017-05-03  Richard Sandiford  <richard.sandiford@linaro.org>
33157         * tree-data-ref.h (SUB_CONFLICTS_IN_A): Wrap SUB argument in brackets.
33158         (SUB_CONFLICTS_IN_B, SUB_LAST_CONFLICT, SUB_DISTANCE): Likewise.
33159         (DDR_A): Wrap DDR argument in brackets.
33160         (DDR_B, DDR_AFFINE_P, DDR_ARE_DEPENDENT, DDR_SUBSCRIPTS): Likewise.
33161         (DDR_LOOP_NEST, DDR_INNER_LOOP, DDR_SELF_REFERENCE): Likewise.
33162         (DDR_REVERSED_P): Likewise.
33164 2017-05-03  Jakub Jelinek  <jakub@redhat.com>
33166         PR tree-optimization/79472
33167         * tree-switch-conversion.c (struct switch_conv_info): Add
33168         contiguous_range and default_case_nonstandard fields.
33169         (collect_switch_conv_info): Compute contiguous_range and
33170         default_case_nonstandard fields, don't clear final_bb if
33171         contiguous_range and only the default case doesn't have the required
33172         structure.
33173         (check_all_empty_except_final): Set default_case_nonstandard instead
33174         of failing if contiguous_range and the default case doesn't have empty
33175         block.
33176         (check_final_bb): Add SWTCH argument, don't fail if contiguous_range
33177         and only the default case doesn't have the required constants.  Skip
33178         virtual phis.
33179         (gather_default_values): Skip virtual phis.  Allow non-NULL CASE_LOW
33180         if default_case_nonstandard.
33181         (build_constructors): Build constant 1 just once.  Assert that default
33182         values aren't inserted in between cases if contiguous_range.  Skip
33183         virtual phis.
33184         (build_arrays): Skip virtual phis.
33185         (prune_bbs): Add DEFAULT_BB argument, don't remove that bb.
33186         (fix_phi_nodes): Don't add e2f phi arg if default_case_nonstandard.
33187         Handle virtual phis.
33188         (gen_inbound_check): Handle default_case_nonstandard case.
33189         (process_switch): Adjust check_final_bb caller.  Call
33190         gather_default_values with the first non-default case instead of
33191         default case if default_case_nonstandard.
33193 2017-05-02  Nathan Sidwell  <nathan@acm.org>
33195         * ggc-page.c (move_ptes_to_front): Replace unsigned >0 with i--
33196         check.  Fix formatting.
33198 2017-05-02  Jan Hubicka  <hubicka@ucw.cz>
33200         * ipa-inline-analysis.c (estimate_node_size_and_time): Allow roundoff
33201         errors when comparing specialized and unspecialized times.
33203 2017-05-02  David Malcolm  <dmalcolm@redhat.com>
33205         * diagnostic-show-locus.c
33206         (layout::should_print_annotation_line_p): Make private.
33207         (layout::print_annotation_line): Make private.
33208         (layout::annotation_line_showed_range_p): Make private.
33209         (layout::show_ruler): Make private.
33210         (layout::print_source_line): Make private.  Pass in line and
33211         line_width, rather than calling location_get_source_line.  Drop
33212         returned value.
33213         (layout::print_leading_fixits): New method.
33214         (layout::print_any_fixits): Rename to...
33215         (layout::print_trailing_fixits): ...this, and make private.
33216         Don't print newline fixits.
33217         (diagnostic_show_locus): Move logic for printing one row into...
33218         (layout::print_line): ...this new function.  Move the
33219         location_get_source_line call and error-handling from
33220         print_source_line to here.  Call print_leading_fixits, and rename
33221         print_any_fixits to print_trailing_fixits.
33222         (selftest::test_fixit_insert_containing_newline): Update now that
33223         newlines are partially supported.
33224         (selftest::test_fixit_insert_containing_newline_2): New test.
33225         (selftest::test_fixit_replace_containing_newline): Update comments.
33226         (selftest::diagnostic_show_locus_c_tests): Call the new test.
33227         * edit-context.c (class added_line): New class.
33228         (class edited_line): Describe newline handling in comment.
33229         (edited_line::actually_edited_p): New method.
33230         (edited_line::print_content): Delete redundant decl.
33231         (edited_line::m_predecessors): New field.
33232         (edited_file::print_content): Call edited_line::print_content.
33233         (edited_file::print_diff): Update to support newlines.
33234         (edited_file::print_diff_hunk): Likewise.
33235         (edited_file::print_run_of_changed_lines): New function.
33236         (edited_file::print_diff_line): Convert to...
33237         (print_diff_line): ...this.
33238         (edited_file::get_effective_line_count): New function.
33239         (edited_line::edited_line): Initialize new field m_predecessors.
33240         (edited_line::~edited_line): Clean up m_predecessors.
33241         (edited_line::apply_fixit): Handle newlines.
33242         (edited_line::get_effective_line_count): New function.
33243         (edited_line::print_content): New function.
33244         (edited_line::print_diff_lines): New function.
33245         (selftest::test_applying_fixits_insert_containing_newline): New
33246         test.
33247         (selftest::test_applying_fixits_replace_containing_newline): New
33248         test.
33249         (selftest::insert_line): New function.
33250         (selftest::test_applying_fixits_multiple_lines): Add example of
33251         inserting a line.
33252         (selftest::edit_context_c_tests): Call the new tests.
33254 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33256         * tree-ssa-loop-ivopts.c (get_scaled_computation_cost_at): Delete
33257         parameter cand.  Update dump information.
33258         (get_computation_cost): Update uses.
33260 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33262         * tree-ssa-loop-ivopts.c (get_computation_aff_1): New.
33263         (get_computation_aff): Reorder parameters.  Use get_computation_aff_1.
33264         (get_computation_at, rewrite_use_address): Update use of
33265         get_computation_aff.
33267 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33269         * tree-ssa-loop-ivopts.c (get_computation_at): Reorder parameters.
33270         (get_computation): Delete.
33271         (get_computation_cost): Implement like get_computation_cost_at.
33272         Use get_computation_at.
33273         (get_computation_cost_at): Delete.
33274         (rewrite_use_nonlinear_expr): Use get_computation_at.
33275         (rewrite_use_compare, remove_unused_ivs): Ditto.
33277 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33279         * tree-ssa-loop-ivopts.c (rewrite_use_address): Simple refactor.
33281 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33283         * tree-ssa-loop-ivopts.c (struct iv_ca): Rename n_regs to n_invs.
33284         (ivopts_global_cost_for_size): Rename parameter and update uses.
33285         (iv_ca_recount_cost): Update uses.
33286         (iv_ca_set_remove_invs, iv_ca_set_no_cp): Record invariants and
33287         candidates seperately in n_invs and n_cands.
33288         (iv_ca_set_add_invs, iv_ca_set_cp, iv_ca_new): Ditto.
33290 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33292         * tree-ssa-loop-ivopts.c (struct walk_tree_data): New.
33293         (find_inv_vars_cb): New.
33294         (find_depends): Renamed to ...
33295         (find_inv_vars): ... this.
33296         (add_candidate_1, force_var_cost): Call find_inv_vars.
33297         (split_address_cost, determine_group_iv_cost_cond): Ditto.
33299 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33301         * tree-ssa-loop-ivopts.c (struct cost_pair): Rename depends_on to
33302         inv_vars.  Add inv_exprs.
33303         (struct iv_cand): Rename depends_on to inv_vars.
33304         (struct ivopts_data): Rename max_inv_id/n_invariant_uses to
33305         max_inv_var_id/n_inv_var_uses.  Move max_inv_expr_id around.
33306         Refactor field used_inv_exprs from has_map to array n_inv_expr_uses.
33307         (dump_cand): Dump inv_vars.
33308         (tree_ssa_iv_optimize_init): Support inv_vars and inv_exprs.
33309         (record_invariant, find_depends, add_candidate_1): Ditto.
33310         (set_group_iv_cost, force_var_cost): Ditto.
33311         (split_address_cost, ptr_difference_cost, difference_cost): Ditto.
33312         (get_computation_cost_at, get_computation_cost): Ditto.
33313         (determine_group_iv_cost_generic): Ditto.
33314         (determine_group_iv_cost_address): Ditto.
33315         (determine_group_iv_cost_cond, autoinc_possible_for_pair): Ditto.
33316         (determine_group_iv_costs): Ditto.
33317         (iv_ca_recount_cost): Update call to ivopts_global_cost_for_size.
33318         (iv_ca_set_remove_invariants): Renamed to ...
33319         (iv_ca_set_remove_invs): ... this.  Support inv_vars and inv_exprs.
33320         (iv_ca_set_no_cp): Use iv_ca_set_remove_invs.
33321         (iv_ca_set_add_invariants):  Renamed to ...
33322         (iv_ca_set_add_invs): ... this.  Support inv_vars and inv_exprs.
33323         (iv_ca_set_cp): Use iv_ca_set_add_invs.
33324         (iv_ca_has_deps): Support inv_vars and inv_exprs.
33325         (iv_ca_new, iv_ca_free, iv_ca_dump, free_loop_data): Ditto.
33326         (create_new_ivs): Remove useless dump.
33328 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33330         * tree-ssa-loop-ivopts.c (get_computation_cost_at): Remove pseudo
33331         iv_cand code.
33332         (determine_group_iv_cost_cond, determine_iv_cost): Ditto.
33333         (iv_ca_set_no_cp, create_new_iv): Ditto.
33335 2017-05-02  Bin Cheng  <bin.cheng@arm.com>
33337         * rtlanal.c (rtx_cost): Handle TRUNCATE between tieable modes.
33339 2017-05-02  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33341         * tree.h (EXPR_CILK_SPAWN): Use macro TREE_CHECK2 instead of
33342         function tree_check2.
33344 2017-05-02  Martin Liska  <mliska@suse.cz>
33346         * doc/gcov.texi: Add missing preposition.
33347         * gcov.c (function_info::function_info): Properly fill up
33348         all member variables.
33350 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33352         * expr.c (expand_expr_real_2): Re-cost if previous costs are the same.
33354 2017-05-02  Tamar Christina  <tamar.christina@arm.com>
33356         * simplify-rtx.c (simplify_binary_operation_1): Add LSHIFTRT case.
33358 2017-05-02  Martin Liska  <mliska@suse.cz>
33360         PR lto/77954.
33361         * lto-streamer-in.c (lto_read_tree_1): Remove
33362         LTO_STREAMER_DEBUG.
33363         * lto-streamer.c (struct tree_hash_entry): Likewise.
33364         (struct tree_entry_hasher): Likewise.
33365         (tree_entry_hasher::hash): Likewise.
33366         (tree_entry_hasher::equal): Likewise.
33367         (lto_streamer_init): Likewise.
33368         (lto_orig_address_map): Likewise.
33369         (lto_orig_address_get): Likewise.
33370         (lto_orig_address_remove): Likewise.
33371         * lto-streamer.h: Likewise.
33372         * tree-streamer-in.c (streamer_alloc_tree): Likewise.
33373         * tree-streamer-out.c (streamer_write_tree_header): Likewise.
33375 2017-05-02  Sebastian Peryt  <sebastian.peryt@intel.com>
33377         * config/i386/avx512fintrin.h (_mm_mask_add_round_sd)
33378         (_mm_maskz_add_round_sd, _mm_mask_add_round_ss)
33379         (mm_maskz_add_round_ss, _mm_mask_sub_round_sd)
33380         (mm_maskz_sub_round_sd, _mm_mask_sub_round_ss)
33381         (mm_maskz_sub_round_ss, _mm_mask_add_sd)
33382         (mm_maskz_add_sd, _mm_mask_add_ss, _mm_maskz_add_ss)
33383         (mm_mask_sub_sd, _mm_maskz_sub_sd, _mm_mask_sub_ss)
33384         (mm_maskz_sub_ss): New intrinsics.
33385         * config/i386/i386-builtin-types.def (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33386         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): New function type aliases.
33387         * config/i386/i386-builtin.def (__builtin_ia32_addsd_mask_round)
33388         (__builtin_ia32_addss_mask_round, __builtin_ia32_subsd_mask_round)
33389         (__builtin_ia32_subss_mask_round): New builtins.
33390         * config/i386/i386.c (V2DF_FTYPE_V2DF_V2DF_V2DF_UQI_INT)
33391         (V4SF_FTYPE_V4SF_V4SF_V4SF_UQI_INT): Handle new types.
33392         * config/i386/sse.md (<sse>_vm<plusminus_insn><mode>3<round_name>):
33393         Renamed to ...
33394         (<sse>_vm<plusminus_insn><mode>3<mask_name><round_name>): ... this.
33395         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_op3>%2, %1, %0|%0, %1, %<iptr>2<round_op3>}):
33396         Changed to ...
33397         (v<plusminus_mnemonic><ssescalarmodesuffix>\t{<round_mask_op3>%2, %1, %0<mask_operand3>|%0<mask_operand3>, %1, %<iptr>2<round_mask_op3>}):
33398         ... this.
33400 2017-05-02  Martin Jambor  <mjambor@suse.cz>
33402         PR tree-optimization/78687
33403         * tree-sra.c (access): New field parent.
33404         (process_subtree_disqualification): New function.
33405         (disqualify_candidate): Call it.
33406         (build_accesses_from_assign): Reset write flag if creating an
33407         assighnment link.
33408         (build_access_subtree): Fill in parent field and also prpagate
33409         down grp_write flag.
33410         (create_artificial_child_access): New parameter set_grp_write, set
33411         grp_write to its value.
33412         (propagate_subaccesses_across_link): Also propagate grp_write flag
33413         values.
33414         (propagate_all_subaccesses): Push the closest parent back to work
33415         queue if add_access_to_work_queue returned true.
33417 2017-05-02  Richard Biener  <rguenther@suse.de>
33419         * common.opt (fstrict-overflow): Alias negative to fwrapv.
33420         * doc/invoke.texi (fstrict-overflow): Remove all traces of
33421         -fstrict-overflow documentation.
33422         * tree.h (TYPE_OVERFLOW_UNDEFINED): Do not test flag_strict_overflow.
33423         (POINTER_TYPE_OVERFLOW_UNDEFINED): Test !flag_wrapv instead of
33424         flag_strict_overflow.
33425         * ipa-inline.c (can_inline_edge_p): Do not test flag_strict_overflow.
33426         * lto-opts.c (lto_write_options): Do not stream it.
33427         * lto-wrapper.c (merge_and_complain): Do not handle it.
33428         * opts.c (default_options_table): Do not set -fstrict-overflow.
33429         (finish_options): Likewise do not clear it when sanitizing.
33430         * simplify-rtx.c (simplify_const_relational_operation): Do not
33431         test flag_strict_overflow.
33433 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33435         * config/alpha/alpha.md (*add<mode>3_ieee): Merge to add<mode>3
33436         using enabled attribute.
33437         (*sub<mode>3_ieee): Merge to sub<mode>3 using enabled attribute.
33438         (*mul<mode>3_ieee): Merge to mul<mode>3 using enabled attribute.
33439         (*div<mode>3_ieee): Merge to div<mode>3 using enabled attribute.
33440         (*sqrt<mode>2_ieee): Merge to sqrt<mode>2 using enabled attribute.
33441         (*fix_truncdfdi_ieee): Merge to *fix_truncdfdi2 using enabled attribute.
33442         (*fix_truncsfdi_ieee): Merge to *fix_truncsfdi2 using enabled attribute.
33443         (*floatdisf_ieee): Merge to floatdisf2 using enabled attribute.
33444         (*floatdidf_ieee): Merge to floatdidf2 using enabled attribute.
33445         (*truncdfsf2_ieee): Merge to truncdfsf2 using enabled attribute.
33446         (*cmpdf_ieee): Merge to *cmpdf_internal using enabled attribute.
33448 2017-05-02  Uros Bizjak  <ubizjak@gmail.com>
33450         * config/i386/i386.c (ix86_code_end): Use {FIRST,LAST}_INT_REG.
33452 2017-05-02  Richard Biener  <rguenther@suse.de>
33454         PR tree-optimization/80591
33455         Revert
33456         2017-04-10  Richard Biener  <rguenther@suse.de>
33458         * tree-ssa-structalias.c (find_func_aliases): Properly handle
33459         asm inputs.
33461 2017-05-02  Richard Biener  <rguenther@suse.de>
33463         PR tree-optimization/80549
33464         * tree-cfgcleanup.c (mfb_keep_latches): New helper.
33465         (cleanup_tree_cfg_noloop): Create forwarders to known loop
33466         headers if they do not have a preheader.
33468 2017-05-02  Martin Liska  <mliska@suse.cz>
33470         PR other/80589
33471         * common.opt: Fix typo.
33472         * doc/invoke.texi: Likewise.
33474 2017-05-01  Jan Beulich  <jbeulich@suse.com>
33476         * config/i386/sse.md (xop_vpermil2<mode>3): Do not allow operand
33477         swapping, add (x,x,m,x,n) alternative.
33479 2017-05-01  Nathan Sidwell  <nathan@acm.org>
33481         * calls.c (combine_pending_stack_adjustment_and_call): Remove
33482         unnecessary unadjusted_alignment check.
33484 2017-05-01  Xi Ruoyao  <ryxi@stu.xidian.edu.cn>
33486         PR c++/80038
33487         * cilk_common.c (expand_builtin_cilk_detach): Move pedigree
33488         operations here.
33489         * gimplify.c (gimplify_cilk_detach): New function.
33490         (gimplify_call_expr, gimplify_modify_expr): Call it as needed.
33491         * tree-core.h: Document EXPR_CILK_SPAWN.
33492         * tree.h (EXPR_CILK_SPAWN): Define.
33494 2017-05-01  David Malcolm  <dmalcolm@redhat.com>
33496         * diagnostic-show-locus.c (layout::get_expanded_location): Rewrite
33497         to use new fixit_hint representation, using the "replace" logic.
33498         (get_line_span_for_fixit_hint): Likewise.
33499         (layout::print_any_fixits): Likewise.
33500         (selftest::test_one_liner_many_fixits): Rename to...
33501         (selftest::test_one_liner_many_fixits_1): ...this, and update
33502         comment and expected output to reflect that the multiple fix-it
33503         hints are now consolidated into one insertion.
33504         (selftest::test_one_liner_many_fixits_2): New test.
33505         (selftest::test_diagnostic_show_locus_one_liner): Update for
33506         above.
33507         (selftest::test_fixit_consolidation): Update for fix-it API
33508         change.
33509         * diagnostic.c (print_parseable_fixits): Likewise.
33510         * edit-context.c (edited_line::m_line_events): Convert from
33511         auto_vec <line_event *> to auto_vec <line_event>.
33512         (class line_event): Convert from abstract base class to a concrete
33513         class, taking over the role of replace_event.
33514         (class insert_event): Delete.
33515         (class replace_event): Rename to class line_event.  Convert to
33516         half-open range.
33517         (edit_context::add_fixits): Reimplement.
33518         (edit_context::apply_insert): Delete.
33519         (edit_context::apply_replace): Rename to...
33520         (edit_context::apply_fixit): ...this.  Convert to half-open range.
33521         (edited_file::apply_insert): Delete.
33522         (edited_file::apply_replace): Rename to...
33523         (edited_file::apply_fixit): ...this.
33524         (edited_line::~edited_line): Drop deletion of events.
33525         (edited_line::apply_insert): Delete.
33526         (edited_line::apply_replace): Rename to...
33527         (edited_line::apply_fixit): ...this.  Convert to half-open range.
33528         Update for change to type of m_line_events.
33529         * edit-context.h (edit_context::apply_insert): Delete.
33530         (edit_context::apply_replace): Rename to...
33531         (edit_context::apply_fixit): ...this.
33533 2017-05-01  Martin Sebor  <msebor@redhat.com>
33535         * gimple-ssa-sprintf.c (format_integer): Set knownrange when it's
33536         known.
33538 2017-05-01  Uros Bizjak  <ubizjak@gmail.com>
33540         PR target/68491
33541         * config/i386/cpuid.h (__get_cpuid): Always return 0 when
33542         __get_cpuid_max returns 0.
33543         (__get_cpuid_count): Ditto.
33545 2017-05-01  Eric Botcazou  <ebotcazou@adacore.com>
33547         * tree.c (substitute_in_expr) <tcc_vl_exp>: Also inline a call if the
33548         replacement expression is another instance of one of its arguments.
33550 2017-05-01  Jakub Jelinek  <jakub@redhat.com>
33552         PR target/79430
33553         * rtlanal.c (reg_set_p): If reg is a stack_pointer_rtx, also
33554         check for stack push/pop autoinc.
33555         * config/i386/i386.c (ix86_agi_dependent): Return false
33556         if the only reason why modified_in_p returned true is that
33557         addr is SP based and set_insn is a push or pop.
33559 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33561         * ipa-inline.c (compute_inlined_call_time): Remove now unnecesary
33562         overflow check.
33564 2017-04-29  Jan Hubicka  <hubicka@ucw.cz>
33566         PR ipa/79224
33567         * ipa-inline-analysis.c (dump_predicate): Add optional parameter NL.
33568         (account_size_time): Use two predicates - exec_pred and
33569         nonconst_pred_ptr.
33570         (evaluate_conditions_for_known_args): Compute both clause and
33571         nonspec_clause.
33572         (evaluate_properties_for_edge): Evaulate both clause and nonspec_clause.
33573         (inline_summary_t::duplicate): Update.
33574         (estimate_function_body_sizes): Caluculate exec and nonconst predicates
33575         separately.
33576         (compute_inline_parameters): Likewise.
33577         (estimate_edge_size_and_time): Update caluclation of time.
33578         (estimate_node_size_and_time): Compute both time and nonspecialized
33579         time.
33580         (estimate_ipcp_clone_size_and_time): Update.
33581         (inline_merge_summary): Update.
33582         (do_estimate_edge_time): Update.
33583         (do_estimate_edge_size): Update.
33584         (do_estimate_edge_hints): Update.
33585         (inline_read_section, inline_write_summary): Stream both new predicates.
33586         * ipa-inline.c (compute_uninlined_call_time): Take uninlined_call_time
33587         as argument.
33588         (compute_inlined_call_time): Cleanup.
33589         (big_speedup_p): Update.
33590         (edge_badness): Update.
33591         * ipa-inline.h (INLINE_TIME_SCALE): Remove.
33592         (size_time_entry): Replace predicate by exec_predicate and
33593         nonconst_predicate.
33594         (edge_growth_cache_entry): Cache both time nad nonspecialized time.
33595         (estimate_edge_time): Return also nonspec_time.
33596         (reset_edge_growth_cache): Update.
33598 2017-04-29  Jakub Jelinek  <jakub@redhat.com>
33600         PR rtl-optimization/80491
33601         * ifcvt.c (noce_process_if_block): When looking for x setter
33602         with missing else_bb, don't check only the insn right before
33603         cond_earliest, but look for the last insn that x is modified in
33604         within the same bb.
33606         PR rtl-optimization/80491
33607         * alias.c (memory_modified_in_insn_p): Return true for CALL_INSNs.
33609 2017-04-29  Marc Glisse  <marc.glisse@inria.fr>
33611         PR tree-optimization/80487
33612         * tree-ssa-alias.c (stmt_kills_ref_p): Handle stpncpy and strncpy.
33614 2017-04-29  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
33616         PR tree-optimization/79697
33617         * tree-ssa-dce.c (mark_stmt_if_obviously_necessary): Check if callee
33618         is BUILT_IN_STRDUP, BUILT_IN_STRNDUP, BUILT_IN_REALLOC.
33619         (propagate_necessity): Check if def_callee is BUILT_IN_STRDUP or
33620         BUILT_IN_STRNDUP.
33621         * gimple-fold.c (gimple_fold_builtin_realloc): New function.
33622         (gimple_fold_builtin): Call gimple_fold_builtin_realloc.
33624 2017-04-28  Martin Sebor  <msebor@redhat.com>
33626         PR tree-optimization/80523
33627         * gimple-ssa-sprintf.c (target_to_host_charmap): New global variable.
33628         (init_target_to_host_charmap, target_to_host, target_strtol10): New
33629         functions.
33630         (maybe_warn, format_directive, parse_directive): Use new functions.
33631         (pass_sprintf_length::execute): Call init_target_to_host_charmap.
33633 2017-04-28  Marc Glisse  <marc.glisse@inria.fr>
33635         * match.pd (X+Z OP Y+Z, X-Z OP Y-Z, Z-X OP Z-Y): New transformations.
33637 2017-04-28  Bernd Edlinger  <bernd.edlinger@hotmail.de>
33639         * configure.ac (SYSTEM_HEADER_DIR, BUILD_SYSTEM_HEADER_DIR,
33640         target_header_dir): Set correctly.
33641         * configure: Regenerated.
33642         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
33643         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
33644         instead of SYSTEM_HEADER_DIR.
33646 2017-04-28  Jan Hubicka  <hubicka@ucw.cz>
33648         * ipa-cp.c (perform_estimation_of_a_value): Turn time to sreal.
33649         (estimate_local_effects): Likewise.
33650         * ipa-inline.c (compute_inlined_call_time, want_inline_small_function_p,
33651         edge_badness, inline_small_functions, dump_overall_stats): LIkewise.
33652         * ipa-inline.h (edge_growth_cache_entry, estimate_time_after_inlining,
33653         estimate_ipcp_clone_size_and_time, do_estimate_edge_time,
33654         do_estimate_edge_time, estimate_edge_time): Likewise.
33655         * ipa-inline-analysis.c (estimate_node_size_and_time,
33656         estimate_ipcp_clone_size_and_time, do_estimate_edge_time): Likewise.
33657         (estimate_time_after_inlining): Remove.
33659 2017-04-28  Martin Liska  <mliska@suse.cz>
33661         * doc/gcov.texi: Enhance documentation of gcov.
33663 2017-04-28  Martin Liska  <mliska@suse.cz>
33665         * doc/gcov.texi: Sort options in alphabetic order.
33666         * doc/gcov-dump.texi: Likewise.
33667         * doc/gcov-tool.texi: Likewise.
33668         * gcov.c (print_usage): Likewise.
33669         * gcov-dump.c (print_usage): Likewise.
33670         * gcov-tool.c (print_merge_usage_message): Likewise.
33671         (print_rewrite_usage_message): Likewise.
33672         (print_overlap_usage_message): Likewise.
33674 2017-04-28  Martin Liska  <mliska@suse.cz>
33676         PR gcov-profile/53915
33677         * gcov.c (format_gcov): Print 'NAN %' when top > bottom.
33679 2017-04-28  Martin Liska  <mliska@suse.cz>
33681         PR gcov-profile/79891
33682         * gcov.c (add_line_counts): Assign BBs to lines just if the BB
33683         is marked by compiler as living on a line.
33684         (get_cycles_count): Remove usage of the union.
33685         (output_intermediate_file): Likewise.
33686         (find_source): Fix GNU coding style.
33687         (accumulate_line_counts): Remove old non-all block mode.
33688         (output_lines): Remove usage of the union.
33689         * profile.c (output_location): Include all BBs, even if
33690         belonging to a same line (and file) as a previous BB.
33692 2017-04-28  Martin Liska  <mliska@suse.cz>
33694         * gcov.c (process_args): Handle new argument 'w'.
33695         (read_graph_file): Assign ID to BBs.
33696         (output_branch_count): Display BB # if verbose flag is set.
33697         (output_lines): Likewise for arcs.
33698         (print_usage): Add '--verbose' option help.
33699         * doc/gcov.texi: Document --verbose (-w) option.
33701 2017-04-28  Martin Liska  <mliska@suse.cz>
33703         * gcov.c (struct block_location_info): New struct.
33704         (process_file): Fill up the new structure.
33705         (read_graph_file): Replace usage of encoding by the newly added
33706         struct.
33707         (add_line_counts): Likewise.
33708         (accumulate_line_counts): Remove usage of the union.
33709         (function_info::function_info): New function.
33710         (function_info::~function_info): Likewise.
33711         (process_file): Call delete instead of release_function.
33712         (release_function): Release the function.
33713         (release_structures): Call delete instead of release_function.
33714         (solve_flow_graph): Replace usage of num_blocks.
33715         (find_exception_blocks): Likewise.
33716         (output_lines): Fix GNU coding style.
33718 2017-04-28  Martin Liska  <mliska@suse.cz>
33720         PR driver/56469
33721         * coverage.c (coverage_remove_note_file): New function.
33722         * coverage.h: Declare the function.
33723         * toplev.c (finalize): Clean if an error has been seen.
33725 2017-04-28  Martin Liska  <mliska@suse.cz>
33727         PR gcov-profile/80031
33728         * gcov-dump.c (tag_blocks): Just print number of basic blocks.
33729         * gcov-io.h (GCOV_TAG_BLOCKS_NUM): Remove unused macro.
33730         * gcov.c (read_graph_file): Read just number of blocks.
33731         * profile.c (branch_prob): Do not stream 0 flags per a basic
33732         block.
33734 2017-04-28  Martin Liska  <mliska@suse.cz>
33736         * gcov-dump.c (tag_*): Add new argument to declarations.
33737         (dump_gcov_file): Likewise.
33738         (tag_blocks): Add and use new argument depth.
33739         (tag_arcs): Likewise.
33740         (tag_lines): Likewise.
33741         (tag_counters): Likewise.
33742         (tag_summary): Likewise.
33743         (dump_working_sets): Use depth to do a proper indentation.
33745 2017-04-28  Jakub Jelinek  <jakub@redhat.com>
33747         PR bootstrap/80531
33748         * cgraph.h (symtab_node::debug_symtab): No longer inline.
33749         * symtab.c (symtab_node::debug_symtab): Move definition here.
33751 2017-04-28  Richard Biener  <rguenther@suse.de>
33753         * lto-streamer.h (LTO_major_version): Bump to 7.
33755 2017-04-28  Richard Biener  <rguenther@suse.de>
33757         * tree-vrp.c (assert_info): New struct.
33758         (add_assert_info): New helper.
33759         (register_edge_assert_for_2): Refactor to add asserts to a vector
33760         of assert_info.
33761         (register_edge_assert_for_1): Likewise.
33762         (register_edge_assert_for): Likewise.
33763         (finish_register_edge_assert_for): New helper actually registering
33764         asserts where live on edge.
33765         (find_conditional_asserts): Adjust.
33766         (find_switch_asserts): Likewise.
33767         (evrp_dom_walker::try_find_new_range): Generalize.
33768         (evrp_dom_walker::before_dom_children): Use register_edge_assert_for.
33770 2017-04-27  Marek Polacek  <polacek@redhat.com>
33772         PR sanitizer/80349
33773         * fold-const.c (fold_binary_loc) <case EQ_EXPR, NE_EXPR>: Convert
33774         arg10 and arg11 to itype.
33776 2017-04-27  Jonathan Wakely  <jwakely@redhat.com>
33778         * doc/extend.texi (Object Size Checking): Improve grammar.
33780 2017-04-27  Richard Earnshaw  <rearnsha@arm.com>
33782         PR target/80530
33783         * config/aarch64/aarch64.c (aarch64_emit_approx_sqrt): Ensure
33784         that the logic for permitting reciprocal estimates matches that
33785         in use_rsqrt_p.
33787 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33789         PR c++/80534
33790         * tree.c (type_cache_hasher::equal): Only compare
33791         TYPE_TYPELESS_STORAGE flag on non-aggregate element types.
33792         (build_array_type_1): Only hash TYPE_TYPELESS_STORAGE flag on
33793         non-aggregate element types.
33794         * tree.h (TYPE_TYPELESS_STORAGE): Fix comment typo, add more details
33795         about the flag on ARRAY_TYPEs in the comment, formatting fix.
33797 2017-04-27  Richard Biener  <rguenther@suse.de>
33799         PR middle-end/80533
33800         * emit-rtl.c (set_mem_attributes_minus_bitpos): When
33801         stripping ARRAY_REFs from MEM_EXPR make sure we're not
33802         keeping a reference to a trailing array.
33804 2017-04-27  Richard Biener  <rguenther@suse.de>
33806         PR middle-end/80539
33807         * tree-chrec.c (chrec_fold_plus_poly_poly): Deal with not
33808         being in loop-closed SSA form conservatively.
33809         (chrec_fold_multiply_poly_poly): Likewise.
33811 2017-04-27  Tamar Christina  <tamar.christina@arm.com>
33813         PR middle-end/79665
33814         * expr.c (expand_expr_real_2): Move TRUNC_MOD_EXPR, FLOOR_MOD_EXPR,
33815         CEIL_MOD_EXPR, ROUND_MOD_EXPR cases.
33817 2017-04-27  Jakub Jelinek  <jakub@redhat.com>
33819         PR target/77728
33820         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): Remove.
33821         (aarch64_function_arg_alignment): Return unsigned int again, but still
33822         ignore TYPE_FIELDS chain decls other than FIELD_DECLs.
33823         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33824         Don't emit -Wpsabi note.
33825         (aarch64_function_arg_boundary): Likewise.
33826         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33827         caller.
33829 2017-04-26  Nathan Sidwell  <nathan@acm.org>
33831         * tree.h (crc32_unsigned_n): Declare.
33832         (crc32_unsigned, crc32_unsigned): Make inline.
33833         * tree.c (crc32_unsigned_bits): Replace with ...
33834         (crc32_unsigned_n): ... this.
33835         (crc32_unsigned, crc32_byte): Remove.
33836         (crc32_string): Remove unnecessary braces.
33838 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33840         * ipa-cp.c (estimate_local_effects): Convert sreal to int.
33841         * ipa-inline-analysis.c (MAX_TIME): Remove.
33842         (account_size_time): Use sreal for time.
33843         (dump_inline_summary): Update.
33844         (estimate_function_body_sizes): Update.
33845         (estimate_edge_size_and_time): Update.
33846         (estimate_calls_size_and_time): Update.
33847         (estimate_node_size_and_time): Update.
33848         (inline_merge_summary): Update.
33849         (inline_update_overall_summary): Update.
33850         (estimate_time_after_inlining): Update.
33851         (inline_read_section): Update.
33852         (inline_write_summary): Update.
33853         * ipa-inline.c (compute_uninlined_call_time): Update.
33854         (compute_inlined_call_time): Update.
33855         (recursive_inlining): Update.
33856         (inline_small_functions): Update.
33857         (dump_overall_stats): Update.
33858         * ipa-inline.h: Include sreal.h.
33859         (size_time_entry): Turn time to sreal.
33860         (inline_summary): Turn self_time nad time to sreal.
33862 2017-04-25  Jan Hubicka  <hubicka@ucw.cz>
33864         * sreal.c: Include backend.h, tree.h, gimple.h, cgraph.h and
33865         data-streamer.h
33866         (sreal::stream_out, sreal::stream_in): New.
33867         * sreal.h (sreal::stream_out, sreal::stream_in): Declare.
33869 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
33871         * Makefile.in (s-options): Invoke opt-gather.awk with LC_ALL=C in the
33872         environment.
33874 2017-04-25  Uros Bizjak  <ubizjak@gmail.com>
33876         PR target/70799
33877         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p):
33878         Handle ASHIFTRT.
33879         (dimode_scalar_chain::compute_convert_gain): Ditto.
33880         (dimode_scalar_chain::make_vector_copies): Ditto.
33881         (dimode_scalar_chain::convert_reg): Ditto.
33882         (dimode_scalar_chain::convert_insn): Ditto.
33883         * config/i386/sse.md (VI24_AVX512BW_1): Remove mode iterator.
33884         (VI248_AVX512BW_1): New mode iterator.
33885         (<mask_codefor>ashr<mode>3<mask_name>): Merge insn pattern with
33886         <mask_codefor>ashrv2di3<mask_name> insn using VI248_AVX512BW_1
33887         mode iterator.
33889 2017-04-25  Martin Sebor  <msebor@redhat.com>
33891         PR tree-optimization/80497
33892         * gimple-ssa-sprintf.c (get_int_range): Avoid assuming all integer
33893         constants are representable in HOST_WIDE_INT.
33894         (parse_directive): Ditto.
33896 2017-04-25  Martin Sebor  <msebor@redhat.com>
33898         PR bootstrap/80486
33899         * dominance.c (dom_info::m_n_basic_blocks): Change type to unsigned.
33900         (new_zero_array): Adjust signature.
33901         (dom_info::dom_init): Used unsigned rather that size_t.
33902         (dom_info::dom_info): Same.
33904 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33905             Jakub Jelinek  <jakub@redhat.com>
33907         PR target/77728
33908         * config/arm/arm.c: Include gimple.h.
33909         (aapcs_layout_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33910         returns negative, increment ncrn only if it returned positive.
33911         (arm_needs_doubleword_align): Return int instead of bool,
33912         ignore DECL_ALIGN of non-FIELD_DECL TYPE_FIELDS chain
33913         members, but if there is any such non-FIELD_DECL
33914         > PARM_BOUNDARY aligned decl, return -1 instead of false.
33915         (arm_function_arg): Emit -Wpsabi note if arm_needs_doubleword_align
33916         returns negative, increment nregs only if it returned positive.
33917         (arm_setup_incoming_varargs): Likewise.
33918         (arm_function_arg_boundary): Emit -Wpsabi note if
33919         arm_needs_doubleword_align returns negative, return
33920         DOUBLEWORD_ALIGNMENT only if it returned positive.
33922 2017-04-25  Marek Polacek  <polacek@redhat.com>
33924         PR sanitizer/80349
33925         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
33926         first argument to type.
33928 2017-04-25  Bill Seurer  <seurer@linux.vnet.ibm.com>
33930         PR target/80482
33931         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Change
33932         type checks to test for compatibility instead of equality.
33934 2017-04-25  Ramana Radhakrishnan  <ramana.radhakrishnan@arm.com>
33935             Jakub Jelinek  <jakub@redhat.com>
33937         PR target/77728
33938         * config/aarch64/aarch64.c (struct aarch64_fn_arg_alignment): New
33939         type.
33940         (aarch64_function_arg_alignment): Return aarch64_fn_arg_alignment
33941         struct.  Ignore DECL_ALIGN of decls other than FIELD_DECL for
33942         the alignment computation, but return their maximum in warn_alignment.
33943         (aarch64_layout_arg): Adjust aarch64_function_arg_alignment caller.
33944         Emit a -Wpsabi note if warn_alignment is 16 bytes, but alignment
33945         is smaller.
33946         (aarch64_function_arg_boundary): Likewise.  Simplify using MIN/MAX.
33947         (aarch64_gimplify_va_arg_expr): Adjust aarch64_function_arg_alignment
33948         caller.
33950 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33952         * config/arc/simdext.md (dmpyh): Fix typo.
33954 2017-04-25  Richard Biener  <rguenther@suse.de>
33956         PR tree-optimization/80492
33957         * alias.c (compare_base_decls): Handle registers with asm
33958         specification conservatively.
33959         * tree-ssa-alias.c (decl_refs_may_alias_p): Handle
33960         compare_base_decls returning dont-know properly.
33962 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33964         * config/arc/arc.c (LEGITIMATE_OFFSET_ADDRESS_P): Delete macro.
33965         (legitimate_offset_address_p): New function.
33966         (arc_legitimate_address_p): Use above function.
33968 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33970         * config/arc/arc.c (arc_output_mi_thunk): Emit PIC calls.
33972 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33974         * config/arc/arc.c (arc_conditional_register_usage): Use ACCL,
33975         ACCH registers whenever they are available.
33977 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33979         * config/arc/arc.c (arc_conditional_register_usage): Make D0, D1
33980         double regs fix when not used.
33982 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33984         * config/arc/arc.h (REGNO_OK_FOR_BASE_P): Consider also extension
33985         core registers.
33986         (REG_OK_FOR_INDEX_P_NONSTRICT): Likewise.
33987         (REG_OK_FOR_BASE_P_NONSTRICT): Likewise.
33989 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33991         * config/arc/arc.c (arc_output_addsi): Check for h-register class
33992         when emitting short ADD instructions.
33994 2017-04-25  Claudiu Zissulescu  <claziss@synopsys.com>
33996         * config/arc/arc.md (cmpsi_cc_insn_mixed): Use 'h' register
33997         constraint.
33998         (cmpsi_cc_c_insn): Likewise.
33999         (cbranchsi4_scratch): Compute proper instruction length using
34000         compact_hreg_operand.
34001         * config/arc/predicates.md (compact_hreg_operand): New predicate.
34003 2017-04-25  Richard Biener  <rguenther@suse.de>
34005         PR middle-end/80509
34006         * passes.c (pass_manager::pass_manager): Initialize
34007         m_name_to_pass_map.
34009 2017-04-25  Richard Biener  <rguenther@suse.de>
34011         PR tree-optimization/79201
34012         * tree-ssa-sink.c (statement_sink_location): Handle calls.
34014 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34016         PR target/80464
34017         * config/s390/vector.md: Split MEM->GPR vector moves for
34018         non-s_operand addresses.
34020 2017-04-25  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
34022         PR target/79895
34023         * config/s390/predicates.md (reload_const_wide_int_operand): New
34024         predicate.
34025         * config/s390/s390.md ("movti"): Remove d/P alternative.
34026         ("movti_bigconst"): New pattern definition.
34028 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34030         PR target/80080
34031         * s390-protos.h (s390_expand_cs_hqi): Removed.
34032         (s390_expand_cs, s390_expand_atomic_exchange_tdsi): New prototypes.
34033         * config/s390/s390.c (s390_emit_compare_and_swap): Handle all integer
34034         modes as well as CCZ1mode and CCZmode.
34035         (s390_expand_atomic_exchange_tdsi, s390_expand_atomic): Adapt to new
34036         signature of s390_emit_compare_and_swap.
34037         (s390_expand_cs_hqi): Likewise, make static.
34038         (s390_expand_cs_tdsi): Generate an explicit compare before trying
34039         compare-and-swap, in some cases.
34040         (s390_expand_cs): Wrapper function.
34041         (s390_expand_atomic_exchange_tdsi): New backend specific expander for
34042         atomic_exchange.
34043         (s390_match_ccmode_set): Allow CCZmode <-> CCZ1 mode.
34044         * config/s390/s390.md ("atomic_compare_and_swap<mode>"): Merge the
34045         patterns for small and large integers.  Forbid symref memory operands.
34046         Move expander to s390.c.  Require cc register.
34047         ("atomic_compare_and_swap<DGPR:mode><CCZZ1:mode>_internal")
34048         ("*atomic_compare_and_swap<TDI:mode><CCZZ1:mode>_1")
34049         ("*atomic_compare_and_swapdi<CCZZ1:mode>_2")
34050         ("*atomic_compare_and_swapsi<CCZZ1:mode>_3"): Use s_operand to forbid
34051         symref memory operands.  Remove CC mode and call s390_match_ccmode
34052         instead.
34053         ("atomic_exchange<mode>"): Allow and implement all integer modes.
34055 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34057         * config/s390/s390.md (define_peephole2): New peephole to help
34058         combining the load-and-test pattern with volatile memory.
34060 2017-04-25  Dominik Vogt  <vogt@linux.vnet.ibm.com>
34062         * config/s390/s390.md ("cstorecc4"): Use load-on-condition and deal
34063         with CCZmode for TARGET_Z196.
34065 2017-04-25  Jakub Jelinek  <jakub@redhat.com>
34067         PR rtl-optimization/80501
34068         * combine.c (make_compound_operation_int): Set subreg_code to SET
34069         even for AND with mask of the sign bit of mode.
34071         PR rtl-optimization/80500
34072         * loop-unroll.c (combine_var_copies_in_loop_exit): Call copy_rtx on
34073         sum's initial value.
34075 2017-04-25  Julian Brown  <julian@codesourcery.com>
34076             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34078         * config/aarch64/thunderx2t99.md (thunderx2t99_crc): New Reservation.
34080 2017-04-25  Marc Glisse  <marc.glisse@inria.fr>
34082         * fold-const.c (tree_single_nonzero_warnv_p): Handle SSA_NAME.
34084 2017-04-25  Julian Brown  <julian@codesourcery.com>
34085             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34087         * config/aarch64/thunderx2t99.md (thunderx2t99_aes): New Reservation.
34088         (thunderx2t99_sha): New Reservation.
34090 2017-04-25  Julian Brown  <julian@codesourcery.com>
34091             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
34093         * config/aarch64/aarch64-simd.md (aarch64_simd_vec_set<mode>): Fix
34094         type for 1-element load.
34096 2017-04-24  Marc Glisse  <marc.glisse@inria.fr>
34098         * match.pd (X/[ex]C CMP Y/[ex]C): New transformation.
34100 2017-04-24  Martin Jambor  <mjambor@suse.cz>
34102         PR tree-optimization/80293
34103         * tree-sra.c (scalarizable_type_p): New parameter const_decl, make
34104         char arrays not totally scalarizable if it is false.
34105         (analyze_all_variable_accesses): Pass correct value in the new
34106         parameter.  Add a statistics counter.
34108 2017-04-24  Jan Hubicka  <hubicka@ucw.cz>
34110         PR middle-end/79931
34111         * ipa-devirt.c (dump_possible_polymorphic_call_targets): Fix ICE.
34113 2017-04-24  Richard Biener  <rguenther@suse.de>
34115         PR tree-optimization/80494
34116         * tree-scalar-evolution.c (analyze_scalar_evolution_1): Bail
34117         out for complex types.
34119 2017-04-24  Richard Biener  <rguenther@suse.de>
34121         * tree-ssa-sccvn.h (run_scc_vn): Adjust prototype.
34122         * tree-ssa-sccvn.c (print_scc): Print SCC size.
34123         (extract_and_process_scc_for_name): Never fail but drop SCC to varying.
34124         (DFS): Adjust and never fail.
34125         (sccvn_dom_walker::fail): Remove.
34126         (sccvn_dom_walker::before_dom_children): Adjust.
34127         (run_scc_vn): Likewise and never fail.
34128         * tree-ssa-pre.c (pass_pre::execute): Adjust.
34129         (pass_fre::execute): Likewise.
34131 2017-04-24  Richard Biener  <rguenther@suse.de>
34133         PR tree-optimization/79725
34134         * tree-ssa-sink.c (statement_sink_location): Return whether
34135         failure reason was zero uses.  Move that check later.
34136         (sink_code_in_bb): Deal with zero uses by removing the stmt
34137         if possible.
34139 2017-04-24  Richard Biener  <rguenther@suse.de>
34141         PR c++/2972
34142         * tree-ssa-uninit.c (warn_uninitialized_vars): Handle some
34143         pointer-based references.
34145 2017-04-24  Richard Biener  <rguenther@suse.de>
34147         PR bootstrap/79814
34148         * pass_manager.h (pass_manager::operator new): Remove.
34149         (pass_manager::operator delete): Likewise.
34150         * passes.c (pass_manager::operator new): Remove.
34151         (pass_manager::operator delete): Likewise.
34152         (pass_manager::pass_manager): Zero individual pass members.
34154 2017-04-23  Uros Bizjak  <ubizjak@gmail.com>
34156         PR target/70799
34157         * config/i386/i386.c (dimode_scalar_to_vector_candidate_p)
34158         <case ASHIFT, case LSHIFTRT>: Also consider variable shifts.
34159         Check "XEXP (src, 1)" operand here.
34160         <case PLUS, case MINUS, case IOR, case XOR, case AND>:
34161         Check "XEXP (src, 1)" operand here.
34162         (dimode_scalar_chain::make_vector_copies): Detect count register
34163         of a shift instruction.  Zero extend count register from QImode
34164         to DImode to satisfy vector shift pattern count operand predicate.
34165         Substitute vector shift count operand with a DImode copy.
34166         (dimode_scalar_chain::convert_reg): Ditto, zero-extend from
34167         vector register.
34169 2017-04-21  Uros Bizjak  <ubizjak@gmail.com>
34171         * config/i386/i386.md (*extzvqi_mem_rex64): Move above *extzv<mode>.
34172         Remove UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34173         (*insvqi_1_mem_rex64): Move above insv<mode>_1.  Remove
34174         UNSPEC_NOREX_MEM tag.  Update corresponding peephole2 pattern.
34175         (UNSPEC_NOREX_MEM): Remove definition.
34177 2017-04-21  Richard Biener  <rguenther@suse.de>
34179         PR tree-optimization/79547
34180         * tree-ssa-structalias.c (find_func_aliases_for_builtin_call):
34181         Handle strlen, strcmp, strncmp, strcasecmp, strncasecmp, memcmp,
34182         bcmp, strspn, strcspn, __builtin_object_size and __builtin_constant_p
34183         without any constraints.
34185 2017-04-21  Richard Biener  <rguenther@suse.de>
34187         PR tree-optimization/78847
34188         * fold-const.c (split_tree): Handle POINTER_PLUS_EXPR.
34190 2017-04-21  Richard Biener  <rguenther@suse.de>
34192         * tree.h (build_qualified_type): Annotate with CXX_MEM_STAT_INFO.
34193         (build_distinct_type_copy): Likewise.
34194         (build_variant_type_copy): Likewise.
34195         * tree.c (build_qualified_type): Pass down mem-stat info.
34196         (build_distinct_type_copy): Likewise.
34197         (build_variant_type_copy): Likewise.
34199 2017-04-21  Richard Biener  <rguenther@suse.de>
34201         PR tree-optimization/80237
34202         * tree-ssa-pre.c (find_leader_in_sets): Add third set argument,
34203         defaulted to NULL.
34204         (phi_translate_1): Also allow a leader in AVAIL_OUT of pred
34205         for a simplified result.
34207 2016-04-21  Richard Biener  <rguenther@suse.de>
34209         * tree-ssa-loop-ivcanon.c (constant_after_peeling): Do not require
34210         sth as strict as a simple_iv but a chrec without symbols and an
34211         operand defined in the loop we are peeling (and not some subloop).
34212         (propagate_constants_for_unrolling): Propagate all constants.
34214 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34216         PR target/79804
34217         * config/i386/i386.c (print_reg): Remove assert for disalowed
34218         regno values, call output_operand_lossage instead.
34220 2017-04-20  Uros Bizjak  <ubizjak@gmail.com>
34222         PR target/78090
34223         * config/i386/constraints.md (Yc): New register constraint.
34224         * config/i386/i386.md (*float<SWI48:mode><MODEF:mode>2_mixed):
34225         Use Yc constraint for alternative 2 of operand 0.  Remove
34226         preferred_for_speed attribute.
34228 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34230         * omp-low.c (lower_lastprivate_clauses): Correct handling of linear and
34231         lastprivate clauses in SIMT case.
34233 2017-04-20  Volker Reichelt  <v.reichelt@netcologne.de>
34235         * doc/invoke.texi (-Wextra-semi): Document new warning option.
34237 2017-04-20  Richard Biener  <rguenther@suse.de>
34239         PR tree-optimization/57796
34240         * tree-vect-stmts.c (vect_model_store_cost): Cost scatters
34241         as N scalar stores.
34242         (vect_model_load_cost): Cost gathers as N scalar loads.
34244 2017-04-20  Richard Biener  <rguenther@suse.de>
34246         * ggc-page.c (ggc_allocated_p): Rename to ...
34247         (safe_lookup_page_table_entry): ... this and return the lookup
34248         result.
34249         (gt_ggc_m_S): Use safe_lookup_page_table_entry.
34251 2017-04-20  Richard Biener  <rguenther@suse.de>
34253         PR tree-optimization/80453
34254         * tree-ssa-sccvn.h (struct vn_phi_s): Add cclhs and ccrhs members.
34255         * tree-ssa-sccvn.c (cond_stmts_equal_p): Use recorded lhs and rhs
34256         from the conditions.
34257         (vn_phi_eq): Pass them down.
34258         (vn_phi_lookup): Record them.
34259         (vn_phi_insert): Likewise.
34261 2017-04-20  Matthew Fortune  <matthew.fortune@imgtec.com>
34263         * config/mips/mips.c (mips_expand_vec_perm_const): Re-fix
34264         uninitialized variable warning to avoid buffer overrun.
34266 2017-04-20  Alexander Monakov  <amonakov@ispras.ru>
34268         PR other/71250
34269         * doc/invoke.texi (-Wmissing-field-initializers): Mention that warning
34270         is suppressed for '{ 0 }' in C.
34272 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34274         * BASE-VER: Set to 8.0.0.
34276 2017-04-20  Thomas Preud'homme  <thomas.preudhomme@arm.com>
34278         * config/arm/arm.c (arm_elf_asm_cdtor): Create non-default
34279         priority .init_array and .fini_array section with SECTION_NOTYPE
34280         flag.
34282 2017-04-20  Jakub Jelinek  <jakub@redhat.com>
34284         PR middle-end/80423
34285         * tree.h (build_array_type): Add typeless_storage default argument.
34286         * tree.c (type_cache_hasher::equal): Also compare
34287         TYPE_TYPELESS_STORAGE flag for ARRAY_TYPEs.
34288         (build_array_type): Add typeless_storage argument, set
34289         TYPE_TYPELESS_STORAGE to it, if shared also hash it, and pass to
34290         recursive call.
34291         (build_nonshared_array_type): Adjust build_array_type_1 caller.
34292         (build_array_type): Likewise.  Add typeless_storage argument.
34294 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34295             Jakub Jelinek  <jakub@redhat.com>
34297         PR tree-optimization/80426
34298         * tree-vrp.c (extract_range_from_binary_expr_1): For an additive
34299         operation on symbolic operands, also compute the overflow for the
34300         invariant part when the operation degenerates into a negation.
34302 2017-04-19  Jakub Jelinek  <jakub@redhat.com>
34304         PR debug/80461
34305         * dwarf2out.c (modified_type_die, gen_type_die_with_usage):
34306         Check for t with zero TYPE_QUALS_NO_ADDR_SPACE.
34308         PR debug/80436
34309         * tree-ssa-loop-manip.c (find_uses_to_rename_def): Ignore debug uses.
34311 2017-04-19  Georg-Johann Lay  <avr@gjlay.de>
34313         PR target/80462
34314         * config/avr/avr.c (tree.h): Include it.
34315         (cgraph.h): Include it.
34316         (avr_encode_section_info): Don't warn for uninitialized progmem
34317         variable if it's just an alias.
34319 2017-04-19  Richard Biener  <rguenther@suse.de>
34321         PR ipa/65972
34322         * auto-profile.c (afdo_vpt_for_early_inline): Update SSA
34323         when needed by AutoPGO.
34325 2017-04-19  Paulo J. Matos  <paulo@matos-sorge.com>
34327         PR lto/50345
34328         * doc/lto.texi: Remove an extra 'that'.
34330 2017-04-19  Segher Boessenkool  <segher@kernel.crashing.org>
34332         PR rtl-optimization/80429
34333         * ira.c (split_live_ranges_for_shrink_wrap): Don't split regs that
34334         are only used in debug insns.
34336 2017-04-19  Eric Botcazou  <ebotcazou@adacore.com>
34337             Vladimir Makarov  <vmakarov@redhat.com>
34339         * config/sparc/predicates.md (input_operand): Add comment.  Return
34340         true for any memory operand when LRA is in progress.
34341         * config/sparc/sparc.c (sparc_expand_move): Minor formatting fix.
34343 2017-04-18  Jeff Law  <law@redhat.com>
34345         PR target/74563
34346         * mips.md ({return,simple_return}_internal): Do not overwrite
34347         operands[0].
34349 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34351         PR tree-optimization/80443
34352         * tree-vrp.c (intersect_ranges): For signed 1-bit precision type,
34353         instead of adding 1, subtract -1 and similarly instead of subtracting
34354         1 add -1.
34356 2017-04-18  Richard Sandiford  <richard.sandiford@arm.com>
34358         PR rtl-optimization/80357
34359         * haifa-sched.c (tmp_bitmap): New variable.
34360         (model_recompute): Handle duplicate use records.
34361         (alloc_global_sched_pressure_data): Initialize tmp_bitmap.
34362         (free_global_sched_pressure_data): Free it.
34364 2017-04-18  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34366         Revert:
34367         2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
34368         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
34369         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
34370         instead of SYSTEM_HEADER_DIR.
34372 2017-04-18  Jeff Law  <law@redhat.com>
34374         PR middle-end/80422
34375         * cfgcleanup.c (try_crossjump_to_edge): Verify SRC1 and SRC2 have
34376         predecessors after walking up the insn chain.
34378 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34380         PR debug/80263
34381         * dwarf2out.c (modified_type_die): Try harder not to emit internal
34382         sizetype type into debug info.
34384 2017-04-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
34386         PR target/80099
34387         * config/rs6000/rs6000.c (rs6000_expand_vector_extract): Eliminate
34388         unneeded test for TARGET_UPPER_REGS_SF.
34389         * config/rs6000/vsx.md (vsx_extract_v4sf_var): Likewise.
34391 2017-04-18  Jakub Jelinek  <jakub@redhat.com>
34393         PR sanitizer/80444
34394         * sancov.c (sancov_pass): Use gsi_start_nondebug_after_labels_bb
34395         instead of gsi_after_labels.
34397 2017-04-18  Jeff Law  <law@redhat.com>
34399         * regcprop.c (maybe_mode_change): Avoid creating copies of the
34400         stack pointer.
34402         Revert:
34403         2017-04-13  Jeff Law  <law@redhat.com>
34404         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34405         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34407 2017-04-18  Georg-Johann Lay  <avr@gjlay.de>
34409         PR target/79453
34410         * config/avr/avr.c (intl.h): Include it.
34411         (avr_pgm_check_var_decl) [reason]: Wrap diagnostic snippets into _().
34413 2017-04-18  Martin Liska  <mliska@suse.cz>
34415         PR gcov-profile/78783
34416         * gcov-tool.c (gcov_output_files): Validate that destination
34417         file is either removed by the tool or by a user.
34419 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34420             Guy Benyei  <guybe@mellanox.com>
34422         * config/arc/arc.c (arc_reorg): Move loop_end_id into a more local
34423         block, and do not negate it, the stored id is already negative.
34425 2017-04-14  Andrew Burgess  <andrew.burgess@embecosm.com>
34427         * config/arc/arc.md (doloop_begin_i): Use @pcl assembler syntax.
34429 2017-04-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
34431         PR target/80098
34432         * config/rs6000/rs6000-cpus.def (OTHER_P9_VECTOR_MASKS): Define
34433         masks of options that should be turned off if the VSX vector
34434         options are turned off.
34435         (OTHER_P8_VECTOR_MASKS): Likewise.
34436         (OTHER_VSX_VECTOR_MASKS): Likewise.
34437         * config/rs6000/rs6000.c (rs6000_option_override_internal): Call
34438         rs6000_disable_incompatible_switches to validate no type switches
34439         like -mvsx.
34440         (rs6000_incompatible_switch): New function to disallow turning on
34441         other vector options if -mno-vsx, -mno-power8-vector, or
34442         -mno-power9-vector are specified.
34444 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34446         * config/arc/arc.h (CRT_CALL_STATIC_FUNCTION): Use long calls.
34448 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34450         * config/arc/arc-protos.h (arc_decl_pretend_args): Remove.
34451         * config/arc/arc.c (arc_decl_pretend_args): Likewise.
34452         * config/arc/arc.h (CFA_FRAME_BASE_OFFSET): Likewise.
34453         (ARG_POINTER_CFA_OFFSET): Likewise.
34455 2017-04-14  Claudiu Zissulescu  <claziss@synopsys.com>
34457         * config/arc/arc.c (arc_mode_dependent_address_p): Relax
34458         conditions to take advantage of various optimizations.
34460 2017-04-13  Jeff Law  <law@redhat.com>
34462         * config/mips.mips.md (zero_extendsidi2): Do not allow SP to appear
34463         in operands[1] if it is a MEM and TARGET_MIPS16 is active.
34464         (zero_extendsidi2_dext): Likewise.
34466 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34468         PR sanitizer/80403
34469         * fold-const.c (fold_ternary_loc): Revert
34470         use op0 instead of fold_convert_loc (loc, type, arg0) part of
34471         2017-04-12 change.
34473 2017-04-13  Vladimir Makarov  <vmakarov@redhat.com>
34475         PR rtl-optimization/80343
34476         * lra-remat.c (update_scratch_ops): Assign original hard reg to
34477         new scratch pseudo.
34479 2017-04-13  Denis Khalikov <d.khalikov@partner.samsung.com>
34481         PR sanitizer/80414
34482         * ubsan.c (ubsan_expand_bounds_ifn): Pass original index
34483         to ubsan_encode_value.
34485 2017-04-13  Jeff Law  <law@redhat.com>
34487         * reload1.c (eliminate_regs_1): Call gen_rtx_raw_SUBREG for SUBREGs
34488         appearing in DEBUG_INSNs.
34490 2017-04-13  Martin Liska  <mliska@suse.cz>
34492         PR gcov-profile/80413
34493         * gcov-io.c (gcov_write_string): Copy to buffer just when
34494         allocated size is greater than zero.
34496 2017-04-13  Jakub Jelinek  <jakub@redhat.com>
34498         PR debug/80321
34499         * dwarf2out.c (decls_for_scope): Ignore declarations of
34500         current_function_decl in BLOCK_NONLOCALIZED_VARS.
34502 2017-04-12  Jan Hubicka  <hubicka@ucw.cz>
34504         PR lto/69953
34505         * ipa-visibility.c (non_local_p): Fix typos.
34506         (localize_node): When localizing symbol in same comdat group,
34507         dissolve the group only when we know external symbols are going
34508         to be privatized.
34509         (function_and_variable_visibility): Do not localize DECL_EXTERNAL.
34511 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34513         PR tree-optimization/79390
34514         * optabs.c (emit_conditional_move): If the preferred op2/op3 operand
34515         order does not result in usable sequence, retry with reversed operand
34516         order.
34518         PR sanitizer/80403
34519         PR sanitizer/80404
34520         PR sanitizer/80405
34521         * fold-const.c (fold_ternary_loc): Use op1 instead of arg1 as argument
34522         to fold_build2_loc.  Convert TREE_OPERAND (tem, 0) to type.  Use
34523         op0 instead of fold_convert_loc (loc, type, arg0).
34525 2017-04-12  Jeff Law  <law@redhat.com>
34527         * genattrtab.c (write_eligible_delay): Verify DELAY_INSN still
34528         has a delay slot in the generated code.
34530         * config/cris/cris.md (cris_preferred_reload_class): Return
34531         GENNONACR_REGS rather than GENERAL_REGS.
34533 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34535         PR c/80163
34536         * expr.c <CASE_CONVERT>: For EXPAND_INITIALIZER determine SIGN_EXTEND
34537         vs. ZERO_EXTEND based on signedness of treeop0's type rather than
34538         signedness of the result type.
34540 2017-04-12  Richard Biener  <rguenther@suse.de>
34541             Jeff Law  <law@redhat.com>
34543         PR tree-optimization/80359
34544         * tree-ssa-dse.c (maybe_trim_partially_dead_store): Do not
34545         trim stores to TARGET_MEM_REFs.
34547 2017-04-12  Richard Biener  <rguenther@suse.de>
34549         PR tree-optimization/79390
34550         * gimple-ssa-split-paths.c (is_feasible_trace): Restrict
34551         threading case even more.
34553 2017-04-12  Segher Boessenkool  <segher@kernel.crashing.org>
34555         PR target/80382
34556         * config/rs6000/sync.md (atomic_load<mode>, atomic_store<mode): Test
34557         for quad_address_p for TImode, instead of just not indexed_address.
34559 2017-04-12  Richard Biener  <rguenther@suse.de>
34560             Bernd Edlinger  <bernd.edlinger@hotmail.de>
34562         PR middle-end/79671
34563         * alias.c (component_uses_parent_alias_set_from): Handle
34564         TYPE_TYPELESS_STORAGE.
34565         (get_alias_set): Likewise.
34566         * tree-core.h (tree_type_common): Add typeless_storage flag.
34567         * tree.h (TYPE_TYPELESS_STORAGE): New macro.
34568         * stor-layout.c (place_union_field): Set TYPE_TYPELESS_STORAGE
34569         for types containing members with TYPE_TYPELESS_STORAGE.
34570         (place_field): Likewise.
34571         (layout_type): Likewise for ARRAY_TYPE.
34572         * lto-streamer-out.c (hash_tree): Hash TYPE_TYPELESS_STORAGE.
34573         * tree-streamer-in.c (unpack_ts_type_common_value_fields): Stream
34574         TYPE_TYPELESS_STORAGE.
34575         * tree-streamer-out.c (pack_ts_type_common_value_fields): Likewise.
34577 2017-04-12  Jakub Jelinek  <jakub@redhat.com>
34579         PR sanitizer/80349
34580         * fold-const.c (fold_binary_loc) <case BIT_IOR_EXPR>: Convert arg0's
34581         first argument to type.
34583 2017-04-11  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
34585         PR target/80376
34586         PR target/80315
34587         * config/rs6000/rs6000.c (rs6000_expand_unop_builtin): Return
34588         CONST0_RTX (mode) rather than const0_rtx where appropriate.
34589         (rs6000_expand_binop_builtin): Likewise.
34590         (rs6000_expand_ternop_builtin): Likewise; also add missing
34591         vsx_xxpermdi_* variants; also fix typo (arg1 => arg2) for
34592         vshasigma built-ins.
34593         * doc/extend.texi: Document that vec_xxpermdi's third argument
34594         must be a constant.
34596 2017-04-11  Uros Bizjak  <ubizjak@gmail.com>
34598         * config/i386/i386.c (dimode_scalar_chain::compute_convert_gain):
34599         Use shift_const cost parameter when calculating gain of STV shifts.
34601 2017-04-11  Vladimir Makarov  <vmakarov@redhat.com>
34603         PR rtl-optimization/70478
34604         * lra-constraints.c (process_alt_operands): Check memory for
34605         disfavoring memory insn operand.
34607 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34609         PR middle-end/80100
34610         * simplify-rtx.c (simplify_binary_operation_1) <case IOR>: Perform
34611         left shift in unsigned HOST_WIDE_INT type.
34613         PR rtl-optimization/80385
34614         * simplify-rtx.c (simplify_unary_operation_1): Don't transform
34615         (not (neg X)) into (plus X -1) for complex or non-integral modes.
34617         PR libgomp/80394
34618         * omp-low.c (scan_omp_task): Don't optimize away empty tasks
34619         if they have any depend clauses.
34621 2017-04-11  Martin Liska  <mliska@suse.cz>
34623         PR ipa/80212
34624         * cgraph.c (cgraph_node::dump): Dump calls_comdat_local.
34625         * ipa-split.c (split_function): Create a local comdat symbol
34626         if caller is in a comdat group.
34628 2017-04-11  Martin Liska  <mliska@suse.cz>
34630         PR ipa/80212
34631         * ipa-cp.c (determine_versionability): Handle calls_comdat_local
34632         flags.
34634 2017-04-11  Martin Sebor  <msebor@redhat.com>
34636         PR middle-end/80364
34637         * gimple-ssa-sprintf.c (get_int_range): Remove second argument and
34638         always use the int type.  Use INTEGRAL_TYPE_P() rather than testing
34639         for INTEGER_TYPE.
34640         (directive::set_width, directive::set_precision, format_character):
34641         Adjust.
34642         (parse_directive): Use INTEGRAL_TYPE_P() rather than testing for
34643         INTEGER_TYPE.
34645 2017-04-11  Richard Earnshaw  <rearnsha@arm.com>
34647         PR target/80389
34648         * config/arm/arm.c (arm_configure_build_target): When -mcpu and -arch
34649         conflict, set target->arch_name instead of target->cpu_name.
34651 2017-04-11  Richard Biener  <rguenther@suse.de>
34653         PR tree-optimization/80374
34654         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Use
34655         build_zero_cst, remove fold_convertible_p check again.
34657 2017-04-11  Martin Liska  <mliska@suse.cz>
34659         PR sanitizer/70878
34660         * ubsan.c (instrument_object_size): Do not instrument register
34661         variables.
34663 2017-04-11  Jakub Jelinek  <jakub@redhat.com>
34665         PR target/80381
34666         * config/i386/i386-builtin-types.def
34667         (V16HI_FTYPE_V16HI_INT_V16HI_UHI_COUNT,
34668         V16HI_FTYPE_V16HI_V8HI_V16HI_UHI_COUNT,
34669         V16SI_FTYPE_V16SI_INT_V16SI_UHI_COUNT,
34670         V16SI_FTYPE_V16SI_V4SI_V16SI_UHI_COUNT,
34671         V2DI_FTYPE_V2DI_INT_V2DI_UQI_COUNT,
34672         V2DI_FTYPE_V2DI_V2DI_V2DI_UQI_COUNT,
34673         V32HI_FTYPE_V32HI_INT_V32HI_USI_COUNT,
34674         V32HI_FTYPE_V32HI_V8HI_V32HI_USI_COUNT,
34675         V4DI_FTYPE_V4DI_INT_V4DI_UQI_COUNT,
34676         V4DI_FTYPE_V4DI_V2DI_V4DI_UQI_COUNT,
34677         V4SI_FTYPE_V4SI_INT_V4SI_UQI_COUNT,
34678         V4SI_FTYPE_V4SI_V4SI_V4SI_UQI_COUNT,
34679         V8DI_FTYPE_V8DI_INT_V8DI_UQI_COUNT,
34680         V8DI_FTYPE_V8DI_V2DI_V8DI_UQI_COUNT,
34681         V8HI_FTYPE_V8HI_INT_V8HI_UQI_COUNT,
34682         V8HI_FTYPE_V8HI_V8HI_V8HI_UQI_COUNT,
34683         V8SI_FTYPE_V8SI_INT_V8SI_UQI_COUNT,
34684         V8SI_FTYPE_V8SI_V4SI_V8SI_UQI_COUNT): New function type aliases.
34685         * config/i386/i386-builtin.def (__builtin_ia32_pslld512_mask,
34686         __builtin_ia32_pslldi512_mask, __builtin_ia32_psllq512_mask,
34687         __builtin_ia32_psllqi512_mask, __builtin_ia32_psrad512_mask,
34688         __builtin_ia32_psradi512_mask, __builtin_ia32_psraq512_mask,
34689         __builtin_ia32_psraqi512_mask, __builtin_ia32_psrld512_mask,
34690         __builtin_ia32_psrldi512_mask, __builtin_ia32_psrlq512_mask,
34691         __builtin_ia32_psrlqi512_mask, __builtin_ia32_psllwi128_mask,
34692         __builtin_ia32_pslldi128_mask, __builtin_ia32_psllqi128_mask,
34693         __builtin_ia32_psllw128_mask, __builtin_ia32_pslld128_mask,
34694         __builtin_ia32_psllq128_mask, __builtin_ia32_psllwi256_mask,
34695         __builtin_ia32_psllw256_mask, __builtin_ia32_pslldi256_mask,
34696         __builtin_ia32_pslld256_mask, __builtin_ia32_psllqi256_mask,
34697         __builtin_ia32_psllq256_mask, __builtin_ia32_psradi128_mask,
34698         __builtin_ia32_psrad128_mask, __builtin_ia32_psradi256_mask,
34699         __builtin_ia32_psrad256_mask, __builtin_ia32_psraqi128_mask,
34700         __builtin_ia32_psraq128_mask, __builtin_ia32_psraqi256_mask,
34701         __builtin_ia32_psraq256_mask, __builtin_ia32_psrldi128_mask,
34702         __builtin_ia32_psrld128_mask, __builtin_ia32_psrldi256_mask,
34703         __builtin_ia32_psrld256_mask, __builtin_ia32_psrlqi128_mask,
34704         __builtin_ia32_psrlq128_mask, __builtin_ia32_psrlqi256_mask,
34705         __builtin_ia32_psrlq256_mask, __builtin_ia32_psrawi256_mask,
34706         __builtin_ia32_psraw256_mask, __builtin_ia32_psrawi128_mask,
34707         __builtin_ia32_psraw128_mask, __builtin_ia32_psrlwi256_mask,
34708         __builtin_ia32_psrlw256_mask, __builtin_ia32_psrlwi128_mask,
34709         __builtin_ia32_psrlw128_mask, __builtin_ia32_psllwi512_mask,
34710         __builtin_ia32_psllw512_mask, __builtin_ia32_psrawi512_mask,
34711         __builtin_ia32_psraw512_mask, __builtin_ia32_psrlwi512_mask,
34712         __builtin_ia32_psrlw512_mask): Use _COUNT suffixed function type
34713         aliases.
34714         * config/i386/i386.c (ix86_expand_args_builtin): Rename last_arg_count
34715         flag to second_arg_count, handle 4 argument function type _COUNT
34716         aliases, handle second_arg_count on second argument rather than last.
34718 2017-04-10  Jeff Law  <law@redhat.com>
34720         PR tree-optimization/80374
34721         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Do not try to
34722         record anything if we can not convert integer_zero_node to the
34723         desired type.
34725 2017-04-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
34727         PR target/80108
34728         * config/rs6000/rs6000.c (rs6000_option_override_internal):
34729         Enhance special handling given to the TARGET_P9_MINMAX option in
34730         relation to certain other options.
34732 2017-04-10  Bin Cheng  <bin.cheng@arm.com>
34734         PR tree-optimization/80153
34735         * tree-ssa-loop-ivopts.c (add_iv_candidate_for_use): Check and
34736         remove POINTER_PLUS_EXPR's base part directly, rather than through
34737         aff_tree.
34739 2017-04-10  Richard Biener  <rguenther@suse.de>
34740             Bin Cheng  <bin.cheng@arm.com>
34742         PR tree-optimization/80153
34743         * tree-affine.c (aff_combination_to_tree): Get base pointer from
34744         the first element of pointer type aff_tree.  Build result expr in
34745         aff_tree's type.
34746         (add_elt_to_tree): Convert to type unconditionally.  Remove other
34747         fold_convert calls.
34748         * tree-ssa-loop-ivopts.c (alloc_iv): Pass in consistent types.
34749         (rewrite_use_nonlinear_expr): Check invariant using iv information.
34751 2017-04-10  Richard Biener  <rguenther@suse.de>
34753         * tree-ssa-structalias.c (find_func_aliases): Properly handle
34754         asm inputs.
34756 2017-04-10  Vladimir Makarov  <vmakarov@redhat.com>
34758         PR rtl-optimization/70478
34759         * lra-constraints.c (curr_small_class_check): New.
34760         (update_and_check_small_class_inputs): New.
34761         (process_alt_operands): Update curr_small_class_check.  Disfavor
34762         alternative insn memory operands.  Check available regs for small
34763         class operands.
34765 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
34767         PR target/80057
34768         * config/mips/mips.opt (-mvirt): Update description.
34769         * doc/invoke.texi (-mvirt): Likewise.
34771 2017-04-10  Richard Biener  <rguenther@suse.de>
34773         PR middle-end/80362
34774         * fold-const.c (fold_binary_loc): Look at unstripped ops when
34775         looking for NEGATE_EXPR in -A / -B to A / B folding.
34777 2017-04-10  Martin Liska  <mliska@suse.cz>
34779         PR gcov-profile/80224
34780         * gcov.c (print_usage): Fix usage string.
34781         (get_gcov_intermediate_filename): Remove.
34782         (output_gcov_file): Use both for normal and intermediate format.
34783         (generate_results): Do not initialize special file for
34784         intermediate format.
34786 2017-04-10  Richard Biener  <rguenther@suse.de>
34788         PR tree-optimization/80304
34789         * tree-ssa-loop-im.c (ref_indep_loop_p_1): Also recurse
34790         for safelen.
34792 2017-04-10  Nathan Sidwell  <nathan@acm.org>
34794         PR target/79905
34795         * config/rs6000/rs6000.c (rs6000_vector_type): New.
34796         (rs6000_init_builtins): Use it.
34798 2016-04-10  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
34800         * config/arm/arm.md (<mrc>): Add mode to SET source.
34801         (<mrrc>): Likewise.
34803 2017-04-10  Richard Biener  <rguenther@suse.de>
34805         PR middle-end/80344
34806         * gimplify.c (is_gimple_mem_rhs_or_call): Allow CLOBBERs.
34808 2017-04-10  Jakub Jelinek  <jakub@redhat.com>
34810         PR target/80324
34811         * config/i386/avx512fintrin.h (_mm512_reduce_add_epi32,
34812         _mm512_reduce_mul_epi32, _mm512_reduce_and_epi32,
34813         _mm512_reduce_or_epi32, _mm512_mask_reduce_add_epi32,
34814         _mm512_mask_reduce_mul_epi32, _mm512_mask_reduce_and_epi32,
34815         _mm512_mask_reduce_or_epi32, _mm512_reduce_min_epi32,
34816         _mm512_reduce_max_epi32, _mm512_reduce_min_epu32,
34817         _mm512_reduce_max_epu32, _mm512_mask_reduce_min_epi32,
34818         _mm512_mask_reduce_max_epi32, _mm512_mask_reduce_min_epu32,
34819         _mm512_mask_reduce_max_epu32, _mm512_reduce_add_ps,
34820         _mm512_reduce_mul_ps, _mm512_mask_reduce_add_ps,
34821         _mm512_mask_reduce_mul_ps, _mm512_reduce_min_ps, _mm512_reduce_max_ps,
34822         _mm512_mask_reduce_min_ps, _mm512_mask_reduce_max_ps,
34823         _mm512_reduce_add_epi64, _mm512_reduce_mul_epi64,
34824         _mm512_reduce_and_epi64, _mm512_reduce_or_epi64,
34825         _mm512_mask_reduce_add_epi64, _mm512_mask_reduce_mul_epi64,
34826         _mm512_mask_reduce_and_epi64, _mm512_mask_reduce_or_epi64,
34827         _mm512_reduce_min_epi64, _mm512_reduce_max_epi64,
34828         _mm512_mask_reduce_min_epi64, _mm512_mask_reduce_max_epi64,
34829         _mm512_reduce_min_epu64, _mm512_reduce_max_epu64,
34830         _mm512_mask_reduce_min_epu64, _mm512_mask_reduce_max_epu64,
34831         _mm512_reduce_add_pd, _mm512_reduce_mul_pd, _mm512_mask_reduce_add_pd,
34832         _mm512_mask_reduce_mul_pd, _mm512_reduce_min_pd, _mm512_reduce_max_pd,
34833         _mm512_mask_reduce_min_pd, _mm512_mask_reduce_max_pd): New intrinsics.
34835 2017-04-08  Vladimir Makarov  <vmakarov@redhat.com>
34837         PR rtl-optimization/70478
34838         * lra-constraints.c: Reverse the last patch.
34840 2017-04-08  Andreas Tobler  <andreast@gcc.gnu.org>
34842         * config/aarch64/aarch64-freebsd.h: Define MCOUNT_NAME.
34843         Add comment for WCHAR_T.
34845 2017-04-08  Martin Liska  <mliska@suse.cz>
34847         Revert:
34848         2017-04-07  Martin Liska  <mliska@suse.cz>
34850         PR ipa/80212
34851         * ipa-split.c (split_function): Add function part to a same comdat
34852         group.
34854 2017-04-08  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
34856         PR target/80358
34857         * config/rs6000/rs6000.c (expand_block_compare): Fix boundary check.
34859 2017-04-07  Pat Haugen  <pthaugen@us.ibm.com>
34861         * rs6000/rs6000.c (vec_load_pendulum): Rename...
34862         (vec_pairing): ...to this.
34863         (power9_sched_reorder2): Rewrite code for pairing vector/vecload insns.
34864         (rs6000_sched_init): Adjust for name change.
34865         (struct rs6000_sched_context): Likewise.
34866         (rs6000_init_sched_context): Likewise.
34867         (rs6000_set_sched_context): Likewise.
34869 2017-04-07  Jakub Jelinek  <jakub@redhat.com>
34871         PR target/80322
34872         PR target/80323
34873         PR target/80325
34874         PR target/80326
34875         * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
34876         intrinsics.
34877         * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
34878         _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
34879         _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
34881 2017-04-07  Andreas Tobler  <andreast@gcc.gnu.org>
34883         * config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
34885 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34887         PR rtl-optimization/70703
34888         * ira-color.c (update_conflict_hard_regno_costs): Use
34889         int64_t instead of HOST_WIDE_INT.
34891 2017-04-07  Vladimir Makarov  <vmakarov@redhat.com>
34893         PR rtl-optimization/70478
34894         * lra-constraints.c (process_alt_operands): Disfavor alternative
34895         insn memory operands.
34897 2017-04-07  Jeff Law  <law@redhat.com>
34899         * config/iq2000/iq2000.c (final_prescan_insn): Do not separate a
34900         CALL and NOTE_INSN_CALL_ARG_LOCATION.
34902 2017-04-07  Martin Liska  <mliska@suse.cz>
34904         PR target/79889
34905         * config/aarch64/aarch64.c (aarch64_process_target_attr):
34906         Show error message instead of an ICE.
34908 2017-04-07  Martin Liska  <mliska@suse.cz>
34910         PR ipa/80212
34911         * ipa-split.c (split_function): Add function part to a same comdat
34912         group.
34914 2017-04-07  Richard Biener  <rguenther@suse.de>
34916         PR middle-end/80341
34917         * tree.c (get_unwidened): Also handle ! for_type case for
34918         INTEGER_CSTs.
34919         * convert.c (do_narrow): Split out from ...
34920         (convert_to_integer_1): ... here.  Do not pass final truncation
34921         type to get_unwidened for TRUNC_DIV_EXPR.
34923 2017-04-07  Richard Biener  <rguenther@suse.de>
34925         * tree-affine.c (wide_int_ext_for_comb): Take type rather
34926         than aff_tree.
34927         (aff_combination_const): Adjust.
34928         (aff_combination_scale): Likewise.
34929         (aff_combination_add_elt): Likewise.
34930         (aff_combination_add_cst): Likewise.
34931         (aff_combination_convert): Likewise.
34932         (add_elt_to_tree): Likewise.  Remove unused argument.
34933         (aff_combination_to_tree): Adjust calls to add_elt_to_tree.
34935 2017-04-07  Sebastian Huber  <sebastian.huber@embedded-brains.de>
34937         * config/arm/arm.h (ARM_DEFAULT_SHORT_ENUMS): Provide default
34938         definition.
34939         * config/arm/arm.c (arm_default_short_enums): Use
34940         ARM_DEFAULT_SHORT_ENUMS.
34941         * config/arm/rtems.h (ARM_DEFAULT_SHORT_ENUMS): Define.
34943 2017-04-06  Jakub Jelinek  <jakub@redhat.com>
34945         PR debug/80234
34946         * dwarf2out.c (gen_member_die): Handle C++17 inline static data
34947         members with redundant out-of-class redeclaration.
34949 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34951         PR target/80286
34952         * config/i386/sse.md (*vec_extractv4si_0_zext_sse4): New pattern.
34953         * config/i386/i386.md (*zero_extendsidi2):
34954         Add (?*x,*x) and (?*v,*v) alternatives.
34956 2017-04-06  Uros Bizjak  <ubizjak@gmail.com>
34958         PR target/79733
34959         * config/i386/i386.c (ix86_expand_builtin)
34960         <case IX86_BUILTIN_K{,OR}TEST{C,Z}{8,16,32,64}>: Determine insn operand
34961         mode from insn data. Convert operands to insn operand mode.
34962         Copy operands that don't satisfy insn predicate to a register.
34964 2017-04-06  Sam Thursfield  <sam.thursfield@codethink.co.uk>
34966         * config/rs6000/x-aix: Increase memory limit for genautomata on AIX.
34967         Update comments.
34969 2017-04-06  Richard Biener  <rguenther@suse.de>
34971         PR tree-optimization/80334
34972         * tree-ssa-loop-ivopts.c (rewrite_use_address): Properly
34973         preserve alignment of accesses.
34975 2017-04-06  Richard Biener  <rguenther@suse.de>
34977         PR tree-optimization/80262
34978         * tree-sra.c (build_ref_for_offset): Preserve address-space
34979         information.
34980         * tree-ssa-sccvn.c (vn_reference_maybe_forwprop_address):
34981         Drop useless address-space information on MEM_REF offsets.
34983 2017-04-05  Andreas Schwab  <schwab@linux-m68k.org>
34985         * builtins.def (BUILT_IN_UPDATE_SETJMP_BUF): Fix type.
34987 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
34989         PR rtl-optimization/70703
34990         * ira-color.c (update_conflict_hard_regno_costs): Use
34991         HOST_WIDE_INT instead of long.
34993 2017-04-05  Uros Bizjak  <ubizjak@gmail.com>
34995         PR target/80298
34996         * config/i386/mmintrin.h: Add -msse target option when __SSE__ is
34997         not defined for x86_64 target.  Add -mmmx target option when __SSE2__
34998         is not defined.
34999         * config/i386/mm3dnow.h: Add -msse target when __SSE__ is not defined
35000         for x86_64 target.  Handle -m3dnowa option.
35002 2017-04-05  Vladimir Makarov  <vmakarov@redhat.com>
35004         PR rtl-optimization/70703
35005         * ira-color.c (update_costs_from_allocno): Use the smallest mode.
35006         (update_conflict_hard_regno_costs): Use long instead of unsigned
35007         arithmetic for cost calculation.
35009 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35010             Bernd Edlinger  <bernd.edlinger@hotmail.de>
35012         PR sanitizer/80308
35013         * asan.c (asan_store_shadow_bytes): Fix location of last_chunk_value
35014         for big endian.
35016 2017-04-05  Eric Botcazou  <ebotcazou@adacore.com>
35018         PR target/78002
35019         * config/aarch64/aarch64.c (aarch64_emit_probe_stack_range): Replace
35020         ptr_mode with Pmode throughout.
35021         * config/aarch64/aarch64.md (probe_stack_range_<PTR:mode): Rename
35022         into probe_stack_range and use DImode.
35024 2017-04-05  Dominik Vogt  <vogt@linux.vnet.ibm.com>
35026         PR target/79890
35027         * config/s390/s390.c (s390_register_info_gprtofpr): Return if
35028         call_eh_return is true.
35030 2017-04-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35032         * config/s390/s390-c.c (s390_resolve_overloaded_builtin):
35033         Initialize last_match_fntype_index.
35035 2017-04-05  Jakub Jelinek  <jakub@redhat.com>
35037         PR target/80310
35038         * tree-nvr.c: Include internal-fn.h.
35039         (pass_return_slot::execute): Ignore internal calls without
35040         direct optab.
35042 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35043             Richard Biener  <rguenther@suse.de>
35045         PR c++/80297
35046         * genmatch.c (capture::gen_transform): For GENERIC unshare_expr
35047         captures used multiple times, except for the last use.
35048         * generic-match-head.c: Include gimplify.h.
35050 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35052         PR tree-optimization/79390
35053         * target.h (struct noce_if_info): Declare.
35054         * targhooks.h (default_noce_conversion_profitable_p): Declare.
35055         * target.def (noce_conversion_profitable_p): New target hook.
35056         * ifcvt.h (struct noce_if_info): New type, moved from ...
35057         * ifcvt.c (struct noce_if_info): ... here.
35058         (noce_conversion_profitable_p): Renamed to ...
35059         (default_noce_conversion_profitable_p): ... this.  No longer
35060         static nor inline.
35061         (noce_try_store_flag_constants, noce_try_addcc,
35062         noce_try_store_flag_mask, noce_try_cmove, noce_try_cmove_arith,
35063         noce_convert_multiple_sets): Use targetm.noce_conversion_profitable_p
35064         instead of noce_conversion_profitable_p.
35065         * config/i386/i386.c: Include ifcvt.h.
35066         (ix86_option_override_internal): Don't override
35067         PARAM_MAX_RTL_IF_CONVERSION_INSNS default.
35068         (ix86_noce_conversion_profitable_p): New function.
35069         (TARGET_NOCE_CONVERSION_PROFITABLE_P): Redefine.
35070         * config/i386/x86-tune.def (X86_TUNE_ONE_IF_CONV_INSN): Adjust comment.
35071         * doc/tm.texi.in (TARGET_NOCE_CONVERSION_PROFITABLE_P): Add.
35072         * doc/tm.texi: Regenerated.
35074 2017-04-04  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35076         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Grammar
35077         correction.
35079 2017-04-04  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35081         PR target/80307
35082         * config/arm/arm.c (thumb1_rtx_costs): Give a cost of 32
35083         instructions for small multiply cores.
35085 2017-04-04  Jeff Law  <law@redhat.com>
35087         * config/mips/mips.c (mips_multi_add): Zero initialize the newly
35088         added member.
35089         (mips_expand_vec_perm_const): Initialize elements in orig_perm
35090         that are not set by the loop over the elements.
35092 2017-04-04  Jakub Jelinek  <jakub@redhat.com>
35094         PR target/80286
35095         * config/i386/i386.c (ix86_expand_args_builtin): If op has scalar
35096         int mode, convert_modes it to mode as unsigned, otherwise use
35097         lowpart_subreg to mode rather than SImode.
35098         * config/i386/sse.md (<mask_codefor>ashr<mode>3<mask_name>,
35099         ashr<mode>3, ashr<mode>3<mask_name>, <shift_insn><mode>3<mask_name>):
35100         Use DImode instead of SImode for the shift count operand.
35101         * config/i386/mmx.md (mmx_ashr<mode>3, mmx_<shift_insn><mode>3):
35102         Likewise.
35104 2017-04-04  Richard Biener  <rguenther@suse.de>
35106         PR middle-end/80281
35107         * match.pd (A + (-B) -> A - B): Make sure to preserve unsigned
35108         arithmetic done for the negate or the plus.  Simplify.
35109         (A - (-B) -> A + B): Likewise.
35110         * fold-const.c (split_tree): Make sure to not negate pointers.
35112 2017-04-04  Segher Boessenkool  <segher@kernel.crashing.org>
35114         PR rtl-optimization/60818
35115         * simplify-rtx.c (simplify_binary_operation_1): Do not replace
35116         a compare of comparisons with the thing compared if this results
35117         in a different machine mode.
35119 2017-04-03  Jonathan Wakely  <jwakely@redhat.com>
35121         * alias.c (base_alias_check): Fix typo in comment.
35122         * cgraph.h (class ipa_polymorphic_call_context): Likewise.
35123         * cgraphunit.c (symbol_table::compile): Likewise.
35124         * collect2.c (maybe_run_lto_and_relink): Likewise.
35125         * config/arm/arm.c (arm_thumb1_mi_thunk): Likewise.
35126         * config/avr/avr-arch.h (avr_arch_info_t): Likewise.
35127         * config/avr/avr.c (avr_map_op_t): Likewise.
35128         * config/cr16/cr16.h (DATA_ALIGNMENT): Likewise.
35129         * config/epiphany/epiphany.c (TARGET_ARG_PARTIAL_BYTES): Likewise.
35130         * config/epiphany/epiphany.md (movcc): Likewise.
35131         * config/i386/i386.c (legitimize_pe_coff_extern_decl): Likewise.
35132         * config/m68k/m68k.c (struct _sched_ib, m68k_sched_variable_issue):
35133         Likewise.
35134         * config/mips/mips.c (mips_save_restore_reg): Likewise.
35135         * config/rx/rx.c (rx_is_restricted_memory_address): Likewise.
35136         * config/s390/s390.c (Z10_EARLYLOAD_DISTANCE): Likewise.
35137         * config/sh/sh.c (sh_rtx_costs): Likewise.
35138         * fold-const.c (fold_truth_andor): Likewise.
35139         * genautomata.c (collapse_flag): Likewise.
35140         * gengtype.h (struct type::u::s): Likewise.
35141         * gensupport.c (has_subst_attribute, add_mnemonic_string): Likewise.
35142         * input.c (FORMAT_AMOUNT): Likewise.
35143         * ipa-cp.c (class ipcp_lattice, agg_replacements_to_vector)
35144         (known_aggs_to_agg_replacement_list): Likewise.
35145         * ipa-inline-analysis.c: Likewise.
35146         * ipa-inline.h (estimate_edge_time, estimate_edge_hints): Likewise.
35147         * ipa-polymorphic-call.c
35148         (ipa_polymorphic_call_context::restrict_to_inner_class): Likewise.
35149         * loop-unroll.c (analyze_insn_to_expand_var): Likewise.
35150         * lra.c (lra_optional_reload_pseudos, lra_subreg_reload_pseudos):
35151         Likewise.
35152         * modulo-sched.c (apply_reg_moves): Likewise.
35153         * omp-expand.c (build_omp_regions_1): Likewise.
35154         * trans-mem.c (struct tm_wrapper_hasher): Likewise.
35155         * tree-ssa-loop-ivopts.c (may_eliminate_iv): Likewise.
35156         * tree-ssa-loop-niter.c (maybe_lower_iteration_bound): Likewise.
35157         * tree-vect-data-refs.c (vect_enhance_data_refs_alignment): Likewise.
35158         * value-prof.c: Likewise.
35159         * var-tracking.c (val_reset): Likewise.
35161 2017-04-03  Richard Biener  <rguenther@suse.de>
35163         PR tree-optimization/80275
35164         * fold-const.c (split_address_to_core_and_offset): Handle
35165         POINTER_PLUS_EXPR.
35167 2017-04-03  Eric Botcazou  <ebotcazou@adacore.com>
35169         * tree-nested.c (get_descriptor_type): Make sure that the alignment of
35170         descriptors is at least equal to that of functions.
35172 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35174         * config/i386/sse.md (movdi_to_sse): Add missing DONE.
35176 2017-04-02  Uros Bizjak  <ubizjak@gmail.com>
35178         PR target/80250
35179         * config/i386/sse.md (mov<IMOD4:mode>): Remove insn pattern.
35180         (mov<IMOD4:mode>): New expander.
35181         (*mov<IMOD4:mode>_internal): New insn and split pattern.
35183 2017-03-31  Segher Boessenkool  <segher@kernel.crashing.org>
35185         PR rtl-optimization/79405
35186         * fwprop.c (propagations_left): New variable.
35187         (forward_propagate_into): Decrement it.
35188         (fwprop_init): Initialize it.
35189         (fw_prop): If the variable has reached zero, stop propagating.
35190         (fwprop_addr): Ditto.
35192 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35194         PR debug/79255
35195         * dwarf2out.c (decls_for_scope): If BLOCK_NONLOCALIZED_VAR is
35196         a FUNCTION_DECL, pass it as decl instead of origin to
35197         process_scope_var.
35199 2017-03-31  Alexander Monakov  <amonakov@ispras.ru>
35201         * config/nvptx/nvptx.c (nvptx_output_softstack_switch): Correct format
35202         string.
35204 2017-03-31  Pat Haugen  <pthaugen@us.ibm.com>
35206         PR target/80107
35207         * config/rs6000/rs6000.md (extendhi<mode>2): Add test for
35208         TARGET_VSX_SMALL_INTEGER.
35210 2017-03-31  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35212         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Add
35213         reference to the OpenPOWER 64-Bit ELF V2 ABI Specification.
35215 2017-03-31  Matthew Fortune  <matthew.fortune@imgtec.com>
35217         * config/mips/mips-msa.md (msa_vec_extract_<msafmt_f>): Update
35218         extraction from odd-numbered MSA register.
35220 2017-03-31  Jakub Jelinek  <jakub@redhat.com>
35222         PR middle-end/80173
35223         * expmed.c (store_bit_field_1): Don't attempt to create
35224         a word subreg out of hard registers wider than word if they
35225         have HARD_REGNO_NREGS of 1 for their mode.
35227         PR middle-end/80163
35228         * varasm.c (initializer_constant_valid_p_1): Disallow sign-extending
35229         conversions to integer types wider than word and pointer.
35231         PR debug/80025
35232         * cselib.h (rtx_equal_for_cselib_1): Add depth argument.
35233         (rtx_equal_for_cselib_p): Pass 0 to it.
35234         * cselib.c (cselib_hasher::equal): Likewise.
35235         (rtx_equal_for_cselib_1): Add depth argument.  If depth
35236         is 128, don't look up VALUE locs and punt.  Increment
35237         depth in recursive calls when walking VALUE locs.
35239 2017-03-31  Bernd Edlinger  <bernd.edlinger@hotmail.de>
35241         * gcov.c (md5sum_to_hex): Fix output of MD5 hex bytes.
35242         (make_gcov_file_name): Use the canonical path name for generating
35243         the MD5 value.
35244         (read_line): Fix handling of files with ascii null bytes.
35246 2017-03-30  Matthew Fortune  <matthew.fortune@imgtec.com>
35248         * config/mips/mips.c (mips_expand_vector_init): Create a const_vector
35249         to initialise a vector register instead
35250         of using a const_int.
35252 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35254         PR translation/80189
35255         * gimplify.c (omp_default_clause): Use %qs instead of %s in
35256         diagnostic messages.
35258 2017-03-30  Peter Bergner  <bergner@vnet.ibm.com>
35260         PR target/80246
35261         * config/rs6000/dfp.md (dfp_dxex_<mode>): Update mode of operand 0.
35262         (dfp_diex_<mode>): Update mode of operand 1.
35263         * doc/extend.texi (dxex, dxexq): Document change to return type.
35264         (diex, diexq): Document change to argument type.
35266 2017-03-30  Martin Jambor  <mjambor@suse.cz>
35268         PR ipa/77333
35269         * cgraph.h (cgraph_build_function_type_skip_args): Declare.
35270         * cgraph.c (redirect_call_stmt_to_callee): Set gimple fntype so that
35271         it reflects the signature changes performed at the callee side.
35272         * cgraphclones.c (build_function_type_skip_args): Make public, renamed
35273         to cgraph_build_function_type_skip_args.
35274         (build_function_decl_skip_args): Adjust call to the above function.
35276 2017-03-30  Jakub Jelinek  <jakub@redhat.com>
35278         PR target/80206
35279         * config/i386/sse.md
35280         (<extract_type>_vextract<shuffletype><extract_suf>_mask): Use
35281         register as dest whenever it is a MEM not rtx_equal_p to the
35282         corresponding dup operand, and when forcing into reg move the
35283         reg into the memory afterwards.
35284         (<extract_type_2>_vextract<shuffletype><extract_suf_2>_mask):
35285         Likewise.  Use <ssehalfvecmode> instead of <ssequartermode>
35286         for the force_reg mode.
35287         (avx512vl_vextractf128<mode>): Use register as dest either
35288         always when a MEM, or when it is a MEM not rtx_equal_p to the
35289         corresponding dup operand, or even not when it is a CONST_VECTOR
35290         depending on the mode and lo vs. hi.
35291         (avx512dq_vextract<shuffletype>64x2_1_maskm): Remove extraneous
35292         parens.
35293         (avx512f_vextract<shuffletype>32x4_1_maskm): Likewise.
35294         (<mask_codefor>avx512dq_vextract<shuffletype>64x2_1<mask_name>):
35295         Likewise.  Require that operands[2] is even.
35296         (<mask_codefor>avx512f_vextract<shuffletype>32x4_1<mask_name>):
35297         Remove extraneous parens.  Require that operands[2] is a multiple
35298         of 4.
35299         (vec_extract_lo_<mode><mask_name>): Don't bother testing if
35300         operands[0] is a MEM if <mask_applied>, the predicates/constraints
35301         disallow memory then.
35303 2017-03-30  Richard Biener  <rguenther@suse.de>
35305         PR tree-optimization/77498
35306         * tree-ssa-pre.c (phi_translate_1): Do not allow simplifications
35307         to non-constants over backedges.
35309 2017-03-29  Segher Boessenkool  <segher@kernel.crashing.org>
35311         PR rtl-optimization/80233
35312         * combine.c (combine_instructions): Only take NONDEBUG_INSN_P insns
35313         as last_combined_insn.  Do not test for BARRIER_P separately.
35315 2017-03-29  Andreas Schwab  <schwab@suse.de>
35317         PR ada/80146
35318         * calls.c (prepare_call_address): Convert funexp to Pmode before
35319         copying to temp reg.
35321 2017-03-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35323         PR tree-optimization/80158
35324         * gimple-ssa-strength-reduction.c (replace_mult_candidate):
35325         Handle possible future case of more than one alternate
35326         interpretation.
35327         (replace_rhs_if_not_dup): Likewise.
35328         (replace_one_candidate): Likewise.
35330 2017-03-28  Vladimir Makarov  <vmakarov@redhat.com>
35332         PR rtl-optimization/80193
35333         * ira.c (ira): Do not check allocation for LRA.
35335 2017-03-28  Alexander Monakov  <amonakov@ispras.ru>
35337         * config/nvptx/nvptx-protos.h (nvptx_output_simt_enter): Declare.
35338         (nvptx_output_simt_exit): Declare.
35339         * config/nvptx/nvptx.c (nvptx_init_unisimt_predicate): Use
35340         cfun->machine->unisimt_location.  Handle NULL unisimt_predicate.
35341         (init_softstack_frame): Move initialization of crtl->is_leaf to...
35342         (nvptx_declare_function_name): ...here.  Emit declaration of local
35343         memory space buffer for omp_simt_enter insn.
35344         (nvptx_output_unisimt_switch): New.
35345         (nvptx_output_softstack_switch): New.
35346         (nvptx_output_simt_enter): New.
35347         (nvptx_output_simt_exit): New.
35348         * config/nvptx/nvptx.h (struct machine_function): New fields
35349         has_simtreg, unisimt_location, simt_stack_size, simt_stack_align.
35350         * config/nvptx/nvptx.md (UNSPECV_SIMT_ENTER): New unspec.
35351         (UNSPECV_SIMT_EXIT): Ditto.
35352         (omp_simt_enter_insn): New insn.
35353         (omp_simt_enter): New expansion.
35354         (omp_simt_exit): New insn.
35355         * config/nvptx/nvptx.opt (msoft-stack-reserve-local): New option.
35357         * internal-fn.c (expand_GOMP_SIMT_ENTER): New.
35358         (expand_GOMP_SIMT_ENTER_ALLOC): New.
35359         (expand_GOMP_SIMT_EXIT): New.
35360         * internal-fn.def (GOMP_SIMT_ENTER): New internal function.
35361         (GOMP_SIMT_ENTER_ALLOC): Ditto.
35362         (GOMP_SIMT_EXIT): Ditto.
35363         * target-insns.def (omp_simt_enter): New insn.
35364         (omp_simt_exit): Ditto.
35365         * omp-low.c (struct omplow_simd_context): New fields simt_eargs,
35366         simt_dlist.
35367         (lower_rec_simd_input_clauses): Implement SIMT privatization.
35368         (lower_rec_input_clauses): Likewise.
35369         (lower_lastprivate_clauses): Handle SIMT privatization.
35371         * omp-offload.c: Include langhooks.h, tree-nested.h, stor-layout.h.
35372         (ompdevlow_adjust_simt_enter): New.
35373         (find_simtpriv_var_op): New.
35374         (execute_omp_device_lower): Handle IFN_GOMP_SIMT_ENTER,
35375         IFN_GOMP_SIMT_ENTER_ALLOC, IFN_GOMP_SIMT_EXIT.
35377         * tree-inline.h (struct copy_body_data): New field dst_simt_vars.
35378         * tree-inline.c (expand_call_inline): Handle SIMT privatization.
35379         (copy_decl_for_dup_finish): Ditto.
35381         * tree-ssa.c (execute_update_addresses_taken): Handle GOMP_SIMT_ENTER.
35383 2017-03-28  Uros Bizjak  <ubizjak@gmail.com>
35385         PR target/53383
35386         * config/i386/i386.c (ix86_option_override_internal): Always
35387         allow -mpreferred-stack-boundary=3 for 64-bit targets.
35389 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35391         * tree-vect-loop.c (optimize_mask_stores): Add bb to the right loop.
35393 2017-03-28  Bin Cheng  <bin.cheng@arm.com>
35395         * tree-vect-loop-manip.c (slpeel_add_loop_guard): New param and
35396         mark new edge's irreducible flag accordign to it.
35397         (vect_do_peeling): Check loop preheader edge's irreducible flag
35398         and pass it to function slpeel_add_loop_guard.
35400 2017-03-28  Richard Sandiford  <richard.sandiford@arm.com>
35402         PR tree-optimization/80218
35403         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
35404         Update block frequencies and counts.
35406 2017-03-28  Richard Biener  <rguenther@suse.de>
35408         PR tree-optimization/78644
35409         * tree-ssa-ccp.c (evaluate_stmt): When we may not use the value
35410         of a simplification result we may not use it at all.
35412 2017-03-28  Richard Biener  <rguenther@suse.de>
35414         PR ipa/80205
35415         * tree-inline.c (copy_phis_for_bb): Do not create PHI node
35416         without arguments, generate default definition of a SSA name.
35418 2017-03-28  Richard Biener  <rguenther@suse.de>
35420         PR middle-end/80222
35421         * gimple-fold.c (gimple_fold_indirect_ref): Do not touch
35422         TYPE_REF_CAN_ALIAS_ALL references.
35423         * fold-const.c (fold_indirect_ref_1): Likewise.
35425 2017-03-28  Martin Liska  <mliska@suse.cz>
35427         PR ipa/80104
35428         * cgraphunit.c (cgraph_node::expand_thunk): Mark argument of a
35429         thunk call as DECL_GIMPLE_REG_P when vector or complex type.
35431 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35432             Thomas Petazzoni  <thomas.petazzoni@free-electrons.com>
35434         * config/arc/arc.h (CPP_SPEC): Add subtarget_cpp_spec.
35435         (EXTRA_SPECS): Define.
35436         (SUBTARGET_EXTRA_SPECS): Likewise.
35437         (SUBTARGET_CPP_SPEC): Likewise.
35438         * config/arc/elf.h (EXTRA_SPECS): Renamed to
35439         SUBTARGET_EXTRA_SPECS.
35440         * config/arc/linux.h (SUBTARGET_CPP_SPEC): Define.
35442 2017-03-28  Claudiu Zissulescu  <claziss@synopsys.com>
35444         * config/arc/simdext.md (vst64_insn): Update pattern.
35445         (vld32wh_insn): Likewise.
35446         (vld32wl_insn): Likewise.
35447         (vld64_insn): Likewise.
35448         (vld32_insn): Likewise.
35450 2017-03-28  Marek Polacek  <polacek@redhat.com>
35452         PR sanitizer/80067
35453         * fold-const.c (fold_comparison): Use protected_set_expr_location
35454         instead of SET_EXPR_LOCATION.
35456 2017-03-28  Markus Trippelsdorf  <markus@trippelsdorf.de>
35458         * tree.c (add_expr): Avoid name lookup warning.
35460 2017-03-27  Jeff Law  <law@redhat.com>
35462         PR tree-optimization/80216
35463         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): Fix typo in
35464         function name.  Limit recursion depth.
35465         (record_temporary_equivalences): Corresponding changes.
35467 2017-03-27  Jonathan Wakely  <jwakely@redhat.com>
35469         * doc/invoke.texi (-Wno-narrowing): Reorder so default behavior is
35470         covered first.
35472 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35474         PR target/80102
35475         * reg-notes.def (REG_CFA_NOTE): Define.  Use it for CFA related
35476         notes.
35477         * cfgcleanup.c (reg_note_cfa_p): New array.
35478         (insns_have_identical_cfa_notes): New function.
35479         (old_insns_match_p): Don't cross-jump in between /f
35480         and non-/f instructions.  If both i1 and i2 are frame related,
35481         verify all CFA notes, their order and content.
35483 2017-03-27  Michael Meissner  <meissner@linux.vnet.ibm.com>
35485         PR target/78543
35486         * config/rs6000/rs6000.md (bswaphi2_extenddi): Combine bswap
35487         HImode and SImode with zero extend to DImode to one insn.
35488         (bswap<mode>2_extenddi): Likewise.
35489         (bswapsi2_extenddi): Likewise.
35490         (bswaphi2_extendsi): Likewise.
35491         (bswaphi2): Combine bswap HImode and SImode into one insn.
35492         Separate memory insns from swapping register.
35493         (bswapsi2): Likewise.
35494         (bswap<mode>2): Likewise.
35495         (bswaphi2_internal): Delete, no longer used.
35496         (bswapsi2_internal): Likewise.
35497         (bswap<mode>2_load): Split bswap HImode/SImode into separate load,
35498         store, and gpr<-gpr swap insns.
35499         (bswap<mode>2_store): Likewise.
35500         (bswaphi2_reg): Register only splitter, combine with the splitter.
35501         (bswaphi2 splitter): Likewise.
35502         (bswapsi2_reg): Likewise.
35503         (bswapsi2 splitter): Likewise.
35504         (bswapdi2): If we have the LDBRX and STDBRX instructions, split
35505         the insns into load, store, and register/register insns.
35506         (bswapdi2_ldbrx): Likewise.
35507         (bswapdi2_load): Likewise.
35508         (bswapdi2_store): Likewise.
35509         (bswapdi2_reg): Likewise.
35511 2017-03-27  Gunther Nikl  <gnikl@users.sourceforge.net>
35513         * system.h (HAVE_DESIGNATED_INITIALIZERS): Fix non C++ case.
35514         (HAVE_DESIGNATED_UNION_INITIALIZERS): Likewise.
35516 2017-03-27  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35518         PR target/80103
35519         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Edit and
35520         add comments.
35521         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35522         special handling for target option conflicts between dform
35523         options (-mpower9-dform, -mpower9-dform-vector,
35524         -mpower9-dform-scalar) and -mno-direct-move.
35526 2017-03-27  Richard Biener  <rguenther@suse.de>
35528         PR tree-optimization/80181
35529         * tree-ssa-ccp.c (likely_value): UNDEFINED ^ X is UNDEFINED.
35531 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35533         * config/arc/predicates.md (move_double_src_operand): Replace the
35534         call to move_double_src_operand with a call to address_operand.
35536 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35538         * config/arc/elf.h (ARGET_ARC_TP_REGNO_DEFAULT): Define.
35539         * config/arc/linux.h (ARGET_ARC_TP_REGNO_DEFAULT): Likewise.
35540         * config/arc/arc.opt (mtp-regno): Use ARGET_ARC_TP_REGNO_DEFAULT.
35542 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35544         * config/arc/predicates.md (long_immediate_loadstore_operand):
35545         Consider scaled addresses cases.
35547 2017-03-27  Claudiu Zissulescu  <claziss@synopsys.com>
35549         * config/arc/arc.c (arc_epilogue_uses): BLINK should be also
35550         restored when in interrupt.
35551         * config/arc/arc.md (simple_return): ARCv2 rtie instruction
35552         doesn't have delay slot.
35554 2017-03-27  Richard Biener  <rguenther@suse.de>
35556         PR ipa/79776
35557         * tree-ssa-structalias.c (associate_varinfo_to_alias): Skip
35558         inlined thunk clones.
35560 2017-03-27  Jakub Jelinek  <jakub@redhat.com>
35562         PR sanitizer/80168
35563         * asan.c (instrument_derefs): Copy over last operand from
35564         original COMPONENT_REF to the new COMPONENT_REF with
35565         DECL_BIT_FIELD_REPRESENTATIVE.
35566         * ubsan.c (instrument_object_size): Likewise.
35568 2017-03-27  Richard Biener  <rguenther@suse.de>
35570         PR tree-optimization/80170
35571         * tree-vect-data-refs.c (vect_compute_data_ref_alignment): Make
35572         sure DR/SCEV didnt fold in constants we do not see when looking
35573         at the reference base alignment.
35575 2017-03-27  Richard Biener  <rguenther@suse.de>
35577         PR middle-end/80171
35578         * gimple-fold.c (fold_ctor_reference): Properly guard against
35579         NULL return value from canonicalize_constructor_val.
35581 2017-03-25  Uros Bizjak  <ubizjak@gmail.com>
35583         PR target/80180
35584         * config/i386/i386.c (ix86_expand_builtin)
35585         <IX86_BUILTIN_RDSEED{16,32,64}_STEP>: Do not expand arg0 between
35586         flags reg setting and flags reg using instructions.
35587         <IX86_BUILTIN_RDRAND{16,32,64}_STEP>: Ditto.  Use non-flags reg
35588         clobbering instructions to zero extend op2.
35590 2017-03-25  Gerald Pfeifer  <gerald@pfeifer.com>
35592         * doc/install.texi (Configuration) <--with-aix-soname>:
35593         Update link to AIX ld.
35595 2017-03-25  Bernd Schmidt  <bschmidt@redhat.com>
35597         PR rtl-optimization/80160
35598         PR rtl-optimization/80159
35599         * lra-assigns.c (must_not_spill_p): Tighten new test to also take
35600         reg_alternate_class into account.
35602 2017-03-24  Vladimir Makarov  <vmakarov@redhat.com>
35604         PR target/80148
35605         * lra-assigns.c (assign_by_spills): Add spilled non-reload pseudos
35606         to consider in curr_insn_transform.
35608 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35610         * genrecog.c (validate_pattern): Add VEC_SELECT validation.
35611         * genmodes.c (emit_min_insn_modes_c): Call emit_mode_nunits
35612         and emit_mode_inner.
35614 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35616         * config/s390/s390-builtins.def: Add VXE builtins.  Add a flags
35617         argument to the overloaded builtin variants.  Use the new flag to
35618         deprecate certain builtin variants.
35619         * config/s390/s390-builtin-types.def: Add new builtin types.
35620         * config/s390/s390-builtins.h: Support new flags field for
35621         overloaded builtins.
35622         * config/s390/s390-c.c (OB_DEF_VAR): New flags field.
35623         (s390_macro_to_expand): Enable vector float data type.
35624         (s390_cpu_cpp_builtins_internal): Indicate support of the new
35625         builtins by incrementing the __VEC__ version number.
35626         (s390_expand_overloaded_builtin): Support expansion of vec_xl and
35627         vec_xst.
35628         (s390_resolve_overloaded_builtin): Emit error messages depending
35629         on the builtin flags.
35630         * config/s390/s390.c (s390_expand_builtin): Support additional
35631         flags argument.  Change error message to match the messages
35632         emitted in s390-c.c.
35633         * config/s390/s390.md: New UNSPEC_* constants.
35634         (op_type): Add new instruction types.
35635         * config/s390/vecintrin.h: Add new builtins and test data class
35636         constants.
35637         * config/s390/vx-builtins.md (V_HW_32_64): Add V4SF.
35638         (V_HW_4, VEC_HW, VECF_HW): New mode iterators.
35639         (VEC_INEXACT, VEC_NOINEXACT): New constants.
35640         ("vec_splats<mode>", "vec_insert<mode>", "vec_promote<mode>")
35641         ("vec_insert_and_zero<mode>", "vec_mergeh<mode>")
35642         ("vec_mergel<mode>"): V_HW -> VEC_HW.
35644         ("vlrlrv16qi", "vstrlrv16qi", "vbpermv16qi", "vec_msumv2di")
35645         ("vmslg", "*vftci<mode>_cconly", "vftci<mode>_intcconly")
35646         ("*vftci<mode>", "vftci<mode>_intcc", "vec_double_s64")
35647         ("vec_double_u64", "vfmin<mode>", "vfmax<mode>"): New definition.
35649         ("and_av2df3", "and_cv2df3", "vec_andc_av2df3")
35650         ("vec_andc_cv2df3", "xor_av2df3", "xor_cv2df3", "vec_nor_av2df3")
35651         ("vec_nor_cv2df3", "ior_av2df3", "ior_cv2df3", "vec_nabs")
35652         ("*vftcidb", "*vftcidb_cconly", "vftcidb"): Remove definition.
35654         ("vec_all_<fpcmpcc:code>v2df", "vec_any_<fpcmpcc:code>v2df")
35655         ("vec_scatter_elementv4si_DI", "vec_cmp<fpcmp:code>v2df")
35656         ("vec_di_to_df_s64", "vec_di_to_df_u64", "vec_df_to_di_u64")
35657         ("vfidb", "*vldeb", "*vledb", "*vec_cmp<insn_cmp>v2df_cconly")
35658         ("vec_cmpeqv2df_cc", "vec_cmpeqv2df_cc", "vec_cmphv2df_cc")
35659         ("vec_cmphev2df_cc", "*vec_cmpeqv2df_cc")
35660         ("*vec_cmphv2df_cc", "*vec_cmphev2df_cc"): Enable new modes as ...
35662         ("vec_all_<fpcmpcc:code><mode>", "vec_any_<fpcmpcc:code><mode>")
35663         ("vec_scatter_element<V_HW_4:mode>_DI")
35664         ("vec_cmp<fpcmp:code><mode>", "vcdgb", "vcdlgb", "vclgdb")
35665         ("vec_fpint<mode>", "vflls")
35666         ("vflrd", "*vec_cmp<insn_cmp><mode>_cconly", "vec_cmpeq<mode>_cc")
35667         ("vec_cmpeq<mode>_cc", "vec_cmph<mode>_cc", "vec_cmphe<mode>_cc")
35668         ("*vec_cmpeq<mode>_cc", "*vec_cmph<mode>_cc")
35669         ("*vec_cmphe<mode>_cc"): ... these.
35671         ("vec_ctd_s64", "vec_ctsl", "vec_ctul", "vec_st2f"): Use rounding
35672         mode constant instead of magic value.
35674 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35676         * config/s390/s390.c (s390_expand_vec_compare): Support other
35677         vector floating point modes than just V2DF.
35678         (s390_expand_vcond): Likewise.
35679         (s390_hard_regno_mode_ok): Allow SFmode values in VRs.
35680         (s390_cannot_change_mode_class): Prevent mode changes between TF
35681         and V1TF in vector registers.
35682         * config/s390/s390.md (DF, SF): New mode attributes.
35683         ("*cmp<mode>_ccs", "add<mode>3", "sub<mode>3", "mul<mode>3")
35684         ("fma<mode>4", "fms<mode>4", "div<mode>3", "*neg<mode>2"): Add
35685         SFmode support for VRs.
35686         * config/s390/vector.md (V_HW, V_HW2, VT_HW, ti*, nonvec): Add new
35687         vector fp modes.
35688         (VFT, VF_HW): New mode iterators.
35689         (vw, sdx): New mode attributes.
35690         ("addv2df3", "subv2df3", "mulv2df3", "divv2df3", "sqrtv2df2")
35691         ("fmav2df4","fmsv2df4", "negv2df2", "absv2df2", "*negabsv2df2")
35692         ("smaxv2df3", "sminv2df3", "*vec_cmp<VFCMP_HW_OP:code>v2df_nocc")
35693         ("vec_cmpuneqv2df", "vec_cmpltgtv2df", "vec_orderedv2df")
35694         ("vec_unorderedv2df"): Adjust the v2df only patterns to support
35695         also the new vector floating point modes.  Renaming to ...
35697         ("add<mode>3", "sub<mode>3", "mul<mode>3", "div<mode>3")
35698         ("sqrt<mode>2", "fma<mode>4", "fms<mode>4", "neg<mode>2")
35699         ("abs<mode>2", "negabs<mode>2", "smax<mode>3")
35700         ("smin<mode>3", "*vec_cmp<VFCMP_HW_OP:code><mode>_nocc")
35701         ("vec_cmpuneq<mode>", "vec_cmpltgt<mode>", "vec_ordered<mode>")
35702         ("vec_unordered<mode>"): ... these.
35704         ("neg_fma<mode>4", "neg_fms<mode>4", "*smax<mode>3_vxe")
35705         ("*smin<mode>3_vxe", "*sminv2df3_vx", "*vec_extendv4sf")
35706         ("*vec_extendv2df"): New insn definitions.
35708 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35710         * config/s390/s390.md ("*adddi3_sign", "*subdi3_sign", "mulditi3")
35711         ("mulditi3_2", "*muldi3_sign"): New patterns.
35712         ("muldi3", "*muldi3", "mulsi3", "*mulsi3"): Add an expander and
35713         rename the pattern definition.
35715 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35717         * config/s390/s390.md ("indirect_jump"): Turn insn definition into
35718         expander.
35719         ("*indirect_jump", "*indirect2_jump"): New pattern definitions.
35721 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35723         * config/s390/s390.c (s390_expand_vec_init): Use vllezl
35724         instruction if possible.
35725         * config/s390/vector.md (vec_halfnumelts): New mode
35726         attribute.
35727         ("*vec_vllezlf<mode>"): New pattern.
35729 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35731         * config/s390/vector.md ("popcountv16qi2", "popcountv8hi2")
35732         ("popcountv4si2", "popcountv2di2"): Rename to ...
35733         ("popcount<mode>2", "popcountv8hi2_vx", "popcountv4si2_vx")
35734         ("popcountv2di2_vx"): ... these and add !TARGET_VXE to the
35735         condition.
35736         ("popcount<mode>2_vxe"): New pattern.
35738 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35740         * common/config/s390/s390-common.c (processor_flags_table): Add
35741         arch12.
35742         * config.gcc: Add arch12.
35743         * config/s390/driver-native.c (s390_host_detect_local_cpu):
35744         Default to arch12 for unknown CPU model numbers.
35745         * config/s390/s390-builtins.def: Add B_VXE builtin flag.
35746         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Adjust
35747         PROCESSOR_max sanity check.
35748         * config/s390/s390-opts.h (enum processor_type): Add
35749         PROCESSOR_ARCH12.
35750         * config/s390/s390.c (processor_table): Add arch12.
35751         (s390_expand_builtin): Add check for B_VXE flag.
35752         (s390_issue_rate): Add PROCESSOR_ARCH12.
35753         (s390_get_sched_attrmask): Likewise.
35754         (s390_get_unit_mask): Likewise.
35755         (s390_sched_score): Enable z13 scheduling for arch12.
35756         (s390_sched_reorder): Likewise.
35757         (s390_sched_variable_issue): Likewise.
35758         * config/s390/s390.h (enum processor_flags): Add PF_ARCH12 and
35759         PF_VXE.
35760         (s390_tune_attr): Use z13 scheduling also for arch12.
35761         (TARGET_CPU_ARCH12, TARGET_CPU_ARCH12_P, TARGET_CPU_VXE)
35762         (TARGET_CPU_VXE_P, TARGET_ARCH12, TARGET_ARCH12_P, TARGET_VXE)
35763         (TARGET_VXE_P): New macros.
35764         * config/s390/s390.md: Add arch12 to cpu attribute.  Add arch12
35765         and vxe to cpu_facility.  Add arch12 and vxe to enabled attribute.
35766         * config/s390/s390.opt: Add arch12 as processor_type.
35768 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35770         * config/s390/s390.md
35771         ("fixuns_truncdddi2", "fixuns_trunctddi2")
35772         ("fixuns_trunc<BFP:mode><GPR:mode>2"): Merge into ...
35773         ("fixuns_trunc<FP:mode><GPR:mode>2"): New expander.
35775         ("fixuns_trunc<BFP:mode><GPR:mode>2", "fixuns_trunc<mode>si2"):
35776         Rename expanders to ...
35778         ("fixuns_trunc<BFP:mode><GPR:mode>2_emu")
35779         ("fixuns_truncdddi2_emu"): ... these.
35781         ("fixuns_trunc<mode>si2_emu"): New expander.
35783         ("*fixuns_truncdfdi2_z13"): Rename to ...
35784         ("*fixuns_truncdfdi2_vx"): ... this.
35786 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35788         * config/s390/2964.md: Remove the single element vector compare
35789         instructions which are no longer used.
35790         * config/s390/s390.c (s390_select_ccmode): Remove handling of
35791         vector CCmodes.
35792         (s390_canonicalize_comparison): Remove handling of DFmode
35793         compares.
35794         (s390_expand_vec_compare_scalar): Remove function.
35795         (s390_emit_compare): Don't call s390_expand_vec_compare_scalar.
35796         * config/s390/s390.md ("*vec_cmp<insn_cmp>df_cconly"): Remove
35797         pattern.
35798         ("*cmp<mode>_ccs"): Add wfcdb instruction.
35800 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35802         * config/s390/s390.md ("mov<mode>_64dfp" DD_DF): Use vleig for loading a
35803         FP zero.
35804         ("*mov<mode>_64" DD_DF): Remove the vector instructions. These
35805         will anyway by matched by mov<mode>_64dfp.
35807 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35809         * config/s390/s390.md ("mov<mode>" SD_SF): Change vleg/vsteg to
35810         vlef/vstef.  Add missing operand to vleif.
35812 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35814         * config/s390/s390.c (s390_expand_vec_init): Enable vector load
35815         pair for all vector types with 64 bit elements.
35816         * config/s390/vx-builtins.md (V_HW_64): Move mode iterator to ...
35817         * config/s390/vector.md (V_HW_64): ... here.
35818         (V_128_NOSINGLE): New mode iterator.
35819         ("vec_init<V_HW:mode>"): Use V_128 as mode iterator.
35820         ("*vec_splat<mode>"): Use V_128_NOSINGLE mode iterator.
35821         ("*vec_tf_to_v1tf", "*vec_ti_to_v1ti"): New pattern definitions.
35822         ("*vec_load_pairv2di"): Change to ...
35823         ("*vec_load_pair<mode>"): ... this one.
35825 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35827         * config/s390/constraints.md: Add comments.
35828         (jKK): Reject element sizes > 8 bytes.
35829         * config/s390/s390.c (s390_split_ok_p): Enable splitting also for
35830         s_operands.
35831         * config/s390/s390.md: Add the s_operand checks formerly in
35832         s390_split_ok_p to various splitters where they are still
35833         required.
35834         * config/s390/vector.md ("mov<mode>" V_128): Add GPR alternatives
35835         for 128 bit vectors.  Plus two splitters.
35837 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35839         * config/s390/s390.md: Rename the cpu facilty vec to vx throughout
35840         the file.
35842 2017-03-24  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
35844         PR target/79893
35845         * config/s390/s390-c.c (s390_adjust_builtin_arglist): Issue an
35846         error if the boundary argument is not constant.
35848 2017-03-24  Jakub Jelinek  <jakub@redhat.com>
35850         PR rtl-optimization/80112
35851         * loop-doloop.c (doloop_condition_get): Don't check condition
35852         if cmp isn't SET with IF_THEN_ELSE src.
35854 2017-03-24  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35856         PR tree-optimization/80158
35857         * gimple-ssa-strength-reduction.c (replace_mult_candidate): When
35858         replacing a candidate statement, also replace it for the
35859         candidate's alternate interpretation.
35860         (replace_rhs_if_not_dup): Likewise.
35861         (replace_one_candidate): Likewise.
35863 2017-03-24  Richard Biener  <rguenther@suse.de>
35865         PR tree-optimization/80167
35866         * graphite-isl-ast-to-gimple.c
35867         (translate_isl_ast_to_gimple::is_valid_rename): Handle default-defs
35868         properly.
35869         (translate_isl_ast_to_gimple::get_rename): Likewise.
35871 2017-03-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35873         * config/rs6000/rs6000.c (rs6000_option_override_internal): Change
35874         handling of certain combinations of target options, including the
35875         combinations -mpower8-vector vs. -mno-vsx, -mpower9-vector vs.
35876         -mno-power8-vector, and -mpower9_dform vs. -mno-power9-vector.
35878 2017-03-23  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
35880         PR target/71436
35881         * config/arm/arm.md (*load_multiple): Add reload_completed to
35882         matching condition.
35884 2017-03-23  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35885             Richard Biener  <rguenth@suse.de>
35887         PR tree-optimization/79908
35888         PR tree-optimization/80136
35889         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
35890         been cast away, gimplify_and_add suffices.
35892 2017-03-23  Markus Trippelsdorf  <markus@trippelsdorf.de>
35894         * tree-vrp.c (identify_jump_threads): Delete avail_exprs.
35896 2017-03-23  Richard Biener  <rguenther@suse.de>
35898         PR tree-optimization/80032
35899         * gimplify.c (gimple_push_cleanup): Forced unconditional
35900         cleanups still have to go to the conditional_cleanups
35901         sequence.
35903 2017-03-22  Jakub Jelinek  <jakub@redhat.com>
35905         PR tree-optimization/80072
35906         * tree-ssa-reassoc.c (struct operand_entry): Change id field type
35907         to unsigned int.
35908         (next_operand_entry_id): Change type to unsigned int.
35909         (sort_by_operand_rank): Make sure to return the right return value
35910         even if unsigned fields are bigger than INT_MAX.
35911         (struct oecount): Change cnt and id type to unsigned int.
35912         (oecount_hasher::equal): Formatting fix.
35913         (oecount_cmp): Make sure to return the right return value
35914         even if unsigned fields are bigger than INT_MAX.
35915         (undistribute_ops_list): Change next_oecount_id type to unsigned int.
35917         PR c++/80129
35918         * gimplify.c (gimplify_modify_expr_rhs) <case COND_EXPR>: Clear
35919         TREE_READONLY on result if writing it more than once.
35921         PR sanitizer/80110
35922         * doc/invoke.texi (-fsanitize=thread): Document that with
35923         -fnon-call-exceptions atomics are not able to throw
35924         exceptions.
35926         PR sanitizer/80110
35927         * tsan.c: Include tree-eh.h.
35928         (instrument_builtin_call): Call maybe_clean_eh_stmt or
35929         maybe_clean_or_replace_eh_stmt where needed.
35930         (instrument_memory_accesses): Add cfg_changed argument.
35931         Call gimple_purge_dead_eh_edges on each block and set *cfg_changed
35932         if it returned true.
35933         (tsan_pass): Adjust caller.  Return TODO_cleanup_cfg if cfg_changed.
35935         PR rtl-optimization/63191
35936         * config/i386/i386.c (ix86_delegitimize_address): Turn into small
35937         wrapper function, moved the whole old content into ...
35938         (ix86_delegitimize_address_1): ... this.  New inline function.
35939         (ix86_find_base_term): Use ix86_delegitimize_address_1 with
35940         true as last argument instead of ix86_delegitimize_address.
35942 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35944         * config/aarch64/aarch64.c (generic_branch_cost): Copy
35945         cortexa57_branch_cost.
35947 2017-03-22  Wilco Dijkstra  <wdijkstr@arm.com>
35949         * config/aarch64/aarch64.c (generic_tunings): Add AES fusion.
35951 2017-03-21  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
35953         PR target/80123
35954         * doc/md.texi (Constraints): Document wA constraint.
35955         * config/rs6000/constraints.md (wA): New.
35956         * config/rs6000/rs6000.c (rs6000_debug_reg_global): Add wA reg_class.
35957         (rs6000_init_hard_regno_mode_ok): Init wA constraint.
35958         * config/rs6000/rs6000.h (RS6000_CONSTRAINT_wA): New.
35959         * config/rs6000/vsx.md (vsx_splat_<mode>): Use wA constraint.
35961 2017-03-22  Cesar Philippidis  <cesar@codesourcery.com>
35963         PR c++/80029
35964         * gimplify.c (is_oacc_declared): New function.
35965         (oacc_default_clause): Use it to set default flags for acc declared
35966         variables inside parallel regions.
35967         (gimplify_scan_omp_clauses): Strip firstprivate pointers for acc
35968         declared variables.
35969         (gimplify_oacc_declare): Gimplify the declare clauses.  Add the
35970         declare attribute to any decl as necessary.
35972 2017-03-22  Thomas Preud'homme  <thomas.preudhomme@arm.com>
35974         PR target/80082
35975         * config/arm/arm-isa.h (isa_bit_lpae): New feature bit.
35976         (ISA_ARMv7ve): Add isa_bit_lpae to the definition.
35977         * config/arm/arm-protos.h (arm_arch7ve): Rename into ...
35978         (arm_arch_lpae): This.
35979         * config/arm/arm.c (arm_arch7ve): Rename into ...
35980         (arm_arch_lpae): This.  Define it in term of isa_bit_lpae.
35981         * config/arm/arm.h (TARGET_HAVE_LPAE): Redefine in term of
35982         arm_arch_lpae.
35984 2017-03-22  Martin Liska  <mliska@suse.cz>
35986         PR target/79906
35987         * config/rs6000/rs6000.c (rs6000_inner_target_options): Show
35988         error message instead of an ICE.
35990 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
35992         * doc/extend.texi (6.11 Additional Floating Types): Revise.
35994 2017-03-21  Kelvin Nilsen  <kelvin@gcc.gnu.org>
35996         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
35997         comments.
35998         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
35999         comments.
36001 2017-03-21  Martin Sebor  <msebor@redhat.com>
36003         * doc/extend.texi: Use "cannot" instead of "can't."
36004         * doc/hostconfig.texi: Same.
36005         * doc/install.texi: Same.
36006         * doc/invoke.texi: Same.
36007         * doc/loop.texi: Same.
36008         * doc/md.texi: Same.
36009         * doc/objc.texi: Same.
36010         * doc/rtl.texi: Same.
36011         * doc/tm.texi: Same.
36012         * doc/tm.texi.in: Same.
36013         * doc/trouble.texi: Same.
36015 2017-03-21  Alexandre Oliva  <aoliva@redhat.com>
36017         PR debug/63238
36018         * dwarf2out.c (struct checksum_attributes): Add at_alignment.
36019         (collect_checksum_attributes): Set it.
36020         (die_checksum_ordered): Use it.
36022 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36024         PR tree-optimization/79908
36025         * tree-stdarg.c (expand_ifn_va_arg_1): Revert the following
36026         change: For a VA_ARG whose LHS has been cast away, use
36027         force_gimple_operand to construct the side effects.
36029 2017-03-21  David Malcolm  <dmalcolm@redhat.com>
36031         PR translation/80001
36032         * omp-offload.c (oacc_loop_fixed_partitions): Make diagnostics
36033         more amenable to translation.
36034         (oacc_loop_auto_partitions): Likewise.
36036 2017-03-21  Marek Polacek  <polacek@redhat.com>
36037             Martin Sebor  <msebor@redhat.com>
36039         PR tree-optimization/80109
36040         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call get_range_info
36041         on INTEGRAL_TYPE_P.
36043 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36044             Segher Boessenkool  <segher@kernel.crashing.org>
36046         PR target/80125
36047         * combine.c (can_combine_p): Revert the 2017-03-20 change, only
36048         check reg_used_between_p between insn and one of succ or succ2
36049         depending on if succ is artificial insn not inserted into insn
36050         stream.
36052 2017-03-21  Martin Liska  <mliska@suse.cz>
36054         PR gcov-profile/80081
36055         * Makefile.in: Add gcov-dump and fix installation of gcov-tool.
36056         * doc/gcc.texi: Include gcov-dump stuff.
36057         * doc/gcov-dump.texi: New file.
36059 2017-03-21  Toma Tabacu  <toma.tabacu@imgtec.com>
36061         PR rtl-optimization/79150
36062         * config/mips/mips.c (mips_block_move_loop): Emit a NOP after the
36063         conditional jump, if the jump is the last insn of the loop.
36065 2017-03-21  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36066             Richard Biener  <rguenth@suse.de>
36068         PR tree-optimization/79908
36069         * tree-stdarg.c (expand_ifn_va_arg_1): For a VA_ARG whose LHS has
36070         been cast away, use force_gimple_operand to construct the side
36071         effects.
36073 2017-03-21  Martin Liska  <mliska@suse.cz>
36075         PR libfortran/79956
36076         * simplify-rtx.c (simplify_immed_subreg): Initialize a variable
36077         to NULL.
36079 2017-03-21  Brad Spengler <spender@grsecurity.net>
36081         PR plugins/80094
36082         * plugin.c (htab_hash_plugin): New function.
36083         (add_new_plugin): Use it and adjust.
36084         (parse_plugin_arg_opt): Adjust.
36085         (init_one_plugin): Likewise.
36087 2017-03-21  Richard Biener  <rguenther@suse.de>
36089         PR tree-optimization/80032
36090         * gimplify.c (gimple_push_cleanup): Add force_uncond parameter,
36091         if set force the cleanup to happen unconditionally.
36092         (gimplify_target_expr): Push inserted clobbers with force_uncond
36093         to avoid them being removed by control-dependent DCE.
36095 2017-03-21  Richard Biener  <rguenther@suse.de>
36097         PR tree-optimization/80122
36098         * tree-inline.c (copy_bb): Do not expans va-arg packs or
36099         va_arg_pack_len when the inlined call stmt requires pack
36100         expansion itself.
36101         * tree-inline.h (struct copy_body_data): Make call_stmt a gcall *.
36103 2017-03-21  Jakub Jelinek  <jakub@redhat.com>
36105         PR sanitizer/78158
36106         * tsan.c (instrument_builtin_call): If the memory model argument
36107         is not a constant, assume it is valid.
36109         PR c/67338
36110         * fold-const.c (round_up_loc): Negate divisor in unsigned type to
36111         avoid UB.
36113 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36115         PR rtl-optimization/79910
36116         * combine.c (can_combine_p): Do not allow combining an I0 or I1
36117         if its dest is used by an insn before I2 (other than the combined
36118         insns themselves, which are properly handled already).
36120 2017-03-20  Segher Boessenkool  <segher@kernel.crashing.org>
36122         Revert:
36123         2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36125         * combine.c (record_used_regs): New static function.
36126         (try_combine): Handle situations where there is an additional
36127         instruction between I2 and I3 which needs to have a LOG_LINK
36128         updated.
36130         Revert:
36131         2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36133         * combine.c (try_combine): Delete redundant i1 test.  Call
36134         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36136 2017-03-20  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
36138         PR target/80083
36139         * config/rs6000/rs6000.md (*movsi_internal1): Fix constraints for
36140         alternatives 13/14.
36142 2017-03-20  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36144         PR tree-optimization/80054
36145         * gimple-ssa-strength-reduction.c (all_phi_incrs_profitable): Fail
36146         the optimization if a PHI or any of its arguments is not dominated
36147         by the candidate's basis.  Use gphi* rather than gimple* as
36148         appropriate.
36149         (replace_profitable_candidates): Clean up a gimple* variable that
36150         should be a gphi* variable.
36152 2017-03-20  Martin Sebor  <msebor@redhat.com>
36154         PR c++/52477
36155         * doc/extend.texi (attribute constructor): Document present limitation.
36157 2017-03-20  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36159         PR target/79963
36160         * config/rs6000/altivec.h (vec_all_ne): Under __cplusplus__ and
36161         __POWER9_VECTOR__ #ifdef control, change template definition to
36162         use Power9-specific built-in function.
36163         (vec_any_eq): Likewise.
36164         * config/rs6000/vector.md (vector_ae_v2di_p): Change the flag used
36165         to control outcomes from this test.
36166         (vector_ae_<mode>p): For VEC_F modes, likewise.
36168 2017-03-20  Ian Lance Taylor  <iant@google.com>
36170         * config/i386/i386.c (ix86_function_regparm): Save an extra
36171         register for -fsplit-stack with DECL_STATIC_CHAIN.
36173 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36175         PR target/79912
36176         * config/riscv/riscv.c (riscv_preferred_reload_class): Remove.
36177         (TARGET_PREFERRED_RELOAD_CLASS): Likewise.
36179 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com>
36181         * config/riscv/riscv.c (riscv_print_operand): Use "fence
36182         iorw,ow".
36183         * config/riscv/sync.mc (mem_thread_fence_1): Use "fence
36184         iorw,iorw".
36186 2017-03-20  Marek Polacek  <polacek@redhat.com>
36188         PR sanitizer/80063
36189         * asan.c (DEF_SANITIZER_BUILTIN): Use do { } while (0).
36191 2017-03-20  Richard Biener  <rguenther@suse.de>
36193         PR tree-optimization/80113
36194         * graphite-isl-ast-to-gimple.c (copy_loop_phi_nodes): Do not
36195         allocate extra SSA name for PHI def.
36196         (add_close_phis_to_outer_loops): Likewise.
36197         (add_close_phis_to_merge_points): Likewise.
36198         (copy_loop_close_phi_args): Likewise.
36199         (copy_cond_phi_nodes): Likewise.
36201 2017-03-20  Martin Liska  <mliska@suse.cz>
36203         PR middle-end/79753
36204         * tree-chkp.c (chkp_build_returned_bound): Do not build
36205         returned bounds for a LHS that's not a BOUNDED_P type.
36207 2017-03-20  Martin Liska  <mliska@suse.cz>
36209         PR target/79769
36210         PR target/79770
36211         * tree-chkp.c (chkp_find_bounds_1): Handle REAL_CST,
36212         COMPLEX_CST and VECTOR_CST.
36214 2017-03-20  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36216         PR target/78857
36217         * config/s390/s390.md ("cmp<mode>_ccs_0"): Add a clobber of the
36218         target operand.  A new splitter adds the clobber statement in case
36219         the target operand is dead anyway.
36221 2017-03-19  Gerald Pfeifer  <gerald@pfeifer.com>
36223         * doc/install.texi (Specific) <sparc-*-linux*>: No longer refer
36224         to age-old versions of binutils and glibc.
36226 2017-03-18  Segher Boessenkool  <segher@kernel.crashing.org>
36228         * doc/contrib.texi (Contributors): Remove duplicate entry for myself.
36230 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36232         * doc/contrib.texi (Contributors): Add Segher Boessenkool.
36234 2017-03-18  Gerald Pfeifer  <gerald@pfeifer.com>
36236         * doc/install.texi (Specific) <arm-*-eabi>: Remove old
36237         requirement for binutils 2.13.
36239 2017-03-17  Jim Wilson  <jim.wilson@linaro.org>
36241         * combine.c (try_combine): Delete redundant i1 test.  Call
36242         prev_nonnote_nondebug_insn instead of prev_nonnote_insn.
36244 2017-03-17  Palmer Dabbelt  <palmer@dabbelt.com
36246         * doc/install.texi (Specific) <riscv32-*-elf>: Add riscv32-*-elf,
36247         riscv32-*-linux, riscv64-*-elf, riscv64-*-linux to the table of
36248         contents.
36249         <riscv64-*-elf>: Re-arrange section
36250         <riscv32-*-elf>: Add a note about requiring binutils 2.28.
36251         <riscv32-*-linux>: Likewise.
36252         <riscv64-*-elf>: Likewise
36253         <riscv64-*-linux>: Likewise.
36255 2017-03-17  Richard Earnshaw  <rearnsha@arm.com>
36257         PR target/80052
36258         * aarch64.opt(verbose-cost-dump): Fix typo.
36260 2017-03-17  Pat Haugen  <pthaugen@us.ibm.com>
36262         PR target/79951
36263         * config/rs6000/rs6000.md (copysign<mode>3_fcpsgn): Test
36264         for VECTOR_UNIT_VSX_P (<MODE>mode) too.
36266 2017-03-17  Bernd Schmidt  <bschmidt@redhat.com>
36268         * reload.c (find_reloads): When reloading a nonoffsettable address,
36269         use RELOAD_OTHER for it and its address reloads.
36271         PR rtl-optimization/79910
36272         * combine.c (record_used_regs): New static function.
36273         (try_combine): Handle situations where there is an additional
36274         instruction between I2 and I3 which needs to have a LOG_LINK
36275         updated.
36277 2017-03-17  Jeff Law  <law@redhat.com>
36279         PR tree-optimization/71437
36280         * tree-vrp.c (simplify_stmt_for_jump_threading): Lookup the
36281         conditional in the hash table first.
36282         (vrp_dom_walker::before_dom_children): Extract condition from
36283         ASSERT_EXPR.  Record condition, its inverion and any implied
36284         conditions as well.
36286 2017-03-17  Marek Polacek  <polacek@redhat.com>
36287             Markus Trippelsdorf  <markus@trippelsdorf.de>
36289         PR tree-optimization/80079
36290         * gimple-ssa-store-merging.c (class pass_store_merging): Initialize
36291         m_stores_head.
36293 2017-03-17  Richard Biener  <rguenther@suse.de>
36295         PR middle-end/80075
36296         * tree-eh.c (stmt_could_throw_1_p): Only handle gimple assigns.
36297         Properly verify the LHS before the RHS possibly claims to be
36298         handled.
36299         (stmt_could_throw_p): Hande gimple conds fully here.  Clobbers
36300         do not throw.
36302 2017-03-17  Martin Jambor  <mjambor@suse.cz>
36304         * doc/invoke.texi (Option Options): Include -fipa-vrp in the list.
36305         (List of -O2 options): Likewise.
36306         (-fipa-bit-cp): Replace "ipa" with "interprocedural."
36307         (-fipa-vrp) New.
36309 2017-03-17  Tom de Vries  <tom@codesourcery.com>
36311         * gcov-dump.c (print_usage): Print bug_report_url.
36313 2017-03-17  Richard Biener  <rguenther@suse.de>
36315         PR middle-end/80050
36316         * genmatch.c (parser::next): Remove pointless check for CPP_EOF.
36317         (parser::peek): Likewise.
36319 2017-03-17  Richard Biener  <rguenther@suse.de>
36321         PR tree-optimization/80048
36322         * sese.c (free_sese_info): Properly release rename_map and
36323         copied_bb_map elements.
36325 2017-03-16  Alexandre Oliva  <aoliva@redhat.com>
36327         * gimple-ssa-store-merging.c (struct imm_store_chain_info):
36328         Add linked-list forward and backlinks.  Insert on
36329         construction, remove on destruction.
36330         (class pass_store_merging): Add m_stores_head field.
36331         (pass_store_merging::terminate_and_process_all_chains):
36332         Iterate over m_stores_head list.
36333         (pass_store_merging::terminate_all_aliasing_chains):
36334         Likewise.
36335         (pass_store_merging::execute): Check for debug stmts first.
36336         Push new chains onto the m_stores_head stack.
36338 2017-03-16  Michael Meissner  <meissner@linux.vnet.ibm.com>
36340         PR target/71294
36341         * config/rs6000/vsx.md (vsx_splat_<mode>, VSX_D iterator): Allow a
36342         SPLAT operation on ISA 2.07 64-bit systems that have direct move,
36343         but no MTVSRDD support, by doing MTVSRD and XXPERMDI.
36345 2017-03-16  Jeff Law  <law@redhat.com>
36347         PR tree-optimization/71437
36348         * tree-ssa-dom.c (dom_opt_dom_walker): Remove thread_across_edge
36349         member function.  Implementation moved into after_dom_children
36350         member function and into the threader's thread_outgoing_edges
36351         function.
36352         (dom_opt_dom_walker::after_dom_children): Simplify by moving
36353         some code into new thread_outgoing_edges.
36354         * tree-ssa-threadedge.c (thread_across_edge): Make static and simplify
36355         definition.  Simplify marker handling (do it here).   Assume we always
36356         have the available expression and the const/copies tables.
36357         (thread_outgoing_edges): New function extracted from tree-ssa-dom.c
36358         and tree-vrp.c
36359         * tree-ssa-threadedge.h (thread_outgoing_edges): Declare.
36360         * tree-vrp.c (equiv_stack): No longer file scoped.
36361         (vrp_dom_walker): New class.
36362         (vrp_dom_walker::before_dom_children): New member function.
36363         (vrp_dom_walker::after_dom_children): Likewise.
36364         (identify_jump_threads):  Setup domwalker.  Use it rather than
36365         walking edges in a random order by hand.  Simplify setup/finalization.
36366         (finalize_jump_threads): Remove.
36367         (vrp_finalize): Do not call identify_jump_threads here.
36368         (execute_vrp): Do it here instead and call thread_through_all_blocks
36369         here too.
36371         PR tree-optimization/71437
36372         * tree-ssa-dom.c (pfn_simplify): Add basic_block argument.  All
36373         callers changed.
36374         (simplify_stmt_for_jump_threading): Add basic_block argument.  All
36375         callers changed.
36376         (lhs_of_dominating_assert): Moved from here into tree-vrp.c.
36377         (dom_opt_dom_walker::thread_across_edge): Remove
36378         handle_dominating_asserts argument.  All callers changed.
36379         (record_temporary_equivalences_from_stmts_at_dest): Corresponding
36380         changes.  Remove calls to lhs_of_dominating_assert.  Other
36381         uses of handle_dominating_asserts turn into unconditional code
36382         (simplify_control_stmt_condition_1): Likewise.
36383         (simplify_control_stmt_condition): Likewise.
36384         (thread_through_normal_block, thread_across_edge): Likewise.
36385         * tree-ssa-threadedge.h (thread_across_edge): Corresponding changes.
36386         * tree-vrp.c (lhs_of_dominating_assert): Move here.  Return original
36387         object if it is not an SSA_NAME.
36388         (simplify_stmt_for_jump_threading): Call lhs_of_dominating_assert
36389         before calling into the VRP specific simplifiers.
36390         (identify_jump_threads): Remove handle_dominating_asserts
36391         argument.
36393 2017-03-16  Jakub Jelinek  <jakub@redhat.com>
36395         PR fortran/79886
36396         * tree-diagnostic.c (default_tree_printer): No longer static.
36397         * tree-diagnostic.h (default_tree_printer): New prototype.
36399 2017-03-16  Tamar Christina  <tamar.christina@arm.com>
36401         * config/aarch64/aarch64-simd.md (*aarch64_simd_mov<mode>)
36402         Change ins into fmov.
36404 2017-03-16  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36406         * config/aarch64/iterators.md (h_con): Return "x" for V4HF and V8HF.
36407         * config/aarch64/aarch64-simd.md (*aarch64_fma4_elt_from_dup<mode>):
36408         Use h_con constraint for operand 1.
36409         (*aarch64_fnma4_elt_from_dup<mode>): Likewise.
36410         (*aarch64_mulx_elt_from_dup<mode>): Likewise for operand 2.
36412 2017-03-15  Jeff Law  <law@redhat.com>
36414         PR tree-optimization/71437
36415         * tree-ssa-dom.c (derive_equivalences_from_bit_ior): New function.
36416         (record_temporary_equivalences): Use it.
36418         PR tree-optimization/71437
36419         * tree-ssa-dom.c (struct cond_equivalence): Moved from here into
36420         tree-ssa-scopedtables.
36421         (lookup_avail_expr, build_and_record_new_cond): Likewise.
36422         (record_conditions, record_cond, vuse_eq): Likewise.
36423         (record_edge_info): Adjust to API tweak of record_conditions.
36424         (simplify_stmt_for_jump_threading): Similarly for lookup_avail_expr.
36425         (record_temporary_equivalences, optimize_stmt): Likewise.
36426         (eliminate_redundant_computations): Likewise.
36427         (record_equivalences_from_stmt): Likewise.
36428         * tree-ssa-scopedtables.c: Include options.h and params.h.
36429         (vuse_eq): New function, moved from tree-ssa-dom.c
36430         (build_and_record_new_cond): Likewise.
36431         (record_conditions): Likewise.  Accept vector of conditions rather
36432         than edge_equivalence structure for first argument.
36433         for the first argument.
36434         (avail_exprs_stack::lookup_avail_expr): New member function, moved
36435         from tree-ssa-dom.c.
36436         (avail_exprs_stack::record_cond): Likewise.
36437         * tree-ssa-scopedtables.h (struct cond_equivalence): Moved here
36438         from tree-ssa-dom.c.
36439         (avail_exprs_stack): Add new member functions lookup_avail_expr
36440         and record_cond.
36441         (record_conditions): Declare.
36443 2017-03-15  Vladimir Makarov  <vmakarov@redhat.com>
36445         PR target/80017
36446         * lra-constraints.c (process_alt_operands): Increase reject for
36447         reloading an input/output operand.
36449 2017-03-15  Michael Meissner  <meissner@linux.vnet.ibm.com>
36451         PR target/79038
36452         * config/rs6000/rs6000.md (float<QHI:mode><IEEE128:mode>2): Define
36453         insns to convert from signed/unsigned char/short to IEEE 128-bit
36454         floating point.
36455         (floatuns<QHI:mode><IEEE128:mode>2): Likewise.
36457 2017-03-15  Uros Bizjak  <ubizjak@gmail.com>
36459         PR target/80019
36460         * config/i386/i386.c (ix86_vector_duplicate_value): Create
36461         subreg of inner mode for values already in registers.
36463 2017-03-15  Bernd Schmidt  <bschmidt@redhat.com>
36465         * config/c6x/c6x.c (hwloop_optimize): Handle case where the old
36466         iteration reg is used after the loop.
36468 2017-03-14  Martin Sebor  <msebor@redhat.com>
36470         PR tree-optimization/79800
36471         * gimple-ssa-sprintf.c (format_floating: Add argument.  Handle
36472         precision in negative-positive range.
36473         (format_floating): Call non-const overload with adjusted precision.
36475 2017-03-14  Michael Meissner  <meissner@linux.vnet.ibm.com>
36477         PR target/79947
36478         * config/rs6000/rs6000.h (TARGET_FRSQRTES): Add check for
36479         -mpowerpc-gfxopt.
36481 2017-03-14  Martin Sebor  <msebor@redhat.com>
36483         PR middle-end/80020
36484         * builtin-attrs.def (ATTR_ALLOC_SIZE_2_NOTHROW_LIST): New macro.
36485         * builtins.def (aligned_alloc): Use it.
36487         PR c/79936
36488         * Makefile.in (GTFILES): Add calls.c.
36489         * calls.c: Include "gt-calls.h".
36491 2017-03-14  Bernd Schmidt  <bschmidt@redhat.com>
36493         PR rtl-optimization/79728
36494         * regs.h (struct target_regs): New field
36495         x_contains_allocatable_regs_of_mode.
36496         (contains_allocatable_regs_of_mode): New macro.
36497         * reginfo.c (init_reg_sets_1): Initialize it, and change
36498         contains_reg_of_mode so it includes global regs as well.
36499         * reload.c (push_reload): Use contains_allocatable_regs_of_mode
36500         rather than contains_regs_of_mode.
36502 2017-03-14  Martin Liska  <mliska@suse.cz>
36504         * doc/invoke.texi: Document options that can't be combined with
36505         -fcheck-pointer-bounds.
36507 2017-03-14  Martin Liska  <mliska@suse.cz>
36509         PR middle-end/79831
36510         * doc/invoke.texi (-Wchkp): Document the option.
36512 2017-03-14  Martin Liska  <mliska@suse.cz>
36514         * Makefile.in: Install gcov-dump.
36516 2017-03-14  Martin Liska  <mliska@suse.cz>
36518         * multiple_target.c (expand_target_clones): Bail out for
36519         an invalid attribute.
36521 2017-03-14  Richard Biener  <rguenther@suse.de>
36523         * alias.c (struct alias_set_entry): Pack properly.
36524         * cfgloop.h (struct loop): Likewise.
36525         * cse.c (struct set): Likewise.
36526         * ipa-utils.c (struct searchc_env): Likewise.
36527         * loop-invariant.c (struct invariant): Likewise.
36528         * lra-remat.c (struct cand): Likewise.
36529         * recog.c (struct change_t): Likewise.
36530         * rtl.h (struct address_info): Likewise.
36531         * symbol-summary.h (function_summary): Likewise.
36532         * tree-loop-distribution.c (struct partition): Likewise.
36533         * tree-object-size.c (struct object_size_info): Likewise.
36534         * tree-ssa-loop-ivopts.c (struct cost_pair): Likewise.
36535         * tree-ssa-threadupdate.c (struct ssa_local_info_t): Likewise.
36536         * tree-vect-data-refs.c (struct _vect_peel_info): Likewise.
36537         * tree-vect-slp.c (struct _slp_oprnd_info): Likewise.
36538         * tree-vect-stmts.c (struct simd_call_arg_info): Likewise.
36539         * tree-vectorizer.h (struct _loop_vec_info): Likewise.
36540         (struct _stmt_vec_info): Likewise.
36542 2017-03-14  Martin Liska  <mliska@suse.cz>
36544         PR target/79892
36545         * multiple_target.c (create_dispatcher_calls): Check that
36546         a target can create a function dispatcher.
36548 2017-03-14  Martin Liska  <mliska@suse.cz>
36550         PR lto/66295
36551         * multiple_target.c (expand_target_clones): Drop local.local
36552         flag for default implementation.
36554 2017-03-14  Richard Biener  <rguenther@suse.de>
36556         PR tree-optimization/80030
36557         * tree-vect-stmts.c (vectorizable_store): Plug memleak.
36559 2017-03-13  Kito Cheng  <kito.cheng@gmail.com>
36561         * config/riscv/riscv.c (riscv_emit_float_compare>: Use
36562         gcc_fallthrough() instead of __attribute__((fallthrough));
36564 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36566         * doc/gcc.texi: Remove "up" link to (DIR).
36567         * doc/gccint.texi: Ditto.
36569 2017-03-13  Gerald Pfeifer  <gerald@pfeifer.com>
36571         * doc/install.texi (Specific) <avr>: Remove reference to
36572         binutils 2.13.
36574 2017-03-13  Jeff Law  <law@redhat.com>
36576         * config/riscv/riscv.c (riscv_emit_float_compare): Use fallthru
36577         attribute rather than comments.
36579         * config/pdp11/pdp11.md (movmemhi): Adjust operand numbers to
36580         match_scratch operand is highest.
36582 2017-03-13  Martin Liska  <mliska@suse.cz>
36584         PR middle-end/78339
36585         * ipa-pure-const.c (warn_function_noreturn): If the declarations
36586         is a CHKP clone, use original declaration.
36588 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36590         * config/arc/arc.c (arc_init): Use multiplier whenever we have it.
36591         (arc_conditional_register_usage): Use a different allocation order
36592         when optimizing for size.
36593         * common/config/arc/arc-common.c (arc_option_optimization_table):
36594         Section anchors default on when optimizing for size.
36596 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36598         * config/arc/arc.md (*tst_bitfield_tst): Fix pattern.
36600 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36602         * config/arc/arc.c (arc_output_addsi): Emit code density adds.
36603         * config/arc/arc.md (cpu_facility): Add cd variant.
36604         (*movqi_insn): Add code density variant.
36605         (*movhi_insn): Likewise.
36606         (*movqi_insn): Likewise.
36607         (*addsi3_mixed): Likewise.
36608         (subsi3_insn): Likewise.
36610 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36612         * config/arc/arc.md (movsi_cond_exec): Update constraint.
36614 2017-03-13  Claudiu Zissulescu  <claziss@synopsys.com>
36616         * config/arc/arc.c (arc_legitimize_pic_address): Handle PIC
36617         expressions with MINUS and UNARY ops.
36619 2017-03-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36621         PR target/79911
36622         * config/arm/neon.md (vec_sel_widen_ssum_lo<VQI:mode><VW:mode>3):
36623         Rename to...
36624         (vec_sel_widen_ssum_lo<mode><V_half>3): ... This. Avoid mismatch
36625         between vec_select and vector argument.
36626         (vec_sel_widen_ssum_hi<VQI:mode><VW:mode>3): Rename to...
36627         (vec_sel_widen_ssum_hi<mode><V_half>3): ... This. Likewise.
36628         (vec_sel_widen_usum_lo<VQI:mode><VW:mode>3): Rename to...
36629         (vec_sel_widen_usum_lo<mode><V_half>3): ... This.
36630         (vec_sel_widen_usum_hi<VQI:mode><VW:mode>3): Rename to...
36631         (vec_sel_widen_usum_hi<mode><V_half>3): ... This.
36633 2017-03-13  Richard Biener  <rguenther@suse.de>
36635         PR other/79991
36636         * params.def (vect-max-peeling-for-alignment): Fix typo.
36638 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36640         * doc/install.texi (Specific) <mips-*-*>: Remove description of
36641         issue that only occurred with binutils below 2.18.
36643 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36645         * doc/install.texi (Specific) <cris-axis-elf>: No longer
36646         refer to binutils 2.11/2.12 minimum.
36648 2017-03-12  Gerald Pfeifer  <gerald@pfeifer.com>
36650         * doc/install.texi (Specific) <powerpc-*-*>: Remove link to
36651         ftp.kernel.org and simplify binutils requirement.
36653 2017-03-11  Gerald Pfeifer  <gerald@pfeifer.com>
36655         * doc/invoke.texi (Warning Options): Fix spelling of link-time
36656         optimization.
36657         (Optimize Options): Ditto.  Also remove redundancy.
36659 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36661         PR translation/79848
36662         * ipa-devirt.c (warn_types_mismatch): Simplify uses of "%<%s%>" to
36663         "%qs".
36664         * ipa-pure-const.c (suggest_attribute): Likewise.  Convert _
36665         to G_ to avoid double translation.
36667 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36669         PR translation/79923
36670         * auto-profile.c (get_combined_location): Convert leading
36671         character of diagnostics to lower case and remove trailing period.
36672         (read_profile): Likewise for various diagnostics.
36673         * config/arm/arm.c (arm_option_override): Remove trailing period
36674         from various diagnostics.
36675         * config/msp430/msp430.c (msp430_expand_delay_cycles): Likewise.
36676         (msp430_expand_delay_cycles): Likewise.
36678 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36680         PR target/79925
36681         * config/aarch64/aarch64.c (aarch64_validate_mcpu): Quote the
36682         full command-line argument, rather than just "str".
36683         (aarch64_validate_march): Likewise.
36684         (aarch64_validate_mtune): Likewise.
36686 2017-03-10  Bernd Schmidt  <bschmidt@redhat.com>
36688         PR rtl-optimization/78911
36689         * lra-assigns.c (must_not_spill_p): New function.
36690         (spill_for): Use it.
36692 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36694         PR tree-optimization/79981
36695         * tree-vrp.c (extract_range_basic): Handle IMAGPART_EXPR of
36696         ATOMIC_COMPARE_EXCHANGE ifn result.
36697         (stmt_interesting_for_vrp, vrp_visit_stmt): Handle
36698         IFN_ATOMIC_COMPARE_EXCHANGE.
36700 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36702         PR driver/79875
36703         * opts.c (parse_sanitizer_options): Add missing question mark to
36704         "did you mean" message.
36706 2017-03-10  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36708         * config/rs6000/rs6000-builtin.def (VMULEUB_UNS): Remove orphaned
36709         built-in.
36710         (VMULEUH_UNS): Likewise.
36711         (VMULOUB_UNS): Likewise.
36712         (VMULOUH_UNS): Likewise.
36713         * config/rs6000/rs6000.c (builtin_function_type): Remove
36714         references to ALTIVEC_BUILTIN_VMUL[EO]U[BH]_UNS.
36716 2017-03-10  David Malcolm  <dmalcolm@redhat.com>
36718         PR bootstrap/79952
36719         * read-rtl-function.c (function_reader::read_rtx_operand): Update
36720         x with result of extra_parsing_for_operand_code_0.
36721         (function_reader::extra_parsing_for_operand_code_0): Convert
36722         return type from void to rtx, returning x.  When reading
36723         SYMBOL_REF with SYMBOL_FLAG_HAS_BLOCK_INFO, reallocate x to the
36724         larger size containing struct block_symbol.
36726 2017-03-10  Segher Boessenkool  <segher@kernel.crashing.org>
36728         * config/rs6000/rs6000.c (rs6000_option_override_internal): Disallow
36729         -mfloat128-hardware without -m64.
36731 2017-03-10  Will Schmidt <will_schmidt@vnet.ibm.com>
36733         PR target/79941
36734         * config/rs6000/rs6000.c (builtin_function_type): Add VMUL*U[HB]
36735         entries to the case statement that marks unsigned arguments to
36736         overloaded functions.
36738 2017-03-10  Kelvin Nilsen  <kelvin@gcc.gnu.org>
36740         * config/rs6000/rs6000.c (rs6000_option_override_internal): Fix
36741         two typographic errors in the handling of TARGET_UPPER_REGS_DI.
36743 2017-03-10  Pat Haugen  <pthaugen@us.ibm.com>
36745         PR target/79907
36746         * config/rs6000/rs6000.c (rs6000_init_hard_regno_mode_ok): Test
36747         TARGET_UPPER_REGS_DI when setting 'wi' constraint regclass.
36749 2017-03-10  Martin Liska  <mliska@suse.cz>
36751         PR target/65705
36752         PR target/69804
36753         * toplev.c (process_options): Enable MPX with LSAN and UBSAN.
36754         * tree-chkp.c (chkp_walk_pointer_assignments): Verify that
36755         FIELD != NULL.
36757 2017-03-10  Olivier Hainque  <hainque@adacore.com>
36759         * tree-switch-conversion (array_value_type): Start by resetting
36760         candidate type to it's main variant.
36762 2017-03-10  Jakub Jelinek  <jakub@redhat.com>
36764         PR rtl-optimization/79909
36765         * combine.c (try_combine): Use simplify_replace_rtx on individual
36766         CALL_INSN_FUNCTION_USAGE elements instead of replace_rtx on copy_rtx
36767         of the whole CALL_INSN_FUNCTION_USAGE.
36769         PR tree-optimization/79972
36770         * gimple-ssa-warn-alloca.c (alloca_call_type): Only call
36771         get_range_info on SSA_NAMEs.  Formatting fixes.
36773 2017-03-10  Richard Biener  <rguenther@suse.de>
36774             Jakub Jelinek  <jakub@redhat.com>
36776         PR tree-optimization/77975
36777         * tree-ssa-loop-niter.c (get_base_for): Allow phi argument from latch
36778         edge to be constant.
36779         (get_val_for): For constant x return it.  Formatting fix.
36780         (loop_niter_by_eval): Avoid pointless looping if the next iteration
36781         would use the same bases as the current one.
36783 2017-03-09  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
36785         * config/rs6000/rs6000.c (rs6000_gen_le_vsx_permute): Use rotate
36786         instead of vec_select for V1TImode.
36787         * conifg/rs6000/vsx.md (VSX_LE): Remove mode iterator that is no
36788         longer needed.
36789         (VSX_LE_128): Add V1TI to this mode iterator.
36790         (*vsx_le_perm_load_<mode>): Change to use VSX_D mode iterator.
36791         (*vsx_le_perm_store_<mode>): Likewise.
36792         (pre-reload splitter for VSX stores): Likewise.
36793         (post-reload splitter for VSX stores): Likewise.
36794         (*vsx_xxpermdi2_le_<mode>): Likewise.
36795         (*vsx_lxvd2x2_le_<mode>): Likewise.
36796         (*vsx_stxvd2x2_le_<mode>): Likewise.
36798 2017-03-09  Michael Eager  <eager@eagercon.com>
36800         Correct failures with --enable-checking=yes,rtl.
36802         * config/microblaze/microblaze.c (microblaze_expand_shift):
36803         Replace GET_CODE test with CONST_INT_P and INTVAL test with
36804         test for const0_rtx.
36805         * config/microblaze/microblaze.md (ashlsi3_byone, ashrsi3_byone,
36806         lshrsi3_byone): Replace INTVAL with test for const1_rtx.
36808 2017-03-09  Richard Biener  <rguenther@suse.de>
36810         PR tree-optimization/79977
36811         * graphite-scop-detection.c (scop_detection::merge_sese):
36812         Handle the case of extra exits to blocks dominating the entry.
36814 2017-03-09  Toma Tabacu  <toma.tabacu@imgtec.com>
36816         * doc/sourcebuild.texi (Effective-Target Keywords, Other attributes):
36817         Document rdynamic.
36819 2017-03-09  Vladimir Makarov  <vmakarov@redhat.com>
36821         PR rtl-optimization/79949
36822         * lra-constraints.c (process_alt_operands): Check memory when
36823         trying to predict a cycle.  Print about the overall increase.
36825 2017-03-09  Richard Biener  <rguenther@suse.de>
36827         PR middle-end/79971
36828         * gimple-expr.c (useless_type_conversion_p): Preserve
36829         TYPE_SATURATING for fixed-point types.
36831 2017-03-09  Richard Biener  <rguenther@suse.de>
36833         PR ipa/79970
36834         * ipa-prop.c (ipa_modify_formal_parameters): Avoid changing
36835         alignment of BLKmode params.
36837 2017-03-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
36839         PR target/79913
36840         * config/aarch64/iterators.md (VALL_F16_NO_V2Q): New mode iterator.
36841         (VALL_NO_V2Q): Likewise.
36842         (VDQF_DF): Delete.
36843         * config/aarch64/aarch64-simd.md
36844         (aarch64_dup_lane_<vswap_width_name><mode>): Use VALL_F16_NO_V2Q
36845         iterator.
36846         (*aarch64_simd_vec_copy_lane_<vswap_width_name><mode>): Use
36847         VALL_NO_V2Q mode iterator.
36848         (*aarch64_vgetfmulx<mode>): Use VDQF iterator.
36850 2017-03-09  Martin Liska  <mliska@suse.cz>
36852         PR tree-optimization/79631
36853         * tree-chkp-opt.c (chkp_is_constant_addr): Call
36854         tree_int_cst_sign_bit just for INTEGER constants.
36856 2017-03-09  Martin Liska  <mliska@suse.cz>
36858         PR target/65705
36859         PR target/69804
36860         * toplev.c (process_options): Disable -fcheck-pointer-bounds with
36861         sanitizers.
36863 2017-03-09  Marek Polacek  <polacek@redhat.com>
36865         PR c++/79672
36866         * tree.c (inchash::add_expr): Handle TREE_VEC.
36868 2017-03-09  Martin Liska  <mliska@suse.cz>
36870         PR ipa/79764
36871         (chkp_narrow_size_and_offset): New function.
36872         (chkp_parse_array_and_component_ref): Support BIT_FIELD_REF.
36873         (void chkp_parse_bit_field_ref): New function.
36874         (chkp_make_addressed_object_bounds): Add case for BIT_FIELD_REF.
36875         (chkp_process_stmt): Use chkp_parse_bit_field_ref.
36877 2017-03-09  Martin Liska  <mliska@suse.cz>
36879         PR ipa/79761
36880         * tree-chkp.c (chkp_get_bound_for_parm): Get bounds for a param.
36881         (chkp_find_bounds_1): Remove gcc_unreachable.
36883 2017-03-09  Jakub Jelinek  <jakub@redhat.com>
36885         PR sanitizer/79944
36886         * asan.c (get_mem_refs_of_builtin_call): For BUILT_IN_ATOMIC* and
36887         BUILT_IN_SYNC*, determine the access type from the size suffix and
36888         always build a MEM_REF with that type.  Handle forgotten
36889         BUILT_IN_SYNC_FETCH_AND_NAND_16 and BUILT_IN_SYNC_NAND_AND_FETCH_16.
36891         PR target/79932
36892         * config/i386/avx512vlintrin.h (_mm256_cmpge_epi32_mask,
36893         _mm256_cmpge_epi64_mask, _mm256_cmpge_epu32_mask,
36894         _mm256_cmpge_epu64_mask, _mm256_cmple_epi32_mask,
36895         _mm256_cmple_epi64_mask, _mm256_cmple_epu32_mask,
36896         _mm256_cmple_epu64_mask, _mm256_cmplt_epi32_mask,
36897         _mm256_cmplt_epi64_mask, _mm256_cmplt_epu32_mask,
36898         _mm256_cmplt_epu64_mask, _mm256_cmpneq_epi32_mask,
36899         _mm256_cmpneq_epi64_mask, _mm256_cmpneq_epu32_mask,
36900         _mm256_cmpneq_epu64_mask, _mm256_mask_cmpge_epi32_mask,
36901         _mm256_mask_cmpge_epi64_mask, _mm256_mask_cmpge_epu32_mask,
36902         _mm256_mask_cmpge_epu64_mask, _mm256_mask_cmple_epi32_mask,
36903         _mm256_mask_cmple_epi64_mask, _mm256_mask_cmple_epu32_mask,
36904         _mm256_mask_cmple_epu64_mask, _mm256_mask_cmplt_epi32_mask,
36905         _mm256_mask_cmplt_epi64_mask, _mm256_mask_cmplt_epu32_mask,
36906         _mm256_mask_cmplt_epu64_mask, _mm256_mask_cmpneq_epi32_mask,
36907         _mm256_mask_cmpneq_epi64_mask, _mm256_mask_cmpneq_epu32_mask,
36908         _mm256_mask_cmpneq_epu64_mask, _mm_cmpge_epi32_mask,
36909         _mm_cmpge_epi64_mask, _mm_cmpge_epu32_mask, _mm_cmpge_epu64_mask,
36910         _mm_cmple_epi32_mask, _mm_cmple_epi64_mask, _mm_cmple_epu32_mask,
36911         _mm_cmple_epu64_mask, _mm_cmplt_epi32_mask, _mm_cmplt_epi64_mask,
36912         _mm_cmplt_epu32_mask, _mm_cmplt_epu64_mask, _mm_cmpneq_epi32_mask,
36913         _mm_cmpneq_epi64_mask, _mm_cmpneq_epu32_mask, _mm_cmpneq_epu64_mask,
36914         _mm_mask_cmpge_epi32_mask, _mm_mask_cmpge_epi64_mask,
36915         _mm_mask_cmpge_epu32_mask, _mm_mask_cmpge_epu64_mask,
36916         _mm_mask_cmple_epi32_mask, _mm_mask_cmple_epi64_mask,
36917         _mm_mask_cmple_epu32_mask, _mm_mask_cmple_epu64_mask,
36918         _mm_mask_cmplt_epi32_mask, _mm_mask_cmplt_epi64_mask,
36919         _mm_mask_cmplt_epu32_mask, _mm_mask_cmplt_epu64_mask,
36920         _mm_mask_cmpneq_epi32_mask, _mm_mask_cmpneq_epi64_mask,
36921         _mm_mask_cmpneq_epu32_mask, _mm_mask_cmpneq_epu64_mask): Move
36922         definitions outside of __OPTIMIZE__ guarded section.
36924         PR target/79932
36925         * config/i386/avx512bwintrin.h (_mm512_packs_epi32,
36926         _mm512_maskz_packs_epi32, _mm512_mask_packs_epi32,
36927         _mm512_packus_epi32, _mm512_maskz_packus_epi32,
36928         _mm512_mask_packus_epi32): Move definitions outside of __OPTIMIZE__
36929         guarded section.
36931 2017-03-09  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
36933         * config/s390/vx-builtins.md ("vfee<mode>", "vfeez<mode>")
36934         ("vfenez<mode>"): Add missing constraints.
36936 2017-03-08  Martin Sebor  <msebor@redhat.com>
36938         PR target/79928
36939         * config/nds32/nds32.c (nds32_option_override):
36940         Fix misspelled diagnostic.
36942 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36944         PR c/79940
36945         * gimplify.c (gimplify_omp_for): Replace index var in outer
36946         taskloop statement with an artificial variable and add
36947         OMP_CLAUSE_PRIVATE clause for it.
36949 2017-03-08  Richard Biener  <rguenther@suse.de>
36951         PR tree-optimization/79955
36952         * tree-ssa-uninit.c (warn_uninitialized_vars): Do not warn
36953         for accesses that are completely outside of the variable.
36955 2017-03-08  Andrew Haley  <aph@redhat.com>
36957         PR tree-optimization/79943
36958         * tree-ssa-loop-split.c (compute_new_first_bound): When
36959         calculating the new upper bound, (END-BEG) should be added, not
36960         subtracted.
36962 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36964         * config/avr/avr.md (setmemhi): Make sure match_dup
36965         operand number comes before match_scratch.
36967 2017-03-08  Richard Biener  <rguenther@suse.de>
36969         PR tree-optimization/79920
36970         * tree-vect-slp.c (vect_create_mask_and_perm): Remove and inline
36971         with ncopies == 1 to ...
36972         (vect_transform_slp_perm_load): ... here.  Properly compute
36973         all element loads by iterating VF times over the group.  Do
36974         not handle ncopies (computed in a broken way) in
36975         vect_create_mask_and_perm.
36977 2017-03-08  Jakub Jelinek  <jakub@redhat.com>
36979         PR sanitizer/79904
36980         * internal-fn.c (expand_vector_ubsan_overflow): If arg0 or arg1
36981         is a uniform vector, use uniform_vector_p return value instead of
36982         building ARRAY_REF on folded VIEW_CONVERT_EXPR to array type.
36984 2017-03-07  Marek Polacek  <polacek@redhat.com>
36986         PR middle-end/79809
36987         * gimple-ssa-warn-alloca.c (pass_walloca::gate): Use HOST_WIDE_INT.
36988         (alloca_call_type): Likewise.
36990 2017-03-07  Martin Liska  <mliska@suse.cz>
36992         * gcov.c (process_args): Put comment to correct location.
36994 2017-03-07  Martin Liska  <mliska@suse.cz>
36996         PR middle-end/68270
36997         * tree-chkp.c (chkp_may_narrow_to_field): Add new argument ref.
36998         Use array_at_struct_end_p instead of DECL_CHAIN (field).
36999         (chkp_narrow_bounds_for_field): Likewise.
37000         (chkp_parse_array_and_component_ref): Pass one more argument to
37001         call.
37003 2017-03-07  Richard Biener  <rguenther@suse.de>
37005         * tree-vect-loop-manip.c (slpeel_add_loop_guard): Preserve
37006         preheaders.
37008 2017-03-07  Segher Boessenkool  <segher@kernel.crashing.org>
37010         * config/i386/i386.c (ix86_local_alignment): Align most aggregates
37011         of 16 bytes and more to 16 bytes, not those of 16 bits and more.
37013 2017-03-07  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37015         PR c/79855
37016         * params.def (PARAM_STORE_MERGING_ALLOW_UNALIGNED): Add full stop
37017         to end of description.
37018         (PARAM_MAX_STORES_TO_MERGE): Likewise.
37020 2017-03-07  Jakub Jelinek  <jakub@redhat.com>
37022         PR rtl-optimization/79901
37023         * config/i386/sse.md (*avx512bw_<code><mode>3<mask_name>): Renamed to
37024         ...
37025         (*avx512f_<code><mode>3<mask_name>): ... this.
37026         (<code><mode>3 with maxmin code iterator): Use VI8_AVX2_AVX512F
37027         iterator instead of VI8_AVX2_AVX512BW.
37029         PR rtl-optimization/79901
37030         * expr.c (expand_expr_real_2): For vector MIN/MAX, if there is no
37031         min/max expander, expand it using expand_vec_cond_expr.
37033         PR sanitizer/79897
37034         * ubsan.c (ubsan_encode_value): Call mark_addressable on the
37035         temporary.
37037 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37039         PR c++/79821
37040         * dwarf2out.h (dw_vec_const): Change array type from unsigned char *
37041         to void * for PCH reasons.
37042         * dwarf2out.c (output_loc_operands, output_die): Cast
37043         v.val_vec.array to unsigned char *.
37045 2017-03-06  John David Anglin  <danglin@gcc.gnu.org>
37047         PR target/77850
37048         * config/pa/pa-64.h (PAD_VARARGS_DOWN): Don't pad down complex and
37049         vector types.
37051 2017-03-06  Vladimir Makarov  <vmakarov@redhat.com>
37053         PR rtl-optimization/79571
37054         * lra-constraints.c (process_alt_operands): Calculate static
37055         reject and subtract it from overall when only addresses will be
37056         reloaded.
37058 2017-03-06  Julia Koval  <julia.koval@intel.com>
37060         PR target/79793
37061         * config/i386/i386.c (ix86_minimum_incoming_stack_boundary): Set
37062         incoming stack boundary to 128 for 64-bit targets.
37064 2017-03-06  Richard Biener  <rguenther@suse.de>
37066         PR tree-optimization/79894
37067         * tree-vectorizer.c (vectorize_loops): Set loop_vectorized_call
37068         to NULL after folding it.
37070 2017-03-06  Richard Biener  <rguenther@suse.de>
37072         PR tree-optimization/79824
37073         * tree-vect-stmts.c (get_group_load_store_type): Fix alignment
37074         check disabling peeling for gaps.
37076 2017-03-06  Toma Tabacu  <toma.tabacu@imgtec.com>
37078         * doc/sourcebuild.texi (Effective-Target Keywords, Environment
37079         attributes): Document gettimeofday.
37081 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37083         * config/s390/s390.c (s390_option_override_internal): Set
37084         PARAM_MIN_VECT_LOOP_BOUND
37086 2017-03-06  Robin Dapp  <rdapp@linux.vnet.ibm.com>
37088         * config/s390/s390.c (s390_asm_output_function_label): Use nopr %r0.
37089         * config/s390/s390.md: Likewise.
37091 2017-03-06  Jakub Jelinek  <jakub@redhat.com>
37093         PR target/79812
37094         * config/i386/sse.md (VI8F_256_512): Remove mode iterator.
37095         (<avx2_avx512>_perm<mode>): Rename to ...
37096         (avx2_perm<mode>): ... this.  Use VI8F_256 iterator instead
37097         of VI8F_256_512.
37098         (<avx512>_perm<mode>_mask): Rename to ...
37099         (avx512vl_perm<mode>_mask): ... this.  Use VI8F_256 iterator instead
37100         of VI8F_256_512.
37101         (<avx2_avx512>_perm<mode>_1<mask_name>): Rename to ...
37102         (avx2_perm<mode>_1<mask_name): ... this.  Use VI8F_256 iterator
37103         instead of VI8F_256_512.
37104         (avx512f_perm<mode>): New define_expand.
37105         (avx512f_perm<mode>_mask): Likewise.
37106         (avx512f_perm<mode>_1<mask_name>): New define_insn.
37107         (<avx512>_vec_dup<mode>_1): Fix up vec_select mode.
37109 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37111         * config/mips/mips-msa.md (msa_fmax_a_<msafmt>, msa_fmin_a_<msafmt>,
37112         msa_max_a_<msafmt>, msa_min_a_<msafmt>): Introduce mode interator for
37113         if_then_else.
37114         (smin<mode>3, smax<mode>3): Change operand print code from 'B' to 'E'.
37116 2017-03-06  Martin Liska  <mliska@suse.cz>
37118         PR sanitize/79783
37119         * asan.c (asan_expand_poison_ifn): Do not expand ASAN_POISON
37120         when having a SSA NAME w/o VAR_DECL assigned to it.
37122 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37124         * config/mips/mips-msa.md (msa_dotp_<su>_d, msa_dpadd_<su>_d,
37125         msa_dpsub_<su>_d): Fix MODE for vec_select.
37127 2017-03-06  Prachi Godbole  <prachi.godbole@imgtec.com>
37129         * config/mips/mips.c (mips_gen_const_int_vector): Change type of last
37130         argument.
37131         * config/mips/mips-protos.h (mips_gen_const_int_vector): Likewise.
37133 2017-03-06  Richard Biener  <rguenther@suse.de>
37135         * lto-streamer.c (lto_check_version): Use %qs in diagnostics.
37136         * plugin.c (register_plugin_info): Likewise.
37137         * tree-chkp.c (chkp_make_static_const_bounds): Likewise.
37139 2017-03-05  Jakub Jelinek  <jakub@redhat.com>
37141         * config/i386/sse.md (sse_storehps, sse_storelps,
37142         avx_<castmode><avxsizesuffix>_<castmode>,
37143         avx512f_<castmode><avxsizesuffix>_<castmode>,
37144         avx512f_<castmode><avxsizesuffix>_256<castmode>): Require
37145         in condition that at least one operand is not a MEM.
37147 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37149         PR middle-end/79805
37150         * internal-fn.def (ATOMIC_BIT_TEST_AND_SET, ATOMIC_BIT_TEST_AND_RESET,
37151         ATOMIC_BIT_TEST_AND_COMPLEMENT, ATOMIC_COMPARE_EXCHANGE): Remove
37152         ECF_NOTHROW.
37153         * gimple-fold.c (fold_builtin_atomic_compare_exchange): Set
37154         gimple_call_nothrow_p flag based on whether original builtin can throw.
37155         If it can, emit following stmts on the fallthrough edge.
37156         * tree-ssa-ccp.c (optimize_atomic_bit_test_and): Similarly, except
37157         don't create new bb if inserting just debug stmts on the edge, try to
37158         insert them on the fallthru bb or just reset debug stmts.
37160 2017-03-03  Segher Boesssenkool  <segher@kernel.crashing.org>
37162         PR target/43763
37163         * config/rs6000/rs6000.c (rs6000_final_prescan_insn): Save and
37164         restore recog_data (including the operand rtxes inside it) around
37165         the call to get_insn_template.
37167 2017-03-03  Martin Sebor  <msebor@redhat.com>
37169         PR tree-optimization/79699
37170         * context.c (context::~context): Free MPFR caches to avoid
37171         a memory leak on program exit.
37173 2017-03-03  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
37175         * config/aarch64/aarch64.c (aarch64_float_const_representable_p):
37176         Use wide_int::ulow () instead of .elt (0).
37178 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37180         * config/i386/i386.md (*pushtf): Change *roF constraint to *roC.
37181         (*pushxf): Limit oF constraint to 32bit targets and add oC
37182         constraint for 64bit targets.
37183         (pushxf splitter): Use PUSH_ROUNDING to calculate stack adjustment.
37184         (*pushdf): Change rmF constraint to rmC.
37186 2017-03-03  Martin Liska  <mliska@suse.cz>
37188         * tree-ssa-loop-prefetch.c (pass_loop_prefetch::execute):
37189         Remove unused variable.
37191 2017-03-03  Jakub Jelinek  <jakub@redhat.com>
37193         PR target/79807
37194         * config/i386/i386.c (ix86_expand_multi_arg_builtin): If target
37195         is a memory operand, increase num_memory.
37196         (ix86_expand_args_builtin): Likewise.
37198 2017-03-03  Jan Hubicka  <jh@suse.cz>
37200         PR lto/79760
37201         * ipa-devirt.c (maybe_record_node): Properly handle
37202         __cxa_pure_virtual visibility.
37204 2017-03-03  Martin Liska  <mliska@suse.cz>
37206         PR tree-optimization/79803
37207         * tree-ssa-loop-prefetch.c (tree_ssa_prefetch_arrays): Remove
37208         assert.
37209         (pass_loop_prefetch::execute): Disabled optimization if an
37210         assumption about L1 cache size is not met.
37212 2017-03-03  Martin Liska  <mliska@suse.cz>
37214         PR rtl-optimization/79574
37215         * gcse.c (struct gcse_expr): Use HOST_WIDE_INT instead of int.
37216         (hash_scan_set): Likewise.
37217         (dump_hash_table): Likewise.
37218         (hoist_code): Likewise.
37220 2017-03-03  Richard Biener  <rguenther@suse.de>
37222         * fixed-value.c (fixed_from_string): Restore use of elt (1)
37223         in place of uhigh ().
37224         (fixed_convert_from_real): Likewise.
37226 2017-03-03  Uros Bizjak  <ubizjak@gmail.com>
37228         PR target/79514
37229         * config/i386/i386.md (*pushxf_rounded): Use Pmode instead of DImode.
37231 2017-03-03  Richard Biener  <rguenther@suse.de>
37233         PR middle-end/79818
37234         * match.pd ( X +- C1 CMP C2 -> X CMP C2 -+ C1): Add missing
37235         TYPE_OVERFLOW_UNDEFINED check.
37237 2017-03-02  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37239         * config/rs6000/vector.md (vector_ne_<mode>_p): Correct operand
37240         numbers.
37241         (vector_ae_<mode>_p): Likewise.
37242         (vector_nez_<mode>_p): Likewise.
37243         (vector_ne_v2di_p): Likewise.
37244         (vector_ae_v2di_p): Likewise.
37245         (vector_ne_<mode>_p): Likewise.
37246         * config/rs6000/vsx.md (vsx_tsqrt<mode>2_fg): Correct operand
37247         numbers.
37248         (vsx_tsqrt<mode>2_fe): Likewise.
37250 2017-03-02  Uros Bizjak  <ubizjak@gmail.com>
37252         PR target/79514
37253         * config/i386/i386.md (*pushxf_rounded): New insn_and_split pattern.
37255 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37257         PR rtl-optimization/79780
37258         * cprop.c (one_cprop_pass): When second and further conditional trap
37259         in a single basic block is turned into an unconditional trap, turn it
37260         into a deleted note to avoid RTL verification failures.
37262 2017-03-02  Richard Biener  <rguenther@suse.de>
37264         * fold-const.c (const_binop): Use ulow () instead of elt (0).
37266 2017-03-02  Richard Biener  <rguenther@suse.de>
37268         PR tree-optimization/79345
37269         PR c++/42000
37270         * tree-ssa-alias.c (walk_aliased_vdefs_1): Take a limit
37271         param and abort the walk, returning -1 if it is hit.
37272         (walk_aliased_vdefs): Take a limit param and pass it on.
37273         * tree-ssa-alias.h (walk_aliased_vdefs): Add a limit param,
37274         defaulting to 0 and return a signed int.
37275         * tree-ssa-uninit.c (struct check_defs_data): New struct.
37276         (check_defs): New helper.
37277         (warn_uninitialized_vars): Use walk_aliased_vdefs to warn
37278         about uninitialized memory.
37279         * fixed-value.c (fixed_from_string): Use ulow/uhigh to avoid
37280         bogus uninitialized warning.
37281         (fixed_convert_from_real): Likewise.
37283 2017-03-02  Bin Cheng  <bin.cheng@arm.com>
37285         PR tree-optimization/66768
37286         * tree-ssa-loop-ivopts.c (find_interesting_uses_address): Skip addr
37287         iv_use if base object can't be determined.
37289 2017-03-02  Jakub Jelinek  <jakub@redhat.com>
37291         PR tree-optimization/79345
37292         * gensupport.h (struct pattern_stats): Add min_scratch_opno field.
37293         * gensupport.c (get_pattern_stats_1) <case MATCH_SCRATCH>: Update it.
37294         (get_pattern_stats): Initialize it.
37295         * genemit.c (gen_expand): Verify match_scratch numbers come after
37296         match_operand/match_dup numbers.
37297         * config/i386/i386.md (<s>mul<mode>3_highpart): Swap match_dup and
37298         match_scratch numbers.
37299         * config/i386/sse.md (avx2_gathersi<mode>, avx2_gatherdi<mode>):
37300         Likewise.
37301         * config/s390/s390.md (trunctdsd2): Likewise.
37303 2017-03-02  Richard Biener  <rguenther@suse.de>
37305         * wide-int.h (wide_int_storage::operator=): Implement in terms
37306         of wi::copy.
37308 2017-03-02  Richard Biener  <rguenther@suse.de>
37310         PR tree-optimization/79777
37311         * tree-ssa-pre.c (eliminate_insert): Give up if we simplify
37312         the to insert expression to sth existing.
37314 2017-03-01  Martin Sebor  <msebor@redhat.com>
37316         PR middle-end/79692
37317         * gimple-ssa-sprintf.c
37318         (directive::known_width_and_precision): New function.
37319         (format_integer): Use it.
37320         (get_mpfr_format_length): Consider the full range of precision
37321         when computing %g output with the # flag.  Set the likely byte
37322         count to 3 rather than 1 when precision is indeterminate.
37323         (format_floating): Correct the lower bound of precision.
37325 2017-03-01  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
37327         * doc/invoke.texi: Document default code model for 64-bit Linux.
37329 2017-03-01  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
37331         PR target/79752
37332         * config/rs6000/rs6000.md (peephole2 for udiv/umod): Should emit
37333         udiv rather than div since input pattern is unsigned.
37335 2017-03-01  Uros Bizjak  <ubizjak@gmail.com>
37337         * config/i386/i386.c (print_reg): Warn for values of
37338         unsupported size in integer register.
37340 2017-03-01  Michael Meissner  <meissner@linux.vnet.ibm.com>
37342         PR target/79439
37343         * config/rs6000/predicates.md (current_file_function_operand): Do
37344         not allow self calls to be local if the function is replaceable.
37346 2017-03-01  Kelvin Nilsen  <kelvin@gcc.gnu.org>
37348         PR target/79395
37349         * config/rs6000/altivec.h (vec_ctz and others): Change the
37350         preprocessor macro that controls conditional compilation from
37351         _ARCH_PWR9 to __POWER9_VECTOR__.
37352         (vec_all_ne): Change parameterization of __altivec_scalar_pred
37353         macro expansion under preprocessor #ifdef __POWER9_VECTOR__
37354         control (instead of _ARCH_PWR9 control) so that template
37355         definition uses power9-specific function.
37356         (vec_any_eq): Likewise.
37357         (vec_all_ne): Change macro definition to use a power9-specific
37358         expansion under #ifdef __POWER9_VECTOR__ control (instead of
37359         _ARCH_PWR9 control).
37360         (vec_any_eq) Likewise.
37361         * config/rs6000/rs6000-builtin.def (CMPNEF): Remove BU_P9V_AV_2
37362         expansion for CMPNEF to remove support for xvcmpnesp instruction.
37363         (CMPNED): Remove BU_P9V_AV2 expansion for CMPNED to remove
37364         support for xvcmpnedp instruction.
37365         (VCMPNEB_P): Replace BU_P9V_AV_P macro expansion with BU_P9V_AV_2
37366         macro expansion so that Power9 implementation of vec_all_ne does
37367         not use the AltiVec predicate framework.
37368         (VCMPNEH_P): Likewise.
37369         (VCMPNEW_P): Likewise.
37370         (VCMPNED_P): Likewise.
37371         (VCMPNEFP_P): Likewise.
37372         (VCMPNEDP_P): Likewise.
37373         (VCMPAEB_P): Add BU_P9V_AV_2 macro expansion to change
37374         implementation of vec_any_eq to not use AltiVec predicate
37375         framework.
37376         (VCMPAEH_P): Likewise.
37377         (VCMPAEW_P): Likewise.
37378         (VCMPAED_P): Likewise.
37379         (VCMPAEFP_P): Likewise.
37380         (VCMPAEDP_P): Likewise.
37381         (VCMPNE_P): Replace BU_P9V_OVERLOAD_P macro expansion with
37382         BU_P9V_OVERLOAD_2 so that Power9 implementation of vec_all_ne does
37383         not use the AltiVec predicate framework.
37384         (VCMPAE_P): Add BU_P9V_OVERLOAD_2 macro to change implementation
37385         of vec_any_eq to not use AltiVec predicate framework.
37386         * config/rs6000/rs6000-c.c (rs6000_target_modify_macros): Add
37387         support for predefined __POWER9_VECTOR__ macro to indicate that
37388         Power9 instruction selection is enabled.
37389         (altivec_overloaded_builtins): Remove extraneous
37390         ALTIVEC_BUILTIN_VEC_CMPNE entry for overloaded
37391         function argument types RS6000_BTI_bool_V16QI and
37392         RS6000_BTI_bool_V16QI.  Remove erroneous ALTIVEC_BUILTIN_VEC_CMPNE
37393         entry for overloaded function argument types RS6000_BTI_bool_V4SI
37394         andRS6000_BTI_bool_V4SI, mapping to P9V_BUILTIN_CMPNEB.  Remove
37395         two entries mapping to P9V_BUITIN_CMPNED and one entry mapping to
37396         P9V_BUILTIN_CMPNEF to force use of instructions not specific to
37397         Power9 for implementations of vec_cmpne.  Change the signature for
37398         all definitions of the overloaded P9V_BUILTIN_VEC_CMPNE_P function
37399         (representing vec_all_ne) to remove the previously described first
37400         argument of type RS6000_BTI_INTSI, as this was an artifact of
37401         reliance on the AltiVec predicate framework, which is no longer
37402         used in the implementation of these functions.  Add
37403         P9V_BUILTIN_VEC_VCMPAE_P entries (representing the vec_anyeq
37404         function) to match all of the P9V_BUILTIN_VEC_VCMNE_P entries
37405         since, unlike the AltiVec predicate framework implementation, we
37406         do not share function descriptors between vec_alle and vec_anyeq.
37407         (altivec_resolve_overloaded_builtin): Add SFmode and DFmode to the
37408         set of modes that receive special treatment even when
37409         TARGET_P9_VECTOR is true.  The special treatment emits code that
37410         does not depend on Power9 instructions.
37411         * config/rs6000/vector.md (vector_ne_<mode>_p): Change this
37412         define_expand to not rely on AltiVec predicate framework.
37413         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37414         function.
37415         (vector_ne_v2di_p): Change this define_expand to not rely on
37416         AltiVec predicate framework.
37417         (vector_ae_v2di_p): New define_expand to represent vec_any_eq
37418         function.
37419         (vector_ne_<mode>_p): Change this define_expand to not rely on
37420         AltiVec predicate framework.
37421         (vector_ae_<mode>p): New define_expand to represent vec_any_eq
37422         function.
37423         * config/rs6000/vsx.md (*vsx_ne_<mode>_p): For modes VSX_EXTRACT_I
37424         (V16QI, V8HI, V4SI), correct a typo in the code emitted for this
37425         define_insn pattern.
37426         (*vsx_ne_<mode>_p): For modes VSX_F (V4SF and V2DF), remove this
37427         define_insn pattern because the xvcmpne<VSs>. instruction is not
37428         supported.
37429         (vcmpne<VSs>): Remove this define_insn because xvcmpne<VSs>
37430         instruction is not supported.
37432 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37434         * config/nvptx/nvptx.c: Include intl.h.
37436 2017-03-01  Martin Jambor  <mjambor@suse.cz>
37438         PR lto/78140
37439         * ipa-prop.h (ipa_bits): Removed field known.
37440         (ipa_jump_func): Removed field vr_known.  Changed fields bits and m_vr
37441         to pointers.  Adjusted their comments to warn about their sharing.
37442         (ipcp_transformation_summary): Change bits to a vector of pointers.
37443         (ipa_check_create_edge_args): Moved to ipa-prop.c, declare.
37444         (ipa_get_ipa_bits_for_value): Declare.
37445         * tree-vrp.h (value_range): Mark as GTY((for_user)).
37446         * ipa-prop.c (ipa_bit_ggc_hash_traits): New.
37447         (ipa_bits_hash_table): Likewise.
37448         (ipa_vr_ggc_hash_traits): Likewise.
37449         (ipa_vr_hash_table): Likewise.
37450         (ipa_print_node_jump_functions_for_edge): Adjust for bits and m_vr
37451         being pointers and vr_known being removed.
37452         (ipa_set_jf_unknown): Likewise.
37453         (ipa_get_ipa_bits_for_value): New function.
37454         (ipa_set_jfunc_bits): Likewise.
37455         (ipa_get_value_range): New overloaded functions.
37456         (ipa_set_jfunc_vr): Likewise.
37457         (ipa_compute_jump_functions_for_edge): Use the above functions to
37458         construct bits and vr parts of jump functions.
37459         (ipa_check_create_edge_args): Move here from ipa-prop.h, also allocate
37460         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37461         exist.
37462         (ipcp_grow_transformations_if_necessary): Also allocate
37463         ipa_bits_hash_table and ipa_vr_hash_table if they do not already
37464         exist.
37465         (ipa_node_params_t::duplicate): Do not copy bits, just pointers to
37466         them.  Fix too long lines.
37467         (ipa_write_jump_function): Adjust for bits and m_vr being pointers and
37468         vr_known being removed.
37469         (ipa_read_jump_function): Use new setter functions to construct bits
37470         and vr parts of jump functions or set them to NULL.
37471         (write_ipcp_transformation_info): Adjust for bits being pointers.
37472         (read_ipcp_transformation_info): Likewise.
37473         (ipcp_update_bits): Likewise.  Fix excessively long lines a trailing
37474         space.
37475         Include gt-ipa-prop.h.
37476         * ipa-cp.c (propagate_bits_across_jump_function): Adjust for bits
37477         being pointers.
37478         (ipcp_store_bits_results): Likewise.
37479         (propagate_vr_across_jump_function): Adjust for m_vr being a pointer.
37480         Do not write to existing jump functions but use a temporary instead.
37482 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37484         PR c++/79681
37485         * fold-const.c (make_bit_field_ref): If orig_inner is COMPONENT_REF,
37486         attempt to use its first operand as BIT_FIELD_REF base.
37488 2017-03-01  Richard Biener  <rguenther@suse.de>
37490         PR middle-end/79721
37491         * tree-chrec.c (chrec_evaluate): Perform computation of Newtons
37492         interpolating formula in wrapping arithmetic.
37493         (chrec_apply): Convert chrec_evaluate return value to wanted type.
37495 2017-03-01  Jakub Jelinek  <jakub@redhat.com>
37497         PR tree-optimization/79734
37498         * tree-vect-generic.c (expand_vector_condition): Optimize
37499         AVX512 vector boolean VEC_COND_EXPRs into bitwise operations.
37500         Handle VEC_COND_EXPR where comparison has different inner width from
37501         type's inner width.
37503 2017-02-28  Sandra Loosemore  <sandra@codesourcery.com>
37505         * doc/invoke.texi (ARC Options): Copy-edit to fix punctuation,
37506         markup, and similar issues.  Remove @opindex entries for things
37507         that aren't options.  Add missing -mmpy-option entries.
37509 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37511         PR tree-optimization/79737
37512         * gimple-ssa-store-merging.c (encode_tree_to_bitpos): If bitlen is
37513         a multiple of BITS_PER_UNIT and !BYTES_BIG_ENDIAN, clear
37514         tmpbuf[byte_size - 1].  Call natice_encode_expr with byte_size - 1
37515         instead of byte_size.  Formatting fix.
37516         (shift_bytes_in_array_right): Formatting fix.
37518 2017-02-28  Eric Botcazou  <ebotcazou@adacore.com>
37520         PR target/79749
37521         * config/sparc/sparc.c (sparc_frame_pointer_required): Add missing
37522         condition on optimize for the leaf function test.
37524 2017-02-28  Martin Liska  <mliska@suse.cz>
37526         PR lto/79625
37527         * read-rtl-function.c (function_reader::handle_unknown_directive):
37528         Bail out when one uses -flto.
37530 2017-02-28  Martin Liska  <mliska@suse.cz>
37532         * common.opt: Replace space with tabular for options of <number>
37533         type.
37534         * config/i386/i386.opt: Show <number> value for
37535         -mlarge-data-threshold.
37536         * opts.c (print_filtered_help): Do not display number in hexadecimal
37537         format.
37539 2017-02-28  Martin Liska  <mliska@suse.cz>
37541         * common.opt: Fix --help=option -Q for options which are of
37542         an enum type.
37544 2017-02-28  Uros Bizjak  <ubizjak@gmail.com>
37546         * config/i386/i386.c (print_reg): Error out for values
37547         of 8-bit size in invalid integer register.
37549 2017-02-28  Martin Sebor  <msebor@redhat.com>
37551         PR tree-optimization/79691
37552         * passes.def (pass_all_optimizations_g): Enable pass_sprintf_length.
37554 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37556         PR target/79729
37557         * config/i386/i386.c (ix86_print_operand) <case 'R'>: Replace
37558         gcc_unreachable with output_operand_lossage.
37560 2017-02-28  Richard Biener  <rguenther@suse.de>
37562         PR tree-optimization/79740
37563         * tree-ssa-sccvn.c (vn_nary_op_insert_into): Allow redundant
37564         inserts.
37565         (visit_nary_op): Insert the nary into the hashtable if we
37566         pattern-matched sth.
37567         * tree-ssa-pre.c (eliminate_insert): Robustify.
37569 2017-02-28  Richard Biener  <rguenther@suse.de>
37571         PR middle-end/79731
37572         * fold-const.c (decode_field_reference): Reject out-of-bound
37573         accesses.
37575 2017-02-28  Jakub Jelinek  <jakub@redhat.com>
37577         * config/i386/i386.c: Include intl.h.
37578         (ix86_option_override_internal): Use cond ? G_("...") : G_("...")
37579         instead of just cond ? "..." : "...".
37580         * config/nvptx/nvptx.c (nvptx_goacc_validate_dims): Likewise.
37581         * coverage.c (read_counts_file): Likewise.
37582         * omp-offload.c: Include intl.h.
37583         (oacc_loop_fixed_partitions): Use cond ? G_("...") : G_("...") instead
37584         of just cond ? "..." : "...".
37585         * gcov.c (read_count_file): Use cond ? N_("...") : N_("...") instead
37586         of just cond ? "..." : "...".
37588 2017-02-28  Richard Earnshaw  <rearnsha@arm.com>
37590         PR target/79742
37591         * config/arm/parsecpu.awk (gen_data): Set tuning target to 'tune for'
37592         entry, if present.
37593         * config/arm/arm-cpus.in (cortex-m0plus.small-multiply): Correct
37594         'tune for' CPU name.
37595         * config/arm/arm-cpu-data.h: Regenerated.
37597 2017-02-28  Richard Biener  <rguenther@suse.de>
37599         PR tree-optimization/79732
37600         * tree-inline.c (expand_call_inline): Do not shadow var.
37602 2017-02-28  Richard Biener  <rguenther@suse.de>
37604         PR tree-optimization/79723
37605         * tree-vect-stmts.c (get_vectype_for_scalar_type_and_size): Preserve
37606         address-space properly.
37608 2017-02-28  Thomas Schwinge  <thomas@codesourcery.com>
37610         * doc/optinfo.texi (Optimization groups): Fix option used for
37611         OPTGROUP_ALL.
37612         * doc/invoke.texi (-fopt-info): Document "omp".
37613         * dumpfile.h: Sort OPTGROUP_OMP before OPTGROUP_VEC.
37614         (OPTGROUP_ALL): Add OPTGROUP_OMP.
37615         * hsa-gen.c (pass_data_gen_hsail): Use OPTGROUP_OMP.
37616         * ipa-hsa.c (pass_data_ipa_hsa): Likewise.
37617         * omp-simd-clone.c (pass_data_omp_simd_clone): Likewise.
37619         * dumpfile.h (OPTGROUP_OPENMP): Rename to OPTGROUP_OMP.  Adjust
37620         all users.
37621         * dumpfile.c (optgroup_options): Instead of "openmp", associate
37622         OPTGROUP_OMP with "omp".
37624 2017-02-27  Pat Haugen  <pthaugen@us.ibm.com>
37626         PR target/79544
37627         * config/rs6000/rs6000-c.c (struct altivec_builtin_types): Use VSRAD
37628         for arithmetic shift of unsigned V2DI.
37630 2017-02-27  Claudiu Zissulescu  <claziss@synopsys.com>
37632         * config.gcc (arc*-): Clean up, use arc/big.h, arc/elf.h, and
37633         arc/linux.h headers.
37634         * config/arc/arc.h (TARGET_OS_CPP_BUILTINS): Remove.
37635         (LINK_SPEC): Likewise.
37636         (ARC_TLS_EXTRA_START_SPEC): Likewise.
37637         (EXTRA_SPECS): Likewise.
37638         (STARTFILE_SPEC): Likewise.
37639         (ENDFILE_SPEC): Likewise.
37640         (LIB_SPEC): Likewise.
37641         (TARGET_SDATA_DEFAULT): Likewise.
37642         (TARGET_MMEDIUM_CALLS_DEFAULT): Likewise.
37643         (MULTILIB_DEFAULTS): Likewise.
37644         (DWARF2_UNWIND_INFO): Likewise.
37645         * config/arc/big.h: New file.
37646         * config/arc/elf.h: Likewise.
37647         * config/arc/linux.h: Likewise.
37648         * config/arc/t-uClibc: Remove.
37650 2017-02-27  Bin Cheng  <bin.cheng@arm.com>
37652         PR tree-optimization/77536
37653         * tree-ssa-loop-manip.c (niter_for_unrolled_loop): New function.
37654         (tree_transform_and_unroll_loop): Use above function to compute the
37655         estimated niter of unrolled loop and use it when scaling profile.
37656         Also use count info rather than frequency if it's non-zero.
37657         * tree-ssa-loop-manip.h niter_for_unrolled_loop(): New declaration.
37658         * tree-vect-loop.c (scale_profile_for_vect_loop): New function.
37659         (vect_transform_loop): Call above function.
37661 2017-02-27  Richard Biener  <rguenther@suse.de>
37663         PR tree-optimization/45397
37664         * tree-ssa-pre.c (eliminate_insert): Handle BIT_AND_EXPR.
37665         * tree-ssa-sccvn.c (valueized_wider_op): New helper.
37666         (visit_nary_op): Add pattern matching for CSEing sign-changed
37667         or truncated operations with wider ones.
37669 2017-02-27  Richard Biener  <rguenther@suse.de>
37671         PR tree-optimization/79690
37672         * tree-vect-stmts.c (vectorizable_store): Use vector type
37673         built from the DR with address-space.
37675 2017-02-26  Gerald Pfeifer  <gerald@pfeifer.com>
37677         * doc/invoke.texi (Optimize Options): Refine the description
37678         of asan-use-after-return.
37680 2017-02-25  Alan Modra  <amodra@gmail.com>
37682         PR rtl-optimization/79584
37683         * lra-constraints.c (base_to_reg): Reload ad->base, the entire
37684         base, not ad->base_term, the reg within base.  Remove assertion
37685         that ad->base == ad->base_term.  Replace gen_int_mode using
37686         bogus mode with const0_rtx.
37688 2017-02-25  Jakub Jelinek  <jakub@redhat.com>
37690         PR middle-end/79396
37691         * tree-eh.c (operation_could_trap_p, stmt_could_throw_1_p): Handle
37692         FMA_EXPR like tcc_binary or tcc_unary.
37694         * tree-ssa-loop-niter.c (number_of_iterations_exit): Simplify warning.
37696         PR debug/77589
37697         * dwarf2out.c (struct dw_loc_list_struct): Add noted_variable_value
37698         bitfield.
37699         (size_of_loc_descr): Handle DW_OP_GNU_variable_value.
37700         (output_loc_operands): Handle DW_OP_call_ref and
37701         DW_OP_GNU_variable_value.
37702         (struct variable_value_struct): New type.
37703         (struct variable_value_hasher): Likewise.
37704         (variable_value_hash): New variable.
37705         (string_types): Remove.
37706         (copy_loc_descr): New function.
37707         (add_loc_descr_to_each): Clarify comment.  Use copy_loc_descr.
37708         (prepend_loc_descr_to_each): New function.
37709         (add_loc_list): Fix comment typo.  Use prepend_loc_descr_to_each
37710         instead of add_loc_descr_to_each if the first argument is single
37711         location list and the second has multiple.
37712         (resolve_args_picking_1): Handle DW_OP_GNU_variable_value.
37713         (loc_list_from_tree_1): For early_dwarf, emit DW_OP_GNU_variable_value
37714         when looking for variable value which doesn't have other location info.
37715         (loc_list_from_tree): Formatting fix.
37716         (gen_array_type_die): Simplify DW_AT_string_length handling.
37717         (adjust_string_types): Remove.
37718         (gen_subprogram_die): Don't call adjust_string_types nor test/set
37719         string_types.  Call resolve_variable_values.
37720         (prune_unused_types_walk_loc_descr): Handle DW_OP_GNU_variable_value.
37721         (resolve_addr_in_expr): Likewise.  Add A argument.
37722         (copy_deref_exprloc): Remove deref argument.  Adjust for the
37723         original expression being DW_OP_GNU_variable_value with optionally
37724         DW_OP_stack_value after it instead of DW_OP_call4 with DW_OP_deref
37725         optionally after it.
37726         (optimize_string_length): Rework for DW_OP_GNU_variable_value.
37727         (resolve_addr): Adjust optimize_string_length and resolve_addr_in_expr
37728         callers.  Set remove_AT_byte_size if removing DW_AT_string_length.
37729         (variable_value_hasher::hash, variable_value_hasher::equal): New
37730         methods.
37731         (resolve_variable_value_in_expr, resolve_variable_value,
37732         resolve_variable_values, note_variable_value_in_expr,
37733         note_variable_value): New functions.
37734         (dwarf2out_early_finish): Call note_variable_value on all toplevel
37735         DIEs.
37737 2017-02-24  Jakub Jelinek  <jakub@redhat.com>
37739         PR c/79677
37740         * opts.h (handle_generated_option): Add GENERATED_P argument.
37741         * opts-common.c (handle_option): Adjust function comment.
37742         (handle_generated_option): Add GENERATED_P argument, pass it to
37743         handle_option.
37744         (control_warning_option): Pass false to handle_generated_option
37745         GENERATED_P.
37746         * opts.c (maybe_default_option): Pass true to handle_generated_option
37747         GENERATED_P.
37748         * optc-gen.awk: Likewise.
37750 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37752         * config/sh/sh.md (tstsi_t): If operands[0] is a SUBREG instead of
37753         a REG, look at the REG it is a SUBREG of.
37754         (splitter for cmpeqsi_t): Ditto.
37756 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37758         * config/pa/pa.c (pa_combine_instructions): Do not share RTL.  Make
37759         the special USEs with the pattern of the insn, not the insn itself.
37761 2017-02-24  Matthew Fortune  <matthew.fortune@imgtec.com>
37763         PR target/79473
37764         * doc/invoke.texi: Document -mload-store-pairs.
37766 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37767             Sandra Loosemore  <sandra@codesourcery.com>
37769         * config/nios2/nios2.c (nios2_simple_const_p): Returns false if the
37770         argument isn't a CONST_INT.
37771         (nios2_alternate_compare_const): Assert op is a CONST_INT.
37772         (nios2_valid_compare_const_p): Assert op is a CONST_INT.
37773         (nios2_validate_compare): Bypass alternate compare logic if *op2
37774         is not a CONST_INT.
37775         (ldstwm_operation_p): Return false if first_base is not a REG or
37776         if first_offset is not a CONST_INT.
37778 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37780         * config/cris/cris.md: Use correct operand in a define_peephole2.
37782 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37784         * config/c6x/c6x.c (predicate_insn): Do not incorrectly share RTL.
37786 2017-02-24  Segher Boessenkool  <segher@kernel.crashing.org>
37788         * config/arc/arc.c (arc_ccfsm_advance): Only take the PATTERN of
37789         this_insn if it is an INSN or JUMP_INSN.
37790         (force_offsettable): Look at base, not at addr.
37791         * config/arc/predicates.md (brcc_nolimm_operator): Don't call INTVAL
37792         on things that aren't necessarily CONST_INTs.
37794 2017-02-24  Uros Bizjak  <ubizjak@gmail.com>
37796         * doc/invoke.texi (x86 Options, -mfpmath=sse): Mention that
37797         -mfpmath=sse is the default also for x86-32 targets with SSE2
37798         instruction set when @option{-ffast-math} is enabled
37800 2017-02-24  Jeff Law  <law@redhat.com>
37802         PR rtl-optimizatoin/79286
37803         * ira.c (update_equiv_regs): Drop may_trap_p exception to
37804         dominance test.
37806 2017-02-24  Richard Biener  <rguenther@suse.de>
37808         PR tree-optimization/79389
37809         * gimple-ssa-split-paths.c (is_feasible_trace): Properly skip
37810         debug insns.
37812 2017-02-24  Aldy Hernandez  <aldyh@redhat.com>
37814         * tree-ssa-loop-niter.c (number_of_iterations_exit): Update
37815         function comment to reflect reality.
37816         (loop_exits_before_overflow): Fix typo in function description.
37818 2017-02-24  Richard Biener  <rguenther@suse.de>
37820         PR tree-optimization/79389
37821         * gimple-ssa-split-paths.c (is_feasible_trace): Verify more
37822         properly that a threading opportunity exists.  Detect conditional
37823         copy/constant propagation opportunities.
37825 2017-02-23  Eric Botcazou  <ebotcazou@adacore.com>
37827         * config/visium/visium.md (type): Add trap.
37828         (b): New mode attribute.
37829         (*btst): Rename into...
37830         (*btst<mode>): ...this and adjust.
37831         (*cbranchsi4_btst_insn): Rename into...
37832         (*cbranch<mode>4_btst_insn): ...this and adjust.
37833         (trap): New define_insn.
37835 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37837         PR tree-optimization/79389
37838         * ifcvt.c (struct noce_if_info): Add rev_cond field.
37839         (noce_reversed_cond_code): New function.
37840         (noce_emit_store_flag): Use rev_cond if non-NULL instead of
37841         reversed_comparison_code.  Formatting fix.
37842         (noce_try_store_flag): Test rev_cond != NULL in addition to
37843         reversed_comparison_code.
37844         (noce_try_store_flag_constants): Likewise.
37845         (noce_try_store_flag_mask): Likewise.
37846         (noce_try_addcc): Use rev_cond if non-NULL instead of
37847         reversed_comparison_code.
37848         (noce_try_cmove_arith): Likewise.  Formatting fixes.
37849         (noce_try_minmax, noce_try_abs): Clear rev_cond.
37850         (noce_find_if_block): Initialize rev_cond.
37851         (find_cond_trap): Call noce_get_condition with then_bb == trap_bb
37852         instead of false as last argument never attempt to reverse it
37853         afterwards.
37855 2017-02-23  Bin Cheng  <bin.cheng@arm.com>
37857         PR tree-optimization/79663
37858         * tree-predcom.c (combine_chains): Process refs in reverse order
37859         only for ZERO length chains, and add explaining comment.
37861 2017-02-23  Jeff Law  <law@redhat.com>
37863         PR tree-optimization/79578
37864         * tree-ssa-dse.c (clear_bytes_written_by): Use OEP_ADDRESS_OF
37865         in call to operand_equal_p.
37867 2017-01-23  Dominique d'Humieres  <dominiq@lps.ens.fr>
37869         PR target/71017
37870         * config/i386/cpuid.h: Fix another undefined behavior.
37872 2017-02-23  Richard Biener  <rguenther@suse.de>
37874         PR tree-optimization/79683
37875         * tree-vect-stmts.c (vect_analyze_stmt): Do not overwrite
37876         vector types for data-refs.
37878 2017-02-23  Martin Liska  <mliska@suse.cz>
37880         * params.def (PARAM_MIN_NONDEBUG_INSN_UID): Change default to 0.
37882 2017-02-23  Jakub Jelinek  <jakub@redhat.com>
37884         PR middle-end/79665
37885         * internal-fn.c (get_range_pos_neg): Moved to ...
37886         * tree.c (get_range_pos_neg): ... here.  No longer static.
37887         * tree.h (get_range_pos_neg): New prototype.
37888         * expr.c (expand_expr_real_2) <case TRUNC_DIV_EXPR>: If both arguments
37889         are known to be in between 0 and signed maximum inclusive, try to
37890         expand both unsigned and signed divmod and use the cheaper one from
37891         those.
37893 2017-02-22  Jeff Law  <law@redhat.com>
37895         PR tree-optimization/79578
37896         * tree-ssa-dse.c (clear_bytes_written_by): Use operand_equal_p
37897         to compare base operands.
37899 2017-02-22  Segher Boessenkool  <segher@kernel.crashing.org>
37901         PR target/79211
37902         * config/rs6000/rs6000.md (*fsel<SFDF:mode><SFDF2:mode>4): Use
37903         gpc_reg_operand instead of fpr_reg_operand.
37905 2017-02-22  Sameera Deshpande  <sameera.deshpande@imgtec.com>
37907         * config/mips/mips.c (mips_return_in_memory): Force FP
37908         vector types to be returned in memory for o32 ABI.
37910 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37912         * dwarf2out.c (gen_variable_die): For -gdwarf-5, use DW_TAG_variable
37913         instead of DW_TAG_member for static data member declarations and don't
37914         set no_linkage_name for static inline data members.
37915         (gen_member_die): For -gdwarf-5 don't change DW_TAG_variable
37916         to DW_TAG_member.
37918 2017-02-22  Martin Liska  <mliska@suse.cz>
37920         * doc/invoke.texi: Replace inequality signs with square brackets
37921         for -Wnormalized.
37923 2017-02-22  Matthew Fortune  <matthew.fortune@imgtec.com>
37925         PR target/78660
37926         * lra-constraints.c (simplify_operand_subreg): Handle
37927         WORD_REGISTER_OPERATIONS targets.
37929 2017-02-22  Jakub Jelinek  <jakub@redhat.com>
37931         PR target/70465
37932         * reg-stack.c (emit_swap_insn): Treat (float_extend:?F (mem:?F))
37933         and (const_double:?F) like (mem:?F) for the purpose of fxch %st(1)
37934         elimination by swapping fld*.
37936 2017-02-22  Richard Biener  <rguenther@suse.de>
37938         PR tree-optimization/79673
37939         * tree-ssa-pre.c (compute_avail): Use wide_int_to_tree to
37940         convert the [TARGET_]MEM_REF offset INTEGER_CST, scrapping off
37941         irrelevant address-space qualifiers and avoiding a
37942         ADDR_SPACE_CONVERT_EXPR from fold_convert.
37944 2017-02-22  Richard Biener  <rguenther@suse.de>
37946         PR tree-optimization/79666
37947         * tree-vrp.c (extract_range_from_binary_expr_1): Make sure
37948         to not symbolically negate if that may introduce undefined
37949         overflow.
37951 2017-02-22  Martin Liska  <mliska@suse.cz>
37953         PR lto/79587
37954         * data-streamer-in.c (streamer_read_gcov_count): Remove assert.
37955         * data-streamer-out.c (streamer_write_gcov_count_stream):
37956         Likewise.
37957         * value-prof.c (stream_out_histogram_value): Make assert more
37958         precise based on type of counter.
37960 2017-02-21  Uros Bizjak  <ubizjak@gmail.com>
37962         PR target/79593
37963         * config/i386/i386.md (standard_x87sse_constant_load splitter):
37964         Use nonimmediate_operand instead of memory_operand for operand 1.
37965         (float-extend standard_x87sse_constant_load splitter): Ditto.
37967 2017-02-21  Jeff Law  <law@redhat.com>
37969         PR tree-optimization/79621
37970         * gimple-ssa-isolate-paths.c (find_implicit_erroneous_behavior): Ignore
37971         blocks with edges to themselves.
37973 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37975         PR target/79633
37976         * tree-chkp-opt.c (chkp_optimize_string_function_calls): Use
37977         is_gimple_call instead of comparing gimple_code with GIMPLE_CALL.
37978         Use gimple_call_builtin_p.
37980         PR target/79570
37981         * sel-sched.c (moveup_expr_cached): Don't call sel_bb_head
37982         on temporarily removed DEBUG_INSNs.
37984         PR tree-optimization/79649
37985         * tree-loop-distribution.c (classify_partition): Give up on
37986         non-generic address space loads/stores.
37988 2017-02-21  Aldy Hernandez  <aldyh@redhat.com>
37990         * doc/loop.texi (Loop manipulation): Remove nonexistent
37991         tree_ssa_loop_version from the documentation.
37992         * cfgloopmanip.c (loop_version): Document CONDITION_BB argument.
37994 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
37996         PR target/79494
37997         * config/i386/i386.c (ix86_expand_split_stack_prologue): Call
37998         make_reg_eh_region_note_nothrow_nononlocal on call_insn.
37999         * config/rs6000/rs6000.c: Include except.h.
38000         (rs6000_expand_split_stack_prologue): Call
38001         make_reg_eh_region_note_nothrow_nononlocal on the call insn.
38003 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38005         PR lto/79579
38006         * ipa-prop.c (ipa_prop_write_jump_functions): Bail out if no edges
38007         have been analyzed.
38009 2017-02-21  Martin Jambor  <mjambor@suse.cz>
38011         * common.opt (-fipa-cp-alignment): Mark as ignored and preserved
38012         for backward compatibility only.
38013         * doc/invoke.texi (Option Summary): Remove all references to
38014         -fipa-cp-alignment.
38016 2017-02-21  Matthew Fortune  <matthew.fortune@imgtec.com>
38018         PR target/78660
38019         Revert:
38020         2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38022         * lra-constraints.c (curr_insn_transform): Handle
38023         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38025 2017-02-21  Martin Liska  <mliska@suse.cz>
38027         * config/i386/i386.opt: Replace -masm-dialect with -masm.
38029 2017-02-21  Thomas Schwinge  <thomas@codesourcery.com>
38031         PR translation/79638
38032         * config/nvptx/nvptx.c (ENTRY_TEMPLATE): Single out "%ntid.y".
38034 2017-02-21  Eric Botcazou  <ebotcazou@adacore.com>
38036         PR ada/67205
38037         * config/arm/arm.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define.
38038         (arm_function_ok_for_sibcall): Return false for an indirect call by
38039         descriptor if all the argument registers are used.
38040         (arm_relayout_function): Use FUNCTION_ALIGNMENT macro to adjust the
38041         alignment of the function.
38043 2017-02-21  Jakub Jelinek  <jakub@redhat.com>
38045         PR tree-optimization/61441
38046         * simplify-rtx.c (simplify_const_unary_operation): For
38047         -fsignaling-nans and sNaN operand, return NULL_RTX rather than
38048         the sNaN unmodified.
38050 2017-02-20  Bernd Edlinger  <bernd.edlinger@hotmail.de>
38052         * Makefile.in (BUILD_SYSTEM_HEADER_DIR): New make variabe.
38053         (LIMITS_H_TEST, if_multiarch, stmp-fixinc): Use BUILD_SYSTEM_HEADER_DIR
38054         instead of SYSTEM_HEADER_DIR.
38056 2017-02-20  Gerald Pfeifer  <gerald@pfeifer.com>
38057             Martin LiÅ¡ka  <mliska@suse.cz>
38059         * doc/invoke.texi (use-after-scope-direct-emission-threshold):
38060         Fix typos and grammar, use active voice, and clarify.
38062 2017-02-20  Marek Polacek  <polacek@redhat.com>
38064         PR middle-end/79537
38065         * gimplify.c (gimplify_expr): Handle unused *&&L;.
38067         PR sanitizer/79558
38068         * ubsan.c (ubsan_type_descriptor): Check if TYPE_MAX_VALUE is null.
38070 2017-02-20  Jakub Jelinek  <jakub@redhat.com>
38072         PR target/79568
38073         * config/i386/i386.c (ix86_expand_builtin): Handle
38074         OPTION_MASK_ISA_AVX512VL and OPTION_MASK_ISA_64BIT in
38075         ix86_builtins_isa[fcode].isa as a requirement of those
38076         flags and any other flag in the bitmask.
38077         (ix86_init_mmx_sse_builtins): Use 0 instead of
38078         ~OPTION_MASK_ISA_64BIT as mask.
38079         * config/i386/i386-builtin.def (__builtin_ia32_rdtsc,
38080         __builtin_ia32_rdtscp, __builtin_ia32_pause, __builtin_ia32_bsrsi,
38081         __builtin_ia32_rdpmc, __builtin_ia32_rolqi, __builtin_ia32_rolhi,
38082         __builtin_ia32_rorqi, __builtin_ia32_rorhi): Likewise.
38084 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38086         PR target/78012
38087         * lra-constraints.c (split_reg): Check requested split mode
38088         is supported by the register.
38090 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38092         * lra-constraints.c (simplify_operand_subreg): Remove early
38093         return false.
38095 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38097         PR target/78660
38098         * lra-constraints.c (curr_insn_transform): Tighten condition
38099         for converting SUBREG reloads from OP_OUT to OP_INOUT.
38101 2017-02-20  Matthew Fortune  <matthew.fortune@imgtec.com>
38103         PR target/78660
38104         * lra-constraints.c (curr_insn_transform): Handle
38105         WORD_REGISTER_OPERATIONS requirements when reloading SUBREGs.
38107 2017-02-19  Uros Bizjak  <ubizjak@gmail.com>
38109         Revert:
38110         2016-05-30  Uros Bizjak  <ubizjak@gmail.com>
38112         * config/i386/sync.md (mfence_nosse): Use "lock orl $0, -4(%esp)".
38114 2017-02-19  Jonathan Wakely  <jwakely@redhat.com>
38116         PR c++/69523
38117         * doc/invoke.texi (C++ Dialect Options) [-Wliteral-suffix]: Update
38118         description.
38120 2017-02-19  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38122         * gimple-pretty-print.c (dump_ternary_rhs): Adjust gimple dump format
38123         for FMA_EXPR.
38125 2017-02-18  Jakub Jelinek  <jakub@redhat.com>
38127         * final.c (last_columnnum, override_columnnum): New variables.
38128         (final_start_function): Set last_columnnum, pass it to begin_prologue
38129         hook and pass 0 to dwarf2out_begin_prologue.
38130         (final_scan_insn): Update override_columnnum.  Pass last_columnnum
38131         to source_line debug hook.
38132         (notice_source_line): Compute last_columnnum and for debug_column_info
38133         return true on column changes.
38134         * debug.h (struct gcc_debug_hooks): Add column argument to
38135         source_line and begin_prologue hooks.
38136         (debug_nothing_int_charstar_int_bool): Remove prototype.
38137         (debug_nothing_int_int_charstar,
38138         debug_nothing_int_int_charstar_int_bool): New prototypes.
38139         (dwarf2out_begin_prologue): Add column argument.
38140         * debug.c (do_nothing_debug_hooks): Adjust source_line and
38141         begin_prologue hooks.
38142         (debug_nothing_int_charstar_int_bool): Remove.
38143         (debug_nothing_int_int_charstar,
38144         debug_nothing_int_int_charstar_int_bool): New functions.
38145         * dwarf2out.c (dwarf2out_begin_prologue): Add column argument, pass it
38146         through to dwarf2out_source_line.
38147         (dwarf2_lineno_debug_hooks): Adjust begin_prologue hook.
38148         (dwarf2out_source_line): Add column argument, emit it if requested.
38149         * sdbout.c (sdbout_source_line, sdbout_begin_prologue): Add column
38150         arguments.
38151         * xcoffout.h (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38152         * xcoffout.c (xcoffout_begin_prologue, xcoffout_source_line): Likewise.
38153         * vmsdbgout.c (vmsdbgout_begin_prologue): Add column argument, pass it
38154         through to dwarf2out_begin_prologue.
38155         (vmsdbgout_source_line): Add column argument, pass it through to
38156         dwarf2out_source_line.
38157         * dbxout.c (dbxout_begin_prologue): Add column argument, adjust
38158         dbxout_source_line caller.
38159         (dbxout_source_line): Add column argument.
38161         * common.opt (gno-column-info, gcolumn-info): New options.
38162         * dwarf2out.c (dwarf2_lineno_debug_hooks): Formatting fix.
38163         (check_die): Also test for multiple DW_AT_decl_column attributes.
38164         (add_src_coords_attributes, dwarf2out_imported_module_or_decl_1): Add
38165         DW_AT_decl_column if requested.
38166         (gen_subprogram_die): Compare and/or add also DW_AT_decl_column
38167         if requested.
38168         (gen_variable_die): Likewise.
38169         (add_call_src_coords_attributes): Add DW_AT_call_column if requested.
38170         * doc/invoke.texi (-gcolumn-info, -gno-column-info): Document.
38172         PR target/79569
38173         * config/i386/i386.opt (m3dnowa): Replace Undocumented with Report.
38174         * common/config/i386/i386-common.c (OPTION_MASK_ISA_3DNOW_A_SET): Define.
38175         (ix86_handle_option): Handle OPT_m3dnowa.
38176         * doc/invoke.texi (-m3dnowa): Document.
38177         * doc/extend.texi (__builtin_ia32_pmulhuw, __builtin_ia32_pf2iw): Use
38178         -m3dnowa instead of -m3dnow -march=athlon.
38180         PR target/79559
38181         * config/i386/i386.c (ix86_print_operand): Use output_operand_lossage
38182         instead of gcc_assert for K, r and R code checks.  Formatting fixes.
38184 2017-02-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
38186         PR target/79261
38187         * config/rs6000/rs6000.c (rs6000_expand_ternop_builtin): Add
38188         support for CODE_FOR_vsx_xxpermdi_v2d[fi]_be.
38189         * config/rs6000/rs6000.md (reload_gpr_from_vsx<mode>): Call
38190         generator for vsx_xxpermdi_<mode>_be.
38191         * config/rs6000/vsx.md (vsx_xxpermdi_<mode>): Remove logic to
38192         force big-endian semantics.
38193         (vsx_xxpermdi_<mode>_be): New define_expand with same
38194         implementation as previous version of vsx_xxpermdi_<mode>.
38196 2017-02-17  Jakub Jelinek  <jakub@redhat.com>
38198         PR tree-optimization/79327
38199         * gimple-ssa-sprintf.c (format_integer): Remove likely_adjust
38200         variable, its initialization and use.
38202 2017-02-17  Julia Koval  <julia.koval@intel.com>
38204         * common/config/i386/i386-common.c (OPTION_MASK_ISA_RDPID_SET): New.
38205         (OPTION_MASK_ISA_PKU_UNSET): New.
38206         (ix86_handle_option): Handle -mrdpid.
38207         * config/i386/cpuid.h (bit_RDPID): New.
38208         * config/i386/driver-i386.c (host_detect_local_cpu):
38209         Detect RDPID feature.
38210         * config/i386/i386-builtin.def (__builtin_ia32_rdpid): New.
38211         * config/i386/i386-c.c (ix86_target_macros_internal):
38212         Handle RDPID flag.
38213         * config/i386/i386.c (ix86_target_string): Add -mrdpid to isa2_opts.
38214         (ix86_valid_target_attribute_inner_p): Add "rdpid".
38215         (ix86_expand_builtin): Handle IX86_BUILTIN_RDPID.
38216         * config/i386/i386.h (TARGET_RDPID, TARGET_RDPID_P): New.
38217         * config/i386/i386.md (define_insn "rdpid"): New.
38218         * config/i386/i386.opt Add -mrdpid.
38219         * config/i386/immintrin.h (_rdpid_u32): New.
38221 2017-02-17  Vladimir Makarov  <vmakarov@redhat.com>
38223         PR rtl-optimization/79541
38224         * lra-constraints.c (curr_insn_transform): Remove wrong asm insn
38225         instead of transforming it into USE.
38227 2017-02-17  Segher Boessenkool  <segher@kernel.crashing.org>
38229         * config/rs6000/rs6000.md (extendsfdf2): Remove default arguments.
38230         If HONOR_SNANS (SFmode) force the input to a register.
38231         (*extendsfdf2_fpr): Add !HONOR_SNANS (SFmode) condition.
38232         (*extendsfdf2_snan): New pattern, used when using SNaNs; it generates
38233         an frsp or similar insn.
38235 2017-02-17  Martin Liska  <mliska@suse.cz>
38237         PR rtl-optimization/79577
38238         * params.def (selsched-max-sched-times): Increase minimum to 1.
38240 2017-02-17  Martin Liska  <mliska@suse.cz>
38242         PR rtl-optimization/79574
38243         * gcse.c (want_to_gcse_p): Prevent integer overflow.
38245 2017-02-17  Martin Liska  <mliska@suse.cz>
38247         PR tree-optimization/79529
38248         * tree-ssa-loop-unswitch.c (is_maybe_undefined): Use
38249         ssa_defined_default_def_p to handle cases which are implicitly
38250         defined.
38251         * tree-ssa.c (ssa_defined_default_def_p): New function.
38252         (ssa_undefined_value_p): Use ssa_defined_default_def_p to handle cases
38253         which are implicitly defined.
38254         * tree-ssa.h (ssa_defined_default_def_p): Declare.
38256 2017-02-17  Richard Biener  <rguenther@suse.de>
38258         PR middle-end/79576
38259         * params.def (max-ssa-name-query-depth): Limit to 10.
38261 2017-02-17  Richard Biener  <rguenther@suse.de>
38263         PR tree-optimization/79552
38264         * tree-ssa-structalias.c (visit_loadstore): Properly verify
38265         default defs.
38267 2017-02-17  Richard Biener  <rguenther@suse.de>
38269         PR bootstrap/79567
38270         * genmatch.c (output_line_directive): Handle DIR_SEPARATOR_2.
38272 2017-02-17  Marek Polacek  <polacek@redhat.com>
38274         PR middle-end/79536
38275         * fold-const.c (fold_negate_expr_1): Renamed from fold_negate_expr.
38276         (fold_negate_expr): New wrapper.
38278 2017-02-16  Sandra Loosemore  <sandra@codesourcery.com>
38280         * doc/invoke.texi (C++ Dialect Options) [-Wno-non-template-friend]: 
38281         Correct terminology and de-emphasize pre-standard behavior.
38283 2017-02-16  Alan Modra  <amodra@gmail.com>
38285         PR rtl-optimization/79286
38286         * ira.c (def_dominates_uses): New function.
38287         (update_equiv_regs): Don't create an equivalence for insns that
38288         may trap where the register def does not dominate the use.
38290 2017-02-16  Vladimir Makarov  <vmakarov@redhat.com>
38292         PR rtl-optimization/78127
38293         * lra.c (lra): Call lra_eliminate before finish the loop after
38294         lra_constraint.
38296 2017-02-16  Richard Biener  <rguenther@suse.de>
38298         * graphite.h: Do not include isl/isl_val_gmp.h, instead include
38299         isl/isl_val.h.
38300         * graphite-isl-ast-to-gimple.c (gmp_cst_to_tree): Remove.
38301         (gcc_expression_from_isl_expr_int): Use generic isl_val interface.
38302         * graphite-sese-to-poly.c: Do not include isl/isl_val_gmp.h.
38303         (isl_val_int_from_wi): New function.
38304         (extract_affine_gmp): Rename to ...
38305         (extract_affine_wi): ... this, take a widest_int.
38306         (extract_affine_int): Just wrap extract_affine_wi.
38307         (add_param_constraints): Use isl_val_int_from_wi.
38308         (add_loop_constraints): Likewise, and extract_affine_wi.
38310 2017-02-15  Jeff Law  <law@redhat.com>
38312         PR middle-end/79521
38313         * ira-costs.c (scan_one_insn): Check have_regs_of_mode before calling
38314         ira_init_register_move_cost_if_necessary.
38316 2017-02-15  Martin Sebor  <msebor@redhat.com>
38318         PR middle-end/32003
38319         * doc/invoke.texi (-fdump-final-insns): Replace option accidentally
38320         removed in a prior commit.
38322 2017-02-15  Bin Cheng  <bin.cheng@arm.com>
38324         PR tree-optimization/79347
38325         * tree-vect-loop-manip.c (vect_do_peeling): Maintain profile
38326         counters during peeling.
38328 2017-02-15  Thomas Schwinge  <thomas@codesourcery.com>
38330         * Makefile.in (site.exp): Remove "set ISLVER".
38332 2017-02-15  Jakub Jelinek  <jakub@redhat.com>
38334         PR target/79487
38335         * real.c (real_from_integer): Call real_convert even for decimal.
38337 2017-02-15  Dominik Vogt  <vogt@linux.vnet.ibm.com>
38339         PR target/79421
38340         * config/s390/s390.c: define TARGET_CUSTOM_FUNCTION_DESCRIPTORS.
38342 2017-02-14  Andrew Pinski  <apinski@cavium.com>
38344         * config/aarch64/aarch64-cores.def (thunderx2t99): Move to under 'C"
38345         cores and change the partno/implementer to be correct.
38346         (thunderx2t99p1): New core which replaces thunderx2t99 and still has
38347         the 'B" as the implementer.
38348         * config/aarch64/aarch64-tune.md: Regenerate.
38350 2017-02-14  Carl Love  <cel@us.ibm.com>
38352         * config/rs6000/rs6000.c: Add case statement entry to make the
38353         xvcvuxdsp built-in argument unsigned.
38354         * config/rs6000/vsx.md: Fix the source and return operand types so they
38355         match the instruction definitions from the ISA document.  Fix typo
38356         in the instruction generation for the (define_insn "vsx_xvcvuxdsp"
38357         statement.
38359 2017-02-14  Vladimir Makarov  <vmakarov@redhat.com>
38361         PR target/79282
38362         * lra-int.h (struct lra_operand_data, struct lra_insn_reg): Add
38363         member early_clobber_alts.
38364         * lra-lives.c (reg_early_clobber_p): New.
38365         (process_bb_lives): Use it.
38366         * lra.c (new_insn_reg): New arg early_clobber_alts.  Use it.
38367         (debug_operand_data): Initialize early_clobber_alts.
38368         (setup_operand_alternative): Set up early_clobber_alts.
38369         (collect_non_operand_hard_regs): Ditto.  Pass early clobber
38370         alternatives to new_insn_reg.
38371         (add_regs_to_insn_regno_info): Add arg early_clobber_alts.  Use
38372         it.
38373         (lra_update_insn_regno_info): Pass the new arg.
38375 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38377         PR middle-end/79505
38378         * omp-offload.c (free_oacc_loop): Release loop->ifns vector.
38379         (new_oacc_loop_raw): Don't clear already cleared fields.
38381         PR target/79481
38382         * config/i386/avx512pfintrin.h (_mm512_prefetch_i32gather_pd,
38383         _mm512_prefetch_i32gather_ps, _mm512_prefetch_i64gather_pd,
38384         _mm512_prefetch_i64gather_ps): New inline functions and macros.
38386 2017-02-14  Uros Bizjak  <ubizjak@gmail.com>
38388         PR target/79495
38389         * config/i386/i386.md (*movxf_internal): Add (o,rC) alternative.
38391 2017-02-14  H.J. Lu  <hongjiu.lu@intel.com>
38393         PR target/79498
38394         * config/i386/i386.c (timode_scalar_chain::convert_insn): Insert
38395         the extra instruction to the right place to store 128-bit constant
38396         when needed.
38398 2017-02-14  Martin Sebor  <msebor@redhat.com>
38400         PR middle-end/79448
38401         * gimple-ssa-sprintf.c (format_directive): Avoid issuing INT_MAX
38402           warning for strings of unknown length.
38404 2017-02-13  Segher Boessenkool  <segher@kernel.crashing.org>
38406         * config.gcc (supported_defaults) [powerpc*-*-*]: Update.
38408 2017-02-14  Jeff Law  <law@redhat.com>
38410         PR target/79404
38411         * ira-costs.c (scan_one_insn): Initialize register move costs
38412         for pseudos seen in USE/CLOBBER insns.
38414         PR tree-optimization/79095
38415         * tree-vrp.c (extract_range_from_binary_expr_1): For EXACT_DIV_EXPR,
38416         if the numerator has the range ~[0,0] make the resultant range ~[0,0].
38417         (extract_range_from_binary_expr): For MINUS_EXPR with no derived range,
38418         if the operands are known to be not equal, then the resulting range
38419         is ~[0,0].
38420         (intersect_ranges): If the new range is ~[0,0] and the old range is
38421         wide, then prefer ~[0,0].
38422         * tree-vrp.c (overflow_comparison_p_1): New function.
38423         (overflow_comparison_p): New function.
38424         * tree-vrp.c (register_edge_assert_for_2): Register additional asserts
38425         if NAME is used in an overflow test.
38426         (vrp_evaluate_conditional_warnv_with_ops): If the ops represent an
38427         overflow check that can be expressed as an equality test, then adjust
38428         ops to be that equality test.
38430 2017-02-14  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
38432         * config/s390/s390-builtin-types.def: Remove flags argument.
38433         * config/s390/s390.c (s390_init_builtins): Likewise.
38435 2017-02-14  Martin Liska  <mliska@suse.cz>
38437         * tree-ssa-loop-unswitch.c (hoist_guard): Release get_loop_body
38438         vector.  Fix trailing white spaces.
38440 2017-02-14  James Greenhalgh  <james.greenhalgh@arm.com>
38442         * config/aarch64/aarch64.c (aarch64_simd_container_mode): Handle
38443         HFmode.
38445 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38447         PR rtl-optimization/68664
38448         * config/arm/arm.c (arm_sched_can_speculate_insn):
38449         New function.  Declare prototype.
38450         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38452 2017-02-14  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38454         PR rtl-optimization/68664
38455         * config/aarch64/aarch64.c (aarch64_sched_can_speculate_insn):
38456         New function.
38457         (TARGET_SCHED_CAN_SPECULATE_INSN): Define.
38459 2017-02-14  Amit Pawar  <amit.pawar@amd.com>
38461         * config/i386/i386.c (znver1_cost): Fix the alignment for function and
38462         max skip bytes for function, loop and jump.
38464 2017-02-14  Prathamesh Kulkarni  <prathamesh.kulkarni@linaro.org>
38466         * gimple-pretty-print.c (dump_unary_rhs): Adjust dump format for
38467         ABS_EXPR for gimple dump.
38469 2017-02-14  Jakub Jelinek  <jakub@redhat.com>
38471         PR target/79462
38472         * config/sh/sh.c (expand_cbranchdi4): Don't clear operands[4].
38474         PR tree-optimization/79408
38475         * tree-vrp.c (simplify_div_or_mod_using_ranges): Handle also the
38476         case when on TRUNC_MOD_EXPR op0 is INTEGER_CST.
38477         (simplify_stmt_using_ranges): Call simplify_div_or_mod_using_ranges
38478         also if rhs1 is INTEGER_CST.
38480 2017-02-14  Richard Biener  <rguenther@suse.de>
38482         PR middle-end/79432
38483         * tree-into-ssa.c (insert_phi_nodes): When the function can
38484         have abnormal edges rewrite SSA names with broken use-def
38485         dominance out of SSA and register them for PHI insertion.
38487 2017-02-13  Martin Sebor  <msebor@redhat.com>
38489         PR middle-end/79496
38490         * gimple-ssa-sprintf.c (pass_sprintf_length::handle_gimple_call): Avoid
38491         clearing info.nowrite flag when snprintf size argument is a range.
38493 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38495         * cprop.c (cprop_jump): Add missing space in string literal.
38496         * tree-ssa-structalias.c (rewrite_constraints): Likewise.
38497         (get_constraint_for_component_ref): Likewise.
38498         * df-core.c (df_worklist_dataflow_doublequeue): Likewise.
38499         * tree-outof-ssa.c (insert_partition_copy_on_edge): Likewise.
38500         * lra-constraints.c (process_alt_operands): Likewise.
38501         * ipa-inline.c (inline_small_functions): Likewise.
38502         * tree-ssa-sccvn.c (visit_reference_op_store): Likewise.
38503         * cgraph.c (cgraph_edge::redirect_call_stmt_to_callee): Likewise.
38504         * trans-mem.c (diagnose_tm_1_op): Likewise.
38505         * omp-grid.c (grid_find_single_omp_among_assignments): Likewise.
38506         (grid_parallel_clauses_gridifiable): Likewise.
38508         * config/nvptx/mkoffload.c (process): Add space in between
38509         , and %d.
38511         * config/i386/i386.h (REG_CLASS_NAMES): Add , in between
38512         "MOD4_SSE_REGS" and "ALL_REGS".
38514         * spellcheck.c (test_data): Add , in between "foo" and "food".
38516 2017-02-13  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38518         PR target/79449
38519         * config/rs6000/rs6000.c (expand_block_compare): Make sure runtime
38520         boundary crossing check and subsequent code generation agree.
38522 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38524         * config/aarch64/aarch64.c (has_memory_op): Delete.
38525         (aarch64_madd_needs_nop): Use contains_mem_rtx_p instead of
38526         has_memory_op.
38528 2017-02-13  Jakub Jelinek  <jakub@redhat.com>
38530         PR rtl-optimization/79388
38531         PR rtl-optimization/79450
38532         * combine.c (distribute_notes): When removing TEM_INSN for which
38533         corresponding dest has last value recorded, invalidate that last
38534         value.
38536 2017-02-13  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38538         * config/arm/arm.c (arm_print_tune_info): Use ASM_COMMENT_START instead
38539         of explicit '@'.  Add missing assembly comment marker on branch costs
38540         printout.
38542 2017-02-13  Nathan Sidwell  <nathan@acm.org>
38544         * gengtype-lex.l (<in_struct>): Add '/'.
38546 2017-02-13  Martin Liska  <mliska@suse.cz>
38548         PR c/79471
38549         * calls.c (expand_call): Replace XALLOCAVEC with XCNEWVEC.
38551 2017-02-13  Richard Biener  <rguenther@suse.de>
38553         * configure.ac (HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS):
38554         Remove.
38555         * configure: Re-generate.
38556         * config.in: Likewise.
38557         * graphite-dependences.c: Simplify as if
38558         HAVE_ISL_OPTIONS_SET_SCHEDULE_SERIALIZE_SCCS was defined.
38559         * graphite-isl-ast-to-gimple.c: Likewise.
38560         * graphite-optimize-isl.c: Likewise.
38561         * graphite-poly.c: Likewise.
38562         * graphite-sese-to-poly.c: Likewise.
38563         * graphite.h: Likewise.
38564         * toplev.c: Include isl/version.h and use isl_version () for
38565         printing the ISL version.
38566         * doc/install.texi: Update ISL requirement.
38568 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38570         * doc/standards.texi (Standards): Update reference to
38571         Objective-C 2.0.
38573 2017-02-12  Gerald Pfeifer  <gerald@pfeifer.com>
38575         * doc/extend.texi (Named Address Spaces): sourceware.org now
38576         defaults to https.
38577         * doc/install.texi (Binaries): Ditto.
38578         (Specific): Ditto.
38580 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38582         * doc/cpp.texi: Replace "stringify"/"stringification" with C
38583         standard terminology "stringize"/"stringizing" throughout.
38584         * doc/cppinternals.texi: Likewise.
38586 2017-02-11  Sandra Loosemore  <sandra@codesourcery.com>
38588         * doc/extend.texi: Fix some spelling mistakes and typos.
38589         * doc/invoke.texi: Likewise.
38591 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38593         PR ipa/79224
38594         * params.def (inline-min-speedup) Change from 10 to 8.
38596 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38598         * doc/invoke.texi (fopenmp): Bump OpenMP version from 4.0 to
38599         4.5.
38601 2017-02-11  Jan Hubicka  <hubicka@ucw.cz>
38603         PR ipa/79224
38604         * ipa-inline-analysis.c (get_minimal_bb): New function.
38605         (record_modified): Use it.
38606         (remap_edge_change_prob): Handle also ancestor functions.
38608 2017-02-11  Gerald Pfeifer  <gerald@pfeifer.com>
38610         * doc/contrib.texi (Contributors): Remove broken link into
38611         the Mauve CVS repository.
38613 2017-02-11  Jakub Jelinek  <jakub@redhat.com>
38615         PR middle-end/79454
38616         * internal-fn.c (expand_vector_ubsan_overflow): Use piece-wise
38617         result computation whenever lhs doesn't have vector mode, not
38618         just when it has BLKmode.
38620 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38622         * doc/makefile.texi (profiledbootstrap): Refer to the
38623         installation instructions only in textual form.
38625 2017-02-10  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
38627         PR target/79295
38628         * config/rs6000/altivec.md (bcd<bcd_add_sub>): Fix constraints.
38630 2017-02-10  Gerald Pfeifer  <gerald@pfeifer.com>
38632         * doc/install.texi (Specific): Use https for blackfin.uclinux.org.
38633         (Specific): Update mingw-w64 reference.
38634         (Binaries): Ditto.
38635         (Specific): Remove broken link to Renesas RX processor.
38637 2017-02-10  Richard Biener  <rguenther@suse.de>
38639         * toplev.c (process_options): Do not mention obsolete graphite
38640         options when printing sorry message about missing graphite support.
38641         Mention -floop-nest-optimize.
38643 2017-02-10  Christophe Lyon  <christophe.lyon@linaro.org>
38645         * config/aarch64/arm_neon.h (vtst_p8): Rewrite without asm.
38646         (vtst_p16): Likewise.
38647         (vtstq_p8): Likewise.
38648         (vtstq_p16): Likewise.
38649         (vtst_p64): New.
38650         (vtstq_p64): Likewise.
38651         * config/arm/arm_neon.h (vgetq_lane_p64): New.
38652         (vset_lane_p64): New.
38653         (vsetq_lane_p64): New.
38655 2017-02-10  Jakub Jelinek  <jakub@redhat.com>
38657         PR tree-optimization/79411
38658         * tree-ssa-reassoc.c (is_reassociable_op): Return false if
38659         stmt operands are SSA_NAMEs used in abnormal phis.
38660         (can_reassociate_p): Return false if op is SSA_NAME used in abnormal
38661         phis.
38663 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38665         PR ipa/70795
38666         * cgraphunit.c (cgraph_node::add_new_function): Set externally_visible
38667         flag if needed.
38669 2017-02-09  Jan Hubicka  <hubicka@ucw.cz>
38671         * tree-ssa-loop-unswitch.c (hoist_guard): Update profile.
38673 2017-02-09  Jakub Jelinek  <jakub@redhat.com>
38675         * omp-offload.c (oacc_loop_auto_partitions): Use || instead of |
38676         to avoid warning.
38678         PR c/79413
38679         * gimplify.h (is_gimple_sizepos): Only test for INTEGER_CST constants,
38680         not arbitrary TREE_CONSTANT.
38682         PR c/79431
38683         * gimplify.c (gimplify_adjust_omp_clauses): Ignore
38684         "omp declare target link" attribute unless is_global_var.
38685         * omp-offload.c (find_link_var_op): Likewise.
38687 2017-02-09  Nathan Sidwell  <nathan@codesourcery.com>
38688             Chung-Lin Tang  <cltang@codesourcery.com>
38690         * gimplify.c (gimplify_scan_omp_clauses): No special handling for
38691         OMP_CLAUSE_TILE.
38692         (gimplify_adjust_omp_clauses): Don't delete TILE.
38693         (gimplify_omp_for): Deal with TILE.
38694         * internal-fn.c (expand_GOACC_TILE): New function.
38695         * internal-fn.def (GOACC_DIM_POS): Comment may be overly conservative.
38696         (GOACC_TILE): New.
38697         * omp-expand.c (struct oacc_collapse): Add tile and outer fields.
38698         (expand_oacc_collapse_init): Add LOC paramter.  Initialize tile
38699         element fields.
38700         (expand_oacc_collapse_vars): Add INNER parm, adjust for tiling,
38701         avoid DIV for outermost collapse var.
38702         (expand_oacc_for): Insert tile element loop as needed.  Adjust.
38703         Remove out of date comments, fix whitespace.
38704         * omp-general.c (omp_extract_for_data): Deal with tiling.
38705         * omp-general.h (enum oacc_loop_flags): Add OLF_TILE flag,
38706         adjust OLF_DIM_BASE value.
38707         (struct omp_for_data): Add tiling field.
38708         * omp-low.c (scan_sharing_clauses): Allow OMP_CLAUSE_TILE.
38709         (lower_oacc_head_mark): Add OLF_TILE as appropriate.  Ensure 2 levels
38710         for auto loops.  Remove default auto determining, moved to
38711         oacc_loop_fixed_partitions.
38712         * omp-offload.c (struct oacc_loop): Change 'ifns' to vector of call
38713         stmts, add e_mask field.
38714         (oacc_dim_call): New function, abstracted out from oacc_thread_numbers.
38715         (oacc_thread_numbers): Use oacc_dim_call.
38716         (oacc_xform_tile): New.
38717         (new_oacc_loop_raw): Initialize e_mask, adjust for ifns vector.
38718         (finish_oacc_loop): Adjust for ifns vector.
38719         (oacc_loop_discover_walk): Append loop abstraction sites to list,
38720         add case for GOACC_TILE fns.
38721         (oacc_loop_xform_loop): Delete.
38722         (oacc_loop_process): Iterate over call list directly, and add
38723         handling for GOACC_TILE fns.
38724         (oacc_loop_fixed_partitions): Determine default auto, deal with TILE,
38725         dump partitioning.
38726         (oacc_loop_auto_partitions): Add outer_assign parm. Assign all but
38727         vector partitioning to outer loops.  Assign 2 partitions to loops
38728         when available. Add TILE handling.
38729         (oacc_loop_partition): Adjust oacc_loop_auto_partitions call.
38730         (execite_oacc_device_lower): Process GOACC_TILE fns,
38731         ignore unknown specs.
38732         * tree-nested.c (convert_nonlocal_omp_clauses): Allow OMP_CLAUSE_TILE.
38733         * tree.c (omp_clause_num_ops): Adjust TILE ops.
38734         * tree.h (OMP_CLAUSE_TILE_ITERVAR, OMP_CLAUSE_TILE_COUNT): New.
38736 2017-02-09  Gerald Pfeifer  <gerald@pfeifer.com>
38738         * configure.ac (ACX_BUGURL): Update.
38739         * configure: Regenerate.
38741 2017-02-09  Richard Biener  <rguenther@suse.de>
38743         PR tree-optimization/69823
38744         * graphite-scop-detection.c (scop_detection::harmful_loop_in_region):
38745         Properly enumerate all BBs in the region.  Use auto_vec/auto_bitmap.
38747 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38749         * config/arc/arc-c.def: Add __NPS400__ definition.
38750         * config/arc/arc.h (CPP_SPEC): Don't define __NPS400__ here.
38751         (TARGET_NPS400): Define.
38753 2017-02-09  Andrew Burgess  <andrew.burgess@embecosm.com>
38755         * config/arc/arc-arch.h (arc_arch_t): Move unchanged to earlier in
38756         file.
38757         (arc_cpu_t): Change base_architecture field, arch, to a arc_arc_t
38758         pointer, arch_info.
38759         (arc_cpu_types): Fill the arch_info field with a pointer into the
38760         arc_arch_types table.
38761         (arc_selected_cpu): Declare.
38762         * config/arc/arc.c (arc_selected_cpu): Make global.
38763         (arc_selected_arch): Delete.
38764         (arc_base_cpu): Delete.
38765         (arc_override_options): Remove references to deleted variables,
38766         update access to arch information.
38767         (ARC_OPT): Update access to arch information.
38768         (ARC_OPTX): Likewise.
38769         * config/arc/arc.h (arc_base_cpu): Remove declaration.
38770         (TARGET_ARC600): Update access to arch information.
38771         (TARGET_ARC601): Likewise.
38772         (TARGET_ARC700): Likewise.
38773         (TARGET_EM): Likewise.
38774         (TARGET_HS): Likewise.
38775         * config/arc/driver-arc.c (arc_cpu_to_as): Update access to arch
38776         information.
38778 2017-02-08  Pat Haugen  <pthaugen@us.ibm.com>
38780         PR target/78604
38781         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Invert
38782         condition/operands for integer GE/LE/GEU/LEU operations.
38784 2017-02-08  Segher Boessenkool  <segher@kernel.crashing.org>
38786         PR translation/79397
38787         * config/rs6000/rs6000.opt (maltivec=le, maltivec=be): Fix spelling
38788         of AltiVec.
38790 2017-02-08  Martin Jambor  <mjambor@suse.cz>
38792         PR ipa/79375
38793         * ipa-prop.c (ipa_alloc_node_params): Make static, return bool
38794         whether allocation happened.
38795         (ipa_initialize_node_params): Do not call ipa_alloc_node_params if
38796         nothing was allocated.
38798 2017-02-08  Jakub Jelinek  <jakub@redhat.com>
38800         PR tree-optimization/79408
38801         * tree-vrp.c (simplify_div_or_mod_using_ranges): If op1 is not
38802         constant, but SSA_NAME with a known integer range, use the minimum
38803         of that range instead of op1 to determine if modulo can be replaced
38804         with its first operand.
38806 2016-02-08  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
38808         * config/riscv/riscv.c (riscv_build_integer_1): Avoid use of INT16_MAX.
38810 2017-02-08  Richard Biener  <rguenther@suse.de>
38812         PR tree-optimization/71824
38813         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
38814         Check all loops contained in the merged region.
38816 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38818         * config/aarch64/aarch64.md (popcount<mode>2): New pattern.
38820 2017-02-07  Andrew Pinski  <apinski@cavium.com>
38822         * config/aarch64/aarch64-cores.def (thunderx): Disable LSE.
38823         (thunderxt88): Likewise.
38824         (thunderxt81): Disable LSE and change v8.1 to v8.
38825         (thunderxt83): Likewise.
38827 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38828             Richard Biener  <rguenther@suse.de>
38830         PR middle-end/79399
38831         * ira-int.h (struct target_ira_int): Change x_max_struct_costs_size
38832         type from int to size_t.
38833         * ira-costs.c (struct_costs_size): Change type from int to size_t.
38835 2017-02-07  Jakub Jelinek  <jakub@redhat.com>
38837         PR rtl-optimization/79386
38838         * cprop.c (bypass_conditional_jumps): Initialize
38839         bypass_last_basic_block already before splitting bbs after
38840         unconditional traps...
38841         (bypass_conditional_jumps): ... rather than here.
38843         PR target/79299
38844         * config/i386/sse.md (xtg_mode, gatherq_mode): New mode attrs.
38845         (*avx512f_gathersi<mode>, *avx512f_gathersi<mode>_2,
38846         *avx512f_gatherdi<mode>, *avx512f_gatherdi<mode>_2): Use them,
38847         fix -masm=intel patterns.
38849 2017-02-07  Richard Biener  <rguenther@suse.de>
38851         PR tree-optimization/79256
38852         PR middle-end/79278
38853         * builtins.c (get_object_alignment_2): Use min_align_of_type
38854         to extract alignment for MEM_REFs to honor BIGGEST_FIELD_ALIGNMENT
38855         and ADJUST_FIELD_ALIGN.
38857         * doc/tm.texi.in (ADJUST_FIELD_ALIGN): Adjust to take additional
38858         type parameter.
38859         * doc/tm.texi: Regenerate.
38860         * stor-layout.c (layout_decl): Adjust.
38861         (update_alignment_for_field): Likewise.
38862         (place_field): Likewise.
38863         (min_align_of_type): Likewise.
38864         * config/arc/arc.h (ADJUST_FIELD_ALIGN): Adjust.
38865         * config/epiphany/epiphany.h (ADJUST_FIELD_ALIGN): Likewise.
38866         * config/epiphany/epiphany.c (epiphany_adjust_field_align): Likewise.
38867         * config/frv/frv.h (ADJUST_FIELD_ALIGN): Likewise.
38868         * config/frv/frv.c (frv_adjust_field_align): Likewise.
38869         * config/i386/i386.h (ADJUST_FIELD_ALIGN): Likewise.
38870         * config/i386/i386.c (x86_field_alignment): Likewise.
38871         * config/rs6000/aix.h (ADJUST_FIELD_ALIGN): Likewise.
38872         * config/rs6000/darwin.h (ADJUST_FIELD_ALIGN): Likewise.
38873         * config/rs6000/freebsd64.h (ADJUST_FIELD_ALIGN): Likewise.
38874         * config/rs6000/linux64.h (ADJUST_FIELD_ALIGN): Likewise.
38875         * config/rs6000/sysv4.h (ADJUST_FIELD_ALIGN): Likewise.
38876         * config/rs6000/rs6000.c (rs6000_special_adjust_field_align_p):
38877          Likewise.
38879         Revert
38880         2017-01-30  Richard Biener  <rguenther@suse.de>
38882         PR tree-optimization/79256
38883         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
38884         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
38885         alignment on TYPE.
38887 2017-02-07  Toma Tabacu  <toma.tabacu@imgtec.com>
38889         * config/mips/mips.c (mips_expand_builtin_insn): Convert the QImode
38890         argument of the pshufh, psllh, psllw, psrah, psraw, psrlh, psrlw
38891         builtins to SImode and emit a zero-extend, if necessary.
38893 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38895         * docs/invoke.texi (RISC-V Options): Alphabetize.
38897 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38899         * doc/invoke.texi (RISC-V Options): Use two spaces to separate
38900         options.
38902 2017-02-06  Palmer Dabbelt  <palmer@dabbelt.com>
38904         * config/riscv/riscv.c: New file.
38905         * common/config/riscv/riscv-common.c: Likewise.
38906         * config.gcc: Likewise.
38907         * config/riscv/constraints.md: Likewise.
38908         * config/riscv/elf.h: Likewise.
38909         * config/riscv/generic.md: Likewise.
38910         * config/riscv/linux.h: Likewise.
38911         * config/riscv/multilib-generator: Likewise.
38912         * config/riscv/peephole.md: Likewise.
38913         * config/riscv/pic.md: Likewise.
38914         * config/riscv/predicates.md: Likewise.
38915         * config/riscv/riscv-builtins.c: Likewise.
38916         * config/riscv/riscv-c.c: Likewise.
38917         * config/riscv/riscv-ftypes.def: Likewise.
38918         * config/riscv/riscv-modes.def: Likewise.
38919         * config/riscv/riscv-opts.h: Likewise.
38920         * config/riscv/riscv-protos.h: Likewise.
38921         * config/riscv/riscv.h: Likewise.
38922         * config/riscv/riscv.md: Likewise.
38923         * config/riscv/riscv.opt: Likewise.
38924         * config/riscv/sync.md: Likewise.
38925         * config/riscv/t-elf-multilib: Likewise.
38926         * config/riscv/t-linux: Likewise.
38927         * config/riscv/t-linux-multilib: Likewise.
38928         * config/riscv/t-riscv: Likewise.
38929         * configure.ac: Likewise.
38930         * doc/contrib.texi: Add Kito Cheng, Palmer Dabbelt, and Andrew
38931         Waterman as RISC-V maintainers.
38932         * doc/install.texi: Add RISC-V entries.
38933         * doc/invoke.texi: Add RISC-V options section.
38934         * doc/md.texi: Add RISC-V constraints section.
38935         * configure: Regenerated.
38937 2017-02-06  Michael Meissner  <meissner@linux.vnet.ibm.com>
38939         PR target/66144
38940         * config/rs6000/vector.md (vcond<mode><mode>): Allow the true and
38941         false values to be constant vectors with all 0 or all 1 bits set.
38942         (vcondu<mode><mode>): Likewise.
38943         * config/rs6000/predicates.md (vector_int_reg_or_same_bit): New
38944         predicate.
38945         (fpmask_comparison_operator): Update comment.
38946         (vecint_comparison_operator): New predicate.
38947         * config/rs6000/rs6000.c (rs6000_emit_vector_cond_expr): Optimize
38948         vector conditionals when the true and false values are constant
38949         vectors with all 0 bits or all 1 bits set.
38951 2017-02-06  Martin Sebor  <msebor@redhat.com>
38953         PR  tree-optimization/79376
38954         * gimple-fold.c (get_range_strlen): Set the minimum length to zero.
38956 2017-02-06  Uros Bizjak  <ubizjak@gmail.com>
38958         * config/i386/sse.md (vector modes -> vec_extract* splitter): Use
38959         explicit subreg RTX with operand 1.  Use VECTOR_MODE_P predicate
38960         to simplify split condition.
38962 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
38964         * omp-expand.c (oxpand_omp_atomic_fetch_op,
38965         expand_omp_atomic_pipeline): Return false if can_atomic_load_p is
38966         false.
38968 2017-02-06  Segher Boessenkool  <segher@kernel.crashing.org>
38970         PR rtl-optimization/68664
38971         * target.def (can_speculate_insn): New hook.
38972         * doc/tm.texi.in (TARGET_SCHED_CAN_SPECULATE_INSN): New hook.
38973         * doc/tm.texi: Regenerate.
38974         * sched-rgn.c (can_schedule_ready_p): Use the new hook.
38975         * config/rs6000/rs6000.c (TARGET_SCHED_CAN_SPECULATE_INSN): New macro.
38976         (rs6000_sched_can_speculate_insn): New function.
38978 2017-02-06  Jakub Jelinek  <jakub@redhat.com>
38980         PR tree-optimization/79284
38981         * tree-vectorizer.h (VECT_SCALAR_BOOLEAN_TYPE_P): Define.
38982         * tree-vect-stmts.c (vect_get_vec_def_for_operand,
38983         vectorizable_mask_load_store, vectorizable_operation,
38984         vect_is_simple_cond, get_same_sized_vectype): Use it instead
38985         of comparing TREE_CODE of a type against BOOLEAN_TYPE.
38986         * tree-vect-patterns.c (check_bool_pattern, search_type_for_mask_1,
38987         vect_recog_bool_pattern, vect_recog_mask_conversion_pattern): Likewise.
38988         * tree-vect-slp.c (vect_get_constant_vectors): Likewise.
38989         * tree-vect-loop.c (vect_determine_vectorization_factor): Likewise.
38990         Remove redundant gimple_code (stmt) == GIMPLE_ASSIGN test after
38991         is_gimple_assign (stmt).  Replace another such test with
38992         is_gimple_assign (stmt).
38994 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
38996         PR target/78883
38997         * config/avr/avr.c (rtl-iter.h): Include it.
38998         (TARGET_LEGITIMATE_COMBINED_INSN): New hook define...
38999         (avr_legitimate_combined_insn): ...and implementation.
39001 2017-02-06  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39003         * config/s390/predicates.md ("larl_operand"): Use macros from hwint.h.
39004         * config/s390/s390.c (s390_const_operand_ok)
39005         (s390_canonicalize_comparison, s390_extract_part)
39006         (s390_single_part, s390_contiguous_bitmask_nowrap_p)
39007         (s390_contiguous_bitmask_p, s390_rtx_costs)
39008         (legitimize_pic_address): Likewise.
39009         * config/s390/s390.md ("clzdi2", "clztidi2"): Likewise.
39010         * config/s390/vx-builtins.md ("vec_genbytemaskv16qi")
39011         ("vec_permi<mode>", "vfae<mode>", "*vfaes<mode>", "vstrc<mode>")
39012         ("*vstrcs<mode>"): Use UINTVAL() to set unsigned HOST_WIDE_INT.
39013         * config/s390/vector.md ("vec_vfenes<mode>"): Likewise.
39015 2017-02-06  Georg-Johann Lay  <avr@gjlay.de>
39017         * config/avr/avr.md (*addhi3_zero_extend): Add alternative where
39018         REGNO($0) == REGNO($1).
39020 2017-02-06  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39022         * config/s390/linux.h(SIZE_TYPE): Add comment.
39024 2017-02-06  Julian Brown  <julian@codesourcery.com>
39025             Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39026             Virendra Pathak  <virendra.pathak@broadcom.com>
39028         * config/aarch64/aarch64-cores.def: Change the scheduler
39029         to Thunderx2t99.
39030         * config/aarch64/aarch64.md: Include thunderx2t99.md.
39031         * config/aarch64/thunderx2t99.md: New file.
39033 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39035         * doc/standards.texi (Go Language): Update link to language
39036         standard.
39038 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39040         * tree-eh.c (lower_resx): Sanitize profile.
39041         (cleanup_empty_eh_move_lp): Likewise.
39043 2017-02-05  Jan Hubicka  <hubicka@ucw.cz>
39045         PR tree-ssa/79347
39046         * cfgloopmanip.c (lv_adjust_loop_entry_edge, loop_version): Add
39047         ELSE_PROB.
39048         * cfgloopmanip.h (loop_version): Update prototype.
39049         * modulo-sched.c (sms_schedule): Update call of loop_version.
39050         * tree-if-conv.c(version_loop_for_if_conversion): Likewise.
39051         * tree-parloops.c (gen_parallel_loop): Likewise.
39052         * tree-ssa-loop-manip.c (tree_transform_and_unroll_loop): Likewise.
39053         * tree-ssa-loop-split.c (split_loop): Likewise.
39054         * tree-ssa-loop-unswitch.c (tree_unswitch_loop): Likewise.
39055         * tree-vect-loop-manip.c (vect_loop_versioning): Likewise.
39057 2017-02-05  Martin Liska  <mliska@suse.cz>
39059         PR bootstrap/78985
39060         * config/s390/s390.c (s390_gimplify_va_arg): Initialize local
39061         variable to NULL.
39062         (print_operand_address): Initialize a struct to zero.
39064 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39066         * doc/contrib.texi (Contributors): Refer to Hans Boehm's
39067         garbage collector only in textual form.
39069 2017-02-05  Gerald Pfeifer  <gerald@pfeifer.com>
39071         * doc/extend.texi (x86 specific memory model extensions for
39072         transactional memory): Simplify a phrase.
39074 2017-02-05  Eric Botcazou  <ebotcazou@adacore.com>
39076         PR target/79353
39077         * config/sparc/sync.md (atomic_loaddi_1): Replace 'U' constraint with
39078         'r', 'm' constraint with 'T' and !TARGET_ARCH64 with TARGET_ARCH32.
39079         (atomic_storedi_1): Likewise.
39081 2017-02-04  Jakub Jelinek  <jakub@redhat.com>
39083         PR tree-optimization/79338
39084         * tree-parloops.c (gather_scalar_reductions): Don't call
39085         vect_analyze_loop_form for loop->inner before destroying loop's
39086         loop_vinfo.
39088 2017-02-03  Martin Sebor  <msebor@redhat.com>
39090         PR tree-optimization/79327
39091         * gimple-ssa-sprintf.c (tree_digits): Avoid adding the base prefix
39092         when precision has resulted in leading zeros.
39093         (format_integer): Adjust the likely counter to assume an unknown
39094         argument that may be zero is non-zero.
39096 2017-02-03  Jason Merrill  <jason@redhat.com>
39098         PR c++/78689
39099         * tree-inline.c (copy_tree_body_r) [COND_EXPR]: Revert change to
39100         avoid copying non-taken branch.
39102 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39104         PR tree-optimization/79340
39105         * tree-vect-loop.c (vectorizable_reduction): Release
39106         vec_defs elements after safe_splicing them into other vectors.
39107         Formatting fixes.
39109         PR tree-optimization/79327
39110         * gimple-ssa-sprintf.c (adjust_range_for_overflow): If returning
39111         true, always set *argmin and *argmax to TYPE_{MIN,MAX}_VALUE of
39112         dirtype.
39113         (format_integer): Use wide_int_to_tree instead of build_int_cst
39114         + to_?hwi.  If argmin is NULL, just set argmin and argmax to
39115         TYPE_{MIN,MAX}_VALUE of argtype.  Simplify and fix computation
39116         of shortest and longest sequence.
39118 2017-02-03  Uros Bizjak  <ubizjak@gmail.com>
39120         * config/i386/i386.c (dimode_scalar_chain::convert_reg):
39121         Use pextrd for TARGET_SSE4_1 when creating scalar copy.
39123 2017-02-03  Walter Lee  <walt@tilera.com>
39125         PR target/78862
39126         * config/tilegx/tilegx.md (tilegx_expand_prologue): Add blockage
39127         after initial stackframe link reg save.
39128         * config/tilepro/tilepro.md (tilepro_expand_prologue): Likewise.
39130 2017-02-03  Jakub Jelinek  <jakub@redhat.com>
39132         PR target/79354
39133         * config/rs6000/rs6000.md (movsi_from_sf): Use wb constraint instead of
39134         wu for stxssp alternative.
39136 2017-02-03  Martin Sebor  <msebor@redhat.com>
39138         PR tree-optimization/79352
39139         * gimple-fold.c (get_range_strlen): Add argument.
39140         (get_range_strlen): Change return type to bool.
39141         (get_maxval_strlen): Pass in a dummy argument.
39142         * gimple-fold.h (get_range_strlen): Change return type to bool.
39143         * gimple-ssa-sprintf.c (get_string_length): Set unlikely counter.
39144         * tree.h (array_at_struct_end_p): Add argument.
39145         * tree.c (array_at_struct_end_p): Handle it.
39147 2017-02-03  Martin Liska  <mliska@suse.cz>
39149         PR lto/66295
39150         * multiple_target.c (create_dispatcher_calls): Redirect edge
39151         from a caller of a dispatcher.
39152         (expand_target_clones): Make the clones local.
39153         (ipa_target_clone): Do both target clones and resolvers.
39154         (ipa_dispatcher_calls): Remove the pass.
39155         (pass_dispatcher_calls::gate): Likewise.
39156         (make_pass_dispatcher_calls): Likewise.
39157         * passes.def (pass_target_clone): Put as very first IPA early
39158         pass.
39160 2017-02-03  Martin Liska  <mliska@suse.cz>
39162         * symtab.c (symtab_node::binds_to_current_def_p): Bail out
39163         in case of a function with ifunc attribute.
39165 2017-02-03  Martin Liska  <mliska@suse.cz>
39167         * cgraph.c (cgraph_node::dump): Dump function version info.
39168         * symtab.c (symtab_node::dump_base): Add missing new line.
39170 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39172         * tree-ssa-ifcombine.c (update_profile_after_ifcombine): New function.
39173         (ifcombine_ifandif): Use it.
39175 2017-02-03  Martin Liska  <mliska@suse.cz>
39177         * doc/invoke.texi: Document default value for
39178         use-after-scope-direct-emission-threshold.
39180 2017-02-03  Martin Liska  <mliska@suse.cz>
39182         PR tree-optimization/79339
39183         * gimple-ssa-sprintf.c (format_floating_max): Call mpfr_clear.
39184         (format_floating): Likewise.
39186 2017-02-03  Martin Liska  <mliska@suse.cz>
39188         PR ipa/79337
39189         * ipa-prop.c (ipa_node_params_t::insert): Remove current
39190         implementation.
39191         (ipa_node_params_t::remove): Likewise.
39192         * ipa-prop.h (ipa_node_params::ipa_node_params): Make default
39193         initialization from removed ipa_node_params_t::insert.
39194         (ipa_node_params::~ipa_node_params): Move from removed
39195         ipa_node_params_t::release.
39196         * symbol-summary.h (symbol_summary::m_released): New member.
39197         Do not release a summary twice.  Do not allow to call finalizer
39198         for types of a summary that live in GGC memory.
39200 2017-02-02  Naveen H.S  <Naveen.Hurugalawadi@cavium.com>
39202         * config/aarch64/aarch64.c (thunderx2t99_tunings): Enable AES and
39203         cmp_branch fusion.
39205 2017-02-02  Martin Sebor  <msebor@redhat.com>
39207         PR middle-end/79275
39208         * gimple-ssa-sprintf.c (get_string_length): Set lower bound to zero.
39209         (format_string): Tighten up the range of output for non-constant
39210         strings and correct the expected range for wide non-constant strings.
39212 2017-02-02  Martin Sebor  <msebor@redhat.com>
39214         * doc/invoke.texi (-maccumulate-args): Fix bad grammar.
39216         PR middle-end/32003
39217         * doc/invoke.texi (-fdump-tree-): Remove pass-specific options from
39218         index.
39219         (-fdump-tree-@var): Add to index and document how to come up
39220         with pass-specific option and dump file names.
39221         (-fdump-passes): Clarify where to look for output.
39223 2017-02-02  Jan Hubicka  <hubicka@ucw.cz>
39225         PR middle-end/77445
39226         * tree-ssa-threadbackward.c (profitable_jump_thread_path): Dump
39227         statistics of the analyzed path; allow threading for speed when
39228         any of BBs along the path are optimized for speed.
39230 2017-02-02  Eric Botcazou  <ebotcazou@adacore.com>
39232         PR middle-end/78468
39233         * emit-rtl.c (init_emit): Add ??? comment for problematic alignment
39234         settings of the virtual registers.
39236         Revert again
39237         2016-08-23  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39239         * explow.c (get_dynamic_stack_size): Take known alignment of stack
39240         pointer + STACK_DYNAMIC_OFFSET into account when calculating the size
39241         needed.
39243 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39245         * config/s390/vx-builtins.md ("vec_ceil", "vec_floor")
39246         ("vec_trunc", "vec_roundc", "vec_round"): Remove expanders.
39248 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39250         * config/s390/s390.md: Add missing comments with the expanded
39251         mnemonics.
39252         * config/s390/vector.md: Likewise.
39253         * config/s390/vx-builtins.md: Likewise.
39255 2017-02-02  Jakub Jelinek  <jakub@redhat.com>
39257         PR target/79197
39258         * config/rs6000/rs6000.md (*fixuns_trunc<mode>di2_fctiduz): Rename to ...
39259         (fixuns_trunc<mode>di2): ... this, remove previous expander.  Put all
39260         conditions on a single line.
39262 2017-02-02  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39264         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39265         __S390_VX__ to __VX__.
39267 2017-02-01  Andrew Pinski  <apinski@cavium.com>
39269         * tree-vect-loop.c (vect_compute_single_scalar_iteration_cost): Pass
39270         stmt_info to record_stmt_cost.
39271         (vect_get_known_peeling_cost): Pass stmt_info if known to
39272         record_stmt_cost.
39273         * config/aarch64/aarch64-protos.h (cpu_vector_cost): Split
39274         cpu_vector_cost field into
39275         scalar_int_stmt_cost and scalar_fp_stmt_cost.  Split vec_stmt_cost
39276         field into vec_int_stmt_cost and vec_fp_stmt_cost.
39277         * config/aarch64/aarch64.c (generic_vector_cost): Update for the
39278         splitting of scalar_stmt_cost and vec_stmt_cost.
39279         (thunderx_vector_cost): Likewise.
39280         (cortexa57_vector_cost): LIkewise.
39281         (exynosm1_vector_cost): Likewise.
39282         (xgene1_vector_cost): Likewise.
39283         (thunderx2t99_vector_cost): Improve after the splitting of the two
39284         fields.
39285         (aarch64_builtin_vectorization_cost): Update for the splitting of
39286         scalar_stmt_cost and vec_stmt_cost.
39288 2017-02-01  Torvald Riegel  <triegel@redhat.com>
39289             Richard Henderson  <rth@redhat.com>
39291         * builtins.c (fold_builtin_atomic_always_lock_free): Make "lock-free"
39292         conditional on existance of a fast atomic load.
39293         * optabs-query.c (can_atomic_load_p): New function.
39294         * optabs-query.h (can_atomic_load_p): Declare it.
39295         * optabs.c (expand_atomic_exchange): Always delegate to libatomic if
39296         no fast atomic load is available for the particular size of access.
39297         (expand_atomic_compare_and_swap): Likewise.
39298         (expand_atomic_load): Likewise.
39299         (expand_atomic_store): Likewise.
39300         (expand_atomic_fetch_op): Likewise.
39301         * testsuite/lib/target-supports.exp
39302         (check_effective_target_sync_int_128): Remove x86 because it provides
39303         no fast atomic load.
39304         (check_effective_target_sync_int_128_runtime): Likewise.
39306 2017-02-01  Richard Biener  <rguenther@suse.de>
39308         * graphite.c: Include tree-vectorizer.h for find_loop_location.
39309         (graphite_transform_loops): Provide opt-info for optimized nests.
39310         * tree-parloop.c (parallelize_loops): Provide opt-info for
39311         parallelized loops.
39313 2017-02-01  Richard Biener  <rguenther@suse.de>
39315         PR middle-end/79315
39316         * tree-cfg.c (move_stmt_op): Never set TREE_BLOCK when it
39317         was not set before.
39319 2017-02-01  Richard Biener  <rguenther@suse.de>
39321         PR tree-optimization/71824
39322         * graphite-scop-detection.c (scop_detection::build_scop_breadth):
39323         Verify the loops are valid in the merged SESE region.
39324         (scop_detection::can_represent_loop_1): Check analyzing the
39325         evolution of the number of iterations in the region succeeds.
39327 2017-01-31  Ian Lance Taylor  <iant@golang.org>
39329         * config/i386/i386.c (ix86_expand_split_stack_prologue): Add
39330         REG_ARGS_SIZE note to 32-bit push insns and call insn.
39332 2017-01-31  David Malcolm  <dmalcolm@redhat.com>
39334         PR preprocessor/79210
39335         * input.c (get_substring_ranges_for_loc): Replace line_width
39336         assertion with error-handling.
39338 2017-01-31  Richard Biener  <rguenther@suse.de>
39340         PR tree-optimization/77318
39341         * graphite-sese-to-poly.c (extract_affine): Fix assert.
39342         (create_pw_aff_from_tree): Take loop parameter.
39343         (add_condition_to_pbb): Pass loop of the condition to
39344         create_pw_aff_from_tree.
39346 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39348         * config/s390/s390.c (s390_asan_shadow_offset): New function.
39349         (TARGET_ASAN_SHADOW_OFFSET): Redefine.
39351 2017-01-31  Michael Meissner  <meissner@linux.vnet.ibm.com>
39353         PR target/78597
39354         PR target/79038
39355         * config/rs6000/rs6000-protos.h (convert_float128_to_int): Delete,
39356         no longer used.
39357         (convert_int_to_float128): Likewise.
39358         * config/rs6000/rs6000.c (convert_float128_to_int): Likewise.
39359         (convert_int_to_float128): Likewise.
39360         * config/rs6000/rs6000.md (UNSPEC_IEEE128_MOVE): Likewise.
39361         (UNSPEC_IEEE128_CONVERT): Likewise.
39362         (floatsi<mode>2, FLOAT128 iterator): Bypass calling
39363         rs6000_expand_float128_convert if we have IEEE 128-bit hardware.
39364         Use local variables for IBM extended format.
39365         (fix_trunc<mode>si2, FLOAT128 iterator): Likewise.
39366         (fix_trunc<mode>si2_fprs): Likewise.
39367         (fixuns_trunc<IEEE128:mode><SDI:mode>2): Likewise.
39368         (floatuns<IEEE128:mode>2, IEEE128 iterator): Likewise.
39369         (fix<uns>_<mode>si2_hw): Rework the IEEE 128-bt hardware support
39370         to know that we can now have integers of all sizes in vector
39371         registers.
39372         (fix<uns>_<mode>di2_hw): Likewise.
39373         (float<uns>_<mode>si2_hw): Likewise.
39374         (fix_<mode>si2_hw): Likewise.
39375         (fixuns_<mode>si2_hw): Likewise.
39376         (float<uns>_<mode>di2_hw): Likewise.
39377         (float_<mode>di2_hw): Likewise.
39378         (float_<mode>si2_hw): Likewise.
39379         (floatuns_<mode>di2_hw): Likewise.
39380         (floatuns_<mode>si2_hw): Likewise.
39381         (xscvqp<su>wz_<mode>): Delete, no longer used.
39382         (xscvqp<su>dz_<mode>): Likewise.
39383         (xscv<su>dqp_<mode>): Likewise.
39384         (ieee128_mfvsrd_64bit): Likewise.
39385         (ieee128_mfvsrd_32bit): Likewise.
39386         (ieee128_mfvsrwz): Likewise.
39387         (ieee128_mtvsrw): Likewise.
39388         (ieee128_mtvsrd_64bit): Likewise.
39389         (ieee128_mtvsrd_32bit): Likewise.
39391 2017-01-31  Martin Liska  <mliska@suse.cz>
39393         PR ipa/79285
39394         * ipa-prop.c (ipa_free_all_node_params): Call release method
39395         instead of ~sumbol_summary to not to trigger double times
39396         dtor of hash_map.
39398 2017-01-31  Aldy Hernandez  <aldyh@redhat.com>
39400         PR tree-optimization/71691
39401         * bitmap.h (class auto_bitmap): New.
39402         * tree-ssa-loop-unswitch.c (tree_may_unswitch_on): Call
39403         is_maybe_undefined instead of ssa_undefined_value_p.
39405 2017-01-31  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
39407         * config/s390/s390-c.c (s390_cpu_cpp_builtins_internal): Rename
39408         __S390_ARCH_LEVEL__ to __ARCH__.
39410 2017-01-31  Jakub Jelinek  <jakub@redhat.com>
39412         PR tree-optimization/79267
39413         * value-prof.c (gimple_ic): Only drop lhs for noreturn calls
39414         if should_remove_lhs_p is true.
39416 2017-01-30  Alexandre Oliva  <aoliva@redhat.com>
39418         PR debug/63238
39419         * dwarf2out.c (clone_as_declaration): Drop DW_AT_alignment.
39420         (add_alignment_attribute): New.
39421         (base_type_die): Add alignment attribute.
39422         (subrange_type_die): Likewise.
39423         (modified_type_die): Likewise.
39424         (gen_array_type_die): Likewise.
39425         (gen_descr_array_type_die: Likewise.
39426         (gen_enumeration_type_die): Likewise.
39427         (gen_subprogram_die): Likewise.
39428         (gen_variable_die): Likewise.
39429         (gen_field_die): Likewise.
39430         (gen_ptr_to_mbr_type_die): Likewise.
39431         (gen_struct_or_union_type_die): Likewise.
39432         (gen_subroutine_type_die): Likewise.
39433         (gen_typedef_die): Likewise.
39434         (base_type_cmp): Compare alignment attribute.
39436 2017-01-30  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
39438         PR target/79170
39439         * config/rs6000/altivec.md (*setb_internal): Rename to setb_signed.
39440         (setb_unsigned) New pattern for setb with CCUNS.
39441         * config/rs6000/rs6000.c (expand_block_compare): Use a different
39442         subfc./subfe sequence to avoid overflow problems.  Generate a
39443         shorter sequence with cmpld/setb for power9.
39444         * config/rs6000/rs6000.md (subf<mode>3_carry_dot2): Add a new pattern
39445         for generating subfc. instruction.
39446         (cmpstrsi): Add TARGET_POPCNTD predicate as the generate sequence
39447         now uses this instruction.
39449 2017-01-30  Ian Lance Taylor  <iant@google.com>
39451         PR debug/79289
39452         * dwarf2out.c (gen_type_die_with_usage): When picking a variant
39453         for FUNCTION_TYPE/METHOD_TYPE, use the first matching one.
39455 2017-01-30  Martin Sebor  <msebor@redhat.com>
39457         * gimple-ssa-sprintf.c (fmtresult::adjust_for_width_or_precision):
39458         Move constant to the right of a relational operator.
39459         (get_mpfr_format_length, format_character, format_string): Ditto.
39460         (should_warn_p, maybe_warn): Same.
39462         * doc/invoke.texi (-Wformat-truncation=1): Fix typo.
39464 2017-01-30  Maxim Ostapenko  <m.ostapenko@samsung.com>
39466         PR lto/79061
39467         * asan.c (get_translation_unit_decl): Remove function.
39468         (asan_add_global): Force has_dynamic_init to zero in LTO mode.
39470 2017-01-30  Martin Liska  <mliska@suse.cz>
39472         PR gcov-profile/79259
39473         * opts.c (common_handle_option): Enable flag_ipa_bit_cp w/
39474         -fprofile-generate.
39476 2017-01-30  Martin Liska  <mliska@suse.cz>
39478         PR bootstrap/78985
39479         * config/aarch64/cortex-a57-fma-steering.c (func_fma_steering::analyze):
39480         Initialize variables with NULL value.
39482 2017-01-30  Richard Earnshaw  <rearnsh@arm.com>
39484         PR target/79260
39485         * config.gcc (arm*-*-*): Add arm/arm-flags.h and arm/arm-isa.h to
39486         tm_p_file.
39487         * arm/arm-protos.h: Don't directly include arm-flags.h and arm-isa.h.
39489 2017-01-30  Richard Biener  <rguenther@suse.de>
39491         PR tree-optimization/79276
39492         * tree-vrp.c (process_assert_insertions): Properly adjust common
39493         when removing a duplicate.
39495 2017-01-30  Richard Biener  <rguenther@suse.de>
39497         PR tree-optimization/79256
39498         * targhooks.c (default_builtin_vector_alignment_reachable): Honor
39499         BIGGEST_FIELD_ALIGNMENT and ADJUST_FIELD_ALIGN to fix up bogus
39500         alignment on TYPE.
39501         * tree.c (build_aligned_type): Set TYPE_USER_ALIGN.
39503 2017-01-30  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39505         PR target/79240
39506         * config/s390/s390.md ("*r<noxa>sbg_<mode>_srl_bitmask")
39507         ("*r<noxa>sbg_<mode>_sll_bitmask")
39508         ("*extzv_<mode>_srl<clobbercc_or_nocc>")
39509         ("*extzv_<mode>_sll<clobbercc_or_nocc>"):
39510         Use contiguous_bitmask_nowrap_operand.
39512 2017-01-29  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
39514         PR target/79268
39515         * config/rs6000/altivec.h (vec_xl): Revise #define.
39516         (vec_xst): Likewise.
39518 2017-01-27  Uros Bizjak  <ubizjak@gmail.com>
39520         * config/i386/i386.c (print_reg): Use REGNO instead of true_regnum.
39522 2017-01-27  Bernd Schmidt  <bschmidt@redhat.com>
39524         PR rtl-optimization/79194
39525         * cprop.c (one_cprop_pass): Move deletion of code after unconditional
39526         traps before call to bypass_conditional_jumps.
39528 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39530         PR tree-optimization/71374
39531         * lra-constraints.c (check_conflict_input_operands): New.
39532         (match_reload): Use it.
39534 2017-01-27  Vladimir Makarov  <vmakarov@redhat.com>
39536         PR target/79131
39537         * lra-assigns.c (find_hard_regno_for_1): Take endianess for into
39538         account to calculate conflict_set.
39540 2017-01-27  Bin Cheng  <bin.cheng@arm.com>
39542         PR rtl-optimization/78559
39543         * combine.c (try_combine): Discard REG_EQUAL and REG_EQUIV for
39544         other_insn in combine.
39546 2017-01-27  Pekka Jääskeläinen  <pekka.jaaskelainen@parmance.com>
39548         * builtin-types.def: Use unsigned_char_type_node for BT_UINT8.  Use
39549         uint16_type_node for BT_UINT16.
39551 2017-01-27  David Malcolm  <dmalcolm@redhat.com>
39553         * doc/sourcebuild.texi (Testsuites): Add "GIMPLE Tests" and
39554         "RTL Tests" to menu.
39555         (GIMPLE Tests): New node.
39556         (RTL Tests): New node.
39558 2017-01-27  Richard Biener  <rguenther@suse.de>
39560         PR tree-optimization/79245
39561         * tree-loop-distribution.c (distribute_loop): Apply cost
39562         modeling also to detected patterns.
39564 2017-01-27  Richard Biener  <rguenther@suse.de>
39566         PR tree-optimization/71433
39567         * tree-vrp.c (register_new_assert_for): Revert earlier changes.
39568         (compare_assert_loc): New function.
39569         (process_assert_insertions): Sort and optimize assert locations
39570         to remove duplicates and push down identical assertions on
39571         edges to their destination block.
39573 2017-01-27  Richard Biener  <rguenther@suse.de>
39575         PR tree-optimization/79244
39576         * tree-vrp.c (remove_range_assertions): Forcefully propagate
39577         out SSA names even if abnormal.
39579 2017-01-27  Jakub Jelinek  <jakub@redhat.com>
39581         * realmpfr.h: Poison MPFR_RND{N,Z,U,D}.
39582         * gimple-ssa-sprintf.c (format_floating_max): Use GMP_RNDN
39583         instead of MPFR_RNDN.
39585 2017-01-27  Richard Earnshaw  <rearnsha@arm.com>
39587         PR target/79239
39588         * arm.c (arm_option_override): Don't call build_target_option_node
39589         until after doing all option overrides.
39590         (arm_valid_target_attribute_tree): Likewise.
39592 2017-01-27  Martin Liska  <mliska@suse.cz>
39594         * doc/invoke.texi (-fprofile-arcs): Document profiling support
39595         for {cd}tors and C++ {cd}tors.
39597 2017-01-27  Dominik Vogt  <vogt@linux.vnet.ibm.com>
39599         * config/s390/s390.md ("*setmem_long_and")
39600         ("*setmem_long_and_31z"): Use zero_extend instead of and.
39602 2017-01-26  Martin Sebor  <msebor@redhat.com>
39604         * gimple-ssa-sprintf.c (format_floating): Simplify the computation
39605         of precision.
39607 2017-01-26  Martin Sebor  <msebor@redhat.com>
39609         * gimple-ssa-sprintf.c (format_floating): Test HAVE_XFmode and
39610         HAVE_DFmode before using XFmode or DFmode.
39611         (parse_directive): Avoid using the z length modifier to avoid
39612         the ISO C++98 does not support the â€˜z’ gnu_printf length modifier.
39614         PR middle-end/78703
39615         * gimple-ssa-sprintf.c (adjust_for_width_or_precision): Change
39616         to accept adjustment as an array.
39617         (get_int_range): New function.
39618         (struct directive): Make width and prec arrays.
39619         (directive::set_width, directive::set_precision): Call get_int_range.
39620         (format_integer, format_floating): Handle width and precision ranges.
39621         (format_string, parse_directive): Same.
39623 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39625         PR debug/79129
39626         * dwarf2out.c (generate_skeleton_bottom_up): For children with
39627         comdat_type_p set, just clone them, but keep the children in the
39628         original DIE.
39630         PR debug/78835
39631         * dwarf2out.c (prune_unused_types): Mark all functions with DIEs
39632         which have direct callers with -fvar-tracking-assignments enabled
39633         in the current TU.
39634         (resolve_addr): Avoid adding skeleton DIEs for DW_AT_call_origin
39635         inside of type units.
39637 2017-01-26  Martin Sebor  <msebor@redhat.com>
39639         PR middle-end/78703
39640         * gimple-ssa-sprintf.c (struct result_range): Add likely and
39641         unlikely counters.
39642         (struct format_result): Replace number_chars, number_chars_min,
39643         and number_chars_max with a single member of struct result_range.
39644         Remove bounded.
39645         (format_result::operator+=): Adjust.
39646         (struct fmtresult): Remove bounded.  Handle likely and unlikely
39647         counters.
39648         (fmtresult::adjust_for_width_or_precision): New function.
39649         (fmtresult:type_max_digits): New function.
39650         (bytes_remaining): Handle likely and unlikely counters.
39651         (min_bytes_remaining): Remove.
39652         (format_percent): Simplify.
39653         (format_integer, format_floating): Set likely and unlikely counters.
39654         (get_string_length, format_character, format_string): Same.
39655         (format_plain, should_warn_p): New function.
39656         (maybe_warn): Call should_warn_p.  Update diagnostic messages
39657         and handle those for all directives, including plain strings.
39658         (format_directive): Handle likely and unlikely counters.
39659         Remove unnecessary quoting from diagnostics.  Add an informational
39660         note.
39661         (add_bytes): Remove.
39662         (pass_sprintf_length::compute_format_length): Simplify.
39663         (try_substitute_return_value): Handle likely and unlikely counters.
39665 2017-01-26  Carl Love  <cel@us.ibm.com>
39667         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Remove
39668         bogus entries for the P8V_BUILTIN_VEC_VGBBD built-ins
39670 2017-01-26  Vladimir Makarov  <vmakarov@redhat.com>
39672         PR target/79131
39673         * lra-assigns.c (setup_live_pseudos_and_spill_after_risky): Take
39674         endianess for subregs into account.
39675         * lra-constraints.c (lra_constraints): Do risky transformations
39676         always on the first iteration.
39677         * lra-lives.c (check_pseudos_live_through_calls): Add arg
39678         last_call_used_reg_set.
39679         (process_bb_lives): Define and use last_call_used_reg_set.
39680         * lra.c (lra): Always continue after lra_constraints on the first
39681         iteration.
39683 2017-01-26  Kirill Yukhin  <kirill.yukhin@gmail.com>
39685         * gcc.target/i386/avx512bw-kshiftlq-2.c: Use unsigned long long
39686         constant.
39687         * gcc.target/i386/avx512bw-kshiftrq-2.c: Ditto.
39689 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39691         * config/i386/avx512fintrin.h (_ktest_mask16_u8,
39692         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): Move to ...
39693         * config/i386/avx512dqintrin.h (_ktest_mask16_u8,
39694         _ktestz_mask16_u8, _ktestc_mask16_u8, _kadd_mask16): ... here.
39695         * config/i386/i386-builtin.def (__builtin_ia32_ktestchi,
39696         __builtin_ia32_ktestzhi, __builtin_ia32_kaddhi): Use
39697         OPTION_MASK_ISA_AVX512DQ instead of OPTION_MASK_ISA_AVX512F.
39698         * config/i386/sse.md (SWI1248_AVX512BWDQ2): New mode iterator.
39699         (kadd<mode>, ktest<mode>): Use it instead of SWI1248_AVX512BWDQ.
39701 2017-01-26  Marek Polacek  <polacek@redhat.com>
39703         PR c/79199
39704         * fold-const.c (operand_equal_p) [COND_EXPR]: Use OP_SAME_WITH_NULL
39705         for the third operand.
39707 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39709         PR middle-end/79236
39710         * omp-low.c (struct omp_context): Add simt_stmt field.
39711         (scan_omp_for): Return omp_context *.
39712         (scan_omp_simd): Set simt_stmt on the non-_simt_ SIMD
39713         context to the _simt_ SIMD stmt.
39714         (lower_omp_for): For combined SIMD with sibling _simt_
39715         SIMD, make sure to use the same decls in _looptemp_
39716         clauses as in the sibling.
39718 2017-01-26  David Sherwood  <david.sherwood@arm.com>
39720         PR middle-end/79212
39721         * gimplify.c (omp_notice_variable): Add GOVD_SEEN flag to variables in
39722         all contexts.
39724 2017-01-26  Jakub Jelinek  <jakub@redhat.com>
39726         PR target/70465
39727         * reg-stack.c (emit_swap_insn): Instead of fld a; fld b; fxchg %st(1);
39728         emit fld b; fld a; if possible.
39730         * brig-builtins.def: Update copyright years.
39731         * config/arm/arm_acle_builtins.def: Update copyright years.
39733 2017-01-25  Michael Meissner  <meissner@linux.vnet.ibm.com>
39735         PR target/79179
39736         * config/rs6000/vsx.md (vsx_extract_<mode>_store): Use wY
39737         constraint instead of o for the stxsd instruction.
39739 2017-01-25  Carl Love  <cel@us.ibm.com>
39741         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Fix order
39742         of entries for ALTIVEC_BUILTIN_VEC_PACKS and P8V_BUILTIN_VEC_VGBBD.
39744 2017-01-25  Jonathan Wakely  <jwakely@redhat.com>
39746         * doc/invoke.texi (C++ Dialect Options): Fix typo.
39748 2017-01-25  Richard Biener  <rguenther@suse.de>
39750         PR tree-optimization/69264
39751         * target.def (vector_alignment_reachable): Improve documentation.
39752         * doc/tm.texi: Regenerate.
39753         * targhooks.c (default_builtin_vector_alignment_reachable): Simplify
39754         and add a comment.
39755         * tree-vect-data-refs.c (vect_supportable_dr_alignment): Revert
39756         earlier changes with respect to TYPE_USER_ALIGN.
39757         (vector_alignment_reachable_p): Likewise.  Improve dumping.
39759 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39761         PR target/79145
39762         * config/arm/arm.md (xordi3): Force constant operand into a register
39763         for TARGET_IWMMXT.
39765 2016-01-25  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
39767         * doc/invoke.texi (-fstore-merging): Correct default optimization
39768         levels at which it is enabled.
39769         (-O): Move -fstore-merging from list to...
39770         (-O2): ... Here.
39772 2017-01-25  Richard Biener  <rguenther@suse.de>
39774         PR debug/78363
39775         * omp-expand.c: Include debug.h.
39776         (expand_omp_taskreg): Make sure to generate early debug before
39777         outlining anything from a function.
39778         (expand_omp_target): Likewise.
39779         (grid_expand_target_grid_body): Likewise.
39781 2017-01-25  Maxim Ostapenko  <m.ostapenko@samsung.com>
39783         PR lto/79061
39784         * asan.c (get_translation_unit_decl): New function.
39785         (asan_add_global): Extract modules file name from globals
39786         TRANSLATION_UNIT_DECL name.
39788 2017-01-24  Eric Botcazou  <ebotcazou@adacore.com>
39790         PR target/77439
39791         * config/arm/arm.c (arm_function_ok_for_sibcall): Add back restriction
39792         for long calls with APCS frame and VFP.
39794 2017-01-24  David Malcolm  <dmalcolm@redhat.com>
39796         * cfg.c (original_copy_tables_initialized_p): New function.
39797         * cfg.h (original_copy_tables_initialized_p): New decl.
39798         * cfgrtl.c (relink_block_chain): Guard the call to
39799         free_original_copy_tables with a call to
39800         original_copy_tables_initialized_p.
39801         * cgraph.h (symtab_node::native_rtl_p): New decl.
39802         * cgraphunit.c (symtab_node::native_rtl_p): New function.
39803         (symtab_node::needed_p): Don't assert for early assembly output
39804         for __RTL functions.
39805         (cgraph_node::finalize_function): Set "force_output" for __RTL
39806         functions.
39807         (cgraph_node::analyze): Bail out early for __RTL functions.
39808         (analyze_functions): Update assertion to support __RTL functions.
39809         (cgraph_node::expand): Bail out early for __RTL functions.
39810         * final.c (rest_of_clean_state): Don't call delete_tree_ssa for
39811         __RTL functions.
39812         * function.h (struct function): Update comment for field
39813         "pass_startwith".
39814         * gimple-expr.c: Include "tree-pass.h".
39815         (gimple_has_body_p): Return false for __RTL functions.
39816         * Makefile.in (OBJS): Add run-rtl-passes.o.
39817         * pass_manager.h (gcc::pass_manager::get_rest_of_compilation): New
39818         accessor.
39819         (gcc::pass_manager::get_clean_slate): New accessor.
39820         * passes.c: Include "insn-addr.h".
39821         (should_skip_pass_p): Add logging.  Update logic for running
39822         "expand" to be compatible with both __GIMPLE and __RTL.  Guard
39823         property-provider override so it is only done for gimple passes.
39824         Don't skip dfinit.
39825         (skip_pass): New function.
39826         (execute_one_pass): Call skip_pass when skipping passes.
39827         * read-md.c (md_reader::read_char): Support filtering
39828         the input to a subset of line numbers.
39829         (md_reader::md_reader): Initialize fields
39830         m_first_line and m_last_line.
39831         (md_reader::read_file_fragment): New function.
39832         * read-md.h (md_reader::read_file_fragment): New decl.
39833         (md_reader::m_first_line): New field.
39834         (md_reader::m_last_line): New field.
39835         * read-rtl-function.c (function_reader::create_function): Only
39836         create cfun if it doesn't already exist.  Set PROP_rtl on cfun's
39837         curr_properties.  Set DECL_INITIAL to a dummy block.
39838         (read_rtl_function_body_from_file_range): New function.
39839         * read-rtl-function.h (read_rtl_function_body_from_file_range):
39840         New decl.
39841         * run-rtl-passes.c: New file.
39842         * run-rtl-passes.h: New file.
39844 2017-01-24  Jeff Law  <law@redhat.com>
39846         * config/microblaze/microblaze.h (ASM_FORMAT_PRIVATE_NAME): Increase
39847         buffer size.
39849 2017-01-24  Bin Cheng  <bin.cheng@arm.com>
39851         PR tree-optimization/79159
39852         * tree-ssa-loop-niter.c (get_cst_init_from_scev): New function.
39853         (record_nonwrapping_iv): Improve boundary using above function if no
39854         value range information.
39856 2017-01-24  Pekka Jääskeläinen  <pekka@parmance.com>
39857             Martin Jambor  <mjambor@suse.cz>
39859         * brig-builtins.def: New file.
39860         * builtins.def (DEF_HSAIL_BUILTIN): New macro.
39861         (DEF_HSAIL_ATOMIC_BUILTIN): Likewise.
39862         (DEF_HSAIL_SAT_BUILTIN): Likewise.
39863         (DEF_HSAIL_INTR_BUILTIN): Likewise.
39864         (DEF_HSAIL_CVT_ZEROI_SAT_BUILTIN): Likewise.
39865         * builtin-types.def (BT_INT8): New.
39866         (BT_INT16): Likewise.
39867         (BT_UINT8): Likewise.
39868         (BT_UINT16): Likewise.
39869         (BT_FN_ULONG): Likewise.
39870         (BT_FN_UINT_INT): Likewise.
39871         (BT_FN_UINT_ULONG): Likewise.
39872         (BT_FN_UINT_LONG): Likewise.
39873         (BT_FN_UINT_PTR): Likewise.
39874         (BT_FN_ULONG_PTR): Likewise.
39875         (BT_FN_INT8_FLOAT): Likewise.
39876         (BT_FN_INT16_FLOAT): Likewise.
39877         (BT_FN_UINT32_FLOAT): Likewise.
39878         (BT_FN_UINT16_FLOAT): Likewise.
39879         (BT_FN_UINT8_FLOAT): Likewise.
39880         (BT_FN_UINT64_FLOAT): Likewise.
39881         (BT_FN_UINT16_UINT32): Likewise.
39882         (BT_FN_UINT32_UINT16): Likewise.
39883         (BT_FN_UINT16_UINT16_UINT16): Likewise.
39884         (BT_FN_INT_PTR_INT): Likewise.
39885         (BT_FN_UINT_PTR_UINT): Likewise.
39886         (BT_FN_LONG_PTR_LONG): Likewise.
39887         (BT_FN_ULONG_PTR_ULONG): Likewise.
39888         (BT_FN_VOID_UINT64_UINT64): Likewise.
39889         (BT_FN_UINT8_UINT8_UINT8): Likewise.
39890         (BT_FN_INT8_INT8_INT8): Likewise.
39891         (BT_FN_INT16_INT16_INT16): Likewise.
39892         (BT_FN_INT_INT_INT): Likewise.
39893         (BT_FN_UINT_FLOAT_UINT): Likewise.
39894         (BT_FN_FLOAT_UINT_UINT): Likewise.
39895         (BT_FN_ULONG_UINT_UINT): Likewise.
39896         (BT_FN_ULONG_UINT_PTR): Likewise.
39897         (BT_FN_ULONG_ULONG_ULONG): Likewise.
39898         (BT_FN_UINT_UINT_UINT): Likewise.
39899         (BT_FN_VOID_UINT_PTR): Likewise.
39900         (BT_FN_UINT_UINT_PTR: Likewise.
39901         (BT_FN_UINT32_UINT64_PTR): Likewise.
39902         (BT_FN_INT_INT_UINT_UINT): Likewise.
39903         (BT_FN_UINT_UINT_UINT_UINT): Likewise.
39904         (BT_FN_UINT_UINT_UINT_PTR): Likewise.
39905         (BT_FN_UINT_ULONG_ULONG_UINT): Likewise.
39906         (BT_FN_ULONG_ULONG_ULONG_ULONG): Likewise.
39907         (BT_FN_LONG_LONG_UINT_UINT): Likewise.
39908         (BT_FN_ULONG_ULONG_UINT_UINT): Likewise.
39909         (BT_FN_VOID_UINT32_UINT64_PTR): Likewise.
39910         (BT_FN_VOID_UINT32_UINT32_PTR): Likewise.
39911         (BT_FN_UINT_UINT_UINT_UINT_UINT): Likewise.
39912         (BT_FN_UINT_FLOAT_FLOAT_FLOAT_FLOAT): Likewise.
39913         (BT_FN_ULONG_ULONG_ULONG_UINT_UINT): Likewise.
39914         * doc/frontends.texi: List BRIG FE.
39915         * doc/install.texi (Testing): Add BRIG tesring requirements.
39916         * doc/invoke.texi (Overall Options): Mention BRIG.
39917         * doc/standards.texi (Standards): Doucment BRIG HSA version.
39919 2017-01-24  Richard Biener  <rguenther@suse.de>
39921         PR translation/79208
39922         * ipa-devirt.c (odr_types_equivalent_p): Fix typo in diagnostic.
39924 2017-01-24  Martin Jambor  <mjambor@suse.cz>
39926         PR bootstrap/79198
39927         * ipa-prop.c (ipa_free_all_node_params): Call summary destructor.
39928         * ipa-prop.c (ipa_node_params_t::insert): Initialize fields known_csts
39929         and known_contexts.
39931 2017-01-24  Aldy Hernandez  <aldyh@redhat.com>
39933         PR middle-end/79123
39934         * gimple-ssa-warn-alloca.c (alloca_call_type): Make sure
39935         casts from signed to unsigned really don't have a range.
39937 2017-01-24  Markus Trippelsdorf  <markus@trippelsdorf.de>
39939         * gimple-ssa-sprintf.c (format_floating): Change MPFR_RNDx to
39940         GMP_RNDx for compatiblity.
39942 2017-01-24  Martin Liska  <mliska@suse.cz>
39944         PR bootstrap/79132
39945         * tree-ssa-reassoc.c (rewrite_expr_tree_parallel): Insert assert
39946         that would prevent us to call alloca with -1 as argument.
39948 2017-01-24  Jakub Jelinek  <jakub@redhat.com>
39950         * dwarf2out.c (output_compilation_unit_header, output_file_names):
39951         Avoid -Wformat-security warning.
39953 2017-01-23  Andrew Pinski  <apinski@cavium.com>
39955         * config/aarch64/aarch64.c (thunderx2t99_addrcost_table): Improve
39956         cost table.
39958 2017-01-23  Martin Sebor  <msebor@redhat.com>
39960         PR middle-end/78703
39961         * gimple-ssa-sprintf.c (warn_level): New global.
39962         (format_integer): Use it here and throughout the rest of the file.
39963         Use the same switch to compute sign as base.
39964         (maybe_warn): New function.
39965         (format_directive): Factor out warnings into maybe_warn.
39966         Add debugging output.  Use warn_level.
39967         (add_bytes): Use warn_level.
39968         (pass_sprintf_length::compute_format_length): Add debugging output.
39969         (try_substitute_return_value): Same.
39970         (pass_sprintf_length::handle_gimple_call): Set and use warn_level.
39972         PR middle-end/78703
39973         * gimple-ssa-sprintf.c (struct format_result): Remove constant member.
39974         (struct fmtresult, format_integer, format_floating): Adjust.
39975         (fmtresult::fmtresult): Set max correctly in two argument ctor.
39976         (get_string_length, format_string,format_directive): Same.
39977         (pass_sprintf_length::compute_format_length): Same.
39978         (try_substitute_return_value): Simplify slightly.
39980         PR middle-end/78703
39981         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust formatting.
39982         (fmtresult::operator+=): Outlined.
39983         (struct fmtresult): Add ctors.
39984         (struct conversion_spec): Rename...
39985         (struct directive): ...to this.  Add and remove data members.
39986         (directive::set_width, directive::set_precision): New functions.
39987         (format_percent): Use fmtresult ctor.
39988         (get_width_and_precision): Remove.
39989         (format_integer): Make naming changes.  Avoid computing width and
39990         precision.
39991         (format_floating): Same.  Adjust indentation.
39992         (format_character, format_none): New functions.
39993         (format_string): Moved character handling to format_character.
39994         (format_directive): Remove arguments, change return type.
39995         (parse_directive): New function.
39996         (pass_sprintf_length::compute_format_length): Move directive
39997         parsing to parse_directive.
39999 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40001         * tree.h (assign_assembler_name_if_neeeded): Rename to ...
40002         (assign_assembler_name_if_needed): ... this.
40003         * tree.c (assign_assembler_name_if_neeeded): Rename to ...
40004         (assign_assembler_name_if_needed): ... this.
40005         (free_lang_data_in_cgraph): Adjust callers.
40006         * cgraphunit.c (cgraph_node::analyze): Likewise.
40007         * omp-expand.c (expand_omp_taskreg, expand_omp_target):
40008         Likewise.
40010 2017-01-23  Richard Biener  <rguenther@suse.de>
40012         PR tree-optimization/79088
40013         PR tree-optimization/79188
40014         * tree-ssa-threadupdate.c (mark_threaded_blocks): Move code
40015         resetting loop bounds after last path deletion.  Reset loop
40016         bounds of the target loop, make code match the comments.
40017         * tree-ssa-threadbackwards.c (pass_early_thread_jumps::execute):
40018         Make sure loops need no fixups.
40020 2017-01-23  Kelvin Nilsen  <kelvin@gcc.gnu.org>
40022         * config/rs6000/rs6000-builtin.def (VSIEDPF): Add scalar insert
40023         exponent support with double type for first argument.
40024         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Changed
40025         type returned by __builtin_vec_extract_sig,
40026         __builtin_vec_extract_sig_sp, and __builtin_vec_extract_sig_dp
40027         functions from "vector int" to "vector unsigned int" or from
40028         "vector long long int" to "vector unsigned long long int".
40029         Changed type returned by __builtin_vec_extract_exp,
40030         __builtin_vec_extract_exp_sp, and __builtin_vec_extract_exp_dp
40031         functions from "vector int" to "vector unsigned int" or from
40032         "vector long long int" to "vector unsigned long long int".
40033         Changed return type of __builtin_vec_test_data_class,
40034         __builtin_vec_test_data_class_sp, and
40035         __builtin_vec_test_data_class_dp from "vector int" to
40036         "vector bool int" or from "vector long long int" to "vector bool
40037         long long int" and changed second argument type from "unsigned
40038         int" to "int".  Added new overloaded function forms "vector float
40039         __builtin_vec_insert_exp (vector float, vector unsigned int)" and
40040         "vector float __builtin_vec_insert_exp_sp (vector float, vector
40041         unsigned int)" and "vector double __builtin_vec_insert_exp (vector
40042         double, vector unsigned long long int)" and "vector double
40043         __builtin_vec_insert_exp_dp (vector double, vector unsigned long
40044         long int)".  Changed return type of
40045         __builtin_scalar_test_data_class and
40046         __builtin_scalar_test_data_class_sp and
40047         __builtin_scalar_test_data_class_dp from "unsigned int" to "bool
40048         int" and changed second argument from "unsigned int" to "int".
40049         Changed type returned by __builtin_scalar_test_neg,
40050         __builtin_scalar_test_neg_sp, and __builtin_scalar_test_neg_dp
40051         from "int" to "bool int".  Added new overloaded function form
40052         "double __builtin_scalar_insert_exp (double, unsigned long long int)".
40053         * config/rs6000/vsx.md (xsiexpdpf): New insn for scalar insert
40054         exponent double-precision with floating point first argument.
40055         * doc/extend.texi (PowerPC AltiVec Built-in Functions): Adjust
40056         documentation of scalar_test_data_class, scalar_test_neg,
40057         scalar_extract_sig, scalar_extract_exp, scalar_insert_exp,
40058         vector_extract_exp, vec_extract_sig, vec_insert_exp, and
40059         vec_test_data_class built-in functions to reflect refinements in
40060         their type signatures.
40062 2017-01-23  Andreas Tobler  <andreast@gcc.gnu.org>
40064         * config/aarch64/aarch64.c (aarch64_elf_asm_constructor): Increase
40065         size of buf.
40066         (aarch64_elf_asm_destructor): Likewise.
40068 2017-01-23  Bernd Schmidt  <bschmidt@redhat.com>
40070         PR rtl-optimization/78634
40071         * config/i386/i386.c (ix86_max_noce_ifcvt_seq_cost): New function.
40072         (TARGET_MAX_NOCE_IFCVT_SEQ_COST): Define.
40073         * ifcvt.c (noce_try_cmove): Add missing cost check.
40075         PR rtl-optimization/71724
40076         * combine.c (if_then_else_cond): Look for situations where it is
40077         beneficial to undo the work of one of the recursive calls.
40079 2017-01-23  Bin Cheng  <bin.cheng@arm.com>
40081         PR tree-optimization/70754
40082         * tree-predcom.c (stmt_combining_refs): New parameter INSERT_BEFORE.
40083         (reassociate_to_the_same_stmt): New parameter INSERT_BEFORE.  Insert
40084         combined stmt before it if not NULL.
40085         (combine_chains): Process refs reversely and compute dominance point
40086         for root ref.
40088 2017-01-23  Martin Liska  <mliska@suse.cz>
40090         PR tree-optimization/79196
40091         * tree-ssa-strlen.c (fold_strstr_to_memcmp): Rename to ...
40092         (fold_strstr_to_strncmp): ... this.  Fold the pattern to strncmp
40093         instead of memcmp.
40094         (strlen_optimize_stmt): Call the renamed function.
40096 2017-01-23  Michael Matz  <matz@suse.de>
40098         PR tree-optimization/78384
40099         * tree-ssa-loop-split.c (patch_loop_exit): Use correct edge.
40101 2017-01-23  Richard Biener  <rguenther@suse.de>
40103         PR tree-optimization/79186
40104         * tree-vrp.c (register_new_assert_for): Make sure we've seen
40105         both incoming edges before moving an assert.
40107 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40109         * ipa-prop.c (load_from_param_1): Removed.
40110         (load_from_unmodified_param): Bits from load_from_param_1 put back
40111         here.
40112         (load_from_param): Removed.
40113         (compute_complex_assign_jump_func): Removed stmt2 and just replaced it
40114         with stmt.  Reverted back to use of load_from_unmodified_param.
40116 2017-01-23  Martin Jambor  <mjambor@suse.cz>
40118         PR ipa/79108
40119         * ipa-prop.h (ipa_param_descriptor): Anotate with with GTY(()).
40120         (ipa_node_params): Annotate with GTY((for_user)).  Make descriptors
40121         field a pointer to garbage collected vector, mark lattices and
40122         ipcp_orig_node with GTY((skip)).
40123         (ipa_get_param_count): Adjust to descriptors being a pointer.
40124         (ipa_get_param): Likewise.
40125         (ipa_get_type): Likewise.
40126         (ipa_get_param_move_cost): Likewise.
40127         (ipa_set_param_used): Likewise.
40128         (ipa_get_controlled_uses): Likewise.
40129         (ipa_set_controlled_uses): Likewise.
40130         (ipa_is_param_used): Likewise.
40131         (ipa_node_params_t): Move into garbage collector.  New methods insert
40132         and remove.
40133         (ipa_node_params_sum): Annotate wth GTY(()).
40134         (ipa_check_create_node_params): Adjust to ipa_node_params_sum being
40135         garbage collected.
40136         (ipa_load_from_parm_agg): Adjust declaration.
40137         * ipa-icf.c (param_used_p): Adjust to descriptors being a pointer.
40138         * ipa-profile.c (ipa_profile): Likewise.
40139         * ipa-prop.c (ipa_get_param_decl_index_1): Likewise.
40140         (ipa_populate_param_decls): Make descriptors parameter garbage
40141         collected.
40142         (ipa_dump_param): Adjust to descriptors being a pointer.
40143         (ipa_alloc_node_params): Likewise.
40144         (ipa_initialize_node_params): Likewise.
40145         (load_from_param_1): Make descriptors parameter garbage collected.
40146         (load_from_unmodified_param): Likewise.
40147         (load_from_param): Likewise.
40148         (ipa_load_from_parm_agg): Likewise.
40149         (ipa_node_params::~ipa_node_params): Removed.
40150         (ipa_free_all_node_params): Remove call to delete operator.
40151         (ipa_node_params_t::insert): New.
40152         (ipa_node_params_t::remove): Likewise.
40153         (ipa_node_params_t::duplicate): Adjust to descriptors being a pointer,
40154         copy known_csts and known_contexts vectors.
40155         (ipa_read_node_info): Adjust to descriptors being a pointer.
40156         (ipcp_modif_dom_walker): Make m_descriptors field garbage
40157         collected.
40158         (ipcp_transform_function): Make descriptors variable garbage
40159         collected.
40161 2017-01-23  Andrew Senkevich  <andrew.senkevich@intel.com>
40163         * config/i386/avx512bwintrin.h: Add k-mask test, kortest intrinsics.
40164         * config/i386/avx512dqintrin.h: Ditto.
40165         * config/i386/avx512fintrin.h: Ditto.
40166         * config/i386/i386.c: Handle new builtins.
40167         * config/i386/i386-builtin.def: Add new builtins.
40168         * config/i386/sse.md (ktest<mode>, kortest<mode>): New.
40169         (UNSPEC_KORTEST, UNSPEC_KTEST): New.
40171 2017-01-23  Jakub Jelinek  <jakub@redhat.com>
40172             Martin Liska  <mliska@suse.cz>
40174         * asan.h: Define ASAN_USE_AFTER_SCOPE_ATTRIBUTE.
40175         * asan.c (asan_expand_poison_ifn): Support stores and use
40176         appropriate ASAN report function.
40177         * internal-fn.c (expand_ASAN_POISON_USE): New function.
40178         * internal-fn.def (ASAN_POISON_USE): Declare.
40179         * tree-into-ssa.c (maybe_add_asan_poison_write): New function.
40180         (maybe_register_def): Create ASAN_POISON_USE when sanitizing.
40181         * tree-ssa-dce.c (eliminate_unnecessary_stmts): Remove
40182         ASAN_POISON calls w/o LHS.
40183         * tree-ssa.c (execute_update_addresses_taken): Create clobber
40184         for ASAN_MARK (UNPOISON, &x, ...) in order to prevent usage of a LHS
40185         from ASAN_MARK (POISON, &x, ...) coming to a PHI node.
40186         * gimplify.c (asan_poison_variables): Add attribute
40187         use_after_scope_memory to variables that really needs to live
40188         in memory.
40189         * tree-ssa.c (is_asan_mark_p): Do not rewrite into SSA when
40190         having the attribute.
40192 2017-01-23  Martin Liska  <mliska@suse.cz>
40194         * asan.c (create_asan_shadow_var): New function.
40195         (asan_expand_poison_ifn): Likewise.
40196         * asan.h (asan_expand_poison_ifn): New declaration.
40197         * internal-fn.c (expand_ASAN_POISON): Likewise.
40198         * internal-fn.def (ASAN_POISON): New builtin.
40199         * sanopt.c (pass_sanopt::execute): Expand
40200         asan_expand_poison_ifn.
40201         * tree-inline.c (copy_decl_for_dup_finish): Make function
40202         external.
40203         * tree-inline.h (copy_decl_for_dup_finish): Likewise.
40204         * tree-ssa.c (is_asan_mark_p): New function.
40205         (execute_update_addresses_taken): Rewrite local variables
40206         (identified just by use-after-scope as addressable) into SSA.
40208 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40210         * doc/install.texi (Specific): opensource.apple.com uses https
40211         now. Remove trailing slash.
40213 2017-01-22  Gerald Pfeifer  <gerald@pfeifer.com>
40215         * README.Portability: Remove note on an Irix compatibility issue.
40217 2017-01-22  Dimitry Andric <dim@FreeBSD.org>
40219         * gcov.c (INCLUDE_ALGORITHM): Define.
40220         (INCLUDE_VECTOR): Define.
40221         No longer include <vector> and <algorithm> directly.
40223 2017-01-21  Gerald Pfeifer  <gerald@pfeifer.com>
40225         * doc/extend.texi (Thread-Local): Change www.akkadia.org reference
40226         to https.
40227         * doc/invoke.texi (Code Gen Options): Ditto.
40229 2017-01-21  Jan Hubicka  <hubicka@ucw.cz>
40231         PR lto/78407
40232         * cfg.c (update_bb_profile_for_threading): Fix updating of probablity.
40234 2017-01-21  Bernd Schmidt  <bschmidt@redhat.com>
40236         rtl-optimization/79125
40237         * cprop.c (local_cprop_pass): Handle cases where we make an
40238         unconditional trap.
40240 2017-01-20  Segher Boessenkool  <segher@kernel.crashing.org>
40242         PR target/61729
40243         PR target/77850
40244         * config/rs6000/rs6000.c (rs6000_gimplify_va_arg): Adjust address to
40245         read from, for big endian.
40247 2017-01-20  Jiong Wang  <jiong.wang@arm.com>
40249         * config/aarch64/aarch64-builtins.c (aarch64_init_builtins): Register
40250         register pauth builtins for LP64 only.
40252 2017-01-20  Marek Polacek  <polacek@redhat.com>
40254         PR c/79152
40255         * gimplify.c (should_warn_for_implicit_fallthrough): Handle consecutive
40256         non-case labels.
40258 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40260         * omp-expand.c (expand_omp_simd): Clear PROP_gimple_lomp_dev regardless
40261         of safelen status.
40262         * omp-offload.c (pass_omp_device_lower::gate): Use PROP_gimple_lomp_dev.
40263         * passes.c (dump_properties): Handle PROP_gimple_lomp_dev.
40264         * tree-inline.c (expand_call_inline): Propagate PROP_gimple_lomp_dev.
40266 2017-01-20  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40268         PR target/71270
40269         * config/arm/arm.c (neon_valid_immediate): Reject vector constants
40270         in big-endian mode when they are not a single duplicated value.
40272 2017-01-20  Richard Biener  <rguenther@suse.de>
40274         * BASE-VER: Bump to 7.0.1.
40276 2017-01-20  Alexander Monakov  <amonakov@ispras.ru>
40278         * omp-low.c (omplow_simd_context): New struct.  Use it...
40279         (lower_rec_simd_input_clauses): ...here and...
40280         (lower_rec_input_clauses): ...here to hold common data.  Adjust all
40281         references to idx, lane, max_vf, is_simt.
40283 2017-01-20  Graham Markall  <graham.markall@embecosm.com>
40285         * config/arc/arc.h (LINK_SPEC): Use arclinux_nps emulation when
40286         mcpu=nps400.
40288 2017-01-20  Martin Jambor  <mjambor@suse.cz>
40290         * hsa.h: Renaed to hsa-common.h.  Adjusted a comment.
40291         * hsa.c: Renaed to hsa-common.c.  Change include of gt-hsa.h to
40292         gt-hsa-common.h.
40293         * Makefile.in (OBJS): Rename hsa.o to hsa-common.o.
40294         (GTFILES): Rename hsa.c to hsa-common.c.
40295         * hsa-brig.c: Change include of hsa.h to hsa-common.h.
40296         * hsa-dump.c: Likewise.
40297         * hsa-gen.c: Likewise.
40298         * hsa-regalloc.c: Likewise.
40299         * ipa-hsa.c: Likewise.
40300         * omp-expand.c: Likewise.
40301         * omp-low.c: Likewise.
40302         * toplev.c: Likewise.
40304 2017-01-20  Marek Polacek  <polacek@redhat.com>
40306         PR c/64279
40307         * doc/invoke.texi: Document -Wduplicated-branches.
40308         * fold-const.c (operand_equal_p): Handle MODIFY_EXPR, INIT_EXPR,
40309         COMPOUND_EXPR, PREDECREMENT_EXPR, PREINCREMENT_EXPR,
40310         POSTDECREMENT_EXPR, POSTINCREMENT_EXPR, CLEANUP_POINT_EXPR, EXPR_STMT,
40311         STATEMENT_LIST, and RETURN_EXPR.  For non-pure non-const functions
40312         return 0 only when not OEP_LEXICOGRAPHIC.
40313         (fold_build_cleanup_point_expr): Use the expression
40314         location when building CLEANUP_POINT_EXPR.
40315         * tree-core.h (enum operand_equal_flag): Add OEP_LEXICOGRAPHIC.
40316         * tree.c (add_expr): Handle error_mark_node.
40318 2017-01-20  Martin Liska  <mliska@suse.cz>
40320         PR lto/69188
40321         * tree-profile.c (init_ic_make_global_vars): Do not call
40322         finalize_decl.
40323         (gimple_init_gcov_profiler): Likewise.
40325 2017-01-20  Martin Liska  <mliska@suse.cz>
40327         PR ipa/71190
40328         * cgraph.h (maybe_create_reference): Remove argument and
40329         update comment.
40330         * cgraphclones.c (cgraph_node::create_virtual_clone): Remove one
40331         argument.
40332         * ipa-cp.c (create_specialized_node): Likewise.
40333         * symtab.c (symtab_node::maybe_create_reference): Handle
40334         VAR_DECLs and ADDR_EXPRs and select ipa_ref_use type.
40336 2017-01-20  Martin Liska  <mliska@suse.cz>
40338         * read-rtl-function.c (function_reader::create_function): Use
40339         build_decl instread of build_decl_stat.
40341 2017-01-20  Andrew Senkevich  <andrew.senkevich@intel.com>
40343         * config/i386/avx512bwintrin.h: Add k-mask registers shift intrinsics.
40344         * config/i386/avx512dqintrin.h: Ditto.
40345         * config/i386/avx512fintrin.h: Ditto.
40346         * config/i386/i386-builtin-types.def: Add new types.
40347         * config/i386/i386.c: Handle new types.
40348         * config/i386/i386-builtin.def (__builtin_ia32_kshiftliqi)
40349         (__builtin_ia32_kshiftlihi, __builtin_ia32_kshiftlisi)
40350         (__builtin_ia32_kshiftlidi, __builtin_ia32_kshiftriqi)
40351         (__builtin_ia32_kshiftrihi, __builtin_ia32_kshiftrisi)
40352         (__builtin_ia32_kshiftridi): New.
40353         * config/i386/sse.md (k<code><mode>): Rename *k<code><mode>.
40355 2017-01-19  Segher Boessenkool  <segher@kernel.crashing.org>
40357         PR target/78875
40358         PR target/79140
40359         * config/rs6000/rs6000.c (TARGET_STACK_PROTECT_GUARD): Unconditionally
40360         define to rs6000_init_stack_protect_guard.
40361         (rs6000_init_stack_protect_guard): New function.
40363 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40364             Yunqiang Su  <yunqiang.su@imgtec.com>
40366         * config.gcc (supported_defaults): Add madd4.
40367         (with_madd4): Add validation.
40368         (all_defaults): Add madd4.
40369         * config/mips/mips.opt (mmadd4): New option.
40370         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40371         mmadd4.
40372         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40373         __mips_no_madd4.
40374         (ISA_HAS_UNFUSED_MADD4): Gate with mips_madd4.
40375         (ISA_HAS_FUSED_MADD4): Likewise.
40376         * doc/invoke.texi (-mmadd4): Document the new option.
40377         * doc/install.texi (--with-madd4): Document the new option.
40379 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40381         * config/aarch64/aarch64-builtins.c (enum aarch64_builtins): New
40382         entries for AARCH64_PAUTH_BUILTIN_XPACLRI,
40383         AARCH64_PAUTH_BUILTIN_PACIA1716, AARCH64_PAUTH_BUILTIN_AUTIA1716.
40384         (aarch64_init_pauth_hint_builtins): New.
40385         (aarch64_init_builtins): Call aarch64_init_pauth_hint_builtins.
40386         (aarch64_expand_builtin): Expand new builtins.
40388 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40390         * reg-notes.def (CFA_TOGGLE_RA_MANGLE): New reg-note.
40391         * combine-stack-adj.c (no_unhandled_cfa): Handle
40392         REG_CFA_TOGGLE_RA_MANGLE.
40393         * dwarf2cfi.c (dwarf2out_frame_debug): Handle REG_CFA_TOGGLE_RA_MANGLE.
40394         * config/aarch64/aarch64.c (aarch64_expand_prologue): Generates DWARF
40395         info for return address signing.
40396         (aarch64_expand_epilogue): Likewise.
40398 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40400         * config/aarch64/aarch64-opts.h (aarch64_function_type): New enum.
40401         * config/aarch64/aarch64-protos.h
40402         (aarch64_return_address_signing_enabled): New declaration.
40403         * config/aarch64/aarch64.c (aarch64_return_address_signing_enabled):
40404         New function.
40405         (aarch64_expand_prologue): Sign return address before it's pushed onto
40406         stack.
40407         (aarch64_expand_epilogue): Authenticate return address fetched from
40408         stack.
40409         (aarch64_override_options): Sanity check for ILP32 and ISA level.
40410         (aarch64_attributes): New function attributes for "sign-return-address".
40411         * config/aarch64/aarch64.md (UNSPEC_AUTI1716, UNSPEC_AUTISP,
40412         UNSPEC_PACI1716, UNSPEC_PACISP, UNSPEC_XPACLRI): New unspecs.
40413         ("*do_return"): Generate combined instructions according to key index.
40414         ("<pauth_mnem_prefix>sp", "<pauth_mnem_prefix1716", "xpaclri"): New.
40415         * config/aarch64/iterators.md (PAUTH_LR_SP, PAUTH_17_16): New integer
40416         iterators.
40417         (pauth_mnem_prefix, pauth_hint_num_a): New integer attributes.
40418         * config/aarch64/aarch64.opt (msign-return-address=): New.
40419         * doc/extend.texi (AArch64 Function Attributes): Documents
40420         "sign-return-address=".
40421         * doc/invoke.texi (AArch64 Options): Documents "-msign-return-address=".
40423 2017-01-19  Matthew Fortune  <matthew.fortune@imgtec.com>
40425         * doc/invoke.texi: Add missing -mlxc1-sxc1 options to
40426         overall option summary.
40428 2017-01-19  Jiong Wang  <jiong.wang@arm.com>
40430         * config/aarch64/aarch64-arches.def: New entry for "armv8.3-a".
40431         * config/aarch64/aarch64.h (AARCH64_FL_V8_3, AARCH64_FL_FOR_ARCH8_3,
40432         AARCH64_ISA_V8_3, TARGET_ARMV8_3): New.
40433         * doc/invoke.texi (AArch64 Options): Document "armv8.3-a".
40435 2017-01-19  Michael Meissner  <meissner@linux.vnet.ibm.com>
40437         * config/rs6000/rs6000-cpus.def (ISA_3_0_MASKS_SERVER): Enable
40438         -mpower9-minmax by default for -mcpu=power9.
40439         (ISA_3_MASKS_IEEE): Require -mvsx-small-integer to enable IEEE
40440         128-bit floating point.
40442 2017-01-20  Alan Modra  <amodra@gmail.com>
40444         * config/rs6000/rs6000.md (cmpstrnsi, cmpstrsi): Fail if
40445         optimizing for size.
40447 2017-01-20  Alan Modra  <amodra@gmail.com>
40449         PR target/79144
40450         * config/rs6000/rs6000.c (expand_strn_compare): Get the asm name
40451         for strcmp and strncmp from corresponding builtin decl.
40453 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40455         * config.gcc (x86_64-*-rtems*): Use i386/rtemself.h
40456         instead of i386/rtems-64.h.
40457         * config/i386/rtems-64.h: Remove.
40459 2017-01-19  Uros Bizjak  <ubizjak@gmail.com>
40461         PR target/78478
40462         Revert:
40463         2013-11-05  Uros Bizjak  <ubizjak@gmail.com>
40465         * config/i386/rtemself.h (LONG_DOUBLE_TYPE_SIZE): New define.
40467 2017-01-19  Tamar Christina  <tamar.christina@arm.com>
40469         * config/aarch64/aarch64.c (aarch64_simd_gen_const_vector_dup):
40470         Change int to HOST_WIDE_INT.
40471         * config/aarch64/aarch64-protos.h
40472         (aarch64_simd_gen_const_vector_dup): Likewise.
40473         * config/aarch64/aarch64-simd.md: Add copysign<mode>3.
40475 2017-01-19  David Malcolm  <dmalcolm@redhat.com>
40477         * langhooks-def.h (lhd_type_for_size): New decl.
40478         (LANG_HOOKS_TYPE_FOR_SIZE): Define as lhd_type_for_size.
40479         * langhooks.c (lhd_type_for_size): New function, taken from
40480         lto_type_for_size.
40482 2017-01-19  Pat Haugen  <pthaugen@us.ibm.com>
40484         * config/rs6000/power9.md (power9-alu): Remove 'cmp' type and add
40485         define_bypass for CR latency.
40486         (power9-cracked-alu): Update bypass latency and remove power9-branch.
40487         (power9-alu2): Add define_bypass for CR latency.
40488         (power9-cmp): New.
40489         (power9-mul): Update insn latency.
40490         (power9-mul-compare): Update insn latency, bypass latency and remove
40491         power9-branch.
40493 2016-01-19  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
40495         * config/aarch64/aarch64-protos.h (aarch64_nopcrelative_literal_loads):
40496         Delete.
40497         * config/aarch64/aarch64.md
40498         (aarch64_reload_movcp<GPF_TF:mode><P:mode>): Delete reference to
40499         aarch64_nopcrelative_literal_loads.
40500         (aarch64_reload_movcp<VALL:mode><P:mode>): Likewise.
40502 2017-01-19  Chenghua Xu  <paul.hua.gm@gmail.com>
40504         * config/mips/mips.h (ISA_HAS_FUSED_MADD4): Enable for
40505         TARGET_LOONGSON_3A.
40506         (ISA_HAS_UNFUSED_MADD4): Exclude TARGET_LOONGSON_3A.
40508 2017-01-19  Doug Gilmore  <doug.gilmore@imgtec.com>
40510         PR target/78176
40511         * config.gcc (supported_defaults): Add lxc1-sxc1.
40512         (with_lxc1_sxc1): Add validation.
40513         (all_defaults): Add lxc1-sxc1.
40514         * config/mips/mips.opt (mlxc1-sxc1): New option.
40515         * config/mips/mips.h (OPTION_DEFAULT_SPECS): Add a default for
40516         mlxc1-sxc1.
40517         (TARGET_CPU_CPP_BUILTINS): Add builtin_define for
40518         __mips_no_lxc1_sxc1.
40519         (ISA_HAS_LXC1_SXC1): Gate with mips_lxc1_sxc1.
40520         * doc/invoke.texi (-mlxc1-sxc1): Document the new option.
40521         * doc/install.texi (--with-lxc1-sxc1): Document the new option.
40523 2017-01-19  Richard Biener  <rguenther@suse.de>
40525         PR tree-optimization/72488
40526         * tree-ssa-sccvn.c (run_scc_vn): When we abort the VN make
40527         sure to restore SSA info.
40528         * tree-ssa.c (verify_ssa): Verify SSA info is not shared.
40530 2017-01-19  Richard Earnshaw  <rearnsha@arm.com>
40532         PR rtl-optimization/79121
40533         * expr.c (expand_expr_real_2, case LSHIFT_EXPR): Look at the signedness
40534         of the inner type when shifting an extended value.
40536 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40538         PR lto/78407
40539         * symtab.c (symtab_node::equal_address_to): Fix comparing of
40540         interposable aliases.
40542 2017-01-18  Peter Bergner  <bergner@vnet.ibm.com>
40544         PR target/78516
40545         * config/rs6000/spe.md (mov_si<mode>_e500_subreg0): Fix constraints.
40546         Use the evmergelohi instruction.
40547         (mov_si<mode>_e500_subreg4_2_le): Likewise.
40548         (mov_sitf_e500_subreg8_2_be): Likewise.
40549         (mov_sitf_e500_subreg12_2_le): Likewise.
40550         (mov_si<mode>_e500_subreg0_2_le): Fix constraints.
40551         (mov_si<mode>_e500_subreg4_2_be): Likewise.
40552         (mov_sitf_e500_subreg8_2_le): Likewise.
40553         (mov_sitf_e500_subreg12_2_be): Likewise.
40555 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40557         * config/rs6000/altivec.md (altivec_vbpermq): Change "type"
40558         attribute from vecsimple to vecperm.
40559         (altivec_vbpermq2): Likewise.
40561 2017-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40563         PR target/79040
40564         * config/rs6000/altivec.h: Fix typo of vec_cntlz to vec_cnttz.
40566 2017-01-18  Aaron Sawdey  <acsawdey@linux.vnet.ibm.com>
40567         * config/rs6000/rs6000-protos.h (expand_strn_compare): Add arg.
40568         * config/rs6000/rs6000.c (expand_strn_compare): Add ability to expand
40569         strcmp. Fix bug where comparison didn't stop with zero byte. Fix
40570         case where N arg is SIZE_MAX.
40571         * config/rs6000/rs6000.md (cmpstrnsi): Args to expand_strn_compare.
40572         (cmpstrsi): Add pattern.
40574 2017-01-18  Michael Meissner  <meissner@linux.vnet.ibm.com>
40576         * config/rs6000/rs6000-c.c (altivec_overloaded_builtins): Add
40577         __builtin_vec_revb builtins.
40578         * config/rs6000/rs6000-builtins.def (P9V_BUILTIN_XXBRQ_V16QI): Add
40579         built-in functions to support generation of the ISA 3.0 XXBR<x>
40580         vector byte reverse instructions.
40581         (P9V_BUILTIN_XXBRQ_V1TI): Likewise.
40582         (P9V_BUILTIN_XXBRD_V2DI): Likewise.
40583         (P9V_BUILTIN_XXBRD_V2DF): Likewise.
40584         (P9V_BUILTIN_XXBGW_V4SI): Likewise.
40585         (P9V_BUILTIN_XXBGW_V4SF): Likewise.
40586         (P9V_BUILTIN_XXBGH_V8HI): Likewise.
40587         (P9V_BUILTIN_VEC_REVB): Likewise.
40588         * config/rs6000/vsx.md (p9_xxbrq_v1ti): New insns/expanders to
40589         generate the ISA 3.0 XXBR<x> vector byte reverse instructions.
40590         (p9_xxbrq_v16qi): Likewise.
40591         (p9_xxbrd_<mode>, VSX_D iterator): Likewise.
40592         (p9_xxbrw_<mode>, VSX_W iterator): Likewise.
40593         (p9_xxbrh_v8hi): Likewise.
40594         * config/rs6000/altivec.h (vec_revb): Define if ISA 3.0.
40595         * doc/extend.texi (RS/6000 Altivec Built-ins): Document the
40596         vec_revb built-in functions.
40598 2017-01-18  Uros Bizjak  <ubizjak@gmail.com>
40600         PR rtl-optimization/78952
40601         * config/i386/i386.md (any_extract): New code iterator.
40602         (*insvqi_2): Use any_extract for source operand.
40603         (*insvqi_3): Use any_shiftrt for source operand.
40605 2017-01-18  Wilco Dijkstra  <wdijkstr@arm.com>
40607         * config/aarch64/aarch64.c (aarch64_sched_adjust_priority)
40608         New function.
40609         (TARGET_SCHED_ADJUST_PRIORITY): Define target hook.
40611 2017-01-18  Matthias Klose  <doko@ubuntu.com>
40613         * doc/install.texi: Allow default for --with-target-bdw-gc-include.
40615 2016-01-18  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40617         * config/rs6000/altivec.h (vec_bperm): Change #define.
40618         * config/rs6000/altivec.md (UNSPEC_VBPERMD): New enum constant.
40619         (altivec_vbpermq2): New define_insn.
40620         (altivec_vbpermd): Likewise.
40621         * config/rs6000/rs6000-builtin.def (VBPERMQ2): New monomorphic
40622         function interface.
40623         (VBPERMD): Likewise.
40624         (VBPERM): New polymorphic function interface.
40625         * config/rs6000/r6000-c.c (altivec_overloaded_builtins_table):
40626         Add entries for P9V_BUILTIN_VEC_VBPERM.
40627         * doc/extend.texi: Add interfaces for vec_bperm.
40629 2017-01-18  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
40631         * config/s390/s390-c.c (s390_expand_overloaded_builtin): Downcase
40632         first letter of error messages.
40633         (s390_resolve_overloaded_builtin): Likewise.
40634         * config/s390/s390.c (s390_expand_builtin): Likewise.
40635         (s390_invalid_arg_for_unprototyped_fn): Likewise.
40636         (s390_valid_target_attribute_inner_p): Likewise.
40637         * config/s390/s390.md ("tabort"): Likewise.
40639 2017-01-18  Toma Tabacu  <toma.tabacu@imgtec.com>
40641         * config/mips/mips.h (ISA_HAS_DIV3): Remove unused macro.
40642         (ISA_AVOID_DIV_HILO): New macro.
40643         (ISA_HAS_DIV): Use new ISA_AVOID_DIV_HILO macro.
40644         (ISA_HAS_DDIV): Likewise.
40646 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40648         * doc/invoke.texi (fabi-version): Correct number of occurrences.
40650 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40652         * doc/invoke.texi (fabi-version): Spelling fix.
40654 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40656         PR c++/70182
40657         * doc/invoke.texi (fabi-version): Mention mangling fix for
40658         operator names.
40660 2017-01-18  Markus Trippelsdorf  <markus@trippelsdorf.de>
40662         PR c++/77489
40663         * doc/invoke.texi (fabi-version): Document discriminator mangling.
40665 2017-01-17  Segher Boessenkool  <segher@kernel.crashing.org>
40667         PR target/78875
40668         * config/rs6000/rs6000-opts.h (stack_protector_guard): New enum.
40669         * config/rs6000/rs6000.c (rs6000_option_override_internal): Handle
40670         the new options.
40671         * config/rs6000/rs6000.md (stack_protect_set): Handle the new more
40672         flexible settings.
40673         (stack_protect_test): Ditto.
40674         * config/rs6000/rs6000.opt (mstack-protector-guard=,
40675         mstack-protector-guard-reg=, mstack-protector-guard-offset=): New
40676         options.
40677         * doc/invoke.texi (Option Summary) [RS/6000 and PowerPC Options]:
40678         Add -mstack-protector-guard=, -mstack-protector-guard-reg=, and
40679         -mstack-protector-guard-offset=.
40680         (RS/6000 and PowerPC Options): Ditto.
40682 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
40684         * config/i386/i386.h (MASK_CLASS_P): New define.
40685         * config/i386/i386.c (inline_secondary_memory_needed): Ensure that
40686         there are no registers from different register sets also when
40687         mask registers are used.  Update function comment.
40688         * config/i386/i386.md (*movsi_internal): Split (*k/*krm) alternative
40689         to (*k/*r) and (*k/*km) alternatives.
40691 2017-01-17  Wilco Dijkstra  <wdijkstr@arm.com>
40693         * config/aarch64/aarch64.md (eh_return): Remove pattern and splitter.
40694         * config/aarch64/aarch64.h (AARCH64_EH_STACKADJ_REGNUM): Remove.
40695         (EH_RETURN_HANDLER_RTX): New define.
40696         * config/aarch64/aarch64.c (aarch64_frame_pointer_required):
40697         Force frame pointer in EH return functions.
40698         (aarch64_expand_epilogue): Add barrier for eh_return.
40699         (aarch64_final_eh_return_addr): Remove.
40700         (aarch64_eh_return_handler_rtx): New function.
40701         * config/aarch64/aarch64-protos.h (aarch64_final_eh_return_addr):
40702         Remove.
40703         (aarch64_eh_return_handler_rtx): New prototype.
40705 2017-01-17  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
40707         * config/rs6000/altivec.h (vec_rlmi): New #define.
40708         (vec_vrlnm): Likewise.
40709         (vec_rlnm): Likewise.
40710         * config/rs6000/altivec.md (UNSPEC_VRLMI): New UNSPEC enum value.
40711         (UNSPEC_VRLNM): Likewise.
40712         (VIlong): New mode iterator.
40713         (altivec_vrl<VI_char>mi): New define_insn.
40714         (altivec_vrl<VI_char>nm): Likewise.
40715         * config/rs6000/rs6000-builtin.def (VRLWNM): New monomorphic
40716         function entry.
40717         (VRLDNM): Likewise.
40718         (RLNM): New polymorphic function entry.
40719         (VRLWMI): New monomorphic function entry.
40720         (VRLDMI): Likewise.
40721         (RLMI): New polymorphic function entry.
40722         * config/rs6000/r6000-c.c (altivec_overloaded_builtin_table): Add
40723         new entries for P9V_BUILTIN_VEC_RLMI and P9V_BUILTIN_VEC_RLNM.
40724         * doc/extend.texi: Add description of vec_rlmi, vec_rlnm, and
40725         vec_vrlnm.
40727 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40729         PR debug/78839
40730         * dwarf2out.c (field_byte_offset): Restore the
40731         PCC_BITFIELD_TYPE_MATTERS behavior for INTEGER_CST DECL_FIELD_OFFSET
40732         and DECL_FIELD_BIT_OFFSET.  Use fold_build2 instead of build2 + fold.
40733         (analyze_variants_discr, gen_variant_part): Use fold_build2 instead
40734         of build2 + fold.
40736 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40738         PR ada/67205
40739         * config/aarch64/aarch64.c (TARGET_CUSTOM_FUNCTION_DESCRIPTORS): Define
40741 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40743         PR debug/71669
40744         * dwarf2out.c (add_data_member_location_attribute): For constant
40745         offset bitfield emit for -gdwarf-5 DW_AT_data_bit_offset attribute
40746         instead of DW_AT_data_member_location, DW_AT_bit_offset and
40747         DW_AT_byte_size attributes.
40749 2017-01-17  Eric Botcazou  <ebotcazou@adacore.com>
40751         * config/rs6000/rs6000.c (rs6000_emit_move): Also use a TOC reference
40752         after forcing to constant memory when the code model is medium.
40754 2017-01-17  Julia Koval  <julia.koval@intel.com>
40756         PR target/76731
40757         * config/i386/avx512fintrin.h
40758         (_mm512_i32gather_ps): Change __addr type to void const*.
40759         (_mm512_mask_i32gather_ps): Ditto.
40760         (_mm512_i32gather_pd): Ditto.
40761         (_mm512_mask_i32gather_pd): Ditto.
40762         (_mm512_i64gather_ps): Ditto.
40763         (_mm512_mask_i64gather_ps): Ditto.
40764         (_mm512_i64gather_pd): Ditto.
40765         (_mm512_mask_i64gather_pd): Ditto.
40766         (_mm512_i32gather_epi32): Ditto.
40767         (_mm512_mask_i32gather_epi32): Ditto.
40768         (_mm512_i32gather_epi64): Ditto.
40769         (_mm512_mask_i32gather_epi64): Ditto.
40770         (_mm512_i64gather_epi32): Ditto.
40771         (_mm512_mask_i64gather_epi32): Ditto.
40772         (_mm512_i64gather_epi64): Ditto.
40773         (_mm512_mask_i64gather_epi64): Ditto.
40774         (_mm512_i32scatter_ps): Change __addr type to void*.
40775         (_mm512_mask_i32scatter_ps): Ditto.
40776         (_mm512_i32scatter_pd): Ditto.
40777         (_mm512_mask_i32scatter_pd): Ditto.
40778         (_mm512_i64scatter_ps): Ditto.
40779         (_mm512_mask_i64scatter_ps): Ditto.
40780         (_mm512_i64scatter_pd): Ditto.
40781         (_mm512_mask_i64scatter_pd): Ditto.
40782         (_mm512_i32scatter_epi32): Ditto.
40783         (_mm512_mask_i32scatter_epi32): Ditto.
40784         (_mm512_i32scatter_epi64): Ditto.
40785         (_mm512_mask_i32scatter_epi64): Ditto.
40786         (_mm512_i64scatter_epi32): Ditto.
40787         (_mm512_mask_i64scatter_epi32): Ditto.
40788         (_mm512_i64scatter_epi64): Ditto.
40789         (_mm512_mask_i64scatter_epi64): Ditto.
40790         * config/i386/avx512pfintrin.h
40791         (_mm512_mask_prefetch_i32gather_pd): Change __addr type to void const*.
40792         (_mm512_mask_prefetch_i32gather_ps): Ditto.
40793         (_mm512_mask_prefetch_i64gather_pd): Ditto.
40794         (_mm512_mask_prefetch_i64gather_ps): Ditto.
40795         (_mm512_prefetch_i32scatter_pd): Change __addr type to void*.
40796         (_mm512_prefetch_i32scatter_ps): Ditto.
40797         (_mm512_mask_prefetch_i32scatter_pd): Ditto.
40798         (_mm512_mask_prefetch_i32scatter_ps): Ditto.
40799         (_mm512_prefetch_i64scatter_pd): Ditto.
40800         (_mm512_prefetch_i64scatter_ps): Ditto.
40801         (_mm512_mask_prefetch_i64scatter_pd): Ditto.
40802         (_mm512_mask_prefetch_i64scatter_ps): Ditto.
40803         * config/i386/avx512vlintrin.h
40804         (_mm256_mmask_i32gather_ps): Change __addr type to void const*.
40805         (_mm_mmask_i32gather_ps): Ditto.
40806         (_mm256_mmask_i32gather_pd): Ditto.
40807         (_mm_mmask_i32gather_pd): Ditto.
40808         (_mm256_mmask_i64gather_ps): Ditto.
40809         (_mm_mmask_i64gather_ps): Ditto.
40810         (_mm256_mmask_i64gather_pd): Ditto.
40811         (_mm_mmask_i64gather_pd): Ditto.
40812         (_mm256_mmask_i32gather_epi32): Ditto.
40813         (_mm_mmask_i32gather_epi32): Ditto.
40814         (_mm256_mmask_i32gather_epi64): Ditto.
40815         (_mm_mmask_i32gather_epi64): Ditto.
40816         (_mm256_mmask_i64gather_epi32): Ditto.
40817         (_mm_mmask_i64gather_epi32): Ditto.
40818         (_mm256_mmask_i64gather_epi64): Ditto.
40819         (_mm_mmask_i64gather_epi64): Ditto.
40820         (_mm256_i32scatter_ps): Change __addr type to void*.
40821         (_mm256_mask_i32scatter_ps): Ditto.
40822         (_mm_i32scatter_ps): Ditto.
40823         (_mm_mask_i32scatter_ps): Ditto.
40824         (_mm256_i32scatter_pd): Ditto.
40825         (_mm256_mask_i32scatter_pd): Ditto.
40826         (_mm_i32scatter_pd): Ditto.
40827         (_mm_mask_i32scatter_pd): Ditto.
40828         (_mm256_i64scatter_ps): Ditto.
40829         (_mm256_mask_i64scatter_ps): Ditto.
40830         (_mm_i64scatter_ps): Ditto.
40831         (_mm_mask_i64scatter_ps): Ditto.
40832         (_mm256_i64scatter_pd): Ditto.
40833         (_mm256_mask_i64scatter_pd): Ditto.
40834         (_mm_i64scatter_pd): Ditto.
40835         (_mm_mask_i64scatter_pd): Ditto.
40836         (_mm256_i32scatter_epi32): Ditto.
40837         (_mm256_mask_i32scatter_epi32): Ditto.
40838         (_mm_i32scatter_epi32): Ditto.
40839         (_mm_mask_i32scatter_epi32): Ditto.
40840         (_mm256_i32scatter_epi64): Ditto.
40841         (_mm256_mask_i32scatter_epi64): Ditto.
40842         (_mm_i32scatter_epi64): Ditto.
40843         (_mm_mask_i32scatter_epi64): Ditto.
40844         (_mm256_i64scatter_epi32): Ditto.
40845         (_mm256_mask_i64scatter_epi32): Ditto.
40846         (_mm_i64scatter_epi32): Ditto.
40847         (_mm_mask_i64scatter_epi32): Ditto.
40848         (_mm256_i64scatter_epi64): Ditto.
40849         (_mm256_mask_i64scatter_epi64): Ditto.
40850         (_mm_i64scatter_epi64): Ditto.
40851         (_mm_mask_i64scatter_epi64): Ditto.
40852         * config/i386/i386-builtin-types.def (V16SF_V16SF_PCFLOAT_V16SI_HI_INT)
40853         (V8DF_V8DF_PCDOUBLE_V8SI_QI_INT, V8SF_V8SF_PCFLOAT_V8DI_QI_INT)
40854         (V8DF_V8DF_PCDOUBLE_V8DI_QI_INT, V16SI_V16SI_PCINT_V16SI_HI_INT)
40855         (V8DI_V8DI_PCINT64_V8SI_QI_INT, V8SI_V8SI_PCINT_V8DI_QI_INT)
40856         (V8DI_V8DI_PCINT64_V8DI_QI_INT, V2DF_V2DF_PCDOUBLE_V4SI_QI_INT)
40857         (V4DF_V4DF_PCDOUBLE_V4SI_QI_INT, V2DF_V2DF_PCDOUBLE_V2DI_QI_INT)
40858         (V4DF_V4DF_PCDOUBLE_V4DI_QI_INT, V4SF_V4SF_PCFLOAT_V4SI_QI_INT)
40859         (V8SF_V8SF_PCFLOAT_V8SI_QI_INT, V4SF_V4SF_PCFLOAT_V2DI_QI_INT)
40860         (V4SF_V4SF_PCFLOAT_V4DI_QI_INT, V2DI_V2DI_PCINT64_V4SI_QI_INT)
40861         (V4DI_V4DI_PCINT64_V4SI_QI_INT, V2DI_V2DI_PCINT64_V2DI_QI_INT)
40862         (V4DI_V4DI_PCINT64_V4DI_QI_INT, V4SI_V4SI_PCINT_V4SI_QI_INT)
40863         (V8SI_V8SI_PCINT_V8SI_QI_INT, V4SI_V4SI_PCINT_V2DI_QI_INT)
40864         (V4SI_V4SI_PCINT_V4DI_QI_INT, VOID_PFLOAT_HI_V16SI_V16SF_INT)
40865         (VOID_PFLOAT_QI_V8SI_V8SF_INT, VOID_PFLOAT_QI_V4SI_V4SF_INT)
40866         (VOID_PDOUBLE_QI_V8SI_V8DF_INT, VOID_PDOUBLE_QI_V4SI_V4DF_INT)
40867         (VOID_PDOUBLE_QI_V4SI_V2DF_INT, VOID_PFLOAT_QI_V8DI_V8SF_INT)
40868         (VOID_PFLOAT_QI_V4DI_V4SF_INT, VOID_PFLOAT_QI_V2DI_V4SF_INT)
40869         (VOID_PDOUBLE_QI_V8DI_V8DF_INT, VOID_PDOUBLE_QI_V4DI_V4DF_INT)
40870         (VOID_PDOUBLE_QI_V2DI_V2DF_INT, VOID_PINT_HI_V16SI_V16SI_INT)
40871         (VOID_PINT_QI_V8SI_V8SI_INT, VOID_PINT_QI_V4SI_V4SI_INT)
40872         (VOID_PLONGLONG_QI_V8SI_V8DI_INT, VOID_PLONGLONG_QI_V4SI_V4DI_INT)
40873         (VOID_PLONGLONG_QI_V4SI_V2DI_INT, VOID_PINT_QI_V8DI_V8SI_INT)
40874         (VOID_PINT_QI_V4DI_V4SI_INT, VOID_PINT_QI_V2DI_V4SI_INT)
40875         (VOID_PLONGLONG_QI_V8DI_V8DI_INT, VOID_QI_V8SI_PCINT64_INT_INT)
40876         (VOID_PLONGLONG_QI_V4DI_V4DI_INT, VOID_PLONGLONG_QI_V2DI_V2DI_INT)
40877         (VOID_HI_V16SI_PCINT_INT_INT, VOID_QI_V8DI_PCINT64_INT_INT)
40878         (VOID_QI_V8DI_PCINT_INT_INT): Remove.
40879         (V16SF_V16SF_PCVOID_V16SI_HI_INT,  V8DF_V8DF_PCVOID_V8SI_QI_INT)
40880         (V8SF_V8SF_PCVOID_V8DI_QI_INT, V8DF_V8DF_PCVOID_V8DI_QI_INT)
40881         (V16SI_V16SI_PCVOID_V16SI_HI_INT, V8DI_V8DI_PCVOID_V8SI_QI_INT)
40882         (V8SI_V8SI_PCVOID_V8DI_QI_INT, V8DI_V8DI_PCVOID_V8DI_QI_INT)
40883         (VOID_PVOID_HI_V16SI_V16SF_INT, VOID_PVOID_QI_V8SI_V8DF_INT)
40884         (VOID_PVOID_QI_V8DI_V8SF_INT, VOID_PVOID_QI_V8DI_V8DF_INT)
40885         (VOID_PVOID_HI_V16SI_V16SI_INT, VOID_PVOID_QI_V8SI_V8DI_INT)
40886         (VOID_PVOID_QI_V8DI_V8SI_INT, VOID_PVOID_QI_V8DI_V8DI_INT)
40887         (V2DF_V2DF_PCVOID_V4SI_QI_INT, V4DF_V4DF_PCVOID_V4SI_QI_INT)
40888         (V2DF_V2DF_PCVOID_V2DI_QI_INT, V4DF_V4DF_PCVOID_V4DI_QI_INT
40889         (V4SF_V4SF_PCVOID_V4SI_QI_INT, V8SF_V8SF_PCVOID_V8SI_QI_INT)
40890         (V4SF_V4SF_PCVOID_V2DI_QI_INT, V4SF_V4SF_PCVOID_V4DI_QI_INT)
40891         (V2DI_V2DI_PCVOID_V4SI_QI_INT, V4DI_V4DI_PCVOID_V4SI_QI_INT)
40892         (V2DI_V2DI_PCVOID_V2DI_QI_INT, V4DI_V4DI_PCVOID_V4DI_QI_INT)
40893         (V4SI_V4SI_PCVOID_V4SI_QI_INT, V8SI_V8SI_PCVOID_V8SI_QI_INT)
40894         (V4SI_V4SI_PCVOID_V2DI_QI_INT, V4SI_V4SI_PCVOID_V4DI_QI_INT)
40895         (VOID_PVOID_QI_V8SI_V8SF_INT, VOID_PVOID_QI_V4SI_V4SF_INT)
40896         (VOID_PVOID_QI_V4SI_V4DF_INT, VOID_PVOID_QI_V4SI_V2DF_INT)
40897         (VOID_PVOID_QI_V4DI_V4SF_INT, VOID_PVOID_QI_V2DI_V4SF_INT)
40898         (VOID_PVOID_QI_V4DI_V4DF_INT, VOID_PVOID_QI_V2DI_V2DF_INT)
40899         (VOID_PVOID_QI_V8SI_V8SI_INT, VOID_PVOID_QI_V4SI_V4SI_INT)
40900         (VOID_PVOID_QI_V4SI_V4DI_INT, VOID_PVOID_QI_V4SI_V2DI_INT)
40901         (VOID_PVOID_QI_V4DI_V4SI_INT, VOID_PVOID_QI_V2DI_V4SI_INT)
40902         (VOID_PVOID_QI_V4DI_V4DI_INT, VOID_PVOID_QI_V2DI_V2DI_INT)
40903         (VOID_QI_V8SI_PCVOID_INT_INT, VOID_HI_V16SI_PCVOID_INT_INT)
40904         (VOID_QI_V8DI_PCVOID_INT_INT): Add.
40905         * config/i386/i386.c (ix86_init_mmx_sse_builtins): Adjust builtin
40906         definitions accordingly.
40908 2017-01-17  Kito Cheng  <kito.cheng@gmail.com>
40909             Kuan-Lin Chen  <kuanlinchentw@gmail.com>
40911         PR target/79079
40912         * internal-fn.c (expand_mul_overflow): Use convert_modes instead of
40913         gen_lowpart.
40915 2017-01-17  Vladimir Makarov  <vmakarov@redhat.com>
40917         PR target/79058
40918         * ira-conflicts.c (ira_build_conflicts): Update total conflict
40919         hard regs for inner regno.
40921 2017-01-17  Martin Liska  <mliska@suse.cz>
40923         PR ipa/71207
40924         * ipa-polymorphic-call.c (contains_type_p): Fix wrong
40925         assumption and add comment.
40927 2017-01-17  Nathan Sidwell  <nathan@acm.org>
40929         * ipa-visibility.c (localize_node): New function, broken out of ...
40930         (function_and_variable_visibility): ... here. Call it.
40932 2017-01-17  Jan Hubicka  <hubicka@ucw.cz>
40934         PR middle-end/77445
40935         * tree-ssa-threadupdate.c (remove_ctrl_stmt_and_useless_edges):
40936         correctly set frequency of oudgoing edge.
40937         (duplicate_thread_path): Fix profile updating.
40939 2017-01-17  Jakub Jelinek  <jakub@redhat.com>
40941         PR other/79046
40942         * configure.ac: Add GCC_BASE_VER.
40943         * Makefile.in (version): Use @get_gcc_base_ver@ instead of cat to get
40944         version from BASE-VER file.
40945         (CFLAGS-gcc.o): Add -DBASEVER=$(BASEVER_s).
40946         (gcc.o): Depend on $(BASEVER).
40947         * common.opt (dumpfullversion): New option.
40948         * gcc.c (driver_handle_option): Handle OPT_dumpfullversion.
40949         * doc/invoke.texi: Document -dumpfullversion.
40950         * doc/install.texi: Document --with-gcc-major-version-only.
40951         * configure: Regenerated.
40953 2017-01-17  Richard Biener  <rguenther@suse.de>
40955         PR tree-optimization/71433
40956         * tree-vrp.c (register_new_assert_for): Merge same asserts
40957         on all incoming edges.
40958         (process_assert_insertions_for): Handle insertions at the
40959         beginning of BBs.
40961 2017-01-17  Gerald Pfeifer  <gerald@pfeifer.com>
40963         * config/i386/cygwin.h (LIBGCJ_SONAME): Remove.
40964         * config/i386/mingw32.h (LIBGCJ_SONAME): Remove.
40966 2017-01-17  Kaz Kojima  <kkojima@gcc.gnu.org>
40968         PR target/78633
40969         * config/sh/sh.md (cmpeqsi_t+1): Call copy_rtx to avoid invalid
40970         RTL sharing.
40972 2017-01-17  Alan Modra  <amodra@gmail.com>
40974         PR target/79066
40975         * config/rs6000/rs6000.md (elf_high, elf_low): Disable when pic.
40976         * config/rs6000/rs6000.c (rs6000_emit_allocate_stack): Don't allow
40977         symbolic stack limit when pic.
40979 2017-01-16  Martin Sebor  <msebor@redhat.com>
40981         PR tree-optimization/78608
40982         * gimple-ssa-sprintf.c (tree_digits): Avoid negating TYPE_MIN.
40984 2017-01-16  Jeff Law  <law@redhat.com>
40986         Revert:
40987         2016-12-02  Tadek Kijkowski  <tkijkowski@gmail.com>
40988         * Makefile.in (PREPROCESSOR_DEFINES): Add a level of indirection
40989         for several include directories that may be relative to sysroot.
40990         * config/i386/x-mingw32 (gplus_includedir): Define.
40991         (gplus_tool_includedir, gplus_backward_include_dir): Likewise.
40992         (native_system_includedir): Likewise.
40993         * config/i386/mingw32.h (STANDARD_STARTFILE_PREFIX_1): Do not
40994         override if TARGET_SYSTEM_ROOT is defined.
40995         (NATIVE_SYSTEM_HEADER_DIR): Likewise.
40997         PR tree-optimization/79090
40998         PR tree-optimization/33562
40999         PR tree-optimization/61912
41000         PR tree-optimization/77485
41001         * tree-ssa-dse.c (compute_trims): Accept STMT argument.  Dump STMT
41002         and computed trims into the dump file.
41004 2017-01-17  Uros Bizjak  <ubizjak@gmail.com>
41006         * config/i386/i386.h (LIMIT_RELOAD_CLASS): Remove.
41008 2017-01-16  Jakub Jelinek  <jakub@redhat.com>
41010         PR c/79089
41011         * gimplify.c (gimplify_init_constructor): If want_value and
41012         object == lhs, unshare lhs to avoid invalid tree sharing.  Formatting
41013         fix.
41015         PR target/79080
41016         * loop-doloop.c (doloop_modify): Call unshare_all_rtl_in_chain on
41017         sequence.  Formatting fixes.
41018         (doloop_optimize): Formatting fixes.
41020         PR driver/49726
41021         * gcc.c (debug_level_greater_than_spec_func): New function.
41022         (static_spec_functions): Add debug-level-gt spec function.
41023         (ASM_DEBUG_SPEC, cpp_options): Use %:debug-level-gt(0) instead of
41024         !g0.
41025         * config/darwin.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41026         * config/darwin9.h (DSYMUTIL_SPEC, ASM_DEBUG_SPEC): Likewise.
41027         * common.opt (g, gcoff, gdwarf, gdwarf-, ggdb, gno-pubnames,
41028         gpubnames, ggnu-pubnames, gno-record-gcc-switches,
41029         grecord-gcc-switches, gno-strict-dwarf, gstrict-dwarf, gstabs,
41030         gstabs+, gtoggle, gvms, gxcoff, gxcoff+): Add Driver flag.
41032 2017-01-16  Uros Bizjak  <ubizjak@gmail.com>
41034         * config/i386/i386.h (HARD_REGNO_CALLER_SAVE_MODE): Apply HImode and
41035         QImode fixups to general and mask registers only.
41037 2017-01-16  Carl Love  <cel@us.ibm.com>
41039         * config/rs6000/rs6000-c (altivec_overloaded_builtins): Add support
41040         for built-in functions
41041         vector signed char vec_nabs (vector signed char)
41042         vector signed short vec_nabs (vector signed short)
41043         vector signed int vec_nabs (vector signed int)
41044         vector signed long long vec_nabs (vector signed long long)
41045         vector float vec_nabs (vector float)
41046         vector double vec_nabs (vector double)
41047         * config/rs6000/rs6000-builtin.def: Add definitions for NABS functions
41048         and NABS overload.
41049         * config/rs6000/altivec.md: New define_expand nabs<mode>2 types
41050         * config/rs6000/altivec.h: New define for vec_nabs built-in function.
41051         * doc/extend.texi: Update the documentation file for the new built-in
41052         functions.
41054 2017-01-16  Martin Sebor  <msebor@redhat.com>
41056         * gimple-ssa-sprintf.c (format_directive): Correct a typo in a warning
41057         message.
41059 2017-01-16  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41061         * config/rs6000/rs6000.c (rtx_is_swappable_p): Change
41062         UNSPEC_VSX__XXSPLTD to require special splat handling.
41064 2017-01-16  David Malcolm  <dmalcolm@redhat.com>
41066         PR bootstrap/78616
41067         * system.h: Poison strndup.
41069 2017-01-16  Alan Modra  <amodra@gmail.com>
41071         PR target/79098
41072         * config/rs6000/rs6000.c (rs6000_legitimate_combined_insn): Don't
41073         use a switch.
41075 2017-01-16  Georg-Johann Lay  <avr@gjlay.de>
41077         * config/avr/avr.h (BRANCH_COST) [reload_completed]: Increase by 4.
41079 2017-01-15  Uros Bizjak  <ubizjak@gmail.com>
41081         * config/i386/i386.c (ix86_legitimate_combined_insn): Do not
41082         call recog here.  Assert that INSN_CODE (insn) is non-negative.
41084 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41086         PR target/72749
41087         * cfgrtl.c (rtl_split_edge): Also patch jump insns that jump to the
41088         fallthrough.
41089         * haifa-sched.c (dump_insn_stream): Don't crash if there is a label
41090         in the currently scheduled RTL fragment.
41092 2017-01-15  Segher Boessenkool  <segher@kernel.crashing.org>
41094         PR rtl-optimization/78751
41095         * ifcvt.c (find_cond_trap): If we generated a non-existing insn,
41096         give up.
41098 2017-01-14  Jeff Law  <law@redhat.com>
41100         PR tree-optimization/79090
41101         * tree-ssa-dse.c (valid_ao_ref_for_dse): Reject zero length and
41102         variable length stores.
41103         (compute_trims): Delete dead assignment to *trim_tail.
41104         (dse_dom_walker::dse_optimize_stmt): Optimize mem* calls with
41105         zero length.
41107 2017-01-14  Bernd Schmidt  <bschmidt@redhat.com>
41109         PR rtl-optimization/78626
41110         PR rtl-optimization/78727
41111         * cprop.c (one_cprop_pass): Collect unconditional traps in the middle
41112         of a block, and split such blocks after everything else is finished.
41114 2017-01-14  Alan Modra  <amodra@gmail.com>
41116         PR target/72749
41117         * combine.c (recog_for_combine_1): Set INSN_CODE before calling
41118         target legitimate_combined_insn.
41119         * config/rs6000/rs6000.c (TARGET_LEGITIMATE_COMBINED_INSN): Define.
41120         (rs6000_legitimate_combined_insn): New function.
41121         * config/rs6000/rs6000.md (UNSPEC_DOLOOP): Delete, and remove
41122         all uses.
41123         (ctr<mode>_internal3): Rename from *ctr<mode>_internal5.
41124         (ctr<mode>_internal4): Rename from *ctr<mode>_internal6.
41125         (ctr<mode>_internal1, ctr<mode>_internal2): Remove '*' from name.
41127 2017-01-14  Gerald Pfeifer  <gerald@pfeifer.com>
41129         * doc/frontends.texi (G++ and GCC): Remove references to Java.
41131 2017-01-13  Jeff Law  <law@redhat.com>
41133         PR tree-optimization/33562
41134         PR tree-optimization/61912
41135         PR tree-optimization/77485
41136         * tree-ssa-dse.c (delete_dead_call): Accept gsi rather than
41137         a statement.
41138         (delete_dead_assignment): Likewise.
41139         (dse_dom_walker::dse_optimize_stmt): Pass in the gsi rather than
41140         statement to delete_dead_call and delete_dead_assignment.
41142 2017-01-13  David Malcolm  <dmalcolm@redhat.com>
41144         PR c/78304
41145         * substring-locations.c (format_warning_va): Strengthen case 1 so
41146         that both endpoints of the substring must be within the format
41147         range for just the substring to be printed.
41149 2017-01-13  Uros Bizjak  <ubizjak@gmail.com>
41151         * config/i386/i386.opt (msgx): Use ix86_isa_flags2 variable.
41152         * config/i386/i386.c (ix86_target_string): Add missing options
41153         to isa_opts and reorder options by implied ISAs.  Rename isa_opts2 to
41154         isa2_opts, ix86_flag_opts to flag2_opts, ix86_target_other to
41155         flags_other and ix86_target_other to flags2_other.  Display unknown
41156         isa2 options.
41157         (ix86_valid_target_attribute_inner_p): Add missing options and
41158         reorder options by implied ISAs, as in ix86_target_string.
41160 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41162         * hash-table.h (hash_table::too_empty_p): New function.
41163         (hash_table::expand): Use it.
41164         (hash_table::traverse): Likewise.
41165         (hash_table::empty_slot): Use sizeof (value_type) instead of
41166         sizeof (PTR) to convert bytes to elements.  Shrink the table
41167         if the current size is excessive for the current number of
41168         elements.
41170 2017-01-13  Richard Sandiford  <richard.sandiford@arm.com>
41172         * ira-costs.c (record_reg_classes): Break from the inner loop
41173         early once alt_fail is known to be true.  Update outer loop
41174         handling accordingly.
41176 2017-01-13  Jeff Law  <law@redhat.com>
41178         * tree-ssa-dse.c (decrement_count): New function.
41179         (increment_start_addr, maybe_trim_memstar_call): Likewise.
41180         (dse_dom_walker::optimize_stmt): Call maybe_trim_memstar_call directly
41181         when we know the partially dead statement is a mem* function.
41183         PR tree-optimization/61912
41184         PR tree-optimization/77485
41185         * tree-ssa-dse.c: Include expr.h.
41186         (maybe_trim_constructor_store): New function.
41187         (maybe_trim_partially_dead_store): Call maybe_trim_constructor_store.
41189         PR tree-optimization/33562
41190         PR tree-optimization/61912
41191         PR tree-optimization/77485
41192         * doc/invoke.texi: Document new dse-max-object-size param.
41193         * params.def (PARM_DSE_MAX_OBJECT_SIZE): New PARAM.
41194         * tree-ssa-dse.c: Include params.h.
41195         (dse_store_status): New enum.
41196         (initialize_ao_ref_for_dse): New, partially extracted from
41197         dse_optimize_stmt.
41198         (valid_ao_ref_for_dse, normalize_ref): New.
41199         (setup_live_bytes_from_ref, compute_trims): Likewise.
41200         (clear_bytes_written_by, maybe_trim_complex_store): Likewise.
41201         (maybe_trim_partially_dead_store): Likewise.
41202         (maybe_trim_complex_store): Likewise.
41203         (dse_classify_store): Renamed from dse_possibly_dead_store_p.
41204         Track what bytes live from the original store.  Return tri-state
41205         for dead, partially dead or live.
41206         (dse_dom_walker): Add constructor, destructor and new private members.
41207         (delete_dead_call, delete_dead_assignment): New extracted from
41208         dse_optimize_stmt.
41209         (dse_optimize_stmt): Make a member of dse_dom_walker.
41210         Use initialize_ao_ref_for_dse.
41212         PR tree-optimization/33562
41213         PR tree-optimization/61912
41214         PR tree-optimization/77485
41215         * sbitmap.h (bitmap_count_bits): Prototype.
41216         (bitmap_clear_range, bitmap_set_range): Likewise.
41217         * sbitmap.c (bitmap_clear_range): New function.
41218         (bitmap_set_range, sbitmap_popcount, bitmap_count_bits): Likewise.
41220 2017-01-13  Martin Liska  <mliska@suse.cz>
41222         PR ipa/79043
41223         * function.c (set_cfun): Add new argument force.
41224         * function.h (set_cfun): Likewise.
41225         * ipa-inline-transform.c (inline_call): Use the function when
41226         strict alising from is dropped for function we inline to.
41228 2017-01-13  Richard Biener  <rguenther@suse.de>
41230         * tree-pretty-print.c (dump_generic_node): Fix inverted condition
41231         for dumping GIMPLE INTEGER_CSTs.
41233 2017-01-13  Rainer Orth  <ro@CeBiTec.Uni-Bielefeld.DE>
41235         * config/sol2.h (TARGET_OS_CPP_BUILTINS): Define __STDC_VERSION__
41236         to 201112L since C++17.
41238 2017-01-13  Maxim Ostapenko  <m.ostapenko@samsung.com>
41240         PR sanitizer/78887
41241         * asan.c (asan_needs_odr_indicator_p): Don't emit ODR indicators
41242         if -fsanitize=kernel-address is present.
41244 2017-01-13  Richard Biener  <rguenther@suse.de>
41246         * tree-pretty-print.c (dump_generic_node): Dump INTEGER_CSTs
41247         as _Literal ( type ) number in case usual suffixes do not
41248         preserve all information.
41250 2017-01-13  Richard Biener  <rguenther@suse.de>
41252         PR tree-optimization/77283
41253         * gimple-ssa-split-paths.c: Include gimple-ssa.h, tree-phinodes.h
41254         and ssa-iterators.h.
41255         (is_feasible_trace): Implement a cost model based on joiner
41256         PHI node uses.
41258 2017-01-12  Michael Meissner  <meissner@linux.vnet.ibm.com>
41260         PR target/79004
41261         * config/rs6000/rs6000.md (FP_ISA3): Do not optimize converting
41262         char or short to __float128/_Float128 directly.
41264 2017-01-12  Martin Sebor  <msebor@redhat.com>
41266         to -Wformat-overflow.
41267         * gimple-ssa-sprintf.c (pass_sprintf_length::gate): Adjust.
41268         (min_bytes_remaining): Same.
41269         (get_string_length): Same.
41270         (format_string): Same.
41271         (format_directive): Same.
41272         (add_bytes): Same.
41273         (pass_sprintf_length::handle_gimple_call): Same.
41275 2017-01-12  Jakub Jelinek  <jakub@redhat.com>
41277         * gimple-ssa-sprintf.c (try_substitute_return_value): Remove
41278         info.nowrite calls with no lhs that can't throw.  Return bool
41279         whether gsi_remove has been called or not.
41280         (pass_sprintf_length::handle_gimple_call): Return bool whether
41281         try_substitute_return_value called gsi_remove.  Formatting fix.
41282         (pass_sprintf_length::execute): Don't use gsi_remove if
41283         handle_gimple_call returned true.
41285         PR bootstrap/79069
41286         * cfgrtl.c (rtl_tidy_fallthru_edge): For any_uncondjump_p that can't
41287         be removed due to side-effects, don't remove following barrier nor
41288         turn the successor edge into fallthru edge.
41290 2017-01-12  Bill Schmidt  <wschmidt@linux.vnet.ibm.com>
41292         PR target/79044
41293         * config/rs6000/rs6000.c (insn_is_swappable_p): Mark
41294         element-reversing loads and stores as not swappable.
41296 2017-01-12  Nathan Sidwell  <nathan@acm.org>
41297             Nicolai Stange  <nicstange@gmail.com>
41299         * combine.c (try_combine): Don't ignore result of overlap checking
41300         loop.  Combine overlap & asm check into single loop.
41302 2017-01-12  Richard Biener  <rguenther@suse.de>
41304         * tree-pretty-print.c (dump_generic_node): Provide -gimple
41305         variant for MEM_REF.  Sanitize INTEGER_CST for -gimple.
41307 2017-01-12  Richard Biener  <rguenther@suse.de>
41309         * tree.c (initialize_tree_contains_struct): Make TS_OPTIMIZATION
41310         and TS_TARGET_OPTION directly derive from TS_BASE.
41311         * tree-core.h (tree_optimization_option): Derive from tree_base.
41312         (tree_target_option): Likewise.
41314 2017-01-11  Uros Bizjak  <ubizjak@gmail.com>
41316         * config/i386/i386.c (memory_address_length): Increase len
41317         only when rip_relative_addr_p returns false.
41319 2017-01-11  Julia Koval  <julia.koval@intel.com>
41321         * common/config/i386/i386-common.c (OPTION_MASK_ISA_SGX_UNSET): New.
41322         (OPTION_MASK_ISA_SGX_SET): New.
41323         (ix86_handle_option): Handle OPT_msgx.
41324         * config.gcc: Added sgxintrin.h.
41325         * config/i386/driver-i386.c (host_detect_local_cpu): Detect sgx.
41326         * config/i386/i386-c.c (ix86_target_macros_internal): Define __SGX__.
41327         * config/i386/i386.c (ix86_target_string): Add -msgx.
41328         (PTA_SGX): New.
41329         (ix86_option_override_internal): Handle new options.
41330         (ix86_valid_target_attribute_inner_p): Add sgx.
41331         * config/i386/i386.h (TARGET_SGX, TARGET_SGX_P): New.
41332         * config/i386/i386.opt: Add msgx.
41333         * config/i386/sgxintrin.h: New file.
41334         * config/i386/x86intrin.h: Add sgxintrin.h.
41336 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41338         PR c++/71537
41339         * fold-const.c (maybe_nonzero_address): Return 1 for function
41340         local objects.
41341         (tree_single_nonzero_warnv_p): Don't handle function local objects
41342         here.
41344         PR c++/72813
41345         * gcc.c (default_compilers): Don't add -o %g.s for -S -save-temps
41346         of c-header.
41348 2017-01-11  David Malcolm  <dmalcolm@redhat.com>
41350         PR driver/78877
41351         * opts.c: Include "spellcheck.h"
41352         (struct string_fragment): New struct.
41353         (struct edit_distance_traits<const string_fragment &>): New
41354         struct.
41355         (get_closest_sanitizer_option): New function.
41356         (parse_sanitizer_options): Offer suggestions for unrecognized arguments.
41358 2017-01-11  Jakub Jelinek  <jakub@redhat.com>
41360         * dwarf2out.c (DWARF_COMPILE_UNIT_HEADER_SIZE): For DWARF5 decrease
41361         by 12.
41362         (DWARF_COMDAT_TYPE_UNIT_HEADER_SIZE): Always
41363         DWARF_COMPILE_UNIT_HEADER_SIZE plus 12.
41364         (DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE): Define.
41365         (calc_base_type_die_sizes): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE
41366         for initial die_offset if dwarf_split_debug_info.
41367         (output_comp_unit): Use DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE for
41368         initial next_die_offset if dwo_id is non-NULL.  Don't emit padding
41369         fields.
41370         (output_skeleton_debug_sections): Formatting fix.  Use
41371         DWARF_COMPILE_UNIT_SKELETON_HEADER_SIZE instead of
41372         DWARF_COMPILE_UNIT_HEADER_SIZE.  Don't emit padding.
41374 2017-01-11  Wilco Dijkstra  <wdijkstr@arm.com>
41376         * config/arm/cortex-a53.md: Add bypasses for
41377         cortex_a53_r2f_cvt.
41378         (cortex_a53_r2f): Only use for transfers.
41379         (cortex_a53_f2r): Likewise.
41380         (cortex_a53_r2f_cvt): Add reservation for conversions.
41381         (cortex_a53_f2r_cvt): Likewise.
41383 2017-01-11  Tamar Christina  <tamar.christina@arm.com>
41385         * config/arm/arm_neon.h: Add __artificial__ and gnu_inline
41386         to all inlined functions, change static to extern.
41388 2017-01-11  Christophe Lyon  <christophe.lyon@linaro.org>
41390         PR target/78253
41391         * config/arm/arm.c (legitimize_pic_address): Handle reference to
41392         weak symbol.
41393         (arm_assemble_integer): Likewise.
41395 2017-01-11  Richard Earnshaw  <rearnsha@arm.com>
41397         * config.gcc: Use new awk script to check CPU, FPU and architecture
41398         parameters for --with-... options.
41399         * config/arm/parsecpu.awk: New file
41400         * config/arm/arm-cpus.in: New file.
41401         * config/arm/arm-opts.h: Include arm-cpu.h instead of processing .def
41402         files.
41403         * config/arm/arm.c: Include arm-cpu-data.h instead of processing .def
41404         files.
41405         * config/arm/t-arm: Update dependency rules.
41406         * common/config/arm/arm-common.c: Include arm-cpu-cdata.h instead
41407         of processing .def files.
41408         * config/arm/genopt.sh: Deleted.
41409         * config/arm/gentune.sh: Deleted.
41410         * config/arm/arm-cores.def: Deleted.
41411         * config/arm/arm-arches.def: Deleted.
41412         * config/arm/arm-fpus.def: Deleted.
41413         * config/arm/arm-tune.md: Regenerated.
41414         * config/arm/arm-tables.opt: Regenerated.
41415         * config/arm/arm-cpu.h: New generated file.
41416         * config/arm/arm-cpu-data.h: New generated file.
41417         * config/arm/arm-cpu-cdata.h: New generated file.
41419 2017-01-11  Maxim Ostapenko  <m.ostapenko@samsung.com>
41421         PR lto/79042
41422         * lto-cgraph.c (lto_output_varpool_node): Pack dynamically_initialized
41423         bit.
41424         (input_varpool_node): Unpack dynamically_initialized bit.
41426 2017-01-11  Eric Botcazou  <ebotcazou@adacore.com>
41428         PR rtl-optimization/79032
41429         * lra-constraints.c (simplify_operand_subreg): In the MEM case, test
41430         the alignment of the adjusted memory reference against that of MODE,
41431         instead of the alignment of the original memory reference.
41433 2017-01-11  Martin Jambor  <mjambor@suse.cz>
41435         * hsa.c (hsa_callable_function_p): Revert addition of DECL_ARTIFICIAL
41436         test.
41437         * ipa-hsa.c (process_hsa_functions): Only duplicate non-artificial
41438         decorated functions.
41440 2017-01-11  Richard Biener  <rguenther@suse.de>
41442         * tree-vrp.c (evrp_dom_walker::before_dom_children): Also
41443         set range/nonnull info for PHI results.  Do not set it on
41444         stmts marked for removal.
41446 2017-01-10  Eric Botcazou  <ebotcazou@adacore.com>
41448         * expr.c (store_field): In the bitfield case, fetch the return value
41449         from the registers before applying a single big-endian adjustment.
41450         Always do a final load for a BLKmode value not larger than a word.
41452 2017-01-10  David Malcolm  <dmalcolm@redhat.com>
41454         PR c++/77949
41455         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41456         that we correctly handle column numbers greater than
41457         LINE_MAP_MAX_COLUMN_NUMBER.
41459 2017-01-10  Martin Sebor  <msebor@redhat.com>
41461         PR middle-end/78245
41462         * gimple-ssa-sprintf.c (get_destination_size): Call
41463         {init,fini}object_sizes.
41464         * tree-object-size.c (addr_object_size): Adjust.
41465         (pass_through_call): Adjust.
41466         (pass_object_sizes::execute): Adjust.
41467         * tree-object-size.h (fini_object_sizes): Declare.
41469 2017-01-10  Martin Sebor  <msebor@redhat.com>
41471         PR tree-optimization/78775
41472         * builtins.c (get_size_range): Move...
41473         * calls.c: ...to here.
41474         (alloc_max_size): Accept zero argument.
41475         (operand_signed_p): Remove.
41476         (maybe_warn_alloc_args_overflow): Call get_size_range.
41477         * calls.h (get_size_range): Declare.
41479 2017-01-10  Joe Seymour  <joe.s@somniumtech.com>
41481         * config/msp430/driver-msp430.c (msp430_mcu_data): Sync with data
41482         from TI's devices.csv file as of September 2016.
41483         * config/msp430/msp430.c (msp430_mcu_data): Likewise.
41485 2017-01-10  Sandra Loosemore  <sandra@codesourcery.com>
41487         * doc/extend.texi: Tweak formatting to fix overfull hbox warnings.
41488         * doc/invoke.texi: Likewise.
41489         * doc/md.texi: Likewise.
41490         * doc/objc.texi: Likewise.
41492 2017-01-10  Joshua Conner  <joshconner@google.com>
41494         * config/arm/fuchsia-elf.h: New file.
41495         * config/fuchsia.h: New file.
41496         * config.gcc (*-*-fuchsia*): Set native_system_header_dir.
41497         (aarch64*-*-fuchsia*, arm*-*-fuchsia*, x86_64-*-fuchsia*): Add to
41498         targets.
41499         * config.host: (aarch64*-*-fuchsia*, arm*-*-fuchsia*): Add to hosts.
41501 2016-01-10  Richard Biener  <rguenther@suse.de>
41503         PR tree-optimization/79034
41504         * tree-call-cdce.c (shrink_wrap_one_built_in_call_with_conds):
41505         Propagate out degenerate PHIs in the joiner.
41507 2017-01-10  Martin Liska  <mliska@suse.cz>
41509         * ipa-icf.c (sort_sem_items_by_decl_uid): New function.
41510         (sort_congruence_classes_by_decl_uid): Likewise.
41511         (sort_congruence_class_groups_by_decl_uid): Likewise.
41512         (sem_item_optimizer::merge_classes): Sort class, groups in these
41513         classes and members in the groups by DECL_UID of declarations.
41514         This would make merge operations stable.
41516 2017-01-10  Martin Liska  <mliska@suse.cz>
41518         * ipa-icf.c (sem_item_optimizer::sem_item_optimizer): Remove
41519         usage of m_classes_vec.
41520         (sem_item_optimizer::~sem_item_optimizer):  Likewise.
41521         (sem_item_optimizer::get_group_by_hash): Likewise.
41522         (sem_item_optimizer::subdivide_classes_by_equality): Likewise.
41523         (sem_item_optimizer::subdivide_classes_by_sensitive_refs): Likewise.
41524         (sem_item_optimizer::verify_classes): Likewise.
41525         (sem_item_optimizer::process_cong_reduction): Likewise.
41526         (sem_item_optimizer::dump_cong_classes): Likewise.
41527         (sem_item_optimizer::merge_classes): Likewise.
41528         * ipa-icf.h (congruence_class_hash): Rename from
41529         congruence_class_group_hash.  Remove declaration of m_classes_vec.
41531 2017-01-10  Andrew Senkevich  <andrew.senkevich@intel.com>
41533         * common/config/i386/i386-common.c (OPTION_MASK_ISA_AVX512VPOPCNTDQ_SET,
41534         OPTION_MASK_ISA_AVX512VPOPCNTDQ_UNSET): New.
41535         * config.gcc: Add avx512vpopcntdqintrin.h.
41536         * config/i386/avx512vpopcntdqintrin.h: New.
41537         * config/i386/cpuid.h (bit_AVX512VPOPCNTDQ): New.
41538         * config/i386/i386-builtin-types.def: Add new types.
41539         * config/i386/i386-builtin.def (__builtin_ia32_vpopcountd_v16si,
41540         __builtin_ia32_vpopcountd_v16si_mask, __builtin_ia32_vpopcountq_v8di,
41541         __builtin_ia32_vpopcountq_v8di_mask): New.
41542         * config/i386/i386-c.c (ix86_target_macros_internal): Define
41543         __AVX512VPOPCNTDQ__.
41544         * config/i386/i386.c (ix86_target_string): Add -mavx512vpopcntdq.
41545         (PTA_AVX512VPOPCNTDQ): Define.
41546         * config/i386/i386.h (TARGET_AVX512VPOPCNTDQ,
41547         TARGET_AVX512VPOPCNTDQ_P): Define.
41548         * config/i386/i386.opt: Add mavx512vpopcntdq.
41549         * config/i386/immintrin.h: Include avx512vpopcntdqintrin.h.
41550         * config/i386/sse.md (define_insn "vpopcount<mode><mask_name>"): New.
41552 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41554         PR middle-end/77484
41555         * predict.def (PRED_CALL): Set to 67.
41557 2017-01-09  Eric Botcazou  <ebotcazou@adacore.com>
41559         * expr.c (store_field): In the bitfield case, if the value comes from
41560         a function call and is of an aggregate type returned in registers, do
41561         not modify the field mode; extract the value in all cases if the mode
41562         is BLKmode and the size is not larger than a word.
41564 2017-01-09  Dominique d'Humieres  <dominiq@lps.ens.fr>
41566         PR target/71017
41567         * config/i386/cpuid.h: Fix undefined behavior.
41569 2017-01-04  Jeff Law  <law@redhat.com>
41571         PR tree-optimization/79007
41572         PR tree-optimization/67955
41573         * tree-ssa-alias.c (same_addr_size_stores_p): Only need to be
41574         conservative for pt.null when flag_non_call_exceptions is on.
41576 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41578         PR translation/79019
41579         PR translation/79020
41580         * params.def (PARAM_INLINE_MIN_SPEEDUP,
41581         PARAM_IPA_CP_SINGLE_CALL_PENALTY,
41582         PARAM_USE_AFTER_SCOPE_DIRECT_EMISSION_THRESHOLD): Fix typos
41583         in descriptions.
41584         * config/avr/avr.opt (maccumulate-args): Likewise.
41585         * config/msp430/msp430.opt (mwarn-mcu): Likewise.
41586         * common.opt (freport-bug): Likewise.
41587         * cif-code.def (CIF_FINAL_ERROR): Likewise.
41588         * doc/invoke.texi (ipa-cp-single-call-penalty): Likewise.
41589         * config/s390/s390.c (s390_invalid_binary_op): Fix spelling in
41590         translatable string.
41591         * config/i386/i386.c (function_value_32): Likewise.
41592         * config/nios2/nios2.c (nios2_valid_target_attribute_rec): Likewise.
41593         * config/msp430/msp430.c (msp430_option_override, msp430_attr):
41594         Likewise.
41595         * config/msp430/driver-msp430.c (msp430_select_hwmult_lib): Likewise.
41596         * common/config/msp430/msp430-common.c (msp430_handle_option):
41597         Likewise.
41598         * symtab.c (symtab_node::verify_base): Likewise.
41599         * opts.c (set_debug_level): Likewise.
41600         * tree.c (verify_type_variant): Likewise.  Fix typo in comment.
41601         * config/rs6000/rs6000-c.c (altivec_resolve_overloaded_builtin): Add
41602         missing whitespace to translatable strings.
41603         * config/avr/avr.md (bswapsi2): Fix typo in comment.
41604         * config/sh/superh.h: Likewise.
41605         * config/i386/xopintrin.h: Likewise.
41606         * config/i386/znver1.md: Likewise.
41607         * config/rs6000/rs6000.c (struct rs6000_opt_mask): Likewise.
41608         * ipa-inline-analysis.c (compute_inline_parameters): Likewise.
41609         * double-int.h (struct double_int): Likewise.
41610         * double-int.c (div_and_round_double): Likewise.
41611         * wide-int.cc: Likewise.
41612         * tree-ssa.c (non_rewritable_mem_ref_base): Likewise.
41613         * tree-ssa-sccvn.c (vn_reference_lookup_3): Likewise.
41614         * cfgcleanup.c (crossjumps_occured): Renamed to ...
41615         (crossjumps_occurred): ... this.
41616         (try_crossjump_bb, try_head_merge_bb, try_optimize_cfg, cleanup_cfg):
41617         Adjust all uses.
41619         PR tree-optimization/78899
41620         * tree-if-conv.c (version_loop_for_if_conversion): Instead of
41621         returning bool return struct loop *, NULL for failure and the new
41622         loop on success.
41623         (versionable_outer_loop_p): Don't version outer loop if it has
41624         dont_vectorized bit set.
41625         (tree_if_conversion): When versioning outer loop, ensure
41626         tree_if_conversion is performed also on the inner loop of the
41627         non-vectorizable outer loop copy.
41628         * tree-vectorizer.c (set_uid_loop_bbs): Formatting fix.  Fold
41629         LOOP_VECTORIZED in inner loop of the scalar outer loop and
41630         prevent vectorization of it.
41631         (vectorize_loops): For outer + inner LOOP_VECTORIZED, ensure
41632         the outer loop vectorization of the non-scalar version is attempted
41633         before vectorization of the inner loop in scalar version.  If
41634         outer LOOP_VECTORIZED guarded loop is not vectorized, prevent
41635         vectorization of its inner loop.
41636         * tree-vect-loop-manip.c (rename_variables_in_bb): If outer_loop
41637         has 2 inner loops, rename also on edges from bb whose single pred
41638         is outer_loop->header.  Fix typo in function comment.
41640 2017-01-09  Martin Sebor  <msebor@redhat.com>
41642         PR bootstrap/79033
41643         * asan.c (asan_emit_stack_protection): Increase local buffer size
41644         to avoid snprintf truncation warning.
41646 2017-01-09  Andrew Pinski  <apinski@cavium.com>
41648         * config/aarch64/aarch64-cores.def: Add thunderx2t99.  Change vulcan
41649         to reference thunderx2t99 for the tuning structure
41650         * config/aarch64/aarch64-cost-tables.h (vulcan_extra_costs):
41651         Rename to ...
41652         (thunderx2t99_extra_costs): This.
41653         * config/aarch64/aarch64-tune.md: Regenerate.
41654         * config/aarch64/aarch64.c (vulcan_addrcost_table): Rename to ...
41655         (vulcan_addrcost_table): This.
41656         (vulcan_regmove_cost): Rename to ...
41657         (thunderx2t99_regmove_cost): This.
41658         (vulcan_vector_cost): Rename to ...
41659         (thunderx2t99_vector_cost): this.
41660         (vulcan_branch_cost): Rename to ...
41661         (thunderx2t99_branch_cost): This.
41662         (vulcan_tunings): Rename to ...
41663         (thunderx2t99_tunings): This and s/vulcan/thunderx2t99 .
41664         * doc/invoke.texi (AARCH64/mtune): Add thunderx2t99.
41666 2017-01-09  Martin Jambor  <mjambor@suse.cz>
41668         PR ipa/78365
41669         PR ipa/78599
41670         * ipa-prop.h (ipa_jump_func): Swap positions of vr_known and m_vr.
41671         * ipa-cp.c (ipa_vr_operation_and_type_effects): New function.
41672         (propagate_vr_accross_jump_function): Use the above function for all
41673         value range computations for pass-through jump functions and type
41674         converasion from explicit value range values.
41675         (ipcp_propagate_stage): Do not attempt to deduce types of formal
41676         parameters from TYPE_ARG_TYPES.
41677         * ipa-prop.c (ipa_write_jump_function): Remove trailing whitespace.
41678         (ipa_write_node_info): Stream type of the actual argument.
41679         (ipa_read_node_info): Likewise. Also remove trailing whitespace.
41681 2017-01-09  Martin Liska  <mliska@suse.cz>
41683         PR pch/78970
41684         * gcc.c (driver_handle_option): Handle OPT_E and set have_E.
41685         (lookup_compiler): Do not show error message with have_E.
41687 2017-01-09  Jakub Jelinek  <jakub@redhat.com>
41689         PR tree-optimization/78938
41690         * tree-vect-stmts.c (vectorizable_condition): For non-masked COND_EXPR
41691         where comp_vectype is VECTOR_BOOLEAN_TYPE_P, use
41692         BIT_{NOT,XOR,AND,IOR}_EXPR on the comparison operands instead of
41693         {EQ,NE,GE,GT,LE,LT}_EXPR directly inside of VEC_COND_EXPR.  Formatting
41694         fixes.
41696 2017-01-09  Kyrylo Tkachov  <kyrylo.tkachov@arm.com>
41698         * tree-ssa-address.c (gen_addr_rtx): Don't handle index if it
41699         is const0_rtx.
41701 2017-01-09  Richard Biener  <rguenther@suse.de>
41703         PR tree-optimization/78997
41704         * tree-vect-slp.c (vect_mask_constant_operand_p): Handle SSA
41705         name condition properly.
41707 2017-01-09  Richard Biener  <rguenther@suse.de>
41709         PR debug/79000
41710         * dwarf2out.c (is_cxx): New overload with context.
41711         (is_naming_typedef_decl): Use it.
41713 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41715         * invoke.texi (Option Summary): Correct spacing in option lists
41716         and add line breaks to fix over-long lines.
41718 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41720         PR middle-end/17660
41722         * extend.texi (Common Variable Attributes): Add xref to GCC
41723         Internals manual to explain mode attribute keywords.
41725 2017-01-08  Sandra Loosemore  <sandra@codesourcery.com>
41727         PR other/16519
41728         * doc/invoke.texi (Option Summary): Move -pthread to Linker Options
41729         and Preprocessor Options.
41730         (Options for Linking): Document -pthread here....
41731         (RS/6000 and PowerPC Options): ...not here.
41732         (Solaris 2 Options): ...or here.
41733         * doc/cppopts.texi: Document -pthread.
41735 2017-01-08  Martin Sebor  <msebor@redhat.com>
41737         PR middle-end/77708
41738         * doc/invoke.texi (Warning Options): Document -Wformat-truncation.
41739         * gimple-ssa-sprintf.c (call_info::reval_used, call_info::warnopt):
41740         New member functions.
41741         (format_directive): Used them.
41742         (add_bytes): Same.
41743         (pass_sprintf_length::handle_gimple_call): Same.
41744         * graphite-sese-to-poly.c (tree_int_to_gmp): Increase buffer size
41745         to avoid truncation for any argument.
41746         (extract_affine_mul): Same.
41747         * tree.c (get_file_function_name): Same.
41749 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41751         PR middle-end/77484
41752         * predict.def (PRED_INDIR_CALL): Set to 86.
41754 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41756         PR preprocessor/54124
41757         * doc/cppopts.texi: Reformat -d subtable to list the full name
41758         of the options.  Add cross-reference to the docs for the general
41759         compiler -d options.
41760         * doc/invoke.texi (Developer Options): Add cross-reference to the
41761         preprocessor-specific -d option documentation.
41763 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41765         PR preprocessor/13498
41766         * doc/cpp.texi (Search Path): Rewrite to remove obsolete and
41767         redudant material, and reflect new command-line options.
41768         (System Headers): Likewise.
41770 2017-01-07  Sandra Loosemore  <sandra@codesourcery.com>
41772         * doc/cppdiropts.texi: Merge documentation of -I, -iquote,
41773         -isystem, and -idirafter.  Copy-edit.
41774         * doc/cppopts.texi: Copy-edit.  Remove contradiction about
41775         default for -ftrack-macro-expansion.  Delete obsolete and
41776         badly-formatted implementation details about -fdebug-cpp output.
41777         * doc/cppwarnopts.texi: Copy-edit.
41779 2017-01-07  David Malcolm  <dmalcolm@redhat.com>
41781         PR c++/72803
41782         * input.c (selftest::test_accessing_ordinary_linemaps): Verify
41783         that the transition from a max line width >= 1<<10 to narrower
41784         lines works correctly.
41786 2017-01-07  Alexandre Oliva  <aoliva@redhat.com>
41788         * doc/options.texi (PerFunction): New.
41789         * opt-functions.awk (switch_flags): Map both Optimization and
41790         PerFunction to CL_OPTIMIZATION.
41791         * opth-gen.awk: Test for PerFunction flag along with
41792         Optimization.
41793         * optc-save-gen.awk: Likewise.  Introduce var_opt_hash and set
41794         it only when the latter is present.  Skip those that don't in
41795         the hash function generator.
41796         * common.opt (fvar-tracking): Mark as PerFunction instead of
41797         Optimization.
41798         (fvar-tracking-assignments): Likewise.
41799         (fvar-tracking-assignments-toggle): Likewise.
41800         (fvar-tracking-uninit): Likewise.
41802 2017-01-07  Jakub Jelinek  <jakub@redhat.com>
41804         PR translation/79018
41805         * params.def (PARAM_MAX_STORES_TO_MERGE): Add missing space between
41806         the and store.
41808 2017-01-06  Mikael Pettersson  <mikpelinux@gmail.com>
41810         PR target/57583
41811         * config/m68k/m68k.opt (LONG_JUMP_TABLE_OFFSETS): New option.
41812         * config/m68k/linux.h (ASM_RETURN_CASE_JUMP): Handle
41813         TARGET_LONG_JUMP_TABLE_OFFSETS.
41814         * config/m68k/m68kelf.h (ASM_RETURN_CASE_JUMP): Likewise.
41815         * config/m68k/netbsd-elf.h (ASM_RETURN_CASE_JUMP): Likewise.
41816         * config/m68k/m68k.h (CASE_VECTOR_MODE): Likewise.
41817         (ASM_OUTPUT_ADDR_DIFF_ELF): Likewise.
41818         * config/m68k/m68k.md (tablejump expander): Likewise.
41819         (*tablejump_pcrel_hi): Renamed from unnamed insn, reject
41820         TARGET_LONG_JUMP_TABLE_OFFSETS.
41821         (*tablejump_pcrel_si): New insn, handle TARGET_LONG_JUMP_TABLE_OFFSETS.
41822         * doc/invoke.texi (M68K options): Add -mlong-jump-table-offsets.
41824 2017-01-06  Edgar E. Iglesias <edgar.iglesias@xilinx.com>
41825             David Holsgrove <david.holsgrove@xilinx.com>
41827         * common/config/microblaze/microblaze-common.c
41828         (TARGET_EXCEPT_UNWIND_INFO): Remove.
41829         * config/microblaze/microblaze-protos.h (microblaze_eh_return):
41830         New prototype.
41831         * config/microblaze/microblaze.c (microblaze_must_save_register)
41832         (microblaze_expand_epilogue, microblaze_return_addr): Handle
41833         calls_eh_return.
41834         (microblaze_eh_return): New function.
41835         * config/microblaze/microblaze.h (RETURN_ADDR_OFFSET)
41836         (EH_RETURN_DATA_REGNO, MB_EH_STACKADJ_REGNUM)
41837         (EH_RETURN_STACKADJ_RTX, ASM_PREFERRED_EH_DATA_FORMAT): New macros.
41838         * config/microblaze/microblaze.md (eh_return): New pattern.
41840 2017-01-06  Jakub Jelinek  <jakub@redhat.com>
41842         * system.h (GCC_DIAGNOSTIC_PUSH_IGNORED, GCC_DIAGNOSTIC_POP,
41843         GCC_DIAGNOSTIC_STRINGIFY): Define.
41845         * read-rtl.c (rtx_reader::read_rtx_code): Avoid -Wsign-compare warning.
41847 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41849         * config/arm/arm.md (<mcrr>): New.
41850         (<mrrc>): New.
41851         * config/arm/arm.c (arm_arch5te): New.
41852         (arm_option_override): Set arm_arch5te.
41853         (arm_coproc_builtin_available): Add support for mcrr, mcrr2, mrrc
41854         and mrrc2.
41855         * config/arm/arm-builtins.c (MCRR_QUALIFIERS): Define to...
41856         (arm_mcrr_qualifiers): ... this. New.
41857         (MRRC_QUALIFIERS): Define to...
41858         (arm_mrrc_qualifiers): ... this. New.
41859         * config/arm/arm_acle.h (__arm_mcrr, __arm_mcrr2, __arm_mrrc,
41860         __arm_mrrc2): New.
41861         * config/arm/arm_acle_builtins.def (mcrr, mcrr2, mrrc, mrrc2): New.
41862         * config/arm/iterators.md (MCRRI, mcrr, MCRR): New.
41863         (MRRCI, mrrc, MRRC): New.
41864         * config/arm/unspecs.md (VUNSPEC_MCRR, VUNSPEC_MCRR2, VUNSPEC_MRRC,
41865         VUNSPEC_MRRC2): New.
41867 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41869         * config/arm/arm.md (<mcr>): New.
41870         (<mrc>): New.
41871         * config/arm/arm.c (arm_coproc_builtin_available): Add
41872         support for mcr, mrc, mcr2 and mrc2.
41873         * config/arm/arm-builtins.c (MCR_QUALIFIERS): Define to...
41874         (arm_mcr_qualifiers): ... this. New.
41875         (MRC_QUALIFIERS): Define to ...
41876         (arm_mrc_qualifiers): ... this. New.
41877         (MCR_QUALIFIERS): Define to ...
41878         (arm_mcr_qualifiers): ... this. New.
41879         * config/arm/arm_acle.h (__arm_mcr, __arm_mrc, __arm_mcr2,
41880         __arm_mrc2): New.
41881         * config/arm/arm_acle_builtins.def (mcr, mcr2, mrc, mrc2): New.
41882         * config/arm/iterators.md (MCRI, mcr, MCR, MRCI, mrc, MRC): New.
41883         * config/arm/unspecs.md (VUNSPEC_MCR, VUNSPEC_MCR2, VUNSPEC_MRC,
41884         VUNSPEC_MRC2): New.
41886 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41888         * config/arm/arm.md (*ldc): New.
41889         (*stc): New.
41890         (<ldc>): New.
41891         (<stc>): New.
41892         * config/arm/arm.c (arm_coproc_builtin_available): Add
41893         support for ldc,ldcl,stc,stcl,ldc2,ldc2l,stc2 and stc2l.
41894         (arm_coproc_ldc_stc_legitimate_address): New.
41895         * config/arm/arm-builtins.c (arm_type_qualifiers): Add
41896         'qualifier_const_pointer'.
41897         (LDC_QUALIFIERS): Define to...
41898         (arm_ldc_qualifiers): ... this. New.
41899         (STC_QUALIFIERS): Define to...
41900         (arm_stc_qualifiers): ... this. New.
41901         * config/arm/arm-protos.h
41902         (arm_coproc_ldc_stc_legitimate_address): New.
41903         * config/arm/arm_acle.h (__arm_ldc, __arm_ldcl, __arm_stc,
41904         __arm_stcl, __arm_ldc2, __arm_ldc2l, __arm_stc2, __arm_stc2l): New.
41905         * config/arm/arm_acle_builtins.def (ldc, ldc2, ldcl, ldc2l, stc,
41906         stc2, stcl, stc2l): New.
41907         * config/arm/constraints.md (Uz): New.
41908         * config/arm/iterators.md (LDCI, STCI, ldc, stc, LDC STC): New.
41909         * config/arm/unspecs.md (VUNSPEC_LDC, VUNSPEC_LDC2, VUNSPEC_LDCL,
41910         VUNSPEC_LDC2L, VUNSPEC_STC, VUNSPEC_STC2, VUNSPEC_STCL,
41911         VUNSPEC_STC2L): New.
41913 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41915         * config/arm/arm.md (<cdp>): New.
41916         * config/arm/arm.c (neon_const_bounds): Rename this ...
41917         (arm_const_bounds): ... this.
41918         (arm_coproc_builtin_available): New.
41919         * config/arm/arm-builtins.c (SIMD_MAX_BUILTIN_ARGS): Increase.
41920         (arm_type_qualifiers): Add 'qualifier_unsigned_immediate'.
41921         (CDP_QUALIFIERS): Define to...
41922         (arm_cdp_qualifiers): ... this. New.
41923         (void_UP): Define.
41924         (arm_expand_builtin_args): Add case for 6 arguments.
41925         * config/arm/arm-protos.h (neon_const_bounds): Rename this ...
41926         (arm_const_bounds): ... this.
41927         (arm_coproc_builtin_available): New.
41928         * config/arm/arm_acle.h (__arm_cdp): New.
41929         (__arm_cdp2): New.
41930         * config/arm/arm_acle_builtins.def (cdp): New.
41931         (cdp2): New.
41932         * config/arm/iterators.md (CDPI,CDP,cdp): New.
41933         * config/arm/neon.md: Rename all 'neon_const_bounds' to
41934         'arm_const_bounds'.
41935         * config/arm/types.md (coproc): New.
41936         * config/arm/unspecs.md (VUNSPEC_CDP, VUNSPEC_CDP2): New.
41937         * doc/extend.texi (ACLE): Add a mention of Coprocessor intrinsics.
41938         * doc/sourcebuild.texi (arm_coproc1_ok, arm_coproc2_ok,
41939         arm_coproc3_ok, arm_coproc4_ok): Document new effective targets.
41941 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41943         * config/arm/arm-builtins.c (arm_unsigned_binop_qualifiers): New.
41944         (UBINOP_QUALIFIERS): New.
41945         (si_UP): Define.
41946         (acle_builtin_data): New. Change comment.
41947         (arm_builtins): Remove ARM_BUILTIN_CRC32B, ARM_BUILTIN_CRC32H,
41948         ARM_BUILTIN_CRC32W, ARM_BUILTIN_CRC32CB, ARM_BUILTIN_CRC32CH,
41949         ARM_BUILTIN_CRC32CW. Add ARM_BUILTIN_ACLE_BASE and include
41950         arm_acle_builtins.def.
41951         (ARM_BUILTIN_ACLE_PATTERN_START): Define.
41952         (arm_init_acle_builtins): New.
41953         (CRC32_BUILTIN): Remove.
41954         (bdesc_2arg): Remove entries for crc32b, crc32h, crc32w,
41955         crc32cb, crc32ch and crc32cw.
41956         (arm_init_crc32_builtins): Remove.
41957         (arm_init_builtins): Use arm_init_acle_builtins rather
41958         than arm_init_crc32_builtins.
41959         (arm_expand_acle_builtin): New.
41960         (arm_expand_builtin): Use 'arm_expand_acle_builtin'.
41961         * config/arm/arm_acle_builtins.def: New.
41963 2017-01-06  Andre Vieira  <andre.simoesdiasvieira@arm.com>
41965         * config/arm/arm-builtins.c (neon_builtin_datum): Rename to ..
41966         (arm_builtin_datum): ... this.
41967         (arm_init_neon_builtin): Rename to ...
41968         (arm_init_builtin): ... this. Add a new parameters PREFIX
41969         and USE_SIG_IN_NAME.
41970         (arm_init_neon_builtins): Replace 'arm_init_neon_builtin' with
41971         'arm_init_builtin'. Replace type 'neon_builtin_datum' with
41972         'arm_builtin_datum'.
41973         (arm_init_vfp_builtins): Likewise.
41974         (builtin_arg): Rename enum's replacing 'NEON_ARG' with
41975         'ARG_BUILTIN' and add a 'ARG_BUILTIN_NEON_MEMORY.
41976         (arm_expand_neon_args): Rename to ...
41977         (arm_expand_builtin_args): ... this. Rename builtin_arg
41978         enum values and differentiate between ARG_BUILTIN_MEMORY
41979         and ARG_BUILTIN_NEON_MEMORY.
41980         (arm_expand_neon_builtin_1): Rename to ...
41981         (arm_expand_builtin_1): ... this. Rename builtin_arg enum
41982         values, arm_expand_builtin_args and add bool parameter NEON.
41983         (arm_expand_neon_builtin): Use arm_expand_builtin_1.
41984         (arm_expand_vfp_builtin): Likewise.
41985         (NEON_MAX_BUILTIN_ARGS): Remove, it was unused.
41987 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
41989         PR middle-end/77484
41990         * predict.def (PRED_POLYMORPHIC_CALL): Set to 59.
41991         * predict.c (tree_estimate_probability_bb): Reverse direction of
41992         polymorphic call predictor.
41994 2017-01-06  David Malcolm  <dmalcolm@redhat.com>
41996         * passes.c (execute_one_pass): Split out pass-skipping logic into...
41997         (determine_pass_name_match): ...this new function and...
41998         (should_skip_pass_p): ...this new function.
42000 2017-01-06  Nathan Sidwell  <nathan@acm.org>
42002         * ipa-visibility.c (function_and_variable_visibility): Reformat
42003         comments and long lines.  Remove extrneous if.
42004         * symtab.c (symtab_node::make_decl_local): Fix code format.
42005         (symtab_node::set_section_for_node): Fix comment typo.
42007 2017-01-06  Martin Liska  <mliska@suse.cz>
42009         PR bootstrap/79003
42010         * lra-constraints.c: Rename invariant to lra_invariant.
42011         * predict.c (set_even_probabilities): Initialize e to NULL.
42013 2017-01-05  Martin Sebor  <msebor@redhat.com>
42015         PR tree-optimization/78910
42016         * gimple-ssa-sprintf.c (tree_digits): Add an argument.
42017         (format_integer): Correct off-by-one error in the handling
42018         of precision with negative numbers in signed conversions..
42020 2017-01-05  Eric Botcazou  <ebotcazou@adacore.com>
42022         * doc/invoke.texi (C Dialect Options): Adjust -fsso-struct entry.
42024 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42026         PR tree-optimization/71016
42027         * tree-ssa-phiopt.c (tree_ssa_phiopt_worker): Pass cond_stmt to
42028         factor_out_conditional_conversion.  Formatting fix.
42029         (factor_out_conditional_conversion): Add cond_stmt argument.
42030         If arg1 is INTEGER_CST, punt if new_arg0 is not any operand of
42031         cond_stmt and if arg0_def_stmt is not the only stmt in its bb.
42032         Formatting fix.
42034 2017-01-05  David Malcolm  <dmalcolm@redhat.com>
42036         * Makefile.in (OBJS): Add read-md.o, read-rtl.o,
42037         read-rtl-function.o, and selftest-rtl.o.
42038         * config/aarch64/aarch64.c: Include selftest.h and selftest-rtl.h.
42039         (selftest::aarch64_test_loading_full_dump): New function.
42040         (selftest::aarch64_run_selftests): New function.
42041         (TARGET_RUN_TARGET_SELFTESTS): Wire it up to
42042         selftest::aarch64_run_selftests.
42043         * config/i386/i386.c
42044         (selftest::ix86_test_loading_dump_fragment_1): New function.
42045         (selftest::ix86_test_loading_call_insn): New function.
42046         (selftest::ix86_test_loading_full_dump): New function.
42047         (selftest::ix86_test_loading_unspec): New function.
42048         (selftest::ix86_run_selftests): Call the new functions.
42049         * emit-rtl.c (maybe_set_max_label_num): New function.
42050         * emit-rtl.h (maybe_set_max_label_num): New decl.
42051         * function.c (instantiate_decls): Guard call to
42052         instantiate_decls_1 with if (DECL_INITIAL (fndecl)).
42053         * function-tests.c (selftest::verify_three_block_rtl_cfg): Remove
42054         "static".
42055         * gensupport.c (gen_reader::gen_reader): Pass "false"
42056         for new "compact" param of rtx_reader.
42057         * print-rtl.c (rtx_writer::print_rtx_operand): Print "(nil)"
42058         rather than an empty string for NULL strings.
42059         * read-md.c: Potentially include config.h rather than bconfig.h.
42060         Wrap include of errors.h with #ifdef GENERATOR_FILE.
42061         (have_error): New global, copied from errors.c.
42062         (md_reader::read_name): Rename to...
42063         (md_reader::read_name_1): ...this, adding "out_loc" param,
42064         and converting "missing name or number" to returning false, rather
42065         than failing.
42066         (md_reader::read_name): Reimplement in terms of read_name_1.
42067         (md_reader::read_name_or_nil): New function.
42068         (md_reader::read_string): Handle "(nil)" by returning NULL.
42069         (md_reader::md_reader): Add new param "compact".
42070         (md_reader::read_md_files): Wrap with #ifdef GENERATOR_FILE.
42071         (md_reader::read_file): New method.
42072         * read-md.h (md_reader::md_reader): Add new param "compact".
42073         (md_reader::read_file): New method.
42074         (md_reader::is_compact): New accessor.
42075         (md_reader::read_name): Convert return type from void to file_location.
42076         (md_reader::read_name_or_nil): New decl.
42077         (md_reader::read_name_1): New decl.
42078         (md_reader::m_compact): New field.
42079         (noop_reader::noop_reader): Pass "false" for new "compact" param
42080         of rtx_reader.
42081         (rtx_reader::rtx_reader): Add new "compact" param.
42082         (rtx_reader::read_rtx_operand): Make virtual and convert return
42083         type from void to rtx.
42084         (rtx_reader::read_until): New decl.
42085         (rtx_reader::handle_any_trailing_information): New virtual function.
42086         (rtx_reader::postprocess): New virtual function.
42087         (rtx_reader::finalize_string): New virtual function.
42088         (rtx_reader::m_in_call_function_usage): New field.
42089         (rtx_reader::m_reuse_rtx_by_id): New field.
42090         * read-rtl-function.c: New file.
42091         * selftest-rtl.c (selftest::assert_rtx_ptr_eq_at): New function.
42092         * selftest-rtl.h (ASSERT_RTX_PTR_EQ): New macro.
42093         (selftest::verify_three_block_rtl_cfg): New decl.
42094         * read-rtl-function.h: New file.
42095         * read-rtl.c: Potentially include config.h rather than bconfig.h.
42096         For host, include function.h, memmodel.h, and emit-rtl.h.
42097         (one_time_initialization): New function.
42098         (struct compact_insn_name): New struct.
42099         (compact_insn_names): New array.
42100         (find_code): Handle insn codes in compact dumps.
42101         (apply_subst_iterator): Wrap with #ifdef GENERATOR_FILE.
42102         (bind_subst_iter_and_attr): Likewise.
42103         (add_condition_to_string): Likewise.
42104         (add_condition_to_rtx): Likewise.
42105         (apply_attribute_uses): Likewise.
42106         (add_current_iterators): Likewise.
42107         (apply_iterators): Likewise.
42108         (initialize_iterators): Guard usage of apply_subst_iterator with
42109         #ifdef GENERATOR_FILE.
42110         (read_conditions): Wrap with #ifdef GENERATOR_FILE.
42111         (md_reader::read_mapping): Likewise.
42112         (add_define_attr_for_define_subst): Likewise.
42113         (add_define_subst_attr): Likewise.
42114         (read_subst_mapping): Likewise.
42115         (check_code_iterator): Likewise.
42116         (rtx_reader::read_rtx): Likewise.  Move one-time initialization
42117         logic to...
42118         (one_time_initialization): New function.
42119         (rtx_reader::read_until): New method.
42120         (read_flags): New function.
42121         (parse_reg_note_name): New function.
42122         (rtx_reader::read_rtx_code): Initialize "iterator" to NULL.
42123         Handle reuse_rtx ids.
42124         Wrap iterator lookup within #ifdef GENERATOR_FILE.
42125         Add parsing support for RTL dumps, mirroring the special-cases in
42126         print_rtx, by calling read_flags, reading REG_NOTE names, INSN_UID
42127         values, and calling handle_any_trailing_information.
42128         (rtx_reader::read_rtx_operand): Convert return type from void
42129         to rtx, returning return_rtx.  Handle case 'e'.  Call
42130         finalize_string on XSTR and XTMPL fields.
42131         (rtx_reader::read_nested_rtx):  Handle dumps in which trailing
42132          "(nil)" values were omitted.  Call the postprocess vfunc on the
42133         return_rtx.
42134         (rtx_reader::rtx_reader): Add new "compact" param and pass to base
42135         class ctor.  Initialize m_in_call_function_usage.  Call
42136         one_time_initialization.
42137         * rtl-tests.c (selftest::test_uncond_jump): Call
42138         set_new_first_and_last_insn.
42139         * rtl.h (read_rtx): Wrap decl with #ifdef GENERATOR_FILE.
42140         * selftest-rtl.c: New file.
42141         * selftest-rtl.h (class selftest::rtl_dump_test): New class.
42142         (selftest::get_insn_by_uid): New decl.
42143         * selftest-run-tests.c (selftest::run_tests): Call
42144         read_rtl_function_c_tests.
42145         * selftest.h  (selftest::read_rtl_function_c_tests): New decl.
42146         * tree-dfa.c (ssa_default_def): Return NULL_TREE for rtl function
42147         dumps.
42149 2017-01-05  Uros Bizjak  <ubizjak@gmail.com>
42151         * config/i386/i386.md (*testqi_ext_3): No need to handle memory
42152         operands in a special way.  Assert that pos+len <= mode precision.
42154 2017-01-05  Jakub Jelinek  <jakub@redhat.com>
42156         * common.opt (fvect-cost-model): Remove RejectNegative flag, use
42157         3 argument Alias with unlimited for the negative form.
42158         (fno-vect-cost-model): Removed.
42160 2017-01-05  Martin Liska  <mliska@suse.cz>
42162         * hsa-gen.c (gen_hsa_divmod): New function.
42163         (gen_hsa_insn_for_internal_fn_call): Use the function for IFN_DIVMOD.
42165 2017-01-05  Martin Liska  <mliska@suse.cz>
42167         PR pch/78970
42168         * gcc.c (lookup_compiler): Reject '-' filename for a precompiled
42169         header.
42171 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42173         * config/s390/s390.c (s390_expand_setmem): Unroll the loop for
42174         small constant length operands.
42176 2017-01-05  Andreas Krebbel  <krebbel@linux.vnet.ibm.com>
42178         * config/s390/s390.c (s390_expand_setmem): Avoid overlapping bytes
42179         between loop iterations.
42181 2017-01-05  Martin Liska  <mliska@suse.cz>
42183         PR sanitizer/78815
42184         * gimplify.c (gimplify_decl_expr): Compare to
42185         asan_poisoned_variables instread of checking flags.
42186         (gimplify_target_expr): Likewise.
42187         (gimplify_expr): Likewise.
42188         (gimplify_function_tree): Conditionally initialize
42189         asan_poisoned_variables.
42191 2017-01-04  Jeff Law  <law@redhat.com>
42193         PR tree-optimizatin/78812
42194         * rtl.h (contains_mem_rtx_p): Prototype.
42195         * ifcvt.c (containts_mem_rtx_p): Move from here to...
42196         * rtlanal.c (contains_mem_rtx_p): Here and remove static linkage.
42197         * gcse.c (prune_expressions): Use contains_mem_rtx_p to discover
42198         and prune MEMs that are not at the toplevel of a SET_SRC rtx.  Look
42199         through ZERO_EXTEND and SIGN_EXTEND when trying to avoid pruning MEMs.
42201 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42203         * input.c (assert_char_at_range): Default-initialize actual_range.
42205 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42207         * df-scan.c (df_ref_create_structure): Make regno unsigned,
42208         to match the caller.
42210 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42212         * cfgexpand.c (expand_gimple_basic_block): Disregard debug
42213         insns after final jump in test to emit dummy move.
42215 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42217         * gimple-iterator.h (gsi_one_nondebug_before_end_p): New.
42218         * tree-eh.c (cleanup_empty_eh): Skip more debug stmts.
42220 2017-01-04  Alexandre Oliva  <aoliva@redhat.com>
42222         * multiple_target.c (create_dispatcher_calls): Init e_next.
42223         * tree-ssa-loop-split.c (split_loop): Init border.
42224         * tree-vect-loop.c (vect_determine_vectorization_factor): Init
42225         scalar_type.
42227 2017-01-04  Michael Meissner  <meissner@linux.vnet.ibm.com>
42229         PR target/71977
42230         PR target/70568
42231         PR target/78823
42232         * config/rs6000/predicates.md (sf_subreg_operand): New predicate.
42233         (altivec_register_operand): Do not return true if the operand
42234         contains a SUBREG mixing SImode and SFmode.
42235         (vsx_register_operand): Likewise.
42236         (vsx_reg_sfsubreg_ok): New predicate.
42237         (vfloat_operand): Do not return true if the operand contains a
42238         SUBREG mixing SImode and SFmode.
42239         (vint_operand): Likewise.
42240         (vlogical_operand): Likewise.
42241         (gpc_reg_operand): Likewise.
42242         (int_reg_operand): Likewise.
42243         * config/rs6000/rs6000-protos.h (valid_sf_si_move): Add declaration.
42244         * config/rs6000/rs6000.c (valid_sf_si_move): New function to
42245         determine if a MOVSI or MOVSF operation contains SUBREGs that mix
42246         SImode and SFmode.
42247         (rs6000_emit_move_si_sf_subreg): New helper function.
42248         (rs6000_emit_move): Call rs6000_emit_move_si_sf_subreg to possbily
42249         fixup SUBREGs involving SImode and SFmode.
42250         * config/rs6000/vsx.md (SFBOOL_*): New constants that are operand
42251         numbers for the new peephole2 optimization.
42252         (peephole2 for SFmode unions): New peephole2 to optimize cases in
42253         the GLIBC math library that do AND/IOR/XOR operations on single
42254         precision floating point.
42255         * config/rs6000/rs6000.h (TARGET_NO_SF_SUBREG): New internal
42256         target macros to say whether we need to avoid SUBREGs mixing
42257         SImode and SFmode.
42258         (TARGET_ALLOW_SF_SUBREG): Likewise.
42259         * config/rs6000/rs6000.md (UNSPEC_SF_FROM_SI): New unspecs.
42260         (UNSPEC_SI_FROM_SF): Likewise.
42261         (iorxor): Change spacing.
42262         (and_ior_xor): New iterator for AND, IOR, and XOR.
42263         (movsi_from_sf): New insns for SImode/SFmode SUBREG support.
42264         (movdi_from_sf_zero_ext): Likewise.
42265         (mov<mode>_hardfloat, FMOVE32 iterator): Use register_operand
42266         instead of gpc_reg_operand.  Add SImode/SFmode SUBREG support.
42267         (movsf_from_si): New insn for SImode/SFmode SUBREG support.
42268         (fma<mode>4): Use gpc_reg_operand instead of register_operand.
42269         (fms<mode>4): Likewise.
42270         (fnma<mode>4): Likewise.
42271         (fnms<mode>4): Likewise.
42272         (nfma<mode>4): Likewise.
42273         (nfms<mode>4): Likewise.
42275 2017-01-04  Marek Polacek  <polacek@redhat.com>
42277         PR c++/64767
42278         * doc/invoke.texi: Document -Wpointer-compare.
42280 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42282         * optc-gen.awk: Emit #error for -W*/-f*/-m* Enum without
42283         RejectNegative.
42285         * dwarf2out.c (output_loc_list): Don't throw away 64K+ location
42286         descriptions for -gdwarf-5 and emit them as uleb128 instead of
42287         2-byte data.
42289 2017-01-04  Kelvin Nilsen  <kelvin@gcc.gnu.org>
42291         PR target/78056
42292         * doc/sourcebuild.texi (PowerPC-specific attributes): Add
42293         documentation of the powerpc_popcntb_ok attribute.
42294         * config/rs6000/rs6000.c (rs6000_option_override_internal): Add
42295         code to issue warning messages if a requested CPU configuration is
42296         not supported by the binary (assembler and loader) toolchain.
42297         (spe_init_builtins): Add two assertions to prevent ICE if attempt is
42298         made to define a built-in function that has been disabled.
42299         (paired_init_builtins): Add assertion to prevent ICE if attempt is
42300         made to define a built-in function that has been disabled.
42301         (altivec_init_builtins): Add comment explaining why definition
42302         of the DST built-in functions is not preceded by an assertion
42303         check.  Add assertions to prevent ICE if attempts are made to
42304         define an altivec predicate or an abs* built-in function that has
42305         been disabled.
42306         (htm_init_builtins): Add comment explaining why definition of the
42307         htm built-in functions is not preceded by an assertion check.
42309 2017-01-04  Jeff Law  <law@redhat.com>
42311         PR tree-optimizatin/67955
42312         * tree-ssa-alias.c (same_addr_size_stores_p): Check offsets first.
42313         Allow any SSA_VAR_P as the base objects.  Use integer_zerop.  Verify
42314         the points-to solution does not include pt_null.  Use DECL_PT_UID
42315         unconditionally.
42317 2017-01-04  Uros Bizjak  <ubizjak@gmail.com>
42319         * config/i386/i386.md (HI/SImode test with imm to QImode splitters):
42320         Use gen_int_mode instead of gen_lopwart for const_int operands.
42322 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42324         PR tree-optimization/71563
42325         * match.pd: Simplify X << Y into X if Y is known to be 0 or
42326         out of range value - has low bits known to be zero.
42328 2017-01-04  Alan Modra  <amodra@gmail.com>
42330         * Makefile.in (aclocal_deps): Update and order as per aclocal.m4.
42331         * configure: Regenerate.
42332         * config.in: Regenerate.
42334 2017-01-04  Jakub Jelinek  <jakub@redhat.com>
42336         PR bootstrap/77569
42337         * input.c (ebcdic_execution_charset::on_error): Don't use strstr for
42338         a substring of the message, but strcmp with the whole message.  Ifdef
42339         ENABLE_NLS, translate the message first using dgettext.
42341 2017-01-03  Jeff Law  <law@redhat.com>
42343         PR tree-optimizatin/78856
42344         * tree-ssa-threadupdate.c: Include tree-vectorizer.h.
42345         (mark_threaded_blocks): Remove code to truncate thread paths that
42346         cross multiple loop headers.  Instead invalidate the cached loop
42347         iteration information and handle case of a thread path walking
42348         into an irreducible region.
42350 2017-01-03  Michael Meissner  <meissner@linux.vnet.ibm.com>
42352         PR target/78900
42353         * config/rs6000/rs6000.c (rs6000_split_signbit): Change some
42354         assertions.  Add support for doing the signbit if the IEEE 128-bit
42355         floating point value is in a GPR.
42356         * config/rs6000/rs6000.md (Fsignbit): Delete.
42357         (signbit<mode>2_dm): Delete using <Fsignbit> and just use "wa".
42358         Update the length attribute if the value is in a GPR.
42359         (signbit<mode>2_dm_<su>ext): Add combiner pattern to eliminate
42360         the sign or zero extension instruction, since the value is always 0/1.
42361         (signbit<mode>2_dm2): Delete using <Fsignbit>.
42363         PR target/78953
42364         * config/rs6000/vsx.md (vsx_extract_<mode>_store_p9): If we are
42365         extracting SImode to a GPR register so that we can generate a
42366         store, limit the vector to be in a traditional Altivec register
42367         for the vextuwrx instruction.
42369 2017-01-03  Ian Lance Taylor  <iant@google.com>
42371         * godump.c (go_format_type): Treat ENUMERAL_TYPE like INTEGER_TYPE.
42373 2017-01-03  Martin Sebor  <msebor@redhat.com>
42375         PR tree-optimization/78696
42376         * gimple-ssa-sprintf.c (format_floating): Correct handling of
42377         precision.  Use MPFR for %f for greater fidelity.  Correct handling
42378         of %g.
42379         (pass_sprintf_length::compute_format_length): Set width and precision
42380         specified by asrerisk to void_node for vararg functions.
42381         (try_substitute_return_value): Adjust dump output.
42383 2017-01-03  David Edelsohn  <dje.gcc@gmail.com>
42385         * doc/invoke.texi (RS6000 options): LRA is enabled by default.
42387 2017-01-03  Eric Botcazou  <ebotcazou@adacore.com>
42389         * doc/invoke.texi (SPARC options): Document -mlra as the default.
42390         * config/sparc/sparc.c (sparc_option_override): Force LRA unless
42391         -mlra/-mno-lra was passed to the compiler.
42393 2017-01-03  James Cowgill  <James.Cowgill@imgtec.com>
42395         PR rtl-optimization/65618
42396         * emit-rtl.c (try_split): Move initialization of "before" and
42397         "after" to just before the call to emit_insn_after_setloc.
42399 2017-01-03  Gerald Pfeifer  <gerald@pfeifer.com>
42401         * doc/md.texi (Standard Names): Remove reference to Java frontend.
42403 2017-01-03  Pierre-Marie de Rodat  <derodat@adacore.com>
42405         * dwarf2out.c (gen_enumeration_type_die): When
42406         -gno-strict-dwarf, add a DW_AT_encoding attribute.
42408 2017-01-03  Jakub Jelinek  <jakub@redhat.com>
42410         PR tree-optimization/78965
42411         * gimple-ssa-sprintf.c (pass_sprintf_length::compute_format_length):
42412         Change first argument from const call_info & to call_info &.  For %n
42413         set info.nowrite to false.
42415         PR middle-end/78901
42416         * gimple-ssa-sprintf.c (try_substitute_return_value): Don't change
42417         possibly throwing calls.
42419         * genmatch.c (dt_node::gen_kids_1): If generic_exprs include SSA_NAME
42420         and exprs_len || fns_len, emit the code for SSA_NAME next to the exprs
42421         and fns handling, rather than in a separate case SSA_NAME.
42423 2017-01-02  Jeff Law  <law@redhat.com>
42425         * config/darwin-driver.c (darwin_driver_init): Const-correctness
42426         fixes for first_period and second_period variables.
42428 2017-01-02  Uros Bizjak  <ubizjak@gmail.com>
42430         PR target/78967
42431         * config/i386/i386.md (UNSPEC_NOREX_MEM): New unspec.
42432         (*insvqi_1): New insn pattern.
42433         (*insvqi_1_mem_rex64): Ditto.
42434         (*insvqi_2): Ditto.
42435         (*insvqi_3): Rename from *insvqi.
42437         (*extzvqi_mem_rex64): Add UNSPEC_NOREX_MEM tag.
42439 2017-01-02  Gerald Pfeifer  <gerald@pfeifer.com>
42441         * doc/cfg.texi (Edges): Remove reference to Java.
42442         (Maintaining the CFG): Ditto.
42444 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42446         PR middle-end/77674
42447         * symtab.c (symtab_node::binds_to_current_def_p): Fix handling of
42448         transparent aliases.
42450 2017-01-01  Jan Hubicka  <hubicka@ucw.cz>
42452         PR middle-end/77484
42453         * predict.def (PRED_CALL): Update hitrate.
42454         (PRED_INDIR_CALL, PRED_POLYMORPHIC_CALL): New predictors.
42455         * predict.c (tree_estimate_probability_bb): Split CALL predictor
42456         into direct/indirect/polymorphic variants.
42458 2017-01-01  Jakub Jelinek  <jakub@redhat.com>
42460         Update copyright years.
42462         * gcc.c (process_command): Update copyright notice dates.
42463         * gcov-dump.c (print_version): Ditto.
42464         * gcov.c (print_version): Ditto.
42465         * gcov-tool.c (print_version): Ditto.
42466         * gengtype.c (create_file): Ditto.
42467         * doc/cpp.texi: Bump @copying's copyright year.
42468         * doc/cppinternals.texi: Ditto.
42469         * doc/gcc.texi: Ditto.
42470         * doc/gccint.texi: Ditto.
42471         * doc/gcov.texi: Ditto.
42472         * doc/install.texi: Ditto.
42473         * doc/invoke.texi: Ditto.
42475 Copyright (C) 2017 Free Software Foundation, Inc.
42477 Copying and distribution of this file, with or without modification,
42478 are permitted in any medium without royalty provided the copyright
42479 notice and this notice are preserved.